The Pennsylvania State University. The Graduate School. College of Agricultural Sciences FORAGE PARTICLE SIZE AND RATION SORTING

Transcription

1 The Pennsylvania State University The Graduate School College of Agricultural Sciences FORAGE PARTICLE SIZE AND RATION SORTING IN LACTATING DAIRY COWS A Dissertation in Animal Science by Daryl D. Maulfair 2011 Daryl D. Maulfair Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2011

2 ii The dissertation of Daryl D. Maulfair was reviewed and approved* by the following: Arlyn J. Heinrichs Professor of Dairy and Animal Science Dissertation Advisor Chair of Committee Chad D. Dechow Associate Professor of Dairy Cattle Genetics Kevin J. Harvatine Assistant Professor of Nutritional Physiology Robert J. Van Saun Professor of Veterinary Science Gabriella A. Varga Distinguished Professor of Animal Science Terry D. Etherton Distinguished Professor of Animal Nutrition Head of the Department of Dairy and Animal Science *Signatures are on file in the Graduate School

3 iii Abstract Three studies were conducted on early to late lactation Holstein dairy cows to examine the effects of forage particle size (FPS) and ration sorting on chewing behavior, ruminal fermentation, and milk yield and components. The objective of the first experiment was to study effects of replacing alfalfa haylage with dry chopped alfalfa hay in the ration on sorting activity and to determine effects on ruminal fermentation, milk production, or milk composition. In addition, a second objective of this study was to compare results of the PSPS and RTPS for the same TMR samples and to determine effects of separation method on particle size distribution. Ration FPS was varied by replacing alfalfa silage with dry chopped alfalfa hay. The levels of hay used were 5, 10, 20, and 40% of forage DM. The results of this study showed that sorting occurred in all rations, but there was only minimal difference in the type or degree of sorting between treatments and only during the first 4 h after feeding. Sorting activity was highest at the beginning of the d and by 24 h after feeding the diets consumed by the cows were not significantly different from the offered diets. There were no negative effects of including dry chopped alfalfa hay in rations up to 23.5% of ration DM on DM intake, milk yield, and rumen fermentation. Small decreases in milk fat and protein content were found to occur with increasing dry hay inclusion. Data from the Penn State and Ro-Tap particle separators were compared, when separating the same TMR samples, and it was determined that data obtained from these 2 methods of particle separation are not directly comparable and that method of particle separation should be considered when interpreting experimental results. The second experiment s objective was to study the interactions between FPS and ruminally fermentable carbohydrates (RFC) for ration sorting, ruminal fermentation, chewing activity, and milk yield and components. This study varied FPS and RFC by feeding 2 lengths of corn silage and 2 grind sizes of corn grain. The results showed that altering RFC had greater

4 iv influence on milk production parameters than FPS; increasing RFC increased milk yield and protein content and decreased milk fat content. Ruminal fermentation was not affected by either FPS or RFC. Ration sorting occurred on all diets as evidenced by the changes in starch, NDF, and particle size composition of the refusals throughout the d and also by selection indices. Diets containing long FPS were sorted to a greater degree than diets containing short FPS, but there was no interaction between FPS and RFC for ration sorting. There was an interaction between FPS and RFC for DMI; DMI decreased with increasing FPS when the diet included low RFC and did change when the diet included high RFC and DMI increased with RFC for the long diets and did not change with RFC on the short diets. Finally, it was determined that approximately 5% of fecal particles were greater than 6.7 mm and that this may be a more accurate estimate of the critical particle size for rumen escape in modern lactating dairy cows. The objective of the final experiment was to induce a bout of SARA in lactating dairy cows that had ad libitum access to 2 distinct diets that varied in FPS and starch fermentability and to determine how SARA affects TMR selection in dairy cows. One diet consisted of long corn silage and dry cracked corn and the other diet consisted of short corn silage and dry fine ground corn. When offered these 2 diets simultaneously cows consumed 18.1% of their total daily intake as long FPS and low RFC diet. However, after a bout of subacute ruminal acidosis, cows increased their intake of the longer ration to 38.3% of total daily intake. The following d long ration intake moderated to 28.0% and 2 d after the acidosis bout intakes were back to normal at 18.6%. These results indicate that cows are able to alter their diet preference for higher physically effective fiber and slower starch fermentability during a bout of subacute ruminal acidosis, and that they can effectively recover from this type of SARA within 72 h when appropriate diets are available.

5 v Table of Contents List of Figures... ix List of Tables... xi Acknowledgements... xiv Chapter 1 Introduction... 1 Chapter 2 Literature Review... 3 Ruminal Acidosis... 3 Fiber Requirements of Dairy Cattle... 5 Forage Particle Size in the Cow... 8 Ration Sorting Critical Particle Size for Rumen Escape The Various Particle Sieving Methods Penn State Particle Separator American Society of Agricultural and Biological Engineers Particle Separator Ro-Tap Particle Separator Z-Box Particle Separator Wet Sieving The Best Separating Method Forage Particle Size and Starch Fermentability Interaction Ruminal Acidosis and Diet Selection Conclusions References Chapter 3 Eating Behavior, Ruminal Fermentation, and Milk Production in Lactating Dairy Cows Fed Rations That Varied in Dry Alfalfa Hay and Alfalfa Silage Content Abstract Introduction Materials and Methods Diets, Cows, and Experimental Design Feed, Refusal, and Particle Size Analysis Chewing Activity Rumen Sampling Milk Production Statistical Analyses Results and Discussion Chemical Composition and Particle Size Distribution Ration Sorting Intake of DM and Particle Fractions Chewing Activity Rumen Characteristics Milk Production and Composition... 57

6 vi Penn State Versus Ro-Tap Particle Separator Conclusions Acknowledgements References Chapter 4 Effects of Varying Forage Particle Size and Fermentable Carbohydrates on Feed Sorting, Ruminal Fermentation, and Milk and Component Yields of Dairy Cows Abstract Introduction Material and Methods Diets, Cows, and Experimental Design Chewing Activity Rumen Parameters Feed, Refusal, and Particle Size Analysis Milk Production Fecal Sampling Statistical Analyses Results and Discussion Chemical Composition and Particle Size Distribution of Diets Chewing Behavior Ruminal Characteristics Intakes, Refusals, and Ration Sorting Milk Yield and Composition Fecal Particle Size Conclusions Acknowledgements References Chapter 5 Effect of Subacute Ruminal Acidosis on Total Mixed Ration Preference in Lactating Dairy Cows Abstract Introduction Materials and Methods Diets, Cows, and Experimental Design Rumen Sampling Feed, Refusal, and Particle Size Analysis Statistical Analyses Results and Discussion Chemical Composition and Particle Size Distribution of Diets Rumen Characteristics TMR Preference, Dry Matter Intake, and Refusals Ration Sorting Conclusions Acknowledgements References

7 Chapter 6 Conclusions Appendix A Technical Note: Evaluation of Procedures for Analyzing Ration Sorting and Rumen Digesta Particle Size in Dairy Cows Abstract Acknowledgements References Appendix B Effect of Feed Sorting on Chewing Behavior, Production, and Rumen Fermentation in Lactating Dairy Cows Abstract Introduction Materials and Methods Diets, Cows, and Experimental Design Chewing Activity Rumen Sampling Feed, Refusal, and Particle Size Analysis Milk Production Statistical Analyses Results and Discussion Chemical Composition and Particle Size Distribution Ration Sorting Intake of DM, NDF, Starch, and Particle Fractions Chewing Activity Rumen Characteristics Milk Production and Composition Conclusions Acknowledgments References Appendix C Effect of Varying TMR Particle Size on Rumen Digesta and Fecal Particle Size and Digestibility in Lactating Dairy Cows Abstract Introduction Materials and Methods Diets, Cows, and Experimental Design Rumen Sampling Fecal Sampling Digestibility Statistical Analyses Results and Discussion Chemical Composition and Particle Size Distribution Rumen Particle Size Fecal Particle Size and Composition Intakes, Fecal Output, and Digestibility Conclusions vii

8 Acknowledgments References viii

9 ix List of Figures Figure 2-1. Effect of the ratio between physically effective NDF (pendf 1.18 ) to ruminally degradable starch from grains (RDSG) in the diet on daily mean ruminal ph Figure 3-1. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on refusal particle size distribution for 19.0 (A), 8.0 (B), 1.18 mm (C) sieves, and pan (D) Figure 3-2. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on cumulative percent of diet daily intake at various times after feeding Figure 3-3. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen ph over time Figure 3-4. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen NH 3 over time Figure 3-5. Particle size distributions of TMR samples separated with the Penn State (PSPS) and Ro-Tap particle separators divided into particle fractions; > 19.0, > 8.0, > 1.18 mm Figure 4-1. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on starch concentration at 0 and 24 h after feeding Figure 4-2. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on NDF concentration at 0 and 24 h after feeding Figure 4-3. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on TMR particle fractions > 26.9 mm (A), > 1.65 mm (B), and pan (C) at 0, 8, 16, and 24 h after feeding Figure 5-1. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on rumen ph over time for baseline, feed restriction, rumen challenge, and recovery d Figure 5-2. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on preference for TMR with long forage (expressed as a percentage of total daily intake) Figure 5-3. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch

10 x on cumulative percent of diet daily intake at various times after feeding for baseline and rumen challenge d Figure B-1. Effect of feeding TMR of increasing particle size on refusal geometric mean particle size Figure B-2. Effect of feeding TMR of increasing particle size on refusal particle distribution as a percentage of original diet. Selected data shown; 26.9-mm sieve (A) and pan (B) Figure B-3. Effect of feeding TMR of increasing particle size on refusal NDF (A) and starch (B) concentration Figure B-4. Effect of feeding TMR of increasing particle size on cumulative particle size selection index. Selected data shown; 26.9-mm sieve (A) and pan (B) Figure B-5. Effect of feeding TMR of increasing particle size on cumulative NDF (A) and starch (B) selection indices Figure B-6. Effect of feeding TMR of increasing particle size on cumulative geometric mean length (X gm ) selection index Figure C-1. Mean rumen digesta particles of all treatments retained on 1.18-, 0.6-, mm screens, soluble fraction, and soluble DM to retained DM ratio throughout the d Figure C-2. Effect of feeding Short (A), Medium (B), Long (C), and Extra Long (D) TMR on rumen digesta particles retained on 9.5-, 6.7-, and 3.35-mm screens throughout the d Figure C-3. Effect of feeding TMR of increasing particle size on fecal NDF (A), indigestible NDF (B), and starch (C) concentration throughout the d Figure C-4. Effect of feeding TMR of increasing particle size on fecal geometric mean particle length (calculated using data from all particle fractions) throughout the d

11 xi List of Tables Table 2-1. Physical effectiveness factors (pef) for NDF in feeds of each physical form classification based on total chewing activity in relation to that elicited by long grass hay Table 3-1. Chemical compositions and particle size distributions determined for corn silage, alfalfa haylage, and dry chopped alfalfa hay Table 3-2. Ingredients, chemical compositions, and particle size distributions for TMR with increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) Table 3-3. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on DMI, feed efficiency, and milk production and components Table 3-4. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on intake of 4 particle size fractions (> 19.0, > 8.0, > 1.18, and < 1.18 mm) Table 3-5. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on chewing behavior Table 3-6. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen fermentation Table 3-7. Particle size distributions of TMR containing 5, 10, 20, and 40% of forage DM as dry chopped alfalfa hay in samples taken at feeding (0 h) and 24 h after feeding and separated with the Penn State and Ro-Tap particle separators Table 4-1. Chemical compositions and particle size distributions determined with the ASABE particle separator for alfalfa haylage and long and short corn silage Table 4-2. Chemical compositions, particle size distributions, and rates of disappearance determined via in situ incubation for dry cracked and dry fine ground corn Table 4-3. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) Table 4-4. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on chewing behavior Table 4-6. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on cumulative selection indices 1 for various particle fractions

12 Table 4-7. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on interval selection indices 1 for various particle fractions Table 4-8. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily DM, NDF, starch, and particle fraction intake Table 4-9. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on milk yield and components Table Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily weighted mean 1 fecal particle size and DM content Table Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily weighted mean 1 ruminal digesta particle size distribution and DM content Table 5-1. Chemical compositions and particle size distributions determined with the ASABE particle separator for alfalfa haylage and long and short corn silage Table 5-2. Chemical compositions, particle size distributions, and rates of disappearance determined via in situ incubation for dry cracked corn, dry fine ground corn, and ground wheat Table 5-3. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) Table 5-4. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on rumen ph and VFA for baseline and rumen challenge d Table 5-5. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) on DMI and refusals for baseline, feed restriction, rumen challenge, and recovery d Table 5-6. Effect of offering 2 free choice TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) on mean selection indices 1 of baseline, rumen challenge, and recovery d (4 d) Table A-1. Percentage of uneaten TMR particles (DM basis) retained on sieves at 8-h intervals after feeding when sampled by 2 different procedures Table A-2. Geometric mean particle length of uneaten TMR and sorting index of the consumed diet 1 obtained with 2 different sampling procedures xii

13 Table A-3. Percentage of rumen digesta particles (DM basis) retained on sieves after wet sieving when digesta samples were prepared with or without being squeezed through cheesecloth Table B-1. Chemical composition and particle size distributions determined with the ASABE particle separator for corn silage, alfalfa haylage, and short (S), medium (M), long (L), or extra long (XL) grass hay Table B-2. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay Table B-3. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on DM, NDF, and starch intake at various times after feeding and total consumption (measured 24 h after feeding) of various sized particles Table B-4. Observed meal characteristics for diets containing short (S), medium (M). long (L), or extra long (XL) grass hay Table B-5. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on chewing behavior as determined by observed meal criteria Table B-6. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on rumen fermentation Table B-7. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on milk production and components Table C-1. Chemical composition and particle size distributions determined with the ASABE particle separator for corn silage, alfalfa haylage, and short (S), medium (M), long (L), or extra long (XL) grass hay Table C-2. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay Table C-3. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on daily weighted means of fecal NDF, indigestible NDF (INDF), starch, ash, DM and X gm Table C-4. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on daily weighted mean fecal particle size distribution Table C-5. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on DMI, indigestible NDF intake (INDFI), fecal output and apparent digestibilities of DM, NDF, and starch xiii

14 xiv Acknowledgements I wish to first thank my advisor Dr. Jud Heinrichs for giving me the great opportunity to attend graduate school and pursue a Ph.D. studying dairy cattle nutrition at one of the greatest institutions in the world. He allowed me the freedom to take my research in a direction of my choosing and was always able to offer sound advice. I hope my future endeavors will bring great respect to his program. Next, I would like to thank my committee members Drs. Chad Dechow, Kevin Harvatine, Robert Van Saun, and Gabriella Varga for their excellent advice, expertise, suggestions, and constructive criticisms; also for taking the time to read my lengthy dissertation, it is very much improved because of them. I also wish to thank everyone in the Heinrichs lab for all of their help and support. Our lab technician, Maria Long, saved me from spending even more hours in the lab and was always able to offer me assistance no matter the procedure. I am also very thankful to the many undergraduates that helped me during my tenure in graduate school: Blair, Catherine, Hilary, Kolby, Laraya, Meghan, Pam, and Peter. Much of my research involved very laborious and tedious tasks, such as measuring hay particles by hand with a ruler or particle separating samples for months on end; these students completed all of their work with enthusiasm and dedication. I want to thank all of the graduate students whose tenure overlapped with mine for their friendship and assistance, but especially to: Dr. Geoff Zanton whose help and advice on anything related to statistics and experimental design was immeasurable, Dr. Gustavo Lascano for showing me the ropes when I first started graduate school, and to Javier Suarez for always happily volunteering anytime I needed an extra hand. I am greatly indebted to Coleen Jones for her excellent editing skills which made me look like a much better writer than I am. I also wish to thank Kyle Heyler, who despite working in a different lab, helped me numerous times during studies and answered many questions; perhaps most importantly for helping me watch football games from the front

15 xv row of Beaver Stadium. I am also grateful to the dairy farm personnel who went out of their way to assist me during my studies, especially Boyd, Dante, Mark, Nadine, Randy, and Travis. I owe a lot of my success to my girlfriend, Suzie Reding, for being supportive of me in everything that I do. Suzie was always willing to help me with my experiments at hours of the day when few others were willing to help. She also happily made lunches and brought me meals when I was living at the dairy barns. Finally spending time with Suzie made graduate school more bearable by taking my mind of my research and studies even if only for a moment. Lastly and most importantly I wish to thank my family. My parents, Dale and Pattie Maulfair, provided me with the best upbringing that is possible; living on a dairy farm. What I learned from my father has directly made me the person I am today. On a daily basis he taught me, by example, the importance of hard work, determination, and honor. I hope to someday be as good a father as he is. My mother has always provided me with the love that only a mother can. My siblings, Jennie and David, were always supportive of my endeavors and their commitment to the family farm made leaving it a little easier.

16 xvi The cow is the foster mother of the human race. From the day of the ancient Hindoo to this time have the thoughts of men turned to this kindly and beneficent creature as one of the chief sustaining forces of human life. William Dempster Hoard

17 Chapter 1 Introduction Forage particle size, relative to the dairy industry, is a very important but also very complex topic. Dairy cows, being ruminants, require adequate fiber for proper rumen function. Fiber is required by the ruminant to maintain a healthy ruminal environment that allows ruminal microorganisms to flourish, which is necessary to achieve optimal digestion and feed efficiency. However, cattle not only have a chemical fiber requirement but also a physical fiber requirement. Cows need physical fiber to maintain the ruminal mat, stimulate chewing, and buffer the rumen. Longer particle size can decrease dry matter and energy intake and lead to sorting, a condition where cattle do not eat the ration as mixed, but rather eat certain parts of the ration and refuse others. It is thought that ration sorting can lead to subacute ruminal acidosis, a condition of abnormally low ruminal ph (< 5.5), because dairy cows generally sort against longer particles and for shorter particles. Not much is known about how cows decide what feed particles to sort for and against and also what factors influence and change feed sorting preference. This sorting behavior leads to a decrease in fiber intake and an increase in starch intake, as generally fiber and starch are positively and negatively associated, respectively, with particle length. It is well known that low ruminal ph has many detrimental effects on not only the rumen, but the whole animal. Acidosis can lead to decreased intake, digestion, and milk fat content and can cause diarrhea and lameness in addition to many other conditions. In addition, forage particle size must be short enough to allow proper fermentation during storage. Shorter particles are necessary to allow for adequate packing of the silage which limits oxygen during storage thus preventing improper fermentation and molding. These conflicting factors make it difficult to describe the optimum particle size distribution for forages to be fed to dairy cattle.

18 2 Another important consideration regarding forage particle size is the method used to measure particle size distribution. Many systems currently exist to measure particle size and even more methods exist to use particle size data to calculate physically effective fiber in rations. However, since there is not a standard method for the dairy industry or dairy researchers several different systems are currently being used and their data are used interchangeably, though their results may not be comparable. Many of the systems attempting to estimate physically effective fiber are based upon the theory that there is a critical size threshold for particles leaving the rumen and that particles above this threshold are effective because they stimulate chewing to promote their particle size reduction and rumen escape. However, the research that the current critical particle size is based on is aged, and the feeding systems that were used when it was conducted were very different than the feeding systems being used in modern dairy production. Dairy cattle nutrition has changed dramatically, even in the last 30 years. In order to take advantage of the great genetic gains available in current dairy breeds a ration that is much higher in energy must be fed. Common ways to increase energy intake are to decrease the forage to concentrate ratio, which increases the energy density of the diet, and to decrease forage particle size, which increases dry matter intake. These factors make cows more susceptible to acidosis and studying forage particle size will help allow dairy nutritionists to push to the limits of energy intake while maintaining ruminal health. This dissertation will attempt to answer some of these questions that currently exist in the area of forage particle size in lactating dairy cows and perhaps ask some new questions that will encourage further research into this exciting field.

19 Chapter 2 Literature Review Ruminal Acidosis The ruminant animal is unique in the animal kingdom because to achieve optimum feed intake and efficiency its ruminal environment must be maintained within certain physiological limits. These limits are required to be maintained to provide a favorable symbiotic relationship between ruminant host and ruminal microorganisms. The ruminant should provide the microorganisms an environment of limited oxygen, relatively neutral ph, constant temperature, relatively continuous influx of water and organic matter, constant removal or neutralization of waste products and indigestible matter, and mean retention time greater than microbial generation time (Van Soest, 1994). The feeding systems necessary in modern dairy cattle production have made it increasingly difficult to provide a ruminal environment that stays within all of these narrow constraints. The enormous energy requirements of high producing dairy cattle require dairy farmers to feed cattle rations of increasing dry matter intakes (DMI) and levels of concentrate feeds. One of the problems associated with this type of feeding system is an increased susceptibility to ruminal acidosis. Ruminal acidosis is a condition where ruminal ph falls below a certain physiological range of which there are 2 distinct types. The first, more severe, condition is referred to as acute ruminal acidosis and it is generally defined as such when ruminal ph drop below 5.0; the second, less severe, condition is referred to as subacute ruminal acidosis (SARA) and it is generally defined as such when ruminal ph falls in the range of 5.0 to 5.5 (Krause and Oetzel, 2006). The decreased ruminal ph that causes acute acidosis is thought to be mainly caused by an increase in

20 4 ruminal lactate, while the decreased ruminal ph that causes SARA is thought to be mainly caused by an accumulation of volatile fatty acids (VFA) (Harmon et al., 1985; Britton and Stock, 1986; Oetzel et al., 1999). Clinical signs of acute acidosis include anorexia, abdominal pain, tachycardia, tachypnea, diarrhea, lethargy, staggering, recumbency, and death (Krause and Oetzel, 2006). Clinical signs of SARA can be delayed for weeks or months after the bout of depressed ruminal ph. There are many negative side effects associated with SARA, including: decreased DMI (Britton and Stock, 1986; Nocek, 1997), decreased milk production and milk fat content (Nocek, 1997), lameness (Nocek, 1997; NRC, 2001; Stone, 2004), decreased feed efficiency (Huntington, 1993; Nocek, 1997), rumenitis (Brent, 1976), and liver abscesses (Brent, 1976; Britton and Stock, 1986). While acute acidosis is a more severe condition, the incidence of SARA is much higher in dairy cattle and thus has a greater economic impact. A study that evaluated 14 Wisconsin dairy herds and tested 154 cows determined that 20.1% of lactating cows had SARA when tested using rumenocentesis (Oetzel et al., 1999). In a case study of a 500-cow dairy in central New York state, Stone (1999) estimated that SARA could cost up to $475 per cow per year in lost production and components only. Therefore, SARA should be heavily focused on for research and prevention. Stone (2004) suggested that there are 4 types of dairy cattle that are at high risk of developing SARA, they are: transition cows, cows with high DMI, cows that experience high variability in ration composition and meal patterns, and cows fed poorly formulated rations. This is closely related to the suggestion of Krause and Oetzel (2006) that there are 3 major causes of SARA in dairy herds: excessive intake of rapidly fermentable carbohydrates, inadequate ruminal adaptation to a highly fermentable diet, and inadequate ruminal buffering caused by inadequate dietary fiber or inadequate physical fiber. Dairy cattle can consume excessive amounts of fermentable carbohydrates in 2 ways, through high levels of concentrate in the ration or moderate levels of concentrates at high DMI (Krause and Oetzel, 2006). The ruminant should be adapted

21 5 slowly to ration changes, especially when going from high forage to low forage diets, to allow the ruminal microorganism profile to adapt (Van Soest, 1994) and ruminal papillae to lengthen (NRC, 2001). The many aspects of dietary and physical fiber will be discussed in greater detail below. Fiber Requirements of Dairy Cattle The National Research Council (NRC; 2001) recommended a minimum NDF level of 25% of ration DM with a forage NDF level of 19% of ration DM for lactating dairy cows. The NRC based its recommendations on NDF as it is the fiber measure that best separates structural from nonstructural carbohydrates and is comprised of most of the compounds that are considered fiber (NRC, 2001). Forage NDF is included in these recommendations because NDF from nonforage sources is estimated to be about 50% as effective at maintaining chewing activity, milk fat content, and ruminal ph; therefore for every 1 percentage unit decrease in forage NDF, total NFD content should be increased by 2 percentage units (NRC, 2001). The NRC (2001) stated that their recommendations are based on cows fed: a TMR, alfalfa or corn silage as the predominant forage, forage with adequate particle size, and dry ground corn as the predominant starch source. These recommendations are therefore limited to rather specific conditions due to the limited data available and because adequate particle size is not defined in a measurable manner. In addition, NDF minimum levels should be increased if corn is replaced by a more readily fermentable starch source (grain starch fermentability: oats > wheat > barley > corn > milo; Herrera-Saldana et al., 1990) or if finely chopped forage is fed. The minimum level of NDF required by dairy cows is a product of cow and ruminal health (NRC, 2001). Forages are the major supplier of NDF in rations and their slower fermentation and physical characteristics are essential for maintaining ruminal health. The decreased digestibility of forage helps to maintain an optimal ruminal environment by

22 6 diluting the effects of large amounts of VFA produced by NFC fermentation. Fiber (NDF) with adequate length is thought to increase chewing in cattle, which increases salivary secretion of NaHCO 3 and buffers the ruminal digesta (Allen, 1997; Nocek, 1997; Krause et al., 2002b). Saliva production and its ability to buffer the rumen is very important to high producing dairy cows. Large amounts of saliva enter the rumen of lactating dairy cows, approximately 98 to 190 L/d (Bailey, 1961a). The primary buffering compounds in saliva are HCO and HPO 4 (Bailey and Balch, 1961; Bailey, 1961b). These compounds will associate with free H + ions in the rumen and decrease ph. HCO - 3 and HPO 2-4 are very strong buffers at higher ph, but when ph drops too low (approximately 5.5) VFA become the primary buffering system in the rumen (Counotte et al., 1979). Bailey (1961a) found that saliva secretion during eating was 2 to 4 times higher than when at rest. Beauchemin et al. (2008) showed that rate (g/min) of salivation stayed constant during eating; however, changes in the rate of eating affected the amount of saliva secreted per unit of DMI when cows were fed barley silage, alfalfa silage, long-stemmed alfalfa hay, or barley straw, these results agree with the previous research of Bailey (1961a). Particle size, DM, and NDF content of forages are factors affecting rate of eating and time spent eating; chewing rate was decreased and thus saliva secreted per unit of DMI increased when ration particle size, DM, and NDF were increased (Bailey, 1961a; Beauchemin et al., 2008). Chewing was probably first suggested as a means of estimating a feed s effectiveness at maintaining ruminal health by Balch (1971). Sudweeks et al. (1981) continued this work with their roughage value index system and since then several methods have been suggested to estimate the effectiveness of fiber. Most methods relate a feed s effectiveness to its ability to stimulate chewing activity in the cow. Mertens (1997) first defined the concept of effective NDF (endf) as the sum total ability of a feed to replace forage or roughage in a ration so that the percentage of fat in milk produced by cows eating the ration is effectively maintained. While maintaining or improving milk fat is a major impetus for trying to define fiber requirements of

23 7 dairy cattle there are many factors that influence milk fat, some not related to diet, making the endf concept broad and hard to measure. For instance, milk fat is heavily affected by stage of lactation and endf would not be able to account for those differences (Allen, 1997). Another interrelated term was also introduced by Mertens (1997) to describe a slightly different characteristic of forage. Physically effective NDF (pendf) is defined as the physical characteristics of fiber (primarily particle size) that influence chewing activity and the biphasic nature of ruminal contents (Mertens, 1997). This measure combines the physical and chemical properties of a feedstuff to predict chewing and is a product of a feed s physical effectiveness factor (pef) and its NDF content. Physically effective NDF differs from other measures of effective fiber (Balch, 1971; Sudweeks et al., 1981) in that it is based on the relative effectiveness of NDF to promote chewing rather than being expressed as min of chewing per kg of DMI (Mertens, 1997). This eliminates animal variation from being attributed to a feed s effectiveness because chewing per unit of feed varies with animal size, breed, and intake (Mertens, 1997). The more specific concept of pendf is easier to measure than endf since pendf is only concerned with the effect of a feed on chewing and the ruminal mat, which are mostly influenced by particle size and NDF content; although fragility and specific gravity probably have a small influence on pendf as well. Mertens (1997) developed a pef system to calculate pendf that ranges from 0 (feed has no effectiveness in promoting chewing) to 1 (feed has maximum effectiveness in promoting chewing); a hypothetical long grass hay with 100% NDF was defined as having a pef of 1 and an estimated 240 min of chewing per kg of DM or NDF for nonlactating cows eating 0.4 to 2.0 times maintenance. A pef table (Table 1-1) was created that classified various feedstuffs by types and physical forms and assigned each feedstuff a pef value that could be multiplied by the NDF content of a corresponding feedstuff to achieve its pendf. This pendf method not only includes NDF content and particle size but differences in NDF composition, specific gravity, and fragility would be partially accounted for by classifying different feedstuffs separately. However,

24 8 Mertens (1997) also developed a laboratory assessment of pendf where feeds are separated via dry vertical shaking and the proportion of the samples retained above a 1.18-mm sieve (1.65-mm sieve diagonal) are multiplied by sample NDF content. This method is based on 3 assumptions: NDF is uniformly distributed over all particles, chewing activity is equal for all particles retained on a 1.18-mm sieve, and fragility is not different among sources of NDF (Mertens, 1997). The first assumption can be eliminated if the portion of samples retained on a 1.18-mm sieve is directly analyzed for NDF. Measurement of pendf has become widely used in dairy cattle nutrition and research, but it is often measured differently from Mertens (1997) procedure. Many instead use the Penn State particle separator (PSPS) and more discussion of this area will follow. Another problem is that the NRC (2001) failed to publish a requirement for pendf because of a stated lack of a standard, validated method to measure effective fiber of feeds or to establish requirements for effective fiber. A weakness of using the latter pendf method is that NDF fractions are not chemically identical for all forages. NDF composition (the ratio of hemicellulose:cellulose:lignin) of forage varies wildly (Van Soest et al., 1991) and is affected by species, maturity, and storage method. This is probably part of the reason for the many contradictions in the literature about effects of pendf on intake, milk production, milk fat content, and chewing behavior. Using the pef system developed by Mertens (1997) that includes values that differ with type of feedstuff would partially correct for differences across NDF compositions and may improve the correlation between pendf and chewing in the literature. Forage Particle Size in the Cow Adequate forage particle size (FPS) is necessary to maintain cow and ruminal health through buffering ruminal ph, but varying FPS also has many other effects. Many of these effects are inconsistent in the literature due to the many interactions that can occur between diet and cow.

25 9 For instance, it is generally accepted that as FPS increases DMI will decrease due to increased gut fill. Kononoff and Heinrichs (2003b), Leonardi et al. (2005b), and Maulfair et al. (2010) all determined that DMI decreased with increasing FPS; major diet ingredients in these studies were: alfalfa haylage and ground corn; oat silage, corn silage, and ground corn; and corn silage, alfalfa haylage, and ground corn respectively. These results are contrary to Yang et al. (2001b), Krause et al. (2002a), Kononoff and Heinrichs (2003a), Beauchemin and Yang (2005), and Yang and Beauchemin (2007) because they showed no effect of FPS on DMI when feeding: barley silage, alfalfa silage, alfalfa hay, and steam-rolled barley grain, alfalfa silage with either high moisture or dry cracked corn, corn silage and ground corn, corn silage and steam-rolled barley grain, and alfalfa silage and steam-rolled barley grain, respectively. Additionally, Allen (2000) reported that only 3 of 20 comparisons, in 13 articles reviewed, where the same source of forage (hay or silage) was chopped to 2 or more lengths reported a significant effect of forage particle length on DMI. Finally, Krause and Combs (2003) found that when feeding rations of alfalfa silage and corn silage of increasing FPS with either dry cracked shelled corn or high-moisture corn DMI actually increased. A possible reason for this discrepancy is that although longer FPS can increase the filling effect of NDF, longer forages also can lead to increased saliva secretion, which may counteract the filling effect by increasing flow out of the rumen (Allen, 2000). Indeed, Froetschel (1995) showed that infusing saliva in the abomasum of steers led to a linear increase (2.3 to 8.3% higher) in reticular contractions and a linear decrease (7.8 to 13.5% lower) in ruminal contents. The author indicated that the infusion rate of 1.5 L/h for 3 h was within physiological range. There is also some inconsistency in the literature regarding effect of FPS on DM digestibility (DMD). Kononoff and Heinrichs (2003a) and Yang and Beauchemin (2005) reported that increasing ration particle size increased DMD when feeding diets of corn silage with ground corn and steam rolled barley grain, respectively. On the contrary, Kononoff and Heinrichs (2003b) and Maulfair et al. (2011) observed that increasing ration particle size decreased DMD

26 10 when rations of alfalfa silage with ground corn and corn silage, alfalfa haylage, and ground corn were fed. In addition, there are several studies that reported no effect of ration particle size on DMD: Krause et al. (2002a) feeding alfalfa silage with either high moisture or dry cracked corn; Yang and Beauchemin (2006a) feeding barley silage with steam rolled barley grain; and Yang and Beauchemin (2007) feeding alfalfa silage with steam rolled barley grain. Clearly the influence of FPS on DMD interacts with other aspects of diet or management. The effect of FPS on digestibility does not become any clearer when digestibilities of NDF (NDFD) and starch (StarchD) are added to the analysis. Comparing some of the previously cited studies, several reported no differences in DMD, NDFD, or StarchD (Yang and Beauchemin, 2006a; 2007) while another study reported no differences in DMD and NDFD but StarchD decreased (Krause et al., 2002a) with increasing ration particle size. In addition, Yang and Beauchemin (2005) reported an increase in DMD and NDFD with no change in StarchD, but Kononoff and Heinrichs (2003a) did not see a change in NDFD with an increase in DMD (StarchD was not determined in this study) when ration particle size was increased. Finally, Maulfair et al. (2011) reported a decrease in DMD with no change in NDFD and StarchD. These differing results are likely caused by interactions between forage type, forage to concentrate ratio (F:C), and starch fermentability with FPS. None of the experiments with steam-rolled barley grain as the main starch source had any effect of ration particle size on StarchD when fed with multiple forage types (alfalfa, barley, and corn silage) (Yang and Beauchemin, 2005; 2006a; 2007). Two studies using corn grain as the main starch source measured StarchD; Maulfair et al. (2011) found that feeding ground corn with corn silage and alfalfa haylage resulted in no change in StarchD while Krause et al. (2002a) determined that StarchD decreased with increasing ration particle size when feeding high-moisture shelled corn and dry cracked shell corn with alfalfa silage. Therefore, it seems that barley grain digestibility is independent of FPS while corn grain digestibility is variable. Forage source did not have consistent results for NDFD with differing

27 11 ration particle size either. Studies feeding an alfalfa silage-based ration had both no effect of ration particle size on NDFD (Krause et al., 2002a; Yang and Beauchemin, 2007) and a decrease in NDFD with increasing ration particle size (Kononoff and Heinrichs, 2003b). Corn silage-based rations were also inconsistent, with 1 study having an increase in NDFD with increasing ration particle size (Yang and Beauchemin, 2005) and 2 studies that had no effect of ration particle size on NDFD (Kononoff and Heinrichs, 2003a; Maulfair et al., 2011). The interactions occurring in these studies between FPS and digestibility are certainly complex and much more research is needed to elucidate these effects. Ration Sorting It has been estimated that the majority (51.1%) of U.S. dairy farms have adopted TMR as a means of feeding lactating cows; additionally, the percentage of dairies with rolling herd averages over 20,000 lb/cow and dairies with over 500 cows estimated to use TMR are 70.7 and 94.1%, respectively (USDA, 2007). The TMR was developed to provide cows a uniform and consistent diet throughout the d, which is beneficial to the ruminal environment. However, dairy cows have been shown to selectively consume or sort their rations when fed a TMR. Cows generally sort against long particles and for finer particles in their ration (Leonardi and Armentano, 2003; Leonardi et al., 2005a; Maulfair et al., 2010). This behavior can create problems because not only are cows reducing the particle size of their consumed diet but also reducing their NDF intake, because generally longer particles of TMR are composed mainly of forages and contain a higher proportion of NDF than the rest of the ration (Leonardi and Armentano, 2003). In the field and in research up until recently, the presence of sorting was usually determined by comparing the particle size distribution of TMR at feeding to its particle

28 12 distribution at the end of the d. While these distributions are still reported in the literature, sorting activity is now more commonly described using a selection index. Leonardi and Armentano (2003) described a selection index as the actual intake of each fraction (Y i ) expressed as a percentage of the expected intake. Expected intake of Y i equals as-fed intake multiplied by the asfed fraction of Y i in the TMR. The resulting values will fall into 3 categories; sorting for (> 100%), sorting against (< 100%), and no sorting (= 100%) for each particle fraction. Sorting can also be described with this same technique using DM instead of as-fed and results are similar (Leonardi et al., 2005a). A potential problem for dealing with sorting in research or the field is the fact that variability of sorting between cows can be very substantial, especially with the longest fraction (Leonardi and Armentano, 2003; Leonardi et al., 2005a). Several factors have been identified that influence sorting behavior in lactating dairy cows. Increasing the proportion of dry hay in the ration, from 20 to 40% of ration DM, has been shown to increase sorting (Leonardi and Armentano, 2003). However, this effect is likely caused by the large change in ration DM (69.3 to 89.9 %). Leonardi et al. (2005a) showed that when feeding a mixed hay (80% alfalfa and 20% grass) based diet with alfalfa silage an increase in ration DM, from 64.4 to 80.8%, increased sorting against long particles and for short particles. Ration DM in these studies however is much higher than silage-based rations typically found on modern dairy farms. In contrast, Miller-Cushon and DeVries (2009) and Felton and DeVries (2010) recently completed studies that looked at effects of ration DM on sorting with diet DM within the normal range and composed of corn silage, alfalfa haylage, and high-moisture corn. Both of these studies found that decreasing ration DM by adding water during or right after mixing actually increased ration sorting when changing from 57.6 to 47.9% DM and from 56.3 to 50.8 and 44.1% DM for Miller-Cushon and DeVries (2009) and Felton and DeVries (2010), respectively. However, both of these experiments were completed during the summer months and heating was noticed by Felton and DeVries (2010) in the lower DM diets. Therefore the authors

29 13 concluded that increased ration sorting was due to diet instability and spoilage and that adding water to diets with < 60% DM may not decrease sorting and depending on environmental conditions may actually increase sorting. Feeding rations of greater particle size have also been shown to increase sorting. Leonardi and Armentano (2003) reported that feeding longer alfalfa hay versus chopped alfalfa hay increased sorting of rations (against long particles and for fine particles); but intake of long particles still increased because of their higher proportion in the diet. Also the authors determined that, surprisingly, there seems to be no difference in sorting between high quality (34.5% NDF) and low quality (44.5% NDF) hay of the same particle size (Leonardi and Armentano, 2003). Other studies that showed increased sorting against long particles and for short particles with increasing FPS are Kononoff et al. (2003b) and Kononoff and Heinrichs (2003a) both feeding corn silage and ground corn. Bhandari et al. (2008) reported that when feeding rolled barley, ration sorting increased with increasing alfalfa silage particle size, but decreased with increasing oat silage particle size. DeVries et al. (2007) determined that when increasing the F:C in a ration containing grass silage, corn silage, and concentrate mash (from approximately 51:49 to 62:38), cows decreased sorting against long particles and for short particles. The authors suggested that an increased proportion of concentrate made it more available and easier for cows to sort for it. This study also showed that it takes dairy cows only 1 d to adjust their TMR sorting behavior when changing from a high forage diet to a low forage diet (DeVries et al., 2007). Leonardi and Armentano (2007) compared feed sorting in tie- versus free-stall barns when feeding a ration that contained forages of 30.2% alfalfa hay, 20.2% corn silage, and 10.3% wheat straw. They found that when housed in a tie-stall barn cows consumed 73.2% of their expected intake of the longest particles, but cows housed in a free-stall consumed only 63.3%, therefore cows in free-stalls exhibited a greater degree of sorting. The authors suggested the

30 14 reason for the difference is because sorting in a tie-stall is limited by the fact that refusals become coarser over time and the cow becomes forced to eat the long particles whereas cows in free-stalls can move to a bunk location that has not been extensively sorted. Additionally, it was discovered that sorting against particles was increased with increasing feed refusal percentage. Leonardi and Armentano (2007) suggested that ration sorting estimates based on individually fed cows likely underestimate feed sorting that would occur in free-stalls. Hosseinkhani et al. (2008) studied the effect of feed bunk competition in close-up dry cows on feeding behavior. Cows were fed a ration containing alfalfa hay, corn silage, and concentrate mash and were assigned to 1 of 2 treatments that either had 1 cow per bin (noncompetitive) or 2 cows per bin (competitive). It was determined that cows on both treatments sort against particles > 19.0 mm and for particles retained on an 1.18-mm sieve and pan; additionally there was no effect of competition level on feed sorting (Hosseinkhani et al., 2008). Cows in the competition group did have much higher feed intake than cows in the noncompetitive group. The influence of feeding frequency on ration sorting was studied by DeVries et al. (2005) by feeding once, twice, and 4 times per d. Feed sorting occurred in all rations as evidenced by increasing levels of refusal NDF throughout the d. It was determined that increasing feeding frequency from once/d to twice/d decreased sorting activity but increasing from twice/d to 4 times/d had no effect on ration sorting in these cows. Recently Maulfair et al. (2010) studied the effect of increasing dry hay particle size in a corn silage and alfalfa hay based ration. They reported large differences in TMR refusal composition (particle size distribution, NDF, and starch) compared to the ration fed as a result of high ration sorting activity; and difference between fed TMR and orts increased with increased ration particle size. However, actual intake of these components after 24 h was similar for all rations, and as a result milk production, milk components, and ruminal characteristics were similar among the rations (Maulfair et al., 2010). Therefore, cows were essentially receiving

31 15 different rations throughout the d, but the final daily outcome was not different. However, while the rations used in this study varied greatly in particle size, they were relatively high in forage and NDF (59 and 34% of ration DM respectively) and low on rapidly fermentable carbohydrates (starch was 27% of ration DM) and therefore unlikely to cause much stress on the ruminal environment. The authors suggested that when measuring sorting activity in lactating dairy cattle it is important to not only consider composition of the orts (which comprise only a small percentage of daily intake) but also actual intakes of various ration components. Another interesting component of this study was that although the diets fed varied greatly in geometric mean particle size, the consumed geometric mean particle size was very similar across treatments (Maulfair et al., 2010). Cows on the shortest diet ate a ration similar in particle size to what was offered, and cows on all other rations ate a shorter ration than what was offered. Maulfair et al. (2010) suggested perhaps cows were sorting to achieve a desired mean particle size and a ration with the proper particle size may be able to limit or eliminate ration sorting by lactating cows. The results of Maulfair et al. (2010) are generally in agreement with the literature about the effect of sorting on ruminal fermentation and milk production and components. Unfortunately many of the studies with objectives specifically looking at sorting do not report ruminal fermentation and milk data (Leonardi and Armentano, 2003; DeVries et al., 2005; DeVries et al., 2007; Hosseinkhani et al., 2008). Of the studies that experienced significant ration sorting and also reported milk data (Kononoff and Heinrichs, 2003a; Kononoff et al., 2003b; Leonardi et al., 2005a; Leonardi and Armentano, 2007; Bhandari et al., 2008; Miller-Cushon and DeVries, 2009; Felton and DeVries, 2010; Maulfair et al., 2010) there were no differences in milk production, milk fat and protein yields, or milk fat and protein concentrations except for the following instances: increased fat % (P = 0.08) and decreased protein % (P = 0.04) and yield (P = 0.03) with increased sorting (Kononoff and Heinrichs, 2003a); quadratically increased fat % (P = 0.03) and yield (P = 0.03) with increased sorting (Kononoff et al., 2003b); decreased fat % with

32 16 increased sorting (P = 0.09; Leonardi et al., 2005a); increased fat yield with increased sorting (P = 0.09, only for oat silage; Bhandari et al., 2008); and decreased protein % with increased sorting (P = 0.07; Felton and DeVries, 2010). Of the studies that experienced significant ration sorting and also reported ruminal fermentation data (Kononoff and Heinrichs, 2003a; Kononoff et al., 2003b; Leonardi et al., 2005a; Bhandari et al., 2008; Maulfair et al., 2010) there were no differences in average: ruminal ph, total VFA concentrations, or NH 3 concentrations except for the following instances: linearly decreased total VFA concentration with increased sorting (P = 0.07; Kononoff et al., 2003b) and quadratically increased ruminal ph with increased sorting (P = 0.07; Maulfair et al., 2010). Therefore the dairy nutrition industry s general consensus that ration sorting causes decreases in milk production and components and ruminal health is not supported by the literature. Critical Particle Size for Rumen Escape The sieve size 1.18 mm has been widely used as the size at which feed particles retained on or above are considered physically effective for dairy cows. This number originated from research of Evans et al. (1973) and Poppi et al. (1980; 1981), where resistance of particles leaving the rumen of cattle and sheep was measured. It was determined that 1.18 mm was a threshold particle size (not mean) for both cattle and sheep for greatly increased resistance to particles leaving the rumen and < 5% of fecal particles are generally retained on a 1.18-mm sieve. It should be noted that a wet sieving technique was used in these studies to measure particle size; differences between results of various particle separators will be discussed later. Some researchers have suggested that the critical particle size for rumen escape in dairy cattle may be larger than 1.18 mm. Yang et al. (2001a) discovered that when feeding cows diets composed of alfalfa silage, barley silage, alfalfa hay, and steam rolled barley their fecal mean

33 17 particle length averaged 1.86 mm and that 24.8% of fecal particles were retained above a mm sieve and 3.1% of particles were above a 3.35-mm sieve. There was no effect of FPS on fecal particle size. Oshita et al. (2004) completed a study with 4 different diets; long alfalfa hay, chopped alfalfa round bale silage, long orchard grass hay, and chopped corn silage and measured fecal particle size; their percentage of fecal particles retained on a 1.0-mm sieve were: 28.0, 25.2, 12.6, and 26.2% respectively. Other studies that reported larger fecal mean particle size than traditionally expected are Kononoff and Heinrichs (2003a; 2003b) where fecal mean particle size averaged 1.13 and 1.03 mm respectively; the rations fed were composed mainly of ground corn with corn silage and alfalfa haylage respectively. The authors also reported that the proportion of fecal particles > 1.18 mm was 48 and 46% of DM respectively and that FPS did not have an effect on fecal mean particle size in either study. Maulfair et al. (2011) fed 4 diets of increasing FPS (achieved via grass hay chop length) and calculated the geometric mean particle size (X gm ) of feces 2 ways; 1. including only particles retained on the smallest sieve and above and 2. including all sample DM by calculating the amount of soluble DM lost during the sieving process. The retained X gm procedure (using only particles retained on 0.15-mm screens) did not result in any differences among rations and retained X gm of all rations averaged 1.13 mm. The total X gm procedure (using all particle fractions) had much lower values than retained X gm and also had a significant linear contrast for fecal X gm to decrease with increasing TMR particle size, decreasing from 0.33 to 0.31 mm for the shortest to the longest ration respectively. The fecal particle distribution resulted in approximately 16% of particles > 3.35 mm and 37% > 1.18 mm as a proportion of DM retained on the 0.15-mm sieve. The distribution had approximately 7% of particles > 3.35 mm and 17% > 1.18 mm as a proportion of total sample DM. These results, and the results of the previously cited studies, are much higher than those observed by Poppi et al. (1981; 1985) where < 5% of particles were > 1.18 mm as a proportion of total sample DM in mature steers fed exclusively forage. The reasons

34 18 for the 3- to 4-fold increase in particles > 1.18 mm escaping the rumen are probably due to large differences in DMI and passage rate of high producing dairy cows compared to steers being fed a maintenance diet. It is clear, based upon all of this data, that 1.18 mm is not the critical threshold for rumen escape in modern lactating dairy cows; however more research is needed to determine the exact size and what factors can lead to variance in this critical size. The Various Particle Sieving Methods Several studies have used the particles retained on the 1.18-mm sieve of the PSPS to determine pendf of TMR (Yang and Beauchemin, 2006b; Yang and Beauchemin, 2007; Bhandari et al., 2008). Also, studies have been conducted that used the 8-mm screen of the PSPS to determine pendf (Calberry et al., 2003; Plaizier, 2004; Yang and Beauchemin, 2005; DeVries et al., 2007). However, the PSPS uses a very different particle separating technique from the one specified by Mertens (1997) pendf procedure. In addition, it should also be noted that when using the 1.18-mm sieve in the PSPS to measure pendf there may be no significant differences in pendf of TMR found, even though there are significant differences in particle size distribution (Yang and Beauchemin, 2006b; Yang and Beauchemin, 2007; Bhandari et al., 2008) and even cow response (Yang and Beauchemin, 2006b). This shows a lack of sensitivity when using this technique to measure pendf, likely caused because when forage chop lengths are varied, most of the differences in particle distribution of the TMR are in particles above this sieve. There currently seems to be no standard TMR and forage particle separating technique for determining pendf and many problems can be created by interchanging pendf values determined by different sieving methods. Several of the most popular particle separating methods will be discussed in more detail.

35 19 Penn State Particle Separator The PSPS is probably considered to be the standard particle separating technique in the dairy cattle industry. The PSPS is a manually operated particle separator that separates as-fed forage and TMR samples via horizontal shaking. Lammers et al. (1996) first developed the PSPS as an easy to use, practical, on-farm tool to mimic Standard S424 of the American Society of Agricultural and Biological Engineers (ASABE), which is the standard method of determining particle size distribution of chopped forages. The first PSPS consisted of 3 particle fractions; > 19.0, > 8.0, and < 8.0 mm. The PSPS was later improved upon by Kononoff et al. (2003a) by adding an 1.18-mm screen to allow for a more accurate characterization of TMR and forages that have a large portion of particles < 8.0 mm. The top 2 screens have circular holes and the screen depth is varied (12.2 and 6.4 mm for the top and middle screens respectively) to provide a 3- dimensional barrier to prevent particles larger than the hole sizes from falling through (Lammers et al., 1996). The bottom sieve is composed of a stainless steel wire cloth that has a nominal screen size of 1.18 mm and a diagonal screen size of 1.67 mm (Kononoff et al., 2003a). Recommended sample size for the PSPS is 1.4 L or ¼ of the ASABE standard sample size since the PSPS has approximately ¼ of the surface area of the ASABE separator (Lammers et al., 1996). The recommended shaking procedure is (on a flat surface) shake the separator horizontally 5 times (at 1.1 Hz with a stroke length of 17 cm; (Kononoff et al., 2003a), then rotate the separator ¼ turn and repeat; a total of 8 sets of 5 shakes should be completed for a total of 40 shakes in 2 full turns (Lammers et al., 1996). Lammers et al. (1996) determined that there was no difference in the results of the PSPS and the ASABE separator in predicting fractions of particles < 19.0 and < 8.0 mm in 21 of the 36 statistical tests. Advantages of the PSPS are its: portability, low cost ($300; Nasco, Fort Atkinson, WI), ease of use, quick results, use of as-fed samples, and good repeatability. It is because of these reasons that it has become popular with dairy farmers

36 20 and nutritionists worldwide. The PSPS can be easily used in a field or barn whenever it is needed without the need for time consuming drying of samples. Some disadvantages of the PSPS are it: determines fewer particle fractions than other methods and uses manual operation. Anytime a procedure requires manual manipulation it introduces a certain amount of human error; however, the ability to rest the PSPS on a smooth, steady surface does a good job of limiting human error. Other disadvantages of the PSPS were reported by Kononoff et al. (2003a), they determined that moisture content of samples and shaking frequency affected particle size distribution and mean particle size. Small losses of moisture caused only minor changes in particle size distributions while complete drying caused large differences, by increasing the amount of particles passing through each sieve (Kononoff et al., 2003a). Therefore it is important to standardize the shaking procedure and consider the effects of moisture when utilizing the PSPS. American Society of Agricultural and Biological Engineers Particle Separator The ASABE or Wisconsin separator is the standard method for determination of particle size distribution of chopped forages (S424.1; ASABE, 2007). It is a very large (> 225 kg) particle separator that is mechanically operated and utilizes a horizontal shaking motion. The ASABE separator consists of a pan and 5 square-hole screens with sizes of 19.0, 12.7, 6.3, 3.96, and 1.17 mm when measured nominally or 26.9, 18.0, 8.98, 5.61, and 1.65 mm when measured diagonally, which are all in frames of mm (length width depth; ASABE, 2007). All of the screens are made of aluminum of varying thickness, increasing with increasing screen size, except the smallest screen, which is wire mesh. Thicknesses of the screens are from top to bottom: 12.7, 9.6, 4.8, 3.1, and 0.64 mm (American Society of Agricultural and Biological Engineers., 2007). The recommended procedure is to use a sample size of 9 to 10 L of uncompressed forage, but samples as small as 2 to 3 L can be used if extra care is taken to

37 21 recover the particles from the screens, and to operate the shaker for 2 min (ASABE, 2007). Several advantages of this separator are it: is mechanically operated, has a moderate number of particle fractions, uses as-fed samples, has screens with more surface area (longer and wider) than PSPS. These advantages help to: reduce human error, more accurately describe particle distribution, eliminate the need for sample drying, and allow for better separation of extremely long particles respectively. Maulfair et al. (2010) found that when using rations of extremely long particle size the PSPS did not adequately separate the particles. The extremely long hay particles would bind together and not allow any particles to fall through the top screen when shaken with the PSPS. The larger screens and more vigorous shaking of the ASABE separator allowed enough movement of the longest particles for the smaller particles to fall through the screens (Maulfair et al., 2010). This situation would not be realized very often though as these diets were very extreme. The disadvantage of this separator is that it is the least portable of all separators; it is very heavy, takes of a lot of space ( cm; length width height), and requires electricity to operate. It is also likely very expensive as they must be custom manufactured. Results of the ASABE particle separator are also susceptible to variation with sample moisture content (ASABE, 2007). Disadvantages of this particle separator strictly limit its use to research. Ro-Tap Particle Separator The Ro-Tap particle separator (RTPS; W.S. Tyler, Mentor, OH) uses a very interesting technique for separating particles. A dried sample is placed on a series of stacked sieves (same sieves used in wet sieving) placed on the machine, which horizontally shakes them while simultaneously a metal arm repeatedly taps the top of the sieve stack (holds 8 to 16 depending on sieve height) to incorporate a vertical shaking element as well. This shaking system could probably be considered obsolete, except it was used for much of the research of Mertens. Mertens

38 22 (1997) developed the concept of pendf and used the RTPS for development of the laboratory assessment of pendf, where the particles retained on a 1.18-mm after shaking are multiplied by the sample NDF content. Mertens (2005) RTPS procedure specifies a sample size of 0.6 L, sieve sizes of 19.0, 13.2, 9.5, 6.7, 4.75, 3.35, 2.36, 1.18, 0.60, and 0.30 mm, and a 10 min operation time. A major factor that creates a difference between the RTPS and other methods is that vertical shaking tends to separate particles by their minimum cross-sectional dimension (usually width in forage particles) whereas horizontal shaking tends to separate particles by their length (Mertens, 1997; Mertens, 2005); this difference is amplified by the fact that the RTPS uses wire screens that have a minimum screen thickness versus the large thicknesses of the PSPS and ASABE separator screens. Since the RTPS utilizes vertical shaking and dried samples it produces results that can be very different from conventional techniques (PSPS and ASABE separator) that use horizontal shaking and as-fed samples. Which shaking technique is optimal may depend on the samples being separated and the hypothesis that is being answered, for instance, separating particles based on their smallest diameter may be more similar to how particles attempt to leave the rumen. Further discussion on the differences between the RTPS and the PSPS can be found in Chapter 3. The other divergence of the RTPS from most conventional techniques is that samples are dried before they are separated. Drying forage samples makes particles become smaller and more fragile, making them more likely to break during the separating process; both of these factors can artificially decrease the resulting particle size distributions (Kononoff et al., 2003a). Drying samples also makes this technique more time consuming as samples are usually dried for at least 24 h (Mertens, 2005). Other disadvantages of the RTPS are: not very portable, expensive ($2,300 2,500 plus sieves; Thermo Fisher Scientific, Waltham, MA), requires electricity, and is extremely loud to operate. Some advantages of the RTPS are that it is mechanically operated, many screens can be used (up to 8 or 16 depending on sieve height), and the screen sizes can be customized for intended use. The characteristics of the RTPS again relegate its use to research.

39 23 Z-Box Particle Separator The Z-Box particle separator was recently developed at the William H. Miner Agricultural Research Institute (Chazy, NY) and was specifically designed to measure pef of asfed forage and TMR samples. The Z-Box was also designed to be highly correlated with the proportion of particles retained above a 1.18-mm sieve when separated via the RTPS. Research and development of this separator involved testing various screen sizes (1.14, 2.38, 3.18, 4.76, and 9.53 mm), shaking motions (horizontal and vertical), and sample sizes (50 and 100 g) (Cotanch and Grant, 2006). Samples of corn silage, hay crop silage, and TMR that varied in pef were separated using the various combinations and the results were compared to the RTPS. Cotanch and Grant (2006) determined that vertical shaking of 50-g samples correlated best with the RTPS particle fraction > 1.18 mm and that the best screen size varied with the type of samples sieved. They suggested that a 3.18-mm screen should be used for corn silage and TMR and a 4.76-mm screen should be used for hay crop silage. The Z-Box is a handheld plastic box ( cm, length width height) that has a removable screen. Cotanch and Grant (2006) recommended the following procedure for Z-Box use: place 50-g sample in box and record weight, insert appropriate sieve, invert box and vigorously shake vertically for 50 shakes (rotating box ¼ turn every 10 shakes), invert box and remove lid and sieve to weigh. Even though Cotanch and Grant (2006) reported low variability both within and between technicians, field observations have proved the opposite. It appears that because of the large requirement for human manipulation the Z-Box does not have very good repeatability. The Z-Box does have the advantages of portability, low cost ($250; William H. Miner Agricultural Research Institute, Chazy, NY), and ease of use (except for having to change screens); however, these factors are overshadowed by its lack of repeatability.

40 24 Wet Sieving There are 2 types of wet sieving reported in the literature. The first type consisted of a series of stacked sieves being completely submersed in a vat of water and moving vertically in the water for a period of time. This type of wet sieving was used by Poppi et al. (1980; 1981; 1985) when 1.18 mm was first suggested as the critical particle size for particles leaving the rumen of cattle and sheep. This type of sieving seemingly has not been used for several decades and would likely be considered obsolete. The other method of wet sieving is the type of procedure used by Beauchemin et al. (1997) and improved upon by Maulfair and Heinrichs (2010). In this procedure a series of stacked sieves of decreasing size have water sprayed onto the top screen and in the middle of the sieve stack. While the water is being sprayed onto the samples in the sieve stack, the entire stack is vibrated via vertical oscillation. The bottom pan in the sieve stack is drained to allow water and soluble matter to flow out. Soluble DM (DM that passes through the smallest sieve) can be determined by calculating the DM lost during the sieving process (Maulfair and Heinrichs, 2010). Six different sieve sizes can be used at 1 time (up to 12 if half-size sieves are used) and the sizes can be customized to suit the intended uses of the separating. This technique lends itself very well to separating samples that have high moisture contents (rumen digesta and fecal samples) because these samples will not separate well using other techniques without drying and drying can change the physical properties of samples. Wet sieving is valuable for research because it most accurately mimics conditions in the rumen as particles pass through the omasal canal. Particles in the rumen are completely water saturated and suspended in fluid when they pass though the omasal canal, and this is the only particle separating method that closely resembles this action. However there are many disadvantages to using this method. This procedure is very time consuming; even with the modifications to increase processing time made by Maulfair and Heinrichs (2010) at least 30 min are required to process a single sample. Wet

41 25 sieving equipment is expensive ($2,900 3,500 plus sieves; Thermo Fisher Scientific, Waltham, MA), not easily portable, and needs running water and electricity to operate. The characteristics of this method make it very valuable for research but impractical for field use. The Best Separating Method Clearly there is no single separator that is best for all uses. The type of sample being used and the hypothesis being questioned influence which particle separator to use. Wet sieving is most likely the best technique when studying particles passing out of the rumen, because rumen digesta and fecal samples can be separated without changing their physical conformation. The separating action of wet sieving also more closely mimics the actions that occur in the rumen; separating on smallest diameter while suspended in fluid. The particle separator that best measures ration pendf is not as easy to define. Since pendf is described as the ability of a feed to stimulate chewing and maintain the rumen mat (Mertens, 1997); the best separator should be the one that best correlates to chewing activity. An as-fed sample may correlate better to chewing because that is the form it is in when presented to the cow. Horizontal separating may correlate better to chewing because it separates on longest diameter (Mertens, 1997; Mertens, 2005) and the cow likely chews until the longest diameter of forage particles are below a certain size. Additionally, repeatability of the separator is extremely important and portability, ease of use, and cost must also be considered if the separator is to be accepted for field use. Therefore, the best particle separator for estimating pendf may be the PSPS, but more research is needed to find the sieve size or combination of sieve sizes that will best correlate to chewing activity or rumen ph.

42 26 Forage Particle Size and Starch Fermentability Interaction Few studies have specifically studied the interaction of FPS and ruminally fermentable carbohydrates (RFC) by altering both simultaneously. Yang et al. (2001b) fed rations that varied extent of grain processing, F:C, and FPS. These factors were altered by feeding: coarse and flat steam-rolled barley grain, F:C of 35:65 and 55:45, and long and short barley silage, alfalfa silage, and alfalfa hay respectively. Yang et al. (2001b) determined that DMI increased with increasing RFC and was not affected by FPS. Average ruminal ph was decreased with increasing RFC and again not affected by FPS. Finally, milk yield, milk fat content, and milk protein content were increased, decreased, and increased, respectively, by increasing RFC; while they were not affected, tended to increase, and tended to increase with increasing FPS. The authors concluded that RFC was the most influential factor affecting milk production while FPS had minimal impact (Yang et al., 2001b). It is important to note that in this study the variation in FPS was not great. The percent of DM retained above the PSPS 19.0-mm sieve for long and short barley silage, long and short alfalfa silage, and long and short alfalfa hay were: 5.6, 0.4, 3.9, 0.3, 20.6, and 0.0%, respectively; therefore even the long hay crop silages were below the current recommendation of 10 to 20% retained on the 19.0-mm sieve when determined with the PSPS (Heinrichs and Kononoff, 2002). The only interaction involving FPS was with F:C for milk fat content (P = 0.06). It was found that when increasing FPS, milk fat content had a higher increase for the high forage diet compared to the low forage diet, likely because the pendf intake increased to a greater extent for the high forage diets (Yang et al., 2001b). Yang et al. (2001b) suggested that ruminal ph and SARA cannot be predicted directly using only the physical characteristics of the diet, as RFC appears to have greater impact on ruminal ph than FPS. Krause et al. (2002a; 2002b) also examined the interactions of FPS and RFC and fed rations that varied FPS with short and long alfalfa silage and varied RFC with dry cracked shelled

43 27 corn and high-moisture corn. It was determined that increasing RFC decreased DMI, while FPS had no effect (Krause et al., 2002a). Krause et al. (2002a) reported that the interaction between FPS and RFC was significant for NDF, ADF, and starch intake; increasing FPS with high RFC decreased NDF and ADF intake and increased starch intake, but increasing FPS with low RFC increased NDF and ADF intake and decreased starch intake. Mean ruminal ph decreased with increasing RFC (5.99 to 5.85) and increased with increasing FPS (5.81 to 6.02), and no interaction between FPS and RFC for mean ruminal ph was present (Krause et al., 2002b). Increasing RFC tended to increase milk yield but had no affect on milk fat or protein content, while increasing FPS had no affect on milk yield, fat, or protein content (Krause et al., 2002a). An interaction between FPS and RFC also occurred (P = 0.06) for milk yield, as milk yield tended to increase with FPS with high RFC and tended to decrease with low RFC (Krause et al., 2002a). The authors suggested that this interaction might be an affect of the shift toward a lower fiber and higher starch intake when FPS was increased with high RFC allowing higher energy intake, whereas the opposite occurred with low RFC. This situation probably also caused the trend towards an interaction (P = 0.09) occurring for milk protein content, as higher energy intake could lead to increased microbial synthesis and result in higher milk protein content (Krause et al., 2002a). Interestingly, it was determined that increasing RFC, by replacing dry cracked shelled corn with high-moisture corn, tended to increase (P = 0.08) ruminating min/d and increased (P = 0.03) ruminating min/kg of NDF intake. Krause et al. (2002b) suggested that since alfalfa silage should be the only diet component that could stimulate rumination, the increase in ruminating activity is a result of an adaptive response by the animals to increased RFC to attenuate low ruminal ph via increased saliva secretion. Also this finding indicates that physical effectiveness of forages is affected by other dietary components such as corn grain moisture and fermentability (Krause et al., 2002b). The authors of this study concluded that diets low in effective fiber and

44 28 high in RFC can be fed to midlactation cows without causing negative effects on cow productivity (Krause et al., 2002a). Finally, the interactions between FPS and RFC were also studied by Krause and Combs (2003). In this study RFC, FPS and F:C were varied by feeding; dry cracked shelled corn or ground high-moisture corn, short or long alfalfa silage, and alfalfa silage as the only forage or a 50:50 mixture of alfalfa and corn silage respectively. It was determined that DMI decreased with increasing RFC and increased with increasing FPS. Mean ruminal ph decreased with increasing RFC and was not affected by FPS, and it should be noted that mean ruminal ph was below 6.0 for all treatments, probably due to the low NDF (24.3 to 26.4%) and high starch (28.4 to 38.7%) contents of the diets (Krause and Combs, 2003). There were significant interactions between FPS and RFC for time below ph 5.8 (h/d) and area below ph 5.8 (h ph units/d); because as FPS increased time and area below ph 5.8 decreased with high RFC but increased with low RFC, but the authors were unable to explain the reasons for these interactions (Krause and Combs, 2003). Finally, milk yield was not affected by RFC but tended to decrease with increasing FPS, milk fat content was decreased with increasing RFC and increased with FPS, and milk protein content was decreased with increasing RFC and not affected by FPS. Krause and Combs (2003) concluded that because of interactions that occurred between FPS and RFC for ruminal fermentation variables, the effects of FPS and RFC are not always additive which complicates the inclusion of these factors in dairy ration formulation and evaluation programs. It is clear from the results of these studies that RFC generally has a greater influence on DMI, milk yield and components, and ruminal fermentation than FPS. Also interactions between FPS and RFC for milk production and ruminal fermentation regularly occur making balancing rations for these components much more difficult. None of these studies that varied both FPS and RFC in order to study their interactions also measured or reported on the effects of these variables on ration sorting or diet selection; so the interactions of FPS and RFC on ration sorting is not

45 29 known. A recent review of 45 studies including 134 different experimental diets examined the influence pendf and ruminally degradable starch from grains on rumen ph and also determined that ruminally degradable starch had a higher impact on ruminal fermentation than pendf (Zebeli et al., 2010). Zebeli et al. (2010) also determined that the ratio of pendf to ruminally degradable starch from grains should be no lower than 1.45 to maintain mean ruminal ph above 6.2 (Figure 2-1). Ruminal Acidosis and Diet Selection The optimal foraging theory of feed selection put forth by Krebs and McCleery (1984) states that animals will select the feed that offers them the greatest potential energy intake rate when given a choice. However, Forbes and Kyriazakis (1995) stated that the ruminant animal is faced with the dilemma of choosing a nutrient dense feed, which allows for it to grow and reach puberty as quickly as possible, or choosing a fibrous feed and supporting a stable and healthy ruminal environment. The work of Cooper et al. (1996) also suggests that ruminants make diet choices that are contrary to the optimal foraging theory by selecting feeds that do not maximize their energy intake rate. Instead they put forth the hypothesis that 1 objective of diet selection in ruminants is to sustain high levels of feed intake by keeping ruminal conditions within certain physiological limits (Cooper et al., 1996). In their study 42 sheep were divided into seven 6 6 Latin squares and were offered a combination of diet choices to study the effects of physical form, carbohydrate source, and NaHCO 3 inclusion rate on feed selection. The feed choice combinations included: low energy density (LED) feeds, long alfalfa hay and alfalfa pellets; high energy density (HED) feeds, barley pellets and sugar beet pulp/barley pellets each with varying NaHCO 3 inclusion rates of 0, 1, 2, 4% (wt/wt). When fed either long or pelleted alfalfa singly, sheep consumed higher amounts of

46 30 alfalfa pellets probably due to an increased rate of passage through the rumen. Of the HED feeds offered singly, sheep consumed more sugar beet pulp/barley pellets than barley pellets. The reason for increased consumption of the sugar beet pulp/ barley pellets is most likely due to its greater buffering capacity which helps maintain higher rumen ph levels as opposed to barley pellets which decrease rumen ph and thus feed intake (Cooper et al., 1996). Sheep tended to select a diet that was supplemented with NaHCO 3 when given a choice; however, there was not a dose dependent response. The 2 most likely reasons for this are that either the NaHCO 3 inclusion rates were too similar and the sheep were not able to differentiate among them or that the highest level of NaHCO 3 inclusion was associated with negative effects on the rumen through increased rumen osmolality (Cooper et al., 1996). When offered the choice between long alfalfa hay and alfalfa pellets, sheep consumed a higher proportion of the pellets. However, the sheep did not eat alfalfa pellets exclusively and chose to consume substantial amounts of long alfalfa hay in order to maintain their rumen health (Cooper et al., 1996). Finally, when offered the choice between either LED feeds and a HED feed, sheep ate a higher proportion of the LED feed when the other feed choice was barley pellets compared to sugar beet pulp/ barley pellets. Cooper et al. (1996) suggested that sheep consumed more LED feed when offered the more highly fermentable barley pellets to minimize the adverse affects on the rumen associated with consuming this type of feed. Also when offered both a LED feed with either HED feed, sheep ate a higher proportion of the LED feed when it was offered as alfalfa pellets, this could be due to the fact that the pellets offer a higher intake rate or that less long hay is needed to be consumed by the sheep in order to maintain certain ruminal conditions (Cooper et al., 1996). Surely there are many factors that influence diet and feed selection in ruminants, but this study shows that a substantial factor is the maintenance of a healthy rumen environment. Castle et al. (1979) completed a study that showed that dairy cattle also do not always follow the optimal foraging theory of Krebs and McCleery (1984). In this study, 3 grass silages of

47 31 different particle lengths were fed simultaneously to 3 pregnant Ayrshire heifers. This study was 3 wk long, at the beginning of each wk the position of silages were changed randomly, and each silage was fed to achieve a 20% refusal rate. Silage intake preferences were measured on the last 4 d of each wk. The heifers consumed 15.9, 31.9, and 52.2% of total DMI as long, medium, and short silages respectively (Castle et al., 1979). One confounding factor in this study was that the 3 silages were chopped differently at harvest and stored separately. The variation of particle sizes altered silage densities and caused differences in silage fermentation; the long silage had the highest ph and butyrate concentration with the lowest lactate concentration (Castle et al., 1979). These factors most likely increased aversion to the long silage and they should be considered when interpreting the results. These heifers clearly showed a preference to consume substantial amounts of longer forages at the expense of maximum feed intake, agreeing with the theory put forth by Cooper et al. (1996). Four models of feed selection in the ruminant have been proposed by Provenza (1995): euphagia, hedyphagia, body morphophysiology and size, and learning through foraging consequences. The euphagia model is described as an animal s innate ability to taste and smell specific nutrients and toxins in feed, which would allow the animal to simultaneously select nutritious feeds while avoiding those that are harmful (Provenza, 1995). The hedyphagia feed selection model states that animals will select feeds that are immediately pleasing to olfactory, gustatory, and tactile senses and avoid feeds that are not (Provenza, 1995). This model relies on the assumption that nutritious compounds taste good and harmful compounds taste bad. The third model, body morphophysiology and size, is based on the fact that ruminant species differ in their ability to ingest forages with different physical and chemical characteristics because of varying morphological and physiological adaptations (Provenza, 1995). Finally, the learning through foraging consequences model involves feedback mechanisms that allow animals to select nutritious and healthy diets when faced with feeds that vary in nutrients and toxins (Provenza,

48 ). This model assumes that diet selection is based upon positive and negative consequences experienced by the animal through ingestion of diverse feeds and includes prominent aspects of the other 3 models. This model relies on neural interactions and feedback mechanisms among taste, smell, and the gastrointestinal tract. These feed selection models may help explain how ruminal acidosis can influence diet selection. In a study by Phy and Provenza (1998b) lambs were fed a meal of rolled barley and then offered a choice of flavored (onion or oregano) rabbit pellets that either contained NaHCO 3 and lasalocid or NaCl. The authors determined that after a grain meal lambs preferred rabbit pellets that contained NaHCO 3 and lasalocid over pellets that contained NaCl. The second part of this study examined the effects of feeding different levels of wheat on the intake of a solution containing NaHCO 3, NaCl, or pure water. First, Phy and Provenza (1998b) determined that lambs increased their consumption of a NaHCO 3 solution when wheat intake was increased but water intake was not affected by wheat intake. In addition, it was determined that lambs increased their intake of a NaHCO 3 solution (186%) to a greater degree than a NaCl solution (140%) when wheat intake was increased. All of these results show that lambs prefer feeds and solutions (NaHCO 3 and lasalocid) that attenuate acidosis after a grain meal to maintain ruminal health (Phy and Provenza, 1998b). However even though clinical acidosis was reduced in the high wheat treatment groups that provided NaHCO 3 compared to NaCl or water (9 versus 26) there was still a substantial number of lambs that had access to NaHCO 3 that showed signs of clinical acidosis (16%), indicating that lambs cannot completely eliminate acidosis problems through feed selection. Another study by Phy and Provenza (1998a) examined what effect eating a meal of rapidly fermentable feed had on the preference for rapidly fermentable feed later on and whether NaHCO 3 and lasalocid influenced this preference change. Lambs fed a lower amount (400 g) of rolled barley for a meal exhibited equal preference for rolled barley and alfalfa pellets (52 and

49 33 48% of total intake respectively) during the next 4 h. However, when a higher amount (1,200 g) of rolled barley was fed the lambs increased their preference for alfalfa pellets over rolled barley (71 and 29% of total intake respectively) during this same time (Phy and Provenza, 1998a). Even though the lambs being fed large amounts of fermentable carbohydrates seemed to adjust their diet preference to maintain rumen health, they were unsuccessful, as all animals in that treatment exhibited signs of clinical acidosis (diarrhea). In the second part of this study, it was determined that lambs had higher intakes when barley was mixed with NaHCO 3 and lasalocid than when barley was offered with NaHCO 3 only, lasalocid only, or no additives (Phy and Provenza, 1998a). These results show that lambs will increase their intake of rapidly fermentable carbohydrates when it is offered with an additive that helps attenuate acidosis. There are only a couple of studies that have examined effects of ruminal acidosis on feed or diet selection in dairy cattle. In one such study, Keunen et al. (2002) studied effects of SARA on the preference of long alfalfa hay versus alfalfa pellets. Four cows were fed a ration that had 25% of ad libitum DMI of TMR replaced with wheat/barley pellets (50% ground wheat and 50% ground barley) for 2 wk (separated by 1 control wk). To determine feed preference cows were offered long alfalfa hay and alfalfa pellets 2 times per d for 30 min each. The preference ratio (amount of hay consumed/ amount of hay + pellets consumed) during the SARA wk was higher than the control wk (0.85 versus 0.60; Keunen et al., 2002). These results suggest that dairy cows will change their feed preference during a bout of SARA to attempt to maintain rumen health. However, because of the design of this experiment, the feed preferences of these cows were only measured for 1 h per d. DeVries et al. (2008) completed another study that examined feed selection and SARA in dairy cattle. This study again used a rumen challenge model to induce SARA, but the challenge consisted of 1 d of feed restriction to 50% of ad libitum intake followed the next morning by 4 kg of ground barley/wheat and then ad libitum access to TMR for the remainder of the d. This model

50 34 was repeated to produce 2 periods. The ration sorting activity of 2 groups of cows, early lactation cows fed a low forage diet and mid-lactation cows fed a high forage diet [high risk (HR) and low risk (LR) to acidosis respectively] were then compared to their sorting activity prior to the rumen challenge. The rumen challenge successfully induced SARA and decreased daily mean rumen ph in HR cows from 5.88 to 5.56 and from 6.25 to 5.88 in LR cows (Dohme et al., 2008). Before the rumen challenge, cows of both groups would sort against long particles (> 19.0mm) and fine particles (< 1.18mm) and sort for medium particles (> 8.0 mm), HR cows sorted against short particles (> 1.18 mm) while LR cows sorted for them (DeVries et al., 2008). In addition HR cows sorted their rations to a greater degree than LR cows. After the rumen challenge, cows in both groups changed their sorting behavior; HR cows generally increased their sorting for medium particles and against short and fine particles, and exhibited no change in sorting long particles, while LR cows exhibited variable responses with sorting activity changing with d and period (DeVries et al., 2008). DeVries et al. (2008) therefore suggested that dairy cattle will alter their ration sorting behavior during a bout of SARA in order to maintain ruminal health. However these results are very difficult to interpret, first because of the many interactions that occur. There was a significant d group effect for long particles, d period effects for all particle fractions, and d group period effects for long, short, and fine particle fractions. In addition, the TMR fed in this study utilized pelleted grain which was contained in the medium particle fraction; therefore the composition of this particle fraction was vastly different than most other studies where the medium fraction is composed mainly of forages. Increasing sorting for the medium particle fraction may not help attenuate acidosis because it contains both forages (fiber) and grain pellets (highly fermentable carbohydrates). There is evidence that a cyclical pattern of intake can occur when ruminants eat grain; when high levels of VFA from starch digestion are produced they can cause malaise (feeling of general discomfort) that decreases intake (Huber, 1976; Provenza, 1995). And while ruminants

51 35 prefer nutritionally dense foods like grains, they will decrease their intake of grains and increase their intake of other feeds when they ingest too much grain (Britton and Stock, 1986; Ortega- Reyes et al., 1992). In fact, animals experiencing malaise will increase the variety of feeds consumed in order to help attenuate their discomfort (Provenza et al., 1994). Feedback mechanisms that alert the ruminant of positive or negative consequences resulting from eating certain feeds are not fully understood. Feedback is possibly related to changes in rumen content composition or blood plasma variables (Keunen et al., 2002). Huber (1976) stated that since hydrogen ion receptors have not been demonstrated to exist in rumen mucosa, there are 3 possible mechanisms for rumen stasis following ruminal digesta acidification: involvement of hydrogen ion receptors elsewhere in the gastrointestinal tract, central inhibition by absorbed acid, and inhibition by absorbed amines or toxins. Provenza et al. (1994) studied mechanisms that allow for postingestive feedback to influence feeding behavior by examining the effects of feeding antiemetic drugs on feed aversions in sheep. Antiemetics are drugs that are effective against vomiting and nausea; the antiemetics used in this study were diphenhydramine hydrochloride, metoclopramide monohydrochloride, and crystalline dexamethasone, which were dosed as a mixture to increase their effectiveness (Provenza et al., 1994). There were 4 treatments in this study, antiemetics plus LiCl (A+L), antiemetics only (A), LiCl only (L), and a control (C). Lithium chloride was included in this study because it is known to induce a malaise that is similar to that caused by excessive ingestion of many compounds (Provenza, 1995). These 4 treatments were applied to 3 different feeds separately in 3 different experiments; feeds were oat grain, wheat grain, and milo. In all 3 cases the feed intakes of the 2 treatments that included antiemetics were higher than the 2 that did not (A+L > L and A > C), also the feed intakes of the treatments that included LiCl were lower than the treatments that did not (A > A+L and C > L) (Provenza et al., 1994). The authors suggested that the reason A consistently had higher intakes than C was because the amount of highly fermentable carbohydrates ingested from grains was able to cause

52 36 mild malaise without the addition of LiCl. Based on these results Provenza et al. (1994) concluded that LiCl and grain overload stimulate the emetic system which induces internal malaise and therefore reduce feed intake. There is substantial evidence in the literature that feed and diet selection is a very complex mechanism and that ruminants have some sort of feedback mechanism(s) in order to maintain ruminal health. Ruminants prefer feeds that maintain macronutrient balance, diminish toxicosis, and attenuate acidosis over feeds that strictly provide high energy intake (Phy and Provenza, 1998b). The overriding principle of ruminant diet selection is probably best summed up by Kyriazakis et al. (1999) who stated that diet selection should be considered within a framework of feeding behavior that views both feed intake and diet selection as an outcome of the animal s internal state and knowledge of the feeding environment. Conclusions A thorough review of the FPS literature leads to the following conclusions: that the general consensus among dairy industry professionals and researchers that ration sorting in lactating dairy cows negatively impacts milk production and components and ruminal fermentation is not supported by the results of the majority of studies reporting ration sorting; that the effects of FPS on DMI, ruminal fermentation, digestibility of DM, NDF, and starch, and milk production and components is very variable and inconsistent; and that RFC has a larger influence on the former variables than FPS and that using a measure that combines FPS and RFC would provide a more accurate and consistent way to predict the affects of diet on animal performance.

53 37 References Allen, M. S Relationship between fermentation acid production in the rumen and the requirement for physically effective fiber. J. Dairy Sci. 80: Allen, M. S Effects of diet on short-term regulation of feed intake by lactating dairy cattle. J. Dairy Sci. 83: American Society of Agricultural and Biological Engineers Method of determining and expressing particle size of chopped forage materials by screening. ANSI/ASAE. S424.1: Bailey, C. B. 1961a. Saliva secretion and its relation to feeding in cattle. Br. J. Nutr. 15: Bailey, C. B. 1961b. Saliva secretion and its relation to feeding in cattle. Br. J. Nutr. 15: Bailey, C. B., and C. C. Balch Saliva secretion and its relation to feeding in cattle. Br. J. Nutr. 15: Balch, C. C Proposal to use time spent chewing as an index of the extent to which diets for ruminants possess the physical property of fibrousness characteristic of roughages. Br. J. Nutr. 26: Beauchemin, K. A., L. Eriksen, P. Norgaard, and L. M. Rode Short Communication: Salivary secretion during meals in lactating dairy cattle. J. Dairy Sci. 91: Beauchemin, K. A., L. M. Rode, and M. V. Eliason Chewing activities and milk production of dairy cows fed alfalfa as hay, silage, or dried cubes of hay or silage. J. Dairy Sci. 80: Beauchemin, K. A., and W. Z. Yang Effects of physically effective fiber on intake, chewing activity, and ruminal acidosis for dairy cows fed diets based on corn silage. J. Dairy Sci. 88: Bhandari, S. K., S. Li, K. H. Ominski, K. M. Wittenberg, and J. C. Plaizier Effects of the chop lengths of alfalfa silage and oat silage on feed intake, milk production, feeding behavior, and rumen fermentation of dairy cows. J. Dairy Sci. 91: Brent, B. E Relationship of acidosis to other feedlot ailments. J. Anim. Sci. 43: Britton, R. A., and R. A. Stock Acidosis, rate of starch digestion and intake. Pages in Proceedings of the Feed Intake Symposium, Oklahoma Agricultural Experiment Station, Norman, OK. Calberry, J. M., J. C. Plaizier, M. S. Einarson, and B. W. McBride Effects of replacing chopped alfalfa hay with alfalfa silage in a total mixed ration on production and rumen conditions of lactating dairy cows. J. Dairy Sci. 86:

54 38 Castle, M. E., W. C. Retter, and J. N. Watson Silage and milk production: comparisons between grass silage of three different chop lengths. Grass and Forage Science. 34: Cooper, S. D., I. Kyriazakis, and J. D. Oldham The effects of physical form of feed, carbohydrate source, and inclusion of sodium bicarbonate on the diet selections of sheep. J. Anim. Sci. 74: Cotanch, K. W., and R. J. Grant Measuring physically effective fiber on-farm to predict cow response. Pages in Proceedings of the Cornell Nutrition Conference for Feed Manufacturers, Cornell University, Ithaca, NY. Counotte, G. H. M., A. T. van't Klooster, J. van der Kuilen, and R. A. Prins An analysis of the buffer system in the rumen of dairy cattle. J. Anim. Sci. 49: DeVries, T. J., K. A. Beauchemin, and M. A. G. von Keyserlingk Dietary forage concentration affects the feed sorting behavior of lactating dairy cows. J. Dairy Sci. 90: DeVries, T. J., F. Dohme, and K. A. Beauchemin Repeated ruminal acidosis challenges in lactating dairy cows at high and low risk for developing acidosis: Feed sorting. J. Dairy Sci. 91: DeVries, T. J., M. A. G. von Keyserlingk, and K. A. Beauchemin Frequency of feed delivery affects the behavior of lactating dairy cows. J. Dairy Sci. 88: Dohme, F., T. J. DeVries, and K. A. Beauchemin Repeated ruminal acidosis challenges in lactating dairy cows at high and low risk for developing acidosis: Ruminal ph. J. Dairy Sci. 91: Evans, E. W., G. R. Pearce, J. Burnett, and S. L. Pillinger Changes in some physical characteristics of the digesta in the reticulo-rumen of cows fed once daily. Br. J. Nutr. 29: Felton, C. A., and T. J. DeVries Effect of water addition to a total mixed ration on feed temperature, feed intake, sorting behavior, and milk production of dairy cows. J. Dairy Sci. 93: Forbes, J. M., and I. Kyriazakis Food preferences in farm animals: why don't they always choose wisely? Proceedings of the Nutrition Society. 54: Froetschel, M. A Effect of abomasal infusion of saliva on reticular motility and ruminal liquid contents of steers. J. Dairy Sci. 78: Harmon, D. L., R. A. Britton, R. L. Prior, and R. A. Stock Net portal absorption of lactate and volatile fatty acids in steers experiencing glucose-induced acidosis or fed a 70% concentrate diet ad libitum. J. Anim. Sci. 60: Heinrichs, A. J. and P. J. Kononoff Evaluating particle size of forages and TMRs using the new Penn State forage particle separator. Penn State Extension Circular DAS

55 Herrera-Saldana, R. E., J. T. Huber, and M. H. Poore Dry matter, crude protein, and starch degradability of five cereal grains. J. Dairy Sci. 73: Hosseinkhani, A., T. J. DeVries, K. L. Proudfoot, R. Valizadeh, D. M. Veira, and M. A. G. von Keyserlingk The effects of feed bunk competition on the feed sorting behavior of close-up dry cows. J. Dairy Sci. 91: Huber, T. L Physiological effects of acidosis on feedlot cattle. J. Anim. Sci. 43: Huntington, G. B Nutritional problems related to the gastro-intestinal tract. Pages in The Ruminant Animal: Digestive Physiology and Nutrition. D. C. Church, ed. Waveland Press, Inc., Long Grove, IL. Keunen, J. E., J. C. Plaizier, L. Kyriazakis, T. F. Duffield, T. M. Widowski, M. I. Lindinger, and B. W. McBride Effects of a subacute ruminal acidosis model on the diet selection of dairy cows. J. Dairy Sci. 85: Kononoff, P. J., and A. J. Heinrichs. 2003a. The effect of corn silage particle size and cottonseed hulls on cows in early lactation. J. Dairy Sci. 86: Kononoff, P. J., and A. J. Heinrichs. 2003b. The effect of reducing alfalfa haylage particle size on cows in early lactation. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and D. R. Buckmaster. 2003a. Modification of the Penn State forage and total mixed ration particle separator and the effects of moisture content on its measurements. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman. 2003b. The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86: Krause, K. M., and D. K. Combs Effects of forage particle size, forage source, and grain fermentability on performance and ruminal ph in midlactation cows. J. Dairy Sci. 86: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002a. Effects of forage particle size and grain fermentability in midlactation cows. I. Milk production and diet digestibility. J. Dairy Sci. 85: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002b. Effects of forage particle size and grain fermentability in midlactation cows. II. Ruminal ph and chewing activity. J. Dairy Sci. 85: Krause, K. M., and G. R. Oetzel Understanding and preventing subacute ruminal acidosis in dairy herds: A review. Anim. Feed Sci. Technol. 126: Krebs, J. R., and H. McCleery Optimisation in behavioural ecology. Page 91 in Behavioural Ecology - An Evolutionary Approach. 2nd ed. J. R. Krebs and N. B. Davies, ed. Blackwell Science Ltd, Oxford, England. 39

56 40 Kyriazakis, I., B. J. Tolkamp, and G. Emmans Diet selection and animal state: an integrative framework. Proceedings of the Nutrition Society. 58: Lammers, B. P., D. R. Buckmaster, and A. J. Heinrichs A simple method for the analysis of particle sizes of forage and total mixed rations. J. Dairy Sci. 79: Leonardi, C., and L. E. Armentano Effect of quantity, quality, and length of alfalfa hay on selective consumption by dairy cows. J. Dairy Sci. 86: Leonardi, C., and L. E. Armentano Short Communication: Feed selection by dairy cows fed individually in a tie-stall or as a group in a free-stall barn. J. Dairy Sci. 90: Leonardi, C., F. Giannico, and L. E. Armentano. 2005a. Effect of water addition on selective consumption (sorting) of dry diets by dairy cattle. J. Dairy Sci. 88: Leonardi, C., K. J. Shinners, and L. E. Armentano. 2005b. Effect of different dietary geometric mean particle length and particle size distribution of oat silage on feeding behavior and productive performance of dairy cattle. J. Dairy Sci. 88: Maulfair, D. D., M. Fustini, and A. J. Heinrichs Effect of varying total mixed ration particle size on rumen digesta and fecal particle size and digestibility in lactating dairy cows. J. Dairy Sci. 94: Maulfair, D. D., and A. J. Heinrichs Technical note: Evaluation of procedures for analyzing ration sorting and rumen digesta particle size in dairy cows. J. Dairy Sci. 93: Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of feed sorting on chewing behavior, production, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 93: Mertens, D. R Creating a system for meeting the fiber requirements of dairy cows. J. Dairy Sci. 80: Mertens, D. R Particle size, fragmentation index, and effective fiber: Tools for evaluating the physical attributes of corn silage. Pages in Proceedings of the Four-State Dairy Nutrition and Management Conference, Wisconsin Agri-Service Association, Madison, WI. Miller-Cushon, E. K., and T. J. DeVries Effect of dietary dry matter concentration on the sorting behavior of lactating dairy cows fed a total mixed ration. J. Dairy Sci. 92: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. Nocek, J. E Bovine acidosis: Implications on laminitis. J. Dairy Sci. 80: Oetzel, G. R., K. V. Nordlund, and E. F. Garrett Effect of ruminal ph and stage of lactation on ruminal lactate concentration in dairy cows. J. Dairy Sci. 82 (Suppl. 1):38 (Abstr.).

57 Ortega-Reyes, L., F. D. Provenza, C. F. Parker, and P. G. Hatfield Drylot performance and ruminal papillae development of lambs exposed to a high concentrate diet while nursing. Small Ruminant Research. 7: Oshita, T., K. Nonaka, S. Kume, and T. Nakui Effects of forage type on particle size distribution of ruminal digesta and faeces of non-lactating cows fed high quality forage. Livest. Prod. Sci. 91: Phy, T. S., and F. D. Provenza. 1998a. Eating barley too frequently or in excess decreases lambs' preference for barley but sodium bicarbonate and lasalocid attenuate the response. J. Anim. Sci. 76: Phy, T. S., and F. D. Provenza. 1998b. Sheep fed grain prefer foods and solutions that attenuate acidosis. J. Anim. Sci. 76: Plaizier, J. C Replacing chopped alfalfa hay with alfalfa silage in barley grain and alfalfabased total mixed rations for lactating dairy cows. J. Dairy Sci. 87: Poppi, D. P., R. E. Hendricksen, and D. J. Minson The relative resistance to escape of leaf and stem particles from the rumen of cattle and sheep. Journal of Agricultural Science, UK. 105:9 14. Poppi, D. P., D. J. Minson, and J. H. Ternouth Studies of cattle and sheep eating leaf and stem fractions of grasses. 3. The retention time in the rumen of large feed particles. Aust. J. Agric. Res. 32: Poppi, D. P., B. W. Norton, D. J. Minson, and R. E. Hendticksen The validity of the critical size theory for particles leaving the rumen. J. Agric. Sci. (Camb.). 94: Provenza, F. D., L. Ortega-Reyes, C. B. Scott, J. J. Lynch, and E. A. Burritt Antiemetic drugs attenuate food aversions in sheep. J. Anim. Sci. 72: Provenza, F. D Postingestive feedback as an elementary determinant of food preference and intake in ruminants. Journal of Range Management. 48:2 17. Stone, W. C The effect of subclinical rumen acidosis on milk components. Pages in Proceedings of the Cornell Nutrition Conference for Feed Manufacturers, Cornell University, Ithaca, NY. Stone, W. C Nutritional approaches to minimize subacute ruminal acidosis and laminitis in dairy cattle. J. Dairy Sci. 87:E13 E26. Sudweeks, E. M., L. O. Ely, D. R. Mertens, and L. R. Sisk Assessing minimum amounts and form of roughages in ruminant diets: Roughage value index system. J. Anim. Sci. 53: USDA Dairy 2007, Part 1: Reference of dairy cattle health and management practices in the United States, No. N USDA-APHIS-VS, CEAH, Fort Collins, CO. 41

58 Van Soest, P. J Nutritional Ecology of the Ruminant. 2nd ed. Comstock Publishing, Ithaca, NY. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, W. Z., and K. A. Beauchemin Effects of physically effective fiber on digestion and milk production by dairy cows fed diets based on corn silage. J. Dairy Sci. 88: Yang, W. Z., and K. A. Beauchemin. 2006a. Increasing the physically effective fiber content of dairy cow diets may lower efficiency of feed use. J. Dairy Sci. 89: Yang, W. Z., and K. A. Beauchemin. 2006b. Physically effective fiber: Method of determination and effects on chewing, ruminal acidosis, and digestion by dairy cows. J. Dairy Sci. 89: Yang, W. Z., and K. A. Beauchemin Altering physically effective fiber intake through forage proportion and particle length: Digestion and milk production. J. Dairy Sci. 90: Yang, W. Z., K. A. Beauchemin, and L. M. Rode. 2001a. Barley processing, forage:concentrate, and forage length effects on chewing and digesta passage in lactating cows. J. Dairy Sci. 84: Yang, W. Z., K. A. Beauchemin, and L. M. Rode. 2001b. Effects of grain processing, forage to concentrate ratio, and forage particle size on rumen ph and digestion by dairy cows. J. Dairy Sci. 84: Zebeli, Q., D. Mansmann, H. Steingass, and B. N. Ametaj Balancing diets for physically effective fibre and ruminally degradable starch: A key to lower the risk of sub-acute rumen acidosis and improve productivity of dairy cattle. Livest. Sci. 127:

59 43 Table 2-1. Physical effectiveness factors (pef) for NDF in feeds of each physical form classification based on total chewing activity in relation to that elicited by long grass hay. From Mertens, D. R Creating a system for meeting the fiber requirements of dairy cows. J. Dairy Sci. 80:

60 44 Figure 2-1. Effect of the ratio between physically effective NDF (pendf 1.18 ) to ruminally degradable starch from grains (RDSG) in the diet on daily mean ruminal ph Ruminal ph = *peNDF:RDSG ratio, if pendf:rdsg ratio < 1.45 ± 0.22, asymptotic plateau of ph = 6.20; root mean square error = 0.15; R2 = 0.41, P < (variables were plotted based on a meta-analysis conducted from 45 studies with a total of 134 different experimental diets). From Zebeli, Q., D. Mansmann, H. Steingass, and B. N. Ametaj Balancing diets for physically effective fibre and ruminally degradable starch: A key to lower the risk of sub-acute rumen acidosis and improve productivity of dairy cattle. Livest. Sci. 127:1 1.

61 Chapter 3 Eating Behavior, Ruminal Fermentation, and Milk Production in Lactating Dairy Cows Fed Rations That Varied in Dry Alfalfa Hay and Alfalfa Silage Content Abstract The objective of this experiment was to evaluate effects of various inclusion levels of dry chopped alfalfa hay and alfalfa silage in lactating dairy cow rations on eating behavior, rumen fermentation, milk yield and components. A second objective of this study was to compare results of the Penn State and Ro-Tap particle separators for the same TMR samples and to determine effects of separation method on particle size distribution. Eight multiparous Holstein cows (79 ± 11 d in milk initially; 647 ± 36 kg body weight) were randomly assigned to a replicated 4 4 Latin square design. During each of the 4 periods, cows were fed 1 of 4 diets that were chemically similar but varied in dry chopped alfalfa hay level. Forage dry matter (DM) content of each ration consisted of 50% corn silage and 5, 10, 20, or 40% dry chopped alfalfa hay. The remaining forage DM content was alfalfa silage (45, 40, 30, and 10% respectively). It was determined that there were minimum effects on sorting early in the d and no effects 4 h after feeding and later with increasing alfalfa hay content. Dry alfalfa hay was included in rations up to 23.5% of ration DM with no negative effects on DM intake, milk yield, and rumen fermentation. Small decreases in milk fat and protein content occurred with increasing dry hay inclusion. Despite changes in total mixed ration refusal particle size distribution throughout the d, by 24 h after feeding no significant ration sorting occurred when measured either by selection indices or actual consumption of various particle size fractions (> 19.0, > 8.0, > 1.18 mm, and pan). Data from the Penn State and Ro-Tap particle separators produced different particle size distributions

62 46 from the same sample. This indicates that data obtained from these 2 methods of particle separation are not directly comparable and that the method of particle separation should be considered when interpreting experimental results. Key Words: chewing, particle size, rumination, sorting Introduction Ration sorting has generally been considered a concern for lactating dairy cow health and feeding. It is believed that ration sorting can lead to SARA because cows usually sort against longer particles and for shorter particles (Leonardi and Armentano, 2003; Kononoff et al., 2003b; DeVries et al., 2007). This type of sorting behavior could lead to decreased NDF intake and physical effectiveness of the diet while starch intake is increased. A decrease in effective fiber can be especially detrimental to high producing dairy cows being fed energy dense rations that rely on longer fiber to increase chewing and saliva secretion to help buffer their rumen (Nocek, 1997; Allen, 1997; Krause et al., 2002). However, Maulfair et al. (2010) determined that drastic ration sorting, when determined by changes in TMR refusal particle size distributions, can occur in diets without any negative effects on milk production, milk components, and rumen fermentation under certain feeding conditions. The authors suggested that the actual consumption of particle size fractions, NDF, and starch should be considered when measuring ration sorting. Therefore there is a need to study ration sorting in greater detail to understand what factors interact to cause negative effects in the cow and develop methods to limit these effects. A main component of forage particle size research is the particle separating equipment. The Penn State particle separator (PSPS) was developed as an inexpensive and easy to use device to characterize particle size distribution of TMR and forages in the field (Lammers et al., 1996; Kononoff et al., 2003a). The PSPS has been increasingly used in research to describe particle size

63 47 distribution of treatment diets and for estimation of physically effective NDF (pendf) by using the proportion of samples particles retained above the 1.18-mm sieve multiplied by their NDF content. The PSPS uses as-fed samples and a horizontal shaking motion to separate the particles. This is in contrast to the Ro-Tap particle separator (RTPS) which uses dried samples and vigorous vertical shaking to separate particles. The RTPS is important to forage particle size research because Mertens (1997) used it to develop the laboratory assessment of pendf, where particles retained on a 1.18-mm sieve after shaking are multiplied by the sample NDF content. One major factor that creates a difference between the PSPS and the RTPS is that vertical shaking tends to separate particles by their minimum cross-sectional dimension, whereas horizontal shaking tends to separate particles by their length (Mertens, 1997; Mertens, 2005). Another factor that could cause different results between these 2 separators are sample drying. Drying can cause particles to shrink and increase their fragility causing them to break; both of which will decrease particle size distributions (Kononoff et al., 2003a). Therefore it is important to understand how the data from these 2 methods of particle separation differ so that their results may be interpreted accurately. The objective of this experiment was to study effects of replacing alfalfa haylage with dry chopped alfalfa hay in the ration on sorting activity and to determine effects on ruminal fermentation, milk production, or milk composition. In addition, a second objective of this study was to compare results of the PSPS and RTPS for the same TMR samples and to determine effects of separation method on particle size distribution.

64 48 Materials and Methods Diets, Cows, and Experimental Design Cows used in this experiment were cared for and maintained according to a protocol approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Eight (4 rumen cannulated) lactating, multiparous, Holstein cows (79 ± 11 DIM initially; 647 ± 36 kg BW) were randomly assigned to a replicated 4 4 Latin square design. There were 4 periods of 21 d, 13 d of adaptation and 8 d of sample collection. Cows were fed 1 of 4 rations each period that were chemically similar and varied only in concentration of chopped alfalfa hay (replacing alfalfa haylage). Dry alfalfa inclusion rates were 5, 10, 20, or 40% of forage DM, representing 2.9, 5.8, 11.7, and 23.5% of total ration DM. Ration ingredients, other than dry chopped alfalfa hay and alfalfa silage, remained similar for all diets except the 40% hay diet. This diet had a decreased amount of canola meal and 0.5% of urea added to maintain similar levels of rumen degradable protein among all rations. Cows were housed in individual tie-stalls in a mechanically ventilated barn and milked twice/d at 0700 and 1900 h. They were fed once/d at 0730 h for ad libitum consumption and a 10% refusal rate to allow maximum opportunity to sort. Feed was pushed up 3 times/d at 1230, 1730, and 2400 h. Rations were balanced to meet or exceed NRC (2001) requirements for cows producing 38.5 kg of milk/d containing 3.75% fat and 3.07% true protein assuming a DMI of 23.9 kg/d and water was available for ad libitum consumption. Feed, Refusal, and Particle Size Analysis Offered TMR and refusals were weighed daily for the duration of the study. On d 20 and 21 of each period feed bunk contents were weighed and sampled at 0, 2, 4, 8, 12, 16, and 24 h

65 49 after feeding to determine particle size distribution and DM content of remaining feed. Particle size distributions of samples were determined with the PSPS according to Kononoff et al. (2003a). Samples were then dried in a forced air oven at 55 C for 48 h to determine DM content. Samples of each TMR and forage were collected on d 16 and 19 of each period, composited by period and analyzed by Cumberland Valley Analytical Services, Inc. (Hagerstown, MD) for CP (AOAC, 2000), ADF (AOAC, 2000), NDF (Van Soest et al., 1991), ash (AOAC, 2000), NFC (Van Soest et al., 1991), and NE L (NRC, 2001). There were 2 procedures used to calculate pendf: pendf 8.0 = % of particles > 8.0 mm NDF of whole sample (top 2 sieves of PSPS) and pendf 1.18 = % of particles > 1.18 mm NDF of whole sample (top 3 sieves of PSPS; Kononoff et al., 2003a). The RTPS was used to separate 95 dried TMR samples comprised of the 4 treatment diets and 0 and 24 h time points to compare to the results of the PSPS. Approximately 0.6 L of dried sample were placed on the top of the sieve stack, which contained sieves of: 9.5, 8.0, 6.7, 4.75, 3.35, 2.36, 1.70, 1.18, 0.60, and 0.15 mm. The RTPS was run for 10 min and particles retained on each sieve were then weighed to determine the proportion of sample DM retained on each sieve. Chewing Activity On d 14 to 18 of each period, eating and rumination behavior were recorded using Institute of Grassland and Environmental Research Behavior Recorders and Graze Jaw Movement Analysis Software (Ultra Sound Advice, London, UK) as described by Rutter (1997; 2000). Chewing was measured for all cows for two 24-h periods including while cows were being milked. These recorders analyze jaw movements of cattle, and the software determines eating or ruminating chews based on amplitude and frequency of jaw movements. This procedure was validated for use with cows housed in tie-stalls by Kononoff et al. (2002).

66 50 Rumen Sampling Rumen sampling was conducted on d 15 of each period at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18, 21.5, and 24.5 h after feeding (Kononoff et al., 2003b). Samples were taken from dorsal, ventral, cranial, caudal, and medial areas of the rumen, mixed thoroughly, and then filtered through 4 layers of cheesecloth. Rumen liquid ph was immediately determined using a handheld ph meter (phtestr 10 BNC, Oakton, Vernon Hills, IL). Approximately 15 ml of filtered liquid was placed into bottles containing 3 ml of 25% metaphosphoric acid and 3 ml of 0.6% 2- ethylbutyric acid (internal standard) and stored at -20 C. After thawing, samples were centrifuged 3 times at 4000 g for 30 min at 4 C to obtain a clear supernatant and were analyzed for NH 3 using a phenol-hypochlorite assay (Broderick and Kang, 1980) and VFA concentration using gas chromatography (Yang and Varga, 1989). Milk Production Milk production was recorded daily and milk samples were taken on d 20 and 21 (4 consecutive milkings). Samples were collected and preserved using 2-bromo-2-nitropropane-1,3 diol. Milk samples were analyzed for fat, true protein, lactose, MUN, and SCC by the Dairy One milk testing laboratory (State College, PA) using infrared spectrophotometry (Foss 605B Milk- Scan; Foss Electric, Hillerod, Denmark). Statistical Analyses Statistical analyses were conducted using PROC MIXED of SAS (Version 9.2, SAS Institute, Cary, NC). Dependent variables were analyzed as a 4 4 Latin square design. All

67 51 denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997) and repeated measurements for rumen samples were analyzed using first-order autoregressive covariance structure (Littell et al., 1998), as well as terms for time and interaction of treatment by time. Because of unequally spaced rumen sampling, weighted mean daily rumen ph, NH 3, and VFA concentrations were determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). A selection index based on refusals was calculated for each of the 4 particle size fractions at 2, 4, 8, 12, 16, and 24 h after feeding. This index was calculated as the actual intake of each fraction (Y i ) expressed as a percentage of the expected intake. Expected intake of Y i equals intake multiplied by the fraction of Y i in the TMR fed (Leonardi and Armentano, 2003). Values > 1.0 indicate cows were sorting for the particle fraction and values < 1.0 indicate cows were sorting against the particle fraction. Sorting indices were calculated using both the expected intake since time point 0 h (cumulative) and the expected intake since the previous time point (interval). The 95% confidence limits were used to determine if selection index was significantly different from 1.0. Chewing behavior and meal criterion was analyzed using the procedure of Maulfair et al. (2010). The data used for calculating the sieve size in the PSPS that is equivalent to the 1.18-mm sieve in the RTPS were natural log transformed to correct for abnormal distribution and improve model fit, which included terms for separator, ration, period, d, time, sieve size, sieve size 2, separator by sieve size, and separator by sieve size 2 along with a random effect of cow. All data are presented as least squares means and treatment effects are considered significant when P 0.05 and a trend when 0.05 < P Means separation tests were conducted using the protected least significant differences (PDIFF) procedure, with significance at P 0.05.

68 52 Results and Discussion Chemical Composition and Particle Size Distribution The chemical compositions and particle size distributions of the forages included in the rations of this study are shown in Table 3-1. Alfalfa silage was replaced with dry chopped alfalfa hay in this study and there were differences in their chemical and physical properties. Dry hay naturally had much higher DM content than silage, 90.4 and 40.0% respectively. This difference led to the treatment TMR differing in DM content as well, but to a lesser degree. Dry chopped alfalfa hay was lower in CP, higher in ADF and NDF, and approximately equal in NFC compared to alfalfa silage. A much greater proportion of particles were retained on the 19.0 mm sieve and the pan for alfalfa hay than silage (4.4 and 4.3 times more for 19.0 mm sieve and pan respectively). Approximately 88% of alfalfa silage particles were retained on the middle 2 screens, compared to approximately 43% of hay particles. Since alfalfa hay had higher fiber levels but fewer particles greater than 8.0 and 1.18 mm than alfalfa silage, pendf values between the 2 forages remained similar. Hay had higher pendf values but was only 17 and 13% higher than silage for pendf 8.0 and pendf 1.18 respectively. Ingredients, chemical compositions, and particle size distributions of the treatment diets used in this study are shown in Table 3-2. Ration DM numerically increased with increasing hay inclusion and was significantly higher in the 40% hay diet as a result of higher DM of hay versus silage, though there was only a 7.1% maximum variation among the diets. Crude protein, NDF, and NFC were not different among diets and averaged 17.9, 34.5, and 36.6% of DM respectively. Forage NDF was increased slightly in the 40% hay ration because of the increased NDF content of the hay over the silage. The pendf values of treatment diets showed mixed results; pendf 1.18 was not different among rations and averaged 29.6% of DM, while pendf 8.0 decreased slightly with increasing alfalfa hay inclusion

69 53 (from 15.1 to 13.7% of DM). Particle size distributions of the treatments were varied; both the 19.0 mm and pan particle fractions generally increased with increasing dry hay inclusion, while the 8.0 mm fraction decreased and the 1.18 mm fraction remained similar among rations. Ration Sorting Refusal particle size distributions for each treatment over the course of 1 d are displayed in Figure 3-1 as the changes in each particle fraction (> 19.0, > 8.0, > 1.18 mm, and pan) over 24 h. The treatments exhibited similar patterns over time in each of the 4 particle fractions. The particles retained on the 19.0 mm sieve decreased after feeding for 3.5 h and then remained stable for the rest of the d; the 40% hay diet had significantly higher values than the other diets for the entire d. All treatments showed a gradual increase in particles retained on the 8.0 mm sieve; the 40% hay was significantly lower than the other treatments and increased at a greater rate. The particles retained on the 1.18 mm screen increased slightly throughout the d and were similar among treatments. Finally, particles retained in the pan decreased in all rations over 24 h; the 40% hay diet started with a higher proportion of these particles but it decreased at a faster rate and was approximately equal to the other rations by 16 h after feeding. Cumulative selection indices for each treatment throughout the d were calculated, and it was determined that at 2 h after feeding, cows in the 40% hay treatment were sorting against particles retained on the 19.0-mm sieve while cows on all other treatments were sorting for these particles. By 4 h after feeding cows on all treatments were sorting for these larger particles and for the remainder of the d did not significantly sort for or against this particle fraction. Particles retained on the 8.0-mm sieve were sorted against by cows on the 20% hay treatment and were not sorted for or against by the other treatments at h 2. For the remainder of the d all treatments were similar and not different from 1.0. The 1.18-mm particle fraction was sorted against by cows on

70 54 the 10% diet at 2 h and on the 5% diet at 4 h; there was no sorting for the other treatments and times. Finally, particles < 1.18 mm were sorted for by cows in the 20% hay treatment at h 2 and were not sorted for or against for the other treatments and times. The selection indices standard errors were quite large due to sorting variation between cows, which made finding significant differences between treatments and from 1.0 difficult. Large variation in ration sorting between animals was also reported by Leonardi and Armentano (2003) and Leonardi et al. (2005), especially in the longer particle fractions. At the end of the d (24 h after feeding) average selection indices for each particle fraction were very close to 1.0: 1.04, 0.98, 0.99, and 1.04 for the 19.0, 8.0, 1.18-mm sieves and pan respectively. This indicates that animals consumed rations that were very close in particle distribution to their offered TMR. These results are in agreement with Maulfair et al. (2010) where it was determined that despite large changes in refusal particle size distribution, by the end of the d ration sorting was not significant when measured via selection indices. In this study ration DM content increased with increasing dry hay inclusion but it had minimal effects on ration sorting behavior during the first 4 h of the d and did not affect sorting later in the d. Leonardi et al. (2005) determined that increasing ration DM increased sorting activity of lactating cows; however, their ration DM increased from 64.4 to 80.8%. Diets in the current study had much less variation (difference between treatments was 7.1 versus 16.4 percentage units) and all had DM contents lower than the lowest DM diet in Leonardi et al. (2005). Two recent studies (Miller-Cushon and DeVries, 2009; Felton and DeVries, 2010) determined that increasing ration DM content actually decreased sorting in lactating cows. The DM content of diets used in these studies ranged from 47.9 to 57.6% and 44.1 to 56.3% for Miller-Cushon and DeVries (2009) and Felton and DeVries (2010) respectively. Miller-Cushon and DeVries (2009) added water to the mixer during diet preparation at a rate of 20% of the diet (DM basis). Felton and DeVries (2010) first mixed the TMR for 10 min and then transferred the diets to a feed cart where water was added at rates of 20 and 44% of the diet (DM basis). The

71 55 authors suggested that summer temperatures experienced during these studies caused heating of the rations that had added water, which contributed to increased sorting behavior. Ration DM contents of diets used in the current study fell in the range of DM where increased sorting was seen with decreasing ration DM; however, the current study took place in November through February and no additional water was added to the diets, potentially increasing stability compared to diets in the cited studies. Intake of DM and Particle Fractions Dry matter intake was not different among treatments and averaged 27.8 kg/d (Table 3-3). Feed efficiency was also similar among diets and averaged 1.41 kg 3.5% FCM/kg DMI. Figure 3-2 shows the cumulative percentage of total daily intake for each ration. There were no differences among treatments and since the cows were fed only once/d, DMI was heavily skewed toward feeding time with the treatments averaging 21.9, 30.5, 49.0, and 70.3% of their daily intake consumed by 2, 4, 8, and 12 h after feeding respectively. Another measure of ration sorting activity, that was determined to be more accurate than refusal particle size distribution (Maulfair et al., 2010), is individual consumption of each particle size fraction. Table 3-4 shows the kg of each particle fraction (> 19.0, > 8.0, > 1.18 mm, and pan) consumed by various time points (2, 4, 8, 12, 16, and 24 h after feeding) throughout the d. These data show that at 24 h after feeding consumption of particles retained on the 19.0-mm sieve and the pan increased with increasing alfalfa hay content. Also, particles retained on the 8.0-mm sieve decreased with increasing hay inclusion while consumption of particles retained on the 1.18-mm sieve were similar across treatments. Trends seen in particle fraction intake are the same trends seen in particle size distributions of the offered TMR; where particles on the 19.0-mm sieve and pan increased, particles on the 8.0-mm sieve decreased and particles on the 1.18-mm sieve did not

72 56 change with increasing dry hay content (Table 3-2).The intake of particle fractions matching changes in particle size distributions of the offered TMR reinforces the conclusion that ration sorting was not significant at the end of the d. Chewing Activity Table 3-5 shows ruminating, eating, and total chewing behavior of the cows in this study. No significant treatment effects on ruminating, eating, and total chewing min/d were found and they averaged (7.9 h), (7.2 h), and (14.9 h) min/d respectively. However there was a tendency for eating time to differ (P = 0.09) as eating time for the 40% diet was higher than for the 10% diet. The reason that larger differences between the treatments for ruminating and eating were not seen is probably because while particles > 19.0 mm increased by 4.9 percentage units with increasing alfalfa hay content, this change was associated with a 10.9 percentage unit decrease in the 8.0-mm particle fraction and a 4.4 percentage unit increase in particles < 1.18 mm. These changes in ration particle size distribution may have effectively canceled each other out, causing chewing behavior to be similar across treatments. Rumen Characteristics There were no differences in rumen ph among treatments; average daily weighted mean, minimum, and maximum ph was 6.29, 5.87, and 7.00 for all treatments (Table 3-6). Figure 3-3 shows rumen ph for each treatment over the course of 1 d. Rumen ph for all treatments decreased immediately after feeding to a nadir between 11.5 and 18.0 h after feeding and then increased to pre-prandial levels. Rumen NH 3 concentration was also not different among treatments for daily weighted mean (averaged 12.1 mg/dl) and minimum; however, maximum

73 57 daily NH 3 concentration showed a trend to increase with increasing dry hay inclusion (Table 3-6). This result is the effect of the inclusion of urea in the 40% hay diet. Figure 3-4 shows rumen NH 3 concentrations for each treatment over the course of 1 d. In all treatments, rumen NH 3 concentrations increased sharply immediately after feeding and peaked 1.5 h later; NH 3 then gradually decreased for the remainder of the d to pre-prandial levels. Finally, daily weighted mean concentrations of acetate, butyrate, valerate, and isobutyrate were not different among rations. There was some minor variation in propionate and isovalerate concentrations among rations that are not practically significant (Table 3-6). Milk Production and Composition Milk production averaged 38.7 kg/d and did not differ among treatments (Table 3-3). Fat, protein, and lactose yields were also not different among treatments; however, fat and protein concentration decreased slightly with increasing dry hay inclusion (3.65 to 3.46% and 3.04 to 2.98% for fat and protein respectively). Finally, MUN concentrations were elevated for the 40% hay treatment, possibly a product of higher maximum rumen NH 3 concentrations (Table 3-6) caused by this treatment being the only one supplemented with urea. Penn State Versus Ro-Tap Particle Separator Particle size distributions of 95 TMR samples from this study separated with both PSPS and RTPS are shown in Table 3-7. Both methods of particle separation produced results that showed significant differences for diet and time effects; however, the actual proportions of particles retained on the various sieves differ dramatically. The large differences between these 2 methods of particle separation can easily been seen in Figure 3-5. The RTPS retained 66.8% of

74 58 particles above the 1.18-mm sieve compared to 88.2% for the PSPS (a 32% increase). This large difference is of importance because generally when calculating pendf the NDF content of the sample is multiplied by this value. Assuming a sample NDF content of 35%, there would be an increase of 7.5 percentage units in the calculated value of pendf when going from the RTPS to PSPS (23.4 to 30.9% pendf respectively). This difference could lead to diet formulations that over estimate pendf and lead to problems associated with lack of fiber. There was an even greater difference in particles retained above an 8.0-mm sieve; the PSPS retained almost 9 times more particles above this threshold than the RTPS. Using natural log transformed data it was determined that for all samples tested, 76.8% (71.9, 82.1; 95% confidence limits) of particles were retained above a 1.18-mm screen in the RTPS. The equivalent screen size in the PSPS that would achieve 76.8% (69.6, 84.8; 95% confidence limits) of particles oversized for these samples was determined to be 5.78 mm. It is clear from these results that the RTPS allows particles to pass through much smaller diameter sieves than the PSPS, and results determined via these separators are not comparable. Since these results are only based on 4 different TMR that were similar in composition and particle size distribution, more study of these methods should be conducted comparing a greater variety of TMR and forages of various types in order to determine the correlation of results from these 2 systems. Conclusions Increasing inclusion of dry chopped alfalfa hay from 10 to 40% of forage DM in a corn silage and alfalfa haylage based TMR had minimal effects on sorting during the first 4 h after feeding and did not change sorting activity of cows later in the d. By 24 h after feeding ration sorting was insignificant when measured via either sorting indices or actual consumption of various particles size fractions, despite changes in TMR refusal particle size distributions over the

75 59 course of the d. It was determined that lactating cows can be fed TMR containing dry chopped alfalfa hay levels up to 23.5% of ration DM without any negative effects on rumen fermentation and milk production. It was also determined that the PSPS and the RTPS produced very different particle size distributions for the same sample, and therefore results from these 2 methods of particle separation should not be used interchangeably. However, more research should be conducted on a greater variety of TMR and also unmixed forages to achieve a more accurate comparison between these 2 methods. Acknowledgements This research was supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture. References Allen, M. S Relationship between fermentation acid production in the rumen and the requirement for physically effective fiber. J. Dairy Sci. 80: Association of Official Analytical Chemists Official Methods of Analysis. 17th ed. AOAC, Arlington, VA. Broderick, G. A., and J. H. Kang Automated simultaneous determination of ammonia and total amino acids in ruminal fluid and in vitro media. J. Dairy Sci. 63: DeVries, T. J., K. A. Beauchemin, and M. A. G. von Keyserlingk Dietary forage concentration affects the feed sorting behavior of lactating dairy cows. J. Dairy Sci. 90: Felton, C. A., and T. J. DeVries Effect of water addition to a total mixed ration on feed temperature, feed intake, sorting behavior, and milk production of dairy cows. J. Dairy Sci. 93: Gaines, W. L The energy basis of measuring milk yield in dairy cows. Illinois Agr. Expt. Sta. Bull. 308.

76 60 Kenward, M. G., and J. H. Roger Small sample inference for fixed effects from restricted maximum likelihood. Biometrics. 53: Kononoff, P. J., A. J. Heinrichs, and D. R. Buckmaster. 2003a. Modification of the Penn State forage and total mixed ration particle separator and the effects of moisture content on its measurements. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman. 2003b. The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86: Kononoff, P. J., H. A. Lehman, and A. J. Heinrichs Technical Note--A comparison of methods used to measure eating and ruminating activity in confined dairy cattle. J. Dairy Sci. 85: Krause, K. M., D. K. Combs, and K. A. Beauchemin Effects of forage particle size and grain fermentability in midlactation cows. II. Ruminal ph and chewing activity. J. Dairy Sci. 85: Lammers, B. P., D. R. Buckmaster, and A. J. Heinrichs A simple method for the analysis of particle sizes of forage and total mixed rations. J. Dairy Sci. 79: Leonardi, C., and L. E. Armentano Effect of quantity, quality, and length of alfalfa hay on selective consumption by dairy cows. J. Dairy Sci. 86: Leonardi, C., F. Giannico, and L. E. Armentano Effect of water addition on selective consumption (sorting) of dry diets by dairy cattle. J. Dairy Sci. 88: Littell, R. C., P. R. Henry, and C. B. Ammerman Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci. 76: Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of feed sorting on chewing behavior, production, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 93: Mertens, D. R Creating a system for meeting the fiber requirements of dairy cows. J. Dairy Sci. 80: Mertens, D. R Particle size, fragmentation index, and effective fiber: Tools for evaluating the physical attributes of corn silage. Pages in Proceedings of the Four-State Dairy Nutrition and Management Conference, Wisconsin Agri-Service Association, Madison, WI. Miller-Cushon, E. K., and T. J. DeVries Effect of dietary dry matter concentration on the sorting behavior of lactating dairy cows fed a total mixed ration. J. Dairy Sci. 92: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. Nocek, J. E Bovine acidosis: Implications on laminitis. J. Dairy Sci. 80:

77 61 Rutter, S. M Graze: A program to analyze recordings of the jaw movements of ruminants. Behav. Res. Meth. Ins. C. 32: Rutter, S. M., R. A. Champion, and P. D. Penning An automatic system to record foraging behaviour in free-ranging ruminants. Appl. Anim. Behav. Sci. 54: Shipley, R. A., and R. E. Clark Tracer Methods for In Vivo Kinetics. Academic Press, New York, NY. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, C.-M. J., and G. A. Varga Effect of three concentrate feeding frequencies on rumen protozoa, rumen digesta kinetics, and milk yield in dairy cows. J. Dairy Sci. 72:

78 62 Table 3-1. Chemical compositions and particle size distributions determined for corn silage, alfalfa haylage, and dry chopped alfalfa hay Item Corn Silage Alfalfa Haylage Alfalfa Hay Composition, % of DM DM, % CP, % ADF, % NDF, % pendf 1 8.0, % pendf , % Ash, % NFC, % NE L, Mcal/kg Particle size, as-fed % retained 19.0 mm mm mm Pan Physically effective NDF 8.0 = % of particles > 8.0 mm NDF of whole sample; top 2 sieves of Penn State particle separator (Kononoff et al., 2003a). 2 Physically effective NDF 1.18 = % of particles > 1.18 mm NDF of whole sample; top 3 sieves of Penn State particle separator (Kononoff et al., 2003a).

79 63 Table 3-2. Ingredients, chemical compositions, and particle size distributions for TMR with increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) Item SEM P-value Ingredients, % of DM Corn silage Alfalfa haylage Alfalfa hay Ground corn Canola meal Roasted soybeans Mineral/ vitamin mix Heat-treated soybean meal Urea 0.5 Composition, % of DM DM, % 49.1 b 49.7 b 51.6 b 56.2 a 1.40 < 0.01 CP ADF NDF Forage NDF 25.5 b 25.8 b 26.5 b 27.9 a 0.82 < pendf a 13.6 b 14.5 ab 13.7 b pendf Ash NFC NE L, Mcal/kg Particle size, as-fed % retained 19.0 mm 4.6 b 3.9 b 6.0 b 9.5 a 1.28 < mm 39.2 a 38.2 a 35.1 a 28.1 b 1.60 < mm 43.4 b 44.5 ab 44.9 ab 45.3 a Pan 12.8 b 13.4 b 14.0 b 17.2 a a b Means within a row with different superscripts differ (P 0.05). 1 Physically effective NDF 8.0 = % of particles > 8.0 mm NDF of whole sample; top 2 sieves of Penn State particle separator (Kononoff et al., 2003a). 2 Physically effective NDF 1.18 = % of particles > 1.18 mm NDF of whole sample; top 3 sieves of Penn State particle separator (Kononoff et al., 2003a).

80 64 Table 3-3. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on DMI, feed efficiency, and milk production and components Item SEM P-value DMI, kg/d Milk yield, kg/d % FCM, kg/d Feed efficiency Fat, % 3.65 a 3.60 ab 3.53 ab 3.46 b Fat, kg/d Protein, % 3.04 a 3.03 ab 3.01 ab 2.98 b Protein, kg/d Lactose, % Lactose, kg/d MUN, mg/dl 16.0 b 16.4 b 15.8 b 17.9 a 0.68 < 0.01 SCC, 1,000 cells/ml a b Means within a row with different superscripts differ (P 0.05) % FCM = (milk kg) (fat kg); (Gaines, 1928). 2 Feed efficiency = 3.5% FCM / DMI

81 65 Table 3-4. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on intake of 4 particle size fractions (> 19.0, > 8.0, > 1.18, and < 1.18 mm) Intake, kg SEM P-value 19.0 mm 2 h h 0.67 b 0.55 b 1.10 ab 1.46 a h 0.82 b 0.68 b 1.29 ab 1.67 a h 0.98 b 0.91 b 1.54 ab 2.26 a 0.26 < h 1.13 b 0.98 b 1.69 b 2.49 a 0.26 < h 1.28 b 1.12 b 1.81 b 2.74 a 0.26 < mm 2 h h h h 7.30 a 7.06 a 6.68 a 5.04 b h 8.49 a 8.36 a 7.94 a 6.04 b h a a 9.56 a 7.46 b 0.63 < mm 2 h h h h h h Pan 2 h 0.39 b 0.66 b 0.66 b 1.26 a h 0.70 b 0.95 ab 1.13 ab 1.44 a h 1.63 b 1.77 b 1.83 b 2.60 a 0.22 < h 2.71 b 2.77 b 3.10 ab 3.65 a h 3.07 b 3.11 b 3.50 b 4.26 a 0.22 < h 3.63 b 3.77 b 4.03 b 4.90 a 0.22 < 0.01 a b Means within a row with different superscripts differ (P 0.05).

82 66 Table 3-5. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on chewing behavior Item, min/d SEM P-value Ruminating Eating ab b ab a Total chewing ab b ab a a b Means within a row with different superscripts differ (P 0.05).

83 67 Table 3-6. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen fermentation Item SEM P-value Rumen ph Weighted mean Minimum Maximum NH 3, mg/dl Weighted mean Minimum Maximum 20.8 b 25.8 ab 22.7 ab 28.3 a VFA weighted mean, µm/ml Acetate Propionate 25.6 a 27.5 b 26.8 ab 26.6 ab Butyrate Valerate Isovalerate 2.36 ab 2.59 a 2.28 ab 2.16 b Isobutyrate a b Means within a row with different superscripts differ (P 0.05). 1 Weighted averages determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972).

84 68 Table 3-7. Particle size distributions of TMR containing 5, 10, 20, and 40% of forage DM as dry chopped alfalfa hay in samples taken at feeding (0 h) and 24 h after feeding and separated with the Penn State and Ro-Tap particle separators Percentage of P-value DM retained 0 h 24 h 0 h 24 h 0 h 24 h 0 h 24 h SEM Diet Time Penn State particle separator 19.0 mm < 0.01 < mm < 0.01 < mm < 0.01 Pan < 0.01 < 0.01 Ro-Tap particle separator 9.5 mm < 0.01 < mm < 0.01 < mm < mm < mm < mm < 0.01 < mm < 0.01 < mm < mm mm < 0.01 Pan < 0.01

85 69 A B

86 70 C D Figure 3-1. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on refusal particle size distribution for 19.0 (A), 8.0 (B), 1.18 mm (C) sieves, and pan (D).

87 71 Figure 3-2. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on cumulative percent of diet daily intake at various times after feeding.

88 Figure 3-3. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen ph over time.

89 73 Figure 3-4. Effect of feeding increasing levels of dry chopped alfalfa hay (5, 10, 20, and 40% of forage DM) on rumen NH 3 over time.

90 74 Figure 3-5. Particle size distributions of TMR samples separated with the Penn State (PSPS) and Ro-Tap particle separators divided into particle fractions; > 19.0, > 8.0, > 1.18 mm.

91 Chapter 4 Effects of Varying Forage Particle Size and Fermentable Carbohydrates on Feed Sorting, Ruminal Fermentation, and Milk and Component Yields of Dairy Cows Abstract Ration sorting is thought to negatively affect ruminal fermentation and yield of milk and components. However, the influence of ruminally degradable starch on ration sorting has not been studied. Therefore the objective of this experiment was to study the interactions between forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) for ration sorting, ruminal fermentation, chewing activity, and milk yield and components. In this study 12 (8 ruminally cannulated) multiparous, lactating Holstein cows were fed TMR that varied in FPS and RFC. Two lengths of corn silage were used to alter FPS and 2 grind sizes of corn grain were used to alter RFC. It was determined that increasing RFC increased ruminating time and did not affect eating time, while increasing FPS increased eating time and did not affect ruminating time. Ruminal fermentation did not differ by altering either FPS or RFC. However, increasing FPS tended to increase mean and maximum ruminal ph and increasing RFC tended to decrease minimum ruminal ph. Particle size distribution and NDF content of refusals increased over time while starch content decreased; indicating that cows were sorting against physically effective NDF and for RFC. Selection indices determined that virtually no interactions occurred between FPS and RFC and that despite significant sorting throughout the d, by 24 h after feeding cows had consumed a ration very similar to what was offered. This theory was reinforced by particle fraction intakes that very closely resembled the proportions of particle fractions in the offered TMR. An interaction between FPS and RFC was seen for DMI, as DMI decreased with

92 76 increasing FPS when the diet included low RFC and did not change when the diet included high RFC and DMI increased with RFC for long diets and did not change with RFC on short diets. Increasing RFC was found to increase milk yield, milk protein content and yield, and lactose content and yield but decrease milk fat content. Increasing FPS did not have as great an impact on milk production as RFC. This study found that there was no significant interaction between FPS and RFC for ration sorting although there was an interaction between FPS and RFC for DMI. RFC had a greater influence on milk yield and components than FPS, but neither affected ruminal fermentation. Key Words: forage particle size, ruminally fermentable starch, sorting Introduction Ration sorting is thought to increase cows susceptibility to SARA. Cows will generally sort for finer particles and against longer particles in their rations, which effectively decreases their fiber intake while increasing their starch intake as fiber and starch are positively and negatively associated, respectively, with longer particles in dairy cow rations (Leonardi and Armentano, 2003; Leonardi et al., 2005). However, Maulfair et al. (2010) showed drastic increases in refusal particle size distribution and NDF content and decreases in starch content throughout the d, the classical determinants of ration sorting, yet found no negative effects on ruminal fermentation and milk production when cows were fed a ration that contained about 34 and 27% of ration DM as NDF and starch content respectively. Ruminally fermentable carbohydrates (RFC) may influence the effective fiber requirement of dairy cows. Yang et al. (2001) suggested that ruminal ph and SARA cannot be predicted using only physical characteristics of rations, because RFC has a greater influence on ruminal ph than forage particle size (FPS). Krause et al. (2002b) determined that the physical

93 77 effectiveness of forages is affected by other dietary components such as corn grain moisture and fermentability. Finally, Krause and Combs (2003) found that significant interactions between FPS and RFC existed for ruminal fermentation and milk production, which indicates that effects of FPS and RFC are not always additive and complicates the formulation of dairy rations. None of these studies measured or determined ration sorting when studying the interaction between FPS and RFC; FPS had been shown to have major influence on ration sorting (Leonardi and Armentano, 2003; Kononoff and Heinrichs, 2003; Kononoff et al., 2003b), but effects of RFC on ration sorting have not been studied. Therefore the objective of this experiment was to study the interactions between FPS and RFC for ration sorting, ruminal fermentation, chewing activity, and milk yield and components. Material and Methods Diets, Cows, and Experimental Design Cows used in this research were cared for and maintained according to a protocol approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Twelve lactating (8 ruminally cannulated), multiparous, Holstein cows averaging 115 ± 49 DIM, weighing 662 ± 64 kg, and with parity of 3.08 ± 0.79 (mean ± SD) were studied. The experimental design consisted of 3 replicated, balanced 4 4 Latin squares with treatments arranged in a 2 2 factorial design; 2 squares were composed of ruminally cannulated cows. Cows were assigned to squares by parity and randomly assigned to 1 of 4 treatments. Treatments were designed to study the effects of 2 lengths of FPS and 2 levels of RFC. Treatment diets varied in FPS by feeding either long (LCS) or short corn silage (SCS) and RFC were varied by feeding either dry cracked corn (CC) or dry fine ground corn (FC). The 4 treatment diets then

94 78 consisted of LCS + CC (LC), LCS + FC (LF), SCS + CC (SC), and SCS + FC (SF). Except for altering corn silage and grain particle size, the 4 treatment diets contained identical ingredients and proportions. Diet ingredients and their percentage of ration DM were: corn silage (42.6), dry ground corn (22.2), alfalfa haylage (15.4), canola meal (9.4), roasted split soybeans (7.1), mineral/vitamin mix (2.5), salt (0.4), and Optigen (Alltech, Nicholasville, KY; 0.4). The study consisted of 4 21-d periods consisting of 14 d of adaptation followed by a 7-d collection period. Corn silage hybrid was Pioneer 34M78 (Pioneer Hi-Bred International, Inc., Johnston, IA) that was planted on 4/19/2010 and harvested on 8/30/2010. Corn silage was harvested with a John Deere 6750 forage harvester (John Deere, Moline, IL) equipped with a kernel processor set at approximately 6.35 mm. The harvester cutterhead used 16 knives (maximum capacity is 48 knives) with the length-of-cut transmission at its highest setting to produce a theoretical length of cut of 47.1 mm. After harvesting, corn silage was ensiled in an Ag-Bag (Ag-Bag, St. Nazianz, WI) and allowed to ferment for 62 d before beginning the study. Corn silage that was removed from the Ag-Bag and mixed into TMR without further processing was considered LCS. A cutand-throw type, single row, forage harvester that was modified to operate on a trailer and be fed manually with a 25 horsepower V-Twin small gas engine was used to reduce the particle size of corn silage to produce SCS. Corn silage was rechopped twice through the custom forage chopper on a daily basis to minimize the chemical variance between LCS and SCS. Dry corn was ground through a Roskamp roller mill (California Pellet Mill Co., Crawfordsville, IN) to produce the CC used in this study. This corn was then ground further with a Case International 1250 grindermixer (Case IH, Racine, WI) using a 3.18 mm screen to produce FC. Diets were mixed separately using an I. H. Rissler model 1050 TMR mixer (E. Rissler Mfg. LLC, New Enterprise, PA). Animals were housed in individual stalls in a mechanically ventilated barn, milked twice/d at 0500 and 1700 h, and fed once/d at approximately 0800 h for ad libitum consumption. Feed refusals were weighed daily and the amount of TMR fed was adjusted daily to maintain a

95 79 10% refusal rate. Feeding once/d at a 10% refusal rate was designed to allow cattle to have increased opportunity to sort rations. Feed was pushed up 3 times/d at 1230, 1730, and 2400 h. Rations were balanced to meet or exceed NRC (2001) requirements for cows producing 52.2 kg of milk/d containing 3.75% fat and 3.07% true protein assuming a DMI of 29.5 kg/d and water was available for ad libitum consumption. Chewing Activity Eating and rumination behavior were recorded on d 15 to 21 of each period, using Institute of Grassland and Environmental Research Behavior Recorders and Graze Jaw Movement Analysis Software (Ultra Sound Advice, London, UK) as described by Rutter (1997; 2000). Chewing was measured for all 12 cows for 2 24-h periods including while cows were being milked. These recorders analyze jaw movements of cattle, and the software determines eating or ruminating chews based on the amplitude and frequency of jaw movements. This procedure was validated for use with cows housed in tie-stalls by Kononoff et al. (2002). Rumen Parameters Ruminal contents were collected from dorsal, ventral, cranial, caudal, and medial areas of the rumens of all 8 ruminally cannulated cows on d 20 of each period at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding (Kononoff et al., 2003b). At each ruminal sampling collected digesta was mixed thoroughly and then separated into 2 equal subsamples. One digesta subsample was strained through 2 layers of cheesecloth. Rumen fluid ph was immediately determined using a handheld ph meter (HI 98121, HANNA Instruments Inc., Woonsocket, RI). Strained ruminal fluid (15 ml) was placed into bottles containing 3 ml of 25%

96 80 metaphosphoric acid and 3 ml of 0.6% 2-ethylbutyric acid (internal standard) and stored at approximately -20 C. After thawing, samples were centrifuged 3 times at 4000 g for 30 min at 4 C to obtain a clear supernatant and were analyzed for VFA concentration using gas chromatography (Yang and Varga, 1989). The second ruminal digesta subsample was utilized for particle size distribution and DM analysis using the procedure of Maulfair et al. (2011) except that samples were not processed in duplicate; particle fractions determined were soluble, > 0.15, > 0.6, > 1.18, > 3.35, > 6.7, and > 9.5 mm. Soluble fraction of samples were calculated as the DM lost during sieving and drying. Data were analyzed using each particle fraction as a percentage of DM retained on 0.15-mm screen (retained) and also as the percentage of DM of the entire sample sieved (total). Finally, rumens of the cannulated cows were completely emptied on d 21 of each period at 5 h after feeding. The weight and volume of ruminal digesta was then recorded, and digesta was sampled for DM analysis. Digesta was then immediately returned to the rumen of each cow. Feed, Refusal, and Particle Size Analysis Feed bunk contents for each animal were weighed and sampled on d 18 and 19 of each period at 0, 8, 16, and 24 h after feeding for DM and particle size analysis. All samples were sieved in the American Society of Agriculture and Biological Engineers (ASABE) forage particle separator, which can determine 6 particle fractions (> 26.9, > 18.0, > 8.98, > 5.61, > 1.65, and < 1.65 mm; screen diagonals; ASABE, 2007). Whole samples were then placed in a forced air oven at 65 C for 48 h to determine DM content. Samples of forages, ground corn, and TMR were taken on d 18 and 9 of each period, composited by period and analyzed by Cumberland Valley Analytical Services, Inc. (Hagerstown, MD) for CP (AOAC, 2000), ADF (AOAC, 2000), NDF (Van Soest et al., 1991), ash (AOAC, 2000), NFC (Van Soest et al., 1991), and NE L (NRC,

97 ). Starch and NDF contents of forages, ground corn, and TMR (at 0 and 24 h after feeding) were determined by drying in a forced-air oven at 65 C for 48 h and grinding (0.5- and 1.0-mm screen for starch and NDF respectively; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ). Starch was then analyzed via the procedure reported by Zanton and Heinrichs (2009) and NDF was analyzed using heat-stable α-amylase and Na 2 SO 3 according to Van Soest (1991). Particle size distributions of forages and TMR were determined via sieving with the ASABE forage particle separator (ASABE, 2007). To determine particle size distributions of ground corn, samples were placed on a series of stacked sieves (sizes 0.15, 0.425, 0.60, 0.85, 1.18, 1.70, 2.36, 3.35, 4.75, and 6.7 mm; VWR, Arlington Heights, IL) contained in a Retsch AS 200 Control sieve shaker (Retsch, Haan, Germany) and were sieved for 10 min at 2.5 mm amplitude. Particles retained on each sieve were then weighed to determine their proportion of total sample DM. There were 2 procedures used to calculate physically effective NDF (pendf): pendf 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of the Penn State particle separator) and pendf 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of the Penn State particle separator; Kononoff et al., 2003a). Corn grain fermentability was determined via in situ bags incubated in quadruplicate in the rumen of 2 lactating cows (each cow incubated 2 bags of each sample for each time point) for 0.5, 1, 2, 4, 6, 8, 12, 16, 24, and 48 h. Approximately 7 g of samples were sealed in nylon bags (10 20 cm, 50 μm pore size; ANKOM, Macedon, NY) attached to a string that was anchored to the rumen cannulae and weighted to locate the bags centrally in the rumen. After removal from the rumen, bags were rinsed in cold water by hand until water was almost clear. Bags were then dried in a forced-air oven at 65 C for 48 h and then weighed to determine remaining DM.

98 82 Milk Production Milk production was recorded daily and milk samples were taken on d 20 and 21 (4 consecutive milkings). Samples were collected and preserved using 2-bromo-2-nitropropane-1,3 diol. Milk samples were analyzed for fat, true protein, lactose, MUN, and SCC by the Dairy One milk testing laboratory (State College, PA) using infrared spectrophotometry (Foss 605B Milk- Scan; Foss Electric, Hillerod, Denmark). Fecal Sampling Fecal samples were taken from all 12 cows at the same time points as rumen sampling (d 20 at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18, 21.5, and 24.5 h after feeding) via grab samples from the rectum. Samples were stored at -20 C until later determination of DM and particle size distribution. Particle size of samples was determined using the same wet sieving technique used for rumen digesta, with the exception of eliminating the top screen (9.5 mm). Geometric mean particle length (X gm ) and standard deviation of particle length (S gm ) were calculated according to ASABE (2007) procedure. X gm was calculated using 2 procedures; the first, retained X gm (X gm Ret), only considered particles retained on the 0.15-mm screen or larger, the second procedure, total X gm (X gm Tot), considered all particle fractions including the soluble fraction that passed through the bottom screen (0.15 mm). Mean particle length of the soluble fraction was assumed to be mm, which is half of the diagonal screen diameter (0.212 mm) of the bottom screen; this is the assumption that ASABE (2007) uses for mean length of particles on the pan. Subsamples were also placed in a forced air oven at 65 C for 48 h to determine DM content.

99 83 Statistical Analyses Statistical analysis was conducted using PROC MIXED of SAS (Version 9.2, SAS Institute, Cary, NC). Dependent variables were analyzed as a 4 4 Latin square design. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997) and repeated measurements for ruminal ph, VFA, and NH 3 concentrations and ground corn DM disappearance were analyzed using the first order autoregressive covariance structure (Littell et al., 1998) as well as terms for time and interaction of treatment by time. Because of unequally spaced rumen and fecal sampling, the weighted mean daily ruminal ph, VFA, and NH 3 concentrations and ruminal digesta and fecal particle size distribution were determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). Area under the curve for the SARA thresholds of 5.8 and 5.5 were also calculated using the trapezoidal rule (Shipley and Clark, 1972). A selection index based on refusals was calculated for each of the 6 particle size fractions at 8, 16, and 24 h after feeding. This index was calculated as the actual intake of each fraction (Y i to pan) expressed as a percentage of the expected intake. Expected intake of Y i equals intake multiplied by the fraction of Y i in the fed TMR (Leonardi and Armentano, 2003). Sorting indices were calculated using both the expected intake since time point 0 (cumulative) and the expected intake since the previous time point (interval). Values > 1.0 indicate cows were sorting for the particle fraction and values < 1.0 indicate cows were sorting against the particle fraction. The 95% confidence limits were used to determine if a selection index was significantly different from 1.0. Chewing behavior and meal criteria was analyzed using the procedure of Maulfair et al. (2010). All data are presented as least squares means and treatment effects are considered significant when P 0.05 and a trend when 0.05 < P Means separation tests were conducted using the protected least significant differences (PDIFF)

100 84 procedure with significance at P 0.05 and are reported when P-value of FPS RFC interaction Results and Discussion Chemical Composition and Particle Size Distribution of Diets Particle size distributions and chemical compositions of forages used in this study are shown in Table 4-1. There was a large difference in particle size distribution between LCS and SCS. When separated with the ASABE particle separator, LCS had many more particles retained on 26.9 and 18.0 mm screens, equal particles on the 8.98 mm screen, and many fewer particles on 5.61 and 1.65 mm screens and the pan than SCS. The approximate equivalency of Penn State particle separator fractions to the ASABE screens are: top ( mm), middle (8.98 mm), lower ( mm), and pan (pan). The particle size distribution of alfalfa haylage was similar to SCS. Chemical compositions of the corn silages were similar and not practically different despite being statistically different for DM and NE L. Sampling error may be responsible for the small differences seen between LCS and SCS since they were taken from the same bag each d as a single batch, with part being re-chopped as the only difference. The pendf measures were, as expected, very different between corn silages, but there was a much greater difference for pendf 8.0 than for pendf The LCS was 1.81 and 1.15 times greater than SCS for pendf 8.0 and pendf 1.18 respectively. The particle size distribution of the corn silage before bagging was analyzed by taking 5 samples evenly spaced over the length of the bag. It was determined that the process of bagging and ensiling corn silage altered its particle size distribution; before ensiling, the proportions retained on each sieve were 30.5 ± 0.73, 24.7 ± 0.70, 24.0 ± 1.05, 9.0 ± 0.10, 8.5 ± 0.44, and 3.4 ± 0.39% (mean ± SEM) for the 26.9, 18.0, 8.98, 5.61,

101 mm sieves, and pan respectively. All particle fractions of the fresh forage were different from ensiled LCS (P < 0.05) except for the pan which tended to be different (P = 0.06). The particle size distributions, chemical compositions, and rates of disappearance for corn grains used in this study are shown in Table 4-2. The particle size distributions of CC and FC were different at all 11 particle fractions. The greatest differences occurred at screen sizes 2.36 mm and larger, where CC had 67.4% and FC had 5.6% of particles retained, and at screen sizes 1.18 mm and smaller, where CC had 18.4% and FC had 78.2% of particles retained. The chemical compositions of CC and FC were similar and not practically different despite being statistically different in DM and CP content. The rates of disappearance of CC and FC were different at every time point except 48 h (P-value = 0.15). The greatest differences between CC and FC were in the first 2 h of incubation, where FC had about 2.1 times more DM disappearance than CC. The disappearance of FC continued to be greater than CC at each time point (except 48 h), but differences between them decreased with increasing incubation time. These data should be interpreted with caution as the impact of eating and rumination on ground corn was not a factor in this analysis, and it is reasonable to assume that chewing would have a larger impact on CC because of its greater potential for further particle size reduction. Particle size distributions and chemical compositions of treatment TMR are shown in Table 4-3. Varying FPS and RFC altered the particle size distribution of diets. The 2 largest fractions were increased with increasing FPS while the 4 other fractions were affected by both FPS and RFC. Increasing FPS increased particles retained on the 8.98-mm sieve and decreased particles retained on the 5.61-, 1.65-mm sieves, and pan. Increasing RFC decreased particles retained on the and 5.61-mm sieves and increased particles on the 1.65-mm sieve and pan. Chemical compositions of TMR were similar and not practically different. The CP, NDF, forage NDF, and starch content of TMR were approximately 16.4, 31.9, 21.4, and 31.0% of DM respectively. The pendf measures were affected by both FPS and RFC effects; increasing FPS

102 86 and decreasing RFC increased both pendf measures. The greatest variation occurred with pendf 8.0, where LC was 2.20 times higher than SF (14.1 versus 6.4%). The LC diet was only 1.33 times higher than SF for pendf 1.18 (27.9 versus 21.0%). Chewing Behavior Ruminating min/d was shown to increase with RFC but was not affected by FPS, this increase was much larger for diets containing short FPS (Table 4-4). The ability of RFC to increase ruminating time may be counterintuitive, but this result was also seen by Krause et al. (2002b). The authors determined that increasing RFC, by replacing dry cracked shelled corn with high-moisture corn in an alfalfa silage diet, tended to increase (P = 0.08) ruminating min/d and increased (P = 0.03) ruminating min/kg of NDF intake. Krause et al. (2002b) suggested that since forage should be the only diet component that could stimulate rumination, the increase in ruminating activity is a result of an adaptive response by the animals to increased RFC to attenuate low ruminal ph via increased saliva secretion. Daily ruminating times varied from 5.9 to 6.7 h/d across treatments. Eating min/d, in contrast to ruminating min/d, were not affected by RFC but increased with FPS. The effect of increased eating time with longer FPS is well known in the literature (Bailey, 1961; Beauchemin et al., 2008). Finally, total chewing time/d was not significantly affected by FPS or RFC, but an increase in either tended to increase total chewing min/d. These results conflict with those reported by Krause et al. (2002b) who found that increasing FPS increased both eating and ruminating min/d and that increasing RFC decreased eating min/d while increasing rumination min/d. These differences in the results might be related to how RFC was increased in the 2 studies; in the current study it was increased by decreasing grind size of dry corn grain whereas Krause et al. (2002b) increased RFC by replacing dry cracked corn with high-moisture corn. However, Krause and Combs (2003) found that increasing

103 87 FPS increased both eating and ruminating time and that RFC did not affect eating or ruminating and this study altered RFC the same way as Krause et al. (2002b). Ruminal Characteristics Daily weighted mean, minimum, and maximum ruminal ph did not differ by varying FPS or RFC; though there was a trend for weighted mean and maximum ph to increase with FPS and for minimum ph to decrease with increasing RFC (Table 4-5). Increasing FPS likely affected ruminal ph through the increased eating time it caused, which has been shown to increase saliva secretion (Bailey, 1961; Beauchemin et al., 2008) and increase ruminal buffering. The increased rumination time caused by increasing RFC did not have the same positive effect of increased eating time on ruminal ph, as ruminal ph actually tended to decrease with increasing RFC. This likely occurred because either increased saliva secretion was unable to compensate for increased RFC or increasing eating time was more effective at elevating saliva secretion than increasing ruminating time. These results conflict with several studies that showed that increasing RFC decreased mean ruminal ph (Yang et al., 2001; Krause et al., 2002b; Krause and Combs, 2003). Perhaps the methods of increasing RFC in these studies (replacing dry cracked with highmoisture corn or replacing coarsely rolled with flatly rolled barley grain) were more effective than the one used in the current study (replacing dry cracked with fine ground corn). Two of these same studies found that FPS had no effect on mean ruminal ph (Yang et al., 2001; Krause and Combs, 2003) and 1 found that ruminal ph increased with FPS (Krause et al., 2002b). Concentrations of NH 3 and lactate did not differ by altering FPS and RFC, though there tended to be an interaction of FPS and RFC for weighted mean NH 3 concentration because NH 3 increased with RFC for the short diets but decreased with increasing RFC for the long diets. Weighted mean concentrations of VFA were also not different when changing FPS or RFC, although there

104 88 were several trends. Increasing FPS tended to decrease acetate, butyrate, and isobutyrate while increasing RFC tended to increase valerate and decrease the acetate to propionate ratio. Also FPS and RFC tended to interact for propionate as it increased with RFC for short FPS diets but did not change for long FPS diets. Finally, ruminal digesta weight and volume were not affected by FPS or RFC. Intakes, Refusals, and Ration Sorting The TMR refusals were analyzed for NDF and starch at 0 and 24 h after feeding and for particle size distribution at 0, 8, 16, and 24 h after feeding. The TMR concentrations of starch and NDF are shown in Figures 4-1 and 4-2 respectively. Starch concentration was lower 24 h after feeding especially for diets that contained long FPS; while NDF concentrations tended to be increased at 24 h after feeding. This indicates that cows were generally sorting for concentrates and against fiber in all treatments. This theory is reinforced by the particle size distribution of TMR over time (Figure 4-3). Particles retained on the and 18.0-mm (data not shown) sieve showed very similar patterns over the course of the d for each diet; these particles increased in diets containing LCS with time after feeding and did not change in diets containing SCS. The particles retained on the and 5.61-mm sieves generally did not change over time for any diets (data not shown), and the amount of particles retained on these sieves relative to each diet remained constant as well. Particles retained on the 1.65-mm sieve decreased in diets containing LCS and did not change in diets containing SCS, while particles retained on the pan decreased in all diets over a 24 h period. These results show that cattle were effectively altering TMR composition (chemical and physical) through sorting, and that cows were able to sort to a higher degree on the long FPS diets.

105 89 Ration sorting was also evaluated via selection indices at 8, 16, and 24 h after feeding. Cumulative selection indices are shown in Table 4-6 and represent actual consumption of each particle fraction at various time points compared to estimated consumption if cows consumed all particles in the proportion offered in the original TMR. These results indicate that ration sorting was affected by FPS for the 26.9-mm particle fraction and by RFC for the and 5.61-mm particle fractions. Generally greater sorting activity was seen at earlier time points and longer particle fractions; no selection index was significantly different from 1.0 at 24 h after feeding for the 3 shortest particle fractions. When significantly affecting selection indices, increasing both FPS and RFC increased sorting activity, but the changes and degrees of sorting were relatively small. Selection indices were also analyzed on an interval basis, which compared actual consumption of each particle fraction at various time points to estimated consumption if cows consumed all particles in the proportion found in TMR at the previous time point (Table 4-7). The interval method allows for a clearer view of how sorting changed throughout the d. For example, cows on LC were sorting against particles retained on the 1.65-mm sieve during the first 8 h of the d, but then sorted for these particles during the last 16 h of the d. Using this method, FPS was much more likely to affect ration sorting than RFC (7 versus 1 significant particle fraction by time point effects). As the d progressed, cows on all treatments increased their sorting against particles retained on the 18.0-mm sieve and generally the 8.98-mm sieve. Cows being fed diets that contained LCS increased their sorting for particles retained on the 5.61-, 1.65-mm sieve, and pan as time after feeding increased. Clearly ration sorting was occurring in all treatments and at various times after feeding, but not to as great a degree as found in other studies (Maulfair et al., 2010). There was a significant interaction between FPS and RFC for DMI (Table 4-8). DMI decreased with increasing FPS when the diet included low RFC and did change when the diet included high RFC; DMI increased with RFC for the long diets and did not change with RFC on

106 90 the short diets. Effects of FPS and RFC on DMI have been variable in the literature: DMI increased with increasing RFC and was not affected by FPS in Yang et al. (2001); DMI decreased with increasing RFC and was not affected by FPS in Krause et al. (2002a); and DMI decreased with increasing RFC and increased with increasing FPS in Krause and Combs (2003). Total NDF intake was not affected by FPS or RFC but starch intake was affected by both. Daily starch intake followed the same pattern of interaction that was found with DMI, and these differences were probably a result of DMI variations. Daily intakes of each particle fraction are also shown in Table 4-8. Intakes of all particle fractions, except 8.98 mm, were affected by FPS, and RFC affected intakes of particle fractions 8.98 mm and smaller. Increasing FPS increased intake of particles retained on the and 18.0-mm sieves and decreased intake of particles retained on the and 1.65-mm sieves, and pan. Increasing RFC decreased intake of particles retained on the and 5.61-mm sieves and increased intake of particles retained in the 1.65-mm sieve and pan. These differences in particle fraction intakes are representative of the differences in their proportions in the offered TMR, indicating that ration sorting was not sufficient to cause differences between the consumed and offered rations. The percentage of total daily intake consumed by 8 and 16 h after feeding was determined and is also shown in Table 4-8. Cows on the longer FPS treatments had consumed a greater percentage of their daily intakes at both 8 and 16 h after feeding compared to cows fed shorter FPS; 62.5 versus 54.6% and 92.1 versus 86.6% of total daily intake for long and short FPS treatments at 8 and 16 h after feeding respectively. This is somewhat surprising as it has been shown that increasing FPS decreases eating rate (Bailey, 1961; Beauchemin et al., 2008). Finally, feed refusal rate was not different among treatments and was successfully managed to a rate of 10.1%.

107 91 Milk Yield and Composition Milk yield and composition data are reported in Table 4-9. Milk yield was not influenced by FPS but was increased with RFC and averaged 43.5 kg/d across treatments. Though increasing RFC increased milk yield it also decreased milk fat content (from 3.54 to 3.35% on average), and therefore 3.5% FCM only tended to be higher with higher RFC levels. Increasing RFC also increased milk protein concentration and yield and lactose concentration and yield. This is likely an effect of increasing available energy in the rumen, which can increase microbial protein synthesis and propionate production. These changes in milk yield and composition were not preceded by changes in ruminal fermentation as measured in this study. A possible explanation for this discrepancy is that only the concentrations of ruminal compounds were measured, and actual production and absorption of VFA and NH 3 are not known. Increasing FPS decreased milk protein concentration and increased MUN levels; this could be due to insufficient available energy for ruminal microorganisms to allow then to effectively utilize available NH 3. Feed efficiency increased with increasing FPS as a result of longer FPS decreasing DMI while maintaining 3.5% FCM. Fecal Particle Size Interestingly RFC has greater influence on fecal particle size distribution than FPS (Table 4-10). Corn silage length had no effect on any particle fractions and X gm when calculated using the retained method, though it did have effects on fecal particle size when calculated using the total method. Increasing RFC influenced virtually every particle fraction and X gm for both methods of calculation. Fecal particle size was decreased with increasing RFC. The reason for this is not clear although it is possible that increased ruminating time/d allowed for a greater

108 92 reduction in digesta particle size via mastication. The X gm Ret for low and high RFC diets averaged 1.49 and 1.15 mm, respectively. The X gm Ret of LF and SF were very close to the results of Maulfair et al. (2011) where the average X gm Ret across all treatments was 1.13 mm. Increasing RFC also decreased X gm Tot from 0.40 to 0.34 mm for low and high RFC diets, respectively. Again X gm Tot of LF and SF agreed with Maulfair et al. (2011) who reported and average of 0.32 mm X gm Tot across all treatments. The current study also determined that up to 6.6 and 3.5% of fecal particles were > 6.7 mm when measures via the retained and total methods, respectively. Particles > 3.35 mm (retained) averaged 26.5 and 16.2% for low and high RFC diets respectively, the latter of which again agreed with Maulfair et al. (2011) where 15.7% of fecal particles were > 3.35 mm (retained). These results agree with the suggestion of Maulfair et al. (2011) that the critical particle size for rumen escape is larger than the commonly held 1.18 mm. This study further suggests that since approximately 5% of fecal particles were retained on a 6.7-mm sieve, this size may be a more accurate estimate of the particle size threshold for increased resistance to ruminal escape. Conclusions It was determined that increasing RFC increased ruminating time and increasing FPS increased eating time. Ruminal fermentation was not affected by either FPS or RFC, though increasing FPS tended to increase mean and maximum ruminal ph and increasing RFC tended to decrease minimum ruminal ph. Refusal particle size distribution and NDF and starch content were observed to change over the course of the d and indicated that cows were sorting against pendf and for RFC. Analysis of selection indices revealed virtually no interaction between FPS and RFC occurred and despite significant sorting throughout the d, by 24 h after feeding cows had consumed a ration very similar to what was offered. This view of sorting was reinforced by

109 93 particle fraction intakes that very closely resembled the proportions of particle fractions in the offered TMR. An interaction between FPS and RFC was seen for DMI, as DMI decreased with increasing FPS when the diet included low RFC and did change when the diet included high RFC and DMI increased with RFC for the long diets and did not change with RFC on the short diets. Increasing RFC was found to increase milk yield, milk protein content and yield, and lactose content and yield but decrease milk fat content. Increasing FPS did not have as great an impact on milk production as RFC. This study therefore concludes that: there was not significant interaction between FPS and RFC for ration sorting although both affected it separately; RFC had greater influence on milk yield and components than FPS; neither FPS of RFC affected ruminal fermentation; and there was an interaction between FPS and RFC for DMI. Finally, it was determined that approximately 5% of fecal particles were greater than 6.7 mm and that this may be a more accurate estimate of the critical particle size for rumen escape in modern lactating dairy cows. Acknowledgements Sincere appreciation is extended to Growmark FS, LLC (Sangerfield, NY) for generously allowing the use of their modified forage harvester for the duration of this trial. This research was supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture. References American Society of Agricultural and Biological Engineers Method of determining and expressing particle size of chopped forage materials by screening. ANSI/ASAE. S424.1:

110 94 Association of Official Analytical Chemists Official Methods of Analysis. 17th ed. AOAC, Arlington, VA. Bailey, C. B Saliva secretion and its relation to feeding in cattle. Br. J. Nutr. 15: Beauchemin, K. A., L. Eriksen, P. Norgaard, and L. M. Rode Short Communication: Salivary secretion during meals in lactating dairy cattle. J. Dairy Sci. 91: Gaines, W. L The energy basis of measuring milk yield in dairy cows. Illinois Agr.Expt.Sta. Bull Kenward, M. G., and J. H. Roger Small sample inference for fixed effects from restricted maximum likelihood. Biometrics. 53: Kononoff, P. J., and A. J. Heinrichs The effect of corn silage particle size and cottonseed hulls on cows in early lactation. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and D. R. Buckmaster. 2003a. Modification of the Penn State forage and total mixed ration particle separator and the effects of moisture content on its measurements. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman. 2003b. The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86: Kononoff, P. J., H. A. Lehman, and A. J. Heinrichs Technical Note--A comparison of methods used to measure eating and ruminating activity in confined dairy cattle. J. Dairy Sci. 85: Krause, K. M., and D. K. Combs Effects of forage particle size, forage source, and grain fermentability on performance and ruminal ph in midlactation cows. J. Dairy Sci. 86: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002a. Effects of forage particle size and grain fermentability in midlactation cows. I. Milk production and diet digestibility. J. Dairy Sci. 85: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002b. Effects of forage particle size and grain fermentability in midlactation cows. II. Ruminal ph and chewing activity. J. Dairy Sci. 85: Leonardi, C., and L. E. Armentano Effect of quantity, quality, and length of alfalfa hay on selective consumption by dairy cows. J. Dairy Sci. 86: Leonardi, C., F. Giannico, and L. E. Armentano Effect of water addition on selective consumption (sorting) of dry diets by dairy cattle. J. Dairy Sci. 88: Littell, R. C., P. R. Henry, and C. B. Ammerman Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci. 76:

111 95 Maulfair, D. D., M. Fustini, and A. J. Heinrichs Effect of varying total mixed ration particle size on rumen digesta and fecal particle size and digestibility in lactating dairy cows. J. Dairy Sci. 94: Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of feed sorting on chewing behavior, production, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 93: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. Rutter, S. M Graze: A program to analyze recordings of the jaw movements of ruminants. Behav. Res. Meth. Ins. C. 32: Rutter, S. M., R. A. Champion, and P. D. Penning An automatic system to record foraging behaviour in free-ranging ruminants. Appl. Anim. Behav. Sci. 54: Shipley, R. A., and R. E. Clark Tracer Methods for In Vivo Kinetics. Academic Press, New York, NY. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, C.-M. J., and G. A. Varga Effect of three concentrate feeding frequencies on rumen protozoa, rumen digesta kinetics, and milk yield in dairy cows. J. Dairy Sci. 72: Yang, W. Z., K. A. Beauchemin, and L. M. Rode Effects of grain processing, forage to concentrate ratio, and forage particle size on rumen ph and digestion by dairy cows. J. Dairy Sci. 84: Zanton, G. I., and A. J. Heinrichs Digestion and nitrogen utilization in dairy heifers limitfed a low or high forage ration at four levels of nitrogen intake1. J. Dairy Sci. 92:

112 96 Table 4-1. Chemical compositions and particle size distributions determined with the ASABE particle separator for alfalfa haylage and long and short corn silage Alfalfa Corn Silage Item Haylage Long Short SEM 1 P-value 1 Particle size, as-fed % retained mm < mm < mm mm mm < 0.01 Pan Composition, % of DM DM < 0.01 CP ADF NDF pendf < pendf < 0.01 Ash NFC Starch NE L, Mcal/kg Associated with corn silages. 2 Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 3 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of Penn State particle separator; Kononoff et al., 2003a). 4 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of Penn State particle separator; Kononoff et al., 2003a).

113 97 Table 4-2. Chemical compositions, particle size distributions, and rates of disappearance determined via in situ incubation for dry cracked and dry fine ground corn Item Cracked Fine Ground SEM P-value Particle size, as-fed % retained 6.70 mm < mm < mm < mm < mm mm < mm < mm < mm < mm < 0.01 Pan Composition, % of DM DM CP ADF NDF Ash NFC NE L, Mcal/kg Rate of disappearance 1, % 0.5 h < h < h < h < h < h < h < h < h < h Nylon bags were incubated in quadruplicate in the rumen of 2 lactating cows (each cow incubated 2 bags of each sample for each time point).

114 98 Table 4-3. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) 1 Treatment P-value Item LC LF SC SF SEM FPS RFC FPS RFC Particle size, as-fed % retained mm < mm < mm < 0.01 < mm < 0.01 < mm < Pan < 0.01 < Composition, % of DM DM, % CP ADF NDF Forage NDF < pendf < pendf < 0.01 < Ash NFC < 0.01 < Starch 29.2 b 30.7 ab 32.8 a 31.3 ab NE L, Mcal/kg a b Means within a row with different superscripts differ (P 0.05). 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. 2 Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 3 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of Penn State particle separator; Kononoff et al., 2003a). 4 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of Penn State particle separator; Kononoff et al., 2003a).

115 99 Table 4-4. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on chewing behavior 1 Treatment P-value Item, min/d LC LF SC SF SEM FPS RFC FPS RFC Ruminating Eating Total chewing LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn.

116 Table 4-5. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on rumen fermentation 1 Treatment P-value Item LC LF SC SF SEM FPS RFC FPS RFC Rumen ph Weighted mean Minimum Maximum AUC 3 < AUC < NH 3, mg/dl Weighted mean Minimum Maximum Lactate, µm/ml Weighted mean Minimum Maximum VFA weighted mean, µm/ml Acetate Propionate Butyrate Valerate Isovalerate Isobutyrate A:P Rumen digesta Volume, L DM Weight, kg LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. 2 Weighted averages determined by calculating the area under the response curve according to the 100 trapezoidal rule (Shipley and Clark, 1972). 3 AUC = Area under curve, ph units min/d (area below ph threshold (5.5 or 5.8) and above ph profiles of cows).

117 101 Table 4-6. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on cumulative selection indices 1 for various particle fractions 2 Treatment P-value Item LC LF SC SF SEM FPS RFC FPS x RFC 26.9 mm 8 h h *0.89 * < h *0.92 * mm 8 h h 0.97 * h 0.97 * * mm 8 h h 0.99 * * h *0.99 * * < mm 8 h h * h mm 8 h h h Pan 8 h * h h *Sorting index is significantly different from 1.00 based on a 95% confidence limit. 1 Values = 1.00 indicate no sorting, values < 1.00 indicate sorting against, and values > 1.00 indicate sorting for. 2 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn; approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan).

118 102 Table 4-7. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on interval selection indices 1 for various particle fractions 2 Treatment P-value Item LC LF SC SF SEM FPS RFC FPS x RFC 26.9 mm 8 h h *0.69 * < h mm 8 h h *0.91 b *0.85 b *0.92 b 1.04 a 0.04 < h *0.89 *0.84 *0.85 * mm 8 h h 0.97 * * h 0.98 *0.95 * mm 8 h h * * < 0.01 < h *1.05 * mm 8 h * h *1.04 * < h *1.10 * < Pan 8 h h *1.14 *1.13 *1.05 * < h *1.15 * a b Means within a row with different superscripts differ (P 0.05). *Sorting index is significantly different from 1.00 based on a 95% confidence limit. 1 Values = 1.00 indicate no sorting, values < 1.00 indicate sorting against, and values > 1.00 indicate sorting for. 2 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn; approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan).

119 103 Table 4-8. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily DM, NDF, starch, and particle fraction intake 1 Treatment P-value Item, kg LC LF SC SF SEM FPS RFC PFS RFC DMI 27.9 b 30.9 a 31.2 a 31.6 a 1.08 < 0.01 < 0.01 < 0.01 NDF Starch 8.2 b 9.6 a 10.2 a 9.8 a 0.36 < < 0.01 Particle fractions mm < mm < mm 5.63 ab 5.42 b 5.77 a 4.95 c < mm 5.67 b 4.42 d 7.20 a 5.10 c 0.19 < 0.01 < 0.01 < mm 6.10 d 7.63 c 9.11 b 9.76 a 0.29 < 0.01 < Pan < 0.01 < Cumulative % of daily intake 8 h < h < Refusal, % a d Means within a row with different superscripts differ (P 0.05). 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. 2 Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan).

120 104 Table 4-9. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on milk yield and components 1 Treatment P-value Item LC LF SC SF SEM FPS RFC FPS RFC Milk yield, kg/d < % FCM, kg/d Feed efficiency a 1.41 ab 1.37 b 1.40 b Fat, % < Fat, kg/d Protein, % Protein, kg/d < Lactose, % 4.75 b 4.83 a 4.78 ab 4.79 ab Lactose, kg/d < MUN, mg/dl < SCC, 1,000 cells/ml 800 a 43 c 242 bc 504 ab < 0.01 a c Means within a row with different superscripts differ (P 0.05). 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn % FCM = (milk kg) (fat kg); (Gaines, 1928). 3 Feed efficiency = 3.5% FCM / DMI.

121 105 Table Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily weighted mean 1 fecal particle size and DM content 2 Treatment P-value Item, % of DM LC LF SC SF SEM FPS RFC FPS x RFC DM < 0.01 < Retained DM mm 5.8 a 4.0 b 6.6 a 3.3 b < mm < mm 17.3 b 17.6 ab 17.0 b 18.7 a mm < mm < X gm 5, mm < S 6 gm, mm < Total DM mm 2.8 a 1.9 b 3.5 a 1.6 b < mm < mm mm < 0.01 < mm < Soluble 52.5 a 52.5 a 49.3 b 51.4 a 0.69 < X gm, mm 0.38 b 0.34 c 0.42 a 0.34 c < S gm, mm < a c Means within a row with different superscripts differ (P 0.05). 1 Weighted means determined by calculating area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). 2 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. 3 Retained DM = Parameters determined from sample retained on sieve stack. 4 Total DM = Parameters determined from total sample including soluble fraction. 5 X gm = geometric mean particle length determined by ASABE (2007). 6 S gm = particle length standard deviation determined by ASABE (2007).

122 106 Table Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on daily weighted mean 1 ruminal digesta particle size distribution and DM content 2 Treatment P-value Item, % of DM LC LF SC SF SEM FPS RFC FPS x RFC DM Retained DM mm < mm mm < 0.01 < mm < mm < mm < Total DM mm < mm mm < 0.01 < mm < mm < mm < Soluble a c Means within a row with different superscripts differ (P 0.05). 1 Weighted means determined by calculating area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). 2 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. 3 Retained DM = Parameters determined from sample retained on sieve stack. 4 Total DM = Parameters determined from total sample including soluble fraction.

123 107 Figure 4-1. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on starch concentration at 0 and 24 h after feeding 1 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn. *time effect P 0.05; overall time effect P < 0.01.

124 108 Figure 4-2. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on NDF concentration at 0 and 24 h after feeding 1 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn; overall time effect P = 0.15.

125 109 A B

126 110 C Figure 4-3. Effect of feeding TMR varying in forage particle size (FPS) and ruminally fermentable carbohydrates (RFC) on TMR particle fractions > 26.9 mm (A), > 1.65 mm (B), and pan (C) at 0, 8, 16, and 24 h after feeding 1 1 LC = long corn silage and dry cracked corn, LF = long corn silage and dry fine ground corn, SC = short corn silage and dry cracked corn, SF = short corn silage and dry fine ground corn.

127 Chapter 5 Effect of Subacute Ruminal Acidosis on Total Mixed Ration Preference in Lactating Dairy Cows Abstract Subacute ruminal acidosis (SARA) is a condition where the ph of the rumen becomes abnormally acidic because of increased and altered production of volatile fatty acids. The objective of this experiment was to determine how SARA affects total mixed ration selection in dairy cows. In this study 8 multiparous, lactating, ruminally cannulated Holstein cows were given a choice between a long forage particle size diet with slow-fermenting starch (LC) and a short forage particle size diet with fast-fermenting starch (SF). Cows were allowed to adapt to this feeding scheme and were then subjected to a rumen challenge to induce a bout of SARA. The rumen challenge successfully decreased rumen ph and altered rumen volatile fatty acid profiles. Daily weighted average rumen ph decreased from 6.02 to 5.77, and average minimum rumen ph decreased from 5.59 to In addition, following the rumen challenge concentrations of acetate, butyrate, and valerate and acetate to propionate ratio increased. In response to the rumen challenge, intake of LC increased from the baseline level of 18.1% of total daily dry matter intake to 38.3% for that d. During the first recovery d after the rumen challenge, LC intake moderated to 28.0% of total daily dry matter intake. On the second recovery d LC intake returned to baseline levels at 18.6%. These results indicate that cows are able to alter their diet preference for higher physically effective fiber and slower starch fermentability during a bout of SARA and that they can effectively fully recover from this type of SARA within 72 h when appropriate diets are available.

128 112 Key Words: acidosis, diet selection, particle size, sorting Introduction Subacute ruminal acidosis (SARA) is a major concern in the modern high producing dairy cow. It is defined as a moderately depressed rumen ph in the range of 5.5 to 5.0 (Nocek, 1997; Krause and Oetzel, 2006). Krause and Oetzel (2006) suggested that there are 3 major causes of SARA in dairy herds: excessive intake of rapidly fermentable carbohydrates, inadequate ruminal adaptation to a highly fermentable diet, and inadequate ruminal buffering caused by inadequate dietary fiber or inadequate physical fiber. The negative effects of SARA are vast and varied; ranging from decreased DMI (Britton and Stock, 1986; Nocek, 1997) and reduced feed efficiency (Huntington, 1993; Nocek, 1997) to decreased milk fat yield (Nocek, 1997) and contributing to lameness (Nocek, 1997; NRC, 2001; Stone, 2004). A study that evaluated 154 cows in 14 Wisconsin dairy herds determined that 20.1% of lactating cows had SARA when tested using rumenocentesis (Oetzel et al., 1999). In a case study of a 500-cow dairy in central New York state, Stone (1999) estimated that SARA could cost up to $475/cow per yr in lost milk production and components only. Clearly, SARA warrants extensive research and management. There are several studies that have examined diet and feed selection changes when sheep or lambs were subjected to acidotic rumen conditions. For example, in a study by Phy and Provenza (1998b) lambs were fed a meal of rolled barley and then offered a choice of flavored (onion or oregano) rabbit pellets that either contained NaHCO 3 and lasalocid or NaCl. The authors determined that after a grain meal lambs preferred rabbit pellets that contained NaHCO 3 and lasalocid over pellets that contained NaCl. Another study by Phy and Provenza (1998a) examined the effect eating a meal of rapidly fermentable feed had on the preference for rapidly

129 113 fermentable feed later in the d. Lambs fed a lower amount (400 g) of rolled barley for a meal exhibited equal preference for rolled barley and alfalfa pellets (52 and 48% of total intake respectively) during the next 4 h. However, when a higher amount (1,200 g) of rolled barley was fed the lambs increased their preference for alfalfa pellets over rolled barley (71 and 29% of total intake respectively) during this same time (Phy and Provenza, 1998a). All of these results show that lambs prefer feeds that attenuate acidosis after a grain meal to maintain ruminal health. In addition, studies have examined the influence of SARA in dairy cows on eating behavior. Keunen et al. (2002) conducted an experiment where 25% of DMI of cows being fed a TMR was replaced by wheat and barley pellets in order to induce SARA. The choice of 2 feeds, long alfalfa hay and alfalfa pellets, was then offered 2 times per d for 30 min each. Cows with SARA increased their consumption of long alfalfa hay over alfalfa pellets when compared to their consumption without SARA; 85 and 60% of test feeds were consumed as long alfalfa hay for SARA and non-sara cows respectively (Keunen et al., 2002). DeVries et al. (2008) used a rumen challenge model to induce SARA in early and mid-lactation Holstein cows. The rumen challenge consisted of restricting feed to 50% of ad libitum DMI for 1 d followed by feeding 4 kg of barley and wheat and then ad libitum access to TMR. Changes in eating behavior were measured by determining the particle size distribution of offered feed and refusals and calculating a selection index for each particle fraction. After the rumen challenge, cows in both groups changed their sorting behavior. DeVries et al. (2008) determined that early lactation cows generally increased their sorting for medium particles and against short and fine particles and exhibited no change in sorting long particles. Mid-lactation cows exhibited variable responses with sorting activity changing with d and period. DeVries et al. (2008) suggested that both early and mid-lactation cows altered their sorting behavior to consume a diet that would help attenuate their bout of SARA.

130 114 Despite there being evidence of dairy cattle altering their eating behavior or diet choice based on their rumen environment, there has been no research published where cattle had access to 2 distinct diets to observe the influence of SARA on diet preference and eating behavior. Therefore the objective of this experiment was to induce a bout of SARA in lactating dairy cows that had ad libitum access to 2 distinct diets that varied in forage particle size and starch fermentability and to determine how SARA affects TMR selection in dairy cows. Materials and Methods Diets, Cows, and Experimental Design Cows used in this research were cared for and maintained according to a protocol approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Eight lactating, multiparous, ruminally cannulated, Holstein cows averaging 219 ± 61 DIM and 44 ± 7 kg/d milk production, weighing 702 ± 56 kg, and with parity of 3.13 ± 0.99 (mean ± SD) were studied. The trial consisted of a 7-d adaptation period followed by an 8-d collection period. For the duration of the study, cows were fed 2 different diets simultaneously: a long particle size diet with slowly fermentable starch (LC) and a short particle size diet with fast starch fermentabilty (SF). Diets were offered to cows in tie-stall feed bunks divided into halves via a plywood panel that eliminated cross contamination of TMR. The side of the feed bunk that the diets were offered was alternated each d to limit the possibility for bias of bunk location or relationship to water bowls. The 2 rations fed contained identical ingredients and proportions, but varied in the particle length of corn silage and the particle size of dry ground corn. The LC diet included long corn silage (LCS) and dry cracked corn (CC) and the SF diet included short corn silage (SCS) and dry fine ground corn (FC). Ingredients and their percentage of ration DM were:

131 115 corn silage (42.6), dry ground corn (22.2), alfalfa haylage (15.4), canola meal (9.4), roasted split soybeans (7.1), mineral/vitamin mix (2.5), salt (0.4), and Optigen (Alltech, Nicholasville, KY; 0.4). Corn silage hybrid was Pioneer 34M78 (Pioneer Hi-Bred International, Inc., Johnston, IA) that was planted on 4/19/2010 and harvested on 8/30/2010. Corn silage was harvested with a John Deere 6750 forage harvester (John Deere, Moline, IL) equipped with a kernel processor set at approximately 6.35 mm. The cutterhead of the harvester used 16 knives (maximum capacity is 48 knives) with the length-of-cut transmission at its highest setting to produce a theoretical length of cut of 47.1 mm. After harvesting, corn silage was ensiled in an Ag-Bag (Ag-Bag, St. Nazianz, WI) and allowed to ferment for 62 d before beginning the study. Corn silage that was removed from the Ag-Bag and mixed into TMR without further processing was considered LCS. A cutand-throw type, single row, forage harvester that was modified to operate on a trailer and be fed manually with a 25 horsepower V-Twin small gas engine was used to reduce the particle size of corn silage to produce SCS. Corn silage was rechopped twice through the custom forage chopper on a daily basis to minimize the chemical variance between LCS and SCS. Dry corn was ground through a Roskamp roller mill (California Pellet Mill Co., Crawfordsville, IN) to produce the CC used in this study. This corn was then ground further with a Case International 1250 grindermixer (Case IH, Racine, WI) using a 3.18 mm screen to produce FC. Diets were mixed separately using an I. H. Rissler model 1050 TMR mixer (E. Rissler Mfg. LLC, New Enterprise, PA). The 4 consecutive d immediately following the adaptation period (d 8 to 11) were designated the baseline for feed preference and rumen conditions. On d 12 feed intakes for each diet were restricted to 75% of baseline intake. Following feed restriction, on d 13 at kg (as-fed) of fine ground wheat was thoroughly mixed into the rumen digesta of each cow via the rumen cannulae to provide a rumen challenge by initiating SARA. Each cow was then allowed ad libitum access to both diets at 0800, the amount of TMR offered allowed for approximately 115%

132 116 of total daily baseline intake to be consumed from either diet offered. Ad libitum TMR feeding continued on d 14 and 15 to monitor recovery from the rumen challenge. Animals were housed in individual stalls, milked twice/d at 0500 and 1700 h, and fed once/d at approximately 0800 h for ad libitum consumption. Cows were fed for a 10% refusal rate except when either treatment diet intake was below 6 kg/d DM, which was set as the minimum amount of feed to be offered to always allow for an opportunity to choose either diet. Feed was pushed up 3 times/d at 1230, 1730, and 2400 h. Rations were balanced to meet or exceed NRC (2001) requirements for cows producing 52.2 kg of milk/d containing 3.75% fat and 3.07% true protein assuming a DMI of 29.5 kg/d and water was available for ad libitum consumption. Rumen Sampling On d 11 of the study, ruminal contents were collected from dorsal, ventral, cranial, caudal, and medial areas of the rumen at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding (Kononoff et al., 2003b) to determine baseline rumen conditions. Rumen sampling also occurred on d 12 (feed restriction) at 11.5, 14.5, 18.0, and 21.5 h after feeding, d 13 (rumen challenge) at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding, and d 14 (recovery) at 3.5, 8.5, 14.5, and 21.5 h after feeding. At each rumen sampling collected digesta was mixed thoroughly, sampled, and filtered through 2 layers of cheesecloth. Rumen liquid ph was immediately determined using a handheld ph meter (HI 98121, HANNA Instruments Inc., Woonsocket, RI). Approximately 15 ml of filtered liquid was placed into bottles containing 3 ml of 25% metaphosphoric acid and 3 ml of 0.6% 2-ethylbutyric acid (internal standard) and stored at approximately 2 C. Within 24 h after collection, samples were centrifuged 3 times at 4000 g for 30 min at 4 C to obtain a clear supernatant and were analyzed for VFA concentration using gas chromatography (Yang and Varga, 1989).

133 117 Feed, Refusal, and Particle Size Analysis Feed bunk contents for each animal were weighed and sampled on d 8 to 15 at 0 and 24 h after feeding to determine particle size distribution and DM content of the remaining feed. Additionally, feed bunk contents were weighed on d 8 to 11 and d 13 at 2, 4, 8, and 16 h after feeding. All samples were sieved in the American Society of Agriculture and Biological Engineers (ASABE) forage particle separator, which can determine 6 particle fractions (> 26.9, > 18.0, > 8.98, > 5.61, > 1.65, and < 1.65 mm; screen diagonal; ASABE, 2007). Whole samples were then placed in a forced air oven at 65 C for 48 h to determine DM content. Samples of forages, ground corn, and TMR were taken on d 11 and 13 and analyzed by Cumberland Valley Analytical Services, Inc. (Hagerstown, MD) for CP (AOAC, 2000), ADF (AOAC, 2000), NDF (Van Soest et al., 1991), ash (AOAC, 2000), NFC (Van Soest et al., 1991), and NE L (NRC, 2001). Starch contents of forages, ground corn, and TMR were determined by grinding (0.5-mm screen; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ) dried samples and then using the starch procedure reported by Zanton and Heinrichs (2009). Particle size distributions of forages and TMR were determined via sieving with the ASABE forage particle separator (ASABE, 2007). To determine particle size distributions of ground corn, samples were placed on a series of stacked sieves (sizes 0.15, 0.425, 0.60, 0.85, 1.18, 1.70, 2.36, 3.35, 4.75, and 6.7 mm; VWR, Arlington Heights, IL) contained in a Retsch AS 200 Control sieve shaker (Retsch, Haan, Germany) and were sieved for 10 min at 2.5 mm amplitude. There was approximately a 41% increase between each sieve screen size, except between the and mm sieves. Particles retained on each sieve were then weighed to determine their proportion of total sample DM. There were 2 procedures used to calculate physically effective NDF (pendf): pendf 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of the Penn State particle separator) and pendf 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3

134 118 sieves of the Penn State particle separator; Kononoff et al., 2003a). Corn grain fermentability was determined via in situ bags incubated in quadruplicate in the rumen of 2 lactating cows (each cow incubated 2 bags of each sample for each time point) for 0.5, 1, 2, 4, 6, 8, 12, 16, 24, and 48 h. After removal from the rumen, bags were rinsed in cold water by hand until water was almost clear. Bags were then dried in a forced-air oven at 65 C for 48 h and then weighed to determine remaining DM. Statistical Analyses Statistical analysis was conducted using PROC MIXED of SAS (Version 9.2, SAS Institute, Cary, NC). Dependent variables were analyzed as a cross over design. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997) and repeated measurements for ruminal ph, ruminal VFA concentrations, and ground corn DM disappearance were analyzed using the first order autoregressive covariance structure (Littell et al., 1998) as well as terms for time and interaction of treatment by time. Because of unequally spaced rumen sampling, the weighted mean daily ph and VFA concentrations were determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). Area under the curve for the SARA thresholds of 5.8 and 5.5 were also calculated using the trapezoidal rule (Shipley and Clark, 1972). For each cow, the 4 baseline d (8, 9, 10, and 11) were averaged before analysis to provide equal number of observations between baseline and rumen challenge d. A selection index based on refusals was calculated for each of the 6 particle size fractions. This index was calculated as the actual intake of each fraction (Y i to pan) expressed as a percentage of the expected intake. Expected intake of Y i equals intake multiplied by the fraction of Y i in the fed TMR (Leonardi and Armentano, 2003). Values > 1.0 indicate cows were sorting for the particle fraction and values < 1.0 indicate cows were sorting against the

135 119 particle fraction. The 95% confidence limits were used to determine if a selection index was significantly different from 1.0. All data are presented as least squares means and treatment effects are considered significant when P < 0.05 and a trend when P < Results and Discussion Chemical Composition and Particle Size Distribution of Diets Particle size distributions and chemical compositions of forages used in this study are shown in Table 5-1. There was a large difference in particle size distribution between LCS and SCS. When separated with the ASABE particle separator, LCS had many more particles retained on 26.9 and 18.0 mm screens, equal particles on the 8.98 mm screen, and many fewer particles on 5.61 and 1.65 mm screens and the pan than SCS. The approximate equivalency of Penn State particle separator fractions to the ASABE screens are: top ( mm), middle (8.98 mm), lower ( mm), and pan (pan). The particle size distribution of alfalfa haylage was similar to SCS. Chemical compositions of the corn silages were similar and not practically different despite some statistically significant differences for DM, ADF, NDF, NFC, and NE L. Sampling error may be responsible for the small differences seen between LCS and SCS since they were taken from the same bag each d as a single batch, with part being re-chopped as the only difference. Rechopping of corn silage could conceivably increase DM content through increased drying rate. The pendf measures were, as expected, very different between corn silages, but there was a much greater difference for pendf 8.0 than for pendf The LCS was 1.81 and 1.15 times greater than SCS for pendf 8.0 and pendf 1.18 respectively. The particle size distributions, chemical compositions, and rates of disappearance for corn grains used in this study are shown in Table 5-2. The particle size distributions of CC and

136 120 FC were different at all 11 particle fractions. The greatest differences occurred at screen sizes 2.36 mm and larger, where CC had 67.4% and FC had 5.6% of particles retained, and at screen sizes 1.18 mm and smaller, where CC had 18.4% and FC had 78.2% of particles retained. The chemical compositions of CC and FC were similar and not practically different despite being statistically different in DM and CP content. The rates of disappearance of CC and FC were different at every time point except 48 h (P-value = 0.15). The greatest differences between CC and FC were in the first 2 h of incubation, where FC had about 2.1 times more DM disappearance than CC. The disappearance of FC continued to be greater than CC at each time point (except 48 h), but the differences between them decreased with increasing incubation time. These data should be interpreted with caution as the impact of eating and rumination on ground corn was not a factor in this analysis and it is reasonable to assume that chewing would have a larger impact on CC because of its greater potential for further particle size reduction. The particle size distributions and chemical compositions of the treatment TMR are shown in Table 5-3. Each particle fraction was different between LC and SF; the 4 largest particle fractions (> 26.9, > 18.0, > 8.98, and > 5.61 mm) were greater for LC, while the 2 smallest particle fractions (> 1.65 mm and pan) were greater for SF. The chemical compositions of the TMR were similar and not practically different. The CP, NDF, forage NDF, and starch content of the TMR were approximately 16.3, 31.8, 21.5, and 30.1% of DM respectively. The pendf measures were very different between LC and SF diets, with the greatest difference occurring with pendf 8.0, where LC was 2.12 times higher than SF (13.8 versus 6.5%). The LC diet was only 1.30 times higher than SF for pendf 1.18 (27.7 versus 21.3%).

137 121 Rumen Characteristics The effect of the rumen challenge model on rumen ph is shown in Figure 5-1. On the baseline d rumen ph gradually decreased after feeding to a low of 5.61 at 11.5 h post- feeding. Rumen ph then gradually increased to pre-prandial levels by 24 h after feeding. The following d (feed restriction d) rumen ph was measured starting at 11.5 h after feeding and rumen ph was not different from baseline levels at 11.5 and 14.5 h after feeding. Rumen ph then increased faster and remained higher than baseline levels for the remainder of the d. The following d (rumen challenge d) ground wheat was mixed into the rumen via the cannulae of all cows 15 min before feeding. Rumen ph, which began at a higher level than baseline, then dropped sharply after feeding until 3.5 h after feeding and remained constant for 18 h after feeding. Rumen challenge d rumen ph had a 3.8-fold larger drop from feeding to 1.5 h after feeding and a 6.1-fold larger drop from feeding to 3.5 h after feeding compared to baseline d rumen ph. Also rumen challenge d rumen ph was lower than baseline d rumen ph at 3.5, 5.5, 8.5, and 18.0 h after feeding. By 21.5 h after feeding on rumen challenge d, rumen ph had returned to baseline levels and stayed at baseline levels for the remainder of the rumen challenge d and the following recovery d, except at 21.5 h after feeding on the recovery d; the cause of this difference is not apparent. Ruminal ph daily weighted average was lower and the area under ruminal ph 5.8 and 5.5 was greater during the rumen challenge d compared to baseline (Table 5-4). This indicates that the rumen challenge was successful in inducing SARA and the drop in average rumen ph (0.25 unit decrease) was comparable to other studies attempting to induce SARA in dairy cattle such as Keunen et al. (2002) and Dohme et al. (2008); 0.14 and 0.35 unit decreases respectively. The area under ruminal ph of 5.8 was increased by 5 fold (50.9 to ph units min/d) on the rumen challenge d and the area under ruminal ph of 5.5 was essentially 0.0 during baseline but increased to 37.3 ph units min/d on the rumen challenge d. Dohme et al. (2008) showed similar

138 122 areas under the curve for their early lactation cows subjected to rumen challenges where areas under 5.8 ruminal ph were 136, 231, and 475 ph units min/d and under 5.5 ruminal ph were 42, 91, and 291 ph units min/d (during the 1 st, 2 nd and 3 rd rumen challenge respectively). In addition, there was more variation in rumen ph on the rumen challenge d as it had a lower minimum (5.28 versus 5.59) and a higher maximum (6.95 versus 6.69) over 24 h. Rumen VFA concentrations were also determined to be different between these 2 d. The daily weighted average concentration for acetate, butyrate, valerate, and isobutyrate, as well as acetate to propionate ratio increased on rumen challenge d. Propionate and isovalerate concentrations were not affected by the rumen challenge. TMR Preference, Dry Matter Intake, and Refusals TMR preference was measured as the amount of LC diet DM consumed divided by total daily DMI and expressed as a percentage. Average LC consumption for all cows over the 4 baseline d was 18.1% of total daily DMI (Figure 5-2). This ratio remained the same for the feed restriction d as both diets were restricted to 75% of baseline intake and there were virtually no ration refusals for either diet (Table 5-5). These results are in agreement with the results of Castle et al. (1979) where 3 grass silages of different particle lengths were fed simultaneously to 3 pregnant Ayrshire heifers. The heifers consumed 15.9, 31.9, and 52.2% of total DMI as long, medium, and short silages respectively. After the rumen challenge, LC intake increased dramatically to 38.3%, followed by 28.0% on the first recovery d. On the second recovery d after rumen challenge, LC intake returned to baseline levels at 18.6% of total daily DMI. These results clearly show that the cows very consistently (small SE values) changed their TMR preference in response to a rumen challenge and also they appear to fully recovered from this rumen challenge within 72 h.

139 123 The DMI and refusals for each diet for the baseline, feed restriction, rumen challenge, and recovery d are shown in Table 5-5. The average daily DMI during the baseline period was 30.7 kg/d (5.3 and 25.4 kg/d for LC and SF respectively). The following d feed was restricted to 75.9% of baseline intake at 23.3 kg/d (4.0 and 19.3 kg/d for LC and SF respectively). Rumen challenge d DMI increased from the baseline for LC by 136% and decreased for SF by 20% for a total daily intake of 32.7 kg/d (excluding ground wheat). Intake of LC recovered to baseline levels by recovery d 1 and SF DMI recovered to baseline levels by recovery d 2. The amount of TMR delivered to the cows was adjusted daily to maintain a refusal rate of 10%, with the exception of the restricted intake d prior to the rumen challenge. However, since the minimum amount of feed offered per diet per d was set at 6 kg of DM and most cows consumed less than this amount of the LC diet, LC refusals were much higher than SF refusals during the baseline period (31.8 versus 9.6%). On the rumen challenge d there was a drastic increase in refusals for both diets because cows were offered approximately 115% of total daily baseline intake for each diet (230% of total daily baseline intake combined) so they had the ability to consume their entire daily intake from only one diet if they preferred. Therefore, LC refusals were 63.0% and SF refusals were 45.5% on the rumen challenge d. Refusal rates remained elevated during the 2 recovery d because larger amounts of feed continued to be offered to allow cows to return to their baseline LC:SF intake ratios without the influence of low diet refusals. Whether TMR was delivered on the left or right side of the feed bunk or whether TMR was delivered to the side of the feed bunk that was adjacent to a water bowl did not affect the percent of LC consumed as a percentage of total daily DMI (P = 0.68 and 0.63, respectively) during the baseline period. The cumulative percentages of daily intakes for each diet for the baseline and rumen challenge d are shown in Figure 5-3. The cows consumed approximately 21.9% of their total daily intake by 2 h after feeding and there were no differences in the cumulative percentages of daily intakes among the diets or d. At 4 h after feeding the cows consumed an average of 32.1%

140 124 of their daily diet intakes; however, there were some small differences among the diets and d. The SF intake was lower on the baseline d compared to LC and SF intake on the rumen challenge d. By 8 and 16 h after feeding the cows had consumed approximately 51.7 and 81.5% respectively of daily diet intakes and there were no differences among diets and d. These results show that the cows consumed the diets simultaneously and in the same ratio throughout the d, independent of which d it was. In other words the cows did not consume a larger proportion of 1 diet at certain times of the d and a larger proportion of the second diet at another time of d. These data also show how heavily a cows daily DMI is skewed toward immediately after feeding when only 1 meal is fed per d in an individual stall housing system, even though ample feed was available before feeding based on consistently having high levels of refusal. Ration Sorting Ration sorting was measured via selection indices calculated by comparing TMR particle size distributions at time of feeding to 24 h after feeding. The selection index uses the actual consumption of a particle fraction divided by the estimated consumption of the particle fraction if no sorting occurred to produce the index value. Index values > 1.0 indicate sorting for a particle fraction and values < 1.0 indicate sorting against. It was found that there were no differences in sorting indices among the baseline, rumen challenge, and recovery d (P all > 0.10); therefore all d were averaged (Table 5-6). Based on these selection indices, sorting occurred in all 6 particle fractions when cows were fed the LC diet. They sorted against particles retained on the 3 largest screens (26.9, 18.0, and 8.98 mm) and for particles in the 3 smallest particle fractions (> 5.61, > 1.65 mm, and pan). All of the LC sorting indices were different from 1.0 based on their 95% confidence limits, even though they did not have very large numerical differences. There was much less ration sorting on the SF diet, as 4 of the 6 particle fractions had no significant sorting

141 125 occurring. The cows fed the SF diet did sort against particles retained on the 18.0 and 8.98 mm screens. In addition, the sorting indices for each particle fraction were different between the diets except for the 18.0 mm screen. It is unlikely that the minimal amount of sorting described in this study influenced cow performance and rumen fermentation, because in a previous study by Maulfair et al. (2010) much greater ration sorting activity was found to have no effects on milk production or rumen fermentation patterns. Conclusions Lactating cows were given the choice between a diet with long forage particle size and slowly fermentable starch and a diet with short forage particle size and rapidly fermentable starch. Cows were allowed to adapt to this 2-diet feeding scheme until the intake ratio of LC:SF remained constant. Cows were then given a rumen challenge to induce a bout of SARA. Results of this study show that dairy cattle can significantly alter their TMR preference, when faced with SARA, to a diet with increased pendf and slower starch fermentability that may help alleviate their acidotic condition. In addition, this study showed that dairy cattle with this severity of a single bout of SARA can fully recover within 72 h after onset. Since cattle were only fed once per d and only subjected to one bout of SARA in this study, further research is warranted to evaluate effects of multiple SARA bouts and different feeding times and feeding systems on diet selection. Acknowledgements Sincere appreciation is extended to Growmark FS, LLC (Sangerfield, NY) for generously allowing the use of their modified forage harvester for the duration of this trial. This research was

142 supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture. 126 References American Society of Agricultural and Biological Engineers Method of determining and expressing particle size of chopped forage materials by screening. ANSI/ASAE. S424.1: Association of Official Analytical Chemists Official Methods of Analysis. 17th ed. AOAC, Arlington, VA. Britton, R. A., and R. A. Stock Acidosis, rate of starch digestion and intake. Pages in Proceedings of the Feed Intake Symposium, Oklahoma Agricultural Experiment Station, Norman, OK. Castle, M. E., W. C. Retter, and J. N. Watson Silage and milk production: comparisons between grass silage of three different chop lengths. Grass and Forage Science. 34: DeVries, T. J., F. Dohme, and K. A. Beauchemin Repeated ruminal acidosis challenges in lactating dairy cows at high and low risk for developing acidosis: Feed sorting. J. Dairy Sci. 91: Dohme, F., T. J. DeVries, and K. A. Beauchemin Repeated ruminal acidosis challenges in lactating dairy cows at high and low risk for developing acidosis: Ruminal ph. J. Dairy Sci. 91: Huntington, G. B Nutritional problems related to the gastro-intestinal tract. Pages in The Ruminant Animal: Digestive Physiology and Nutrition. D. C. Church, ed. Waveland Press, Inc., Long Grove, IL. Kenward, M. G., and J. H. Roger Small sample inference for fixed effects from restricted maximum likelihood. Biometrics. 53: Keunen, J. E., J. C. Plaizier, L. Kyriazakis, T. F. Duffield, T. M. Widowski, M. I. Lindinger, and B. W. McBride Effects of a subacute ruminal acidosis model on the diet selection of dairy cows. J. Dairy Sci. 85: Kononoff, P. J., A. J. Heinrichs, and D. R. Buckmaster. 2003a. Modification of the Penn State forage and total mixed ration particle separator and the effects of moisture content on its measurements. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman. 2003b. The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86:

143 127 Krause, K. M., and G. R. Oetzel Understanding and preventing subacute ruminal acidosis in dairy herds: A review. Anim. Feed Sci. Technol. 126: Leonardi, C., and L. E. Armentano Effect of quantity, quality, and length of alfalfa hay on selective consumption by dairy cows. J. Dairy Sci. 86: Littell, R. C., P. R. Henry, and C. B. Ammerman Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci. 76: Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of feed sorting on chewing behavior, production, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 93: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. Nocek, J. E Bovine acidosis: Implications on laminitis. J. Dairy Sci. 80: Oetzel, G. R., K. V. Nordlund, and E. F. Garrett Effect of ruminal ph and stage of lactation on ruminal lactate concentration in dairy cows. J. Dairy Sci. 82 (Suppl. 1):38. (Abstr.) Phy, T. S., and F. D. Provenza. 1998a. Eating barley too frequently or in excess decreases lambs' preference for barley but sodium bicarbonate and lasalocid attenuate the response. J. Anim. Sci. 76: Phy, T. S., and F. D. Provenza. 1998b. Sheep fed grain prefer foods and solutions that attenuate acidosis. J. Anim. Sci. 76: Shipley, R. A., and R. E. Clark Tracer Methods for In Vivo Kinetics. Academic Press, New York, NY. Stone, W. C The effect of subclinical rumen acidosis on milk components. Pages in Proceedings of the Cornell Nutrition Conference for Feed Manufacturers, Cornell University, Ithaca, NY. Stone, W. C Nutritional approaches to minimize subacute ruminal acidosis and laminitis in dairy cattle. J. Dairy Sci. 87:E13 E26. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, C.-M. J., and G. A. Varga Effect of three concentrate feeding frequencies on rumen protozoa, rumen digesta kinetics, and milk yield in dairy cows. J. Dairy Sci. 72: Zanton, G. I., and A. J. Heinrichs Digestion and nitrogen utilization in dairy heifers limitfed a low or high forage ration at four levels of nitrogen intake. J. Dairy Sci. 92:

144 Table 5-1. Chemical compositions and particle size distributions determined with the ASABE particle separator for alfalfa haylage and long and short corn silage 128 Alfalfa Corn Silage Item Haylage Long Short SEM 1 P-value 1 Particle size, as-fed % retained mm < mm < mm mm mm < 0.01 Pan Composition, % of DM DM < 0.01 CP ADF NDF pendf < pendf < 0.01 Ash NFC Starch NE L, Mcal/kg Associated with corn silages. 2 Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 3 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of Penn State particle separator; Kononoff et al., 2003a). 4 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of Penn State particle separator; Kononoff et al., 2003a).

145 Table 5-2. Chemical compositions, particle size distributions, and rates of disappearance determined via in situ incubation for dry cracked corn, dry fine ground corn, and ground wheat 129 Ground Ground Corn Item Wheat Cracked Fine SEM 1 P-value 1 Particle size, as-fed % retained 6.70 mm < mm < mm < mm < mm mm < mm < mm < mm < mm < 0.01 Pan Composition, % of DM DM CP ADF NDF Ash NFC NE L, Mcal/kg Rate of disappearance 2, % 0.5 h < h < h < h < h < h < h < h < h < h Associated with ground corn. 2 Nylon bags were incubated in quadruplicate in the rumen of 2 lactating cows (each cow incubated 2 bags of each sample for each time point).

146 130 Table 5-3. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) Item LC SF SEM P-value Particle size, as-fed % retained mm < mm < mm < mm mm < 0.01 Pan Composition, % of DM DM, % CP ADF NDF Forage NDF pendf < pendf < 0.01 Ash NFC < 0.01 Starch NE L, Mcal/kg Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of Penn State particle separator; Kononoff et al., 2003a). 3 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of Penn State particle separator; Kononoff et al., 2003a).

147 131 Table 5-4. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on rumen ph and VFA for baseline and rumen challenge d Item Baseline Challenge SEM P-value Rumen ph Weighted average < 0.01 Minimum < 0.01 Maximum < 0.01 AUC 2 < 5.8, ph units min/d < 0.01 AUC < 5.5, ph units min/d < 0.01 VFA weighted average, µm/ml Acetate < 0.01 Propionate Butyrate < 0.01 Valerate Isovalerate Isobutyrate Acetate: Propionate Weighted averages determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). 2 AUC = Area under curve (area below ph threshold (5.5 or 5.8) and above ph profiles of cows.

148 132 Table 5-5. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) on DMI and refusals for baseline, feed restriction, rumen challenge, and recovery d DMI, kg Refusal, % Day LC SF LC SF Baseline 5.3 bc 25.4 a 31.8 c 9.6 d Feed restriction 4.0 c 19.3 c 4.8 d 0.5 e Rumen challenge 12.5 a 20.2 c 63.0 a 45.5 a Recovery b 21.6 bc 42.7 bc 27.1 b Recovery c 23.5 ab 49.5 b 18.4 c SEM P-value < 0.01 < 0.01 < 0.01 < 0.01 a e Means within a column with different superscripts differ (P 0.05).

149 133 Table 5-6. Effect of offering 2 free choice TMR containing long forage and slowly fermentable starch (LC) or short forage and rapidly fermentable starch (SF) on mean selection indices 1 of baseline, rumen challenge, and recovery d (4 d) Screen, mm 2 LC SF SEM P-value < < < 0.01 Pan < Values = 1.00 indicate no sorting, values < 1.00 indicate sorting against, and values > 1.00 indicate sorting for. 2 Approximate equivalency to Penn State particle separator: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 3 Sorting index is not significantly different from 1.00 based on a 95% confidence limit.

150 134 Figure 5-1. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on rumen ph over time for baseline, feed restriction, rumen challenge, and recovery d.

151 135 Figure 5-2. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on preference for TMR with long forage (expressed as a percentage of total daily intake). a c Means with different superscripts differ (P 0.05).

152 136 Figure 5-3. Effect of rumen challenge while offering 2 free choice TMR containing long forage and slowly fermentable starch or short forage and rapidly fermentable starch on cumulative percent of diet daily intake at various times after feeding for baseline and rumen challenge d. a b Means within a time point with different superscripts differ (P 0.05).

153 Chapter 6 Conclusions The first study of this dissertation concluded that replacing alfalfa silage with dry chopped alfalfa hay at levels of 5, 10, 20, and 40% of forage DM had only minimal influence on sorting behavior in lactating dairy cows and only during the first 4 h after feeding. Significant sorting occurred early in the d but by 24 h after feeding cattle had consumed rations that were similar in composition, when measured via sorting indices or actual particle fraction consumption, to their offered TMR indicating that cows were changing their sorting behavior throughout the d. Dry chopped alfalfa hay was included at levels up to 23.5% of ration DM without negative effects on milk production and rumen fermentation. These results suggest that refusal particle size distribution is not a good measure to determine if sorting is a problem in dairy cows. This study also determined that the Penn State and Ro-Tap particle separators produce different results when separating the same samples; indicating that data obtained from these 2 methods of particle separation should not be used interchangeably. The second study investigated the interaction of forage particle size (FPS) and ruminally fermentable carbohydrates (RFC). Four diets were fed that varied in corn silage particle size and corn grain grind size. It was determined that RFC increased milk yield and milk protein content while decreasing milk fat content. Ruminal ph, NH 3, lactate, VFA, and volume were not affected by either FPS or RFC. Changes in starch, NDF, and particle size composition of the refusals throughout the d and selection indices indicated that ration sorting was occurring and diets containing long FPS and high RFC were sorted to a greater degree than diets containing short FPS but no interaction between FPS and RFC was present. There was, however, an interaction between FPS and RFC for DMI. It was shown that DMI decreased with increasing FPS when the

154 138 diet included low RFC and did not change when the diet included high RFC, and DMI increased with RFC for the long diets and did not change with RFC on the short diets. The final study of this dissertation fed lactating cows 2 diets simultaneously and allowed ad libitum consumption of both rations. After cattle were adapted to this feeding system the effects of inducing subacute ruminal acidosis (SARA) with a rumen challenge on diet preference were studied. After adaptation, cattle consumed 18.1% of their total daily intake as the long forage particle size and slowly fermentable starch diet versus a short forage particle size and rapidly fermentable starch diet. When faced with a bout of SARA, cows drastically increased their consumption of the long forage particle size and slowly fermentable starch diet to 38.3% of total daily intake, possibly to help attenuate the SARA. These cattle were fully recovered within 72 h after the initial rumen challenge. This dissertation concludes the following: that ration sorting in lactating dairy cows, despite the general consensus by the majority of dairy cattle nutritionists and researchers, is not of major concern because negative effects seldom occur; that the critical particle size for rumen escape is larger than the previously held 1.18 mm and it is probably close to 6.7 mm; that RFC have a greater influence than FPS on DMI, ruminal fermentation and milk yield and components; and that dairy cattle can alter their diet preference during a bout of subacute ruminal acidosis to consume more physically effective fiber and less rapidly fermentable starch, possibly to attenuate the acidosis. A major finding of this dissertation is that the critical particle size for rumen escape is much greater than 1.18 mm and is likely close to 6.7 mm. The implication of a critical particle size that is much greater than previously thought is that the method for calculating physically effective NDF (pendf) may have to be revised. Particles that are 1.18 mm in length may not be as effective at stimulating chewing activity in lactating dairy cows as previously thought and this may a reason for the many inconsistencies in the literature about the effects of pendf. The use of

155 139 the 8.0-mm sieve of the Penn State particle separator to calculate pendf therefore may be a more accurate method than using the 1.18-mm sieve of the same separator. Further research should be conducted analyzing alterations in the proportions of forages and diets greater than 6.7 mm on chewing activity, ruminal fermentation, and milk composition. Another major finding of this dissertation is that lactating dairy cattle can alter their diet selection when faced with a bout of SARA. These results seem to lead to a lot of exciting research. The mechanism or mechanisms that allow cows to quickly recognize variation in their ruminal environment and the need for changes in feed choice and then correctly identify the feedstuff that would best attenuate the ruminal condition is an area that holds great potential for increasing our knowledge of ruminants. The more practical aspects of this behavior also hold much potential for further research. How successfully cows can correct imbalances in their ruminal environment through diet selection needs to be determined. If cows can use diet selection or feed sorting to bring about improvements in ruminal health then practical applications should be researched to determine if this behavior can be used commercially to the advantage of dairy farmers. Perhaps feeding free choice long hay could decrease the incidence of SARA on farms. Feeding 2 TMR that vary in pendf and monitoring intake of each may provide a sign of increased susceptibility to SARA if intake of the high pendf ration increases. Several results of this dissertation point to the possibility of cows altering their feeding behavior to seemingly improve their ruminal environment or its consistency. The feeding behavior of dairy cattle and its many interactions with feedstuffs, ration compositions, and feeding management have seen very little research. As the physiological systems of the dairy cow are continued to be pushed to their limits with ever increasing energy intake and milk production; how and why cows alter their feed sorting, diet selection, eating, and ruminating behavior may hold the key to increasing animal performance and health in the coming decades.

156 Appendix A Technical Note: Evaluation of Procedures for Analyzing Ration Sorting and Rumen Digesta Particle Size in Dairy Cows Journal of Dairy Science Vol. 93 No. 8, , 2010 D. D. Maulfair and A. J. Heinrichs Abstract Collecting total mixed ration (TMR) samples throughout the d to measure sorting activity of dairy cows may cause changes to sorting behavior of cows or may make it more difficult to elucidate effects of sorting on TMR particle size distributions. Also, forage particle size research commonly includes analysis of the solid portion of rumen digesta for particle size distribution after digesta has been squeezed through several layers of cheesecloth. Therefore, the first objective of this experiment was to determine if collecting TMR samples throughout the d affected sorting behavior of cows and resulted in a different particle size distribution than when TMR was not artificially altered during the d. The second objective of this experiment was to determine if squeezing rumen digesta samples through cheesecloth changed particle size distribution when analyzed by a wet sieving technique. It was determined that small, significant differences existed in particle size distribution between the 2 sampling methods of TMR for sorting behavior. These differences were more likely to occur at time points later in the d. This resulted in small changes in sorting indices calculated from these data; sampling and mixing TMR throughout the d reduced the degree of sorting. Squeezing rumen digesta through 4 layers

157 141 of cheesecloth had no effect on particle size distribution of particles > 0.15 mm but reduced the amount of rumen fluid-associated dry matter contained in the sample. Key words: dairy cow, feeding behavior, particle size, sorting When collecting TMR samples to analyze ration sorting it is necessary to thoroughly mix the remaining TMR in order to collect representative samples. Several studies (Hosseinkhani et al., 2008; Kononoff et al., 2003; Leonardi and Armentano, 2003) have sampled the remaining TMR several times throughout the d. This method may lead to incorrect conclusions about sorting because any sorting that had occurred up to sampling time would be nullified during sample collection. It has not been determined if mixing TMR to take samples affects further ration sorting behavior of dairy cows. Therefore, the first objective of this experiment was to determine if collecting TMR samples throughout the d affected sorting behavior of cows and resulted in a different particle size distribution than when TMR was not artificially altered during the d. When taking rumen samples from fistulated dairy cows during feeding studies, it is common procedure to squeeze rumen digesta through several layers of cheesecloth to obtain the fluid fraction for analysis (Kononoff et al., 2003). The solid fraction retained on the cheesecloth sometimes is used for particle size distribution analysis via wet sieving. However, it is not known if squeezing through cheesecloth affects particle size distribution of the solid fraction. Therefore, the second objective of this experiment was to determine if squeezing rumen digesta samples through cheesecloth changed particle size distribution when analyzed by a wet sieving technique. Data for this paper were collected during the final period of a feeding trial designed to study the effects of varying forage particle size on ration sorting in lactating dairy cows (Maulfair et al., 2010). Cows were cared for and maintained according to a procedure approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Eight lactating, multiparous, Holstein cows averaging 90 ± 32 DIM, weighing 642 ± 82 kg, and with parity of 2.25 ± 0.46 (mean ± SD) were randomly assigned to replicated 4 4 Latin squares; 1 square of

158 142 cows was rumen fistulated. The periods were 21 d in length, with a 13-d adaptation period followed by an 8-d collection period. During each of the 4 periods, cows were fed 1 of 4 rations that contained identical feed ingredients and proportions. Ration ingredients and their percentage of ration DM were: corn silage (29.4), alfalfa haylage (17.6), grass hay (11.8), ground corn (22.9), roasted soybeans (6.7), canola meal (5.7), heat-treated soybean meal (3.2), mineral/vitamin mix (2.4), and salt (0.3). Rations contained 15.9% CP, 34.0% NDF, and 1.65 Mcal/kg NE L and varied only in chop length of the dry grass hay included in the ration. Particle sizes (geometric mean ± SD, mm) of the rations were: short (4.46 ± 3.02), medium (5.10 ± 3.56), long (5.32 ± 3.92), and extra long (5.84 ± 4.39) as determined by ASABE (2007). All diets were mixed separately using an I. H. Rissler model 1050 TMR mixer (E. Rissler Mfg. LLC, New Enterprise, PA). Animals were housed in individual stalls in a mechanically ventilated barn, milked twice per d at 0700 and 1900 h, and fed once per d at approximately 0730 h for ad libitum consumption and a 10% refusal rate to allow for maximum opportunity to sort the ration. Feed was pushed up but not mixed at 1230, 1730, and 2400 h. All rations were balanced to meet or exceed NRC (2001) requirements, and water was available ad libitum. To examine whether mixing and sampling remaining feed affected sorting behavior of dairy cows, TMR samples were taken on d 20 and 21 at 0, 2, 4, 8, 12, 16, and 24 h after feeding (Mixed). In addition, samples were taken at 0 and 8 h (d 19 and 22); 0 and 16 h (d 23 and 24); and 0 and 24 h (d 25 and 26) after feeding (Unmixed). During the study, average DMI was ± 0.88 kg/d and refusals averaged ± 0.70% of DMI; DMI and refusals were not different between treatments (P > 0.24 and 0.22, respectively). At each sampling point TMR was removed from the feed bunk, weighed, thoroughly mixed, sampled, and then returned to the cow, which is standard procedure in feeding studies. All samples were sieved in the American Society of Agriculture Engineers forage particle separator, which can determine 6 particle fractions (> 26.9, > 18.0, > 8.98, > 5.61, > 1.65, and < 1.65 mm; screen diagonal; (ASABE, 2007). Geometric mean

159 143 particle length (X gm ) and standard deviation of particle length (S gm ) were calculated according to the ASABE (2007) procedure. Since > 1% of material was retained on the top screen, 3 samples of each diet were randomly selected, and all particles retained on the top screen were measured manually (with a ruler) before drying. Measured mean particle sizes for the top screen were: ± 3.6, ± 9.1, 84.5 ± 2.6, 74.8 ± 6.6 (mean ± SD, mm) for the extra long, long, medium, and short diets, respectively. Whole samples were then placed in a forced air oven at 55 C for 48 h to determine DM content. Sorting indices based on refusals were calculated for particle size fractions at 8, 16, and 24 h after feeding. Actual intake of each particle fraction was divided by expected intake of each particle fraction (Leonardi and Armentano, 2003). Values > 1 indicate selective consumption and values < 1 indicate selective refusal of the DM retained on an individual sieve. Additionally, X gm sorting indices were calculated for the same time points by dividing the X gm of TMR consumed up to each time point by X gm at time 0. Values > 1 indicate cows were consuming rations with longer particle size and values < 1 indicate cows were consuming rations with shorter particle size than the diets fed. Statistical analysis was conducted using the MIXED procedure of SAS (2006). The model included sampling method, time, and diet as fixed effects, cow as a random effect, and the interaction of sampling method and time. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997). All data are presented as least squares means and sampling method effects are considered significant when P < 0.05 and a trend when P < It was found that TMR did differ slightly when particle distribution was expressed by individual screens (Table A-1). There were some small, significant differences in several of the particle fractions, and sampling method generated more significant and larger differences as the time after feeding increased. In general, where there were significant differences, the Unmixed sampling protocol had a longer particle size than Mixed. When particle distribution of the unconsumed TMR was expressed as geometric mean particle length (Table A-2) there were no

160 144 differences for 0 and 8 h. However, there was a trend at 16 h for Unmixed to be longer than Mixed, and at 24 h samples of remaining TMR collected with the Unmixed protocol had significantly longer particles than the Mixed sampling scheme. In Table A-2 the X gm and S gm values are from the unconsumed TMR or what is left in the feed bunk at that time point. They are the same data that are found in Table A-1 converted from individual screens to an average particle size. There was a significant linear contrast for the sampling procedure by time interaction, indicating that the X gm of the uneaten diet increased to a greater extent for Unmixed than for Mixed with increasing time. When sorting index was calculated using all 6 particle fractions, virtually no significant differences in sorting were found between sampling methods at any of the time points (data not shown). Using the sorting index calculated with X gm (Table A-2), cows were eating shorter rations than they were fed when analyzed using both sampling procedures at all time points. There were no differences at 8 and 16 h, but at 24 h the Unmixed sampling procedure resulted in a lower selection index than the Mixed sampling procedure. Again, the linear contrast for the sampling procedure by time interaction was significant because sorting index increased for the Mixed and decreased for the Unmixed sampling method over time. Diet effects on sorting behavior were found to be significant and are discussed in Maulfair et al. (2010). Based on these results, mixing TMR several times throughout the d to obtain a representative sample caused the particle size of the unconsumed TMR to be smaller and biased conclusions about sorting behavior toward less sorting than what actually occurred. To determine if squeezing rumen digesta through cheesecloth affected particle size distribution obtained via wet sieving, rumen samples were taken on d 15 at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding. Samples were taken from 5 rumen locations (dorsal, ventral, cranial, caudal, and medial areas), mixed thoroughly, and then separated into 2 equal parts. One part was squeezed though 4 layers of cheesecloth, and the solid fraction retained on the cheesecloth was stored in a -20 C freezer. The second part was stored the same way, but

161 145 without the initial squeezing. To determine particle size distribution the 2 samples were then wet sieved using a procedure modified from Beauchemin (1997). Sub-samples (approximately 30 g) were placed on a series of stacked sieves (sizes 0.15, 0.6, 1.18, 3.35, 6.7, 9.5 mm; VWR, Arlington Heights, IL) contained in a Retsch AS 200 Control sieve shaker (Retsch, Haan, Germany) and sieved in duplicate. The samples were sieved for 10 min at 2.5 mm amplitude with cold water flow rate at approximately 1.5 to 2.0 L/min to ensure particles were separated thoroughly. Contents retained on the sieves were rinsed with cold water into a funnel with rumen in situ bags (5 10 cm, 53 μm pore size; ANKOM, Macedon, NY) attached to the stem to collect the sample. Bags were then dried in a forced air oven at 55 C for 24 h and weighed to determine DM retained on each sieve. A portion of each sample was also dried at 55 C for 24 h in a forced air oven without sieving to determine the DM content of the original sample. The rumen fluidassociated fraction of the sample was calculated as the DM lost during the sieving and drying process. Statistical analysis was conducted using the MIXED procedure of SAS (2006). The model included sampling method, time, and diet as fixed effects and cow as a random effect. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997). All data are presented as least squares means, and treatment effects are considered significant when P < 0.05 and a trend when P < There were no significant differences found between the 2 sampling techniques for any of the fractions retained on screens (Table A-3). There was significantly more (46.22 vs %) rumen fluid-associated DM per unit of solid-associated DM for samples that were not squeezed through cheesecloth. If the proportion of rumen fluid-associated DM is of importance to the objective of the experiment, then rumen digesta should not be squeezed before wet sieving. However, if only particles retained on screens are of importance, using rumen digesta after squeezing will have no effect on results.

162 146 In conclusion, mixing TMR to take representative samples several times throughout the d had a small effect of decreasing the particle size of uneaten feed, which may lead to the conclusion that cows are sorting their ration to a lesser extent (when cows are sorting against longer particles and for shorter particles) than if sorting samples were taken only at the end of the sampling interval. Also, squeezing rumen digesta through 4 layers of cheesecloth had no effect on particle fractions > 0.15 mm, but it reduced the proportion of rumen fluid-associated DM per unit of solid-associated DM. Acknowledgements Sincere appreciation is extended to Geoff Zanton (Penn State, University Park, PA) for statistical advice and support. This research was supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture. References American Society of Agricultural and Biological Engineers Method of Determining and Expressing Particle Size of Chopped Forage Materials by Screening. ANSI/ASAE S242.1: Beauchemin, K. A., L. M. Rode, and M. V. Eliason Chewing Activities and Milk Production of Dairy Cows Fed Alfalfa as Hay, Silage, or Dried Cubes of Hay or Silage. J. Dairy Sci. 80(2): Hosseinkhani, A., T. J. DeVries, K. L. Proudfoot, R. Valizadeh, D. M. Veira, and M. A. G. von Keyserlingk The Effects of Feed Bunk Competition on the Feed Sorting Behavior of Close-Up Dry Cows. J. Dairy Sci. 91(3): Kenward, M. G., and J. H. Roger Small Sample Inference for Fixed Effects from Restricted Maximum Likelihood. Biometrics 53(3): Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman The Effect of Corn Silage Particle Size on Eating Behavior, Chewing Activities, and Rumen Fermentation in Lactating Dairy Cows. J. Dairy Sci. 86(10):

163 Leonardi, C., and L. E. Armentano Effect of Quantity, Quality, and Length of Alfalfa Hay on Selective Consumption by Dairy Cows. J. Dairy Sci. 86(2): Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of Feed Sorting on Chewing Behavior, Production, and Rumen Fermentation in Lactating Dairy Cows. J. Dairy Sci. 93: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. SAS Institute SAS User's Guide: Statistics. Version in SAS Inst. Inc., Cary, NC. 147

164 Table A-1. Percentage of uneaten TMR particles (DM basis) retained on sieves at 8-h intervals after feeding when sampled by 2 different procedures 148 Item 1 Mixed 2 Unmixed 3 SE P-value Hour mm mm mm mm mm Pan Hour mm mm mm mm mm Pan Hour mm mm < mm < mm mm Pan <0.01 Hour mm mm mm < mm < mm Pan Pore size of screens. 2 TMR was mixed, sampled, and returned to cow at 2, 4, 8, 12, 16, and 24 h after feeding. 3 TMR was not mixed until sample collections at the respective time point.

165 Table A-2. Geometric mean particle length of uneaten TMR and sorting index of the consumed diet 1 obtained with 2 different sampling procedures 149 Item Mixed 2 Unmixed 3 SE P-value Hour 0 X 4 gm, mm S 5 gm, mm Hour 8 X gm, mm S gm, mm Index Hour 16 X gm, mm S gm, mm <0.01 Index Hour 24 X gm, mm S gm, mm Index < X gm treatment time interaction linear contrast P = 0.01, quadratic contrast P = 0.58; index treatment time interaction linear contrast P = 0.03, quadratic contrast P = TMR was mixed, sampled, and returned to cow at 2, 4, 8, 12, 16, and 24 h after feeding. 3 TMR was not mixed until sample collections at the respective time point. 4 X gm = geometric mean particle length determined by ASABE (2007). 5 S gm = particle length standard deviation determined by ASABE (2007). 6 Index = (X gm consumed up to time i ) / (X gm at time 0).

166 Table A-3. Percentage of rumen digesta particles (DM basis) retained on sieves after wet sieving when digesta samples were prepared with or without being squeezed through cheesecloth 150 Item 1 Squeezed Non-Squeezed SE P-value 9.5 mm mm mm mm mm mm Fluid-assoc/ Solid-assoc < Pore size of screens. 2 Rumen fluid-associated DM per unit of solid-associated DM.

167 Appendix B Effect of Feed Sorting on Chewing Behavior, Production, and Rumen Fermentation in Lactating Dairy Cows Journal of Dairy Science Vol. 93 No. 10, , 2010 D. D. Maulfair, G. I. Zanton, M. Fustini, and A. J. Heinrichs Abstract Ration sorting is thought to allow cows to effectively eat different rations throughout the d causing fluctuations in rumen fermentation patterns that can be detrimental to production and possibly animal health. The objective of this experiment was to study the effects of varying total mixed ration (TMR) particle size on sorting behavior of lactating dairy cows and to evaluate effects on chewing behavior, milk yield, milk components, and rumen fermentation. Eight multiparous, Holstein cows (90 ± 32 d in milk; 4 rumen cannulated) were randomly assigned to replicated 4 4 Latin squares. Cows were fed diets that varied in the chop length of dry grass hay. The diet consisted of 29.4% corn silage, 22.9% ground corn, 17.6% alfalfa haylage, and 11.8% dry grass hay on a dry matter basis. The percentage of hay particles > 26.9 mm was 4.2, 34.1, 60.4, and 77.6% for the short (S), medium (M), long (L), and extra long (XL) hays respectively. This resulted in the TMR of each diet having 1.5 (S), 6.5 (M), 8.6 (L), and 11.7% (XL) of particles > 26.9 mm. Daily ruminating min/kg dry matter intake (DMI; 19.3, 19.2, 22.4, and 21.3; S, M, L, and XL) and eating min/kg DMI (13.9, 14.6, 17.2, and 16.1; S, M, L, and XL) increased linearly as TMR particle size increased. Daily DMI decreased linearly as TMR particle size increased and was 26.9 (S), 27.0 (M), 24.1 (L), and 25.1 (XL) kg/d. No differences were

168 152 found in rumen volatile fatty acids and NH 3 and there were only slight changes in rumen ph. Milk production and milk components were also similar among diets. Despite large differences in particle size among these diets and certain chewing and ruminating differences, there were no changes in rumen fermentation, milk production, or milk components found in this study. Key Words: chewing, particle size, rumination, sorting Introduction The NRC (2001) recommends a minimum NDF level of 25% of DM and a forage NDF level of 19% of DM for lactating dairy cows. However, the NRC states that these values are based on cows fed: a TMR, alfalfa or corn silage as the predominant forage, forage with adequate particle size, and dry ground corn as the predominant starch source. These recommendations are therefore limited to rather specific conditions due to the limited data available and they do not define adequate particle size in a measurable manner. Fiber with adequate length is thought to increase chewing in cattle, which increases salivary secretion of NaHCO 3 and buffers the rumen digesta (Nocek, 1997; Allen, 1997; Krause et al., 2002b). Beauchemin et al. (2008) showed that rate (g/min) of salivation stayed constant during eating; however, changes in the rate of eating affected the amount of saliva secreted per unit of DMI when cows were fed barley silage, alfalfa silage, long-stemmed alfalfa hay, or barley straw. Particle size, DM, and NDF content of forages are factors affecting rate of eating and time spent eating (Bailey, 1961; Beauchemin et al., 2008) and it has been suggested that time spent chewing is a good measure of a feed s physical effectiveness (Balch, 1971; Sudweeks et al., 1981). Physically effective NDF (pendf), which combines the physical and chemical properties of a feedstuff, is commonly defined as the NDF concentration multiplied by the percentage of particles retained on a 1.18 mm sieve (1.65 mm;

169 153 screen diagonal) and greater. This definition presumes that the cows consume the ration as formulated. Dairy cows have been shown to selectively consume or sort their rations when fed a TMR. Cows generally sort against long particles and for finer particles (Leonardi and Armentano, 2003; Kononoff et al., 2003; DeVries et al., 2007). This is thought to create problems because, not only are they reducing the particle size of the diet consumed, but also reducing NDF intake because the longer particles of the TMR contain a higher proportion of NDF than the rest of the ration (Leonardi and Armentano, 2003). Feeding longer alfalfa hay versus chopped alfalfa hay increased sorting of rations, but intake of long particles still increased when fed the long alfalfa hay because of the higher concentration in the diet (Leonardi and Armentano, 2003). A potential problem for dealing with sorting on dairy farms is the fact that variability of sorting among cows can be very substantial, especially with the longest fraction (Leonardi and Armentano, 2003; Leonardi et al., 2005a). Therefore, the objective of this experiment was to study the effects of varying TMR particle size on sorting behavior and to evaluate its effects on chewing behavior, milk yield, milk components, and rumen fermentation in lactating dairy cows. Materials and Methods Diets, Cows, and Experimental Design Cows used in this research were cared for and maintained according to a protocol approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Eight lactating, multiparous, Holstein cows averaging 90 ± 32 DIM, weighing 642 ± 82 kg, and with parity of 2.25 ± 0.46 (mean ± SD) were randomly assigned to replicated 4 4 Latin squares;

170 154 1 square contained rumen cannulated cows. The periods were 21 d in length, with a 13-d adaptation period followed by an 8-d collection period. During each of the 4 periods, cows were fed 1 of 4 rations that contained identical feed ingredients and proportions but varied in the length of the dry grass hay included in the ration. Ingredients and their percentage of ration DM were: corn silage (29.4), ground corn (22.9), alfalfa haylage (17.6), grass hay (11.8), roasted soybeans (6.7), canola meal (5.7), heat-treated soybean meal (3.2), mineral/vitamin mix 1 (2.4), and salt (0.3). Grass hay inclusion level (20% of forage DM) was chosen based on previous research that showed it allowed for rations to be properly balanced while still creating adequate variations in particle size distributions between rations. Grass hay lengths of short (S), medium (M), long (L), and extra long (XL) were produced using several bale choppers. The XL and L hay was chopped once and twice, respectively, with a Case IH model 8610 bale processor (Case IH, Racine, WI). The M and S hay was chopped once and 3 times, respectively, with a Roto Grind model 760 tub grinder (Burrows Enterprises Inc., Greeley, CO); the S hay was additionally run once through a New Holland model 718 forage harvester (New Holland Ag, Racine, WI). All diets were mixed separately using an I. H. Rissler model 1050 TMR mixer (E. Rissler Mfg. LLC, New Enterprise, PA). Animals were housed in individual stalls, milked twice/d at 0700 and 1900 h and fed once/d at approximately 0730 h for ad libitum consumption and for 10% refusal to allow extensive opportunity to sort the ration. Feed was pushed up 3 times/d at 1230, 1730, and 2400 h. All rations were balanced to meet or exceed NRC (2001) requirements and water was available for ad libitum consumption. 1 Mineral and vitamin mix contained 12.2% Ca, 0.41% P, 3.88% Mg, 0.48% K, 0.37% S, 3.54% Na, 5.46% Cl, 222 mg/kg of Fe, 1,379 mg/kg of Zn, 455 mg/kg of Cu, 1,363 mg/kg of Mn, 11.2 mg/kg of Se, 7.33 mg/kg of Co, 18.5 mg/kg of I, 298 KIU/kg of vitamin A, 73.9 KIU/kg of vitamin D, 2,853 IU/kg of vitamin E.

171 155 Chewing Activity Eating and ruminating activity were recorded on d 14 through 18 of each period using Institute of Grassland and Environmental Research Behavior Recorders and Graze Jaw Movement Analysis Software (Ultra Sound Advice, London, UK) as described by Rutter (1997; 2000). These recorders analyze jaw movements of cattle and the software can determine eating or ruminating chews based on the amplitude and frequency of jaw movements. This procedure has been validated for use with cows housed in tie-stalls by Kononoff et al. (2002). Chewing was measured for all cows for two 24-h periods including while cows were being milked. Inter-meal intervals were separated from intra-meal intervals by analysis of the 2 d of chewing data (minimum interval 4 s) by a modification of the methodology reported by Tolkamp et al. (1998) and Yeates et al. (2001). Initial analysis of the probability density functions (PDF) of all data revealed that inter- and intra-meal histograms were each skewed toward the point that these histograms crossed. Yeates et al. (2001) reported that a Weibull distribution fitted to the last population of intervals adequately accounted for the skewness observed in that data set, whereas skewness in the first population was subdivided into 2 populations of intervals associated with drinking and non-drinking within-meal intervals. To account for the skewness present in the current data set, a Weibull distribution was fit to both populations to avoid potential overparameterization and allow for a skewed representation of the data. As the methodology employed is based upon the concept of satiety and the treatments administered in the current experiment were hypothesized to affect meal responses, a meal criterion estimate for the treatments that could be evaluated for statistical differences would be of value. Frequently, meal criteria are estimated from pooled or individual cow data; however, a Latin square experimental design with treatments applied over periods and cows limits the appropriateness of pooling the data. Since parameter estimates resulting from analysis of the individual cow within period

172 156 replicates are not estimated from the data without error or correlation, a nonlinear mixed model methodology was employed to estimate parameters while explicitly accounting for the design of the experiment within the framework of the nonlinear estimation procedure. The following Weibull, mixture model was fit to the observed cumulative frequencies (grouped by 0.1 log e second intervals; CDF) using the nonlinear mixed procedure in SAS (2006) using adaptive Gaussian quadrature with the Laplace approximation to the marginal likelihood: where: and ; t = log e (interval time); i= 1,2 = coefficient of the shape parameter of the Weibull distribution for the first or second population, respectively; m i= 1,2 = coefficient of the time point of inflection of the CDF and mode (maximum frequency) of the PDF for the first or second population, respectively; t c = coefficient of the meal criterion; jkl = random error component ~N[0,V 2 ], with V 2 = ; that is, the residual standard deviation is weighted by the squared root of the PDF. Additionally, each coefficient is the sum of the overall mean parameter estimate across period, treatment, and cow; the fixed effects of period and treatment; and the random effect of cow:

173 157 ; where: Z jkl = generic coefficient of the Weibull mixture model Z = estimate of coefficient Z across periods, treatments, and cows ZP j = fixed effect of period j on Z (j = 1, 2, 3, 4) subject to the constraint that ZP j = 0 ZT k = fixed effect of treatment k on Z (k = 1, 2, 3, 4) subject to the constraint that ZT k =0 zc l = random effect of cow l on Z (l=1,,8) ~N[0, 2 c ]. The model used to calculate meal criteria in this experiment was different from that used by Yeates et al. (2001) in 3 ways: 1) the CDF was fit with the nonlinear mixed procedure based on maximum likelihood estimation of the parameters; 2) the scale parameter of the Weibull distribution was replaced by the expected value parameter of the mode of the PDF or the time point of inflection of the CDF; 3) the parameter estimate of p (the proportion of intervals in the first population), was replaced by the expected value parameter for the time when the PDF of the 2 populations intersected (that is, the meal criterion: t c ) so that the meal criteria could be explicitly estimated concurrently with the remaining coefficients of the model. Additionally, the studentized residuals were observed to be heteroscedastic across time and appeared to vary in association with the PDF. Thus, the variance was weighted by the PDF, which removed the heteroscedasticity. The parameter 2 was observed to be a far-from-linear parameter according to the Hougaard skewness calculation; a substitution of Log e 2 for 2 was found to make the estimate close-to-linear and improve the estimates of the parameters. The variance-covariance matrix of the random effects was initially considered to be unstructured; however, the only covariance parameter estimate that significantly contributed to model fit (by the Bayesian information criterion) was between m 1 and 1, thus only this covariance parameter was retained in the final fitting of the model. An overall test of the significance of a treatment effect on each of

174 158 the parameters of the model was carried out by fitting the full and reduced model and using the likelihood ratio test. Predicted values for t c were computed using the parameter estimates and empirical Bayes estimates of the random effects; the number of meals was then calculated as the sum of intervals exceeding the predicted meal criterion within cow and period. Least squares means and standard errors of the within treatment parameter estimates were calculated from the solutions and the variance-covariance matrix for the nonlinear mixed model, respectively. Results were back-transformed differences between treatments evaluated using the 95% confidence intervals of the least squares means. Meal criteria intervals of 5 and 7 min were evaluated in addition to the calculated meal criteria. The 5-min interval was used for comparison to studies that used manual observation (Maekawa et al., 2002; Beauchemin et al., 2003; Leonardi et al., 2005b) or video observation (Bhandari et al., 2008) at 5-min intervals to determine chewing activity. The 7-min meal criterion was used because it is the default inter-meal interval for the Graze program (Rutter, 2000), and it is similar to research from several studies (Dado and Allen, 1993; Mooney and Allen, 2007) that used 7.5 min. Rumen Sampling On d 15 of each period, ruminal contents were collected from dorsal, ventral, cranial, caudal, and medial areas of the rumen at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18, 21.5, and 24.5 h after feeding (Kononoff et al., 2003). Collected digesta was mixed thoroughly, sampled, and filtered through 4 layers of cheesecloth. Rumen liquid ph was immediately determined using a handheld ph meter (phtestr 10 BNC, Oakton, Vernon Hills, IL). Approximately 15 ml of filtered liquid was placed into bottles containing 3 ml of 25% metaphosphoric acid and 3 ml of 0.6% 2-ethylbutyric acid (internal standard) and stored at -20 C. After thawing, samples were

175 159 centrifuged 3 times at 4000 g for 30 min at 4 C to obtain a clear supernatant and were analyzed for NH 3 using a phenol-hypochlorite assay (Broderick and Kang, 1980) and VFA concentration using gas chromatography (Yang and Varga, 1989). Feed, Refusal, and Particle Size Analysis Feed bunk contents for each animal were weighed and sampled on d 20 and 21 at 0, 2, 4, 8, 12, 16, and 24 h after feeding to determine particle size distribution and DM of the remaining feed. At 0, 8, 16, and 24 h after feeding refusals were also analyzed for NDF and starch content to determine intake of these components between each time point. All samples were sieved in the American Society of Agriculture and Biological Engineers forage particle separator, which can determine 6 particle fractions (> 26.9, > 18.0, > 8.98, > 5.61, > 1.65, and < 1.65 mm; screen diagonal; ASABE, 2007). Since > 1% of material was retained on the top screen, 3 samples of each diet were randomly selected, and all particles retained on the top screen were measured manually (with ruler) before drying. Whole samples were then placed in a forced air oven at 55 C for 48 h to determine DM content. Geometric mean particle length (X gm ) and standard deviation of the particle length (S gm ) were calculated according to ASABE (2007) procedure. Samples were then ground (1 mm screen; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ) to determine NDF using heat-stable α-amylase and Na 2 SO 3 according to Van Soest (1991) and ground (0.5 mm screen; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ) to analyze starch using a modified procedure from Knudsen (1997). Samples of forages and TMR were taken 3 times/wk, composited by period, and analyzed by Cumberland Valley Analytical Services, Inc. (Hagerstown, MD) for CP, ADF, NDF, ash, NFC, and NE L. There were 2 procedures used to calculate pendf; pendf 8.98 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of PSPS) and pendf 1.65 = % of particles > 1.65 mm NDF of whole sample (similar to

176 160 top 3 sieves of PSPS). A sorting index based on the refusals was calculated for the particle size fractions at 2, 4, 8, 12, 16, and 24 h after feeding and for NDF and starch at 8, 16, and 24 h after feeding. Sorting activity was calculated as the actual intake of each fraction (Y 1 to pan) expressed as a percentage of the expected intake. Expected intake of Y i equals intake multiplied by the fraction of Y i in the TMR (Leonardi and Armentano, 2003). Sorting indices were calculated using both the expected intake since time point 0 (cumulative) and the expected intake since the previous time point (interval). Additionally, X gm sorting indices were calculated for the same time points by dividing the X gm of TMR consumed up to each time point by X gm at time 0. Values > 1 indicate cows were consuming rations with longer particle size and values < 1 indicate cows were consuming rations with shorter particle size than the diets fed. Milk Production Milk production was recorded and samples were taken on d 20 and 21 at morning and evening milkings. Samples were collected and preserved using 2-bromo-2-nitropropane-1,3 diol. Milk samples were analyzed for fat, true protein, lactose, MUN, and SCC by the Dairy One milk testing laboratory (State College, PA) using infrared spectrophotometry (Foss 605B Milk-Scan; Foss Electric, Hillerod, Denmark). Statistical Analyses Statistical analysis was conducted using PROC MIXED of SAS (2006). Dependent variables were analyzed as a 4 4 Latin square design. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997) and repeated measurements for rumen samples and refusal particle size, NDF, and starch were analyzed using the first order

177 161 autoregressive covariance structure (Littell et al., 1998), as well as terms for time and interaction of treatment by time. Because of unequally spaced rumen sampling, the weighted mean daily ph, NH 3, and VFA concentrations were determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). The data were analyzed for orthogonal contrasts using the fed TMR X gm that was corrected for unequal spacing according to Robson (1959). All data are presented as least squares means and treatment effects are considered significant when P 0.05 and a trend when P Results and Discussion Chemical Composition and Particle Size Distribution The chemical composition, particle size distribution, and X gm of forages included in the rations are shown in Table B-1. Particle size was determined with the ASABE forage particle separator because particle length of some diets was so great that the Penn State Particle Separator (PSPS) did not adequately separate samples. The PSPS particle fractions and their approximate equivalent ASABE separator screens are: top ( mm), middle (8.98 mm), lower ( mm), and pan (pan). The grass hays had large differences in particle size, particularly with the particles retained on the 26.9 mm screen although all particle fractions had differences among the hays. In addition, the X gm increased greatly from the shortest to the longest ration, with a 13- fold difference between S grass hay and XL grass hay. The M hay had lower ADF and NDF and higher NFC values than other hay lengths; this was probably due to individual bale variation. Although all bales were from the same field and cutting, each length of hay was composed of different bales. These differences however did not affect TMR chemical composition. Particle size distribution of the fed TMR also varied greatly (Table B-2). The greatest differences were in

178 162 the particle fraction > 26.9 mm. The only particle fraction that did not show differences among diets was particles retained on the pan. Measured mean particle lengths for the top screen to calculate X gm were: 74.8 ± 6.6, 84.5 ± 2.6, ± 9.1, and ± 3.6 (mean ± SD, mm) for short (S), medium (M), long (L), and extra long (XL) diets, respectively. Particle lengths (geometric mean ± SD, mm; ASABE, 2007) of the fed rations were: 4.46 ± 0.13, 5.10 ± 0.13, 5.32 ± 0.13, and 5.84 ± 0.13 for S, M, L, and XL diets, respectively. The X gm of the rations were approximately equally spaced with differences averaging 0.46 mm between each ration from S to XL; S gm increased linearly with increasing ration particle size. Chemical compositions were similar among the rations with only slight differences in DM, linearly increased with increasing particle size. It is interesting to note that although there were large differences in mean particle length among rations pendf 1.65 remained constant. This occurs because all particles greater than 1.65 mm are weighted equally regardless of length, a weakness of calculating pendf this way. There was a linear trend for pendf 8.98 to increase with increasing TMR particle size but the numerical difference was small. Ration Sorting Figure B-1 shows the X gm of refusals increased in all rations throughout the d. The amount of change varied by diet; the shortest diet changed very little between feeding and removal of orts while the longest diet had a very drastic change in X gm during the same time period. This effect can also be seen in Figure B-2, where the percentage of particle fractions in refusals in relation to their percentage in fed TMR is shown. The top screen percentages increased in all diets with very large increases in the longest 2 rations (107, 157, 193, and 283%; S, M, L, and XL). The pan percentages decreased in all rations, again with greater changes between TMR fed and TMR refused as the TMR particle size increased. The pan percentages after 24 h were:

179 163 82, 74, 61, and 49% of the amount in the fed TMR for the S, M, L, and XL rations respectively. Only graphs of the top screen and pan are shown due to space constraints, but the 2 largest screens showed very similar patterns, the middle 2 screens did not show substantial differences among the rations or from the original diet, and the bottom 2 fractions showed similar patterns. These results are supported by the finding that NDF concentration in refusals increased more in the longer rations throughout the d than the shorter rations (Figure B-3). In addition, the level of starch decreased in the 2 longest rations and remained unchanged or even increased in the 2 shorter rations. These data would lead to the conclusion that animals consumed very different amounts of starch and NDF during the d due to sorting activity. However, when the amount of these components consumed per d was calculated, rations had similar levels of NDF and starch intakes (Table B-3). Cumulative sorting indices for particle size, NDF, and starch intake expressed as the actual intake of each component divided by the predicted intake of that component are shown in Figures B-4 and B-5. Sorting indices of S and M rations for the top screen were higher than L and XL sorting indices at 8 h and less, and after 8 h there were no differences among the rations. Pan sorting indices showed that at 2 h L and XL were highest, M was intermediate, and S was lowest. After 2 h the differences diminished and eventually disappeared by 12 h. Only the top screen and pan fractions are shown for space saving, but the top 3 screens showed similar patterns, the fourth screen did not show substantial differences among the rations, and the bottom 2 fractions showed similar patterns. The S and M rations had higher NDF sorting indices and lower starch sorting indices than L and XL rations at 8 h after feeding. By 24 h after feeding there were no longer differences among the 4 rations for NDF or starch sorting indices. Figure B-6 shows the cumulative X gm selection index which combines all six particle fractions to make an easier comparison. The S and M rations had much higher selection indices for the first 8 h than the L and XL rations; the ration being consumed was longer than the ration fed for S and M and shorter for L and XL. After 8 h the rations became much

180 164 closer in values but remained different, at 24 h the longest 3 rations remained below 1.0 and S was equal to 1.0. Intake of DM, NDF, Starch, and Particle Fractions There was a linear trend for decreased DMI as TMR particle size increased (Table B-3); this trend was probably due to increased gut fill associated with the bulkier diets, as has been noted previously (Kononoff and Heinrichs, 2003; Leonardi et al., 2005b). These results are contrary to other studies (Krause et al., 2002a; Beauchemin and Yang, 2005) that showed no effect of forage particle size on DMI. The diets had NDF and starch intakes that were not different despite very different sorting characteristics among the rations throughout the d. Analysis of intake of individual particle fractions revealed that intake of particles retained on the 26.9-mm sieve increased linearly from 0.39 to 2.43 kg/d with increasing ration particle size, as seen in Table B-3. In contrast, intake of particles retained on the and 8.98-mm sieves showed a linear decrease as particle size of the ration increased. Intake of particles retained on the 5.61-mm sieve was similar among rations. Intake of particles retained on the 1.65-mm screen and the pan were also not different among rations, most likely influenced by the equal concentrate fed to all groups. The consumed X gm was much closer among rations than the X gm fed. Consumed X gm for all rations was between 4.44 and 5.10 mm. This probably occurred because cows on the shorter rations made up for being offered fewer particles > 26.9 mm by increasing their intake of particles retained on the and 8.98-mm sieves. Also, intake of NDF and starch remained similar among rations despite different intakes of particle fractions because the particles retained on the top 3 sieves that varied in intakes were primarily grass hay, and thus had similar composition. Finally, refusal percentages met or slightly exceeded the goal of 10% and were not different among rations.

181 165 Chewing Activity Observed meal criteria (Table B-4) were 7.6, 13.8, 10.5, and 11.2 min for S, M, L, and XL rations, respectively, with S meal criterion being significantly less than the other rations. The modes of the intra-meal intervals were found to be 13.7, 13.5, 14.1, and 12.0 s for S, M, L, and XL respectively and were not different from each other. The modes of the inter-meal intervals were 51.8, 72.1, 58.7, and 68.4 min for S, M, L, and XL, with S having a shorter interval than the other diets. The DMI per meal was determined to be similar among diets and averaged 2.35 kg/meal. There were no differences among diets for ruminating, eating, or total chewing time per d for all meal criteria and averaged 515, 388, and 903 min/d respectively (Table B-5). When chewing activity was expressed as time/kg DMI there were significant linear contrasts for increased ruminating, eating, and total chewing time/kg DMI as TMR particle size increased and averaged 20.6, 15.5, and 36.0 min/kg respectively. Ruminating, eating, and total chewing activity values expressed as min/d were higher than those reported in other studies (Kononoff and Heinrichs, 2003; Beauchemin et al., 2003; Beauchemin and Yang, 2005). However, when these data were expressed as min/kg DMI they were found to be similar to the data reported in these same studies. This may be explained by the fact that the DMI in this study was on average 4.4 kg/d higher than these 3 other studies increasing the total amount of daily chewing activity. Chewing activity expressed as min/d was probably not different among rations because DMI decreased linearly while chewing activity (min/kg DMI) increased linearly with increasing TMR particle size effectively nullifying the changes. The mean total time that chewing activity was recorded was not different among rations and was 23.8 h. Chewing activity was also expressed as number of chews/d and chews/kg DMI. Again there were no differences among diets for ruminating, eating, and total chewing when calculated on a daily basis. There were no changes in number of ruminating chews/kg DMI, but there was a linear trend for number of eating and total

182 166 chews/kg DMI to increase as TMR particle size increased. There were no differences among diets in number of meals eaten/d. Similar to chewing time/d, number of boli/d showed no differences among diets, but when expressed as boli/kg DMI there was a linear increase with increasing TMR particle size. Chewing data were also analyzed using 5-min and 7-min meal criteria. Eating and total chewing time increased slightly as length of the meal criteria interval increased. The number of eating and total chews increased slightly as length of the meal criteria interval increased. Number of meals/d decreased as length of the meal criteria increased. Therefore, these data show that exact meal criterion used is not important as there are only small changes in values of variables and there are no changes to conclusions made based on these values. Rumen Characteristics The weighted mean rumen ph was similar among rations (Table B-6) but there was a trend for a quadratic contrast for these data. Kononoff and Heinrichs (2003) found a similar quadratic contrast when increasing alfalfa haylage particle size in TMR. The average weighted mean for all diets was This was similar to results from some studies (Krause et al., 2002b; Beauchemin et al., 2003), higher than some (Beauchemin and Yang, 2005), and lower than others (Kononoff et al., 2003; Leonardi et al., 2005b). There were no differences found in the minimum rumen ph among rations and average minimum ph for all rations was There was a linear tendency for decreasing maximum rumen ph with increasing ration particle size. Perhaps the intake of starch at any one time was not great enough to overcome the buffering capacity of the rumen with a large amount of forage still retained from the previous d. This slow digesting fiber allows the cow to create a more uniform rumen environment than the actual intake of feed would allow otherwise. The weighted mean, minimum, and maximum NH 3 concentrations were found

183 167 to have no significant contrasts among rations and the mean averaged 8.3 mg/dl. These results are similar to what was found by Kononoff et al. (2003). The concentrations of acetate, propionate, butyrate, valerate, isovalerate, isobutyrate were also shown to have no significant contrasts among diets. Concentrations of all VFA measured were similar to those found by Kononoff and Heinrichs (2003) Milk Production and Composition Milk production averaged 38.7 kg/d and the rations had no effect on milk, FCM, fat, protein, or lactose yield (Table B-7). Milk fat percentage was similar among diets, as was milk protein percentage. These results are in agreement with some studies that found that changes in forage particle size did not affect milk production or components (Krause et al., 2002a; Beauchemin et al., 2003; Bhandari et al., 2008) but are in disagreement with others that found that changes in forage particle size influence milk production or components (Kononoff et al., 2003; Leonardi et al., 2005b). There were linear trends for lactose and MUN to decrease with increasing ration particle size, but there were only slight numerical differences. There was a significant quadratic contrast for SCC but the reason for tendency is not apparent. One cow was removed from the milk production and composition analysis because of chronically high SCC. Removing this cow from the analysis decreased average SCC by 135,350 cells/ml and increased average percent fat by 0.09; however, it did not change the conclusions for any production parameters.

184 168 Conclusions In this experiment 4 diets that varied only in the particle size of their grass hay were fed to lactating dairy cows to measure differences in sorting activity and the effect of these differences on production parameters. Great differences were observed among rations when sorting activity was determined by the change in composition of refusals (particle size, NDF, and starch) compared to the ration fed. However, actual intake of these components after 24 h was similar for all rations and as a result milk production, milk components and rumen characteristics were similar among the rations. Therefore, cows were essentially receiving different rations throughout the d, but the final daily outcome was not different. When measuring sorting activity in lactating dairy cattle it is important not to only consider composition of the orts (which comprise only a small percentage of the daily intake) but also actual intakes of various ration components. In addition, although the diets fed varied greatly in X gm, the X gm of what was consumed by cows were very similar. Cows on the S ration ate a ration similar in X gm to what was offered, and cows on all other rations ate a shorter ration than what was offered. Since the ration the cows actually consumed had similar X gm and the cows sorted the ration that was offered, perhaps these cows were sorting to achieve a desired X gm. If this is the case, a ration with the proper X gm may be able to limit or eliminate ration sorting by lactating cows. Acknowledgments This research was supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture.

185 169 References Allen, M. S Relationship between fermentation acid production in the rumen and the requirement for physically effective fiber. J. Dairy Sci. 80: American Society of Agricultural and Biological Engineers Method of determining and expressing particle size of chopped forage materials by screening. ANSI/ASAE. S424.1: Bailey, C. B Saliva secretion and its relation to feeding in cattle. Br. J. Nutr. 15: Balch, C. C Proposal to use time spent chewing as an index of the extent to which diets for ruminants possess the physical property of fibrousness characteristic of roughages. Br. J. Nutr. 26: Beauchemin, K. A., L. Eriksen, P. Norgaard, and L. M. Rode Short Communication: Salivary secretion during meals in lactating dairy cattle. J. Dairy Sci. 91: Beauchemin, K. A., and W. Z. Yang Effects of physically effective fiber on intake, chewing activity, and ruminal acidosis for dairy cows fed diets based on corn silage. J. Dairy Sci. 88: Beauchemin, K. A., W. Z. Yang, and L. M. Rode Effects of particle size of alfalfa-based dairy cow diets on chewing activity, ruminal fermentation, and milk production. J. Dairy Sci. 86: Bhandari, S. K., S. Li, K. H. Ominski, K. M. Wittenberg, and J. C. Plaizier Effects of the chop lengths of alfalfa silage and oat silage on feed intake, milk production, feeding behavior, and rumen fermentation of dairy cows. J. Dairy Sci. 91: Broderick, G. A., and J. H. Kang Automated simultaneous determination of ammonia and total amino acids in ruminal fluid and in vitro media. J. Dairy Sci. 63: Dado, R. G., and M. S. Allen Continuous computer acquisition of feed and water intakes, chewing, reticular motility, and ruminal ph of cattle. J. Dairy Sci. 76: DeVries, T. J., K. A. Beauchemin, and M. A. G. von Keyserlingk Dietary forage concentration affects the feed sorting behavior of lactating dairy cows. J. Dairy Sci. 90: Kenward, M. G., and J. H. Roger Small sample inference for fixed effects from restricted maximum likelihood. Biometrics. 53: Knudsen, K. E. B Carbohydrate and lignin contents of plant materials used in animal feeding. Anim. Feed Sci. Technol. 67: Kononoff, P. J., and A. J. Heinrichs The effect of reducing alfalfa haylage particle size on cows in early lactation. J. Dairy Sci. 86:

186 170 Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86: Kononoff, P. J., H. A. Lehman, and A. J. Heinrichs Technical Note--A comparison of methods used to measure eating and ruminating activity in confined dairy cattle. J. Dairy Sci. 85: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002a. Effects of forage particle size and grain fermentability in midlactation cows. I. Milk production and diet digestibility. J. Dairy Sci. 85: Krause, K. M., D. K. Combs, and K. A. Beauchemin. 2002b. Effects of forage particle size and grain fermentability in midlactation cows. II. Ruminal ph and chewing activity. J. Dairy Sci. 85: Leonardi, C., and L. E. Armentano Effect of quantity, quality, and length of alfalfa hay on selective consumption by dairy cows. J. Dairy Sci. 86: Leonardi, C., F. Giannico, and L. E. Armentano. 2005a. Effect of water addition on selective consumption (sorting) of dry diets by dairy cattle. J. Dairy Sci. 88: Leonardi, C., K. J. Shinners, and L. E. Armentano. 2005b. Effect of different dietary geometric mean particle length and particle size distribution of oat silage on feeding behavior and productive performance of dairy cattle. J. Dairy Sci. 88: Littell, R. C., P. R. Henry, and C. B. Ammerman Statistical analysis of repeated measures data using SAS procedures. J. Anim Sci. 76: Maekawa, M., K. A. Beauchemin, and D. A. Christensen Effect of concentrate level and feeding management on chewing activities, saliva production, and ruminal ph of lactating dairy cows. J. Dairy Sci. 85: Mooney, C. S., and M. S. Allen Effect of Dietary Strong Ions on Chewing Activity and Milk Production in Lactating Dairy Cows. J. Dairy Sci. 90: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC. Nocek, J. E Bovine acidosis: Implications on laminitis. J. Dairy Sci. 80: Robson, D. S A simple method for constructing orthogonal polynomials when the independent variable is unequally spaced. Biometrics. 15: Rutter, S. M Graze: A program to analyze recordings of the jaw movements of ruminants. Behav. Res. Meth. Ins. C. 32: Rutter, S. M., R. A. Champion, and P. D. Penning An automatic system to record foraging behaviour in free-ranging ruminants. Appl. Anim. Behav. Sci. 54:

187 171 SAS InstituteSAS User's Guide: Statistics. Version SAS Inst. Inc., Cary, NC, Shipley, R. A., and R. E. Clark Tracer Methods for In Vivo Kinetics. Academic Press, New York, NY. Sudweeks, E. M., L. O. Ely, D. R. Mertens, and L. R. Sisk Assessing minimum amounts and form of roughages in ruminant diets: Roughage value index system. J. Anim. Sci. 53: Tolkamp, B. J., D. J. Allcroft, E. J. Austin, B. L. Nielsen, and I. Kyriazakis Satiety splits feeding behaviour into bouts. J. Theor. Biol. 194: Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, C.-M. J., and G. A. Varga Effect of three concentrate feeding frequencies on rumen protozoa, rumen digesta kinetics, and milk yield in dairy cows. J. Dairy Sci. 72: Yeates, M. P., B. J. Tolkamp, D. J. Allcroft, and I. Kyriazakis The use of mixed distribution models to determine bout criteria for analysis of animal behaviour. J. Theor. Biol. 213:

188 172 Table B-1. Chemical composition and particle size distributions determined with the ASABE particle separator for corn silage, alfalfa haylage, and short (S), medium (M), long (L), or extra long (XL) grass hay Corn Alfalfa Grass hay Item silage haylage S M L XL SEM P-value Particle size, as-fed % retained mm d 34.1 c 60.4 b 77.6 a 4.10 < mm a 12.9 a 11.5 a 6.83 b mm a 15.7 a 10.4 b 5.30 c 1.38 < mm a 9.64 b 6.21 c 3.66 d 0.65 < mm a 12.7 b 6.54 c 4.17 c 0.82 < 0.01 Pan a 15.0 b 4.91 c 2.43 c 1.61 < 0.01 X 2 gm, mm c 14.6 c 38.0 b 65.4 a 3.67 < 0.01 S 3 gm, mm c 4.93 a 4.22 b 3.43 c 0.18 < 0.01 Composition, % of DM DM a 89.8 ab 90.1 ab 89.4 b CP ADF NDF pendf pendf Ash NFC NE L, Mcal/kg a d Means within a row with different superscripts differ (P 0.05). 1 Approximate equivalency to PSPS: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 X gm = geometric mean particle length determined by ASABE (2007). 3 S gm = particle length standard deviation determined by ASABE (2007). 4 Physically effective NDF 8.98 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of PSPS). 5 Physically effective NDF 1.65 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of PSPS).

189 173 Table B-2. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay Item S M L XL SEM Linear Quadratic Particle size, as-fed % retained mm < mm < mm < mm < mm < Pan X 2 gm, mm < S 3 gm, mm < Composition, % of DM DM, % CP ADF NDF Forage NDF pendf pendf Ash Starch NE L, Mcal/kg Approximate equivalency to PSPS: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 X gm = geometric mean particle length determined by ASABE (2007). 3 S gm = particle length standard deviation determined by ASABE (2007). 4 Physically effective NDF 8.98 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of PSPS). 5 Physically effective NDF 1.65 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of PSPS.

190 174 Table B-3. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on DM, NDF, and starch intake at various times after feeding and total consumption (measured 24 h after feeding) of various sized particles Item S M L XL SEM Linear Quadratic DMI, kg 8 h h h NDF, kg 8 h h h Starch, kg 8 h h h Particles consumed, kg mm < mm < mm < mm mm Pan X 2 gm, mm < S 3 gm, mm < Refusal, % Approximate equivalency to PSPS: top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 X gm = geometric mean particle length determined by ASABE (2007). 3 S gm = particle length standard deviation determined by ASABE (2007).

191 Table B-4. Observed meal characteristics for diets containing short (S), medium (M). long (L), or extra long (XL) grass hay 175 Item S M L XL SEM 1 Intra-meal interval, s 2 Mode % CI 16.8, , , ,9.8. Inter-meal interval, min 3 Mode 51.8 b 72.1 a 58.7 a 68.4 a. 95% CI 70.1, , , ,50.5. Meal criterion, min 7.6 b 13.8 a 10.5 a 11.2 a. 95% CI 10.0, , , ,8.4. DMI/meal, kg a b Means within a row with different superscripts differ after transforming (P 0.05). 1 For model output, back-transformed 95% confidence intervals are shown. 2 Intra-meal interval = bout of no eating within meals. 3 Inter-meal interval = bout of no eating outside of meals. 4 Calculated based on daily DMI

192 176 Table B-5. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on chewing behavior as determined by observed meal criteria 1 Item S M L XL SEM Linear Quadratic Min/d Ruminating Eating Total chewing Total time recorded 1,424 1,434 1,425 1, Min/kg Ruminating Eating Total chewing Chews/d Ruminating 23,690 23,874 20,711 25,100 2, Eating 19,699 20,322 19,462 21,775 1, Total chews 43,388 44,196 40,173 46,874 3, Chews/kg Ruminating Eating Total chews. 1,612 1,615 1,857 1, Meals/d Boli, number/d Boli, number/kg DMI Observed meal criteria use intervals predicted from current dataset.

193 177 Table B-6. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on rumen fermentation Item S M L XL SEM Linear Quadratic Rumen ph Weighted average Minimum Maximum NH 3, mg/dl Weighted average Minimum Maximum VFA Weighted average, µm/ml Acetate Propionate Butyrate Valerate Isovalerate Isobutyrate Weighted averages determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972).

194 178 Table B-7. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on milk production and components 1 Item S M L XL SEM Linear Quadratic Milk yield, kg/d % FCM, kg/d Fat, % Fat, kg/d Protein, % Protein, kg/d Lactose, % Lactose, kg/d MUN, mg/dl SCC, 1,000 cells/ml One cow was removed from analysis due to chronic high SCC.

195 179 Figure B-1. Effect of feeding TMR of increasing particle size on refusal geometric mean particle size.

196 180 A B Figure B-2. Effect of feeding TMR of increasing particle size on refusal particle distribution as a percentage of original diet. Selected data shown; 26.9-mm sieve (A) and pan (B).

197 181 A B Figure B-3. Effect of feeding TMR of increasing particle size on refusal NDF (A) and starch (B) concentration.

198 182 A B Figure B-4. Effect of feeding TMR of increasing particle size on cumulative particle size selection index. Selected data shown; 26.9-mm sieve (A) and pan (B).

199 183 A B Figure B-5. Effect of feeding TMR of increasing particle size on cumulative NDF (A) and starch (B) selection indices.

200 184 Figure B-6. Effect of feeding TMR of increasing particle size on cumulative geometric mean length (X gm ) selection index.

201 Appendix C Effect of Varying TMR Particle Size on Rumen Digesta and Fecal Particle Size and Digestibility in Lactating Dairy Cows Journal of Dairy Science Vol. 94 No. 7, , 2011 D. D. Maulfair, M. Fustini, and A. J. Heinrichs Abstract The objective of this experiment was to evaluate the effects of feeding rations of different particle sizes on rumen digesta and fecal matter particle size. Four rumen cannulated, multiparous, Holstein cows (104 ± 15 d in milk) were randomly assigned to a 4 4 Latin square. The diets consisted of 29.4% corn silage, 22.9% ground corn, 17.6% alfalfa haylage, and 11.8% dry grass hay (20% of forage dry matter) on a dry matter basis. Dry grass hay was chopped to 4 different lengths to vary the total mixed ration particle size. Geometric mean particle sizes of the rations were 4.46, 5.10, 5.32, and 5.84 mm for Short, Medium, Long, and Extra Long diets respectively. The ration affected rumen digesta particle size for particles 3.35 mm, and had no effect on distribution of particles < 3.35 mm. All rumen digesta particle size fractions varied by time after feeding; with soluble particle fractions increasing immediately after feeding while 0.15, 0.6, and 1.18 mm particle size fractions decreased slightly after feeding. Particle fractions > 1.18 mm had ration by time interactions. Fecal neutral detergent fiber and indigestible neutral detergent fiber concentrations decreased with increasing total mixed ration particle size. Fecal particle size expressed as total geometric mean particle length followed this same tendency. Fecal particle size expressed as retained geometric mean particle length averaged 1.13 mm with greater

202 186 than 36% of particle being larger than 1.18 mm. All fecal nutrient concentrations measured were significantly affected by time after feeding with NDF and INDF increasing after feeding and peaking at about 12 h later and then decreasing to pre-prandial levels. Starch concentrations were determined to have the opposite effect. Additionally, apparent digestibility of diet nutrients was analyzed and dry matter digestibility tended to decrease with increasing total mixed ration particle size, while other nutrient digestibilities were not different among rations. These results show that the critical size for increased resistance to rumen escape is larger than 1.18 mm and this critical size is constant throughout the d. This study also concludes that, when using average quality grass hay to provide the range of particle sizes fed, dry matter digestibility tends to decrease with increasing ration particle size. Key Words: digestibility, particle size, rumen escape Introduction The sieve size 1.18 mm has been widely used as the size in which feed particles retained on or above are considered physically effective for dairy cows. This number originated from research of Evans et al. (1973) and Poppi et al. (1980; 1981), where resistance of particles leaving the rumen of cattle and sheep was measured. It was determined that 1.18 mm was a threshold particle size for both cattle and sheep for greatly increased resistance to particles leaving the rumen and < 5% of fecal particles are generally retained on a 1.18-mm sieve (Poppi et al., 1980, 1981). It should be noted that a wet sieving technique was used in these studies to measure particle size and this procedure is very different from the dry vertical sieving procedure used by Mertens (1997) to develop the physical effectiveness factor of feeds (using particles retained on a 1.18-mm sieve) that is used by some ration formulation software today. Therefore it should not be assumed that these two different sieving methods will produce similar results. Some researchers

203 187 have suggested that the critical particle size for rumen escape in dairy cattle may be larger than 1.18 mm (Yang et al., 2001; Oshita et al., 2004), however determining this has proven difficult. Also little is known if diet particle size or time after feeding affects this critical particle size for passage from the rumen. There is some controversy regarding effect of ration particle size on DM digestibility (DMD). Kononoff and Heinrichs (2003a) and Yang and Beauchemin (2005) reported that increasing ration particle size increased DMD; however, Kononoff and Heinrichs (2003b) observed that increasing ration particle size decreased DMD. In addition there are several studies that reported no effect of ration particle size on DMD (Krause et al., 2002; Yang and Beauchemin, 2006; 2007). Clearly this effect is variable based on other aspects of the diet or management. Therefore, the objective of this experiment was to study effects of varying TMR particle size on rumen digesta and fecal particle size in lactating dairy cows to determine the critical size for particles leaving the rumen and if rumen and fecal particle size change throughout the d and according to diet particle size. Materials and Methods Diets, Cows, and Experimental Design Cows used in this research were cared for and maintained according to a protocol approved by The Pennsylvania State University Institutional Animal Care and Use Committee. Four lactating, multiparous, rumen cannulated, Holstein cows averaging 104 ± 15 DIM, weighing 659 ± 88 kg, and with parity of 2.25 ± 0.50 (mean ± SD) were randomly assigned to a 4 4 Latin square. Periods were 21 d in length, with a 13-d adaptation period followed by an 8-d collection period. During each of the 4 periods, cows were fed 1 of 4 rations that contained identical feed

204 188 ingredients and proportions but varied in the length of dry grass hay included in the ration. Ingredients and their percentage of ration DM were: corn silage (29.4), ground corn (22.9), haylage (17.6), grass hay (11.8), roasted soybeans (6.7), canola meal (5.7), heat-treated soybean meal (3.2), mineral/vitamin mix (2.4), and salt (0.3). More detailed information regarding diets was reported by Maulfair et al. (2010). Rations were designated as short (S), medium (M), long (L), and extra long (XL). Animals were housed in individual stalls, milked twice/d at 0700 and 1900 h and fed once/d at approximately 0730 h for ad libitum consumption with 12% refusal to allow maximum opportunity to sort the ration. Feed was pushed up 3 times/d at 1230, 1730, and 2400 h. All rations were balanced to meet or exceed NRC (2001) requirements, and water was available ad libitum. Rumen Sampling On d 15 of each period, ruminal contents were collected from dorsal, ventral, cranial, caudal, and medial areas of the rumen at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding (Kononoff et al., 2003b). Collected digesta was mixed thoroughly, sampled, and filtered through 4 layers of cheesecloth. Solid portions of digesta samples retained on cheesecloth were stored at -20 C for later determination of particle size distribution via the wet sieving technique of Maulfair and Heinrichs (2010). Maulfair and Heinrichs (2010) determined that squeezing rumen digesta through cheesecloth before wet sieving had no effect on particle size distribution of particles > 0.15 mm but reduced the amount of soluble DM contained in the sample. Representative samples (approximately 30 g) were mixed in 1 L of water and soaked for 10 min. Samples were then placed on a series of stacked sieves (sizes 0.15, 0.6, 1.18, 3.35, 6.7, 9.5 mm; VWR, Arlington Heights, IL) contained in a Retsch AS 200 Control sieve shaker (Retsch, Haan, Germany) and sieved in duplicate. Samples were sieved for 10 min at 2.5 mm

205 189 amplitude with the sprayer ring located between and 1.18-mm screens and cold water flow rate at approximately 1.5 to 2.0 L/min to ensure particles were separated thoroughly. Sieve contents were rinsed into a funnel with rumen in situ bags (5 x 10 cm, 53 μm pore size; ANKOM, Macedon, NY) attached to the stem to collect the sample. Bags were then dried in a forced air oven at 55 C for 24 h and weighed to determine DM retained on each sieve. A portion of each sample was also dried at 55 C for 24 h in a forced air oven without sieving to determine DM content of the original sample. Soluble fraction of the sample was calculated as the DM lost during sieving and drying. Data were analyzed using each particle fraction as a percentage of DM retained on 0.15-mm screen (retained) and also as the percentage of DM of the entire sample sieved (total). Fecal Sampling Fecal sampling occurred at the same time points as rumen sampling (d 15 at 0.0, 1.5, 3.5, 5.5, 8.5, 11.5, 14.5, 18.0, 21.5, and 24.5 h after feeding) via grab samples from the rectum. Samples were stored at -20 C until later determination of particle size distribution and concentration of DM, NDF, indigestible NDF (INDF), starch, and ash. Particle size of subsamples was determined using the same wet sieving technique used for rumen digesta, with the exception of eliminating the top screen (9.5 mm). Geometric mean particle length (X gm ) and standard deviation of particle length (S gm ) were calculated according to American Society of Agricultural and Biological Engineers (ASABE) (2007) procedure. X gm was calculated using 2 procedures; the first, retained X gm (X gm Ret), only considered particles retained on the 0.15-mm screen or larger, the second procedure, total X gm (X gm Tot), considered all particle fractions including the soluble fraction that passed through the bottom screen (0.15 mm). Mean particle length of the soluble fraction was assumed to be mm, which is half of the diagonal screen

206 190 diameter (0.212 mm) of the bottom screen; this is the assumption that ASABE (2007) uses for mean length of particles on the pan. Subsamples were also placed in a forced air oven at 55 C for 48 h to determine DM content and were then ground (1-mm screen; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ) to determine NDF using heat-stable α-amylase and Na 2 SO 3 according to Van Soest (1991) and ground (0.5-mm screen; Wiley Mill, Arthur H. Thomas Co. Inc., Swedesboro, NJ) to analyze starch using a modified procedure from Knudsen (1997; modification included 150 mg of sample, 45 units of amyloglucosidase, and analysis of released glucose monomers with procedure no. 1075, Stanbio Laboratory Inc.) For INDF determination, subsamples were enclosed in F57 filter bags (ANKOM Technology, Macedon, NY) in sextuplicate, then incubated in the rumen of 2 cows (each cow incubated 3 bags of each sample) for 12 d. After removal from the rumen, bags were rinsed in cold water by hand until water was almost clear. Bags were then dried in a forced air oven at 55 C for 48 h and later processed using the same procedure used for NDF determination. Ash was determined by combustion at 600 C for 6 h (AOAC, 1990). Digestibility Dry matter intakes were recorded daily and feed bunk contents were sampled at 0 and 24 h after feeding on d 21 and 22 and were analyzed for DM, NDF, INDF, and starch using identical procedures to those of fecal samples. Intakes of NDF, INDF, and starch were determined by subtracting the amount of each in refusals (refused TMR weight refused TMR concentration) from the amount of each fed (fed TMR weight fed TMR concentration). Fecal output was calculated by dividing intake of INDF by INDF concentration (24.5 h weighted mean) in feces. Since intake is based on a 24 h period and fecal output is based on 24.5 h period it must be assumed that the 30 min difference will not significantly affect the results. Apparent

207 191 digestibilities for all parameters were calculated by the following formula: (intake (24.5 h weighted mean concentration in feces fecal output)) intake. Statistical Analyses Statistical analysis was conducted using PROC MIXED of SAS (2006). Dependent variables were analyzed as a 4 4 Latin square design. All denominator degrees of freedom for F-tests were calculated according to Kenward and Roger (1997) and repeated measurements for rumen and fecal samples were analyzed using the first order autoregressive covariance structure (Littell et al., 1998), as well as terms for time and interaction of treatment by time. Because of unequally spaced rumen and fecal sampling, weighted mean daily concentrations and proportions were determined by calculating the area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). Data were analyzed for orthogonal contrasts using the fed TMR X gm that was corrected for unequal spacing according to Robson (1959). All data is presented as least squares means; treatment effects are considered significant when P 0.05 and a trend when P Results and Discussion Chemical Composition and Particle Size Distribution Chemical composition, particle size distribution, and X gm of forages included in the rations are shown in Table C-1. The M hay had lower ADF and NDF and higher NFC values than other hay lengths; this was probably due to individual bale variation. Although all bales were from the same field and cutting, each length of hay was composed of different bales. These

208 192 differences however did not affect TMR chemical composition because the 0.86% expected change in TMR NDF concentration was probably masked by sampling and lab error (Table C-2). Particle size was determined with the ASABE forage particle separator because particle length of some diets was so great that the Penn State Particle Separator (PSPS) did not adequately separate samples and small particles were improperly retained on the top screen. The PSPS particle fractions and their approximate equivalent ASABE separator screens are: top ( mm), middle (8.98 mm), lower ( mm), and pan (pan). Particle size distribution of the fed TMR varied greatly among treatments, but chemical compositions were similar (Table C-2). More detailed information regarding forages and diets was reported by Maulfair et al. (2010). Rumen Particle Size There were no differences in study conclusions between analysis of particle fractions as percentage of retained or total DM, therefore discussion and graphs of rumen digesta will be based on total DM. Particles that passed through the 3.35-mm screen were affected by time after feeding but not by ration; these particle fractions were 1.18, 0.6, 0.15 mm, and soluble, while particles retained on 9.5-, 6.7-, and 3.35-mm screens were affected by both time and ration. This finding is similar to Kononoff and Heinrichs (2003a) where rumen digesta particles retained on and 6.7-mm sieves increased with increasing ration particle size while particles retained on 0.6- and 0.15-mm sieves were not affected by ration particle size. However, Kononoff and Heinrichs (2003a) determined that particles retained on the 1.18-mm sieve decreased with increasing ration particle size in contrast to the present study. This contradiction may be caused by differences in forages used in these studies, the present study used corn silage, alfalfa haylage, and dry grass hay, while Kononoff and Heinrichs (2003a) used only corn silage. Evans et al. (1973) also determined that coarse particles retained on the largest screen (2.4 mm) responded to

209 193 effects of time and feeding and smaller particles had less response. These particle fractions, and additionally the soluble DM to retained DM ratio, are shown in Figure C-1 and expressed as the mean of the 4 treatments. Particles in the soluble fraction and the soluble DM to retained DM ratio markedly increased after feeding, remained elevated, and began to decrease slowly at 11.5 h after feeding, eventually returning to pre-prandial levels just prior to the next feeding. Particles retained on the 0.15-mm screen exhibited the opposite effect, decreased after feeding, remained lowered, and began to slowly increase at about 11.5 h after feeding to pre-prandial levels. Particles retained on 1.18-and 0.6-mm screens had less substantial changes compared to the other particle fractions. These fractions followed a similar pattern as the 0.15-mm fraction, as they decreased after feeding and began to slowly increase at about 11.5 h after feeding. Figure C-2 shows that the 6.7-mm particle fraction was least abundant for all rations. The most abundant fraction for S was 3.35 mm, L and XL was 9.5 mm, and M alternated between 3.35 and 9.5 mm. The M diet started with 3.35 mm being most abundant; by 8.5 h after feeding 9.5 mm became most abundant; finally at 24.5 h after feeding 3.35 mm was again the most abundant particle fraction. There were ration by time interactions as the 9.5-mm fraction increased after feeding in S and L diets while it decreased in M and XL diets. The 3.35-mm particle fraction increased in XL diets, decreased in S and L diets, and maintained its level in M diets. It seems that the 9.5-mm and 3.35-mm particle fractions acted inversely of each other after feeding. The 6.7-mm particle fraction did not have substantial changes over time after feeding. Fecal Particle Size and Composition The weighted means for fecal concentrations of NDF, INDF, starch, ash, and DM are shown in Table C-3. The weighted mean represents the average value over the course of the d even though sampling time points were not equally spaced. There was a significant linear contrast

210 194 for fecal NDF concentrations to decrease (from 50.7 to 47.2%) with increasing TMR particle size (from S to XL) even though NDF intake was not different across treatments (Maulfair et al., 2010). Fecal INDF concentration also followed this tendency, decreasing from 30.0 to 27.4% with increasing TMR particle size. There were no differences in weighted means for starch, ash, and DM. When determining fecal particle size distribution no particles were retained on the 6.7- mm screen. Fecal particle size was expressed as X gm using 2 different procedures. The X gm Ret procedure (using only particles retained on 0.15-mm screens) did not result in any differences among rations for weighted means, and X gm of all rations averaged 1.13 mm. These values agree with results of Kononoff and Heinrichs (2003a; 2003b) that reported fecal X gm averaged 1.13 and 1.03 mm respectively and did not change based on ration particle size. These fecal particle size data are lower than those reported by Yang et al. (2001), which averaged 1.86 mm and also did not differ due to ration particle size. The X gm Tot procedure (using all particle fractions) had much lower values than X gm Ret and had a significant linear contrast for fecal X gm to decrease with increasing TMR particle size, decreasing from 0.33 to 0.31 mm for S to XL respectively. This effect was caused by the increasing proportion of the soluble DM fraction with increasing ration particle size while all other particle fractions exhibited no effect of ration (Table C-4). One possible explanation for increased soluble DM in feces is because chewing min/kg of DMI increased with TMR of larger particle size (Maulfair et al., 2010) possibly increasing saliva secretion, therefore increasing liquid passing out of the rumen and causing a greater proportion of particles < 0.15 mm to leave the rumen (Owens and Isaacson, 1977). Another possible cause of increased soluble DM in feces is increased hind gut fermentation, leading to higher numbers of bacteria which would be included the soluble fraction. The fecal particle distribution resulted in approximately 16% of particles > 3.35 mm and 37% > 1.18 mm as a proportion of DM retained on the 0.15-mm sieve. The distribution had approximately 7% of particles > 3.35 mm and 17% > 1.18 mm as a proportion of total sample DM. These results are similar to Kononoff and Heinrichs

211 195 (2003a; 2003b), who reported that 48 and 46% respectively of fecal particles were > 1.18 mm as a proportion of DM retained on a 0.15-mm sieve; however, they are much higher than those observed by Poppi et al. (1981; 1985) where < 5% of particles were > 1.18 mm as a proportion of total sample DM in mature steers fed exclusively forage. The reasons for the 3- to 4-fold increase in particles > 1.18 mm escaping the rumen are probably due to large differences in DMI and passage rate of high producing dairy cows compared to steers being fed a maintenance diet. When fecal nutrients were analyzed over time it was determined that NDF, INDF, starch, ash, and DM concentrations were all affected by time after feeding (Figure C-3). In all rations both NDF and INDF concentrations increased after feeding to a peak at about 11.5 h after feeding, and then decreased to pre-prandial levels. Fecal starch concentrations however exhibited the opposite tendency with starch concentrations decreased in all rations after feeding to a low at about 11.5 h and then increased to pre-prandial levels. In all rations fecal ash concentrations followed a pattern over time similar to NDF and INDF concentrations, and fecal DM concentrations followed a pattern over time similar to starch concentrations. Neither X gm Ret nor X gm Tot (Figure C-4) was affected by time after feeding; however several individual particle size fractions did change significantly over time. The fractions that were affected by time after feeding were 0.6 and 0.15 mm using the retained procedure and 0.15 mm and soluble using the total procedure (data not shown). Figure C-4 also shows that generally X gm Tot decreased with increasing TMR particle size. Intakes, Fecal Output, and Digestibility Dry matter intakes ranged from 23.6 to 27.1 kg/d and were not affected by treatment (Table C-5). This effect was also present for INDF intake and fecal output. Dry matter digestibility averaged 61.6% and decreased linearly (P = 0.08) as ration particle size increased.

212 196 This effect was also seen by Kononoff and Heinrichs (2003b), where DMD decreased from 66.5 to 63.1% with increasing ration particle size. However this effect is in contrast to Kononoff and Heinrichs (2003a) and Yang and Beauchemin (2005), where DMD increased with increasing ration particle size. Digestibility of NDF and starch averaged 45.6 and 94.8% respectively, and neither was different among rations in this current study. There are many conflicting results comparing changes in DMD with NDF digestibility (NDFD) and starch digestibility (StarchD) when ration particle size is increased. Some studies reported no differences in DMD, NDFD, or StarchD (Yang and Beauchemin, 2006; 2007) while another study reported no differences in DMD and NDFD but StarchD decreased (Krause et al., 2002) with increasing ration particle size. In addition, Yang and Beauchemin (2005) reported an increase in DMD and NDFD with no change in StarchD, but Kononoff and Heinrichs (2003a) did not see a change in NDFD with an increase in DMD (StarchD was not determined in this study) when ration particle size was increased. These differing results are likely caused by interactions between forage type, forage to concentrate ratio, and starch fermentability with forage particle size. None of the experiments with steam-rolled barley grain as the main starch source had any effect of ration particle size on StarchD when fed with multiple forage types (alfalfa, barley, and corn silage) (Yang and Beauchemin, 2005; 2006; 2007). Only one of these studies using corn grain as the main starch source measured StarchD and it was determined that StarchD decreased with increasing ration particle size when feeding high moisture shelled corn and dry cracked shell corn with alfalfa silage (Krause et al., 2002). Therefore it seems that barley grain digestibility is independent of forage particle size while corn grain digestibility may not be. Forage source did not have consistent results for NDFD with differing ration particle size. Studies feeding an alfalfa silage based ration had both no effect of ration particle size on NDFD (Krause et al., 2002; Yang and Beauchemin, 2007) and a decrease in NDFD with increasing ration particle size (Kononoff and Heinrichs, 2003b). Corn silage based rations were also inconsistent with one study having an

213 197 increase in NDFD with increasing ration particle size (Yang and Beauchemin, 2005) and one study had no effect of ration particle size on NDFD (Kononoff and Heinrichs, 2003a). There are probably many factors that are influencing these differences in NDFD within each forage source. Conclusions In this experiment, 4 diets that varied in particle size were fed to lactating dairy cows. It was determined that rumen digesta particle size increased with increasing ration particle size for sieves 3.35 mm and remained the same for sieves < 3.35 mm. Fecal particle size was not different among rations and averaged 1.13 mm with more than 36% of particles being retained on an 1.18-mm sieve or larger. Therefore it can be concluded that the critical size threshold for increased resistance to rumen escape is larger than 1.18 mm in modern high producing dairy cows. In addition, this critical size is constant throughout the d as fecal particle size fractions > 1.18 mm were not affected by time after feeding. This study also concludes that for the range of TMR particle sizes fed, which was achieved using various lengths of dry grass hay, dry matter digestibility tends to decrease with increasing ration particle size. Acknowledgments This research was supported in part by agricultural research funds administered by The Pennsylvania Department of Agriculture. References American Society of Agricultural and Biological Engineers Method of determining and expressing particle size of chopped forage materials by screening. ANSI/ASAE. S424.1:

214 198 Association of Official Analytical Chemists Official Methods of Analysis. 15th ed. AOAC, Arlington, VA. Evans, E. W., G. R. Pearce, J. Burnett, and S. L. Pillinger Changes in some physical characteristics of the digesta in the reticulo-rumen of cows fed once daily. Br. J. Nutr. 29: Kenward, M. G., and J. H. Roger Small sample inference for fixed effects from restricted maximum likelihood. Biometrics. 53: Knudsen, K. E. B Carbohydrate and lignin contents of plant materials used in animal feeding. Anim. Feed Sci. Technol. 67: Kononoff, P. J., and A. J. Heinrichs. 2003a. The effect of corn silage particle size and cottonseed hulls on cows in early lactation. J. Dairy Sci. 86: Kononoff, P. J., and A. J. Heinrichs. 2003b. The effect of reducing alfalfa haylage particle size on cows in early lactation. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and D. R. Buckmaster. 2003a. Modification of the Penn State forage and total mixed ration particle separator and the effects of moisture content on its measurements. J. Dairy Sci. 86: Kononoff, P. J., A. J. Heinrichs, and H. A. Lehman. 2003b. The effect of corn silage particle size on eating behavior, chewing activities, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 86: Krause, K. M., D. K. Combs, and K. A. Beauchemin Effects of forage particle size and grain fermentability in midlactation cows. I. Milk production and diet digestibility. J. Dairy Sci. 85: Littell, R. C., P. R. Henry, and C. B. Ammerman Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci. 76: Maulfair, D. D., and A. J. Heinrichs Technical note: Evaluation of procedures for analyzing ration sorting and rumen digesta particle size in dairy cows. J. Dairy Sci. 93: Maulfair, D. D., G. I. Zanton, M. Fustini, and A. J. Heinrichs Effect of feed sorting on chewing behavior, production, and rumen fermentation in lactating dairy cows. J. Dairy Sci. 93: Mertens, D. R Creating a system for meeting the fiber requirements of dairy cows. J. Dairy Sci. 80: National Research Council Nutrient Requirements of Dairy Cattle. 7th rev. ed. Natl. Acad. Sci., Washington, DC.

215 199 Oshita, T., K. Nonaka, S. Kume, and T. Nakui Effects of forage type on particle size distribution of ruminal digesta and faeces of non-lactating cows fed high quality forage. Livest. Prod. Sci. 91: Owens, F. N., and H. R. Isaacson Ruminal microbial yields: Factors influencing synthesis and bypass. Fed. Proc. 36: Poppi, D. P., R. E. Hendricksen, and D. J. Minson The relative resistance to escape of leaf and stem particles from the rumen of cattle and sheep. Journal of Agricultural Science, UK. 105:9 14. Poppi, D. P., D. J. Minson, and J. H. Ternouth Studies of cattle and sheep eating leaf and stem fractions of grasses. 3. The retention time in the rumen of large feed particles. Aust. J. Agric. Res. 32: Poppi, D. P., B. W. Norton, D. J. Minson, and R. E. Hendticksen The validity of the critical size theory for particles leaving the rumen. J. Agric. Sci. (Camb.). 94: Robson, D. S A simple method for constructing orthogonal polynomials when the independent variable is unequally spaced. Biometrics. 15: SAS InstituteSAS User's Guide: Statistics. Version SAS Inst. Inc., Cary, NC, Shipley, R. A., and R. E. Clark Tracer Methods for In Vivo Kinetics. Academic Press, New York, NY. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: Yang, W. Z., and K. A. Beauchemin Effects of physically effective fiber on digestion and milk production by dairy cows fed diets based on corn silage. J. Dairy Sci. 88: Yang, W. Z., and K. A. Beauchemin Increasing the physically effective fiber content of dairy cow diets may lower efficiency of feed use. J. Dairy Sci. 89: Yang, W. Z., and K. A. Beauchemin Altering physically effective fiber intake through forage proportion and particle length: Digestion and milk production. J. Dairy Sci. 90: Yang, W. Z., K. A. Beauchemin, and L. M. Rode Barley processing, forage:concentrate, and forage length effects on chewing and digesta passage in lactating cows. J. Dairy Sci. 84:

216 200 Table C-1. Chemical composition and particle size distributions determined with the ASABE particle separator for corn silage, alfalfa haylage, and short (S), medium (M), long (L), or extra long (XL) grass hay Corn Alfalfa Grass hay Item silage haylage S M L XL SEM P-value Particle size, as-fed % retained mm d 34.1 c 60.4 b 77.6 a 4.10 < mm a 12.9 a 11.5 a 6.8 b mm a 15.7 a 10.4 b 5.3 c 1.38 < mm a 9.6 b 6.2 c 3.7 d 0.65 < mm a 12.7 b 6.5 c 4.2 c 0.82 < 0.01 Pan a 15.0 b 4.9 c 2.4 c 1.61 < 0.01 X 2 gm, mm c 14.6 c 38.0 b 65.4 a 3.67 < 0.01 S 3 gm, mm c 4.9 a 4.2 b 3.4 c 0.18 < 0.01 Composition, % of DM DM a 89.8 ab 90.1 ab 89.4 b CP ADF NDF pendf pendf Ash NFC NE L, Mcal/kg a d Means within a row with different superscripts differ (P 0.05). 1 Approximate equivalency to Penn State Particle Separator (PSPS): top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 X gm = geometric mean particle length determined by ASABE (2007). 3 S gm = particle length standard deviation determined by ASABE (2007). 4 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of PSPS) (Kononoff et al., 2003a). 5 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of PSPS) (Kononoff et al., 2003a). 6 NE L = Net energy of lactation, as described by NRC (2001).

217 201 Table C-2. Chemical composition and particle size distributions determined with the ASABE particle separator for TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay Item S M L XL SEM Linear Quadratic Particle size, as-fed % retained mm < mm < mm < mm < mm < Pan X 2 gm, mm < S 3 gm, mm < Composition, % of DM DM, % CP ADF NDF Forage NDF pendf pendf Ash Starch NE L, Mcal/kg Approximate equivalency to Penn State Particle Separator (PSPS): top sieve ( mm), middle sieve (8.98 mm), lower sieve ( mm), and pan (pan). 2 X gm = geometric mean particle length determined by ASABE (2007). 3 S gm = particle length standard deviation determined by ASABE (2007). 4 Physically effective NDF 8.0 = % of particles > 8.98 mm NDF of whole sample (similar to top 2 sieves of PSPS) (Kononoff et al., 2003a). 5 Physically effective NDF 1.18 = % of particles > 1.65 mm NDF of whole sample (similar to top 3 sieves of PSPS) (Kononoff et al., 2003a). 6 NE L = Net energy of lactation, as described by NRC (2001).

218 Table C-3. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on daily weighted means of fecal NDF, indigestible NDF (INDF), starch, ash, DM and X gm Item, % of DM 1 S M L XL SEM Linear Quadratic NDF INDF Starch Ash DM, % X gm Ret S gm Ret X gm Tot S gm Tot Weighted means determined by calculating area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972). 2 X gm Ret = geometric mean particle length determined by ASABE (2007) using data from screens 0.15 mm. 3 S gm Ret = particle length standard deviation determined by ASABE (2007) using data from screens 0.15 mm. 4 X gm Tot = geometric mean particle length determined by ASABE (2007) using data from all particle fractions and assuming a mean particle length of mm for particles passing through bottom screen. 5 S gm Tot = particle length standard deviation determined by ASABE (2007) using data from all particle fractions and assuming a mean particle length of mm for particles passing through bottom screen.

219 203 Table C-4. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on daily weighted mean fecal particle size distribution. Screen, mm 1 S M L XL SE Linear Quadratic Retained, % of DM Total, % of DM Soluble Weighted means determined by calculating area under the response curve according to the trapezoidal rule (Shipley and Clark, 1972).

220 204 Table C-5. Effect of feeding TMR containing short (S), medium (M), long (L), or extra long (XL) grass hay on DMI, indigestible NDF intake (INDFI), fecal output and apparent digestibilities of DM, NDF, and starch Item S M L XL SEM Linear Quadratic DMI, kg INDFI, kg Feces, kg DMD 1, % NDFD 2, % StarchD 3, % DMD = DM digestibility. 2 NDFD = NDF digestibility. 3 StarchD = starch digestibility.

221 205 Figure C-1. Mean rumen digesta particles of all treatments retained on 1.18-, 0.6-, 0.15-mm screens, soluble fraction, and soluble DM to retained DM ratio throughout the d.

222 206 A B

223 207 C D Figure C-2. Effect of feeding Short (A), Medium (B), Long (C), and Extra Long (D) TMR on rumen digesta particles retained on 9.5-, 6.7-, and 3.35-mm screens throughout the d.

224 208 A B

225 209 C Figure C-3. Effect of feeding TMR of increasing particle size on fecal NDF (A), indigestible NDF (B), and starch (C) concentration throughout the d.

226 210 Figure C-4. Effect of feeding TMR of increasing particle size on fecal geometric mean particle length (calculated using data from all particle fractions) throughout the d.