What Do Primates Hear? A Meta-analysis of All Known Nonhuman Primate Behavioral Audiograms

Transcription

1 Int J Primatol (2009) 30:55 91 DOI /s What Do Primates Hear? A Meta-analysis of All Known Nonhuman Primate Behavioral Audiograms Mark N. Coleman Received: 1 May 2008 / Accepted: 29 September 2008 / Published online: 31 January 2009 # Springer Science + Business Media, LLC 2009 Abstract Research on the hearing abilities of nonhuman primates dates back >70 yr and there are audiograms graphs showing hearing sensitivity over a range of frequencies for 29 different species including representatives from almost every major group. However, the methods used to obtain the audiograms have been nearly as varied as the number of species tested. I sought to determine the degree to which one can directly compare the audiograms by examining several factors that could have a significant impact on the results: the behavioral conditioning procedure employed to train and test the subjects, the type of transducer used to deliver the test tones, the procedure used to calibrate the amplitude of the test tones, the acoustic enclosure used to minimize ambient noise, and the method used to determine the final threshold values. Audiograms produced using speakers cannot be compared directly with those produced using headphones, and in some cases the calibration procedure and testing chamber may also limit the potential for interspecific comparisons. Based on the findings, I provide 2 lists of optimal primate audiograms: 1 for speaker-derived audiograms and the other for headphone-derived audiograms. I measured a set of audiometric variables on each of the optimal audiograms, and phylogenetic comparisons of the data show that superfamilies of primates display unique patterns of hearing sensitivity, particularly at frequencies in the lower range. Lastly, I discuss the implications for behavioralists investigating primate vocalizations in the field. Keywords audiograms. high-frequency sensitivity. low-frequency sensitivity. primate hearing. primate vocalizations M. N. Coleman (*) Department of Anatomy, Midwestern University, Glendale, AZ 85308, USA mcolem@midwestern.edu

2 56 M.N. Coleman Introduction Scientific investigation on hearing performance in nonhuman primates dates back >70 yr to the works of Elder (1934) and Wendt (1934). They used behavioral conditioning to determine absolute auditory thresholds (lowest detectable level) in chimpanzees and several species of monkeys, respectively. Researchers generally consider auditory thresholds the most fundamental measure of hearing performance (Harris 1943; Heffner 2004; Jackson et al. 1999; Lonsbury-Martin and Martin 1981; Stebbins 1975), though they have also investigated other parameters of hearing sensitivity such as localization ability and frequency and amplitude discrimination in a handful of species. Here, hearing sensitivity refers solely to differences in absolute auditory thresholds. Researchers have now investigated most major groups of primates and produced audiograms bivariate graphs showing absolute auditory thresholds plotted against frequency for ca. 30 species (Table I). However, there have been nearly as many variations in testing procedures as there have been species investigated. Before it is possible to compare the results from the studies, it is first necessary to determine the effects of differences in testing procedures on the final threshold values. The main aspects of testing procedures that could have substantial impacts on the results include the conditioning procedure employed to train the test subjects, the psychophysical technique used to determine the threshold value, the type of transducer used to deliver test stimuli, and the calibration of the sound pressure level of the test tones. I sought to determine the degree to which one can compare audiograms produced by different laboratories and then use the findings to establish a set of optimal primate audiograms. The final step was to measure a suite of audiometric variables on the data sets that I briefly discuss and offer for future analyses. Methods Testing Hearing Sensitivity Behaviorally determined audiograms are produced by presenting pure tones of varying intensity and frequency in an acoustically controlled environment, i.e., soundproof chamber, and training the test subjects to respond when they detect the sounds. Test stimuli are presented via either loudspeakers, headphones, or insert earphones after calibrating the sound field. Catch trials, wherein no stimulus is presented but all other aspects of the trial are the same as during a stimulus trial, are often intermixed with stimulus trials and are used to estimate the guess rate of the subject. Subjects are trained and tested via either operant or classical conditioning. In operant conditioning, behaviors in response to stimuli are shaped by either reinforcing a positive response (responding to the detection of a stimulus) with positive reinforcement, e.g., food or water, or discouraging a negative response (not responding to an unmistakably detectable stimulus) with negative reinforcement, e.g., shock. Conditioned suppression is a commonly used subtype of operant

3 What Do Primates Hear? 57 Table I All primate species for which behavioral audiograms have been determined Investigators Year Species Technique Transducer M/F Ages Elder 1934 Pan troglodytes PR Headphones 2/1 3 7 Wendt 1934 Ateles paniscus, Cercocebus torquatus, Homo PR Speaker 1/1, rest 0/1? sapiens, Macaca mulatta, Papio anubis Elder 1935 H. sapiens, Pan troglodytes PR Speaker 1/2 Children, 3 7 Harris 1943 M. mulatta, Macaca sinica NR Speaker? <2 Seiden 1957 Callithrix jacchus, H. sapiens NR Speaker 4/1 Juvenile Adult,? Clack and Herman 1963 M. mulatta NR Speaker 0/6 Adult Semenoff and Young 1964 H. sapiens, Macaca nemestrina GSR Headphones 3/0,3 Y. Adult, Adult Fujita and Eliot 1965 Macaca fascicularis, M. mulatta, Saimiri scuireus NR/PR Speaker 9,4,3 Y. Adults Behar et al H. sapiens, M. mulatta NR Speaker 3/ , Adol. Adult Farrer and Prim 1965 H. sapiens, P. troglodytes NR Headphones? 4 6,5 34 Stebbins et al M. fascicularis, M. nemestrina PR Headphones 3/0,1/0 Y. Adol. Clack 1966 H. sapiens, M. mulatta NR Headphones, Speaker 0/7,0/2? Dalton 1968 Cebus capucinus, M. mulatta CS, GSR, ECR Headphones, Speaker 5,3? Dalton et al M. mulatta CS Headphones 4 Adol. Bragg and Dreher 1969 Cebus capucinus NR Earphones 7/7 Adult Heffner et al. 1969a Galago senagalensis CS Speaker 2 Adult Heffner and Masterson 1970 Nycticebus coucang, Perodicticus potto CS Speaker 1/1, 1/1 Y. Adults, Y. Adults Gourevich 1970 M. nemestrina PR with shock Headphones?? Mitchell et al Eulemur fulvus, Eulemur macaco, Lemur catta PR with shock Speaker 1/1, 1/1, 1/1 3 Adult, 3 Adol. Mitchell 1970 L. catta NR Speaker 4/1 3 7 Mitchell et al L. catta NR Speaker 4/1 Adol. Adult Green 1971, 1975 Saimiri scuireus NR/PR Headphones 4/0 Adult Gillette et al L. catta NR Speaker 2/2 3 6 Pugh et al M. nemestrina PR Headphones 1/0? Stebbins 1973 Chlorocebus aethiops, Erythrocebus patas, Macaca arctoides, M. fascicularis, M. mulatta, M. nemestrina, Papio papio PR Headphones Male 3 7 Beecher 1974a S. scuireus PR Speaker 2/0? Beecher 1974b Aotus trivergatus PR Speaker 2/0? Pfingst et al H. sapiens, M. mulatta PR Headphones 2/1,2/ ,3 5

4 58 M.N. Coleman Table I (continued) Investigators Year Species Technique Transducer M/F Ages Pfingst et al M. mulatta PR Headphones 13? Niaussat and Molin 1978 Phaner furcifer PR, ECR, PO Speaker 0/2? Lonsbury-Martin and Martin 1981 M. mulatta PR Headphones 1-Jul Adult Hienz et al Papio cynocephalus PR Speaker 4 Adult Bennett et al M. mulatta PR Speakers 2/ Brown and Waser 1984 Cercopithecus mitis, H. sapiens PR Speaker 1/1 Juvenile, Y. Adults Brown 1986 Lophocebus albigena PR Speaker 1/1 Adult Smith et al Erythrocebus patas PR Headphones 4/0 Juvenile Owren et al Chlorocebus aethiops, Cercopithecus neglectus, H. sapiens, Macaca fuscata PR Headphones 3, 5, 5 Juvenile Kojima 1990 H. sapiens, P. troglodytes PR Headphones 0/2 42,5 7 Smith and Olszyk 1997 M. fuscata PR Earphones 4, 0 Juvenile Jackson et al H. sapiens, M. fuscata CS Speakers 3, Lasky et al M. mulatta PR Speakers 11 Adult Heffner 2004 E. fulvus N/A Speaker N/A N/A PR = positive reinforcement; NR = negative reinforcement; CS = conditioned suppression; GSR = galvanic skin response; ABR = auditory brain stem response; ECR = evoked cortical response; OE = otoacoustic emission; PO = pinna orientation. For all studies that investigated squirrel monkeys and owl monkeys, the exact species designation may not be the same as stated in the original publications. At the time of these studies, all squirrel monkeys were designated as Saimiri scuireus and all owl monkeys as Aotus trivergatus. However, since these studies, the genera have been split up into multiple species (Costello et al. 1993; Hershkovitz 1984) and consequently are referred to in the text as Saimiri sp. and Aotus sp.

5 What Do Primates Hear? 59 conditioning wherein the desired action is the cessation of an operant behavior, such as licking a water spout, in response to a stimulus to avoid receiving shock. In classical conditioning, used in only 2 primate studies, a conditioned reflex is established by pairing an existing unconditioned reflex with a neutral stimulus that initially has no detectable response. After a number of pairings, the conditioned stimulus is capable of eliciting a conditioned reflex in the absence of the unconditioned stimulus. Once researchers have trained a subject using one of the conditioning methods and certain performance criteria are met, they use standard psychophysical techniques to test the subject. The general approach is to vary the intensity of a stimulus and record the responses. Stebbins (1970) examined the influence of different psychophysical techniques method of limits, method of constant stimuli, and the tracking procedure for presenting test tones and found little difference in the results (threshold values) regardless of the technique employed. The final threshold values for each frequency tested are most commonly determined as the threshold level that can be detected 50% of the time; a few studies also factor in the guess rate of the subject. Extracting Threshold Values from Published Audiograms I extracted threshold values from audiograms obtained from the literature via the following method. I first digitally scanned audiograms and then imported them into Sigma Scan Pro 5 image measurement software. I standardized intensity level by setting the measurement distance equal to the decibel levels along the ordinate. I then drew a horizontal line along the 0 db SPL point and measured thresholds for each frequency as either positive or negative deviations from this line (Fig. 1). To evaluate the accuracy of this method, I compared extracted threshold values with published values from tables in 14 audiogram studies. Extracted threshold values were the same as published values in 74% (124/168) of the measurements. In 24% of cases (40/168) the values deviated by 1 db and in 2% of cases (4/168) the difference was 2 db. Fig. 1 Threshold values were extracted from published audiograms by measuring deviations from 0 db SPL at individual frequencies. More details are in the text.

6 60 M.N. Coleman For certain comparisons, I interpolated threshold values for intermediate frequencies that were not tested in the original study using the formula: db int ¼ db 1 þ ððfreq int Freq 1 Þ= ðfreq h Freq 1 ÞÞ*dB ð h db 1 Þ wherein db int is the intensity and Freq int is the frequency for the threshold to be interpolated; db l is the intensity of the lower frequency (Freq l ); and db h is the intensity of the higher frequency (Freq h ) with known values. I performed the calculation only when a study did not test a common frequency, e.g., 500 Hz, but tested frequencies that were very close to the desired frequency, e.g., 256 and 512 Hz. Previous Research on Procedural Differences Researchers sought to investigate the effects of conditioning procedures on primate audiograms in 2 studies. The first was that of Fujita and Elliot (1965), who tested the hearing in squirrel monkeys (Saimiri sp.) and 2 species of macaques (Macaca fascicularis and M. mulatta), using either positive or negative reinforcement (Fig. 2). The authors found fair agreement in the auditory characteristics of the species, regardless of the procedure used (Fujita and Elliot 1965, p. 143), though they did note that the mid-frequency dip was more pronounced in the data obtained via positive reinforcement. This is well illustrated in the comparison involving Macaca fascicularis (Fig. 2A), though the specific procedure does not seem to have much impact on the dip in Macaca mulatta (Fig. 2B). In fact, the audiograms of Macaca mulatta are nearly identical at most frequencies regardless of procedure, with an average difference of only 2.3 db across all frequencies. Even with the discrepancy in the midrange frequencies for Macaca fascicularis, the mean difference for all frequencies amounts to only 3.6 db. The audiograms for squirrel monkeys (Fig. 2C) also show a considerable level of agreement, with a mean difference for all frequencies of 4.7 db, though the negative reinforcement curve does not show the prominent peaks in sensitivity evident in the positive reinforcement curve. The second conditioning procedure comparison was that of Green (1971, 1975), who sought to compare the effects of aversive (negative reinforcement using shock) and appetitive (positive reinforcement using food) techniques on auditory sensitivity in squirrel monkeys (only 1 subject completed both tests). In the audiograms produced by the 2 procedures (Fig. 3), the threshold values are similar in the khz region with a mean difference of 2.1 db. However, there is some discrepancy at the highest and lowest frequencies tested and including all frequencies the mean difference between threshold values is 6.5 db. Nonetheless, Green states, in general, the results indicated that absolute threshold was invariant with the two procedures (1971: p. 86) and further that sensory data are not influenced by the type of reinforcer used to maintain behavior during testing (1975, p.261). Both comparisons suggest that the conditioning procedure only marginally affects the auditory thresholds. Sivian and White (1933) first documented differences produced by different transducer types in their classic study on human hearing. They noted that thresholds produced by loudspeakers were consistently lower throughout the entire frequency

7 What Do Primates Hear? 61 Fig. 2 Audiograms from Fujita and Elliot (1965) comparing different conditioning procedures. (A) Mean audiograms for Macaca fascicularis. (B) Mean audiograms for Macaca mulatta. (C) Mean audiograms for Saimiri sp.

8 62 M.N. Coleman Fig. 3 Audiograms for Saimiri from Green (1971, 1975) comparing negative and positive reinforcement conditioning procedures for the single subject (no. 4) that completed both tests. range than those produced using headphones. The phenomenon is often termed the missing 6 db (Munson and Wiener 1952) because the average difference between open field, via speakers, and closed field, via headphones or earphones, audiograms is usually ca. 6 db. Despite the fact that the problem has received considerable attention in the human audiology literature, researchers have scarcely investigated it in nonhuman primates. Only 2 studies have directly collected both open and closed field data from the same subjects (Clack 1966; Dalton 1968), but the results were equivocal and uninformative and therefore I do not discuss them. Jackson et al. (1999) compared the free field audiogram for Japanese macaques (Macaca fuscata) obtained in their study with those from 2 other studies that tested subjects in closed field environments (Fig. 4E). They found good agreement between the open and closed field audiograms at middle and high frequencies, but noted that their loudspeaker-derived audiogram illustrated considerably lower thresholds at frequencies <1 khz. They attributed the differences to the difficulties in calibrating headphones. Their studies demonstrate the need for more research on the effects of both transducer type and conditioning procedure on thresholds. Analysis Effects of Transducer Type and Conditioning Procedure More than one laboratory has tested the hearing of 4 different nonhuman primate species via various combinations of transducer type and conditioning procedure: Saimiri sp. (Beecher 1974a; Fujita and Elliot 1965; Green 1971); Macaca fascicularis (Fujita and Elliot 1965; Stebbins et al. 1966); M. fuscata (Jackson et al. 1999; Owren et al. 1988; Smith and Olszyk 1997); and M. mulatta (Behar et al. 1965; Bennett et al. 1983; Clack; 1966; Clack and Herman 1963; Dalton 1968; Dalton et al. 1969; Fujita and Elliot 1965; Harris 1943; Lasky et al. 1999; Lonsbury- Martin and Martin 1981; Pfingst et al. 1975, 1978; Wendt 1934). For the comparisons of Macaca mulatta, I modified or excluded several of the audiograms.

9 What Do Primates Hear? 63 Fig. 4 (A) Mean audiograms for Saimiri obtained using speakers and headphones. (B) Same audiograms coded for negative and positive reinforcement conditioning procedures. (C) Mean audiograms for Macaca fascicularis obtained using speakers and headphones. (D) Same audiograms coded for negative and positive reinforcement conditioning procedures. (E) Mean audiograms for Macaca fuscata obtained using speakers, headphones, and insert-earphones. (F) Same audiograms coded for negative and positive reinforcement conditioning procedures. (G) Mean audiograms for Macaca mulatta obtained using speakers and headphones. (H) Same audiograms coded for conditioned suppression, negative, and positive reinforcement conditioning procedures.

10 64 M.N. Coleman I adjusted the audiogram by Wendt (1934) via the ISO (International Organization for Standardization) human threshold values to calibrate the sound pressure levels instead of the values from Sivian and White (1933), originally used by Wendt. I did not consider the audiograms from Dalton (1968) and Pfingst et al. (1975) because the same laboratories produced subsequent audiograms of Macaca mulatta based on larger sample sizes. I also excluded the audiograms produced by Clack (1966) and Lasky et al. (1999) from the comparisons based on the finding that the threshold values are completely outside of the range of values for all other macaques. Several studies testing nonhuman primates also tested human subjects, generally to validate sound pressure levels (Behar et al. 1965; Brown and Waser 1984; Clack 1966; Jackson et al. 1999; Kojima 1990; Owren et al. 1988; Pfingst et al. 1975; Seiden 1957). I also examine the human threshold values but limited them to comparing transducer type because humans are never tested via negative reinforcement. I also consider 5 other human audiograms that researchers commonly use in comparisons of primate hearing: open and closed-field (Sivian and White 1933); Dadson and King (1952); free-field standard (ISO 1961); and closed-field standard (ANSI 1969). Examining first the audiograms for squirrel monkeys (Saimiri sp.), the thresholds from studies using speakers average 15.7 db lower than those using headphones (Fig. 4A). The differences are greatest at frequencies <1 khz (24.3 db), less at frequencies >8 khz (13.3 db), and least at frequencies between 1 and 8 khz (10.5 db). There is a significant difference (1-tailed, Mann-Whitney U test) at nearly every frequency: 125, 250, and 500 Hz (p=0.010); 1 and 2 khz (p=0.016); 4 khz (p=0.038); 16 khz (p=0.010); 32 khz (p=0.029). In this comparison, the audiograms obtained using headphones did not show a distinct mid-frequency dip, whereas those from speakers had a rise of 5.1 db relative to the mean for the 2 peaks in sensitivity. Figure 4B contains the same audiograms except separated by either negative reinforcement or positive reinforcement. The mean difference between the thresholds for both conditioning procedures is only 0.5 db. At frequencies <1 khz the differences averaged 1.7 db; at frequencies >8 khz the differences averaged 0.8 db; and at intermediate frequencies the differences were 0.5 db. There is no significant difference at any frequency tested. In the audiograms for crab-eating macaques (Macaca fascicularis), the thresholds from the studies that used speakers averaged 8.5 db lower than those that used headphones (Fig. 4C). The differences are greatest at frequencies >8 khz (15.5 db), intermediate at frequencies between 1 and 8 khz (7.3 db), and lowest at frequencies <1 khz (6.1 db). There are significant differences at 0.25, 4, and 8 khz (p=0.036). Opposite from the squirrel monkey comparison, the headphone audiogram showed a more pronounced mid-frequency dip (14 db) compared with the mean speaker audiogram (7.3 db). Figure 4D illustrates that the single audiogram that used negative reinforcement averaged 4.6 db lower than the positive reinforcement audiograms. The differences are greatest at midrange frequencies (6.3 db), slightly less at frequencies >8 khz (6 db), and smallest at frequencies <1 khz (2.5 db). There are significant differences at 2 and 4 khz (p=0.036), reflecting the pronounced mid-frequency dip evident in the positive reinforcement audiograms. The speaker audiogram for Japanese macaques (Macaca fuscata) averages 6.7 db lower than the headphone audiogram and 13.2 db lower than the insert-earphone

11 What Do Primates Hear? 65 audiogram (Fig. 4E). When the headphone and insert-earphone data are considered together, the results from the transducer and conditioning comparisons are identical because the one study that used speakers was also the only one to use negative reinforcement (Jackson et al. 1999). The speaker/negative reinforcement audiograms (Fig. 4F) average 9.9 db lower than the mean for the other audiograms, though the nonspeaker/positive reinforcement audiograms are actually 5.3 db lower at frequencies >8 khz. The greatest differences are at frequencies below 1 khz (24.8 db) whereas the frequencies between 1 and 8 khz show a mean difference of 6.4 db. There are significant differences at 0.25, 0.5, 2, and 4 khz (p=0.036) and at 16 and 32 khz (p=0.036) despite the relatively small difference in the means at higher frequencies (>8 khz). The mid-frequency dip is most distinct in the headphone audiogram (12.5 db). Similar to the previous comparisons, the audiograms for rhesus macaques (Macaca mulatta) derived using speakers were lower on average (7.1 db) than the audiograms produced using headphones (Fig. 4G). Frequencies <1 khz showed the largest differences (12 db), frequencies between 1 and 8 khz are less affected (4.8 db), and frequencies >8 khz even less so (4.3 db). The negative reinforcement audiograms have lower thresholds than the positive reinforcement audiograms (Fig. 4H), though the difference averages only 2.2 db. The differences in the mean thresholds are greatest at frequencies >8 khz (6 db) with less differences at frequencies between 1 and 8 khz (4.2 db) and <1 khz (1.6 db). The mid-frequency dip was ca. 6 db higher in the positive reinforcement curves. When the conditioned suppression audiogram is included with the negative reinforcement audiograms, the results remain virtually unchanged (mean difference of 2.7 db). The thresholds from the human studies (Fig. 5) using speakers average 8.6 db lower than those that used headphones. The largest differences are at frequencies >4 khz (9.8 db). Frequencies <1 khz are only slightly less divergent (9.2 db) and the frequencies between 1 and 4 khz show the least differences (7.2 db). There are Fig. 5 Mean human audiograms obtained in studies of nonhuman primates comparing those that used speakers vs. those that used headphones to present the test tones. Audiograms by Brown and Waser (1984) and Owren et al. (1988) are singled out to investigate the possibility that the threshold values were elevated due to excessive ambient noise in their testing chambers (see text).

12 66 M.N. Coleman significant differences between transducer types at 125 Hz (p=0.001), 250 Hz (p= 0.004), and 4 khz (p<0.001). Another noteworthy difference between transducer types is that the lowest threshold (best frequency) for the average of the speaker studies is at 4 khz, while that of the headphone studies is at 1 khz, though both mean audiograms show a broad region of good sensitivity between 1 and 4 khz with the thresholds separated by 5 db. Table II provides a summary of the results presented in the transducer and conditioning procedure comparisons. In all of the studies cited here, hearing thresholds are 7 16 db lower (mean 10 db) when signals were presented using speakers than in studies in which signals were presented using headphones. Though the differences are generally greater at frequencies <1 khz, frequencies >8 khz show the greatest variability in 3 cases and in 1 case frequencies between 1 and 8 khz are most variable. The low-frequency (<1 khz) results agree with the findings of Jackson et al. (1999), but go further to show that all regions of the audiogram are affected by transducer type, similar to the conclusions reached by Sivian and White (1933). Packer (1983) suggested that using headphones may cause physiological noise partially to mask frequencies <1 khz, resulting in auditory thresholds with higher sound pressure levels. However, it does not explain the elevated thresholds at midrange and high-frequencies. The differences may be the result of excluding the acoustic effects of the outer ear (and body) and the fact that most headphone calibrations are conducted with the probe microphone close to the entrance of the ear canal and not at the eardrum itself (Jackson et al. 1999; Killion 1978; Rudmose 1982; Shaw 1974). The mid-frequency dip appears to be unrelated to transducer type because the comparisons are divided as to which type produced a more pronounced loss in sensitivity at the midrange frequencies. The finding does not support the suggestion Table II Results from transducer and conditioning procedure comparisons Species Transducer (db) Variability Dip Conditioning (db) Variability Dip Homo sapiens 8.6 (S) Hi > low > n/a n/a n/a n/a mid Saimiri sp (S) Low > hi > S > H 0.5 (NR) Low > hi > PR > NR mid mid Macaca 8.5 (S) Hi > mid > H > S 4.6 (NR) Mid > hi > low PR > NR fascicularis low Macaca fuscata 9.9 (S) Low > mid > H>S, 9.9 (NR) Low > mid > PR > NR hi E hi Macaca mulatta 7.1 (S) Low > mid > S > H 2.2 (NR) Hi > mid > PR > NR hi low Mean 9.96 (S) Low Split 4.3 (NR) Low PR The second column shows the mean decrease in intensity for all frequencies between audiograms obtained with different transducer types and the letter in parentheses shows the type that produced the lower values. The fifth column shows the mean decrease in intensity for all frequencies between audiograms obtained with different conditioning procedures and the letters in parentheses show the procedure that produced the lower values. S = speakers; H = headphones; E = insert-earphones; NR = negative reinforcement; PR = positive reinforcement.

13 What Do Primates Hear? 67 by Kojima (1990) that the midrange dip, common to most monkey audiograms, is a byproduct of using a closed-field testing environment. The lack of association between the presence of a midrange dip and transducer type is strengthened further by the fact that the dip occurs in most monkey audiograms regardless of whether they were tested using speakers (Beecher 1974a, b) or headphones (Owren et al. 1988). This also highlights the unusual characteristics of the few monkey audiograms, such as those produced for Cercopithecus mitis (Brown and Waser 1984) and Lophocebus albigena (Brown 1986), that lack a distinct midrange dip. The differences related to conditioning procedure are much less marked than in the transducer comparison, though the trend is for the negative reinforcement audiograms to be slightly lower than the positive reinforcement audiograms. The one case in which there is a relatively large difference (Macaca fuscata) may be just as likely due to transducer type because the specific groupings for both comparisons are the same. Differences of this magnitude (<5 db) fall within the intraspecific range found in most studies (Table IV), suggesting that conditioning procedure has an insignificant effect on threshold values. The finding supports evidence from other studies that have found relatively good agreement between audiograms regardless of the conditioning procedure (Fujita and Elliot 1965; Green 1975; Prosen et al. 1978; Stebbins 1971). In contrast, the mid-frequency dip is always more accentuated in the negative reinforcement studies, which supports observations of previous investigators (Fujita and Elliot 1965; Green 1971, 1975). It is interesting that there is a general trend for positive reinforcement studies to use headphones and for negative reinforcement studies to use speakers. For the studies that employed positive reinforcement, 13 used headphones and 9 used speakers whereas 8 of the negative reinforcement studies used speakers vs. only 3 that used headphones. The comparisons support the hypothesis that threshold values derived using speakers are lower than those measured using either headphones or insert earphones. It also shows that there is not a universal adjustment standard, applicable to all species, that can be applied to closed-field audiograms to make them comparable to open-field audiograms. Therefore, one cannot directly compare threshold data obtained using different transducer types. In contrast, there is sparse evidence for a difference in threshold values resulting from using different conditioning procedures positive vs. negative reinforcement suggesting that it is not a limiting factor when comparing primate audiograms. Based on the findings, I separate the audiograms into speaker derived and non-speaker-derived for the remainder of the study. Effects of Testing Chamber and Calibration Technique One of the next potential problems to address is the difficulty of accurately measuring the sound pressure levels and the relationship to the perceived intensity by the test subject. Stebbins (1971, p. 161) expressed the magnitude of the problem by stating, It is always extremely difficult to evaluate the characteristics of the sound field in relation to the physical position of the subject in any experimental arrangement. Precise measurement of the stimuli in physical units appropriate to the source of energy is essential if we are to make any quantitative statements about the sensory acuity of an organism. Regrettably, this is also one of the most difficult

14 68 M.N. Coleman obstacles to overcome because there are few standardized protocols, and various researchers have employed a wide array of different approaches. Table III summarizes the information on the different types of acoustic enclosures and calibration procedures employed in each of the studies considered here. Fortunately, researchers recognized relatively early in studies of animal hearing that the testing chamber must be as free from extraneous noise as possible so that Table III Type of sound attenuating testing chamber used in each study and the method used to calibrate the absolute sound pressure levels Investigators Testing chamber Calibration procedure Elder 1934 Semi-soundproof room Audiometer Wendt 1934 Sound attenuating chamber Human standard Elder 1935 Semi-soundproof room Noncalibrated Harris 1943 Sound attenuating chamber Multiple-position average Seiden 1957 Sound attenuating chamber Multiple-position average Clack and Herman 1963 Sound attenuating booth? Semenoff and Young 1964 Soundproof refrigerator (30 db) Voltage readings Fujita and Eliot 1965 Soundproof rooms Multiple-position average Behar et al Double-walled soundproof room Hot-spots (45 db) Farrer and Prim 1965 Double-walled chamber (40 db)? Stebbins et al Double-walled soundproof room Probe tube Clack 1966 Sound attenuating booth Pinna position Dalton 1968?? Dalton et al. 1969?? Bragg and Dreher 1969 Double-walled soundproof room Noncalibrated Heffner et al Sound-treated room Head position Heffner and Masterson 1970 Double-walled chamber Head position Gourevich 1970 Double-walled chamber Probe tube Mitchell et al Double-walled chamber Pinna position with dummy animal Mitchell 1970 Double-walled chamber Pinna position with dummy animal Mitchell et al Double-walled chamber Pinna position with dummy animal Green 1971, 1975 Double-walled chamber (40 60 db) Probe tube Gillette et al Double-walled chamber Hot-spots Pugh et al Double-walled chamber Probe tube Stebbins 1973 Double-walled chamber Probe tube Beecher 1974a Double-walled chamber with cotton Head position Beecher 1974a Double-walled chamber with cotton Head position Pfingst et al Double-walled chamber Probe tube Pfingst et al Double-walled chamber Probe tube Niaussat and Molin 1978 Soundproof room? Lonsbury-Martin and Martin Double-walled chamber Probe tube 1981 Hienz et al Double-walled chamber Pinna position Bennett et al Sound attenuating booth (23 db) Pinna position with dummy animal Brown and Waser 1984 Semi-anechoic room (51 db) Head position Brown 1986 Semi-anechoic room (51 db) Head position Smith et al Double-walled chamber 6-cm 3 coupler Owren et al Single-walled chamber Probe tube Kojima 1990 Double-walled chamber 6-cm 3 coupler Smith and Olszyk 1997 Double-walled chamber Artificial ear Jackson et al Double-walled chamber Hot-spots Lasky et al Sound attenuating booth Head position Heffner 2004??

15 What Do Primates Hear? 69 low-level test signals are not masked and therefore undetectable. The general standard has become the usage of a double-walled commercially constructed sound attenuating chamber; Industrial Acoustics Corporation is by far the most widely used. Some researchers have taken additional steps to minimize unwanted sounds by adding supplementary sound absorbing materials (Beecher 1974a) and isolating the subjects from substrate vibrations (Heffner et al. 1969a; Jackson et al. 1999). However, some of the earlier studies as well as a few recent ones utilized less than ideal testing enclosures. The techniques used to calibrate the intensity levels of test stimuli have not achieved the same level of uniformity as with the usage of double-walled sound attenuating chambers (Table III). Considering only the free-field studies (speakerderived), researchers have employed numerous approaches such as averaging the intensity from numerous positions, using only the value from the most intense position (hot-spots), measuring from the approximate position of the center of the head, measuring from the approximate position of the pinna, and using models of the subjects with microphones placed at the ear position. Though the last approach might seem intuitively to be the most precise because it incorporates sound field disturbances caused by the subject itself, only 2 laboratories have used it (Bennett et al. 1983; Mitchell 1970; Mitchell et al. 1970, 1971). A potential problem with the hot-spot approach is that it assumes that the subject will sample a sound from the most intense area; hence sampling away from the area will produce threshold estimates that are too high. Even if the subjects are fixed in a single position, their ears are still free to move, causing variations in perceived intensity. The calibration techniques for studies using headphones have not been quite as varied, though there is still no universally applied method. The most prevalent approach has been to use probe tube calibration, which measures sound pressure levels with a microphone placed directly in a coupler that rests against the ear cushions. In studies of human hearing, a 6-cm 3 coupler has become the industry standard. However, the coupler approximates the volume of air enclosed within the human outer ear and may therefore not be suitable for use with subjects with substantially smaller or larger ears than is typical in humans. One monkey audiogram (Erythrocebus patas) used the 6-cm 3 coupler (Smith et al. 1987) and the possible effects related to it are discussed below (Fig. 13b). The study by Kojima (1990) onpan troglodytes also used a 6-cm 3 coupler to calibrate intensity levels. However, because the chimpanzee outer ear is larger than the human ear (Schultz 1969), the threshold values from the study may have slightly underestimated the chimpanzees true sensitivity. It is difficult to evaluate the Japanese macaque audiogram produced by Smith and Olszyk (1997) because they used a unique transducer (insert earphones) and novel calibration procedure (artificial ear). Evaluating the effects of using non-double-walled sound attenuating chambers and disparate calibration techniques is not a straightforward task. However, comparing the results from one study with those from other studies that tested the same species (holding transducer-type constant) may help identify data that appear to be unusually divergent. Beyond the general approach, 2 extensively tested primate taxa serve as standards for comparing the results from different investigators: humans, which were the first and have been the most frequently tested animal in auditory research, provide the best calibration standard available. Macaca mulatta is

16 70 M.N. Coleman the most commonly tested nonhuman primate, and is used as a benchmark for evaluating other nonhuman primate audiograms that also tested the species. To evaluate audiogram studies that did not test humans, rhesus macaques or the same species in another laboratory, I give critical consideration to the range of intraspecific variability and other potentially confounding factors. Estimating the Average Range of Intraspecific Variability Before evaluating the interlaboratory variability in audiograms, it is of interest to estimate the average range of intra-species variability that one can expect based on the results from individual laboratories (Table IV). The mean range of absolute values for intraspecific thresholds of all studies is 8.4 db, with a range for individual studies of db. In light of the finding that the study with the lowest value tested only 2 individuals (Beecher 1974a), it seems possible that lower levels of variability may be associated with only testing a few individuals. However, the study that presented data on 6 individuals (Fujita and Elliot 1965) has a mean range of 8.8 db, which is very close to the average and not near the upper limit, as would be expected if more individuals resulted in higher variability. Further, the range for studies that tested only 2 individuals was db, excluding the GSR threshold data obtained for Macaca mulatta in the study by Dalton (1968). This suggests that intersubject variability is not directly related to the number of individuals tested. When one considers the mean for all studies, the most variability occurred at frequencies >8 khz (9.2 db), frequencies between 1 and 8 khz showed the same value as the mean for all frequencies, and frequencies <1 khz showed the lowest variability (7.9 db): only marginally different from the mean for all frequencies (8.4 db). The data suggest that on average thresholds vary by ca. ±4.2 db around the mean with slightly more variability at higher frequencies and slightly less at lower frequencies. The value is not substantially higher than the average threshold criterion (6 db) that is used to establish the final threshold values or the average intersession variability (3.5 db) for individual subjects (Table V). Consequently, when comparing intraspecific differences between different laboratories, it would seem that differences <8.4 db may not be meaningful but that differences greater than this may be considered above the average intraspecific variation. Human Audiogram Reference Comparisons In total, 10 studies, 5 speaker-derived and 5 headphone-derived, tested humans together with the primates from each study (Fig. 5). The only 2 human audiograms obtained in studies of nonhuman primates that are not included are those of Wendt (1934) and Semenoff and Young (1964). I exclude the audiogram from Wendt (1934) because the intensity levels in the study were calibrated using a human audiogram (Sivian and White 1933), and therefore their human audiogram is uninformative as to the sound pressure levels of their testing environment. I also exclude the audiograms from Semenoff and Young (1964), both for Homo sapiens and Macaca nemestrina, because their results are published in terms of root-meansquare voltage readings instead of SPL measurements. I use the remaining human audiograms as a baseline to evaluate the nonhuman audiograms from the studies.

17 What Do Primates Hear? 71 Table IV Intersubject variability for all studies that published individual audiograms Species No. of individuals Mean range Low range (<1 khz) Middle range (1 8 khz) High range (>8 khz) Study Aotus sp Beecher 1974b Ateles sp Wendt 1934 Callithrix Seiden 1957 jacchus Cebus Dalton 1968 capucinus 1 Chlorocebus Owren et al aethiops Cercopithecus mitis Brown and Waser 1984 Cercopithecus Owren et al neglectus Erythrocebus Smith et al patas Galago sengalensis Heffner et al. 1969a Lemur catta Gillette et al L. catta Mitchell et al Eulmur fulvus Mitchell et al Eulemur macaco Mitchell et al Macaca fascicularis Stebbins et al M. fascicularis Fujita and Elliot 1965 M. fascicularis Fujita and Elliot 1965 Macaca Jackson et al fuscata M. fuscata Owren et al M. fuscata Smith and Olszyk 1997 Macaca mulatta Behar et al M. mulatta Bennett et al M. mulatta Clack 1966 M. mulatta Dalton 1968 M. mulatta Dalton et al M. mulatta Harris 1943 M. mulatta Pfingst et al M. mulatta Fujita and Elliot 1965 M. mulatta Fujita and Elliot 1965 Macaca Gourevich 1970 nemestrina Macaca sinica Harris 1943 Nycticebus coucang Heffner and Masterson 1970 Pan troglodytes Elder 1934 P. troglodytes Kojima 1990 Papio cynocephalus Hienz et al. 1982

18 72 M.N. Coleman Table IV (continued) Species No. of individuals Mean range Low range (<1 khz) Middle range (1 8 khz) High range (>8 khz) Study Perodicticus potto Heffner and Masterson 1970 Phaner furcifer 2 5 Niaussat and Molin 1978 Saimiri sp Beecher, 1974a Saimiri sp Fujita and Elliot 1965 Saimiri sp Fujita and Elliot 1965 Saimiri sp Fujita and Elliot 1965 Saimiri sp Green 1971 Mean All data obtained via traditional behavioral response testing methods unless otherwise noted. 1 = galvanic skin response used to determine thresholds; 2 = evoked cortical response used to determine thresholds. Two different groups investigating nonhuman primate hearing used acoustic enclosures other than the standard double-walled sound attenuating chambers, and the authors suggested that the reported threshold values might be elevated owing to excessive ambient noise. The first of the studies was by Brown and Waser (1984) on Cercopithecus mitis. Their testing chamber was simply a sound treated semianechoic room that had ambient noise levels that appear to be higher than those for the few studies that presented the information (the number in parentheses in the second column of Table III). The human data from Brown and Waser (1984) are highlighted in Fig. 5. The position of the audiogram, particularly at the lower frequencies, Table V Studies that published the acceptable range in intensity values (db) for determining thresholds or measured the variability in final threshold estimates for >1 session The number in parentheses in the second column is the number of trials used to determine the final threshold and the number in parentheses in the third column is the mean for the range in that cell. Study Threshold criterion (db) Intersession variability (db) Elder Fujita and Elliot (5) Behar et al (4.2) Clack, Dalton (1) Mitchell et al (3) Green (3) Pugh et al <3 Beecher 1974 <3 Pfingst et al Bennett et al (2) 5.2 Smith et al (8) Owren et al (5) Jackson et al (2) Lasky et al (2) Mean 6 3.5

19 What Do Primates Hear? 73 overlaps the studies that used headphones, though the study used a loudspeaker, supporting their notion that the low-frequency thresholds may be slightly higher than the actual values. Therefore, it would seem unwise to compare the absolute threshold levels of Cercopithecus mitis to those of other studies. Brown (1986) also tested Lophocebus albigena, in the same laboratory, so the reservations should also apply to the audiogram. The other study that potentially suffered from poor acoustic isolation was that of Owren et al. (1988) on Cercopithecus neglectus, Chlorocebus aethiops, and Macaca fuscata. They used a single-walled sound attenuating chamber and headphones to deliver the acoustic stimuli. I also singled out the human audiogram produced in their study in Fig. 5, and it falls in the middle of the human audiograms produced using headphones at the lower and higher ranges, but fluctuates between transducertype groupings at the middle frequencies. Thus, the thresholds in the midrange appear slightly lower than anticipated and not elevated, as would be expected if their testing chamber had higher than normal ambient noise levels. Because all other aspects of the study (probe calibration, etc.) used standard techniques, there does not appear to be substantial evidence to exclude the data from Owren et al. (1988) from comparisons with other species. Because the Macaca fuscata audiogram from Owren et al. (1988) appears normal in most respects, I compare it with the audiogram of the same species produced by Smith and Olszyk (1997). The audiogram from Smith and Olszyk (1997) used insert-earphones and a unique calibrated approach: the earphones were calibrated by inserting each one into an artificial ear canal (based on diameter and length measurements from a Macaca fuscata cadaveric head) and measuring SPLs at the tympanic membrane end of the canal. Comparing the audiograms (Fig. 6), it can be seen that the audiogram from Smith and Olszyk (1997) appears to be much less sensitive at most of the lower frequencies tested. In fact, the mean difference for all frequencies of 1 khz and lower is 13.4 db. The finding, plus the fact that the Fig. 6 Audiograms for Macaca fuscata produced by Owren et al. (1988) and Smith and Olszyk (1997). The study by Owren et al. (1988) used headphones whereas the study by Smith and Olszyk used insert earphones. These differences plus calibration procedures are likely responsible for the discrepancies in the low-frequency region.

20 74 M.N. Coleman compatibility of the calibration technique with other methods is unclear and that the researchers used a type of transducer not used by others, suggests that one should not compare the Macaca fuscata from Smith and Olszyk (1997) with other species. Jackson et al. (1999) also tested the hearing in Macaca fuscata, but they used loudspeakers to present the stimuli. Though no other researchers used speakers to test the species, Jackson et al. also assessed the hearing in 7 humans. Their human audiogram demonstrates among the lowest values for most of the frequencies tested but is still close to the other human audiograms produced using speakers. In fact, compared to the ISO human standard, the human audiogram from Jackson et al. (1999) has a difference for all frequencies of only 5 db. Because this presents no real difficulties and all other aspects of the study were normal, the audiogram appears suitable for interspecific comparisons with other speaker-derived audiograms. The final audiogram to evaluate using the human reference standard is that of Callithrix jacchus from Seiden (1957). Compared with the other speaker-derived audiograms, the human audiogram produced in their study falls toward the top of the range, though it is not unacceptably high, even though Seiden did not use a doublewalled chamber. The finding is not entirely surprising because Seiden compared his human audiogram with the American Standards Association reference curve (Licklider 1951) for validation. However, the most troubling aspect of the audiogram for Callithrix jacchus relates to the high-frequency thresholds. Seiden s marmosets show the least highfrequency sensitivity of any monkey tested to date (Fig. 15A) and they are also the smallest in body mass [346 g (Smith and Jungers, 1997)]. This is unusual because most small-bodied mammals generally show good high-frequency hearing most likely related to the difficulty in localizing all but high-frequency sounds for individuals with relatively small heads (Heffner et al. 2001; Masterson et al. 1969). Though Seiden went to great lengths to try to elicit a response from the subjects at higher frequencies, the frequency range is beyond that which was validated using the human subjects. Accordingly, one should probably consider with caution the highfrequency thresholds illustrated by the audiogram until an addition audiogram is produced for the species. Macaque Audiogram Reference Comparisons Three laboratories have produced Macaca mulatta audiograms using headphones (Fig. 7A). The audiograms produced by Pfingst et al.(1978) and Lonsbury-Martin and Martin (1981) are extremely similar, but the audiogram from Dalton et al. (1969) illustrates much lower threshold values and is most sensitive at the mid-frequency dip evident in the other 2 audiograms. When one considers all 3 audiograms together, the mean range for all frequencies is 11.7 db, but the value drops to 2.4 db when the results of Dalton et al. (1969) are excluded from the calculation. Therefore, the best estimate of a headphone-derived audiogram for Macaca mulatta is based on the mean thresholds from Pfingst et al. (1978) and Lonsbury-Martin and Martin (1981). In total, 8 different laboratories have produced speaker-derived audiograms for Macaca mulatta (Fig. 7B) and there is considerable variability among the thresholds and contours, with a mean variability for all frequencies of 25.8 db. However, several audiograms present reasons to be suspicious of the results. The audiogram of

21 What Do Primates Hear? 75 Fig. 7 (A) Audiograms for Macaca mulatta obtained using headphones from Dalton et al. (1969), Pfingst et al.(1978), and Lonsbury-Martin and Martin (1981). (B) Mean audiograms for Macaca mulatta obtained using speakers from Wendt (1934), Harris (1943), Clack and Herman (1963), Behar et al. (1965), Fujita and Elliot (1965), Clack (1966), Bennett et al. (1983), and Lasky et al. (1999). Lasky et al. (1999) falls well above all other rhesus macaque audiograms and is obviously uncharacteristic for the species. The results from both Clack and Herman (1963) and Clack (1966) may also be dubious because in the former case there was no mention of calibrating the test signals and in the latter the values were significantly below the humans in their study, which is an atypical finding when comparing macaques to humans. Stebbins (1971) suggested that Harris s (1943) data are anomalously low at 8 khz (and seemingly at 1 and 2 khz also), as can be seen in Fig. 7B. In addition, Harris apparently did not use a double-walled sound chamber. Removing the audiograms lowers the mean difference slightly to db, still a high level of variability. Because the audiogram of Macaca mulatta from Harris appears divergent at the aforementioned frequencies, it also seems practical to exclude Macaca sinica, the other species tested in the study, from interspecific comparisons. The next audiogram of Macaca mulatta to consider is that produced by Wendt (1934), which is shown in Fig. 8 together with the other species tested in that study.

22 76 M.N. Coleman Wendt determined the threshold values by calibrating the relative thresholds produced by human subjects with the values from Sivian and White (1933). The macaque audiogram shows a peak at 8 khz that is below the lowest audible frequency of the humans and it can also be seen that the audiograms for Papio anubis and Ateles paniscus illustrate an exaggerated mid-frequency notch. That both of the qualities are unlike those demonstrated by any other primate audiogram suggests that the audiograms are not suited for interspecific comparisons. Nevertheless, even though the absolute thresholds are doubtful, the relative differences between the species tested by Wendt may prove informative for certain comparisons. This leaves the audiograms for Macaca mulatta from Behar et al. (1965), Fujita and Elliot 1965, and Bennett et al. (1983), which have a mean range for all frequencies of 11.6 db (Fig. 9). Though the audiograms are in close agreement in the lower and middle frequencies, the audiogram from Fujita and Elliot (1965) shows much lower thresholds at 8 khz and above (mean difference 19.8 db). Considering only the frequencies from 63 Hz to 4 khz from all 3 studies, the mean variability is 8 db. The unusually low thresholds at the higher frequencies in the study of Fujita and Elliot (1965) does not seem to be related to the soundproof rooms they used to attenuate external sounds because excess ambient noise would be expected to produce elevated thresholds, not lower ones. Accordingly, the mean speaker audiogram of Macaca mulatta (Fig. 9) is based on all 3 studies at the low and middle frequencies, but only on the thresholds from Behar et al. (1965) and Bennett et al. (1983) at frequencies >8 khz. One of the other taxa tested by Fujita and Elliot was Saimiri. They tested 9 squirrel monkeys, which were the most extensively scrutinized of the 3 species investigated (3 different conditioning procedures were explored). The mean audiogram for Saimiri from the study is compared with that from Beecher (1974b) in Fig. 10 (the testing procedures for Beecher s squirrel monkeys are described below in the section on owl monkeys). The mean difference between the audiograms is 10.2 db, but this is largely driven by the difference at 16 khz (19.3 db). Both Fig. 8 Audiograms from Wendt (1934) for humans (Homo sapiens), macaques (Macaca mulatta), spider monkeys (Ateles paniscus), baboons (Papio anubis), and mangabeys (Cercocebus torquatus).

23 What Do Primates Hear? 77 Fig. 9 Speaker-derived audiograms for Macaca mulatta from Behar et al. (1965), Fujita and Elliot (1965), and Bennett et al. (1983). Speaker mean audiogram for Macaca mulatta (solid line) derived from all 3 studies at lower and middle frequencies but excluding data from Fujita and Elliot (1965) at higher frequencies. audiograms reflect a similar overall shape, but the frequency of best sensitivity is at a higher frequency in the audiogram from Fujita and Elliot (1965), which explains the large difference at 16 khz. Also note that the sound pressure level at the best frequency is similar in the 2 studies, unlike the comparison between the Macaca mulatta of Fujita and Elliot (1965) and those from Behar et al. (1965) and Bennett et al. (1983). Therefore, considering the large sample size for squirrel monkeys in Fujita and Elliot s study and attention to procedures, the best estimate of hearing in Saimiri is a composite (Fig. 10) between the audiograms from Beecher (1974a) and Fujita and Elliot (1965). The third species tested by Fujita and Elliot was Macaca fascicularis. Unlike the other 2 species, no other laboratory tested the species using speakers. However, Fig. 10 Audiograms for Saimiri obtained using speakers from Fujita and Elliot (1965) and Beecher (1974a) and mean audiogram (solid line) derived from the studies.

24 78 M.N. Coleman considering the moderately good agreement between the other audiograms from Fujita and Elliot (1965) and other researchers results, the audiogram will tentatively be considered as the best approximation of speaker-based auditory thresholds in Macaca fascicularis. The final macaque to consider is Macaca nemestrina, which is represented (Fig. 11) by the audiograms from Stebbins et al. (1966), Gourevich (1970), and Pugh et al. (1973). The thresholds at midrange frequencies are very similar but there is considerable discrepancy at higher and lower frequencies, resulting in a mean range of 14.2 db. Removing the audiogram of Pugh et al. (1973), which is based on only 1 individual of unknown age, from the variability computation results in a value of only 5.6 db. Therefore, the best estimate for a headphone audiogram for Macaca nemestrina is the mean (Fig. 11) of the threshold values presented by Stebbins et al. (1966) and Gourevich (1970). Because the audiogram for Macaca nemestrina from Stebbins et al. (1966) agrees well with that of Gourevich (1970) and the fact that standard methods and equipment were used, the other audiogram produced by Stebbins in this study for Macaca fascicularis also appears valid. Comparisons of Pan troglodytes Three laboratories have contributed threshold data for Pan troglodytes. Though the chimpanzee audiogram by Elder (1934) was produced during the early stages of nonhuman primate auditory studies (the first in fact), I include it here because it is suspected that the audiogram of Kojima (1990) may have slightly underestimated the threshold values due to the usage of a 6-cm 3 coupler for calibration. Figure 12 contains both audiograms and there is similarity in the contours, though Elder s audiogram has lower thresholds at all frequencies except 8 khz. Because the difference is in the expected direction and the fact that the mean difference is 11 db, I averaged the thresholds. Also included in the figure are the high-frequency cutoff points determined by Elder (1935) and Farrer and Prim (1965). The high-frequency Fig. 11 Audiograms for Macaca nemestrina obtained using headphones from Stebbins et al. (1966), Gourevich (1970), and Pugh et al. (1973). Mean audiogram for Macaca nemestrina (solid line) based on data from Stebbins et al. (1966) and Gourevich (1970) only.

25 What Do Primates Hear? 79 Fig. 12 Audiograms for Pan troglodytes from Elder (1934) and Kojima (1990). Mean audiogram for Pan troglodytes (solid line) based on the studies in addition to high-frequency limit data form Elder (1935) and Farrer and Prim (1965). limits determined by Elder, assuming an 80 db SPL, and Kojima are only ca. 1 khz apart though the value determined by Farrer and Prim is ca. 4 khz lower. However, because the exact SPL of the high-frequency limit from Elder is unknown, I take the high-frequency cutoff, at 80 db SPL, as the average of the 3 values. Additional Specific Comparisons Numerous hearing studies investigated a species that was not tested along with either humans or rhesus macaques, and was not tested in other laboratories: 3 platyrrhines (Aotus sp., Cebus capucinus, Saimiri sp.), 2 Old World monkeys (Erythrocebus patas, Papio cynocephalus), 4 lemuroids (Eulemur fulvus, Eulemur macaco, Lemur catta, Phaner furcifer), and 3 lorisoids (Galago senegalensis, Nycticebus coucang, and Perodicticus potto). New World Monkeys: Ceboids There appears to be no methodological reasons to reject the audiogram for Aotus produced by Beecher (1974b). He used standard psychophysical methods, transducers (speakers), and calibration procedures and employed a double-walled sound attenuating chamber. In fact, the individual audiograms for the 2 focal subjects produced the lowest variability for all frequencies of any study examined here (Table IV). The same basic arguments apply to the audiogram of Saimiri sp. produced in Beecher s laboratory (Beecher 1974a) that was combined with the data from Fujita and Elliot (1966). The study by Bragg and Dreher (1969) on capuchin monkeys (Cebus capucinus) offered the potential to provide an audiogram for one of the most common New World monkeys. However, the authors did not determine the exact sound pressure levels; thus the threshold values are not comparable with the values provided in other studies.

26 80 M.N. Coleman The study on Saimiri by Green (1971, 1975) used a double-walled attenuating chamber, probe calibration, standard psychophysical techniques, and common transducers (circumaural headphones). In addition, the main goal of Green s study was to test the difference between conditioning procedures positive vs. negative reinforcement and concluded that there is insignificant influence on the final thresholds. Therefore, this audiogram presents no obvious difficulties for interspecific comparisons with other headphone-derived audiograms. Old World Monkeys: Cercopithicoids As previously mentioned, the study on Erythrocebus patas by Smith et al. (1987) was the only study on monkeys to use the 6-cm 3 coupler to calibrate the test tones. Figure 13 shows the pretreatment specific mean audiogram for Erythrocebus patas (this study actually compared the pre- and posttreatment effect of ototoxic antibiotics on the outer hair cells of the inner ear) along with the composite cercopithecine audiogram from Stebbins (1973). Both sets of data employed similar conditioning procedures and used headphones, although the Stebbins study did not use the 6-cm 3 coupler. The audiograms are quite similar at 16 khz and above (mean range of variability from 16 to 40 khz=3 db) but the patas monkey audiogram becomes increasingly less sensitive as frequency decreases. The mean variability for all frequencies is 23.4 db, but the range for frequencies from 63 to 500 Hz averages 45.5 db and becomes as large as 52 db at 63 and 125 Hz. One possible explanation for these discrepancies is that they represent a true species distinction. However, the Stebbins audiogram also included patas monkeys and it was stated that the interspecific variability between the Old World monkeys investigated was no greater than the intraspecific variation (Stebbins, 1973). Another possibility relates to the calibration procedures that were used in the Smith et al. (1987) study. The use of a 6-cm 3 coupler can produce artificially low thresholds when applied to an outer ear with less volume than that typical of humans. However, this is exactly opposite the trend observed, so it is unlikely that the differences are Fig. 13 Audiogram for Erythrocebus patas from Smith et al. (1987) compared with cercopithecoid mean audiogram from Stebbins (1973). Note the large differences at lower and middle frequencies.

27 What Do Primates Hear? 81 attributable solely to the calibration technique. One final explanation is related to the fact that the patas monkey audiograms were produced using pulsed tones instead of steady tones as in the Stebbins (1973) study, and in all other studies producing audiograms. Regardless of the reason for the differences at the lower and middle frequencies between this audiogram and those of other Old World monkeys, it would appear this audiogram is unsuitable for further comparisons. The study by Hienz et al. (1982) examined the hearing in 4 adult yellow baboons (Papio cynocephalus). Because this study utilized standard conditioning procedures, transducers (wide-range speaker), sound enclosures, and calibration techniques there are no methodological reasons to reject the species mean audiogram. In addition, the intraspecific variation was 8.1 db, just below the average for all species. The authors commented on the fact that their study produced thresholds that averaged ca. 5 db lower than those from the studies on macaques by Stebbins et al. (1966) and Pfingst et al. (1978), which used similar methods, but this is not surprising because these studies used headphones whereas that of Hienz et al. (1982) used speakers. Malagasy Lemurs: Lemuroids Taken at face value, there appear to be few reasons to reject the audiogram for Eulemur macaca produced in the study by Mitchell et al. (1970). They employed a fairly standard conditioning procedure and psychophysical techniques and used an extended range speaker to deliver the test tones. The calibration method was somewhat novel (they used a lemur-size paper-maché model with a microphone placed in the approximate position of the ear), but this alone does not seem to warrant exclusion. Further, the mean variability for all frequencies between the 2 test subjects was 7.4 db. However, in a subsequent publication (Mitchell et al. 1971) the authors retested Lemur catta (also tested in this study) and found the thresholds in the second study were significantly lower for almost every frequency vs. those produced in the first study. They attributed the difference to the fact that all falsepositive responses were punished with shock (the standard practice is to punish false-positives with a timeout period) causing conservative responses by the test subjects and therefore elevated thresholds. Therefore, I take this finding to conclude that the audiograms from the initial study are not amenable to interspecific comparisons. The preceding argument for Eulemur macaco would also seem to apply to the audiogram for Eulemur fulvus because both were determined in the same study. Still, as a further evaluation of these data, the audiogram for Eulemur fulvus will be compared with the limited audiometric data presented in Heffner (2004) for the species. Heffner (2004) presented 3 data points for Eulemur fulvus from an unpublished study by D. Sutherland and R. B. Masterson that included the lowfrequency cutoff, the frequency of best sensitivity, and the high-frequency cutoff. The specific mean audiogram for Eulemur fulvus from Mitchell et al. (1970), as well as the 3 data points provided by Heffner are illustrated in Fig. 14. Although Mitchell et al. (1970) acknowledged that their initial study underestimated the sensitivity of their subjects at most frequencies, the highest and lowest frequencies tested appeared to be unaffected. Yet, the high- and low-frequency regions from the 2 studies are quite different. The extrapolated high-frequency cutoff (at 60 db SPL) for the

28 82 M.N. Coleman Fig. 14 Audiogram for Eulemur fulvus from Mitchell et al. (1970) compared with the values for lowfrequency cutoff, best frequency, and high-frequency cutoff provided in Heffner (2004). Mitchell et al. data is 66 khz whereas that from Heffner is 43 khz. The extrapolated low-frequency cutoff (at 60 db SPL) for the data of Mitchell et al. is 46 Hz whereas that from Heffner is 73 Hz. These results further underscore the questionable data for Eulemur macaco and Eulemur fulvus of Mitchell et al. (1970). The final audiogram presented in Gillette et al. (1973) for Lemur catta represents the end product of the previous studies by Mitchell (1970), Mitchell et al. (1970), and Mitchell et al. (1971). The only significant methodological differences between Gillette et al. (1973) and the Mitchell et al. (1971) was that the shock rate and intensity were lower (only intratrial false-positives were punished) and the method used to calibrate the sound pressure levels was changed. The new method utilized hot-spots, or areas of maximum intensity, as the estimate of the true intensity value (Gillette et al. 1973). The authors discovered 2 hot-spots of equal intensity, which were preferred sampling spots for most of the experimental subjects. One potential reservation with the audiogram relates to the fact that the intersubject variability for all frequencies was quite high at 18.5 db (Table IV). Although the variability was greatest for frequencies >8 khz (26.3 db), the variability for the midrange frequencies (14.7 db) and frequencies <1 khz (13.8 db) were also above the average for all species. However, considering the rigorous attention to methodological issues, extensive testing, and the fact that it represents one of the only estimates of hearing in lemurs, this audiogram is considered as a provisional estimate of hearing in Lemur catta. The other lemur audiogram to evaluate is that for Phaner furcifer (Niaussat and Molin 1978). Although they presented only limited information on some of the testing parameters, e.g., type of testing chamber and calibration technique, the authors appeared to adhere to rigorous conditioning standards: only 2 of the 4 subjects met the performance criterion. In addition, the 2 subjects that generated the final audiogram show threshold values that were 5 db of each other (Niaussat and Molin 1978). The validity of these results is further strengthened by the finding that the behavioral audiogram essentially paralleled an audiogram produced using an electrophysiological technique (auditory evoked potentials). Although the behavioral

29 What Do Primates Hear? 83 audiogram did show lower threshold values, this is a typical finding when comparing behavioral vs. electrophysiological audiograms (Lasky et al. 1999). However, because of the lack of information on the accuracy of the SPLs one should consider this audiogram a first approximation of hearing sensitivity for Phaner furcifer. Lorisoids Researchers have tested the hearing sensitivity of 3 species of lorisoids, and because the same group of researchers, consisting of Henry Heffner, Bruce Masterson, and Richard Ravizza, produced the audiograms via similar procedures, I consider them together. In general, there appears to be nothing about the conditioning procedure, psychophysical methods, transducer, or calibration technique that would cause concern when evaluating the audiograms. However, Ravizza et al. (1969) tested Galago senegalensis in a sound-treated room instead of a burlap-draped acoustic chamber (Heffner and Masterson 1970) used with the other 2 species. However, the fact that Galago had the lowest intersubject variability of the group (5.2 db vs. 7.3 db for Nycticebus coucang and 13.6 db for Perodicticus potto) suggests that the acoustic enclosure did not cause excessive ambient noise. The authors commented on the relatively high intersubject variability for Perodicticus but stated that it represented the true variation of hearing in Perodicticus potto because both individuals had a low false positive rate (Heffner and Masterson 1970, p. 179). In view of the low values for the other species and the attention to detail Heffner and Masterson illustrated, all 3 audiograms are valid for interspecific comparisons with other speaker-derived audiograms. Discussion The final list of speaker-derived audiograms that met the critical evaluation include those for Aotus sp., Callithrix jacchus, Galago senagalensis, Lemur catta, Macaca fascicularis, M. fuscata, M. mulatta, Nycticebus coucang, Papio cynocephalus, Perodicticus potto, Phaner furcifer, and Saimiri sp. The audiograms are shown in Fig. 15A along with the ISO free-field human standard and the audiogram for tree shrews (Tupaia glis) from Heffner et al. (1969b). The final list for the headphonederived audiograms includes Cercopithecus neglectus, Chlorocebus aethiops, Macaca fascicularis, M. fuscata, M. mulatta, M. nemestrina, Pan troglodytes, and Saimiri sp. and are shown in Fig. 15B. Several phylogenetic trends are evident from the graphs. Apes and Old World monkeys (catarrhines) have the best low-frequency sensitivity of all primates tested, followed by New World monkeys (platyrrhines), with lemuroids and lorisoids illustrating the least low-frequency sensitivity (Fig. 15A). In fact, the audiograms for Phaner furcifer and Galago senegalensis are little differentiated, at most frequencies, from tree shrews, which are all animals of relatively similar size. On the high-frequency side, apes show the least sensitivity but there is less separation among the other taxa. Evidently, galagos, lemurs, and tree shrews have the best high-frequency sensitivity but the other lorisoids (Nycticebus

30 84 M.N. Coleman Fig. 15 (A) Final speaker-derived audiograms for hominoids (Homo sapiens), cercopithecoids (Macaca fascicularis, M. fuscata, M. mulatta, Papio cynocephalus), ceboids (Aotus sp., Callithrix jacchus, Saimiri sp.), lemuroids (Lemur catta, Phaner furcifer), and lorisoids (Galago senegalensis, Nycticebus coucang, Perodicticus potto). An audiogram for tree shrews (Tupaia glis) is also included. Note the reduced highfrequency sensitivity and accentuated mid-frequency notch for Callithrix jacchus. (B) Final headphonederived audiograms for hominoids (Pan troglodytes), cercopithecoids (Cercopithecus neglectus, Chlorocebus aethiops, Macaca fascicularis, M. fuscata, M. mulatta, M. nemestrina), and ceboids (Saimiri sp.). See text for details on threshold data used to define species mean audiograms (see online version for color figures). and Perodicticus) are no more sensitive than the platyrrhines, with the possible exception of the suspect high-frequency data from Callithrix, and only moderately more sensitive than Old World monkeys. In the midrange frequencies, lorisoids illustrate a single peak in maximum sensitivity at relatively higher frequencies, similar to tree shrews. Monkeys (both Old and New World) show a double peak of maximum sensitivity separated by a midfrequency notch, although it is less accentuated in Old World monkeys, whereas humans show a maxima of sensitivity in the region of the midfrequency notch. Lemur catta is seemingly similar to most New World monkeys at the midrange frequencies with the exception of Callithrix jacchus. The exaggerated extent of the

31 What Do Primates Hear? 85 midfrequency notch in Callithrix is another unusual characteristic of this audiogram, which casts additional doubt on the validity of the threshold values. On a more limited scale, these phylogenetic patterns are also visible in the headphone-derived audiograms (Fig. 15B), although note the presence of the midfrequency notch in the audiogram for chimpanzees, which is generally absent from human audiograms regardless of transducer type. Several researchers have noted various aspects of these phylogenetic patterns. The seminal primate hearing study by Elder (1934) was the first to note the presence of the midfrequency notch, which subsequent researchers confirmed in other nonhuman anthropoids (Fujita and Elliot 1965; Harris 1943; Stebbins et al. 1966; Wendt 1934). Elder was also the first to propose that humans have less high-frequency sensitivity than chimpanzees (1935), and later studies confirmed this supposition as well (Farrer and Prim 1965; Kojima 1990). Soon researchers realized that all other catarrhines have better high-frequency sensitivity than humans do (Behar et al. 1965; Fujita and Elliot 1965; Green 1975; Stebbins et al. 1966; Wendt 1934). Once researchers began to test prosimians they discovered that in general these primates have better high-frequency hearing, less low-frequency hearing, and a higher best frequency of sensitivity than anthropoids do (Heffner et al. 1969a; Mitchellet al. 1970, 1971; Stebbins 1971, 1973). The one characteristic all of the investigators appear to have overlooked is the intermediate position of platyrrhines, falling between catarrhines and prosimians in low-frequency sensitivity. Although Beecher noted that Old World monkeys (macaques) are slightly more sensitive at lower frequencies than the New World monkeys he tested, he went on to suggest that it is fair to speak of a general monkey audibility function without specifying a particular species (Beecher 1974b:197). The true magnitude of the difference between platyrrhines and catarrhines in low-frequency sensitivity was probably obscured to Beecher because he compared his monkeys, tested using speakers, with the macaques that Stebbins et al. (1966) tested via headphones. Although it might seem that many of the differences, particularly in lowfrequency sensitivity, are simply a size-related phenomenon, 2 examples provide evidence that this is not strictly the case. The first is a narrow allometric comparison presented by Coleman and Ross (2004) that contrasted the audiograms of lorisoids with those of platyrrhines. The mean body mass for the species in the 2 groups the same ones considered here was nonsignificantly different, although there was a significant difference in low-frequency sensitivity. Coleman and Ross (2004) also found similar differences when the analysis was limited to the smallest species in each group: Galago senegalensis vs. Callithrix jacchus. The second example comes from the audiograms presented by Wendt (1934). The absolute threshold values from the study are questionable but the relative interspecific differences may still provide insightful clues. In Wendt s study, the New World monkeys (Ateles) illustrated less low-frequency sensitivity than any of the catarrhines examined (Fig. 8). Because spider monkeys are among the largest New World monkeys and are within the size range of the Old World monkeys that were tested, the distinctions in low-frequency sensitivity appear to be unrelated to body size. To quantify the differences in hearing sensitivity between various primates, I measured a suite of auditory parameters on the audiograms (Fig. 16). I defined the

32 86 M.N. Coleman Fig. 16 Audiometric parameters measured on final audiogram data sets. Complete descriptions of each measurement are in the text. high- and low-frequency cutoff points as the highest and lowest audible frequencies at 60 db SPL. For species not tested up to this threshold level, I used extrapolated values as long as the species were tested at 49 db SPL. Even with this reduced threshold criterion, it was not possible to derive cutoff points for all taxa; lowfrequency cutoff was still only available for fewer than half of the species. Therefore, I derived 2 additional measures of high- and low-frequency sensitivity that could be taken on all taxa: threshold (db) at 32 khz and threshold at 250 Hz. I assessed the region of best sensitivity by taking the frequency and intensity of the lowest point (frequency with the lowest threshold) on the audiogram. For primate audiograms that take the form of a W shape (2 peaks), I estimated the region of best sensitivity by taking the frequency and intensity of both peaks (low and high) as well as the midfrequency dip surrounding the peaks. I estimated the audible area by measuring the area bounded by the audiogram below 50 db SPL, via Sigma Scan 5.0 image measurement software. I scaled each audiogram to the same ordinate and abscissa ranges and standardized each audiogram using the intensity values along the ordinate axis. I subdivided the total audible area into high (>8 khz), middle (1 8 khz), and low (<1 khz) regions. I chose the divisions based on the observation that primate audiograms generally start to roll off (decrease in sensitivity) at ca. 1 khz on the low-frequency side and at ca. 8 khz on the high-frequency side. One advantage to using areal subdivisions such as these over more traditional measures of regional sensitivity is that they incorporate data from multiple threshold values and not just a single arbitrarily defined point. I used extrapolated values for species tested to 10 db of the upper bound (i.e., db SPL). I measured the total range of sensitivity as the number of octaves between the highest and lowest audible frequency at 50 db SPL. The values derived from these measurements are in Table VI. Although I present the quantitative data primarily for future analyses, a few aspects of the numerical values merit comment. The first relates to the fact that many behavioral primatologists often make the implicit assumption that their subjects generally hear similarly to humans, particularly in terms of assessing the audibility of vocalizations. However, the data show that commonly not to be the case.