Differential outcomes facilitate same/different concept learning

Transcription

1 Anim Cogn (2010) 13: DOI /s SHORT COMMUNICATION Differential outcomes facilitate same/different concept learning Kelly A. Schmidtke Jeffrey S. Katz Anthony A. Wright Received: 24 September 2009 / Accepted: 29 October 2009 / Published online: 24 November 2009 Ó Springer-Verlag 2009 Abstract Commonly recognized, the training procedure one employs often affects the results they obtain. Here, we demonstrate for the first time that abstract-concept learning is affected by employing a differential-outcomes procedure. The differential-outcome effect has been shown to occur for item-specific strategies but has not been established for relational strategies. To test whether differentoutcome expectancies can facilitate a relational strategy, eight pigeons were trained and tested in a two-item same/ different task with pictures. After pecking an upper picture, they pecked a lower picture if the pictures were the same or a white rectangle if the pictures were different. Two groups of pigeons were rewarded with either different outcomes (sounds and food amounts) or same outcomes. Both groups were trained to criterion with successively larger picture sets (8 1,024 items) and were transfer tested with novel pictures following each acquisition. With the smallest training sets, neither group showed any novel-stimulus transfer. But after acquiring the task with 32 pictures, the different-outcomes group responded more accurately to novel pictures than the same-outcome group. As the training set-size increased, both groups transfer performance converged and became equivalent to training performance. These results show for the first time that training K. A. Schmidtke (&) J. S. Katz Department of Psychology, Auburn University, 226 Thach Hall, Auburn, AL 36849, USA schmike@auburn.edu A. A. Wright Medical School, Department of Neurobiology and Anatomy, University of Texas Health Science Center at Houston, Houston, TX, USA with different outcomes facilitates abstract-concept learning. Keywords Different outcomes Concept learning Same/different Pigeons Introduction In most training procedures, all correct conditional discrimination choices are followed by the same reinforcement, a same-outcomes procedure. Alternatively, the subject may receive training where each correct comparison choice is correlated with a unique outcome, a differentoutcomes procedure. The differential-outcome effect (DOE) describes the improved acquisition rate, retention, and/or terminal accuracy that occur(s) when each sample stimulus and its comparison stimulus are correlated with a unique outcome (DeMarse and Urcuioli 2005; Edwards et al. 1982; Grant and Kelly 2001; Kruse et al. 1983; Maki et al. 1995; Urcuioli 2005). In one of the earliest papers describing the DOE, rats learned two conditional discriminations that involved hearing either a clicker or tone sample and then making a choice response to either a right or left lever comparison (Trapold 1970). Rats in the same-outcome group received consistently similar outcomes for both correct choice responses, either always pellets or always sucrose. Rats in the different-outcomes group received consistently different outcomes for each correct choice response. Trapold found that rats in the different-outcomes group acquired the discriminations faster and maintained higher terminal accuracy than rats in the same-outcome group. He suggested that rats that experienced consistently different outcomes develop unique expectancies linked to each

2 584 Anim Cogn (2010) 13: sample which provided them with an additional prospective, expectancy cue to guide their choice response (for an expanded discussion on how the DOE might relate to stimulus class formation see, Sidman 2000). Since Trapold s (1970) experiment, the DOE has been obtained with different species, outcome types, and methodologies. The different nonhuman species include: rats (Carlson and Wielkiewicz 1976; Trapold 1970), pigeons (Brodigan and Peterson 1976; Urcuioli and Lionello- DeNolf 2001), dogs (Overmier et al. 1971), and horses (Miyashita et al. 2000). The DOE has also been demonstrated in various human populations, such as languagedeficient children (Hewitt 1965; Jansen and Guess 1978), typical children (Estévaz et al. 2001), adults with alcoholinduced amnesia (Hochhalter et al. 2000), and college students (Miller et al. 2002). The DOE has been obtained with a variety of outcome types: reinforcers with different qualitative natures (food vs. water, Trapold 1970), different probabilities of one reinforcer (0.2 vs. 1.0, Urcuioli 1990a, b), different quantities of one reinforcer (one vs. five food pellets, Carlson and Wielkiewicz 1976), and conditional reinforcers (hopperlight vs. food, Urcuioli and Lionello- DeNolf 2001; token colors, Estévez et al. 2003). The DOE has been demonstrated with both between-subjects designs (Brodigan and Peterson 1976; Trapold 1970) and withinsubject designs (Alling et al. 1991). All these findings suggest that the DOE is a general process, common across a wide variety of species within a variety of procedures. However, while many different procedures have produced the DOE, all the procedures used have encouraged subjects to use an item-specific strategy. When subjects use an item-specific strategy, their choice responses are bound to the absolute properties of the training items. For example, subjects may learn a matching-to-sample discrimination with red and green items. In this task, if the sample is red, then the subject should choose the comparison that is red; but when the sample is green, the subject should choose the green comparison. Using an item-specific strategy, to respond correctly, subjects might memorize the correct responses as two if then rules, but then when novel items are presented, say a yellow sample and then a yellow and blue comparison, they will perform no better than chance. By contrast, if subjects use a relational strategy, their choice responses will be based on the relationship between the items, and that relationship can control behavior regardless of the specific items presented. Hence, the subjects ability to respond accurately when novel items are presented makes procedures that encourage relational strategies more flexible than procedures that encourage item-specific strategies. Recent research shows that nonhuman subjects can use relational strategies for the abstract concepts of matching and/or same/different (e.g., Blaisdell and Cook 2005; Bodily et al. 2008; Bovet and Vauclair 2001; Brooks and Wasserman 2008; Cook et al. 1997; Katz and Wright 2006; Katz et al. 2002; Wright et al. 2003). For the purpose of the present experiment, it is important to note that such abstract-concept learning can depend on the number of items used in the training set (Bodily et al. 2008; Katz and Wright 2006; Katz et al. 2002, 2007; Nakamura et al. 2009; Wright and Katz 2006; Wright et al. 2003). In our two-item same/different (S/D) task after a pigeon responded to a picture on a computer screen, a picture appeared below it with a white rectangle to the right. If the two pictures were the same, a peck to the lower picture was reinforced. If the two pictures were different, a peck to the white rectangle was reinforced. Hence, the S/D task is different than a MTS task because the different response item (i.e., the white rectangle) occurs on every trial (Wright 2006). In our S/D task, when subjects were trained with a small set-size (8 items), they show no concept learning, as evidenced by their performance with novel items being equivalent to chance (transfer performance is 50%). When more items were added to the training set-size (32, 64, 128), subjects showed increasing partial concept learning, as evidenced by their performance accuracy with novel items being gradually higher than chance but lower than their performance accuracy with training items. When subjects were trained with larger set-sizes (C256), full concept learning emerged, as evidenced when their performance with novel items was statistically similar to their performance with training items. In addition, we have also shown that this high level of transfer cannot be attributed to stimulus generalization (Bodily et al. 2008; Wright and Katz 2007). In summary, transfer to novel items was tested after each set-size expansion and transfer performance increased as the set-size grew, meaning that all the pigeons demonstrated partial concept learning before they all demonstrated full concept learning. In the present experiment, pigeons were trained on the S/D task with set-size expansion and received either the same outcome (like Katz and Wright 2006) or different outcomes following correct choices. For the same-outcome group, each type of correct response was reinforced with the same duration of grain access accompanied by the same sound. For the different-outcomes group, one type of correct choice (e.g., same response) was reinforced with 5-s of grain access accompanied by one sound, and the other type of correct choice (e.g., different response) was reinforced with 1-s of grain access accompanied by a different sound (or vice versa). The purpose of this experiment was to explore whether different outcomes would facilitate acquisition of the S/D task and relational learning of the task. Facilitated acquisition of the task would be evidenced if the pigeons trained with the different-outcomes procedure acquired the task

3 Anim Cogn (2010) 13: with fewer training sessions than those trained with the same-outcomes procedure. Facilitated relational learning would be evidenced if the different-outcomes group transferred better to novel items than the same-outcome group, at least at some training set-size. At the smallest training set (8 items), there would likely be no transfer by either group, based upon the results of previous studies (Katz and Wright 2006; Nakamura et al. 2009). As the setsize expanded, both groups would likely at some point transfer to novel stimuli with accuracy equivalent to their baseline performance. Therefore, group differences would most likely appear at some intermediate size of the training set. Method Subjects Eight experimentally naïve White Carneaux pigeons (Columba livia) from the Palmetto Pigeon Plant in Sumter, South Carolina were maintained at 85% of their freefeeding weight. Grit and water were available at all times when pigeons were in their individual home cage. The colony room was maintained on a 14/10 h light dark cycle. Apparatus Chambers Pigeons were tested in a wooden chamber that was 35.9-cm wide cm deep cm high. A fan (Dayton 5C115A, Niles, IL) located at the back wall of the chamber provided ventilation and white noise. A 28 V, 0.07 A, 1.96 W (No. 1829, Chicago Miniature, Hackensack, NJ) houselight was located in the center of the ceiling of the chamber. Mixed grain was delivered from a custom-built hopper, through a cm opening centered 3.8 cm above the floor in the front panel. An infrared touch screen (17 00 Unitouch, Carroll Round Touch, Round Rock, TX) detected the pigeons pecks. The touch screen was pressure fit into a cutout in the front center of the chamber ( cm) 7.7 cm from the chamber s ceiling. Items The items were digitized travel-slide color pictures. Stimuli were presented on a 40.3-cm color monitor (Eizo T550; Ishikawa, Japan, pixel resolution). Stimulus displays consisted of two travel-slide color pictures (each cm) and a white rectangle ( cm) on a black background. The pictures were vertically aligned with a 1.28-cm gap between them. The top picture was centered cm from the left edge and cm from the top of the front panel. The bottom of the white rectangle was horizontally aligned with the bottom of the lower picture with a 1.4-cm gap between them. The white rectangle always occurred on the right side of the lower picture. Same and different trials were composed by quasirandomly selecting pictures from an 8-item set referred to as: Apples, Buildings, Cat, Woman s Face, Flower, Glass and Pitcher, Keys, and Orangutan and can be seen along with examples of the displays in color in Wright and Katz (2006). There were a total of 64 stimulus display configurations (8 same and 56 different). Sounds Sound stimuli were emitted simultaneously from two speakers to the left and right of the food hopper. For the same-outcome group, the sound stimulus was always a 0.5-s, Hz tone. For the different-outcomes group, sound 1 was the 5-s Windows Ò opening theme and sound 2 was a 1-s ditty. Experimental control Events were controlled and recorded with custom software written in Visual Basic on a Pentium personal computer. A video card (ATI 3D Rage Pro AGP 2X, Ontario, Canada) controlled the monitor. A computer-controlled relay interface (Model no. PI0-12, Metrabyte, Taunton, MA) operated the food hopper and houselight. Training procedure Preliminary training All pigeons were first trained to eat from the hopper. Next, their responses were autoshaped to a white rectangle (later to become the different response area) and another similar white rectangle in the position where the lower picture would appear during the S/D task. The rectangles were presented on separate trials and randomly occurred 50 times each in a 100-trial session. A rectangle was presented for 10 s. If a peck to a rectangle occurred within the 10-s presentation, then the rectangle was removed and food presented. Either a peck or the 10-s presentation was followed by the 0.5-s, Hz tone and the illuminated food hopper for 2 7 s (depending on the individual pigeon s metabolism and weight). A 50-s intertrial interval (ITI) followed reinforcement. Once a pigeon consistently responded to the rectangles (1 7 sessions), S/D training began.

4 586 Anim Cogn (2010) 13: Same/different training Initial training on the S/D task began with presentation of the upper item on a black background. Over the first 10 training sessions, the number of responses to the upper picture was gradually increased from 1 to 10. Following completion of the peck requirement to the upper item, the lower item and the white rectangle were simultaneously presented along with the upper item. If the two items were the same, a peck to the lower item was correct and was rewarded. If the two items were different, a peck to the white rectangle was correct and was rewarded. After the choice response, the display was immediately extinguished. Correct choices resulted in a particular duration of grain access and a particular sound, both of which varied according to the pigeon s group. The four pigeons in the same-outcome group always experienced the same outcome within a session, access to grain and the tone, for both correct choice types. The duration of grain access was between 2 and 5 s (depending on the individual pigeon s metabolism and weight) and was always constant within a session. For two pigeons in the different-outcomes group, correct same choice responses were followed by 5-s access to grain and sound 1. Correct different choice responses were followed by 1-s access to grain and sound 2. These outcomes were reversed for the remaining two pigeons in the different-outcomes group. All reinforcements were followed by a 15-s ITI. Starting on the fifth training session, incorrect choices were also followed by a repeat of the incorrect trial (correction procedure). On correction trials, a darkened 15-s time-out preceded the ITI. An incorrect trial was repeated until it was correct. Accuracy was based on first trial performance only. Performance on correction trials did not figure in any analyses presented in this article. Sessions consisted of 100 trials (50 same/50 different). The sequence of same and different trials was randomly constructed and varied from day to day. The items used to construct the displays were selected with replacement from the 8-item set. If a session was not completed within 2 h, it was stopped and resumed the next day. Training continued until performance was 80% or better over three consecutive sessions. The correction procedure was then removed and training continued until the same criterion was again met. After having acquired the task at each set-size, transfer testing began (except at the 16-item set). Transfer testing Transfer testing was conducted for six consecutive sessions. Each testing session contained 100 trials (90 baseline trials and 10 transfer trials). There were 5 same and 5 different novel transfer trials each session. A novel picture item was used only once during transfer testing. Thus, the transfer test consisted of 90 trial-unique items (15 pictures per session for 6 sessions). The 10 transfer trials within each session were pseudorandomly placed between trials 7 and 92. Performance on transfer trials was reinforced identically to baseline trials. Set-size expansion At the completion of transfer testing, the item set-size was increased from 8 to 16 and then to 32, 64, 128, 256, 512, and finally 1,024 pictures. To be consistent with the method from our previous research, no testing followed the acquisition of set-size 16 (Katz and Wright 2006; Katz et al. 2002; Wright et al. 2003). Instead, after reaching criterion at the 16-item set-size, the birds were immediately trained with the 32-item set-size. Each set-size increase retained the items used from the previous training set and an equal number of new ones. For each session, the items used to construct the displays were randomly selected from the training session s item set-size. The performance criterion for 16 items, following at least two training sessions with a correction procedure, was 85% correct. The correction procedure was then removed and training resumed until 85% correct performance was again obtained. No correction procedure was used for 32, 64, 128, 256, 512, and 1,024-item sets (unless performance fell below 70% correct). Training was conducted with these sets for a minimum of three sessions along with a performance criterion of 85% correct on a single session prior to transfer testing. These criteria were identical to those used in our prior research. Transfer testing was identical to that following 8-item acquisition except the transfer stimuli were all novel and the baseline stimuli were taken from the just acquired set-size. All training and transfer stimuli can be seen, in color, in Wright and Katz (2006). Results Acquisition The differential-outcomes procedure did not hasten the acquisition rate of the S/D task for training with the 8-item set-size. An independent t-test showed that the same-outcome group (M = 2,325.00, SEM = ) and differentoutcomes group 1 (M = 2,366.67, SEM = ) acquired 1 For acquisition, we compared the different-outcomes subgroups at each set-size to see if they could be combined into a single group using a two-way repeated measures ANOVA of Set-Size 9 Group on trials to criterion. Similar analyses were conducted on training trial accuracy and novel item trial test accuracy for the transfer test sessions. In all cases, no interactions were found. Therefore, for the acquisition and transfer analyses, the different-outcomes subgroups were combined into a single different-outcomes group.

5 Anim Cogn (2010) 13: the 8-item set-size after a similar number of trials, t 6 = 0.22, P = In fact, there were no differences in acquisition rate at any set-size, as confirmed by a two-way repeated measures analysis of variance (ANOVA) with Group (different-outcomes, same-outcome) and Set-Size (8, 16, 32, 64, 128, 256, 512, 1,024) as factors on trials to criterion, which yielded no significant effects (Group, F 1,6 = 0.356, P = 0.572; Set-Size, F 7,42 = 1.8, P = 0.106; Group 9 Set-Size, F 7,42 = 1.100, P = 0.381). Testing For both groups, transfer performance was initially at chance but became equivalent to baseline performance at the 128-item set-size. Figure 1 shows each group s performance on baseline and transfer trials during test sessions at each set-size up to full concept learning. The differentoutcomes procedure did have an effect on transfer performance at the 32-item set, as confirmed by a three-way repeated measures ANOVA of Group (different-outcomes, same-outcome) 9 Trial Type (baseline, transfer) 9 Set- Size (8, 32, 64, 128 1,024), which yielded main effects of Set-Size, F 3,18 = 26.38, P \ 0.001, and Trial Type, F 1,6 = , P \ 0.001, a Trial Type 9 Set-Size interaction, F 3,18 = 60.8, P \ 0.001, and a Group 9 Set- Size interaction, F 3,18 = 5.3, P = Post hoc tests indicated this latter interaction was due to the 13.3% increase in transfer performance for the different-outcomes group relative to the same-outcome group at the 32-item set-size, t 6 = 3.189, P \ (different-outcomes group: M = , SD = 4.977, same-outcome group: M = 58.75, SD = ). The difference in baseline performance of the two groups at 32-item set-size Percent Correct Same Outcome Baseline Same Outcome Transfer Different Outcomes Baseline Different Outcomes Transfer Set Size Fig. 1 Mean percentage correct for baseline and transfer performance for the same- and different-outcomes groups over the six transfer sessions at each set-size. Data were collapsed across set-sizes 128 1,024 because this was the point where full concept learning occurred. The bars represent the standard error of the mean approached but was not significant, t 6 = 2.211, P = 0.069, and one might expect that increase due to relational learning. But to take into account the different-outcomes group s increased performance on baseline items, the differences in baseline and transfer performance at the 32-item set-size were calculated to yield a difference score for each pigeon s transfer performance (baseline accuracy - transfer accuracy). The difference score indicates an absolute level of transfer, where a score of zero indicates identical baseline and transfer performance and full abstract-concept learning. Comparing the groups using the difference scores at the 32-item set-size, the difference between the groups was significant, t 6 = 3.188, P \ 0.019, with the different-outcomes group s difference score being closer to zero (M = , SD = 1.931) than the sameoutcome group s score (M = , SD = 4.597). This result means that the difference in baseline performance does not explain the difference in transfer performance; hence, facilitation of abstract-concept learning at the 32-item set-size. Additional analyses indicated transfer performance was stable across the six consecutive test session days at each set-size, as revealed by separate ANOVAs with Test Session Day (1 6) and Group (different-outcomes, same-outcome) as factors, F s range from 0.52 to General discussion In the present experiment, different outcomes were associated with correct responses to same and different choices. The different-outcomes procedure did not affect acquisition rate across the expanding set-size, but did facilitate transfer by evidencing greater partial concept learning (13.3%) than the same-outcomes procedure at the 32-item set-size. As the training set-size increased beyond 32 items, both groups transfer performance converged and became equivalent to training performance (i.e., full concept learning). These results show for the first time that training with different outcomes can facilitate control by relational associations in pigeons. The present results further expand the literature on abstract-concept learning. Here, we have replicated our previous research showing that the number of items used to train an abstract concept is a critical factor as to whether subjects can apply the concept. In addition, the experiment provides evidence that although more training items facilitate concept formation, fewer exemplars are necessary in order to observe a substantial level of intermediate concept learning. This is an important result in that it demonstrates that there is not a fixed number of training exemplars to obtain a given level of novel-stimulus transfer. Thus, somewhat different outcome expectancies

6 588 Anim Cogn (2010) 13: associated with the two parts of an abstract relation (same and different in this case) can promote abstract-concept learning. The finding that the different-outcomes procedure facilitates transfer performance after training with fewer exemplars is worth mentioning for several reasons. To start, these results are the first to find a DOE with a relational strategy in any nonhuman species. By comparison, Cook et al. (1995) explicitly used the differential-outcomes procedure to train their pigeons in an S/D task. Qualitatively different seed types where used to reward each correct choice type, same and different, so that one correct choice type resulted in peanut hearts and the other correct choice type resulted in safflower seeds. After reversing the location of the outcomes and replacing both the outcomes with mixed grain, they concluded that the differential outcomes did not affect the pigeons performance. In contrast to their results, the present experiment suggests that the differential-outcomes procedure can affect pigeons performance. The present experiment findings suggest for the first time that the differential-outcomes procedure can affect and facilitate pigeons relational learning. This result is interesting because the DOE literature emphasizes the conditioned connection between the specific samples and outcomes. In our task, the outcomes are not correlated with specific samples; the outcomes are correlated with abstract rules. In humans, only one study has found a DOE on a relational strategy (Estévez et al. 2007). In Estévez et al. s first experiment, participants were presented with positive number relations, such as 4.33 \ 4.09, on a computer screen. Participants learned to press the K key if the relation expressed by the sign was correct and the J key if the relation was incorrect. Correct responses to a greaterthan or less-than relation were followed consistently by either a short melody or the word Great. Estévez et al. found that the differential-outcomes procedure enabled participants to respond faster; however, it did not affect participants accuracy. In a second experiment, the participants were presented with positive and negative numbers alone or in combination. In their experiment, they found that the differential-outcomes procedure enabled participants to respond more accurately when two negative numbers were presented (e.g., -1 \ -3). But, the previous effect on reaction time disappeared. While these experiments demonstrate a relational strategy for humans, it did not require the subjects to learn that relation. Instead, the humans entered the experiment with the greater-than and less-than concept. In contrast, the pigeons in our experiment entered the experiment without any explicit training in the concept of same or different. Our set-size expansion procedure allowed us to examine how subjects learn to use a relational strategy. The present experiment shows the differential-outcomes procedure can hasten relational learning and did so with a nonhuman species the pigeon a species that has long been regarded as deficient in relational concept learning (e.g., Premack 1978). Why did the differential-outcomes procedure not enhance acquisition rate, as measured by trials to criterion? A likely reason for this result is that unlike other different-outcome experiments that encourage item-specific strategies perhaps the same/different task is learned relationally from its outset. That is, the pigeons are learning the task relationally (not item-specifically) even at the small 8-item set-size. The reason the pigeons initially failed transfer to novel items is not because they lack a same/different concept after training with the small set-size, but because the application of the same/different concept is limited to specific items. In such a case, the pigeons would be learning the task relationally, but without abstraction. The standard test of abstract-concept learning is successful transfer to novel items. The common interpretation of a lack of novel-stimulus transfer is that the subjects must have been learning the task itemspecifically. But that interpretation (i.e., speculation) may be incorrect, at least in these similar S/D tasks with picture stimuli (Elmore et al. 2009; Nakamura et al. 2009; Wright and Katz 2009). To elaborate, one method we used was to train pigeons and monkeys with only a portion of the 64 possible training displays constructed from the 8-item set and testing them with the untrained novel combinations. If pigeons learned the task itemspecifically, they would fail transfer to the untrained set. They passed transfer to the untrained set but not transfer to novel pictures. Such results indicated relational learning but not abstract-concept learning. The equivalent rates of acquisition in the present experiment support the idea that the pigeons are learning the task relationally and not item-specifically. Consider the alternative, if the pigeons of the present experiment were learning the task item-specifically, then the different-outcomes group would be expected to acquire the task faster then the same-outcome group in acquisition, a standard DOE. But there was no acquisition advantage for the different-outcomes group. What does result from differential outcomes appears to be a more rapid expansion of the restricted-domain relational learning for the different-outcomes group over the same-outcome group, producing better novel-stimulus transfer at the intermediate set-size of 32 items. Acknowledgments This research was supported by NIH grants, MH , MH , and NSF grant IBN This experiment was conducted following the relevant ethical guidelines for animal research (IACUC approved and conducted in AAALAC approved facilities). This research was part of Kelly A. Schmidtke s Masters Thesis. We thank Kent Bodily, Michelle Hernandez, John

7 Anim Cogn (2010) 13: Magnotti, and Brad Sturz for their careful assistance in conducting this experiment. References Alling K, Nickel M, Poling A (1991) The effects of differential and nondifferential outcomes on response rates and accuracy under a delayed-matching-to-sample procedure. Psychol Rec 41: Blaisdell AP, Cook RG (2005) Two-item same-different concept learning in pigeons. Learn Behav 33:67 77 Bodily KD, Katz JS, Wright AA (2008) Matching-to-sample abstractconcept learning by pigeons. J Exp Psychol Anim B 34: Bovet D, Vauclair J (2001) Judgment of conceptual identity in monkeys. Psychon Bull Rev 8: Brodigan DL, Peterson GB (1976) Two-choice conditional discrimination performance of pigeons as a function of reward expectancy, prechoice delay, and domesticity. Anim Learn Behav 4: Brooks DI, Wasserman EA (2008) Same/different discrimination learning with trial-unique stimuli. Psychon Bull Rev 15: Carlson JG, Wielkiewicz RM (1976) Mediators of the effects of magnitude of reinforcement. Learn Motiv 7: Cook RG, Cavoto KK, Cavoto BR (1995) Same-different texture discrimination and concept learning in pigeons. J Exp Psychol Anim B 21: Cook RG, Katz JS, Cavoto BR (1997) Pigeon same-different concept learning with multiple stimulus classes. J Exp Psychol Anim B 23: DeMarse TB, Urcuioli PJ (2005) Control of matching by differential outcome expectancies in the absence of differential sampleoutcome associations: a serial compound view. J Exp Psychol Anim B 34: Edwards CA, Jagielo JA, Zentall TR, Hogan DE (1982) Acquired equivalence and distinctiveness in matching to sample by pigeons: mediation by reinforcer-specific expectancies. J Exp Psychol Anim B 8: Elmore LC, Wright AA, Rivera JJ, Katz JS (2009) Individual differences: either relational learning or item-specific learning in a same/different task. Learn Behav 37: Estévaz AF, Fuentes L, Marý-Beffa P, Gonzalez C, Alvarez D (2001) The differential outcome effect as a useful tool to improve conditional discrimination leaning in children. Learn Motiv 32:48 64 Estévez AF, Fuentes LJ, Overmier BJ, González C (2003) Differential outcomes effect in children and adults with Down syndrome. Am J Ment Retard 108: Estévez AF, Vivas AB, Alonso D, Marý-Beffa P, Fuentes LJ, Overmier BJ, González C (2007) Enhancing challenged students recognition of mathematical relations through differential outcomes training. Q J Exp Psychol 60: Grant DS, Kelly R (2001) Anticipation and short-term retention in pigeons. In: Cook RG (ed) Avian visual cognition. Available via Accessed 1 Aug 2009 Hewitt FM (1965) Teaching speech to an autistic child through operant conditioning. Am J Orthopsychiatry 35: Hochhalter AK, Sweeney WA, Bakke BL, Holub RJ, Overmier JB (2000) Improving face recognition in alcohol dementia. Clin Gerontol 22:3 18 Jansen C, Guess D (1978) Use of function as a consequence in training receptive labeling to severely and profoundly retarded individuals. AAESPH 3: Katz JS, Wright AA (2006) Same/different abstract-concept learning by pigeons. J Exp Psychol Anim B 32:80 86 Katz JS, Wright AA, Bachevalier J (2002) Mechanisms of same/ different abstract-concept learning by rhesus monkeys (Macaca mulatta). J Exp Psychol Anim B 28: Katz JS, Wright AA, Bodily KD (2007) Issues in the comparative cognition of abstract-concept learning. Comp Cogn Behav Rev 2:79 92 Kruse JM, Overmier JB, Konz WA, Rokke E (1983) Pavlovian CSs effects upon instrumental choice behavior are reinforcer specific. Learn Motiv 14: Maki P, Overmier BJ, Delos S, Gutmann AJ (1995) Expectancies as factors influencing conditional discrimination performance of children. Psychol Rec 45:45 72 Miller O, Waugh KM, Chambers K (2002) Differential outcomes effect: increased accuracy in adults learning kanji with stimulusspecific rewards. Psychol Rec 51: Miyashita Y, Nakajima S, Imada H (2000) Differential outcomes effect in horses. J Exp Anal Behav 74: Nakamura T, Wright AA, Katz JS, Bodily KD, Sturz BR (2009) Abstract-concept learning carryover effects from the initial training set in pigeons (Columba livia). J Comp Psychol :79 89 Overmier JB, Bull JA III, Trapold MA (1971) Discriminative cue properties of different fears and their role in response selection in dogs. J Comp Physiol 76: Premack D (1978) On the abstractness of human concepts: why it would be difficult to talk to a pigeon. In: Hulse SH, Fowler H, Honig WK (eds) Cognitive processes in animal behavior. Hillsdale, New Jersey, pp Sidman M (2000) Equivalence relations and the reinforcement contingency. J Exp Anal Behav 74: Trapold MA (1970) Are expectancies based upon different positive reinforcing events discriminably different? Learn Motiv 1: Urcuioli PJ (1990a) Differential outcomes and many-to-one matching: effects of correlation with correct choice. Anim Learn Behav 18: Urcuioli PJ (1990b) Some relationships between outcome expectancies and sample stimuli in pigeons delayed matching. Anim Learn Behav 18: Urcuioli PJ (2005) Behavioral and associative effects of differential outcomes in discrimination learning. Learn Behav 33:1 21 Urcuioli PJ, Lionello-DeNolf KM (2001) Some tests of the anticipatory mediated generalization model of acquired sample equivalence in pigeons many-to-one matching. Anim Learn Behav 29: Wright AA (2006) Memory processing. In: Wasserman EA, Zentall TR (eds) Comparative cognition: experimental exploration of animal intelligence. Oxford University Press, New York, pp Wright AA, Katz JS (2006) Mechanisms of same/different concept learning in primates and avians. Behav Process 72: Wright AA, Katz JS (2007) Generalization hypothesis of abstractconcept learning: learning strategies and related issues in Macaca mulatta, Cebus apella, and Columba livia. J Comp Psychol 121: Wright AA, Katz JS (2009) A case for restricted-domain relational learning. Psychon Bull Rev 16: Wright AA, Rivera JJ, Katz JS, Bachevalier J (2003) Abstractconcept learning and list-memory processing by capuchin and rhesus monkeys. J Exp Psychol Anim B 29: