- Email: [email protected]
Wether ewe know me or not: the discrimination of individual humans by sheep1
Behavioural Processes 43 (1998) 27 – 32
Wether ewe know me or not: the discrimination of individual humans by sheep1 Hank Davis *, Christina Norris, Allison Taylor Department of Psychology, Uni6ersity of Guelph, Guelph, Ontario NIG 2W1, Canada Received 12 June 1997; received in revised form 7 October 1997; accepted 7 October 1997
Abstract A growing literature suggests that animals of various species can discriminate between individual humans. In the present study, 15 experimentally naive sheep were rewarded for making a nosepress response in the presence of one handler (S +) and non-reinforced for this behavior in the presence of a second person (S − ). All animals responded significantly more to the S + handler (PB0.001) during non-reinforced test trials. Furthermore, sheep engaged in clearly different behavior during a 10-s pre-trial period, depending on which handler was present. Control conditions preclude discrimination based on order effects, temporal cues, or a win-stay/lose-shift learning set. The ability to differentiate between individual humans, regardless of its sensory basis, suggests that animals can use humans to predict the delivery of hedonic events that routinely occur in research settings. The resultant human-based operant and Pavlovian conditioning has implications for the design of research protocols and the interpretation of results. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Sheep; Discrimination; Human recognition
1. Introduction Recent work from our laboratory suggests that a variety of animals including rats, cows, seals, llamas and chickens have the ability to discrimi-
* Corresponding author. 1 Spelling of all words in the title is intentional. A wether refers to a male sheep castrated before maturity and a ewe is a female sheep.
nate between individual human beings (Taylor and Davis, 1996; Davis et al., 1997). These findings are consistent with reports by other investigators detailing human recognition in prairie dogs (Slobodchikoff et al., 1991); pigs (Tanida et al., 1994); chimpanzees (Boysen and Berntson, 1986); and domestic dogs (Settle et al., 1994). The ability to differentiate between humans is fundamental to the suggestion by Davis and Balfour (1992) that repeated contact with scientists or technicians might itself be an experimental vari-
0376-6357/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PII S 0 3 7 6 - 6 3 5 7 ( 9 7 ) 0 0 0 8 2 - X
28
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
able in much animal research. McGuigan (1963) emphasized the importance of the search for experimenter effects in behavioral research and described it as ‘a long, energy consuming project.’ To that end, Davis and Balfour (1992) offered preliminary evidence that animals may use individual persons to predict hedonic events routinely involved in many kinds of research. Such events could include, but are not limited to, being fed, shocked, petted, or subjected to surgical procedures. If the animal associates these experiences with a specific human, the presence of that human could result in anticipatory behavioral or physiological responses (Summerlee, 1992). This possibility is consistent with Pavlov’s observation that ‘person’ could serve as a specific conditioned stimulus (CS) for measurable conditioned responses. Such human-based Pavlovian conditioning has been demonstrated in dogs by Gantt et al. (1966) and others. Similarly, an individual person might function as a discriminative stimulus (SD) by providing a context in which certain forms of operant behavior are rewarded or punished. Given the growing body of evidence to suggest both human recognition and its predictive function for many animals, it is surprising that Kendrick (1990) has argued that ‘...sheep cannot distinguish between humans, their sex, what they are wearing, (regardless of) whether the back or front view is presented...’. In the following experiment, techniques were applied which had been previously used with cows and rabbits to investigate the ability of domestic sheep to discriminate between two different humans and to use this information to predict whether responding would be rewarded by food.
2. Method
2.1. Subjects The subjects involved in this study consisted of 16 experimentally-naive intact male Arcott sheep. All animals were housed at the University of Guelph for subsequent use in the veterinary teaching hospital. The 10-month-old sheep were kept in
pairs in adjacent pens measuring 351× 303 cm. The sheep remained on a regular forage-based diet and had free access to water throughout the experiment. A mineral supplement (Sodium molybdate topdress) was used in 10 g quantities as a reinforcement. Two wire gates, each measuring 170× 83 cm, were hinged together and used to enclose the individual sheep within the home pen for training and testing.
2.2. Procedure The sheep were randomly divided into two groups. The first group was trained and tested 1 month prior to the second. Each group of eight sheep was housed in pairs in four adjacent pens along one side of the barn. Two female handlers served as both S+ and S− for half of the animals. The assignment of handler to S+ and S− conditions was counterbalanced across pair of animals. The same person was used as S+ for the two sheep that were penned together. Handlers differed in height (1.7 vs. 1.65 m) and hair color (brown vs. blond). Both handlers wore standard green coveralls and black rubber boots during training and testing. Neither handler wore makeup or jewellery and their individual hygiene regimen (e.g. soap, deodorant) was held constant throughout training and testing. The handlers only spoke to indicate the total number of responses at 30-s intervals for data recording. Training and testing consisted of 10-min sessions per sheep each day for 5 days a week. The test animal was separated from his pen mate using the gate, creating a 170 cm2 training area within the home pen. In order to minimize disruption during training and testing, no sheep was denied visual, auditory, or olfactory contact with his pen mate or neighboring sheep. The handler stood in one corner of the training area facing the sheep. Food (topdress) was kept in the right front pocket of the handler’s coveralls to repeatedly replace the 10 g portions used as a reinforcement. During training and test sessions, the handler (regardless of whether she was serving as S+ or S−) stood with her left arm extended straight
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
down in front of her body with her left hand in a fist. The handler’s right arm hung at a 45° angle adjacent to the right side of her body, with her right hand in a fist containing approximately 10 g of food reinforcement.
2.3. Training The sheep were initially shaped by the S+ handler, using successive approximations, to nosepress the handler’s left fist in order to receive food reinforcement presented by the handler’s right hand. Shaping required an average of 7 sessions to complete (range 6 – 8). A nosepress was defined as the sheep using his nose to make contact with the handler’s left fist. Responses were clearly defined, discrete actions (i.e. touch and withdraw). Inter-observer reliability of response measurement during selected video-taped sessions exceeded 0.9. After the response had been shaped and the sheep was responding continuously, variable ratio (VR) and discrimination training began. Each subsequent session included both VR and discrimination training. The ratio schedule was gradually increased from continuous reinforcement (CRF) to VR3. When the sheep was responding at a consistent rate, usually within three sessions, its schedule was increased to a VR5 so that each sheep was reinforced on average for every fifth response in the presence of the S+ handler. The increase to VR 3 and VR 5 schedules was made at the discretion of the handler without reference to a formal criterion. The VR5 schedule remained in effect for the balance of training. The schedule was based upon a random number table of 122 values (range 1 – 10) that averaged 4.98. Discrimination training took place during the same sessions. A successive discrimination procedure was employed. The sheep were reinforced for responding only in the presence of the S+ person. Initial discrimination sessions included alternating S + trials of 2 min and S − trials of 30 s throughout the 10-min session. The S − trials were gradually increased to 2 min as the sheep learned to withhold responses in S − and work continuously in S+ . The sheep eventually worked on alternating S+ and S − trials of 2 min each for a total session length of 10 min. The
29
changeover between stimulus individuals (i.e. the transition between S+ and S− ) involved practiced, standardized movements so as to reduce, to the greatest extent possible, extraneous cues. The number of nosepress responses was recorded at each 30 s interval in both S+ and S− . One animal did not complete training and was therefore not included in the test sessions. No data were collected or analyzed for this animal, leaving a total of 15 subjects.
2.4. Testing Once the rate of responding in S+ was three times that recorded in S−, the sheep was given two consecutive test sessions. Data collection took place during extinction; i.e. no responses, whether in S+ or S− , were reinforced. Each test session lasted approximately 10 min. S + and S − trials of random length were presented in random sequence in order to preclude discrimination based on temporal or order cues. On occasions in which two consecutive trials involved the same individual (e.g. S + followed by S + ), the cue person engaged in the same behavior (leaving, re-entering) that was involved in a change of cue sequence. The test sessions included three randomly inserted 30-s S+ trials during which responding was reinforced in order to minimize extinction of nosepress responding during testing. However, data from these trials were not included in the analysis. In order to maximize data collection during extinction test sessions, shortened S+ and S− trials, varying in length from 30 to 90 s, were used. Equivalent S + and S − time periods (total= 5.5 min) were used to compare response frequencies statistically. Each trial during both training and testing was preceded by a 10-s pre-trial exposure. During this time the handler remained standing in the corner of the training area with both hands behind her back (free of reinforcement and unavailable for response) for 10 s. After the pre-trial exposure, the handler obtained food from her pocket (whether serving as S+ or S− ) and placed her hands in the appropriate position (as described above) for the subsequent trial. The handler’s
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
30
right hand therefore always contained grain (regardless of whether she served as S+ or S− ) in order to avoid discrimination based on olfactory cues. There were no ITIs in which the two experimenters were not visible to the sheep being tested. Behavior during a random sample of 10-s pretrial exposures was classified according to the following categories: (1) looking away from the handler; (2) watching handler; (3) sniffing handler; or (4) reaching for handler.
3. Results All animals (n=15) learned to nosepress on a VR5 schedule of reinforcement and to discriminate between S + and S − in an average of 6.9 sessions (range 6–8). Data from the two test sessions were pooled and a x 2 analysis was computed for each sheep comparing responding during extinction in the presence of S+ and S− handlers. These results are shown in Table 1. All subjects showed a significantly greater response frequency (P B 0.001) in the presence of the S+ handler. Analysis of behavioral data during the 10-s pre-trial exposures revealed a consistent pattern Table 1 x 2 Comparisons of response frequencies to S+ and S− handlers during equivalent unreinforced test periods Subject
S+
S−
x 2*
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15
77 128 74 84 107 121 111 132 99 80 101 89 76 99 93
0 1 1 3 2 3 5 3 2 4 7 1 0 1 3
77.00 125.04 71.06 75.42 101.14 112.30 96.86 123.26 93.16 68.76 81.82 86.04 76.00 96.04 84.38
* PB0.001. x 2crit (1) = 10.83.
across all animals. Subjects reliably oriented toward, sniffed and reached toward the S+ handler during this period even though nosepress responses were precluded. In contrast, these behaviors did not occur during pre-trial exposure to the S− handler. Instead, subjects occasionally turned their bodies and walked away from the S− handler. Informal observation suggested an identical distinction in the behavior of the second sheep in each pair when they were not being tested. While ‘waiting their turn’ or immediately following testing, these animals occasionally approached or shyed away from the test area, depending on whether the S+ or S− person was present.
4. Discussion All animals responded at a higher rate in the presence of the S+ handler, previously associated with delivery of reinforcement. Responding was virtually eliminated in the presence of the S− person. These data, along with consistent behavioral differences observed during the pre-trial periods, suggest that sheep were able to discriminate between individual humans and associate them with different reinforcement histories. Because testing occurred during extinction, with randomly ordered S+ and S− trials of unequal duration, discrimination is unlikely to reflect extraneous factors such as order or temporal cues. Test results are also unlikely to stem from a win-stay/lose-shift learning set. Using such a strategy, subjects would have made between the average and maximum number of responses required for reinforcement on the VR5 schedule and ceased responding if reinforcement did not occur. However, subjects in this study made, on average, 17.8 responses (range 7–36) in the first 30 s of unreinforced S+ test trials, a number nearly twice the maximum required for VR reinforcement. On average, sheep made less than one response (range 0–3) in the first 30 s of S− test trials, thus suggesting that discrimination was made on the basis of stimulus characteristics of the handlers, rather than the consequences of responding. Although Kendrick (1990) does present some anecdotal evidence to the contrary, his conclusion
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
that sheep cannot distinguish between individual humans is surprising. Certainly, the present research demonstrates that sheep possess this ability. Although the sensory basis for the ability to discriminate between humans has not been explored, existing knowledge of sheep physiology and behavior strongly suggests that both visual and olfactory cues played a primary role in their performance (Baldwin and Meese, 1977; Piggins and Phillips, 1996). The present results extend current knowledge of perception in sheep. Earlier discriminative work showed sheep to be capable of distinguishing between the shape of geometric objects (Baldwin, 1981), although they failed to detect differences in color (Bazely and Ensor, 1989). Recognition of conspecifics by sheep has been demonstrated in a number of studies (Shillito and Alexander, 1975; Alexander and Shillito, 1977; Kendrick et al., 1995) and Kendrick (1992) has suggested that sheep may have specialized neural circuits for responding to particular aspects of conspecifics such as horn size, breed and facial features. Although the response rate differential between S + and S− reveals that individual humans may serve as discriminative stimuli for operant behavior in sheep, the difference in pre-trial behaviour (sniffing and reaching only for the S+ handler) demonstrates a Pavlovian component of this discrimination as well. Such human-based Pavlovian effects may take various forms, including skeletalmotor approach behavior, such as that observed in the present experiment or previously demonstrated in rats using conspecifics as CSs (Timberlake and Grant, 1975). Similarly, Summerlee (1992) reported cortical arousal in rabbits differentially elicited by individual persons in a research environment. Collectively, these findings support the notion of ‘person as CS’ (Gantt et al., 1966) and underscore the possibility that associations made between specific humans and hedonic events in a test environment may directly influence the results of animal research (Davis and Balfour, 1992). While such human-based associative effects may be a source of confounding in animal research, they may also enhance the research process in a variety of ways. For example, Reinhardt (1992) has used
31
the presence of a particular human to facilitate blood and tissue sampling in rhesus monkeys. Similarly, Oden and Thompson (1992), Pepperberg (1992) and Schusterman et al. (1992) have used repeated contact between the researcher and subject to enhance motivation and attention in studies of animal cognition using species as diverse as chimpanzees, parrots and sea lions. Interactions between researcher and animal subject are likely to persist and form the basis for unintended conditioned effects involving both behavior and physiology (Summerlee, 1992). Warnings about such possibilities have remained largely unheeded and empirically unexplored until very recently (McGuigan, 1963). Estep and Hetts (1992) have surveyed the manner in which humans and animals interact within the research domain and concluded ‘‘This relationship can be ignored, or studied and used to advantage, but it will not cease to exist. The scientist who acknowledges the existence of these relationships and understands how they are formed can use this information to produce better, more efficient and more humane research’’.
Acknowledgements This research was supported in part by a grant to HD from the Natural Sciences and Engineering Research Council of Canada. All animals were treated in compliance with the provisions of the Canadian Council on Animal Care. The authors thank Jim Mottin and David Piggins for their critical assistance.
References Alexander, G., Shillito, E. E., 1977. Importance of visual cues from various body regions in maternal recognition of the young in merino sheep (o6is aries). Appl. Anim. Ethol. 3, 137 – 143. Baldwin, B. A., 1981. Shape discrimination in sheep and calves. Anim. Behav. 29, 830 – 834. Baldwin, B. A., Meese, G. B., 1977. The ability of sheep to distinguish between conspecifics by means of olfaction. Physiol. Behav. 18, 803 – 808.
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
32
Bazely, D. R., Ensor, C. V., 1989. Discrimination learning in sheep with cues varying in brightness and hue. Appl. Anim. Behav. Sci. 23, 293–299. Boysen, S. T., Berntson, G. G., 1986. Cardiac correlates of individual recognition in the chimpanzee (Pan troglodytes). J. Comp. Psychol. 100, 321–324. Davis, H. Balfour, D. (Eds.), 1992. The Inevitable Bond: Examining Scientist-Animal Interactions.Cambridge University Press, New York. Davis, H., Taylor, A., Norris, C., 1997. Preference for familiar humans by rats. Psychon. Bull. Rev. 4, 118–120. Estep, D.Q., Hetts, S., 1992. Interactions, relationships, and bonds: the conceptual basis for scientist-animal relations. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 6–26. Gantt, W. H., Newton, J. E., Royer, F. L., Stephens, J. H., 1966. Effect of person. Conditional Reflex 1 (1), 18–35. Kendrick, K., 1992. Cognition. In: Phillips, C. D., Piggins, D. (Eds.), Farm Animals and the Environment. Cambridge University Press, Cambridge, pp. 209–231. Kendrick, K., 1990. Through a sheep’s eye. New Sci. 1716, 62– 65. Kendrick, K. M., Atkins, K., Hinton, M. R., Broad, K. D., Fabre-Nys, C., Keverne, B., 1995. Facial and vocal discrimination in sheep. Anim. Behav. 49, 1665–1676. McGuigan, F. J., 1963. The experimenter: a neglected stimulus object. Psychol. Bull. 60, 421–428. Oden, D. L., Thompson, K. R., 1992. The role of social bonds in motivating chimpanzee cognition. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining ScientistAnimal Interactions. Cambridge University Press, New York, pp. 218 – 231. Pepperberg, I.M., 1992. Social interaction as a condition for learning in avian species: a synthesis of the disciplines of ethology and psychology. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 178– 204.
.
Piggins, D., Phillips, C. J. C., 1996. The eye of the domesticated sheep with implications for vision. Anim. Sci. 62, 301 – 308. Reinhardt, V., 1992. Improved handling of experimental rhesus monkeys. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 171 – 177. Schusterman, R. J., Gisiner, R., Hanggi, E. B., 1992. Imprinting and other aspects of pinniped-human interactions. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 334 – 356. Settle, R. H., Sommerville, B. A., McCormick, J., Broom, D. M., 1994. Human scent matching using specially trained dogs. Anim. Behav. 48, 1443 – 1448. Shillito, E., Alexander, G., 1975. Mutual recognition amongst ewes and lambs of four breeds of sheep (o6is aries). Appl. Anim. Ethol. 1, 151 – 165. Slobodchikoff, C. N., Kiriazis, J., Fischer, C., Creef, E., 1991. Semantic information distinguishing individual predators in the alarm calls of Gunnison’s prairie dogs. Anim. Behav. 42, 713 – 719. Summerlee, A. J. S., 1992. Behavioral arousal and its effect on the experimental animal and the experimenter. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 132 – 152. Tanida, H., Miura, A., Tanaka, T., Yoshimoto, T., 1994. The role of handling in communication between humans and weanling pigs. Appl. Anim. Behav. Sci. 40, 219 – 228. Taylor, A., Davis, H., 1996. The response of llamas (Lama glama) to familiar and unfamiliar humans. Int. J. Comp. Psychol. 9, 42 – 50. Timberlake, W., Grant, D. L., 1975. Auto-shaping in rats to the presentation of another rat predicting food. Science 190, 690 – 692.
Wether ewe know me or not: the discrimination of individual humans by sheep1 Hank Davis *, Christina Norris, Allison Taylor Department of Psychology, Uni6ersity of Guelph, Guelph, Ontario NIG 2W1, Canada Received 12 June 1997; received in revised form 7 October 1997; accepted 7 October 1997
Abstract A growing literature suggests that animals of various species can discriminate between individual humans. In the present study, 15 experimentally naive sheep were rewarded for making a nosepress response in the presence of one handler (S +) and non-reinforced for this behavior in the presence of a second person (S − ). All animals responded significantly more to the S + handler (PB0.001) during non-reinforced test trials. Furthermore, sheep engaged in clearly different behavior during a 10-s pre-trial period, depending on which handler was present. Control conditions preclude discrimination based on order effects, temporal cues, or a win-stay/lose-shift learning set. The ability to differentiate between individual humans, regardless of its sensory basis, suggests that animals can use humans to predict the delivery of hedonic events that routinely occur in research settings. The resultant human-based operant and Pavlovian conditioning has implications for the design of research protocols and the interpretation of results. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Sheep; Discrimination; Human recognition
1. Introduction Recent work from our laboratory suggests that a variety of animals including rats, cows, seals, llamas and chickens have the ability to discrimi-
* Corresponding author. 1 Spelling of all words in the title is intentional. A wether refers to a male sheep castrated before maturity and a ewe is a female sheep.
nate between individual human beings (Taylor and Davis, 1996; Davis et al., 1997). These findings are consistent with reports by other investigators detailing human recognition in prairie dogs (Slobodchikoff et al., 1991); pigs (Tanida et al., 1994); chimpanzees (Boysen and Berntson, 1986); and domestic dogs (Settle et al., 1994). The ability to differentiate between humans is fundamental to the suggestion by Davis and Balfour (1992) that repeated contact with scientists or technicians might itself be an experimental vari-
0376-6357/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PII S 0 3 7 6 - 6 3 5 7 ( 9 7 ) 0 0 0 8 2 - X
28
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
able in much animal research. McGuigan (1963) emphasized the importance of the search for experimenter effects in behavioral research and described it as ‘a long, energy consuming project.’ To that end, Davis and Balfour (1992) offered preliminary evidence that animals may use individual persons to predict hedonic events routinely involved in many kinds of research. Such events could include, but are not limited to, being fed, shocked, petted, or subjected to surgical procedures. If the animal associates these experiences with a specific human, the presence of that human could result in anticipatory behavioral or physiological responses (Summerlee, 1992). This possibility is consistent with Pavlov’s observation that ‘person’ could serve as a specific conditioned stimulus (CS) for measurable conditioned responses. Such human-based Pavlovian conditioning has been demonstrated in dogs by Gantt et al. (1966) and others. Similarly, an individual person might function as a discriminative stimulus (SD) by providing a context in which certain forms of operant behavior are rewarded or punished. Given the growing body of evidence to suggest both human recognition and its predictive function for many animals, it is surprising that Kendrick (1990) has argued that ‘...sheep cannot distinguish between humans, their sex, what they are wearing, (regardless of) whether the back or front view is presented...’. In the following experiment, techniques were applied which had been previously used with cows and rabbits to investigate the ability of domestic sheep to discriminate between two different humans and to use this information to predict whether responding would be rewarded by food.
2. Method
2.1. Subjects The subjects involved in this study consisted of 16 experimentally-naive intact male Arcott sheep. All animals were housed at the University of Guelph for subsequent use in the veterinary teaching hospital. The 10-month-old sheep were kept in
pairs in adjacent pens measuring 351× 303 cm. The sheep remained on a regular forage-based diet and had free access to water throughout the experiment. A mineral supplement (Sodium molybdate topdress) was used in 10 g quantities as a reinforcement. Two wire gates, each measuring 170× 83 cm, were hinged together and used to enclose the individual sheep within the home pen for training and testing.
2.2. Procedure The sheep were randomly divided into two groups. The first group was trained and tested 1 month prior to the second. Each group of eight sheep was housed in pairs in four adjacent pens along one side of the barn. Two female handlers served as both S+ and S− for half of the animals. The assignment of handler to S+ and S− conditions was counterbalanced across pair of animals. The same person was used as S+ for the two sheep that were penned together. Handlers differed in height (1.7 vs. 1.65 m) and hair color (brown vs. blond). Both handlers wore standard green coveralls and black rubber boots during training and testing. Neither handler wore makeup or jewellery and their individual hygiene regimen (e.g. soap, deodorant) was held constant throughout training and testing. The handlers only spoke to indicate the total number of responses at 30-s intervals for data recording. Training and testing consisted of 10-min sessions per sheep each day for 5 days a week. The test animal was separated from his pen mate using the gate, creating a 170 cm2 training area within the home pen. In order to minimize disruption during training and testing, no sheep was denied visual, auditory, or olfactory contact with his pen mate or neighboring sheep. The handler stood in one corner of the training area facing the sheep. Food (topdress) was kept in the right front pocket of the handler’s coveralls to repeatedly replace the 10 g portions used as a reinforcement. During training and test sessions, the handler (regardless of whether she was serving as S+ or S−) stood with her left arm extended straight
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
down in front of her body with her left hand in a fist. The handler’s right arm hung at a 45° angle adjacent to the right side of her body, with her right hand in a fist containing approximately 10 g of food reinforcement.
2.3. Training The sheep were initially shaped by the S+ handler, using successive approximations, to nosepress the handler’s left fist in order to receive food reinforcement presented by the handler’s right hand. Shaping required an average of 7 sessions to complete (range 6 – 8). A nosepress was defined as the sheep using his nose to make contact with the handler’s left fist. Responses were clearly defined, discrete actions (i.e. touch and withdraw). Inter-observer reliability of response measurement during selected video-taped sessions exceeded 0.9. After the response had been shaped and the sheep was responding continuously, variable ratio (VR) and discrimination training began. Each subsequent session included both VR and discrimination training. The ratio schedule was gradually increased from continuous reinforcement (CRF) to VR3. When the sheep was responding at a consistent rate, usually within three sessions, its schedule was increased to a VR5 so that each sheep was reinforced on average for every fifth response in the presence of the S+ handler. The increase to VR 3 and VR 5 schedules was made at the discretion of the handler without reference to a formal criterion. The VR5 schedule remained in effect for the balance of training. The schedule was based upon a random number table of 122 values (range 1 – 10) that averaged 4.98. Discrimination training took place during the same sessions. A successive discrimination procedure was employed. The sheep were reinforced for responding only in the presence of the S+ person. Initial discrimination sessions included alternating S + trials of 2 min and S − trials of 30 s throughout the 10-min session. The S − trials were gradually increased to 2 min as the sheep learned to withhold responses in S − and work continuously in S+ . The sheep eventually worked on alternating S+ and S − trials of 2 min each for a total session length of 10 min. The
29
changeover between stimulus individuals (i.e. the transition between S+ and S− ) involved practiced, standardized movements so as to reduce, to the greatest extent possible, extraneous cues. The number of nosepress responses was recorded at each 30 s interval in both S+ and S− . One animal did not complete training and was therefore not included in the test sessions. No data were collected or analyzed for this animal, leaving a total of 15 subjects.
2.4. Testing Once the rate of responding in S+ was three times that recorded in S−, the sheep was given two consecutive test sessions. Data collection took place during extinction; i.e. no responses, whether in S+ or S− , were reinforced. Each test session lasted approximately 10 min. S + and S − trials of random length were presented in random sequence in order to preclude discrimination based on temporal or order cues. On occasions in which two consecutive trials involved the same individual (e.g. S + followed by S + ), the cue person engaged in the same behavior (leaving, re-entering) that was involved in a change of cue sequence. The test sessions included three randomly inserted 30-s S+ trials during which responding was reinforced in order to minimize extinction of nosepress responding during testing. However, data from these trials were not included in the analysis. In order to maximize data collection during extinction test sessions, shortened S+ and S− trials, varying in length from 30 to 90 s, were used. Equivalent S + and S − time periods (total= 5.5 min) were used to compare response frequencies statistically. Each trial during both training and testing was preceded by a 10-s pre-trial exposure. During this time the handler remained standing in the corner of the training area with both hands behind her back (free of reinforcement and unavailable for response) for 10 s. After the pre-trial exposure, the handler obtained food from her pocket (whether serving as S+ or S− ) and placed her hands in the appropriate position (as described above) for the subsequent trial. The handler’s
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
30
right hand therefore always contained grain (regardless of whether she served as S+ or S− ) in order to avoid discrimination based on olfactory cues. There were no ITIs in which the two experimenters were not visible to the sheep being tested. Behavior during a random sample of 10-s pretrial exposures was classified according to the following categories: (1) looking away from the handler; (2) watching handler; (3) sniffing handler; or (4) reaching for handler.
3. Results All animals (n=15) learned to nosepress on a VR5 schedule of reinforcement and to discriminate between S + and S − in an average of 6.9 sessions (range 6–8). Data from the two test sessions were pooled and a x 2 analysis was computed for each sheep comparing responding during extinction in the presence of S+ and S− handlers. These results are shown in Table 1. All subjects showed a significantly greater response frequency (P B 0.001) in the presence of the S+ handler. Analysis of behavioral data during the 10-s pre-trial exposures revealed a consistent pattern Table 1 x 2 Comparisons of response frequencies to S+ and S− handlers during equivalent unreinforced test periods Subject
S+
S−
x 2*
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15
77 128 74 84 107 121 111 132 99 80 101 89 76 99 93
0 1 1 3 2 3 5 3 2 4 7 1 0 1 3
77.00 125.04 71.06 75.42 101.14 112.30 96.86 123.26 93.16 68.76 81.82 86.04 76.00 96.04 84.38
* PB0.001. x 2crit (1) = 10.83.
across all animals. Subjects reliably oriented toward, sniffed and reached toward the S+ handler during this period even though nosepress responses were precluded. In contrast, these behaviors did not occur during pre-trial exposure to the S− handler. Instead, subjects occasionally turned their bodies and walked away from the S− handler. Informal observation suggested an identical distinction in the behavior of the second sheep in each pair when they were not being tested. While ‘waiting their turn’ or immediately following testing, these animals occasionally approached or shyed away from the test area, depending on whether the S+ or S− person was present.
4. Discussion All animals responded at a higher rate in the presence of the S+ handler, previously associated with delivery of reinforcement. Responding was virtually eliminated in the presence of the S− person. These data, along with consistent behavioral differences observed during the pre-trial periods, suggest that sheep were able to discriminate between individual humans and associate them with different reinforcement histories. Because testing occurred during extinction, with randomly ordered S+ and S− trials of unequal duration, discrimination is unlikely to reflect extraneous factors such as order or temporal cues. Test results are also unlikely to stem from a win-stay/lose-shift learning set. Using such a strategy, subjects would have made between the average and maximum number of responses required for reinforcement on the VR5 schedule and ceased responding if reinforcement did not occur. However, subjects in this study made, on average, 17.8 responses (range 7–36) in the first 30 s of unreinforced S+ test trials, a number nearly twice the maximum required for VR reinforcement. On average, sheep made less than one response (range 0–3) in the first 30 s of S− test trials, thus suggesting that discrimination was made on the basis of stimulus characteristics of the handlers, rather than the consequences of responding. Although Kendrick (1990) does present some anecdotal evidence to the contrary, his conclusion
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
that sheep cannot distinguish between individual humans is surprising. Certainly, the present research demonstrates that sheep possess this ability. Although the sensory basis for the ability to discriminate between humans has not been explored, existing knowledge of sheep physiology and behavior strongly suggests that both visual and olfactory cues played a primary role in their performance (Baldwin and Meese, 1977; Piggins and Phillips, 1996). The present results extend current knowledge of perception in sheep. Earlier discriminative work showed sheep to be capable of distinguishing between the shape of geometric objects (Baldwin, 1981), although they failed to detect differences in color (Bazely and Ensor, 1989). Recognition of conspecifics by sheep has been demonstrated in a number of studies (Shillito and Alexander, 1975; Alexander and Shillito, 1977; Kendrick et al., 1995) and Kendrick (1992) has suggested that sheep may have specialized neural circuits for responding to particular aspects of conspecifics such as horn size, breed and facial features. Although the response rate differential between S + and S− reveals that individual humans may serve as discriminative stimuli for operant behavior in sheep, the difference in pre-trial behaviour (sniffing and reaching only for the S+ handler) demonstrates a Pavlovian component of this discrimination as well. Such human-based Pavlovian effects may take various forms, including skeletalmotor approach behavior, such as that observed in the present experiment or previously demonstrated in rats using conspecifics as CSs (Timberlake and Grant, 1975). Similarly, Summerlee (1992) reported cortical arousal in rabbits differentially elicited by individual persons in a research environment. Collectively, these findings support the notion of ‘person as CS’ (Gantt et al., 1966) and underscore the possibility that associations made between specific humans and hedonic events in a test environment may directly influence the results of animal research (Davis and Balfour, 1992). While such human-based associative effects may be a source of confounding in animal research, they may also enhance the research process in a variety of ways. For example, Reinhardt (1992) has used
31
the presence of a particular human to facilitate blood and tissue sampling in rhesus monkeys. Similarly, Oden and Thompson (1992), Pepperberg (1992) and Schusterman et al. (1992) have used repeated contact between the researcher and subject to enhance motivation and attention in studies of animal cognition using species as diverse as chimpanzees, parrots and sea lions. Interactions between researcher and animal subject are likely to persist and form the basis for unintended conditioned effects involving both behavior and physiology (Summerlee, 1992). Warnings about such possibilities have remained largely unheeded and empirically unexplored until very recently (McGuigan, 1963). Estep and Hetts (1992) have surveyed the manner in which humans and animals interact within the research domain and concluded ‘‘This relationship can be ignored, or studied and used to advantage, but it will not cease to exist. The scientist who acknowledges the existence of these relationships and understands how they are formed can use this information to produce better, more efficient and more humane research’’.
Acknowledgements This research was supported in part by a grant to HD from the Natural Sciences and Engineering Research Council of Canada. All animals were treated in compliance with the provisions of the Canadian Council on Animal Care. The authors thank Jim Mottin and David Piggins for their critical assistance.
References Alexander, G., Shillito, E. E., 1977. Importance of visual cues from various body regions in maternal recognition of the young in merino sheep (o6is aries). Appl. Anim. Ethol. 3, 137 – 143. Baldwin, B. A., 1981. Shape discrimination in sheep and calves. Anim. Behav. 29, 830 – 834. Baldwin, B. A., Meese, G. B., 1977. The ability of sheep to distinguish between conspecifics by means of olfaction. Physiol. Behav. 18, 803 – 808.
H. Da6is et al. / Beha6ioural Processes 43 (1998) 27–32
32
Bazely, D. R., Ensor, C. V., 1989. Discrimination learning in sheep with cues varying in brightness and hue. Appl. Anim. Behav. Sci. 23, 293–299. Boysen, S. T., Berntson, G. G., 1986. Cardiac correlates of individual recognition in the chimpanzee (Pan troglodytes). J. Comp. Psychol. 100, 321–324. Davis, H. Balfour, D. (Eds.), 1992. The Inevitable Bond: Examining Scientist-Animal Interactions.Cambridge University Press, New York. Davis, H., Taylor, A., Norris, C., 1997. Preference for familiar humans by rats. Psychon. Bull. Rev. 4, 118–120. Estep, D.Q., Hetts, S., 1992. Interactions, relationships, and bonds: the conceptual basis for scientist-animal relations. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 6–26. Gantt, W. H., Newton, J. E., Royer, F. L., Stephens, J. H., 1966. Effect of person. Conditional Reflex 1 (1), 18–35. Kendrick, K., 1992. Cognition. In: Phillips, C. D., Piggins, D. (Eds.), Farm Animals and the Environment. Cambridge University Press, Cambridge, pp. 209–231. Kendrick, K., 1990. Through a sheep’s eye. New Sci. 1716, 62– 65. Kendrick, K. M., Atkins, K., Hinton, M. R., Broad, K. D., Fabre-Nys, C., Keverne, B., 1995. Facial and vocal discrimination in sheep. Anim. Behav. 49, 1665–1676. McGuigan, F. J., 1963. The experimenter: a neglected stimulus object. Psychol. Bull. 60, 421–428. Oden, D. L., Thompson, K. R., 1992. The role of social bonds in motivating chimpanzee cognition. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining ScientistAnimal Interactions. Cambridge University Press, New York, pp. 218 – 231. Pepperberg, I.M., 1992. Social interaction as a condition for learning in avian species: a synthesis of the disciplines of ethology and psychology. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 178– 204.
.
Piggins, D., Phillips, C. J. C., 1996. The eye of the domesticated sheep with implications for vision. Anim. Sci. 62, 301 – 308. Reinhardt, V., 1992. Improved handling of experimental rhesus monkeys. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 171 – 177. Schusterman, R. J., Gisiner, R., Hanggi, E. B., 1992. Imprinting and other aspects of pinniped-human interactions. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 334 – 356. Settle, R. H., Sommerville, B. A., McCormick, J., Broom, D. M., 1994. Human scent matching using specially trained dogs. Anim. Behav. 48, 1443 – 1448. Shillito, E., Alexander, G., 1975. Mutual recognition amongst ewes and lambs of four breeds of sheep (o6is aries). Appl. Anim. Ethol. 1, 151 – 165. Slobodchikoff, C. N., Kiriazis, J., Fischer, C., Creef, E., 1991. Semantic information distinguishing individual predators in the alarm calls of Gunnison’s prairie dogs. Anim. Behav. 42, 713 – 719. Summerlee, A. J. S., 1992. Behavioral arousal and its effect on the experimental animal and the experimenter. In: Davis, H., Balfour, D. (Eds.), The Inevitable Bond: Examining Scientist-Animal Interactions The Inevitable Bond: Examining Scientist-Animal Interactions. Cambridge University Press, New York, pp. 132 – 152. Tanida, H., Miura, A., Tanaka, T., Yoshimoto, T., 1994. The role of handling in communication between humans and weanling pigs. Appl. Anim. Behav. Sci. 40, 219 – 228. Taylor, A., Davis, H., 1996. The response of llamas (Lama glama) to familiar and unfamiliar humans. Int. J. Comp. Psychol. 9, 42 – 50. Timberlake, W., Grant, D. L., 1975. Auto-shaping in rats to the presentation of another rat predicting food. Science 190, 690 – 692.