- Email: [email protected]
Configural learning in human Pavlovian conditioning: acquisition of a biconditional discrimination
Biological Psychology 59 (2002) 163– 168 www.elsevier.com/locate/biopsycho
Brief report
Configural learning in human Pavlovian conditioning: acquisition of a biconditional discrimination Klaus Lober *, Harald Lachnit Fachbereich Psychologie, Philipps-Uni6ersita¨t Marburg, Gutenbergstraße 18, D-35032 Marburg, Germany Received 3 September 2001; accepted 7 January 2002
Abstract Previous studies of conditioning have shown that non-human animals are able to master discrimination problems which cannot be solved on the basis of elemental associations. Most of these discrimination problems, however, have not yet been investigated in human Pavlovian conditioning. In a skin conductance conditioning experiment we therefore assessed whether humans can solve a biconditional discrimination. Participants underwent conditioning with an AB +, CD+, AD−, CB− design and successfully mastered this discrimination. Thus, they were able to learn about the relationship of specific configurations of stimuli with the reinforcer. The results extend previous findings of configural learning in humans. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Configural learning; Humans; Pavlovian conditioning; Skin conductance response
1. Introduction Most theories of associative learning assume that each element of a stimulus compound is represented separately and builds its own association with the reinforcer (Mackintosh, 1975; Pearce and Hall, 1980; Rescorla and Wagner, 1972). According to those elemental theories, the associative strength of a compound is based on the sum of the associative strengths of its elements. * Corresponding author. Tel.: + 49-6421-2823439; fax: +49-6421-2826621. E-mail addresses: [email protected] (K. Lober), [email protected] (H. Lachnit). 0301-0511/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 1 - 0 5 1 1 ( 0 2 ) 0 0 0 0 4 - 2
164
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
A serious problem for elemental theories is that organisms are able to solve certain configural discrimination problems that are not consistent with the summation principle. One of these problems is the so-called biconditional discrimination (Saavedra, 1975). In this problem, animals are trained with an AB+ , CD + , AD − , CB − discrimination in which each element is reinforced on half of its presentations and non-reinforced on the other half. To solve it, animals have to learn to associate a reinforcer with specific configurations of stimuli. The summation principle fails to predict successful differentiation because summation of elemental associations should lead to intermediate levels of responding to each compound. There are several ways to overcome this problem. First, one can retain the summation principle but assume that the presentation of a compound not only activates the representations of its elements but also the representation of a hypothetical unique cue (Wagner, 1971). This cue codes the simultaneous presentation of a specific configuration of stimuli. Thus, an AB compound will not only activate representations of A and B but also an additional representation of a unique cue, say X. For a biconditional problem, it is assumed that each compound will activate its own unique cue. These cues will then acquire either positive (for AB+ and CD +) or negative (for AD− and CB − ) associative strengths during the course of learning. Thereby, they will allow to successfully master the discrimination. Second, configural theories abandon the summation principle. Instead, they assume that compound stimuli are represented as configurations and that associations will develop between these configural representations and the unconditioned stimulus (US; e.g. Pearce, 1987, 1994). Since different compounds are represented separately, these theories have no difficulties to account for the successful acquisition of configural discrimination problems. Given their theoretical importance, it is surprising that of the configural problems only negative patterning (A+ , B +, AB − ) has been investigated in human Pavlovian conditioning (e.g. Lachnit and Kimmel, 1993; Lachnit and Lober, 2001). In particular, we are not aware of any report of a biconditional discrimination in humans although it has been demonstrated in various animals including such diverse species as rabbits (Saavedra, 1975), pigeons (Rescorla et al., 1985), honeybees (Hellstern et al., 2002), and monkeys (Saunders and Weiskrantz, 1989). The present report seeks to provide evidence that humans can solve a biconditional discrimination in a Pavlovian conditioning preparation.
2. Methods
2.1. Participants Thirty-two University of Marburg students (10 males, 22 females) took part in the experiment. The data of an additional participant were excluded from the analyses because she did not show unconditioned skin conductance responses (SCRs) on more than 25% of the reinforced trials.
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
165
2.2. Stimuli and apparatus The letters B, G, T and X were used as conditioned stimuli (CSs). They were presented in compounds on a computer screen positioned 1.5 m in front of the participants. Each letter was approximately 4 cm high and 3 cm wide. The background of the screen was black and the letters were white. One letter of each compound was presented on the left and the other one on the right side of the centre of the screen. Left– right allocation remained constant throughout the experiment. The distance between the letters of a compound was approximately 1.5 cm. The CSs were presented for 8 s on each trial. A DC electric shock served as the US. The shock was delivered via Ag/AgCl electrodes to the volar surface of the participants’ left arm from an isolated transformer – condensor shock generator (Kimmel et al., 1980). The intensity of the shock was adjusted individually so that it would be ‘definitely unpleasant but not really painful’. Shock duration was approximately 20 ms. On reinforced trials, the shock was delivered simultaneously with CS-offset. The intertrial interval (CS-onset to CS-onset) was 2493 s. Palmar skin conductance was picked from the thenar and hypothenar eminences of the participants’ right hand by Ag/AgCl electrodes. These were 0.8 cm in diameter and were filled with an electrolytic medium (Unibase, with 0.05 mol NaCl). Skin conductance was measured via a constant voltage bridge (Lykken and Venables, 1971) and sampled by a computer at 20 Hz.
2.3. Procedure Participants were run individually in a sound-attenuated room. They sat in a cushioned chair facing towards the computer screen. After their skin was cleaned with alcohol, the electrodes were attached and the US intensity was adjusted. Then participants were given the written instructions, which stated the purpose of the experiment as measuring their physiological responses to various stimuli (letters on a screen as well as occasional electric shocks), and that participants should try to relax and avoid unnecessary movement and heavy breathing. The training consisted of 24 trials, six presentations of each compound. The stimuli were counterbalanced across participants so that the compounds TG and XB were reinforced and TB and XG were non-reinforced for half of the participants whereas these contingencies were reversed for the remaining participants. Orthogonally to the CS contingencies, we manipulated whether the discrimination training began with a reinforced or a non-reinforced CS. The remaining CSs were presented in a pre-determined random order with the exception that no more than three reinforced or three non-reinforced stimuli were administered in direct succession.
2.4. Dependent 6ariables Two SCR indices were used as dependent variables. The first interval response (FIR) was defined as the maximum change in conductance beginning within 1–4 s
166
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
after CS onset (i.e. the difference between the skin conductance level at the peak of the response and the value just at its onset). The second interval response (SIR) was defined as the maximum conductance change beginning during the interval 4– 9 s after CS onset. Conductance changes were converted to logarithmic values (after adding 1) and then multiplied by 1000.
3. Results and discussion Stated probability levels for statistical analyses are based on the Greenhouse– Geisser adjustment of degrees of freedom where appropriate (Greenhouse and Geisser, 1959). Fig. 1 shows mean SCR magnitudes during reinforced (labelled AB+
Fig. 1. Mean electrodermal FIR (top panel) and SIR (bottom panel) during discrimination training. Compounds AB and CD were reinforced with a US whereas compounds AD and CB were non-reinforced. Responses are shown in blocks of two trials.
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
167
and CD + ) and non-reinforced trials (labelled AD− and CB− ). For this figure and the statistical analyses, trials with the same contingency were collapsed in blocks of two according to their order of presentation during the acquisition. As can be seen in Fig. 1, participants showed larger FIRS and SIRS to reinforced compounds than to non-reinforced compounds. This was confirmed by two separate 2× 6 repeated-measures analyses of variance, which included the factors contingency (reinforced vs. non-reinforced) and block of trials. For both FIR and SIR, successful response differentiation was evident in significant main effects of contingency, F(1, 31)=9.36, MSE = 5342.21, PB 0.006, and F(1, 31)= 9.18, MSE = 14585.70, P B0.006, respectively. Furthermore, significant Contingency× Block interactions were found, F(5, 155) = 3.14, MSE = 2772.07, PB 0.03, and F(5, 155)= 3.12, MSE = 4338.04, PB 0.02. For SIR, this interaction indicates increasing differentiation as training progressed. For FIR, it indicates that responses to reinforced compounds were only significantly larger on Blocks 2 and 5. The factor block was not significant, both Fs B2.54. Additional analyses were run to test whether responding differed across compounds with the same contingency. No main effect of compound was found, all Fs B 2.50, indicating similar SCRs. To our knowledge, this has been the first demonstration of the acquisition of a biconditional discrimination in human Pavlovian conditioning. Thus, our results extend findings from conditioning experiments with other species and add to previous evidence showing that humans are able to solve negative patterning, which also requires a configural solution. We would, however, like to stress that the ability to solve configural discriminations does not mean that humans will always process stimuli configurally. Instead, recent evidence from another domain of human associative learning suggests that humans can process stimulus compounds flexibly. In causal learning experiments, it has been found that whether compounds are processed elementally or configurally may, for example, depend on participants’ past experience (Williams and Braker, 1999; Williams et al., 1994). In those experiments, participants showed successful learning of discrimination problems that are usually solved elementally when they had experienced a pre-training that also encouraged an elemental solution but not after a different pre-training that encouraged a configural solution. In a recent study, we could extend these findings to electrodermal Pavlovian conditioning (Lober et al., 2002). Further research will have to investigate whether prior experience will similarly influence the subsequent acquisition of configural problems like negative patterning or biconditional discrimination.
Acknowledgements The research reported in this article was supported by grant La 564/12-1 from the German Science Foundation (Deutsche Forschungsgemeinschaft: DFG) to Harald Lachnit. We thank Metin U8 ngo¨ r, Anne Meinhardt, and Sarah Schreckenberg for their help with the experiment, and Frauke Melchers, Ottmar Lipp, and three anonymous reviewers for valuable comments on earlier versions of this paper.
168
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
References Greenhouse, S.W., Geisser, S., 1959. On methods in the analysis of profile data. Psychometrika 24, 95 – 112. Hellstern, F., Giurfa, M., Lachnit, H., Menzel, R., 2002. Biconditional discrimination in Pavlovian olfactory conditioning with honeybees. Manuscript in preparation. Kimmel, H.D., King, J., Hudy, J.J., Gardner, K.A., 1980. A mutual inductance shocker. Behavior Research Methods, Instruments & Computers 12, 605 – 606. Lachnit, H., Kimmel, H.D., 1993. Positive and negative patterning in human classical skin conductance response conditioning. Animal Learning & Behavior 21, 314 – 326. Lachnit, H., Lober, K., 2001. What is learned in patterning discriminations? Further tests of configural accounts of associative learning in human electrodermal conditioning. Biological Psychology 56, 45 – 61. Lober, K., Lachnit, H., Shanks, D.R., 2002. Past experience influences the processing of compound cues in human Pavlovian conditioning. Manuscript in preparation. Lykken, D.T., Venables, P.H., 1971. Direct measurement of skin conductance: A proposal for standardization. Psychophysiology 8, 656 –672. Mackintosh, N.J., 1975. A theory of attention: Variations in the associability of stimuli with reinforcement. Psychological Review 82, 276 –298. Pearce, J.M., 1987. A model for stimulus generalization in Pavlovian conditioning. Psychological Review 94, 61 – 73. Pearce, J.M., 1994. Similarity and discrimination: A selective review and a connectionist model. Psychological Review 101, 587 –607. Pearce, J.M., Hall, G., 1980. A model for Pavlovian learning: Variations in the effectiveness of conditioned but not of unconditioned stimuli. Psychological Review 87, 532 – 552. Rescorla, R.A., Grau, J.W., Durlach, P.J., 1985. Analysis of the unique cue in configural discriminations. Journal of Experimental Psychology: Animal Behavior Processes 11, 356 – 366. Rescorla, R.A., Wagner, A.R., 1972. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In: Black, A.H., Prokasy, W.F. (Eds.), Classical Conditioning II: Current Theory and Research. Appleton-Century-Crofts, New York, pp. 64 – 99. Saavedra, M.A., 1975. Pavlovian compound conditioning in the rabbit. Learning and Motivation 6, 314 – 326. Saunders, R.C., Weiskrantz, L., 1989. The effects of fornix transection and combined fornix transection, mammillary body lesions and hippocampal ablations on object-pair association memory in the rhesus monkey. Behavioural Brain Research 35, 85 – 94. Wagner, A.R., 1971. Elementary associations. In: Kendler, H.H., Spence, J.T. (Eds.), Essays in Neobehaviorism: A Memorial Volume to Kenneth W. Spence. Appleton-Century-Crofts, New York, pp. 187 – 213. Williams, D.A., Braker, D.S., 1999. Influence of past experience on the coding of compound stimuli. Journal of Experimental Psychology: Animal Behavior Processes 25, 461 – 474. Williams, D.A., Sagness, K.E., McPhee, J.E., 1994. Configural and elemental strategies in predictive learning. Journal of Experimental Psychology: Learning, Memory and Cognition 20, 694 – 709.
Brief report
Configural learning in human Pavlovian conditioning: acquisition of a biconditional discrimination Klaus Lober *, Harald Lachnit Fachbereich Psychologie, Philipps-Uni6ersita¨t Marburg, Gutenbergstraße 18, D-35032 Marburg, Germany Received 3 September 2001; accepted 7 January 2002
Abstract Previous studies of conditioning have shown that non-human animals are able to master discrimination problems which cannot be solved on the basis of elemental associations. Most of these discrimination problems, however, have not yet been investigated in human Pavlovian conditioning. In a skin conductance conditioning experiment we therefore assessed whether humans can solve a biconditional discrimination. Participants underwent conditioning with an AB +, CD+, AD−, CB− design and successfully mastered this discrimination. Thus, they were able to learn about the relationship of specific configurations of stimuli with the reinforcer. The results extend previous findings of configural learning in humans. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Configural learning; Humans; Pavlovian conditioning; Skin conductance response
1. Introduction Most theories of associative learning assume that each element of a stimulus compound is represented separately and builds its own association with the reinforcer (Mackintosh, 1975; Pearce and Hall, 1980; Rescorla and Wagner, 1972). According to those elemental theories, the associative strength of a compound is based on the sum of the associative strengths of its elements. * Corresponding author. Tel.: + 49-6421-2823439; fax: +49-6421-2826621. E-mail addresses: [email protected] (K. Lober), [email protected] (H. Lachnit). 0301-0511/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 1 - 0 5 1 1 ( 0 2 ) 0 0 0 0 4 - 2
164
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
A serious problem for elemental theories is that organisms are able to solve certain configural discrimination problems that are not consistent with the summation principle. One of these problems is the so-called biconditional discrimination (Saavedra, 1975). In this problem, animals are trained with an AB+ , CD + , AD − , CB − discrimination in which each element is reinforced on half of its presentations and non-reinforced on the other half. To solve it, animals have to learn to associate a reinforcer with specific configurations of stimuli. The summation principle fails to predict successful differentiation because summation of elemental associations should lead to intermediate levels of responding to each compound. There are several ways to overcome this problem. First, one can retain the summation principle but assume that the presentation of a compound not only activates the representations of its elements but also the representation of a hypothetical unique cue (Wagner, 1971). This cue codes the simultaneous presentation of a specific configuration of stimuli. Thus, an AB compound will not only activate representations of A and B but also an additional representation of a unique cue, say X. For a biconditional problem, it is assumed that each compound will activate its own unique cue. These cues will then acquire either positive (for AB+ and CD +) or negative (for AD− and CB − ) associative strengths during the course of learning. Thereby, they will allow to successfully master the discrimination. Second, configural theories abandon the summation principle. Instead, they assume that compound stimuli are represented as configurations and that associations will develop between these configural representations and the unconditioned stimulus (US; e.g. Pearce, 1987, 1994). Since different compounds are represented separately, these theories have no difficulties to account for the successful acquisition of configural discrimination problems. Given their theoretical importance, it is surprising that of the configural problems only negative patterning (A+ , B +, AB − ) has been investigated in human Pavlovian conditioning (e.g. Lachnit and Kimmel, 1993; Lachnit and Lober, 2001). In particular, we are not aware of any report of a biconditional discrimination in humans although it has been demonstrated in various animals including such diverse species as rabbits (Saavedra, 1975), pigeons (Rescorla et al., 1985), honeybees (Hellstern et al., 2002), and monkeys (Saunders and Weiskrantz, 1989). The present report seeks to provide evidence that humans can solve a biconditional discrimination in a Pavlovian conditioning preparation.
2. Methods
2.1. Participants Thirty-two University of Marburg students (10 males, 22 females) took part in the experiment. The data of an additional participant were excluded from the analyses because she did not show unconditioned skin conductance responses (SCRs) on more than 25% of the reinforced trials.
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
165
2.2. Stimuli and apparatus The letters B, G, T and X were used as conditioned stimuli (CSs). They were presented in compounds on a computer screen positioned 1.5 m in front of the participants. Each letter was approximately 4 cm high and 3 cm wide. The background of the screen was black and the letters were white. One letter of each compound was presented on the left and the other one on the right side of the centre of the screen. Left– right allocation remained constant throughout the experiment. The distance between the letters of a compound was approximately 1.5 cm. The CSs were presented for 8 s on each trial. A DC electric shock served as the US. The shock was delivered via Ag/AgCl electrodes to the volar surface of the participants’ left arm from an isolated transformer – condensor shock generator (Kimmel et al., 1980). The intensity of the shock was adjusted individually so that it would be ‘definitely unpleasant but not really painful’. Shock duration was approximately 20 ms. On reinforced trials, the shock was delivered simultaneously with CS-offset. The intertrial interval (CS-onset to CS-onset) was 2493 s. Palmar skin conductance was picked from the thenar and hypothenar eminences of the participants’ right hand by Ag/AgCl electrodes. These were 0.8 cm in diameter and were filled with an electrolytic medium (Unibase, with 0.05 mol NaCl). Skin conductance was measured via a constant voltage bridge (Lykken and Venables, 1971) and sampled by a computer at 20 Hz.
2.3. Procedure Participants were run individually in a sound-attenuated room. They sat in a cushioned chair facing towards the computer screen. After their skin was cleaned with alcohol, the electrodes were attached and the US intensity was adjusted. Then participants were given the written instructions, which stated the purpose of the experiment as measuring their physiological responses to various stimuli (letters on a screen as well as occasional electric shocks), and that participants should try to relax and avoid unnecessary movement and heavy breathing. The training consisted of 24 trials, six presentations of each compound. The stimuli were counterbalanced across participants so that the compounds TG and XB were reinforced and TB and XG were non-reinforced for half of the participants whereas these contingencies were reversed for the remaining participants. Orthogonally to the CS contingencies, we manipulated whether the discrimination training began with a reinforced or a non-reinforced CS. The remaining CSs were presented in a pre-determined random order with the exception that no more than three reinforced or three non-reinforced stimuli were administered in direct succession.
2.4. Dependent 6ariables Two SCR indices were used as dependent variables. The first interval response (FIR) was defined as the maximum change in conductance beginning within 1–4 s
166
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
after CS onset (i.e. the difference between the skin conductance level at the peak of the response and the value just at its onset). The second interval response (SIR) was defined as the maximum conductance change beginning during the interval 4– 9 s after CS onset. Conductance changes were converted to logarithmic values (after adding 1) and then multiplied by 1000.
3. Results and discussion Stated probability levels for statistical analyses are based on the Greenhouse– Geisser adjustment of degrees of freedom where appropriate (Greenhouse and Geisser, 1959). Fig. 1 shows mean SCR magnitudes during reinforced (labelled AB+
Fig. 1. Mean electrodermal FIR (top panel) and SIR (bottom panel) during discrimination training. Compounds AB and CD were reinforced with a US whereas compounds AD and CB were non-reinforced. Responses are shown in blocks of two trials.
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
167
and CD + ) and non-reinforced trials (labelled AD− and CB− ). For this figure and the statistical analyses, trials with the same contingency were collapsed in blocks of two according to their order of presentation during the acquisition. As can be seen in Fig. 1, participants showed larger FIRS and SIRS to reinforced compounds than to non-reinforced compounds. This was confirmed by two separate 2× 6 repeated-measures analyses of variance, which included the factors contingency (reinforced vs. non-reinforced) and block of trials. For both FIR and SIR, successful response differentiation was evident in significant main effects of contingency, F(1, 31)=9.36, MSE = 5342.21, PB 0.006, and F(1, 31)= 9.18, MSE = 14585.70, P B0.006, respectively. Furthermore, significant Contingency× Block interactions were found, F(5, 155) = 3.14, MSE = 2772.07, PB 0.03, and F(5, 155)= 3.12, MSE = 4338.04, PB 0.02. For SIR, this interaction indicates increasing differentiation as training progressed. For FIR, it indicates that responses to reinforced compounds were only significantly larger on Blocks 2 and 5. The factor block was not significant, both Fs B2.54. Additional analyses were run to test whether responding differed across compounds with the same contingency. No main effect of compound was found, all Fs B 2.50, indicating similar SCRs. To our knowledge, this has been the first demonstration of the acquisition of a biconditional discrimination in human Pavlovian conditioning. Thus, our results extend findings from conditioning experiments with other species and add to previous evidence showing that humans are able to solve negative patterning, which also requires a configural solution. We would, however, like to stress that the ability to solve configural discriminations does not mean that humans will always process stimuli configurally. Instead, recent evidence from another domain of human associative learning suggests that humans can process stimulus compounds flexibly. In causal learning experiments, it has been found that whether compounds are processed elementally or configurally may, for example, depend on participants’ past experience (Williams and Braker, 1999; Williams et al., 1994). In those experiments, participants showed successful learning of discrimination problems that are usually solved elementally when they had experienced a pre-training that also encouraged an elemental solution but not after a different pre-training that encouraged a configural solution. In a recent study, we could extend these findings to electrodermal Pavlovian conditioning (Lober et al., 2002). Further research will have to investigate whether prior experience will similarly influence the subsequent acquisition of configural problems like negative patterning or biconditional discrimination.
Acknowledgements The research reported in this article was supported by grant La 564/12-1 from the German Science Foundation (Deutsche Forschungsgemeinschaft: DFG) to Harald Lachnit. We thank Metin U8 ngo¨ r, Anne Meinhardt, and Sarah Schreckenberg for their help with the experiment, and Frauke Melchers, Ottmar Lipp, and three anonymous reviewers for valuable comments on earlier versions of this paper.
168
K. Lober, H. Lachnit / Biological Psychology 59 (2002) 163–168
References Greenhouse, S.W., Geisser, S., 1959. On methods in the analysis of profile data. Psychometrika 24, 95 – 112. Hellstern, F., Giurfa, M., Lachnit, H., Menzel, R., 2002. Biconditional discrimination in Pavlovian olfactory conditioning with honeybees. Manuscript in preparation. Kimmel, H.D., King, J., Hudy, J.J., Gardner, K.A., 1980. A mutual inductance shocker. Behavior Research Methods, Instruments & Computers 12, 605 – 606. Lachnit, H., Kimmel, H.D., 1993. Positive and negative patterning in human classical skin conductance response conditioning. Animal Learning & Behavior 21, 314 – 326. Lachnit, H., Lober, K., 2001. What is learned in patterning discriminations? Further tests of configural accounts of associative learning in human electrodermal conditioning. Biological Psychology 56, 45 – 61. Lober, K., Lachnit, H., Shanks, D.R., 2002. Past experience influences the processing of compound cues in human Pavlovian conditioning. Manuscript in preparation. Lykken, D.T., Venables, P.H., 1971. Direct measurement of skin conductance: A proposal for standardization. Psychophysiology 8, 656 –672. Mackintosh, N.J., 1975. A theory of attention: Variations in the associability of stimuli with reinforcement. Psychological Review 82, 276 –298. Pearce, J.M., 1987. A model for stimulus generalization in Pavlovian conditioning. Psychological Review 94, 61 – 73. Pearce, J.M., 1994. Similarity and discrimination: A selective review and a connectionist model. Psychological Review 101, 587 –607. Pearce, J.M., Hall, G., 1980. A model for Pavlovian learning: Variations in the effectiveness of conditioned but not of unconditioned stimuli. Psychological Review 87, 532 – 552. Rescorla, R.A., Grau, J.W., Durlach, P.J., 1985. Analysis of the unique cue in configural discriminations. Journal of Experimental Psychology: Animal Behavior Processes 11, 356 – 366. Rescorla, R.A., Wagner, A.R., 1972. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In: Black, A.H., Prokasy, W.F. (Eds.), Classical Conditioning II: Current Theory and Research. Appleton-Century-Crofts, New York, pp. 64 – 99. Saavedra, M.A., 1975. Pavlovian compound conditioning in the rabbit. Learning and Motivation 6, 314 – 326. Saunders, R.C., Weiskrantz, L., 1989. The effects of fornix transection and combined fornix transection, mammillary body lesions and hippocampal ablations on object-pair association memory in the rhesus monkey. Behavioural Brain Research 35, 85 – 94. Wagner, A.R., 1971. Elementary associations. In: Kendler, H.H., Spence, J.T. (Eds.), Essays in Neobehaviorism: A Memorial Volume to Kenneth W. Spence. Appleton-Century-Crofts, New York, pp. 187 – 213. Williams, D.A., Braker, D.S., 1999. Influence of past experience on the coding of compound stimuli. Journal of Experimental Psychology: Animal Behavior Processes 25, 461 – 474. Williams, D.A., Sagness, K.E., McPhee, J.E., 1994. Configural and elemental strategies in predictive learning. Journal of Experimental Psychology: Learning, Memory and Cognition 20, 694 – 709.