Changes in stimulus salience as a result of stimulus preexposure: evidence from aversive and appetitive testing procedures

Learning & Behavior2003, 31 (2), 185-191 Changes in stimulus salience as a result of stimulus preexposure: Evidence from aversive In two experiments, rats received preexposure to three compound flavor stimuli, AX, BX, and CX, where X represents a saline solution. AX and BX were presented in alternation; CX, on a separate block of trials. The value of X was then modified, being devalued by aversive conditioning in Experiment 1, and rendered valuable by the induction of a state of salt need in Experiment 2. When given a choice be- tween BX and CX, the rats consumed more of BX than of CX in Experiment 1, and more of CX than of BX in Experiment 2, suggesting that B and C differed in their ability to modulate the response governed by the X element. It was suggested that blocked preexposure to CX reduces the salience of the C stim- ulus but that the salience of B is maintained by preexposure in which BX is alternated with AX. The im- plications of this result for the phenomenon of perceptual learning are discussed.
Our standard theories of classical conditioning (see, salience can be explained in terms of known associative e.g., Wagner, 1981) assume the existence of a represen- tational node for each of the events that can be shown to The experimental evidence on which Hall's (2003) ac- be effective as a conditioned or unconditioned stimulus count is based comes, in part, from a recent series of (CS or US). Learning is held to occur when two nodes studies reported by Blair and Hall (2003) examining gen- are activated concurrently and to consist of changes in eralization in flavor aversion learning after preexposure the strength of associative links between them. Since the to the stimuli. Our basic procedure made use of three fla- degree of change will depend on the extent to which each vors, A, B, and C, that were rendered similar by the ad- of the nodes is activated, it is important to be able to dition of an explicit element (X) common to each. (Here specify the relationship between the external stimulus A represents a lemon solution; B and C represent sucrose and the level of activity that its presentation induces in its and saline solutions, counterbalanced; and X, a solution node. The usual assumption has been that the level of ac- of quinine.) One of these compounds (AX) was trained tivity in a node will be determined directly by the inten- as a CS in the flavor-aversion paradigm; the aim of the sity (or salience) of the stimulus that is applied. But Hall experiment was to see how different schedules of pre- (2003) has challenged this assumption, arguing, on the exposure to the stimuli would influence generalization basis of evidence from experiments on perceptual learn- from AX to the test compounds BX and CX. In the pre- ing, that the ability of a given stimulus to activate its node exposure phase, there were four presentations of each of can change with experience-that mere exposure to a the three stimuli. Presentations of CX occurred in one stimulus can engage a learning process that produces block of trials; presentations of AX and BX, however, a change in the effective salience of the stimulus. Hall were intermixed, occurring on alternate trials. It was suggests that repeated exposure to a stimulus will nor- found (Blair & Hall, 2003, Experiment 1) that the aver- mally result in a loss of salience but that exposure to in- sion established to AX generalized less readily to BX termixed (alternating) presentations of two similar stim- than to CX. In explaining these results, we argued that uli will maintain, or even enhance, the effective salience the aversion displayed on the test trials in this procedure of their distinctive features. The experiments to be de- would be largely determined by the ability of the condi- scribed in this article were designed to explore the latter tioned element, X, to evoke its conditioned response suggestion, and, in particular, to assess whether the evi- (CR) and that the differential responding shown to BX dence that has been interpreted as indicating changes in and CX indicated that the B and C differed in their ability to interfere with the expression of this CR. We suggested that the effective salience of B and C had been modified This work was supported by a studentship from the Biotechnology and during the preexposure phase-that the salience of C had Biological Science Research Council to Chris Blair. We thank C. Bonardi, declined during the block of exposure trials with CX but E. Mondragón, and M. Symonds for helpful discussion. Correspondence that alternating trials with AX and BX had maintained or concerning this article should be addressed to either C. A. J. Blair or G. Hall, who are both at the Department of Psychology, University of enhanced the salience of their unique features (and in York, York YO10 5DD, England (e-mail: GH1@york.ac.uk).
particular of B). The less salient C stimulus would be less Copyright 2003 Psychonomic Society, Inc.
likely to interfere with the ability of the conditioned el- compound as in the previous experiment) was used as ement X to evoke its CR than would the more salient B the CS. When tested with BX and CX, the animals again stimulus, producing the result obtained.
consumed less of CX than of BX. This is the result to be The interpretation offered by Hall (2003) implies the expected if the test performance depends on a difference existence of a novel (or, at least, nonassociative) learn- between B and C in their ability to interfere with the ex- ing process that modulates the perceptual effectiveness pression of the CR governed by X; the more salient B of stimuli. But before we commit ourselves to this con- will be more effective in this regard than the less salient clusion, it is important to be sure that the effects on C, both when X has been conditioned in compound with which it is based cannot be explained in terms of stan- A, as in the previous experiment, and when X has been dard associative mechanisms. And, as Hall (2003) has conditioned alone. McLaren and Mackintosh (2000), acknowledged, the associative account of perceptual however, predict that there should be no difference on learning proposed by McLaren and Mackintosh (2000; test when X has been conditioned alone. According to see also McLaren, Kaye, & Mackintosh, 1989) can sup- their account, BX differs from CX in that B is able to in- ply an explanation for the results reported by Blair and hibit the A representation whereas C is not; but this dif- ference will be relevant only when the conditioning pro- McLaren and Mackintosh (2000) point out that expo- cedure is one that has endowed A with some associative sure to a compound stimulus will result in the formation strength (as when conditioning is given with AX); it of excitatory associations among the elements of the should not affect performance when conditioning has compound (Rescorla & Cunningham, 1978). The pre- exposure procedure used by Blair and Hall (2003) can Given that the results of Blair and Hall's (2003) Ex- thus be expected to establish associations between A and periment 5 appear to have theoretically important impli- X, B and X, and C and X. That between A and X could cations, we attempted, in the experiments to be reported contribute to the magnitude of the CR observed on the here, to confirm and extend them. In the present Experi- test trial with CX. Conditioning with AX as the CS will ment 1, we adopted essentially the same design and pro- endow both A and X with associative strength, and the cedure, but we changed the nature of the flavors used; in aversion shown to CX on the test will no doubt be deter- particular, we used saline rather than quinine as the X el- mined largely by the strength of the conditioned X ele- ement. To anticipate, we succeeded in replicating the re- ment. But although A is not presented on the test, its sult obtained by Blair and Hall in their Experiment 5, thus strength could still play a role, because the existence of extending the generality of the effect. More important, the A-X association would allow the X element of the the outcome of Experiment 1 formed the basis for a fur- test stimulus to contact the US representation by way of ther investigation of the phenomenon. According to the the chain X-A-US. Such a process can be expected to interpretation offered by Hall (2003), the difference be- operate when CX is the test stimulus, but when BX is the tween BX and CX on test derives from a difference in the test stimulus another factor must be taken into account.
ability of B and C to interfere with the responding gov- According to standard associative theory, preexposure erned by the conditioned X element. In the experiments consisting of alternating trials with AX and BX will not described so far, X has been subjected to aversive condi- only establish excitatory associations between the ele- tioning so that the effect of presenting the salient B ele- ments of each compound, but also result in the formation ment along with X has been to attenuate the suppression of inhibitory associations between their unique features.
of consumption generated by X. But if Hall's (2003) in- In particular, once the excitatory, within-compound, terpretation is correct, B should be more effective than C A-X association has been established, presentation of in attenuating any response controlled by the X ele- BX will activate the representation of the absent A ele- ment-if, for instance, X is particularly valued by the rat, ment (by way of the X-A association), a circumstance then the presence of B in the BX compound can be ex- that, according to Wagner (1981), will result in the for- pected to lead to a reduction in the amount consumed mation of inhibitory links between B and A and between (compared with the amount of CX consumed).
X and A. The excitatory link between X and A will be Using saline as the X element allows a test of this pre- reestablished on the next AX trial, but B, being a unique diction. Injecting rats with a mixture of furosemide and predictor of the absence of A, will acquire net inhibitory deoxycorticosterone acetate (Furo-Doca) induces a state strength. As a consequence, on test trials with BX, acti- of salt need and renders them eager to consume a saline vation of A by way of the X-A link will be inhibited by solution (Cruz, Perelle, & Wolf, 1977). Animals in this the presence of B, and the associative strength acquired state should readily consume compounds such as BX by A will not be able to contribute to the CR.
and CX, which contain saline. But if alternating pre- In an attempt to test the validity of this account, we exposure renders B relatively salient (with respect to the conducted a further study (Blair & Hall, 2003, Experi- C element preexposed in a block of CX trials), then they ment 5) in which, as before, rats received exposure con- will be less ready to consume the BX compound (in sisting of alternating presentations of AX and BX and a which the presence of B will interfere with the percep- separate block of CX trials. They then received aversion tion of X) than the CX compound. This prediction was conditioning in which the X element (rather than the AX EXPERIMENT 1
were given unrestricted access to water on both drinking sessions.
The second conditioning trial, given in the morning session of the In this experiment, all subjects (rats) received an ini- next day, was identical to the first and was followed by a further re- tial preexposure phase consisting of trials with three covery day. The third conditioning trial was identical to the second.
compound flavor stimuli, AX, BX, and CX. A, B, and C Water was again available in the afternoon session following this conditioning trial, and 1 further recovery day preceded the test were commercially produced food flavorings; X was a saline solution. This was followed by aversion condi- On the following morning session, subjects were given a 30-min tioning with X as the CS. All then received a two-bottle choice test, receiving access to 30 ml of BX and to 30 ml of CX in test in which the compound BX and CX were presented.
two separate drinking tubes presented simultaneously. The two (In the experiment by Blair & Hall, 2003, on which this tubes were inserted into the cage on either side of the aperture used study was based, the test procedure consisted of a series for presentations of the single tube given during earlier stages of training. The two spouts were separated by a distance of 5 cm. The of single-bottle tests. The use of a two-bottle test in the position of the tubes was counterbalanced in such a way that half present experiment was dictated by the desire to match the rats were presented with BX on the right, and half with CX on the procedure with that to be used in Experiment 2, in the right. Water was made available for 30 min in the afternoon ses- which the use of a two-bottle test was dictated by the par- sion. Over the next 4 days, four further test sessions were given, the ticular procedures employed.) If this procedure should procedure being identical to that just described.
generate an effect equivalent to that demonstrated by Blair and Hall (2003, Experiment 5), one would expect Results and Discussion
to find that the subjects would consume more of BX than The rats consumed all of the fluid that was made avail- able on the preexposure sessions. The conditioning pro- cedure successfully established an aversion to X. Group mean consumption was 9.31 ml on Trial 1, 9.32 ml on Subjects and Apparatus. The subjects were 8 experimentally
Trial 2, and 4.08 ml on Trial 3. All subjects showed a re- naive male hooded Lister rats with a mean ad-lib weight of 367 g at duction in consumption from Trial 1 to Trial 3.
the start of the experiment. The rats were housed singly, with con- Group mean consumption scores for the test sessions tinuous access to food, in a colony room that was artificially lit are presented in Figure 1. The rats consumed rather lit- from 8:00 a.m. to 8:00 p.m. each day. Access to water was restricted tle of either solution on the first test session, but con- as detailed below. Presentations of the various solutions used as thestimuli in these experiments were given in the home cages.
sumption increased over sessions, presumably as a result The solutions used as experimental stimuli were administered at of extinction of the aversion conditioned to X. The sub- room temperature in 50-ml plastic centrifuge tubes equipped with jects consumed more of BX than of CX on each of the a rubber stopper to which was fitted a stainless steel, ball-bearing- five tests, with the difference between the means for con- tipped spout. The following solutions were used: A compound con-sisting of 0.08 M saline (NaCl) and almond (2% v/v almond flavor-ing supplied by Supercook, Leeds, U.K.); a compound of 0.08 M saline and vanilla (1% v/v Supercook vanilla flavoring) sulfate; anda compound of 0.08 M saline and peppermint (0.5% v/v Supercook peppermint flavoring). Consumption was measured by weighingthe tubes before and after trials, to the nearest 0.1 g. The US for the conditioning trials was an intraperitoneal injection of 0.3 M lithiumchloride (LiCl) at 10 ml/kg of body weight.
Procedure. A schedule of water deprivation was initiated by re-
moving the standard water bottles overnight. On each of the fol-lowing 3 days, access to water was restricted to two daily sessions of 30 min, at 11:00 a.m. and 5:00 p.m. Presentation of fluids con-tinued to be given at these times throughout the experiment.
Over the next 6 days (the preexposure phase), all subjects received four presentations of each of the three compound flavors AX, BX,and CX. Half the animals were first given 4 days of intermixed ac- cess to flavors AX and BX, with 10 ml of AX being presented dur-ing the first daily drinking session and 10 ml of BX during the sec- ond. This was followed by 2 days of blocked presentations of CX,in which 10 ml of this flavor was made available in both morningand afternoon drinking sessions. The remainder of the subjects re- ceived the blocked presentations of CX on the first 2 days of thephase followed by 4 days of AX and BX. For all animals, flavor A was peppermint and flavor X was saline. The critical test flavors, Band C, were counterbalanced with half the animals receiving vanilla Figure 1. Experiment 1: Mean consumption of flavor com-
as B and almond as C, and half receiving the reverse arrangement.
pounds BX and CX over five two-bottle test trials. The rats had
Three conditioning trials followed. The first was given in the all received preexposure consisting of a block of trials with CX
morning session the day after preexposure ended. It consisted of a and alternating trials with AX and BX, prior to aversion condi-
30-min presentation of 9 ml of X, followed immediately by an in- tioning with X. Error bars represent within-subjects standard
jection of LiCl. The rats were given free access to water in the af- error, computed on scores adjusted for variation between sub-
ternoon session. The next day was a recovery day on which animals jects (Bakeman & McArthur, 1996).
sumption of the two flavors increasing throughout. This saline. We gave rats preexposure consisting of alternating description of the results was confirmed by an analysis trials with AX and BX and a separate block of CX trials.
of variance (ANOVA) conducted on the data summa- We then gave them an injection of Furo-Doca followed rized in the figure, with test stimulus (BX or CX) and by a test session in which BX and CX were presented. We trial as the factors. There was no significant main effect expected that the rats in a state of salt need would tend to of test stimulus [F(1,7) = 2.18; here and elsewhere a sig- consume both of these compounds readily, to the extent nificance level of p <.05 was adopted], but there was a that they were able to perceive the saline component. The significant effect of trial [F(4,28) = 16.32], and, criti- question of interest was whether the supposedly more cally, a significant interaction between trial and stimulus salient B element would interfere with this tendency more [F(4,28) = 4.54]. Pairwise comparisons using Tukey's than the less salient C element would, resulting in a test revealed that the scores for the two flavors differed greater consumption of CX than of BX.
It will be noted that, in this experiment, the looked-for The data presented in the figure pool the results for pattern on the test session (more consumption of CX two subgroups, one presented with CX on the first block than of BX) was the opposite of that predicted (and of preexposure trials, the other presented with CX on the found) in Experiment 1. To find this pattern would allow second block of preexposure trials. With only 4 animals us to rule out some possible, but theoretically uninter- in each of these subgroups, our ability to detect any dif- esting, explanations for the result of that experiment.
ference between them will be limited. There was some One of these was the possibility that the preexposure indication, however, that the effect of interest (the low schedules used in these experiments were differentially level of consumption of CX on test) was more substan- effective in their ability to produce habituation. If habit- tial in the subgroup given CX in the second block. Pool- uation occurs more readily with the alternating schedule ing over all test trials showed that this group drank a than with the blocked arrangement, then the BX compound mean of 5.7 ml of BX and 1.6 ml of CX; the equivalent would be less likely to evoke neophobia than would the scores for the group given CX first were 6.0 ml and CX compound, resulting in greater consumption of BX 3.8 ml. This difference between the subgroups was not, than CX-the result obtained in Experiment 1. But this however, statistically reliable. An ANOVA paralleling account cannot predict greater consumption of CX than that just described, but including subgroup as a variable, of BX, the result anticipated in this experiment.
yielded neither a significant main effect of this variable An alternative interpretation of the results of Experi- [F(1,6) = 1.49] nor any significant interaction-for the ment 1 can be derived from the fact that in that experiment, subgroup 3 stimulus interaction, F < 1; for the triple presentations of BX were always given during afternoon sessions in preexposure, whereas CX also occurred dur- Although this experiment made use of different fla- ing morning sessions (but see Blair & Hall, 2003, Ex- vors as the stimuli and a different test procedure, its re- periment 1B). Aversion conditioning to X occurred dur- sults were entirely consistent with those reported by ing a morning session. Accordingly, if time of day can Blair and Hall (2003, Experiment 5). Rats that had re- serve as a cue that mediates generalization, it might be ceived aversion conditioning with flavor X drank more argued that generalization would be greater to CX than of it on test when it was presented in compound with B to BX, producing the effect seen in Experiment 1. But than when it was presented in compound with C. This this argument does not apply to the present experiment- outcome is thus in accord with the suggestion that the given that the procedure does not involve a conditioning schedule of preexposure used in the first phase of the ex- phase, there is no basis on which the time-of-day factor periment rendered stimulus B effectively more salient can play a role and therefore no reason to predict any dif- than stimulus C and thus better able to disrupt the ex- ference in consumption between CX and BX on the test.
pression of the aversion governed by X.
Finally, it could be argued that the results of Experi- ment 1, in which the difference between BX and CX de- EXPERIMENT 2
veloped only over the course of several nonreinforced test sessions, might be a consequence of a difference be- Our proposal that the results of Experiment 1 reflect tween the two test stimuli in the rate at which extinction differences in salience between B and C depends on the occurs. Given that the aversion shown on test will be (surely plausible) assumption that a more salient stimulus largely a consequence of the associative strength gov- will be more effective than a less salient stimulus in in- erned by a stimulus element, X, that is common to both terfering with the response elicited by some other stimu- test compounds, it is not clear how such a difference lus with which it is compounded. In Experiment 1, this might arise. Nonetheless, it is worth noting that differ- response was the aversion to X established by an aversive ential extinction could not easily explain the anticipated conditioning procedure in which X served as the CS. But results of the present experiment, in which the test con- if our interpretation is correct, B and C should differ in sisted of a single presentation of BX and CX.
their ability to modulate any response controlled by X. In the present experiment, we made use of the fact that X was a solution of saline and that an injection of Furo- The subjects were 16 experimentally naive male hooded Lister Doca would increase the readiness of rats to consume rats with a mean ad-lib weight of 352 g at the start of the experi- ment. The experiment was run in two replications, each using 8 an- ference between the groups was quite marked. The ef- imals. In the first replication, 1 animal became ill prior to the test fect shown in the figure was almost entirely a conse- phase so that data were available for 15 subjects. The flavors used quence of the behavior of the subgroup (n = 8) given CX as stimuli were those described for Experiment 1. The treatment first. This group consumed 22.6 ml of CX and 7.1 ml of used to induce a sodium appetite was a subcutaneous injection of 0.5 ml of a mixture of 10 mg furosemide (Furo) and 5 mg of de- BX. The subgroup given CX second (n = 7) consumed oxycorticosterone acetate (Doca) dispersed in 20 ml of distilled rather less overall, and the difference between the scores for BX (11.2 ml) and CX (11.9 ml) was negligible. An The subjects initially received exposure to AX, BX, and CX, the ANOVA conducted on these data revealed a significant procedure being identical to that described for the preexposure main effect of test stimulus [BX or CX; F(1,13) = 6.25] phase of Experiment 1. One hour after the end of the final pre- and no significant main effect of subgroup [F(1,13) = exposure session (the afternoon session on Day 6 of preexposure), all animals received an injection of Furo-Doca. The food was then 3.76], but there was a significant interaction between removed from the home cages in the colony room, and the subjects these variables [F(1,13) = 5.24]. Analysis of simple ef- were given free access to distilled water overnight. On the follow- fects showed that the scores for BX and CX differed sig- ing day, the distilled water was removed from the cages 3 h prior to nificantly in the subgroup given CX first [F(1,13) = the test session (the morning drinking session). On test, the subjects 12.29], but not in the subgroup given CX second (F < 1).
were given a free-access choice test, receiving 30 ml of flavor BX The theoretical implications of this difference be- and an identical volume of flavor CX, in the same manner as in Ex- tween the subgroups will be taken up in the General Dis- periment 1. Only one such test was given (the effects of the Furo- cussion. For the time being, we will simply note that the overall pattern of the results (as shown in Figure 2) ac- Results and Discussion
cords with the interpretation that we offered for the re- The rats consumed all of the fluid made available on sults of Experiment 1. In that experiment, the stimulus each of preexposure sessions, with no evidence of neo- element X was made aversive and the animals consumed less of CX than of BX on test, a result we interpreted as The results of the test session, group means for con- showing that B was more effective than C in modulating sumption of BX and CX, are presented in Figure 2. It the response controlled by X. In this experiment, stimu- shows that subjects consumed more of CX than of BX.
lus element X was made desirable by the administration A within-subjects, one-way ANOVA confirmed that the of Furo-Doca just after preexposure. The resulting salt difference between the two test stimuli was statistically need meant that when given a choice, subjects were mo- tivated to consume the flavor compound that they per- As in the previous experiment, we looked separately at ceived as being richer in salt. Our prediction, therefore, the performance of the two counterbalanced subgroups, was that the perceptually salient B element would be which were given CX presentations as the first or as the more likely to interfere with perception of the desirable second block of preexposure trials. In this case, the dif- salt than would the less salient C element, leading to greater consumption of CX than of BX-the result that As we pointed out in the introduction to this experi- ment, the finding of greater consumption of CX than of BX in the present experiment rules out some, relatively trivial, explanations that might be offered for the results of Experiment 1. It also allows us to address a possible explanation that can be derived from an elaboration of the theory proposed by McLaren et al. (1989). Accord- ing to that theory, it will be recalled, the preexposure schedule employed in these experiments will result in the formation of excitatory links between X and each of A, B, and C; it will also allow the formation of inhibitory links between A and B. When animals are conditioned with X (as in Experiment 1), the excitatory X- C link should generate activity in the representation of stimulus C; but the existence of inhibitory A- B links might limit the extent to which the representations of these stimuli can be activated. Finally, if it is accepted that the asso- Figure 2. Experiment 2: Mean consumption of flavor com-
ciatively activated representation of a stimulus can un- pounds BX and CX on the choice test. The rats had all received
dergo conditioning (see Hall, 1996), it follows that C preexposure consisting of a block of trials with CX and alternat-
will be more likely than A or B to acquire associative ing trials with AX and BX, followed by the induction of a salt need
strength during reinforced X trials. Differences in the ac- immediately prior to the test. Error bars represent within-subjects
standard error, computed on scores adjusted for variation be-
quisition of strength by C could then explain why the an- tween subjects (Bakeman & McArthur, 1996).
imals showed a greater aversion to CX than to BX in Ex- periment 1. Resting, as it does, on a number of debatable exposure phase of these experiments will also undergo assumptions, this explanation may seem implausible, but habituation and thus become less effective at eliciting there is nothing in the results of Experiment 1 to rule it whatever unobserved covert response they would nor- out entirely. It cannot apply to the results of the present mally evoke in the animal; it is this loss of effectiveness experiment, however, since the mechanism proposed de- that we have equated with a change in stimulus salience.
pends critically on events occurring during the condi- The real problem is to explain why this loss of effective- tioning trials with X. The central feature of this experi- ness should be attenuated or reversed for A and B when ment was that the value of X was modified by means of preexposure consists of alternating trials with AX and a procedure that did not involve presentations of X.
According to Gibson's (1969) account of perceptual GENERAL DISCUSSION
learning, exposure to similar stimuli will engage a dif- ferentiation process that will enhance the perceptual ef- Taken together with the results reported by Blair and fectiveness of unique stimulus features. Differentiation Hall (2003), the experiments described here allow the is held to occur more readily when the subject has a following conclusions. Preexposure to the compound chance to compare the relevant stimuli. It might be sug- stimuli AX, BX, and CX will, among other things, mod- gested, then, that alternating preexposure is particularly ify the properties of their unique features, A, B, and C.
effective in maintaining or strengthening the effective- When the preexposure consists of alternating trials with ness of A and B because these stimuli are presented ac- AX and BX, and a separate block of CX trials, the con- cording to a schedule that promotes stimulus compari- sequence appears to be that the effective salience of the son. The obvious problem with this suggestion is that B element (and presumably also of the A element, al- although AX and BX were indeed presented in alterna- though this was not tested directly) is greater than that of tion in these experiments, the interval between succes- the C element. This difference is evidenced by the effects sive trials was at least 5 h, making it unlikely that any ac- of compounding B and C with another element (X). When tivation produced by presentation of the first stimulus X is aversive, compounding it with B results in less of an would still be ongoing when the second stimulus oc- aversion than does compounding it with C. When X has curred. What does seem possible, however, is that the a positive value, compounding it with B results in less central representation of each of the critical stimulus fea- being consumed than when it is compounded with C.
tures will be activated associatively on each trial (after This pattern of results is parsimoniously explained in the first) in the alternating schedule. Alternating trials terms of the proposal that the more salient B is more with AX and BX can be expected to establish and main- likely to interfere with the ability of X to evoke its re- tain excitatory connections between X and A and be- tween X and B, allowing A to be activated associatively The proposal that alternating exposure to two similar on BX trials and B to be activated associatively on AX stimuli will enhance the salience of their unique ele- trials. Hall (2003) has advanced the hypothesis that the ments (or at least results in a lesser loss than is produced associative activation of a stimulus representation in the by repeated presentation of just one stimulus) has the ca- absence of direct presentation of the stimulus itself will pability to explain a wide range of perceptual learning act to reverse the loss of salience that occurs when the effects. The central feature of these effects (for a recent review, see Hall, 2001) is that appropriately scheduled Comparison of the counterbalanced subgroups of the exposure to similar stimuli can increase the ease with present experiments (those given CX preexposure in the which they can subsequently be discriminated. The first block vs. those given CX preexposure in the second essence of discrimination learning is that behavior must block) allows a test of this hypothesis. For both sub- come to be controlled by the unique, distinguishing fea- groups, the alternating schedule used for presentations tures of the stimuli rather than by features they hold in of AX and BX ensures that associative activation of the common. A learning process that enhances the (relative) representation of B will occur regularly on AX trials.
salience of the unique features of similar stimuli would According to the hypothesis, the salience of B should thus be capable of generating the observed effects.
therefore be maintained or enhanced in both. The fate of It remains to explain the nature of the learning process the C element will, however, differ in the two subgroups.
responsible for changes in stimulus salience, and here we When CX is presented second during preexposure, there have little more than speculation to offer. That repeated will be no opportunity for the associative activation of C exposure to a given stimulus (such as CX in the present in the absence of the stimulus, and its decline in salience experiments) might cause a loss of salience (in both C should proceed unhindered. But when CX is presented and X) is not problematic. The phenomenon of habitua- first, the formation of the within-compound C-X asso- tion demonstrates that repeated presentation of a stimu- ciation during the first block of trials will mean that C lus that initially evokes an overt response will reduce the can be activated associatively on the subsequent AX and effectiveness of the stimulus in this respect. Although BX trials (at least until the point at which these trials re- they do not evoke any obvious overt response, it is rea- sult in extinction of the X- C association). C should sonable to assume that the stimuli presented in the pre- therefore lose salience less readily in the latter condition.
It follows that the difference in effectiveness between B sults are challenging for the hypothesis offered by Hall and C on the test should be greater in the subgroup that (2003), but, importantly for our present concern, the received CX during the second block of preexposure tri- basic pattern obtained is not to be easily explained by es- als than in the subgroup that received CX in the first tablished associative theories of perceptual learning as Unfortunately, the results of the present experiments are equivocal on this issue. In Experiment 1, the means REFERENCES
for the two subgroups were entirely in accord with our prediction; the subgroup given CX first in preexposure Bakeman, R., & McArthur, D. (1996). Picturing repeated measures: Comments on Loftus, Morrison, and others. Behavior Research drank more of CX on the test than did the subgroup given Methods, Instruments, & Computers, 28, 584-589.
CX second, implying that the salience of C was greater Blair, C. A. J., & Hall, G. (2003). Perceptual learning in flavor aver- in the former group than in the latter. That the difference sion: Evidence for learned changes in stimulus effectivenes between the subgroups was small (and not statistically Cruz, C. E., Perelle, I. B., & Wolf, G. (1977). Methodological as- reliable) would also be expected, given that the CX-first arrangement supplies conditions that are less than opti- mal for the associative activation of C (and thus for pro- Gibson, E. J. (1969). Principles of perceptual learning and develop- ducing the increase in effective salience envisaged by ment. New York: Appleton-Century-Crofts.
our interpretation). In this subgroup, C may be activated Hall, G. (1996). Learning about associatively activated stimulus rep- resentations: Implications for acquired equivalence and perceptual associatively during the initial presentations of AX and BX that follow exposure to CX, but, in the absence of Hall, G. (2001). Perceptual learning: Association and differentiation.
further CX trials, it is to be expected that extinction of In R. R. Mowrer & S. B. Klein (Eds.), Handbook of contemporary the within-compound association between X and C will learning theories (pp. 367-407). Mahwah, NJ: Erlbaum.
Hall, G. (2003). Learned changes in the sensitivity of stimulus repre- limit this effect just to these initial trials. This would ex- sentations: Associative and nonassociative mechanisms. plain why consumption of CX was only slightly higher in the subgroup for which CX was presented first, and why, McLaren, I. P. L., Kaye, H., & Mackintosh, N. J. (1989). An asso- overall, the means indicate a clear perceptual learning ciative theory of the representation of stimuli: Applications to per- effect (significantly higher consumption of BX than CX) ceptual learning and latent inhibition. In R. G. M. Morris (Ed.), Par-allel distributed processing: Implications for psychology and neurobiology (pp. 102-130). Oxford: Oxford University Press, The results for the equivalent analysis of Experiment 2 are more problematic. Here the subgroup given CX first McLaren, I. P. L., & Mackintosh, N. J. (2000). An elemental model (but not that given CX second) showed a profound dif- of associative learning: I. Latent inhibition and perceptual learning.
Animal Learning & Behavior, 28, 211-246.
ference in consumption of the two test solutions-the re- Rescorla, R. A., & Cunningham, C. L. (1978). Within-compound fla- verse of the effect predicted (and obtained in Experi- ment 1). We can see no obvious explanation for this outcome, and we will need to carry out further work to Wagner, A. R. (1981). SOP: A model of automatic memory processing confirm its reliability. However this may turn out, it is in animal behavior. In N. E. Spear & R. R. Miller (Eds.), Information processing in animals: Memory mechanisms (pp. 5-47). Hillsdale, encouraging that the overall findings of Experiment 2 are largely in accord with conclusions drawn from a body of work that has, until now, relied on the condi- tioned flavor aversion procedure. Some aspects of the re- revision accepted for publication January 22, 2003.)

Source: http://www-users.york.ac.uk/~gh1/pdf/2003Blair&Hall.pdf