An Analysis of Altering the Value of a Classically Conditioned Predictive Cue

Theoretically , predictive cues have a maJor role in conditioning. Both conditional stimuli and operant responses function in a way which provide information to a subject about what is to occur under a particular set of circumstances . Any available cue will have a particular relationship with any given outcome stimulus (S0 ) based on the probabilities of cooccurrences and non co-occurrences, whereby the resulting predictive value will impact on the behavioral consequences. A stimulus can produce behavioral effects it never caused prior to conditioning. Are those behavioral effects the result of the predictive value of the stimulus alone, or does the motivational value of the stimulus change? Could the resulting behavior be based on some combination of both? The value both incentive value as well as predictive value of the predictive cue itself is of significance. Knowing about the manner in which cues are affected will have an effect on how they are used, how they will affect behavior and how they function in applied settings. Despite an abundance of valuable information pertaining to such events, several crucial issues remain to be examined . First, exactly how does the predictive value of the predictive cue influence behavior? Cues which are highly reliable should influence behavior in a very systematic manner. Depending upon the motivational value of the outcome predicted, a con sistent approach or withdrawal behavior should be exhibited. Additionally, cues which are unreliable should have less systematic influence on behavior. Since such cues provide no reliable information regarding outcomes, no reliable behavioral consequences should be observed. The present research was designed to separate the predictive and incentive values of the predictive cue itself by examining the behavioral consequences of altering the incentive value of outcome stimuli on the incentive value of the cues that predict them . This design was employed to examine if the relative tendency of the animal to approach or withdraw from any particular predictor changed when the relative tendency to approach or withdraw from the outcome stimulus was altered . The questions specifically investigated were: in regard to a classically conditioned behavioral effect, to what extent does a stimulus retain or take on the initial value of the stimulus it predicts, and to what extent is it based upon the current (altered) value of the stimulus it predicts? The design employed provides a measure of the motivational value of predictive cues relative to the current motivational value of the outcome stimulus. In all circumstances it is the value of the outcome stimulus that was manipulated, while responding to or for the predictive cue (S2) was measured. Thus, the behavioral consequences of the value of outcome stimuli on the value of the predictive cues was assessed. Subjects were 64 male Sprague-Dawley albino rats weighing 250 350 gm. All subjects were hungry and thirsty throughout the entire experiment. Subjects were randomly assigned to one of eight groups (eight subjects per group) and run through the sequence of the experiment as determined by the particular group . Group membership is indicated by a three unit code pertaining to the Sensory Preconditioning Configuration Sensory Preconditioning Stimulus Presentation Revaluation phase sequences. For example, subjects in group APL received Sensory Preconditioning Configuration A, Stimulus Presentation P, and Revaluation L. As a result of the above manipulations, the following should be true for both Sensory Preconditioning Configurations: 1) The RN group has no predictive value and provides information regarding the original , unaltered motivational value of S1 and S2 . 2) The RL group has no predictive value and provides information regarding the original, unaltered motivational value of S2 and information regarding the devalued motivational value of S 1. 3) Group PN evidences the ·predictive value of S2 and the inherent , unaltered motivational value of S 1. 4) The PL group provides information regarding the motivational value of S2 after St has been devalued , as well as the devalued motivational value of S1 and the predictive value of S2. A 2x2x2 Multivariate Analysis of Variance (MANOVA) with eight dependent measures (T1P, T1S2, T2S1, T2H2O, T3S2, T3H2O, T4S1, T4S2) was conducted . The MANOVA produced the following results : Configuration (C), F(8,47) = 22.45 , p<.001; Stimulus Presentation (StP), F(8,47) =0.61, n.s.; Revaluation, (R) F(8,4 7) = 19.54, p<.001; Configuration x Stimulus Presentation , F(8,47) =0.8, n.s.; Configuration x Revaluation, F(8,47) = 10.36, p<.001; Stimulus Presentation x Revaluation, F(8,47) =0.93, n.s.; Configuration x Stimulus Presentation x Revaluation, F(8,47) = 1.32, n.s. Follow-up ANOVAs and Tukeys revealed an unexpected pattern of results. The data suggests that illness is what produced the quinine-water discrimination , but that prior to experiencing the salient contingent event of illness to quinine, water and quinine were not responded to differentially. Such a finding has strong implications for the Sensory Preconditioning phase. Since all animals had access to water in their home cages daily during all phases (except Revaluation and Testing) , and were exposed to water in the training apparatus during Acclimation , animals (both Paired and Random) experienced presentations of saccharin and water in a random fashion. In essence , there was no contingent relationship in place for the Paired subjects and results indicate that all subjects responded as if no learning occurred dur ing Sensory Preconditioning. Consequently , no predictive relationship was established so habituation to the taste of saccharin took place. Also, as a result of the absence of this predictive relationship , none of the differential experimental conditions were in place and subjects were unable to predict either the original or altered value of the CS.

which are unreliable should have less systematic influence on behavior.
Since such cues provide no reliable information regarding outcomes, no reliable behavioral consequences should be observed. The present research was designed to separate the predictive and incentive values of the predictive cue itself by examining the behavioral consequences of altering the incentive value of outcome stimuli on the incentive value of the cues that predict them . This design was employed to examine if the relative tendency of the animal to approach or withdraw from any particular predictor changed when the relative tendency to approach or withdraw from the outcome stimulus was altered . The questions specifically  (T1P, T1S2, T2S1, T2H2O, T3S2, T3H2O, T4S1, T4S2) was conducted . The MANOVA produced the following results : Configuration ( Tukeys revealed an unexpected pattern of results. The data suggests that illness is what produced the quinine-water discrimination , but that prior to experiencing the salient contingent event of illness to quinine, water and quinine were not responded to differentially. Such a finding has strong implications for the Sensory Preconditioning phase. Since all animals had access to water in their home cages daily during all phases (except Revaluation and Testing) , and were exposed to water in the training apparatus during Acclimation , animals (both Paired and Random) experienced presentations of saccharin and water in a random fashion. In essence , there was no contingent relationship in place for the Paired subjects and results indicate that all subjects responded as if no learning occurred dur ing Sensory Preconditioning.
Consequently , no predictive relationship was established so habituation to the taste of saccharin took place . Also, as a result of the absence of this predictive relationship , none of the differential experimental conditions were in place and subjects were unable to predict either the original or altered value of the CS.
goodness!) when I needed help running animals . They are my family and my friends. They are respected, appreciated, and deeply loved.
My friends Tony Riccitelli and Sandy Smith have endured the loss of their friend from their lives over the past year and a half, and the entrance of a shadow, that occasionally showed up and expected everything to be the same. And it was! Tony has been supportive in every possible sense. He has always given me encouragement, praise, his keen critical analysis, and especially his loyal, enduring, and loving friendship. Sandy has perhaps felt my absence more than anyone yet she has remained my caring and loving friend. She has consisten tly kept me in touch with my emotional and spiritual side, regardless of how much I try to set it aside. I hope that I can be the friend to each of them that they are to me.
During the running of this project, I have been a fixture in the halls of Chafee . My fellow graduate students -my friends -Jim Arruda, Susan Curley, and Sherri Gold shared the same area of the building and became my direct and immediate support system . They understood! They were always there (as tends to be the habit with grad students) to offer me encouragement, listen to me complain (which I did -and still do -often), endure my days of bothersome chatter (that happens when you spend eight hours a day alone in the dark) and to remind me of the light up ahead .
Sue Curley is simply the most positive person I know. She always has a smile and kind word. Her warmth , genuineness, and clear thinking was (and remains) a bright spot in any day I encounter her.
In addition, Sue was kind enough to use her artist ic abilities to draw a figure for this document for me.
Sherri is someone I have come to know and appreciate in a slow but secure fashion. Sherri is yet another friend who helped make those days of running research bearable ; she spent many afternoons sitting with me in the dark . Her sincere, honest manner was a comfort. To Dr. Silverstein, and each of you, my most sincere thanks . (S1 /(S1 + H20)) on Test Day 2 2x2 ANOV A Source Table for Test 2: Suppression Ratio (S1 /(S1 + H20)) 2x2x2 ANOVA Source Association Presidential Address, in American Psychologist ) started a new discussion of Pavlovian conditioning in terms of " ... the learning of relations among events so as to allow the organism to represent its environment." (p.151). In essence, Rescorla was introducing a new type of cognitive view of conditioning based on "information" though, it is a view which has theoretical roots m the work of E.C. Tolman (e.g., Tolman, 1932).
This new view of conditioning incorporates traditional empirical findings with some more recent results that Rescorla and his colleagues Rescorla, , 1985 have produced .      (1987,1988 ) work is that in both operant and Pavlovian learning, the animal learns to predict the probable occurrence or non-occurrence of an outcome stimulus (SO), either an operant reinforcer /punisher or a Pavlovian unconditioned stimulus , which has a particular motivational valuation (usually a strong one ). In the Pavlovian case it is a signal, the conditional stimulus, that provides the animal with the predictive cue . In the operant case it is the occurrence of the operant re sponse clas s (R), that provide s the subject with the predictive cue .
Theoretically, since the reliability of the relationship between a predictive cue and an outcome establishes predictive value, one of the factor s that help s to guide behavior is the predictive value of the available cues. Any available cue will have a particular relationship with any given S 0 based on the probabilities of co-occurrences and non co-occurrences whereby the resulting predictive value will impact on the behavioral consequences.
In addition, one might infer from the above stated premise that in both operant and Pavlovian conditioning the animal may also be learning to change the valuation of the predictive cues. That is, the cue may change from some relatively neutral initial value prior to conditioning to an incentive or aversive value based, in some way, on the outcome it now predicts. If presentation of the predictive cue has behavioral consequences that it did not have prior to conditioning (predicting a particular so with a particular motivational value) then perhaps the motivational value of the predictive cue has changed along with its predictive value. If so, it is possible that whenever the motivational value of an outcome stimulus is altered by any procedure, the motivational valuation of the predictive cue may also change . Such a change in valuation could be demonstrated by appropriate alterations in the animal's behavior.
In the Pavlovian situation the appropriate alteration would be based upon a change in the probability of some approach or withdrawal response. In the operant situation the appropriate alteration would be indicated by a change in the probability of the operant response class itself .
Researchers have addressed "what is learned" in a multitude of ways over the years by probing to find general laws of learning.
The new cognitive perspective is actually a reorganization of some not so new information , but it is different from the "old" cognitive views because it is more empirically grounded . Earlier cognitive views of conditioning ) met with much criticism principally due to the lack of clearly specific empirical support rather than because they were inadequate theories.
Work regarding "what is learned" which provides support for a cognitive perspective of learning can be found in some of the most innovative and valuable contributions to the body of literature regarding learning. Leon  research on "the blocking effect", and the Rescorla/Wagner model of conditioning  examine how the association of stimuli (CSs and USs) takes place (through autoshaping).
Both , and , set out to investigate a variation of the "US -reduction hypothesis" (Domjan & Burkard, 1982, p. 95), an hypothesis that viewed Pavlovian conditioning as a US's ability to produce conditioning. Initially, a US has its full potential of conditionabi lity, but as conditioning trials progress (the CS is paired with the US) it seem to have less additional conditionability potential.   This came to be known as "the blocking effect". Further research on "the blocking effect" shed light on circumstances which enhanced conditioning as well as blocked it.  found that blocking did not occur in situations when the second stimulus was not redundant and/or where it provided new or "surprising " information.
What Kamin's research (1968Kamin's research ( ,1969 and the Rescorla/Wagner model  indicate goes beyond mere conditionability of a particular response and the blocking effect. The notion of "surprisingness" has to do with contingency and predictability . Kami n's (1968Kami n's ( , 1969) and Rescorla's and Wagner's     , revealed that cues which were redundant yet informative also produced effective learnin g.
To more clearly examine this "informational" aspect of cues, Cohen , Cali sto, and Lentz (1979) successfully attempted to separate the informational component of a contingency from the reinforcing component of it. Using chained schedules and secondary reinforcers, the se researchers found that stimuli were only effective at maintaining behavior (a nd could serve as secondary reinforcers) when they were informational , otherwise they needed to be tied to primary reinforcers in order to maintain behavior over a long chain of reinforcement schedules.  was one of the earliest and most insightful proponents of the need for a more empirical orientation to the possibility of the cognitive/informational perspective.
In an analytically challenging article entitled, ' "What is learned?" -an empirical enigma',   There are two basic views that had been considered: 1) that a special connection, of some sort, is established between the CS and a response (commonly known as an S-R association) such that the CS, after conditioning, brings about the CR . Thus, the CS is no longer affected by the current 9 motivational value of the US, or 2) that a "functional equivalence" is established between the CS and the US, such that the CS operates as a signal for the US and is always influenced by the motivational value of the US.
The S-R perspective on "what is learned" need not consider the motivational value of the US in regard to the CS since it is the particular response being conditioned that is crucial. The US is of significance only in the respect that it produces the UR, which is what the CS gets associated with , and why the CR subsequently occurs. Once the S-R association is established, it is only affected by new S-R associations. Changes in the CR are produced by the CS being associated with a new UR -as elicited by a US of a different motivational value. But the original S-R association is not affected by the current valuation of the US. The CS simply "evokes" the particular R.
The S-R perspective differs significantly from a "functional equivalence" view. Proponents of a functional equivalence approach maintain that the CS comes to function in the same way that the US does. Functional equivalence means that a CS takes on the motivational value of the US and, for that reason, comes to "evoke" the same response as the US  . Functional equivalence suggests that "what is learned" is an equivalent relationship between the CS and US, so that the valence of the CS is based on the current · motivational value of the US . A more recent version of the functional equivalence hypothesis based on the work of Rescorla and his colleagues Rescorla, , 1982, maintains that a "functional equivalence" between the CS and US is due to the contingent relationship between them which makes the CS a reliable and informative predictor of the US. should occur. Results indicated that when S1 had been devalued (by making subjects ill after they taste it) , a decrease in responding (consumption) to S2 (the CS) occurred. Colwill and Rescorla (1985)  The classical conditioning paradigm of sensory preconditioning involves an initial pha se of presenting two (relatively neutral) stimuli (S2 and S 1) in a reliable and predictive manner (i.e., S2 -S 1 ) (Brogden,1939;. Subsequent to this pairing, S1 is used in a typica l Pavlovian arrangement such that S 1 becomes a CS predictive of a US with a particular motivational value (S 1 -US) and eventually produces the CR. The S 2 and S 1 relationship is then examined in a third phase by evaluating the behavioral consequences of S2 (S2 -?). If S2 produces the CR, then S 2 and S 1 have been preconditioned and a predictive relationship has been established.
That is , learning took place between two relatively neutral stimuli with no overt behavioral consequences observed .

19
Taste aversion learning is a phenomenon that demonstrates a "prepared" response  in classical conditioning.
Prepared responses are responses that are learned more quickly, with fewer errors, and extinguish more slowly than other learned responses . A conditioned taste aversion (CTA) is an avoidance response to a particular taste/flavor that has been associated with gastric illness . Taste aversions are learned quickly -very often in one trial -can be learned with unusually long duration s between the taste and illness interval (interstimulus interval), are affected by familiarity of the stimuli, and especially the interoc eptive/exteroceptive nature of the stimuli (Bond & DiGuisto,1976; ). In an attempt to assess the motivational value of predictive cues after altering the motivational value (devaluing) of the outcome stimuli they predict , a classical conditioning procedure was employed.
During the Revaluation phase subjects were exposed to conditions which was designed either to devalue the outcome stimulus or In addition, there was a manipulation of Stimulus Configuration during the Sensory Preconditioning phase.
In the present study, S2 and S1 were both gustatory stimuli: a saccharin solution and a quinine solution. S2 and S1 solution configurations were manipulated as an independent variable; that is, half of the groups received saccharin as SI and quinine as S2 (Configuration  A) while the remainder received quinine as SI and as saccharin S2 (Configuration B) . This was incorporated into the design in order to allow ge neralization of the stimuli to be examined. This manipulation of Stimulus Configuration examines the possibility that the initial motivational values of CSs and USs may affect learning. Also, it provides information regarding the precise S2 and SI value configuration since saccharin and quinine are not equally desirable pre -exper iment ally. That is, with initial value configurations of "greate r to lesser " and "lesser to greater" .
The Test phases were used to exa mine whether th e value of th e predictive cue had chan ged. Testing probed whether the post-Revaluation value of the predictive cue was based on 1) the initial value of the outcome stimulus (i.e., it was purely predictive value that was learned), 2) the current (revalued) value of the outcome stimulus (i.e., it was purely motivational value that was learned) or 3) some combination of predictive and motivational value. In addition, assessment of original motivational value of the CSs and USs employed could be established as well as any alteration in these values after Sensory Preconditioning without devaluation of S1.
The most crucial test, of the several employed, assessed the motivational value of the predictive cue by presenting it contingently as outcome for a response class whose topography was different from anything subjects did during the conditioning phase.
In such a test, the predictive cue is presented to the animal if, and only if, it performs the operant. Such a pattern of presentation of stimuli parallels a conditioned reinforcement or conditioned punishment paradigm.
In the conditioned punishment paradigm, the predictive cue, which predicts a newly devalued outcome, is presented contingent upon the response, while in the conditioned reinforcement paradigm , the predictive cue, which predicts a newly positively valued outcome, is presented contingent upon the operant. If the animal does not perform the operant it receives no presentations of the stimulus. Responding in this situation is reflective of the current motivational value of the predictive cue.
Thus for predictive cues with sos which have been devalued, a decrease in responding and/or a more rapid extinction should be observed than with control groups; whereas such a decrease should not be observed to cues which have not been devalued.
An overall schematic representation of the experimental design and the differential treatments in each group can be found in Table I. Such a design facilitates a more direct examination of the predictive cues and presents more parsimonious information as well as eliminates alternate explanations for the results.
INSERT  As a result of the above manipulations, the following should be true for both Configuration s A and B : 1) The RN group has no predictive value for S2 and provides information regarding the original, unaltered motivational value of SJ and S2 .
2) The RL group has no predictive value for S2 and provides information regarding the inherent, unaltered motivational value of S2 and information regarding the devalued motivational value of SI .
3) Group PN reveals the predictive value of S2 and the inherent, unaltered motivational value of SI .
4) The PL group provides information regarding the motivational value of S2 after SJ has been devalued, as well as the devalued motivational value of S 1 and the predictive value of S2.  The three fluid tastes dispensed were tap water, a .00006-M quinine monohydrochloride solution , and a .15% w/v sodium saccharin solution  . Both the quinine and saccharin solutions were made with bottled Natural Spring water.
All injections were ip inject ions using a 25 x 5/8 gage needle and a 1.0 cc siringe. Injections were either a .9% w/v sod ium chloride (NaCl) solution (saline)  or a 3.0 M lith ium chloride (LiCI)   During all phases of the experiment , any time an animal was in the operant chamber being run though a particular sequence , the house light and fan in the apparatus were on.
Acclimat ion . All groups were exposed to this phase in Preconditioning Configuration A, in which S1 and S2 were saccharin and quinine , respectively , while the other half of the subjects received Sensory Precondition ing Conf iguration B in which S1 and S2 were quinine and saccharin , respectively . Subjects in these groups (Configuration A and B) differed only in the solution assignment (S1 or S2) . All other manipulations during this phase occurred in the same fashion to both subjects experiencing Configuration A and Configuration B.
During this phase, all subjects received five sessions (one session per day for five consecutive days) with 20 presentations of S2 and 20 presentations of S1 in each session. Half of the subjects in each configuration group were exposed to 20 paired (P) presentations of S2 and S1 (S2-S 1 ), while the remainder were exposed to 20 truly random presentations of S2 and 20 truly random presentations S1 (S2/S1 ). Exposure indicates that the animals were in the operant chamber while stimuli were presented .
Presentations consisted of the appropriate taste fluid spout being introduced into the apparatus and being available for consumption.
For subjects in the P groups , a 1 0 s presentation of S2 was followed by a 1 0 s presentation of S1 with a 20 s interstimulus interval (ISi) and a 50 s intertrial interval (ITI) for an overall trial interval of 90 s. Subjects in the R groups were exposed to 10 s presentations of S2 and 10 s presentations of S1 in a truly random fashion, with ISi and ITI ranges of 20 -50 s, such that the overall trial interval was also maintained at 90 s. The amounts of S1 and S2 consumed by each subject during each session was recorded .

Revaluation . Two days after the completion of the Sensory
Preconditioning phase, Revaluation began. Revaluation consisted of one 20 min session per day for three consecutive days. All subjects were given 20 min free access to their respective S1 solution in the operant chamber with the fluid dispenser and spout panel operating. Animals were not given water in their home cages.
Access to the S1 solution was the only drinking opportunity they received during this phase . When removed from the operant chamber each subject was administered an ip injection of either Animals were given 20 min free access to both S1 and tap water, simultaneously . Again , the amount of each fluid consumed was recorded for each subject (T3S2 & T3H2O) .
Test 4 (T 41-One day after Test 3, Test 4 was conducted . Under the same apparatus circumstances as the previous two tests , Test 4 offered a two bottle choice test of S1 and S2 , w ith 20 min free access to both fluids . Consumption was recorded (T 4S1 & T 4S2) .

Results
Means and standard deviations were computed for each group on all dependent measures and are presented in Table 2. Although consumption levels were quite low for some measures standard dev iations ind icate that the sca le was sens itive enough to detect st ill lower consumption , th e reby avoid ing floor effects .  source table is presented in Table 3).
Insert Tables 2 & 3 about here   34 2x2x2 ANOVAs (univariate analysis for each dependent measure) were generated (see Table 3  Comparisons were then conducted on the basis of apriori predictions made on theoretical grounds, in order to examine the specific pattern of results for particular behavioral effects . To investigate the conditioned taste aversion effect (L vs N), a transformation of T2S1 and T2H2O data was performed such that the suppression ratio of the amount of S1 consumed to total amount of fluid consumed [S1 / (S1 + H2O)] could be examined . A 2x2 ANOVA using Configuration x Revaluation yielded significant differences for both Configuration and Revaluation as well as an interaction (see Table 4): C, F(1,59) = 7.43, p<.01; R, F(1,59) = 49.11 , p<.001 ; C x R, F(1,59) = 10.16, p<.01, (Table 4 provides the mean suppression ratios for these groups and Table 5 the source table for this   analysis). Tukey follow-ups indicated significant differences (i.e ., p<.05) between group BN (where quinine was S 1 and the animal experienced no illness) and all other groups, as well as group AN (where saccharin was S 1 and the animal experienced no illness) and all other groups . That is, N (NaCl , nonpoisoned) subjects consumed significantly more S 1 relative to total fluid intake than did L (LiCI, poisoned ) subjects and , BN subjects consumed significantly more S 1 (quinine) relative to total fluid intake than did AN subjects (for whom S1 was saccharin) . See Figure 2 for this data . In addition , analyses on the absolute amounts of S1 and H2O consumed_ on Test Day 2 indicated , for C, F(1,59)=7.0, p<.05, for R, F(1,59)=0.15, n.s., and for C x R, F(1,59)=18 .55, p<.001. Tukey follow up analyses (p<.05) showed that animals in the AN group consumed significantly more water than all other groups as well as more water than saccharin or quinine than any other group. There were no differences in the amount of water consumed by any of the remaining groups, that is, groups AL, BL, and BN did not differ in the amount of water each consumed . Also to be noted, was the finding that there was no difference in water consumption for groups BL and BN. Nor was there a difference in the amount of water and quinine consumed by group BN. However, group BL consumed significantly less quinine than water . Additionally , group BL consumed significantly less quinine than group BN.
Since therewas a strong , theoretical reason to examine the effect of Sensory Preconditioning (or the lack of such an effect , as indicated by the MANOVA) for subjects receiving Paired vs Random stimuli presentat ions , a 2x2x2 AN OVA (C x StP x R) for T 1 P was conducted (Source Table is presented in Table 6) . Lever pressing contingently reinforced by S2 should be indicative of the predictive and motivational value of S2. The following results were found for were presented randomly, and no illness was experienced) pressed significantly more than all other groups and that there were no significant differences among any other groups (see Figure 3) . Thus the interaction of random vs predictive presentation of S2 and S1 with illness vs no illness was entirely the result of the extremely high frequency of lever pressing in the BAN group as compared to all other groups .
In addition, a 2x2x2 ANOVA (C x StP x R) was conducted to examine the P vs R Sensory Preconditioning effect of T 382 and T 3H2O . A standard ratio transformation to examine suppression was calculated [S2/(S2+ H2O)] to investigate the effect due to Stimulus Presentation (P vs R). S2 consumption relative to total fluid intake should reflect differences due to the contingency between S1 and S2 (see Table 7 for means and sds  Table 8 for ANOVA source table). These data show no differences due to Stimulus Presentation (P vs R), suggesting the need to further examine these results to assess whether the absence of significant differences was due to all animals learning or none learning the relationship between S2 and S1. This is investigated in detail with post hoc analyses (see below). In order to examine the pattern of significant results, Tukey follow-up analyses were conducted and indicated that group APN consumed significantly (p<.05) more S2 relative to total fluid intake than groups ARL, BAL, BPN, BAN , and BPL. In addition, group ARN consumed significantly (p<.05) more S2 relative to total fluid intake than groups BAL, BPN, BAN, and BPL.
No other d ifferences were significant. Perhaps more interesting was the ANOVA of the absolute amounts of S2 and H2O yielding: C,  Table 9). Follow-up analyses were most reveal ing in the pattern of nonsignificant results . No differences were found for any A group, that is no differences were found for comparisons of water and quinine where quinine was S2, regardless of the Stimulus Presentation (i.e., P or R), nor illness or not to S1 (saccharin) . This pattern of results leads to a number of post hoc analyses which are presented later in this sect ion.
Insert Tables 4 , 5 (Table 1 O presents Source Table) . Tukey follow-up tests indicated groups AS2 and 881 consumed significantly more (p<.05) than groups AS1 and 882. That is, across Configuration , subjects consumed significantly more quinine (groups AS2 and 881) than sacchar in (groups AS1 and 882) . All subjects clearly preferred quinine .
Insert Table 10 about here In light of the absence of significant differences between the Paired and Random groups, post hoc analyses were conducted .
These analyses were employed in order to ascertain whether all groups learned a predictive relationship between S1 and S2 or if none of the groups established a predictive relationship between S1 and S2. graphic presentation of the means is presented in Figure 2 . Groups and measures indicating quinine consumption were APS2, ARS2 , BPS1, and BRS1 (because quinine was S2 in the A groups and quinine was S1 in the B groups) . Significant results were found for Day, F(4,52)=123.76, p<.001, and Day x C, F(4,52)=4.82 , p<.01. Figure 3 presents the means for quinine consumption across days .
Differences due to the differing predictive value should emerge across Sensory Preconditioning days for a particular taste stimulus (either saccharin or quinine) as the predictive value of the Paired S2 stimu lus is altered while the value of the same taste stimulus remains unaltered in the Random groups. As Figures 5 and 6 demonstrate , no such pattern developed .
Insert Tables 11 & 12 and   Figures 5 & 6 about here   Discussion The results of this experiment provide ample support for the prediction regarding the primary effect of the devaluation of the US. Subjects exposed to LiCI during Revaluation consumed significantly less S 1 than groups that were not exposed to S 1 40 devaluation (i.e ., NaCl groups) for both the saccharin and quinine showed differences between BN and AN groups . The only difference between these groups was that S 1 was quinine for the B groups and saccharin for the A groups . Neither group was made ill therefore, the original value of S 1 (the US) was not altered . This finding indicates that B subjects con sumed significantly more quinine relative to their total fluid intake (quinine and water) than A subjects consumed of saccharin relative to their total fluid intake (i.e., saccharin and water), and supports the conclusion regarding a greater motivational value for the quinine solution .
Additionally, analyses from Test Day 2 using the absolute amounts of water , saccharin and quinine, as appropriate for each respective group, show that animals in AL, AN, and BL groups discriminated their respective S 1 fluid from water, while group BN did not. This finding is important because it demonstrates that saccharin is clearly discriminable from water but that quinine only appears to be discriminable from water when subjects have been made ill to it. Group BL clearly shows a difference in the amount of quinine and water consumed , whereas group BN does not. This finding indicates that quinine and water are equally preferred , and either 1) that animals can not distinguish quinine from water unless they have been made ill to quinine, or 2) that the animals are able to discriminate but do not demonstrate any behavioral changes unless they are made ill to quinine. This information also is germane to the assessment of the data regarding the Stimulus However, data from Test Day 2 (S 1 vs H 2 0) also showed that there were no differences between the amounts of qu inine and water consumed when animals were not made ill to quinine . Moreover , on Test Day 3 , when quinine was S 2 , there were no differences among any of the A groups regarding quin ine and water consumption. On both Test Days 2 and 3 all animals show no evidence of a discrimination between quinine and water except for animals that were specifically made ill to the tast e of qu inine (i.e ., the BL group) .
These results , taken collectively , suggest that illness to quinine is what produced the quinine-water discrimination , but that in the absence of experiencing the salient contingency of illness to qu inin e, water and quinine were not differentially responded to. Such a find ing has strong impl ications for the Sen sory Precond ition ing pha se results of thi s study. Since all animals had access to water in their home cages daily during all phases (except Revaluation and Testing), and were exposed to water in the training apparatus during Acclimation, animals in both the Why did animals that were not made ill to the taste of quinine not demon strate discriminatory responding to quinine and water?
The data demo nstrate th at the animals can discriminate between the two tastes but that they did not do so until the y had been made ill to the tas te of quinine.  , for a detailed discussion of these topics).
The findings of the current research although not designed to deal with such issues, are consistent with the above cited research regarding latent discrimination and learning. Signal detection theory and latent discrimination learning, ultimately, lend support to a cognitive-behavioral synthesis .
Some of the answers to  empirical construction on the question of "what is learned" still need to be explored. To date, information regarding the extent to which a CS retains or alters its original incentive value as a function of the initial motivational value of the US it predicts, and the extent to which that value is affected by the altered value of the stimulus it predicts (the US), is still incomplete. What remains is the need to investigate the motivational value of a predictive cue in an altered setting as an outcome to a free operant, under a variety of motivational conditions with a variety of predictive relationships employing empirically, discriminable stimuli.
This may be accomplished through the use of the basic design employed in this study with the following adjustments: 1) Use stimuli of more biological/emotional significance with an overt behavioral response . This will allow for an examination of the establishment of the predictive versus random relationship during acquisition through an examination of the learning curve during conditioning.
2) Maintain use of a Te st phase which examines the incentive value of the CS as an outcome for a new response and comparing it with non-contingent presentations of the CS. This is a crucial test of stimulus value due to its direct nature. However, maximize the possibility of the test situation and create a more sensitive measure by employing a leaner reinforcement schedule and a longer session length. A leaner reinforcement schedule would allow for a longer time interval and more behavior to occur without the delivery of the outcome (and consummatory behaviors) disrupting the response rate. In addition a longer session length would allow for more observations to be made and a more stable response rate to emerge . S2=SACCHA RIN S1=SACCHARIN S1=QUININE STIMULUS PRESENTATION PAIRED RANDOM PAIRED RANDOM S2-S1 S2/S1 S2 -S1