Compound-component and conditional discrimination of colors and odors by honeybees: Further tests of a continuity model

In experiments previously reported, individual honeybees were trained in a variety of problems to discriminate color-odor compounds. The results could be modeled accurately on the assumption that the associative strength of each component of a compound stimulus changes independently with reinforcement or nonreinforcement of the compound (independence rule) and that the associative strength of a compound is equal to the sum of the strengths of its components (summation rule). In the present experiments, which were designed to challenge the model, honeybees were trained in compound-component problems (choosing between compounds and their separate components) and in conditional problems (choosing between colors on the basis of a common odor or between odors on the basis of a common color). The new data, together with all of the previous data, could be modeled accurately on the further assumption that interaction among the components of a compound generates a new, compound-unique component that gains and loses associative strength in the same way as other components and contributes in the same way to the strength of the compound; the independence and summation rules continue to apply.     

Performance of honeybees in reversal and ambiguous-cue problems: Tests of a choice model

Individual honeybees visiting the laboratory regularly for food were trained in a variety of discrimination-reversal and ambiguous-cue problems to choose between two differently scented targets, of which one or the other contained sucrose solution. The training was simulated with a set of equations for predicting choice on the basis of associative strength, and the accuracy of prediction was indexed by the root-mean-square deviation of simulated data from obtained data. Developed earlier on the basis of performance in other problems (Couvillon & Bitterman, 1985), the model proved adequate for the new problems as well.     

Analysis of choice in honeybees

Free-flying honeybees were trained in a set of four problems to choose between two differently scented targets, one or the other of which contained sucrose solution. The training was simulated quantitatively, always with the same simple linear equations for computing changes produced by reinforcement and nonreinforcement in the strength of association between each target and the sucrose, but with a diverse array of functions for predicting choice on the basis of relative strength. Accuracy of prediction was indexed by the root-mean-square (RMS) deviation of simulated data from real data. The results provide some good approximations of what is tentatively assumed to be the true choice function, setting the stage for further development of the associative features of the model to encompass more complex phenomena of honeybee learning in choice situations.     

Learning in honeybees as a function of amount of reward: Rejection of the equal-asymptote assumption

Foraging honeybees were trained individually with successively presented targets differing in odor, one containing 5 µl and the other 20 µl of a 50% sucrose solution, after which preferences were measured in choice tests. In Experiment 1, there were either 8 training trials with each target, 16 trials with each, or 8 trials with the 20-µ1 target and 16 trials with the 5-µl target. In Experiments 2 and 3, the odor-amount relation was reversed after either 24 or 16 trials with each target. In Experiment 4, differential reward was introduced only after two, four, or six feedings-to-repletion on each target. All of the results could be simulated quantitatively and with considerable accuracy on the assumption that the attractiveness of an odor is given by the strength of its association with sucrose; that asymptotic associative strength is an increasing function of amount of reward; and that choice between two odors is determined by their relative associative strength.     

Transverse patterning in pigeons

Pigeons were trained to choose between colored lights (A, B, C, and D), first in a two-pair ambiguouscue problem (A+B?, B+C?), and then, with all colors nondifferentially reinforced, in a three-pair loop problem (A+B?, B+C?, C+A?) followed by a four-pair loop problem (A+B?, B+D?, D+C?, C+A?). Systematic efforts were made to simulate the data with a variety of models incorporating one or another of three conceptions of stimulus compounding prominent in the literature on compound conditioning. One conception is that the components of a compound stimulus are altered by interaction; the second is that they are not altered, but only supplemented with a new (compound-unique) component generated in the interaction; and the third is that the components entirely disappear in a configurational transformation. The ambiguous-cue data could be simulated accurately with each of the models, but the loop data with none of them. A convincing explanation of performance in loop problems remains to be found.     

Learning in honeybees as a function of amount and frequency of reward

In a series of four experiments with free-flying honeybees, individual foragers were trained with targets of two different colors that contained 5 or 20 μl of 50% sucrose solution. The two targets were singly presented in quasi-random sequences on each visit, with the amount of reward to be found on each target perfectly predictable from its color. The number of training visits (4–32) was varied both within and between experiments, and so also was the relative frequency of trials with the 5- and 20-μl targets (1:1, 2:1, 3:1, and 9:1). At the conclusion of training under each condition, unrewarded responses to the targets were measured in a 10-min extinction test, with the targets presented either separately to two different groups of animals (Experiment 1) or as a pair (Experiments 2–4). When the number of training trials with each target was the same (Experiments 1 and 2), the animals responded more in extinction to the 20-μl target than to the 5-μl target, although there was a decline in the overall level of responding to both targets (an overlearning-extinction effect) as the number of training trials increased. After nine times as many, or only three times as many, training trials with the5- μl target as with the 20-μl target, the animals responded more in extinction to the 5-μl target (Experiment 3); after twice as many training trials with the 5-μl target as with the 20-μl target, there was equal responding to both (Experiment 4). The preferences shown in the choice tests of Experiments 2–4 could be simulated rather accurately on the assumptions of a model previously developed to deal with the discrete-trials choice behavior of honeybees and the further assumption that associative strength grows at a rate increasing with amount of reward to an asymptote independent of amount of reward.     

Compound conditioning in honeybees: Blocking tests of the independence assumption

Six experiments with color-odor compounds failed to produce convincing evidence of blocking in honeybees even when the possibility of masking by within-compound association could be discounted. The parsimonious assumption that the components of a compound stimulus gain and lose associative strength independently with reinforcement and nonreinforcement of the compounds (which the experiments were designed to challenge) remains tenable for color-odor compounds, although perhaps not for intramodal compounds.     

Reward and learning in honeybees: analysis of an overshadowing effect

Previous experiments have shown that honeybees trained with colored targets baited with 5- versus 20-µl drops of sucrose solution fail to develop a preference for the 20-µl color when the location of the drop on each target is marked by a white dot (dot-color overshadowing) but that discrimination is not impaired by dots when the targets differ in odor rather than in color. In Experiments 1–3, dot-color overshadowing failed to appear with differences in concentration rather than amount of sucrose (50% vs. 20% or 0%), but it did appear in Experiments 4 and 5 with a difference in probability of reward (consistent vs. partial). Experiment 6 showed no dot-odor overshadowing with a difference in probability of reward. The results are not generally predictable from the Rescorla-Wagner principle of shared associative strength, but point instead (in conjunction with those of earlier experiments) to competition for visual attention.     

Overshadowing and stimulus intensity

Two experiments on conditioned suppression in rats examined overshadowing between visual and auditory components of a compound conditioned stimulus. In the first experiment, when one component was markedly more salient than the other, the more salient overshadowed the less salient, but the latter, although acquiring significant associative strength, did not overshadow the former. When the two components were of approximately equal salience, each overshadowed the other. In the second experiment, reciprocal overshadowing was again observed between two equally salient stimuli, but only when their absolute intensities were relatively low. The failure to observe reciprocal overshadowing under all conditions raises problems for those theories of stimulus selection which assume that stimuli compete for some strictly limited resource. It was suggested, instead, that overshadowing might occur when animals fail to learn to attend to, or actually learn to ignore, stimuli that are not uniquely successful predictors of reinforcement.     

Covariation in conditioned response strength between stimuli trained in compound

Conditioned lick suppression by water-deprived rats was used to elaborate on recent evidence that the attenuated conditioned response elicited by an overshadowed stimulus may be enhanced by extinction of the overshadowing stimulus with which it had been trained in simultaneous compound. Using a modified serial stimulus arrangement in which a light coexisted with the last half of a tone that terminated with footshock, it was found in Experiment 1 that the tone overshadowed the light. Extinction of the tone-shock association resulted in a virtually complete recovery of the response to the overshadowed light. Using this serial overshadowing procedure, the possibility that the strength of a conditioned response to an element trained in compound covaries as a function of the strength of the response to the other element was tested in Experiment 2. Following overshadowing training similar to that of Experiment 1, independent reinforcement of the overshadowed light, that is, associative inflation, was found to have no deleterious effect on the response to the overshadowing tone. This suggests that the effects of postconditioning extinction and inflation of one element do not have symmetrical effects upon responding to the other element. The results of Experiment 2 were replicated in Experiment 3 using a simultaneous compound stimulus as opposed to the serial compound of the previous studies. These results are discussed in terms of various associative and cognitive models of learning and performance.     

Transfer of facilitation in the rat

In two experiments, rats solved two concurrent discrimination problems in which one stimulus (i.e., a facilitator) signaled the reinforcement of another stimulus (i.e., a target). Then a transfer test assessed the capacity of facilitators trained in one problem to promote responding to targets trained in the other. Experiment 1 found that a facilitator promoted as much responding to such a transfer target as to the target with which it was originally trained. Transfer was not obtained with a pseudofacilitator that was uninformative, in training, about the reinforcement of its target. Experiment 2 manipulated the stimulus modality of the targets and facilitators. Its results indicated that transfer performance was not due to generalization between training and transfer targets or facilitators. These results parallel those from comparable autoshaping paradigms with pigeons, and they agree with the view that facilitators promote responding by lowering the threshold for activation of the US representation.     

The effectiveness of inhibitors in human predictive judgments depends on the strength of the positive predictor

We tested whether the development of inhibitory strength, as measured by a summation test, is proportional to the strength of the positive cue (P) against which the inhibitory cue (I) is trained. P predicted the outcome, whereas the co-occurrence of P with I (PI) predicted no outcome. In Experiments 1, 2, and 3, we compared the latter design against a version in which P was overshadowed by another cue (X). In this design, the compound P^oX predicted the outcome, but P^oI^o predicted no outcome. In all three experiments, overshadowed cue I^o was less inhibitory than I. In Experiment 4, a P produced by fewer training trials also supported weaker inhibitory learning. Overall, the results were consistent with associative learning theories, especially Pearce’s (1994) configural model. Contingency models need to make additional assumptions to accommodate this property of inhibitory learning.     

Intramodal blocking in honeybees

Foraging honeybees were trained in a concurrent blocking design with a compound stimulus (AX) reinforced and one of its components (A) either reinforced for a blocking group or nonreinforced for a control group. In Experiment 1, a compound of two colors was used; in Experiment 2, a compound of two odors was used; in Experiment 3, a color-position compound, with position defined in terms of proximity to a distinctive visual landmark, was used; and, in Experiment 4, an odor-position compound was used. In each of the first three experiments, the blocking group responded less than did the control group in a subsequent test with X; in the fourth experiment, the two groups did not differ. The results are in accord with expectations based on those of previous experiments with honeybees in which the independence assumption was found to hold for intermodal compounds but not for intramodal compounds.     

Changes in the retrievability of associations to elements of the compound CS determine the expression of overshadowing

In two experiments, rats were presented with a taste conditioned stimulus (CS) alone, an odor CS alone, or an odor-taste compound followed by lithium chloride injection. When tested 1 day following conditioning, there was evidence that the odor cue overshadowed conditioning to the taste; however, there was no indication of overshadowing following a longer (21-day) retention interval, despite undiminished strength of the aversion in animals conditioned with only the single element (taste). The overshadowing observed at the 1-day retention interval was not reciprocal. Rats conditioned with the odor CS alone or with the compound CS expressed odor aversions of comparable strength—that is, no overshadowing. However, in contrast to the taste aversion, overshadowing of conditioning to the odor by taste was evident following a 21-day retention interval. Rather than reflecting a failure of the overshadowed stimulus to acquire associative strength, these data suggest that overshadowing may be expressed, or not expressed, as a result of changes in the relative retrievability of learned associations over time.     

Summation: Assessment of a configural theory

In five experiments, rats were given Pavlovian pairings of auditory and visual stimuli with delivery of food pellets. Experiment 1 found greater responding to an AB compound after training with the individual A and B stimuli, compared with responding both to the A and B elements and to a separately trained CD compound. Experiment 2 found this enhanced responding to depend on the associative strengths of A and B. In Experiment 3, responding was greater to a CD compound than to the other compounds after an AB-, AD+, BC+ training procedure. In Experiment 4, responding to an AB compound was greater than that to the elements after A was reinforced on a 100% schedule and B on a 50% schedule. In Experiment 5, responding to an AC compound was greater than that to either A or C after an AB+, CD+, A-training procedure. A configural theory, such as that proposed by Pearce (1987), anticipates summation in none of these procedures, unless the conditioned context is assumed to have a salience greater than zero. In order to predict summation in Experiments 3, 4, and 5, a context salience greater than that of the elements must be assumed. However, such an assumption also anticipates that extinction of a 100% stimulus should eliminate responding to a 50% stimulus. The results of Experiment 3 contradicted that prediction. These results conform better to the expectations of elemental models of conditioning.     

Generalization decrements: further support for flexibility in stimulus processing

Thorwart and Lachnit (2009) found reliable symmetrical decrements in two generalization tasks: Removing an already trained component from a compound did not result in larger decrements than adding a new one did. In two contingency learning experiments, we investigated first whether time pressure during stimulus processing, as well as the degree of perceptual grouping, was effective in controlling the symmetry of the decrements (Experiment 1); and second, whether the symmetry was affected by the causal versus predictive nature of the relationship between the cue and the outcome (Experiment 2). The experiments generated unexpected results, since both revealed asymmetrical decrements independent of the manipulations introduced. They therefore demonstrate that more research is needed in order to understand the variables influencing stimulus representation in human associative learning.     

Performance factors in associative learning: Assessment of the sometimes competing retrieval model

Previous simulations revealed that the sometimes competing retrieval model (SOCR; Stout & Miller, Psychological Review, 114, 759-783, 2007), which assumes local error reduction, can explain many cue interaction phenomena that elude traditional associative theories based on total error reduction. Here, we applied SOCR to a new set of Pavlovian phenomena. Simulations used a single set of fixed parameters to simulate each basic effect (e.g., blocking) and, for specific experiments using different procedures, used fitted parameters discovered through hill climbing. In simulation 1, SOCR was successfully applied to basic acquisition, including the overtraining effect, which is context dependent. In simulation 2, we applied SOCR to basic extinction and renewal. SOCR anticipated these effects with both fixed parameters and best-fitting parameters, although the renewal effects were weaker than those observed in some experiments. In simulation 3a, feature-negative training was simulated, including the often observed transition from second-order conditioning to conditioned inhibition. In simulation 3b, SOCR predicted the observation that conditioned inhibition after feature-negative and differential conditioning depends on intertrial interval. In simulation 3c, SOCR successfully predicted failure of conditioned inhibition to extinguish with presentations of the inhibitor alone under most circumstances. In simulation 4, cue competition, including blocking (4a), recovery from relative validity (4b), and unblocking (4c), was simulated. In simulation 5, SOCR correctly predicted that inhibitors gain more behavioral control than do excitors when they are trained in compound. Simulation 6 demonstrated that SOCR explains the slower acquisition observed following CS-weak shock pairings.     

Restricted processing of simultaneously presented brightness and pattern stimuli in pigeons

Performance during simultaneous matching-to-sample was assessed in pigeons presented with element and compound visual samples. In Experiment 1, pigeons were trained with a symbolic matching procedure, in which different pairs of colored comparison cues presented on side keys were mapped onto a bright or dim houselight as one pair of sample stimuli and onto vertical and horizontal lines on the center key as a second pair of sample stimuli. They were then tested with houselight-line compound samples. It was found that matching accuracy for lines was significantly diminished with compound samples relative to element samples. Conversely, house-light intensities were matched as well with compound samples as with element samples. In Experiment 2, a similar effect was found with pigeons that had been trained to match only line samples. In Experiment 3, it was discovered that sample duration had no influence on the matching deficit found with lines following compound samples in birds either trained or not trained to match houselight intensities. These results, taken in combination with recent findings from experiments with auditory-visual compounds, suggest a restricted processing account of pigeon processing of simultaneously presented stimuli from different sources.     

Normalization between stimulus elements in a model of Pavlovian conditioning: Showjumping on an elemental horse

Harris and Livesey. Learning & Behavior, 38, 1-26, (2010) described an elemental model of associative learning that implements a simple learning rule that produces results equivalent to those proposed by Rescorla and Wagner (1972), and additionally modifies in "real time" the strength of the associative connections between elements. The novel feature of this model is that stimulus elements interact by suppressively normalizing one another's activation. Because of the normalization process, element activity is a nonlinear function of sensory input strength, and the shape of the function changes depending on the number and saliences of all stimuli that are present. The model can solve a range of complex discriminations and account for related empirical findings that have been taken as evidence for configural learning processes. Here we evaluate the model's performance against the host of conditioning phenomena that are outlined in the companion article, and we present a freely available computer program for use by other researchers to simulate the model's behavior in a variety of conditioning paradigms.     

Stimulus generalization decrement in latent inhibition to a compound following exposure to the elements or the compound

In two experiments using the taste-aversion paradigm, we attempted to replicate a result reported by Holland and Forbes (1980), in which exposure to the elements of a compound produced more interference with future conditioning (latent inhibition) to the compound than did exposure to the compound itself. In our first experiment, a compound of HC1 and sucrose was used and the amount of fluid consumed during exposure and the first conditioning trial was controlled. Rather than finding enhanced interference produced by exposure to the elements, we found reduced interference relative to exposure to the compound. In Experiment 2, a compound of NaCl and sucrose was used and a method similar to that used by Holland and Forbes was employed. We replicated the result of our Experiment 1. We interpret these results as posing problems for some associative accounts of latent inhibition but as being easily explained as an instance of stimulus generalization decrement.