Transfer of oddity learning in the pigeon

Two pigeons were trained on a six-key modified oddity-from-sample procedure. The stimuli were olor pictures of birds, butterflies, and human faces. Initially, the third peck on the sample key which presented one of three different bird pictures) lit only one comparison key. Every three dditional pecks on the sample illuminated another comparison key. Fifteen sample pecks produced he maximum of five comparison stimuli. A peck on the comparison key that presented the non-atching bird picture produced grain. Pecks on matching keys turned off all the comparison keys nd repeated the trial. The birds learned to peck each sample until the non-matching comparison timulus was produced, and then to peck that key. After acquisition (70%–90% accuracy), the hree bird stimuli were replaced by a new set of three bird pictures. Subsequent phases provided ew sets of bird, butterfly, and human face stimuli. Both birds showed transfer of oddity learning o the novel samples. The data suggest that the birds may have been engaging in conceptual-type oddity learning, rather than learning discrete five-key discriminations or a series of two component chains.     

Failure of visual prototype learning in the pigeon

Pigeons did not show generalization gradients along a distortion level after single stimulus training with dot patterns produced on a computer-controlled CRT. After discrimination training between a triangle made of six dots versus patterns made up of six random dots, the birds showed orderly generalization gradients as function of degree of distortion of the original triangle. When they were trained with distorted exemplars of the prototype triangle, they responded more often to triangles at this level of distortion than to the prototype triangle. This observation suggests a lack of abstraction of the prototype from its distorted samples. Such failure of prototype learning was also obtained after training with a different S? (a square made up of six dots) or with a different S+ (horizontally arranged dots).     

Ordinal position in the serial learning of rats

Rats were runway trained on each of two, three-trial series consisting of different varieties of reward (X, Y, and Z) and nonreward (N) serving as trial outcomes. The two series are represented as XNY and ZNN. Distinguishing the two series were different brightness and texture cues on the runway floor. Transfer tests, conducted after the rats had developed faster running for rewarded trials than for nonrewarded trials and slower running on Trial 2 of ZNN than on Trial 2 of XNY, provided evidence that trial position, rather than item memories, was controlling the discriminations. In Experiment 1, reversing the floor cues completely reversed the discriminations. In Experiment 2, transfer to NNN did not change the routine patterns of approach that had been established.     

Generalization of learning in three-and-a-half-month-old infants on the basis of amodal relations

Infants of 3.5 months (N = 124) were given the opportunity to learn to relate two objects and their natural, distinctive sounds during a training phase. The objects and sounds were united by temporal synchrony and amodal audiovisual information specifying object composition. Infants then participated in one of three types of transfer tests (requiring low, moderate, or high degrees of generalization) to measure the extent to which intermodal knowledge generalized to a new task and across events (familiar events; change in color/shape; change in substance, motion, and color/shape). Results indicated that infants tested with the familiar events and with events of a new color/shape showed learning and transfer of knowledge. In contrast, infants tested with events of a new substance, motion, and color/shape showed no generalization of learning. Thus, infants of 3.5 months appear to show a moderate degree of generalization of intermodal knowledge across events. Although this knowledge is not restricted to the events of original learning, it cannot yet be flexibly extended across a variety of contexts.     

Generalization of causal efficacy judgments after evaluative learning

In three experiments, we examined the effect of response-outcome relations on human ratings of causal efficacy and demonstrated that such efficacy ratings transfer to novel situations through derived stimulus relations. Causal efficacy ratings were higher, and probability of an outcome given a response was lower, for a differential reinforcement of high rate schedule than for either a differential reinforcement of low rate schedule (Experiment 1) or a variable interval schedule (Experiment 2). In Experiment 3, we employed schedules that were equated for outcome probability and noted that ratings of causal efficacy and the rate of response were higher on a variable ratio than on a variable interval schedule. For participants in all three experiments, causal efficacy ratings transferred to the stimulus present during each schedule and generalized to novel stimuli through derived relations. The results corroborate the view that schedules are a determinant of both response rates and causal efficacy ratings. In addition, the novel demonstration of a mechanism of generalization of these ratings via derived relations has clinical implications.     

Effects of interrun interval on serial learning

Two groups of five rats each received a decreasing quantity of food reward, 14-7-3-1-0 .045-g food pellets, over successive runs in a runway. The interrun interval (IRI) separating runs within each of two daily pattern repetitions (trials) was 10 sec (short, S) or 4–5 min (long, L) and varied over four successive phases of training in the order indicated by the group names, that is, Groups SLSS and LLLS. Anticipation of the 0-pellet element developed more rapidly in Group SLSS than in Group LLLS, but did eventually occur at the long IRI. Anticipation was eliminated by the increase in IRI experienced by Group SLSS in Phase 2 and by the decrease in IRI experienced by Group LLLS in Phase 4. The results are discussed with reference to the effects of changes in stimulus context accompanying IRI shifts on retrieval of task-specific knowledge and with reference to the possible signal value established to IRI-specific stimuli.     

Stimulus control of anticipatory responding in instrumental learning as revealed in serial learning tasks

Items in a fixed series A-B-C can be anticipated employing either item information (e.g., ItemA cues signal Item B) or position information (e.g., Position 2 cues signal Item B). Identical fixed series of rewarded and nonrewarded trials were employed in investigations reported by Capaldi and Miller (1988), Capaldi, Alptekin, Miller, and Birmingham (1997), and Burns, Dunkman, and Detloff (1999). These series were learned employing item cues according to Capaldi and colleagues and position cues according to Burns et al. (1999). Burns et al.’s (1999) conclusions were based on rats’ behavior when shifted to novel series, which could reveal position learning but not item learning. Employing the acquisition conditions of Burns and co-workers but shifts to series that could reveal item learning, evidence for item learning was obtained in two experiments. The most reasonable interpretation of the above set of findings is that rats employ a compound of both types of cues to signal items. To our knowledge, no set of prior findings so strongly suggests a similar conclusion.     

Generalization of the CS- preexposure effect transfers to taste- aversion learning

Rats repeatedly exposed to a maple, vanilla, banana, or water solution were subsequently poisoned with lithium chloride after drinking the maple solution. Subjects preexposed to water showed the strongest aversions to maple, followed, in descending order, by rats preexposed to banana, vanilla, and maple. These results suggest that the effects of prepoisoning experience with a novel flavor generalized to the maple CS and interfered with the development of the maple aversion. The technique reported here is potentially applicable to study of gustatory discrimination processes.     

An analysis of serial pattern learning by rats

When rats receive a sequence of rewards of different magnitudes for traversing a runway, they learn to “track” the sequence, showing anticipation of the forthcoming reward by appropriate running speed. There is disagreement as to whether this behavior depends on rats’ encoding and recalling a complete sequence of foregoing hedonic events or just the immediately preceding one. The present experiments showed that rats can remember more hedonic events than the most recent one. In Experiment 1, when exposed concurrently to the sequences 10-1-0 (pellets) and 0-1-10, they were faster on Run 3 of the increasing than of the decreasing sequence, a discrimination which cannot be made on the basis of the preceding (1-pellet) reward alone. Experiment 2 showed that this behavior reflects genuine anticipation of the Run 3 reward, not simultaneous contrast or other simple aftereffects of Runs 1 and 2. It is argued, however, that these results, together with related findings by Capaldi and Verry (1981), show merely that rats can recall a hedonic event other than the most recent one, not that a sequence of such events is fully recalled in order.     

Integration and representation in rats’ serial pattern learning in the T-maze

Rats were exposed to three-trial series consisting of reinforced (R) trials and one nonreinforced (N) trial in a fixed order, RRN and RNR (Experiments 1 and 2) or NRR and RRN (Experiment 3), on extended visually distinct runways in a T-maze. When initially presented with the same sequence on each series in a session (separate presentations) with the same runway on all trials within a series (Experiments 1 and 3), all the rats developed slower running speeds on N than on R trials. When a runway was sometimes changed between the first and next two trials during separate presentations training (Experiment 2) or both sequences were later intermixed within each session in each experiment, only rats exposed to each sequence on a specific runway maintained these serial running patterns. Rats displayed serial running patterns on a test RNN sequence similar to that on the RNR sequence (Experiment 2), as would be predicted by an intertrial association model of serial pattern learning (Capaldi & Molina, 1979), but responded on test RRR and NRN sequences (Experiment 3) as would be predicted by an ordinal-trialtag/intratrial association model (Burns, Wiley, & Payne, 1986). Results from test series of free-choice trials in Experiments 1 and 2 failed to support a prediction of the intratrial association model that these rats would integrate RRN and RNR sequences. Rather than always selecting a baited runway on both the second and the third free-choice trials, the rats only selected a baited runway on the third trial on the basis of their choice on the second trial, as would be predicted by the intertrial association model. Only after experiencing all possible outcome sequences during forced-choice training in Experiment 3 did these rats predominantly select a baited runway on every free-choice trial.     

Formal structure and pattern length in serial pattern learning by rats

When rats learn to anticipate a sequence of stimulus events, such as a serial pattern of different food quantities, they are sensitive to the rule-based formal structure relating the magnitude of successive stimuli. Earlier research has shown that if formal structure is simple (e.g., if a single “less than” rule relates the size of each successive quantity), patterns are learned faster than if formal structure is complex (e.g., if two or more rules such as “less than” and “greater than” relate successive pattern quantities). Two experiments tested the hypothesis that pattern length modulates the role of pattern complexity. We predicted that pattern length and pattern complexity interact in determining pattern difficulty. That is to say, long complex patterns should be learned more slowly than short complex patterns. However, long simple patterns should be learned faster than short simple patterns. In Experiment 1, rats ran a straight runway to receive repeated sequences of food quantities. The long-monotonic group received a formally simple 18-10-6-3-1-0 pattern, in which each number represents a quantity of food pellets. The long-nonmonotonic group received a formally complex 10-1-3-6-18-0 pattern. Similarly, the short-monotonic and short-nonmonotonic groups received 18-1-0 and 1-18-0 patterns. Pattern tracking—fast and slow running in anticipation of large and small quantities of food, respectively—was taken as an index of pattern learning. In Experiment 2, comparable patterns were used, but rats leverpressed in a discrete-trial procedure; response latencies measured pattern tracking. In both experiments, rats learned formally simple patterns faster than they did formally complex patterns. In Experiments 1 and 2, but less clearly in Experiment 2, the predicted interaction was obtained. The results support and generalize the idea that rats encode and use some representation of the formal rule structure of serial patterns as they learn them.     

Memory,serial anticipation pattern learning,and transfer in rats

Learning & Behavior - Three experiments tested hypotheses about whether rats respond appropriately to, or track, an orderly series of reward magnitudes terminating in nonreward by encoding the...     

Difficult serial anticipation learning in rats: Rule-encoding vs. memory

Rats, trained in a runway, were asked to anticipate, while running slowly, the last two events in repeating series of .045-g food pellets. The series were either weakly monotonic (14, 5, 5, 1, and then 0 pellets/run) or nonmonotonic (5-5-14-1-0). While the terminal 0-pellet event was better anticipated in the weakly monotonic series, the reverse was the case for the next-to-last 1-pellet event. These findings were expected from a memory-discrimination learning hypothesis of serial learning, which suggests that the memory of one event in a series can be used to signal the next event. However, the better anticipation of the 1-pellet event by the nonmonotonic group was inconsistent with the recently stated rule-encoding position of Hulse (1980). According to that view, difficult series of the sort employed in the present investigation are learned by encoding the rule structure of the series, with events in the series with the simple rule structure (the weakly monotonic series in this investigation) being better anticipated than events in the series with the complex rule structure.     

Associative learning and elemental representation: II. Generalization and discrimination

This paper follows on from an earlier companion paper (McLaren & Mackintosh, 2000), in which we further developed the elemental associative theory put forward in McLaren, Kaye, and Mackintosh (1989). Here, we begin by explicating the idea that stimuli can be represented as patterns of activation distributed across a set of units and that different stimuli activate partially overlapping sets (the degree of overlap being proportional to the similarity of the stimuli). A consequence of this view is that the overall level of activity of some of the units representing a stimulus may be dependent on the nature of the other stimuli present at the same time. This allows an elemental analysis in which provision for the representation of configurations of stimuli is made. A selective review of studies of generalization and discrimination learning, including peak shift, transfer along a continuum, configural discrimination, and summation, suggests that the principles embodied in this class of theory deserve careful consideration and will form part of any successful model of associative learning in humans or animals. There are some phenomena that require an elemental/associative explanation.     

Observational learning in the pigeon: Effects of model’s rate of response and percentage of reinforcement

Seventeen pigeons observed a model peck an illuminated key at either a high or a low rate and obtain a high or low percentage of possible reinforcement. Observers were subsequently placed on an automaintenance schedule. Although there was no difference among groups in the number of trials to the first peck, when the present data were compared with other researchers’ automaintenance-acquisition data, there was some indication that modeling reduced the number of trials to the first peck over nonmodeled birds. However, by the end of 20 sessions, the birds that had observed a model pecking at a high rate and with consistent reinforcement pecked significantly faster than those that had observed the model peck at a slow rate or obtain infrequent reinforcement. The conclusion is that two types of information are transferred via a model: first, a correlation between the stimulus and the reinforcer, and second, the specific or minute attributes of the schedule that may produce reinforcement.     

An examination into some variables said to affect serial learning

In each of three serial learning investigations, rats in a runway were given varying numbers of 0.045-g food pellets in a fixed order. Serial learning was indexed by faster running to larger than to smaller reward quantities. It has been suggested by Hulse (Animal Learning & Behavior, 1980,8, 689–690) and by Roitblat (Behavioral Brain Sciences, 1982,5, 353–371) that differences between two or more serial learning groups that have been obtained under one set of specified experimental conditions may be completely reversed or eliminated under another set of specified experimental conditions. In each of the three investigations reported here, we examined series that had been compared in previous investigations, employing, however, experimental conditions that, according to Hulse and to Roitblat, should produce results different from those obtained previously. In each of the three investigations reported here, the groups differed as they had previously. The findings obtained in this report suggest that none of the following variables is critical to the results obtained in serial learning investigation in the sense suggested by Hulse and Roitblat: the number of items contained in the series, the number of times the series is presented each day, the temporal interval elapsing between items, and the temporal interval elapsing between series presentations. The implications of the present findings for the rule-encoding view of Hulse and his coworkers and for the memory-discrimination learning view of Capaldi and his coworkers are examined.     

Oddity learning in the pigeon: Effect of negative instances,correction, and number of incorrect alternatives

Pigeons have difficulty learning a standard oddity task involving two colors and three stimulus positions. In Experiment 1, performance on standard noncorrection trials was compared with performance on (1) rerun correction trials in which errors resulted in trial repetition, (2) noncorrection trials with added “negative instance” trials involving presentation of three stimuli, all of which matched, and (3) a combination of correction and added negative instance trials. Results indicated that negative instances, but not correction trials, significantly facilitated oddity performance. In Experiment 2, Phase 1, number of stimulus positions lit (three or five) was factorially manipulated with number of positions on which the odd stimulus could appear (three or five). An increase in number of positions lit, but not number of positions that could be odd, facilitated performance. In Phase 2, birds transferred from trials with five positions lit to four positions lit performed significantly better than controls; but in Phase 3, the same birds did not perform significantly better than controls when transferred to trials with three positions lit. In both experiments, analysis of performance as a function of response position indicated better performance at the end of each display than in the middle. These results, together with the group performance differences in Experiment 2, suggest that oddity learning in pigeons involves a size, or number, discrimination.     

Chromatic adaptation in the pigeon

This study presents evidence that adaptation to colored light alters the apparent hue of subsequently presented stimuli in pigeons. During training, right and left keypecks were reinforced following responses to colored and nominally achromatic slides, respectively. During test sessions, subjects continued to observe and report on the two classes of slides while 6-min components alternated, such that the experimental chamber was illuminated with either a green flood lamp or a nominally white bulb. The proportion of right keypecks following achromatic slides was much higher during green components than during white components, indicating that the achromatic slides appeared more like the chromatic slides.     

The effects of intertrial and feature-target intervals on operant serial feature-positive discrimination learning

The effects of intertrial interval (ITI) and feature-target interval (FTI) on the nature of learning in discrete-trial operant serial feature-positive (feature → target+ / target?) discrimination training were examined in two experiments with rats. Discrimination performance was acquired more rapidly with longer ITIs and shorter FTIs. In contrast, transfer to a separately trained target was greater with shorter ITIs regardless of FTI. Persistence of discrimination performance after feature extinction was greater with longer ITIs. Only the last of these performance measures showed evidence for invariance with constant ITI/FTI ratios. The results are discussed in the context of theories of occasion setting.