Connectionist models of conditioning: A tutorial

Models containing networks of neuron-like units have become increasingly prominent in the study of both cognitive psychology and artificial intelligence. This article describes the basic features of connectionist models and provides an illustrative application to compound-stimulus effects in respondent conditioning. Connectionist models designed specifically for operant conditioning are not yet widely available, but some current learning algorithms for machine learning indicate that such models are feasible. Conversely, designers for machine learning appear to have recognized the value of behavioral principles in producing adaptive behavior in their creations.     

Handling time and choice in pigeons

According to optimal foraging theory, animals should prefer food items with the highest ratios of energy intake to handling time. When single items have negligible handling times, one large item should be preferred to a collection of small ones of equivalent total weight. However, when pigeons were offered such a choice on equal concurrent variable-interval schedules in a shuttlebox, they preferred the side offering many small items per reinforcement to that offering one or a few relatively large items. This preference was still evident on concurrent fixed-cumulative-duration schedules in which choosing the alternative with longer handling time substantially lowered the rate of food intake.     

Concurrent access to two concentrations of orally delivered phencyclidine: effects of feeding conditions.

Two experiments addressed the effects of food satiation and deprivation on oral self-administration of two concurrently available phencyclidine concentrations. In the first experiment, 8 rhesus monkeys self-administered either of two concentrations of phencyclidine ("PCP, angel dust") and water under concurrent fixed-ratio 16 schedules. One concentration was always held constant (0.25 mg/mL) while a series of other phencyclidine concentrations, ranging from 0 (water) to 1.0 mg/mL, was presented in a nonsystematic order. Initially the monkeys were tested while food satiated, and the procedure was then repeated during food deprivation. The monkeys usually selected the higher concentration within the first few minutes of the session, indicating that taste and/or other immediate postingestional effects were important factors. Contrary to a number of previous reports, there were no consistent differences across subjects in the mean number of liquid deliveries or mean drug intake (mg/kg) during food satiation and deprivation. However, for all monkeys the within-session time course of responding during food satiation consistently differed from that during deprivation. A second experiment assessed whether the failure to find consistent differences in drug intake during food satiation and deprivation had been due to the history of concurrent access to different phencyclidine concentrations or to the extended experience with phencyclidine under food-satiation conditions. Six additional monkeys (Group 2) were exposed to the phencyclidine self-administration procedure (during food satiation and deprivation) for the same length of time as the monkeys in Experiment 1 (Group 1), except they received only concurrent access to phencyclidine (0.25 mg/mL) and water. Both groups then received concurrent access to phencyclidine and water during five repeated cycles of food deprivation and satiation. There were also marked individual differences in Group 2: During food satiation, 2 of the monkeys' responding increased, 1 showed no change, and 3 decreased. Examination of a number of historical variables indicated that the greater the percentage of total sessions spent during food satiation with phencyclidine available (before these experiments began), the greater the amounts of phencyclidine consumed during food satiation and the smaller the differences in phencyclidine intake when the two feeding conditions were compared.     

Relative allocation on concurrent schedules can depend on schedule parameters when behavioral parameters are constant

We develop a simple model of switching between the initial links of a concurrent-chain procedure. Behavior is determined by four parameters μ₁, μ₂, q₁, and q₂. The first two are the basic rates of switching from Schedule 1 and Schedule 2, respectively. The second two are the probabilities of leaving Schedule 1 and Schedule 2 after the corresponding terminal link has been completed. We show that for fixed values of these four parameters, the relative allocation on the initial links may change as a result of changes in initial-link schedules. The effect can be quite large if the switching rates are low. An implication is that relative allocation is not necessarily a good measure of behavior.     

More on concurrent interval-ratio schedules: a replication and review.

It has been suggested that the failure to maximize reinforcement on concurrent variable-interval, variable-ratio schedules may be misleading. Inasmuch as response costs are not directly measured, it is possible that subjects are optimally balancing the benefits of reinforcement against the costs of responding. To evaluate this hypothesis, pigeons were tested in a procedure in which interval and ratio schedules had equal response costs. On a concurrent variable time (VT), variable ratio-time (VRT) schedule, the VT schedule runs throughout the session and the VRT schedule is controlled by responses to a changeover key that switches from one schedule to the other. Reinforcement is presented independent of response. This schedule retains the essential features of concurrent VI VR, but eliminates differential response costs for the two alternatives. It therefore also eliminates at least one significant ambiguity about the reinforcement maximizing performance. Pigeons did not maximize rate of reinforcement on this procedure. Instead, their times spent on the alternative schedules matched the relative rates of reinforcement, even when schedule parameters were such that matching earned the lowest possible overall rate of reinforcement. It was further shown that the observed matching was not a procedural artifact arising from the constraints built into the schedule.     

Preference for less segmented fixed-time components in concurrent-chain schedules of reinforcement

A concurrent-chain procedure was used to examine choice between segmented and less segmented response-independent schedules of reinforcement. A pair of independent, concurrent variable-interval 60-s schedules were presented in the initial link, along with a 1.5-s changeover delay. A chained fixed-interval fixed-time and its corresponding tandem schedule constituted the terminal links. The length of the fixed-interval schedule in the terminal link was varied between 5 s and 30 s while that of the fixed-time schedule was kept at 5 s over conditions. The first components of both terminal-link schedules were accompanied by the same stimulus. Except in the baseline condition, the onset of the second component of the terminal-link chained schedule was accompanied by either a localized (key color) or a nonlocalized (dark houselight) stimulus change. Stimulus conditions were constant during the terminal-link tandem schedule. With three exceptions, pigeons demonstrated a slight preference for the tandem over the chained schedule in the terminal link. Furthermore, this preference varied inversely with the length of the first component. In general, these results are consistent with previous studies that reported an adverse effect on choice by segmenting an interval schedule into two or more components, but they are inconsistent with studies that reported preference for signaled over unsignaled delay of reinforcement.     

Behavior regulation and learned performance: Some misapprehensions and disagreements

The behavior-regulation approach to learned performance has been the subject of misapprehension and disagreement concerning: (1) the nature and importance of behavior regulation, (2) the definition and role of behavioral set-points, (3) the relation of optimal schedule performance to behavioral set-points, and (4) the question of whether deviations from total responding or from response patterns are the primary determinant of molar responding under schedule constraint. After clarifying the nature and role of behavior regulation and set-points, this paper shows that the data used to question optimal schedule performance (Allison, 1981a) actually strongly support the general behavior-regulation approach. These data also indicate a role for response-pattern set-points in determining schedule behavior, but contradict the hypothesis that deviations from response-pattern characteristics are the primary determinant of molar schedule effects.     

Imitation among infants in a day-care center and the development of locomotion

To examine the relationship between imitation and locomotor development among peers during infancy, we observed free play in a class of infants in a day-care center over a one-year period. The analysis included 16 infants (mean age was 8.3 months (SD=2.3) at the beginning of the observation) and 5 caregivers. The longitudinal data were divided into the crawling, cruising, and walking periods, and the imitation of object manipulation that took place among the infants during each period was analyzed. With the development of locomotion, peer-to-peer imitation became frequent. Infants who had acquired the ability to walk demonstrated imitation more frequently in triadic interactions than in dyadic interactions. Imitation took place between infants, but it was often mediated by caregivers rather than directly between infants. The results suggest that the acquisition of locomotor skills is a link in a developmental cascade from motor development to infant peer interaction. However, it is reasonable to interpret the results as indicating that the development of locomotion does not directly affect infant interactions in isolation, but rather that walking increases both the sharing and manipulation of objects with caregivers and proximity to peers, and their combined effects form a cascade.     

Social support from weak ties: Insight from the literature on minimal social interactions

When people face difficult life events, such as cancer or bereavement, they fare better when they receive support. These “experiencers” are likely to seek and receive support from a wide range of people, from close others to acquaintances. Indeed, the social support literature has long acknowledged the value of having a diverse support network. Research suggests that experiencers often perceive “weak ties” (i.e., acquaintances) as sources of support, and that experiencers sometimes prefer to get support from weak ties rather than strong ties. Providing support can be challenging for all kinds of supporters, however weak ties may be more likely than strong ties to allow worries about their inability to provide effective support to stop them from providing any support at all, thus depriving experiencers of opportunities for additional support. In this paper, we focus on the fact that often the provision of support occurs via a social interaction. We draw on the social psychology literature on minimal social interactions to suggest reasons why potential weak tie supporters might doubt their ability to provide effective support, and to generate advice to encourage potential supporters to reach out. Finally, we suggest future areas of research, with the ultimate goal of helping to expand the support that is available for people experiencing difficult life events.