The effect of feedback delay and feedback type on perceptual category learning: the limits of multiple systems

Evidence that learning rule-based (RB) and information-integration (II) category structures can be dissociated across different experimental variables has been used to support the view that such learning is supported by multiple learning systems. Across 4 experiments, we examined the effects of 2 variables, the delay between response and feedback and the informativeness of feedback, which had previously been shown to dissociate learning of the 2 types of category structure. Our aim was twofold: first, to determine whether these dissociations meet the more stringent inferential criteria of state-trace analysis and, second, to determine the conditions under which they can be observed. Experiment 1 confirmed that a mask-filled feedback delay dissociated the learning of RB and II category structures with minimally informative (yes/no) feedback and also met the state-trace criteria for the involvement of multiple latent variables. Experiment 2 showed that this effect is eliminated when a less similar, fixed pattern mask is presented in the interval between response and feedback. Experiment 3 showed that the selective effect of feedback delay on II learning is reduced with fully informative feedback (in which the correct category is specified after an incorrect response) and that feedback type did not dissociate RB and II learning. Experiment 4 extended the results of Experiment 2, showing that the differential effect of feedback delay is eliminated when a fixed pattern mask is used. These results pose important challenges to models of category learning, and we discuss their implications for multiple learning system models and their alternatives.     

Can semi-supervised learning explain incorrect beliefs about categories?

Three experiments with 88 college-aged participants explored how unlabeled experiences—learning episodes in which people encounter objects without information about their category membership—influence beliefs about category structure. Participants performed a simple one-dimensional categorization task in a brief supervised learning phase, then made a large number of unsupervised categorization decisions about new items. In all three experiments, the unsupervised experience altered participants’ implicit and explicit mental category boundaries, their explicit beliefs about the most representative members of each category, and even their memory for the items encountered during the supervised learning phase. These changes were influenced by both the range and frequency distribution of the unlabeled stimuli: mental category boundaries shifted toward the middle of the range and toward the trough of the bimodal distribution of unlabeled items, whereas beliefs about the most representative category members shifted toward the modes of the unlabeled distribution. One consequence of this shift in representations is a false-consensus effect (Experiment 3) where participants, despite receiving very disparate training experiences, show strong agreement in judgments about representativeness and boundary location following unsupervised category judgments.     

Bridging in social networks: Who are the people in structural holes and why are they there?

The current paper examines the psychological characteristics of people who link tightly connected cliques to each other with friendship ties. Based on the work of sociologist Georg Simmel, I argue that these people are characterized by two separate motivations, which I call entrepreneurial and relationship-building motivations. Using a complete network from a heterogeneous class of Jewish-, Arab-, and Druze-Israelis, I provide the first empirical evidence of their existence. Gould and Fernandez's brokerage roles were calculated for each participant, and triad proportions from each brokerage type were correlated with well-established psychological questionnaires. Results indicate the existence of two psychologically distinct network orientations, easily interpretable in terms of Simmel's analysis of processes in triads.     

The Wisdom of Individuals: Exploring People's Knowledge About Everyday Events Using Iterated Learning

Determining the knowledge that guides human judgments is fundamental to understanding how people reason, make decisions, and form predictions. We use an experimental procedure called 'iterated learning,' in which the responses that people give on one trial are used to generate the data they see on the next, to pinpoint the knowledge that informs people's predictions about everyday events (e.g., predicting the total box office gross of a movie from its current take). In particular, we use this method to discriminate between two models of human judgments: a simple Bayesian model ( Griffiths & Tenenbaum, 2006 ) and a recently proposed alternative model that assumes people store only a few instances of each type of event in memory (Min K ; Mozer, Pashler, & Homaei, 2008 ). Although testing these models using standard experimental procedures is difficult due to differences in the number of free parameters and the need to make assumptions about the knowledge of individual learners, we show that the two models make very different predictions about the outcome of iterated learning. The results of an experiment using this methodology provide a rich picture of how much people know about the distributions of everyday quantities, and they are inconsistent with the predictions of the Min K model. The results suggest that accurate predictions about everyday events reflect relatively sophisticated knowledge on the part of individuals.     

Studies in spatial learning. I. Orientation and the short-cut. 1946.

1. The original rough formulation of the expectancy theory is difficult to distinguish from the alternative stimulus-response doctrines. Part of this difficulty results from the fact that implicit in this rough formulation, is a definition of the matrix "x expects a goal at location L," which makes it equivalent to the matrix "x runs down the practiced path," when certain conditions are fulfilled. Because of this difficulty, we have rejected this definition. 2. We have suggested instead a definition of the matrix "x expects a goal at location L" which makes it equivalent to the matrix "x runs down the path which points directly to the location L," when certain conditions are fulfilled. 3. To determine whether rats will run down such a path, whenever the original path is blocked, we have run 56 female rats in a situation which conformed to these conditions. 4. Thirty-six percent of the rats chose the path which pointed directly towards the location of the goal. The remaining rats were distributed over the other paths in a chance fashion. 5. We have concluded (1) that rats do learn to expect goals in specific locations, (2) that there are important similarities between this behavior and human symbolic behavior, and (3) that these similarities justify our using the word 'expectation' as a name for the disposition to short-cut when the original patch is blocked.     

Property content guides children’s memory for social learning episodes

How do children’s interpretations of the generality of learning episodes affect what they encode? In the present studies, we investigated the hypothesis that children encode distinct aspects of learning episodes containing generalizable and non-generalizable properties. Two studies with preschool (N = 50) and young school-aged children (N = 49) reveal that their encoding is contingent on the generalizability of the property they are learning. Children remembered generalizable properties (e.g., morphological or normative properties) more than non-generalizable properties (e.g., historical events or preferences). Conversely, they remembered category exemplars associated with non-generalizable properties more than category exemplars associated with generalizable properties. The findings highlight the utility of remembering distinct aspects of social learning episodes for children’s future generalization.     

Revealing human inductive biases for category learning by simulating cultural transmission

We explored people’s inductive biases in category learning—that is, the factors that make learning category structures easy or hard—using iterated learning. This method uses the responses of one participant to train the next, simulating cultural transmission and converging on category structures that people find easy to learn. We applied this method to four different stimulus sets, varying in the identifiability of their underlying dimensions. The results of iterated learning provide an unusually clear picture of people’s inductive biases. The category structures that emerge often correspond to a linear boundary on a single dimension, when such a dimension can be identified. However, other kinds of category structures also appear, depending on the nature of the stimuli. The results from this single experiment are consistent with previous empirical findings that were gleaned from decades of research into human category learning.     

The dimensionality of perceptual category learning: A state-trace analysis

State-trace analysis was used to investigate the effect of concurrent working memory load on perceptual category learning. Initial reanalysis of Zeithamova and Maddox (2006, Experiment 1) revealed an apparently two-dimensional state-trace plot consistent with a dual-system interpretation of category learning. However, three modified replications of the original experiment found evidence of a single resource underlying the learning of both rule-based and information integration category structures. Follow-up analyses of the Zeithamova and Maddox data, restricted to only those participants who had learned the category task and performed the concurrent working memory task adequately, revealed a one-dimensional plot consistent with a single-resource interpretation and the results of the three new experiments. The results highlight the potential of state-trace analysis in furthering our understanding of the mechanisms underlying category learning.     

Drift in children's categories: when experienced distributions conflict with prior learning

Psychological intuitions about natural category structure do not always correspond to the true structure of the world. The current study explores young children's responses to conflict between intuitive structure and authoritative feedback using a semi‐supervised learning (Zhu et al., 2007) paradigm. In three experiments, 160 children between the ages of 4 and 8 learned a one‐dimensional decision criterion for distinguishing yummy and yucky ‘alien fruits’. They then categorized a large number of new fruits without corrective feedback. The distribution of the new fruits was manipulated such that the natural boundary in the stimuli did not always correspond to the learned boundary. Children changed their decision criteria to reflect the structure of the new stimuli, effectively unlearning the original boundary. Younger children were especially swayed by the distributional information, being relatively insensitive to feedback that the original non‐natural boundary was, in fact, still correct. Results are discussed in terms of children's ability to selectively attend to specific information (i.e. feedback vs. distribution), and their interests in forming generally useful representations of experience.