On the dominance of unidimensional rules in unsupervised categorization

In several experiments, observers tried to categorize stimuli constructed from two separable stimulus dimensions in the absence of any trial-by-trial feedback. In all of the experiments, the observers were told the number of categories (i.e., two), they were told that perfect accuracy was possible, and they were given extensive experience in the task (i.e., 800 trials). When the boundary separating the contrasting categories was umdimensional, the accuracy of all observers improved significantly over blocks (i.e., learning occurred), and all observers eventually responded optimally. When the optimal boundary was diagonal, none of the observers responded optimally. Instead they all used some sort of suboptimal unidimensional rule. In a separate feedback experiment, all observers responded optimally in the diagonal condition. These results contrast with those for supervised category learning; they support the hypothesis that in the absence of feedback, people are constrained to use unidimensional rules.     

The tight coupling between category and causal learning

The main goal of the present research was to demonstrate the interaction between category and causal induction in causal model learning. We used a two-phase learning procedure in which learners were presented with learning input referring to two interconnected causal relations forming a causal chain (Experiment 1) or a common-cause model (Experiments 2a, b). One of the three events (i.e., the intermediate event of the chain, or the common cause) was presented as a set of uncategorized exemplars. Although participants were not provided with any feedback about category labels, they tended to induce categories in the first phase that maximized the predictability of their causes or effects. In the second causal learning phase, participants had the choice between transferring the newly learned categories from the first phase at the cost of suboptimal predictions, or they could induce a new set of optimally predictive categories for the second causal relation, but at the cost of proliferating different category schemes for the same set of events. It turned out that in all three experiments learners tended to transfer the categories entailed by the first causal relation to the second causal relation.     

Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback—an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Implicit learning of conjunctive rule sets: An alternative to artificial grammars

A single experiment is reported that investigated implicit learning using a conjunctive rule set applied to natural words. Participants memorized a training list consisting of words that were either rare-concrete and common-abstract or common-concrete and rare-abstract. At test, they were told of the rule set, but not told what it was. Instead, they were shown all four word types and asked to classify words as rule-consistent words or not. Participants classified the items above chance, but were unable to verbalize the rules, even when shown a list that included the categories that made up the conjunctive rule and asked to select them. Most participants identified familiarity as the reason for classifying the items as they did. An analysis of the materials demonstrated that conscious micro-rules (i.e., chunk knowledge) could not have driven performance. We propose that such materials offer an alternative to artificial grammar for studies of implicit learning.     

Phantom ROC: Recollection rejection in a hybrid conjoint recognition signal detection model

The present experiments test predictions of dual process models with regards to recollection rejection using a model called phantom ROC. Participants studied pictures and then took a recognition test in which they were presented with targets (i.e., the exact picture presented), related lures (i.e., the same object but from a different angle), and unrelated lures (i.e., objects that had not been shown). For each item, participants answered both standard recognition and meaning recognition questions. In Experiment 1 participants studied pictures under either full attention or divided attention. In Experiment 2 some participants were told that no object was shown twice (i.e., mutual exclusivity instructions), while others were told that both camera angles might have been shown for some objects (no mutual exclusivity instructions). The present experiments provide three converging measures, all of which are consistent with dual process models that propose a recollection rejection mechanism.     

The effect of post‐identification feedback on the recall of crime and perpetrator details

The present study examines the effect of identification feedback on the quantity and accuracy of crime event details recalled, willingness to attempt misleading questions and confidence in the accuracy of these details. All participants (N = 60) viewed a short video clip of a staged building society robbery and then made a false identification of the robber from a target‐absent photospread. Eyewitnesses were next given confirming feedback (i.e. told that they had identified the suspect), disconfirming feedback (i.e. told that they had failed to identify the suspect) or no feedback. All eyewitnesses then attempted a series of short‐answer questions relating to details about the robber, accomplice, victim, building society, theft and getaway. Disconfirming feedback significantly reduced eyewitness confidence in recall accuracy but there was no significant effect of feedback on the overall quantity and accuracy of details recalled or willingness to attempt misleading questions. The theoretical implications of these results are discussed. Copyright © 2005 John Wiley & Sons, Ltd.     

Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback--an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Monitoring of learning at the category level when learning a natural concept: Will task experience improve its resolution?

Researchers have recently begun to investigate people's ability to monitor their learning of natural categories. For concept learning tasks, a learner seeks to accurately monitor learning at the category level — i.e., to accurately judge whether exemplars will be correctly classified into the appropriate category on an upcoming test. Our interest was in whether monitoring resolution at the category level would improve as participants gain task experience across multiple study-test blocks, as well as within each block. In four experiments, exemplar birds (e.g., American Goldfinch, Cassin's Finch) paired with each family name (e.g., Finch) were studied, and participants made a judgment of learning (JOL) for each exemplar. Of most interest, before and after studying the exemplars, participants made category learning judgments (CLJs), which involved predicting the likelihood of correctly classifying novel birds into each family. Tests included exemplars that had been studied or exemplars that had not been studied (novel). This procedure was repeated for either one or two additional blocks. The relative accuracy of CLJs did not improve across blocks even when explicit feedback was provided, whereas item-by-item JOL accuracy improved across blocks. Category level resolution did improve from pre-study to post-study on an initial block, but it did not consistently increase within later blocks. The stable accuracy of CLJs across blocks poses a theoretical and empirical challenge for identifying techniques to improve people's ability to judge their learning of natural categories.     

Shaping data collection congruent with experimental hypotheses

The influence of instructions and feedback from an experimenter on observational recordings of disruptive behavior was evaluated. Four subject-observers recorded four categories of disruptive behavior from videotapes of children in a classroom setting. Two sets of videotapes, labelled "baseline" and "treatment", were matched for rates of disruptive behavior in each category. The observers were told that two target behaviors were expected to decrease and the other two control behaviors were not expected to change during the treatment condition. During observational recording of treatment videotapes, the observers were given positive feedback when they reported decreases in the two target behaviors, and negative feedback when they reported either to change or increases in the two control behaviors. The target behaviors were recorded less frequently by observers, while recordings of control behaviors were unaffected during the treatment phase. These results suggest that contingent feedback to observers should be avoided as a possible source of bias in behavioral recordings.     

Ad hoc category restructuring

Participants learned to classify seemingly arbitrary words into categories that also corresponded to ad hoc categories (see, e.g., Barsalou, 1983). By adapting experimental mechanisms previously used to study knowledge restructuring in perceptual categorization, we provide a novel account of how experimental and preexperimental knowledge interact. Participants were told of the existence of the ad hoc categories either at the beginning or the end of training. When the ad hoc labels were revealed at the end of training, participants switched from categorization based on experimental learning to categorization based on preexperimental knowledge in some, but not all, circumstances. Important mediators of the extent of that switch were the amount of performance error experienced during prior learning and whether or not prior knowledge was in conflict with experimental learning. We present a computational model of the trade-off between preexperimental knowledge and experimental learning that accounts for the main results.     

Auditory feedback and memory for music performance: Sound evidence for an encoding effect

Research on the effects of context and task on learning and memory has included approaches that emphasize processes during learning (e.g., Craik & Tulving, 1975) and approaches that emphasize a match of conditions during learning with conditions during a later test of memory (e.g., Morris, Bransford, & Franks, 1977; Proteau, 1992; Tulving & Thomson, 1973). We investigated the effects of auditory context on learning and retrieval in three experiments on memorized music performance (a form of serial recall). Auditory feedback (presence or absence) was manipulated while pianists learned musical pieces from notation and when they later played the pieces from memory. Auditory feedback during learning significantly improved later recall. However, auditory feedback at test did not significantly affect recall, nor was there an interaction between conditions at learning and test. Auditory feedback in music performance appears to be a contextual factor that affects learning but is relatively independent of retrieval conditions.     

Effects of auditory feedback on gait behavior,gaze patterns and outcome performance in long jumping

In the current study, we conducted two experiments to investigate the impact of concurrent, action-induced auditory feedback on gait patterns, gaze behavior and outcome performance in long jumping. In Experiment 1, we examined the effects of present vs. absent auditory feedback on gait, gaze and performance outcome measures. Results revealed a significant interaction effect between condition (present vs. absent auditory feedback) and phase (acceleration vs. zeroing-in phase) on participants' step lengths indicating that the absence (rather than the presence) of auditory feedback led to facilitatory effects in terms of a more prototypical gait pattern (i.e., shorter steps in the acceleration phase and longer steps in the zeroing-in phase). Similarly, the absent auditory feedback led to a higher gaze stability in terms of less switches between areas of interest (AOIs). However, there was no effect on jumped distance. In Experiment 2, we scrutinized the influence of concurrent vs. delayed auditory feedback on all three performance parameters. In contrast to concurrent feedback, delayed auditory feedback negatively affected all three measures: participants showed (i) dysfunctional deviations from their prototypical gait pattern (i.e., shorter steps across both phases of the run-up), (ii) less stable, maladaptive gaze patterns (i.e., more switches between AOIs) and (iii) poorer jumping performance (i.e., shorter jumped distances). Together, the two experiments provide clear evidence for the impact of concurrent, action-induced auditory feedback on the coordination of complex, rhythmical motor tasks such as the long jump.     

Aging preserves the ability to perceive 3D object shape from static but not deforming boundary contours

A single experiment investigated how younger (aged 18-32 years) and older (aged 62-82 years) observers perceive 3D object shape from deforming and static boundary contours. On any given trial, observers were shown two smoothly-curved objects, similar to water-smoothed granite rocks, and were required to judge whether they possessed the "same" or "different" shape. The objects presented during the "different" trials produced differently-shaped boundary contours. The objects presented during the "same" trials also produced different boundary contours, because one of the objects was always rotated in depth relative to the other by 5, 25, or 45 degrees. Each observer participated in 12 experimental conditions formed by the combination of 2 motion types (deforming vs. static boundary contours), 2 surface types (objects depicted as silhouettes or with texture and Lambertian shading), and 3 angular offsets (5, 25, and 45 degrees). When there was no motion (static silhouettes or stationary objects presented with shading and texture), the older observers performed as well as the younger observers. In the moving object conditions with shading and texture, the older observers' performance was facilitated by the motion, but the amount of this facilitation was reduced relative to that exhibited by the younger observers. In contrast, the older observers obtained no benefit in performance at all from the deforming (i.e., moving) silhouettes. The reduced ability of older observers to perceive 3D shape from motion is probably due to a low-level deterioration in the ability to detect and discriminate motion itself.     

Examining recognition criterion rigidity during testing using a biased-feedback technique: Evidence for adaptive criterion learning

Recognition models often assume that subjects use specific evidence values (decision criteria) to adaptively parse continuous memory evidence into response categories (e.g., "old," "new"). Although explicit pretest instructions influence criterion placement, these criteria appear extremely resistant to change once testing begins. We tested criterion sensitivity to local feedback using a novel biased-feedback technique designed to tacitly encourage certain errors by indicating they are the correct choices. Experiment 1 demonstrated that fully correct feedback had little effect on criterion placement, whereas biased feedback during Experiments 2 and 3 yielded prominent, durable, and adaptive criterion shifts, with observers reporting that they were unaware of the manipulation in Experiment 3. These data suggest that recognition criteria can be easily modified during testing through a form of feedback learning that operates independently of stimulus characteristics and observers' awareness of the nature of the manipulation. This mechanism may be fundamentally different from criterion shifts following explicit instructions and warnings, or shifts linked to manipulations of stimulus characteristics combined with feedback highlighting those manipulations.     

Category number impacts rule-based but not information-integration category learning: further evidence for dissociable category-learning systems

Category number effects on rule-based and information-integration category learning were investigated. Category number affected accuracy and the distribution of best-fitting models in the rule-based task but had no effect on accuracy and little effect on the distribution of best-fining models in the information-integration task. In the 2 category conditions, rule-based learning was better than information-integration learning, whereas in the 4 category conditions, unidimensional and conjunctive rule-based learning was worse than information-integration learning. Rule-based strategies were used in the 2-category/rule-based condition, but about half of the observers used rule-based strategies in the 4-category unidimensional and conjunctive rule-based conditions. Information-integration strategies were used in the 4-category/ information-integration condition and by the end of training were used in the 2-category/information-integration condition.     

One or two dimensions in spontaneous classification: a simplicity approach

When participants are asked to spontaneously categorize a set of items, they typically produce unidimensional classifications, i.e., categorize the items on the basis of only one of their dimensions of variation. We examine whether it is possible to predict unidimensional vs. two-dimensional classification on the basis of the abstract stimulus structure, by employing Pothos and Chater's simplicity model of spontaneous categorization [Pothos, E. M., & Chater, N. (2002). A simplicity principle in unsupervised human categorization. Cognitive Science, 26, 303-343]. The simplicity model provides a quantitative measure of how intuitive a particular classification is. With objects represented in two dimensions, we propose that a unidimensional classification will be preferred if it is more intuitive than all possible two-dimensional ones, and vice versa. Empirical results supporting this proposal are reported. Implications for Goodman's paradox are discussed.     

Novel Labels Increase Category Coherence,But Only When People Have the Goal to Coordinate

From infancy, we recognize that labels denote category membership and help us to identify the critical features that objects within a category share. Labels not only reflect how we categorize, but also allow us to communicate and share categories with others. Given the special status of labels as markers of category membership, do novel labels (i.e., non‐words) affect the way in which adults select dimensions for categorization in unsupervised settings? Additionally, is the purpose of this effect primarily coordinative (i.e., do labels promote shared understanding of how we categorize objects)? To address this, we conducted two experiments in which participants individually categorized images of mountains with or without novel labels, and with or without a goal of coordination, within a non‐communicative paradigm. People who sorted items with novel labels had more similar categories than people who sorted without labels only when they were told that their categories should make sense to other people, and not otherwise. We argue that sorters' goals determine whether novel labels promote the development of socially coherent categories.     

Effector-dependent learning by observation of a finger movement sequence

Can observational learning be effector dependent? In 3 experiments, observers watched a model respond to a 6-item unique sequence in a serial reaction time task. Their sequence knowledge was then compared with that of controls who had performed an unrelated task or observed a model responding to random targets. Observational learning was indicated when the introduction of a new sequence was associated with more reaction time elevation in observers than in controls. The authors found evidence of observational learning only when observers used the finger movement sequence that they observed during training, not when they responded at the same sequence of locations using different digits. Free generation and recognition tests also detected observational learning. These results imply that observational learning can be both explicit and effector dependent.     

Dual-learning systems during speech category learning

Dual-system models of visual category learning posit the existence of an explicit, hypothesis-testing reflective system, as well as an implicit, procedural-based reflexive system. The reflective and reflexive learning systems are competitive and neurally dissociable. Relatively little is known about the role of these domain-general learning systems in speech category learning. Given the multidimensional, redundant, and variable nature of acoustic cues in speech categories, our working hypothesis is that speech categories are learned reflexively. To this end, we examined the relative contribution of these learning systems to speech learning in adults. Native English speakers learned to categorize Mandarin tone categories over 480 trials. The training protocol involved trial-by-trial feedback and multiple talkers. Experiments 1 and 2 examined the effect of manipulating the timing (immediate vs. delayed) and information content (full vs. minimal) of feedback. Dual-system models of visual category learning predict that delayed feedback and providing rich, informational feedback enhance reflective learning, while immediate and minimally informative feedback enhance reflexive learning. Across the two experiments, our results show that feedback manipulations that targeted reflexive learning enhanced category learning success. In Experiment 3, we examined the role of trial-to-trial talker information (mixed vs. blocked presentation) on speech category learning success. We hypothesized that the mixed condition would enhance reflexive learning by not allowing an association between talker-related acoustic cues and speech categories. Our results show that the mixed talker condition led to relatively greater accuracies. Our experiments demonstrate that speech categories are optimally learned by training methods that target the reflexive learning system.