Feature distribution and background knowledge in category learning

Three experiments examined the interactions of category structure and prior knowledge in category learning.Experiment 1 examined the distribution of atypical or 'crossover' features in category learning. In real categories, crossover features may be unevenly distributed found primarily in very unusual examples of a category (like whales or ostriches). In contrast, in many psychology experiments, each item has exactly one crossover feature. Even versus uneven distribution of crossover features did not affect category learning when the features were neutral. However, when the features were connected by prior knowledge, it was much harder for subjects to learn the structure with the uneven distribution of crossover features. Experiments 2 and 3 found similar results with a slightly less uneven condition. We conclude that learning is a function of the interaction of category structure and prior knowledge rather than either one alone. Furthermore, knowledge benefits learning even when the category contains contradictions of the knowledge, so long as the contradictions are not very salient.     

The acquisition of category structure in unsupervised learning

Four experiments examined the extent to which prior knowledge influences the acquisition of category structure in unsupervised learning conditions. Prior knowledge is general knowledge about a broad domain that explains why an object has the features it does. Category structure refers to the statistical regularities of features within and across categories. Subjects viewed items and then divided them up into the categories that seemed most natural. Each item had one feature that was related to prior knowledge and five features that were not. The results showed that even this small amount of prior knowledge helped subjects to discover the category structure. In addition, prior knowledge enhanced the learning of many of the category's features, and not just the features that were directly relevant to the knowledge. The results suggest that prior knowledge may help to integrate the features of a category, thereby improving the acquisition of category structure.     

Prior sensitivity in theory testing: An apologia for the Bayes factor

A commonly voiced concern with the Bayes factor is that, unlike many other Bayesian and non-Bayesian quantitative measures of model evaluation, it is highly sensitive to the parameter prior. This paper argues that, when dealing with psychological models that are quantitatively instantiated theories, being sensitive to the prior is an attractive feature of a model evaluation measure. This assertion follows from the observation that in psychological models parameters are not completely unknown, but correspond to psychological variables about which theory often exists. This theory can be formally captured in the prior range and prior distribution of the parameters, indicating which parameter values are allowed, likely, unlikely and forbidden. Because the prior is a vehicle for expressing psychological theory, it should, like the model equation, be considered as an integral part of the model. It is argued that the combined practice of building models using informative priors, and evaluating models using prior sensitive measures advances knowledge.     

Considering prior knowledge when using elaborative interrogation

Previous studies have promoted the use of elaborative interrogation (a “why”-questioning strategy) for the acquisition of factual information. One assumption in these studies is that prior knowledge influences when students will be able to use the strategy, with greater prior knowledge leading to higher recall performance. In the studies reviewed here, the effect of prior knowledge on strategy effectiveness was investigated. Specifically, students' performances were compared for materials about which they possessed substantial prior knowledge, little prior knowledge, inconsistent prior knowledge, or shared prior knowledge. In general, the data support the use of elaborative interrogation when studying alone or in dyads especially when learners possess some relevant prior knowledge about the new information.     

Learning of nondomain facts in high- and low-knowledge domains

Research on expertise has repeatedly documented that experts learn new information better than do novices, but only when the information is relevant to the expert's domain. It was found in Experiment 1 that participants showed superior learning and recall of a large quantity of new, non-domain-relevant facts about concepts within their domain of high knowledge than about concepts for which they had low domain knowledge. Experiment 2 investigated whether the participants' superior recall of new facts related to concepts within their domain of high knowledge was due to the number of prior facts associated with the concept or to the prior frequency of repetition of those concepts. It was found that participants' recall of new facts was better for concepts with 5 prior associated facts than for concepts with a single prior association but that the number of previous repetitions of each concept did not affect the level of recall for the new facts.     

Visuo‐spatial abilities in autism: A review

Individuals with autism show various signs of heightened abilities in visuo‐spatial functioning. First, it is long‐established that they excel on embedded figures and block design tests relative to comparison participants. Second, some evidence suggests that processing global features is affected by inappropriate processing of the constituent elements of a stimulus. Third, they are more accurate than comparison participants in judging the shape of a slanted circle in a context in which ambient visual cues are eliminated. This suggests that their perception of the shape is less influenced by prior knowledge. Fourth, they are fast at searching for feature and conjunctive targets in a visual array. Contrary to earlier reports, however, they are susceptible to visual illusions. Also, they do show evidence of utilising prior knowledge when pairing a colour with an atypically coloured target. Accordingly, we conclude that there is something distinctive about autistic visuo‐spatial functioning, but not necessarily in ways that are predicted by the ‘weak central coherence’ hypothesis. Copyright © 2004 John Wiley & Sons, Ltd.     

Category learning with minimal prior knowledge

In 6 experiments, the authors examined the use of prior knowledge in category learning. Previous studies of the effects of knowledge on category learning have used categories in which knowledge was related to all of the category's features. However, people's knowledge of real-world categories often consists of many "rote" features that are not related to their prior knowledge. Five experiments found that even minimal prior knowledge (1 knowledge-relevant feature and 5 rote features per exemplar) can facilitate category learning. Posttests revealed that although the knowledge aided learning, subjects also acquired the rote features that were not related to knowledge, contradicting predictions of an attentional explanation of the knowledge effect. The results of Experiment 6 suggested that subjects attempt to link even rote features to their knowledge.     

The role of prior knowledge in error correction for younger and older adults

Previous work has demonstrated that, when given feedback, younger adults are more likely to correct high-confidence errors compared with low-confidence errors, a finding termed the hypercorrection effect. Research examining the hypercorrection effect in both older and younger adults has demonstrated that the relationship between confidence and error correction was stronger for younger adults compared with older adults. However, recent work suggests that error correction is largely related to prior knowledge, while confidence may primarily serve as a proxy for prior knowledge. Prior knowledge generally remains stable or increases with age; thus, the current experiment explored how both confidence and prior knowledge contributed to error correction in younger and older adults. Participants answered general knowledge questions, rated how confident they were that their response was correct, received correct answer feedback, and rated their prior knowledge of the correct response. Overall, confidence was related to error correction for younger adults, but this relationship was much smaller for older adults. However, prior knowledge was strongly related to error correction for both younger and older adults. Confidence alone played little unique role in error correction after controlling for the role of prior knowledge. These data demonstrate that prior knowledge largely predicts error correction and suggests that both older and younger adults can use their prior knowledge to effectively correct errors in memory.     

Effects of salience and time pressure on ratings of social causality

A model is proposed to account for the effects of a target person's salience on judgments of that target. It is argued that salience leads to more extreme inferences in the direction implied by prior knowledge that is relevant to the judgment. This knowledge may include both specific information about the target being rated and general information about the class of stimuli to which the target belongs. Two experiments supported these hypotheses. When subjects were under time pressure to make judgments of a target person's influence in a social situation, their judgments increased with the salience of the target when they had prior knowledge that the target was generally high in social influence. However, their judgments decreased with the target's salience when subjects had prior knowledge that the target was generally low in social influence. When subjects were given ample time to make their judgments, however, the effects of target salience were attenuated. Possible implications of these findings for prior research on salience effects are discussed.     

Perceiving the sequence of speech and non-speech stimuli

The task is to estimate on which syllable of a spoken sentence a click was superimposed. Experiment I confirms Ladefoged and Broadbent's finding of a systematic tendency to prepose the click (negative displacement), but shows also that the tendency is decreased when prior knowledge of the sentence is provided. Experiment II shows that acoustic prior knowledge is not necessary to produce the decrease and that it occurs also with textual prior knowledge. Experiment III shows that the negative displacement is not eliminated by short-term practice on the task, as Fodor and Bever contended. The effect of prior knowledge is inconsistent with the explanation of negative displacement in terms of attention demands suggested by Ladefoged and Broadbent. It is argued that this explanation was unnecessary, and that negative displacement can be expected in a system which analyses speech by discrete units.     

Feature-based control of attention: evidence for two forms of dimension weighting.

In three experiments, I examined whether prior knowledge of a target feature dimension is useful for guiding spatial attention to the target in a variety of tasks: visual search (Experiments 1A and 1B), texture segregation (Experiment 2), and visual enumeration (Experiment 3). Experiment 1A used a simple search task and found that reaction times for blocks in which a target was defined within a single feature dimension were shorter than those for blocks in which a target was defined across dimensions (within-dimension facilitation: WDF). Intertrial facilitation (ITF; Müller, Heller, & Ziegler, 1995), a dimension-based priming effect from one trial to the immediately following one, was also observed. Both WDF and ITF disappeared when the same stimuli were used under a compound search task (Experiment 1B), in which participants responded to an attribute of the target in a feature dimension different from its defining dimensions. Experiments 2 and 3 showed that WDF and ITF are not necessarily contingent upon each other: In a texture discrimination task, only WDF was found; in an enumeration task for six or seven targets, only ITF was found. These results show that the two forms of dimension weighting, WDF and ITF, are mediated by different mechanisms. WDF was eliminated when focal attention to targets was required, suggesting that feature-based modulation is limited as a source for controlling spatial attention (Kumada, 1999). ITF was correlated with type of response, suggesting dimension-specific response mechanisms (Cohen & Shoup, 1997).     

Feature-based control of attention: Evidence for two forms of dimension weighting

In three experiments, I examined whether prior knowledge of a target feature dimension is useful for guiding spatial attention to the target in a variety of tasks: visual search (Experiments 1A and 1B), texture segregation (Experiment 2), and visual enumeration (Experiment 3). Experiment 1A used a simple search task and found that reaction times for blocks in which a target was defined within a single feature dimension were shorter than those for blocks in which a target was defined across dimensions (within-dimension facilitation; WDF). Intertrial facilitation (ITF; Müller, Heller, & Ziegler, 1995), a dimension-based priming effect from one trial to the immediately following one, was also observed. Both WDF and ITF disappeared when the same stimuli were used under a compound search task (Experiment 1B), in which participants responded to an attribute of the target in a feature dimension different from its defining dimensions. Experiments 2 and 3 showed that WDF and ITF are not necessarily contingent upon each other: In a texture discrimination task, only WDF was found; in an enumeration task for six or seven targets, only ITF was found. These results show that the two forms of dimension weighting, WDF and ITF, are mediated by different mechanisms. WDF was eliminated when focal attention to targets was required, suggesting that feature-based modulation is limited as a source for controlling spatial attention (Kumada, 1999). ITF was correlated with type of response, suggesting dimension-specific response mechanisms (Cohen & Shoup, 1997).     

Mechanisms for precuing superiority in visual recognition

Superior recognition, when similar alternatives are studied prior to brief target pictures of common objects, is usually attributed to selective feature analysis, which is beneficial only when the target is not readily discriminable in the alternatives. The present study found equal precuing superiority, however, for similar and dissimilar alternatives with abstract geometric stimuli, and with pictures of common objects, precuing superiority occurred only when each alternative represented the same semantic category. Precuing effects did not appear when different categories were represented in the alternatives, and performance was similar to other conditions in which only a categorical match between target and alternatives was possible. These results suggest that selective feature analysis and precuing superiority only occur when the target cannot be identified by its semantic category. With dissimilar alternatives activation of the target's category can be used to select a response, and there is no benefit from precuing.     

Integrating episodic memories and prior knowledge at multiple levels of abstraction

Prior knowledge can have a large influence on recall when the memory for the original event is error prone or incomplete. We investigated the interaction between memory and prior knowledge in a recall task involving natural objects such as fruits and vegetables. We first quantified prior knowledge for the sizes of objects in a norming experiment. We then assessed the influence of prior knowledge in a memory experiment in which we compared the actual size of objects shown during a study phase with the reconstructed size of an object during the test phase. Recall was biased both by the mean size of the specific object studied and by the mean size of all objects in the category. This result suggests that the influence of prior knowledge can come from multiple, hierarchically related levels of representation, such as the object-category and superordinate-category levels.     

Further evidence for feature correlations in semantic memory.

The role of feature correlations in semantic memory is a central issue in conceptual representation. In two versions of the feature verification task, participants were faster to verify that a feature (< is juicy >) is part of a concept (grapefruit) if it is strongly rather than weakly intercorrelated with the other features of that concept. Contrasting interactions between feature correlations and SOA were found when the concept versus the feature was presented first. An attractor network model of word meaning that naturally learns and uses feature correlations predicted those interactions. This research provides further evidence that semantic memory includes implicitly learned statistical knowledge of feature relationships, in contrast to theories such as spreading activation networks, in which feature correlations play no role.     

Effects of amount and kind of prior target-uncertainty reduction in the repeated presentations paradigm

Single letters were presented for varying numbers of repeated brief exposures. The S reported on the target after each presentation, identifying only those symbols perceived with certainty. A d’ analysis of the results revealed that target-uncertainty reduction produced significant facilitation in the average level of perceptual sensitivity only in a condition in which the target symbol was exposed prior to each trial. Prior knowledge of the target symbol without prior exposure produced a measurable, but nonsignificant, facilitation in sensitivity. The data suggested that the growth curve normally associated with the repeated-presentations paradigm may be the result of a progressive liberalization of the S’s criterion for reporting in some studies. The data further showed that prior exposure and perhaps prior knowledge can significantly alter the shape of the repetition curve when the target is a single letter. These findings were interpreted as an indication that both amount and kind of target-uncertainty reduction can produce significant changes in the repetition effect for single letters.     

A knowledge-resonance (KRES) model of category learning

This article introduces a connectionist model of category learning that takes into account the prior knowledge that people bring to new learning situations. In contrast to connectionist learning models that assume a feedforward network and learn by the delta rule or backpropagation, this model, the knowledge-resonance model, or KRES, employs a recurrent network with bidirectional symmetric connection whose weights are updated according to a contrastive Hebbian learning rule. We demonstrate that when prior knowledge is represented in the network, KRES accounts for a considerable range of empirical results regarding the effects of prior knowledge on category learning, including (1) the accelerated learning that occurs in the presence of knowledge, (2) the better learning in the presence of knowledge of category features that are not related to prior knowledge, (3) the reinterpretation of features with ambiguous interpretations in light of error-corrective feedback, and (4) the unlearning of prior knowledge when that knowledge is inappropriate in the context of a particular category.     

Partial word knowledge in the absence of recall

Attributes of words can be known even when the words are not currently retrievable. Although repeatedly demonstrated for semantic and contextual dimensions, the evidence is ambiguous for structural characteristics. The present research demonstrates significant above-chance first-letter knowledge across four ordinal levels of retrieval confidence for nonretrieved words—tip of the tongue (TOT), high familiar, low familiar, unfamiliar. Contrary to prior research, there was minimal evidence for syllable number knowledge, even at highest confidence levels. Initial letter recognition in the absence of retrieval resembles the recognition without identification in episodic memory (Cleary, Current Directions in Psychological Science 17: 353–357, 2008), and such implicit familiarity may contribute more generally to confidence assessments of word knowledge in both semantic and episodic memory domains. Furthermore, this outcome suggests that word feature priming in the form of partial phonological activation may occur to some extent for all words during a retrieval attempt, and even for ones that are judged to be unknown.     

Age differences in dimension weighting on visual feature search

Two experiments examined age‐difference effects in visual feature search for two types of prior information (knowledge about target‐defining features and intertrial transition of target‐defining features). In Experiment 1, younger participants had shorter reaction times searching for a target in blocks where the target was defined by a single feature dimension compared to where the target was defined across dimensions (within‐dimension facilitation; WDF). However, WDF was not found in older participants. Intertrial facilitation (ITF), a dimension‐based carryover effect in consecutive trials, was observed both for young and older participants. Experiment 2 provided participants with probabilistic knowledge about target‐defining features. Again, young participants showed a facilitatory effect related to this knowledge, but this effect was reduced in older participants. However, intertrial facilitation was observed for young and older adults to the same degree. These results are not consistent with general age‐related decrement in the maintenance of dimension weighting, but support a dual mechanism of dimension weighting (Kumada, 2001). The results suggest that knowledge‐based control of dimension weighting declined with ageing, but involuntary shift of dimension weighting by intertrial transition of target‐defining features is maintained in ageing.     

Unpriming: the deactivation of thoughts through expression

Unpriming is a decrease in the influence of primed knowledge following a behavior expressing that knowledge. The authors investigated strategies for unpriming the knowledge of an answer that is activated when people are asked to consider a simple question. Experiment 1 found that prior correct answering eliminated the bias people normally show toward correct responding when asked to answer yes-no questions randomly. Experiment 2 revealed that prior answering intended to be random did not unprime knowledge on subsequent attempts to answer randomly. Experiment 3 found that exposure to the correct answer did not influence the knowledge bias but that exposure to the incorrect answer increased bias. Experiment 4 revealed that merely expressing the answer for oneself was sufficient to unprime knowledge. Experiment 5 found that each item of activated knowledge needs to be unprimed specifically, in that correctly answering 1 question does not reduce the knowledge bias in randomly answering another.