Visual feature learning in artificial grammar classification

The Artificial Grammar Learning task has been used extensively to assess individuals' implicit learning capabilities. Previous work suggests that participants implicitly acquire rule-based knowledge as well as exemplar-specific knowledge in this task. This study investigated whether exemplar-specific knowledge acquired in this task is based on the visual features of the exemplars. When a change in the font and case occurred between study and test, there was no effect on sensitivity to grammatical rules in classification judgments. However, such a change did virtually eliminate sensitivity to training frequencies of letter bigrams and trigrams (chunk strength) in classification judgments. Performance of a secondary task during study eliminated this font sensitivity and generally reduced the contribution of chunk strength knowledge. The results are consistent with the idea that perceptual fluency makes a contribution to artificial grammar judgments.     

Rules-plus-exception tasks: A problem for exemplar models?

Erickson and Kruschke (2002b) have shown that human subjects generalize category knowledge in a rule-like fashion when exposed to a rule-plus-exception categorization task. This result has remained a challenge to exemplar models of category learning. We show that these models can account for such performance, if they are augmented with exemplar-specific specificity or exemplar-specific attention. This result, however, is only achieved if the choice rule that converts evidence for competing categories into probabilities is sensitive to small differences between evidence values close to 0. Exemplar-specific attention provided the best overall approximation of the data. Exemplar-specific specificity provided a slightly worse approximation, but it predicted better the rule-like generalization pattern observed.     

Alphabetic and nonalphabetic L1 effects in English word identification: a comparison of Korean and Chinese English L2 learners

Different writing systems in the world select different units of spoken language for mapping. Do these writing system differences influence how first language (L1) literacy experiences affect cognitive processes in learning to read a second language (L2)? Two groups of college students who were learning to read English as a second language (ESL) were examined for their relative reliance on phonological and orthographic processing in English word identification: Korean students with an alphabetic L1 literacy background, and Chinese students with a nonalphabetic L1 literacy background. In a semantic category judgment task, Korean ESL learners made more false positive errors in judging stimuli that were homophones to category exemplars than they did in judging spelling controls. However, there were no significant differences in responses to stimuli in these two conditions for Chinese ESL learners. Chinese ESL learners, on the other hand, made more accurate responses to stimuli that were less similar in spelling to category exemplars than those that were more similar. Chinese ESL learners may rely less on phonological information and more on orthographic information in identifying English words than their Korean counterparts. Further evidence supporting this argument came from a phoneme deletion task in which Chinese subjects performed more poorly overall than their Korean counterparts and made more errors that were phonologically incorrect but orthographically acceptable. We suggest that cross-writing system differences in L1s and L1 reading skills transfer could be responsible for these ESL performance differences.     

A ROWS is a ROSE: Spelling,sound, and reading

Skilled readers generally are assumed to make little or no use of words’ phonological features in visual word identification. Contrary to this assumption, college students’ performance in the present reading experiments showed large effects of stimulus word phonology. In Experiments 1 and 2, these subjects produced larger false positive error rates in a semantic categorization task when they responded to stimulus foils that were homophonic to category exemplars (e.g., ROWS for the category A FLOWER) than when they responded to spelling control foils. Additionally, in Experiment 2, this homophony effect was found under brief-exposure pattern-masking conditions, a result consistent with the possibility that phonology is an early source of constraint in word identification. Subjects did, however, correctly reject most homophone foils in Experiments 1 and 2. Experiment 3 investigated the source of this ability. The results of Experiment 3 suggest that subjects detected homophone impostors, such as ROWS, by verifying target foil spellings against their knowledge of the correct spellings of category exemplars, such as ROSE.     

A little labeling goes a long way: Semi‐supervised learning in infancy

There is considerable evidence that labeling supports infants' object categorization. Yet in daily life, most of the category exemplars that infants encounter will remain unlabeled. Inspired by recent evidence from machine learning, we propose that infants successfully exploit this sparsely labeled input through “semi‐supervised learning.” Providing only a few labeled exemplars leads infants to initiate the process of categorization, after which they can integrate all subsequent exemplars, labeled or unlabeled, into their evolving category representations. Using a classic novelty preference task, we introduced 2‐year‐old infants (n = 96) to a novel object category, varying whether and when its exemplars were labeled. Infants were equally successful whether all exemplars were labeled (fully supervised condition) or only the first two exemplars were labeled (semi‐supervised condition), but they failed when no exemplars were labeled (unsupervised condition). Furthermore, the timing of the labeling mattered: when the labeled exemplars were provided at the end, rather than the beginning, of familiarization (reversed semi‐supervised condition), infants failed to learn the category. This provides the first evidence of semi‐supervised learning in infancy, revealing that infants excel at learning from exactly the kind of input that they typically receive in acquiring real‐world categories and their names.     

Inadvertent plagiarism in young and older adults: The role of working memory capacity in reducing memory errors

Two experiments examined inadvertent plagiarism in young and older adults. Young and older adults took turns generating category exemplars in small groups, and after a short retention interval recall was tested and subjects were asked to generate new exemplars (i.e., exemplars not initially generated). When asked to generate new exemplars, older adults were more likely to repeat exemplars that had been generated earlier by others (i.e., generate-new plagiarism). When asked to recall the exemplars they had generated earlier, older adults were more likely to claim that they had generated exemplars that had been generated by others (i.e., recall-own plagiarism), and were also more likely to falsely recall exemplars that had not been generated at all. There were no age differences in confidence for items that were plagiarized on the generate-new task. Hierarchical regression analyses indicated that age differences in generate-new plagiarism and false recall were entirely mediated by measures of episodic recall and working memory capacity. We conclude that inadvertent plagiarism errors result from the failure of systematic decision processes, and that controlled attention is important for avoiding memory errors.     

A probabilistic account of exemplar and category generation

People are capable of imagining and generating new category exemplars and categories. This ability has not been addressed by previous models of categorization, most of which focus on classifying category exemplars rather than generating them. We develop a formal account of exemplar and category generation which proposes that category knowledge is represented by probability distributions over exemplars and categories, and that new exemplars and categories are generated by sampling from these distributions. This sampling account of generation is evaluated in two pairs of behavioral experiments. In the first pair of experiments, participants were asked to generate novel exemplars of a category. In the second pair of experiments, participants were asked to generate a novel category after observing exemplars from several related categories. The results suggest that generation is influenced by both structural and distributional properties of the observed categories, and we argue that our data are better explained by the sampling account than by several alternative approaches.     

Electrophysiological correlates of exemplar-specific processes in implicit and explicit memory

The present ERP study investigated the retrieval of task-irrelevant exemplar-specific information under implicit and explicit memory conditions. Subjects completed either an indirect memory test (a natural/artificial judgment) or a direct recognition memory test. Both test groups were presented with new items, identical repetitions, and perceptually different but conceptually similar exemplars of previously seen study objects. Implicit and explicit memory retrieval elicited clearly dissociable ERP components that were differentially affected by exemplar changes from study to test. In the indirect test, identical repetitions, but not different exemplars, elicited a significant ERP repetition priming effect. In contrast, both types of repeated objects gave rise to a reliable old/new effect in the direct test. The results corroborate that implicit and explicit memory fall back on distinct cognitive representation and, more importantly, indicate that these representations differ in the type of stimulus information stored. Implicit retrieval entailed obligatory access to exemplar-specific perceptual information, despite its being task irrelevant. In contrast, explicit retrieval proved to be more flexible with conceptual and perceptual information accessed according to task demands.     

Generalization hypothesis of abstract-concept learning: learning strategies and related issues in Macaca mulatta, Cebus apella, and Columba livia

The generalization hypothesis of abstract-concept learning was tested with a meta-analysis of rhesus monkeys (Macaca mulatta), capuchin monkeys (Cebus apella), and pigeons (Columba livia) learning a same/different (S/D) task with expanding training sets. The generalization hypothesis states that as the number of training items increases, generalization from the training pairs will increase and could explain the subjects' accurate novel-stimulus transfer. By contrast, concept learning is learning the relationship between each pair of items; with more training items subjects learn more exemplars of the rule and transfer better. Having to learn the stimulus pairs (the generalization hypothesis) would require more training as the set size increases, whereas learning the concept might require less training because subjects would be learning an abstract rule. The results strongly support concept or rule learning despite severely relaxing the generalization-hypothesis parameters. Thus, generalization was not a factor in the transfer from these experiments, adding to the evidence that these subjects were learning the S/D abstract concept.     

Available but unattended conceptual information in working memory: temporarily active semantic content or persistent memory for prior operations?

In 2 experiments, the authors investigated available but unattended information following working memory task demands. The experimental task presented a memory set containing exemplars from 2 conceptual categories. Following instructions to focus attention on only 1 category, priming of both categories was assessed with category comparisons of exemplar pairs. Priming was greatest for the focused category and for exemplars from the memory set (direct priming). Priming also extended to new exemplars of both categories (indirect priming) and showed little decline over more than 1 min of intervening processes. Finally, changing between category exemplars and features across memory set and comparison phases eliminated the indirect priming. These results support a persistent, operation-specific, procedural account of available but unattended conceptual information in working memory.     

On the relationship between categorical frequency estimation and cued recall

In a categorical frequency-estimation task, subjects have to indicate the number of exemplars from particular categories that have been shown on a list. In three experiments, the relationship between this task and the recall of exemplars with category names as cues was investigated. Three variables (generation, blocking of category exemplars, and provision of extralist cues) were shown to have identical effects on the two tasks. These results support a model of categorical frequency estimation in which subjects use a category name to retrieve exemplars and then base their frequency estimate on a count of the exemplars that have been retrieved.     

The generalized polymorphous concept account of graded structure in abstract categories

Abstract categories present with graded structure. The extent to which feature commonality between exemplars and category provides a satisfying account of this graded structure varies from one abstract category to the other (Hampton, 1981). We investigate whether the incorporation of features that exemplars share with external categories yields an improved account of abstract categories' graded structures. In doing so, we follow the suggestion that abstract categories are relational in nature (Goldstone, 1996; Wiemer-Hastings & Xu, 2005). The generalized polymorphous concept model, which incorporates both types of features, is found to improve the account of typicality and category membership in three of seven studied abstract categories. These three categories are found to be the most abstract, suggesting that it is appropriate to think of abstract categories as varying along a continuum of abstractness/interrelatedness rather than as a distinct type of category altogether.     

How does prior knowledge affect implicit and explicit concept learning?

Two experiments investigated the effect of prior knowledge on implicit and explicit learning. Implicit as opposed to explicit learning is sometimes characterized as unselective or purely statistical. During training, one group of participants was presented with category exemplars whose features could be tied together by integrative knowledge, whereas another group saw category exemplars with unrelated feature combinations. Half of the participants in each group learned these categories under a secondary-task condition (meant to discourage explicit learning), and the remaining half performed the categorization task under a single-task condition (meant to favour explicit learning). In a test phase, participants classified only the individual features of the training exemplars. Secondary- as opposed to single-task conditions impaired explicit but not implicit knowledge (as determined by subjective measures). Importantly, prior knowledge resulted in increased amounts of both implicit and explicit knowledge.     

A comparison of the category fluency deficits associated with Alzheimer's and Huntington's disease

The supermarket verbal fluency test of the Dementia Rating Scale (DRS) was administered to 20 patients with mild Alzheimer's disease (Mi-DAT), 20 patients with moderately severe Alzheimer's disease (Mo-DAT), 20 patients with Huntington's disease (HD), and 40 normal control subjects. The findings confirmed previous reports that Mo-DAT patients retrieved fewer words per category of supermarket items sampled and had a greater propensity to generate category labels (superordinates) than did intact controls. Similar disruptions of the structure of semantic knowledge were also noted in the fluency performances of the Mi-DAT and HD patients. The Mi-DAT patients' tendency to generate few exemplars for each category sampled suggested that a significant disruption in the structure of semantic knowledge occurred even in the earliest stages of DAT. When the present fluency findings for the HD patients were considered with previous reports of linguistic changes in this disorder, it appeared that HD patients' deterioration in semantic knowledge involved associative changes rather than the bottom-up breakdown associated with DAT.     

The role of exemplar distribution in infants'' differentiation of categories

Three experiments investigated how the frequency of exposure to particular exemplars influenced 10-month-old infants' differentiation of land and sea animals in an object-examining task. In Experiments 1 and 2, one category exemplar was presented more frequently than the others during familiarization (i.e., that exemplar was presented on 6 of 12 familiarization trials, and 3 other exemplars were each presented on 2 familiarization trials). For half of the infants, the frequent exemplar was similar to other category exemplars (e.g., a zebra if the familiarization category was land animals), and for half the frequent exemplar was not similar to many other category exemplars (e.g., a rabbit). Infants who frequently experienced the similar exemplar formed an exclusive category, and differentiated land and sea animals. Infants who frequently experienced a dissimilar exemplar, in contrast, formed an inclusive category, and failed to differentiate between land and sea animals. In Experiment 3, infants received frequent experience with a set of similar or dissimilar exemplars, and the same pattern was observed. Thus, 10-month-old infants are sensitive to the distribution of the exemplars to which they are exposed, and they form different category boundaries depending on that distribution.     

Two routes for activation in the priming of categorical coordinates

The authors operationalized category priming as participants' recognition facilitation of nonstudied, low semantically similar exemplars by studied exemplars in a category. The existing literature either does not examine the effect of studied exemplars on nonstudied exemplars in a category or fails to show an appreciable amount of category priming. In 2 experiments, the authors demonstrated a unique process to account for the category priming effect and distinguish it from the semantic priming effect, facilitation of semantically similar exemplars, in the context of a category. In Experiment 1A, the authors used a multidimensional scaling technique to examine participants' internal structure of different categories. In Experiment 1B, the authors used a lexical decision task that used these internal structures to show that semantic encoding of category exemplars causes activation of existing category knowledge in memory. Consequently, participants easily recognized nonstudied, low semantically similar exemplars in a category. However, recognition facilitation between high semantically similar exemplars did not require category knowledge activation.     

类别学习中两种学习模式的比较研究：分类学习与推理学习

采用学习迁移任务范式,使用基于单一特征的类别判断技术,比较了非线性分离结构下,分类学习和推理学习的学习效率、学习过程与策略和学习结果。结果表明：在学习效率上,分类学习比推理学习更好地习得了含有较多样例的类别知识,分类学习的速度上显著快于推理学习。在学习的过程与策略上,推理学习比分类学习更为关注类别内不同特征的相关,但在分类策略的运用上不如分类学习灵活。在学习的结果上,推理学习倾向于原型记忆,分类学习倾向于进行样例记忆,分类学习比推理学习更好地掌握了类别原型     

Distributional expectations and the induction of category structure

Previous research on how categories are learned from observation of exemplars has largely ignored the possible role of prior expectations concerning how exemplars will be distributed. The experiments reported here explored this issue by presenting subjects with category-learning tasks in which the distributions of exemplars defining the categories were varied. In Experiments 1 and 2 the distributional form of a category was found to affect speed of learning. Learning was faster when a category's distribution was normal than when it was multimodal. Also, subjects in the early stages of learning a multimodal category responded as if it were unimodal. These results suggested that subjects enter category-learning tasks with expectations of unimodal, possibly normal, distributions of exemplars. Experiments 3 and 4 attempted to manipulate subjects' prior expectations by varying the distribution of exemplars in the first of two consecutive category-learning tasks. Learning a multimodal category was influenced by the shape of a previously learned distribution and was facilitated when the earlier distribution was either multimodal or skewed, rather than normal. These results are interpreted as support for a dual-process model of category learning that incorporates the effects of prior expectations concerning exemplar distributions.     

Recognition vs. classification of ill-defined category exemplars

In Experiment 1, subjects were trained on three ill-defined categories represented by 4, 8, and 12 exemplars. Learning was followed by either a classification test on old, new, and random exemplars or a recognition test in which each learning exemplar was presented with two foils, both equidistant to the learning exemplar, but one more similar to the category prototype than the other. In Experiment 2, category size was manipulated between subjects, followed by the recognition test. Recognition data indicated that increasing the number of learning exemplars did not increase the total amount of exemplar information retained for each category. In addition, the more prototypical foil was incorrectly recognized more often than the less prototypical foil. Classification data replicated previous findings that increasing category size in learning enhances classification of new exemplars. The results were interpreted in terms of the relative amounts of categorical and idiosyncratic information that were retained from the original learning stimuli. It was proposed that categorical information can become a more dominant determinant of performance as category size increases.