An integrated account of generalization across objects and features

Humans routinely make inductive generalizations about unobserved features of objects. Previous accounts of inductive reasoning often focus on inferences about a single object or feature: accounts of causal reasoning often focus on a single object with one or more unobserved features, and accounts of property induction often focus on a single feature that is unobserved for one or more objects. We explore problems where people must make inferences about multiple objects and features, and propose that people solve these problems by integrating knowledge about features with knowledge about objects. We evaluate three computational methods for integrating multiple systems of knowledge: the output combination approach combines the outputs produced by these systems, the distribution combination approach combines the probability distributions captured by these systems, and the structure combination approach combines a graph structure over features with a graph structure over objects. Three experiments explore problems where participants make inferences that draw on causal relationships between features and taxonomic relationships between animals, and we find that the structure combination approach provides the best account of our data.     

Feature-based versus category-based induction with uncertain categories

Previous research has suggested that when feature inferences have to be made about an instance whose category membership is uncertain, feature-based inductive reasoning is used to the exclusion of category-based induction. These results contrast with the observation that people can and do use category-based induction when category membership is known. The present experiments examined the conditions that drive feature-based and category-based strategies in induction under category uncertainty. Specifically, 2 experiments investigated whether reliance on feature-based inductive strategies is a product of the lack of coherence in the categories used in previous research or is due to the use of a decision-only induction procedure. Experiment 1 found that feature-based reasoning remained the preferred strategy even when categories with relatively high internal coherence were used. Experiment 2 found a shift toward category-based reasoning when participants were trained to classify category members prior to feature induction. Together, these results suggest that an appropriate conceptual representation must be formed through experience with a category before it is likely to be used as a basis for feature induction.     

Perceived distributions of the characteristics of in-group and out-group members: empirical evidence and a computer simulation

This research studied 2 properties of perceived distributions of the characteristics of social category members: the probability of differentiating (making distinctions) among category members and the perceived variability (variance) of category members. The results of 4 experiments supported the hypothesis that greater familiarity with a social group leads to greater perceived differentiation and variability regarding that group. In-group members formed more differentiated and variable distributions for groups defined by age and more differentiated distributions for groups defined by nationality. For gender (where students were roughly equally familiar with people of both genders), no in-group--out-group differences occurred. Also, students perceived greater differentiation and variability among classmates over the course of a semester. To explain these results, we developed PDIST, a multiple exemplar model that assumes that people form perceived distributions by activating a set of category exemplars and then judging the relative likelihoods of different feature values on the basis of the relative activation strengths of these feature values. The results of a computer simulation experiment indicated that PDIST is sufficient to explain the results of our 4 experiments. According to the perceived distributions formed by PDIST, increasing familiarity leads to greater differentiation and variability, has a concave impact, and has greater impact on differentiation than on variability.     

Feature inference with uncertain categorization: Re-assessing Anderson’s rational model

A key function of categories is to help predictions about unobserved features of objects. At the same time, humans are often in situations where the categories of the objects they perceive are uncertain. In an influential paper, Anderson (Psychological Review, 98(3), 409–429, 1991) proposed a rational model for feature inferences with uncertain categorization. A crucial feature of this model is the conditional independence assumption—it assumes that the within category feature correlation is zero. In prior research, this model has been found to provide a poor fit to participants’ inferences. This evidence is restricted to task environments inconsistent with the conditional independence assumption. Currently available evidence thus provides little information about how this model would fit participants’ inferences in a setting with conditional independence. In four experiments based on a novel paradigm and one experiment based on an existing paradigm, we assess the performance of Anderson’s model under conditional independence. We find that this model predicts participants’ inferences better than competing models. One model assumes that inferences are based on just the most likely category. The second model is insensitive to categories but sensitive to overall feature correlation. The performance of Anderson’s model is evidence that inferences were influenced not only by the more likely category but also by the other candidate category. Our findings suggest that a version of Anderson’s model which relaxes the conditional independence assumption will likely perform well in environments characterized by within-category feature correlation.     

Inference using categories

How do people use category membership and similarity for making inductive inferences? The authors addressed this question by examining the impact of category labels and category features on inference and classification tasks that were designed to be comparable. In the inference task, participants predicted the value of a missing feature of an item given its category label and other feature values. In the classification task, participants predicted the category label of an item given its feature values. The results from 4 experiments suggest that category membership influences inference even when similarity information contradicts the category label. This tendency was stronger when the category label conveyed class inclusion information than when the label reflected a feature of the category. These findings suggest that category membership affects inference beyond similarity and that category labels and category features are 2 different things.     

The influence of facial characteristics on children's age perceptions

To examine the impact of age-related variations in facial characteristics on children's age judgments, two experiments were conducted in which craniofacial shape and facial wrinkling were independently manipulated in stimulus faces as sources of age information. Using a paired-comparisons task, children between the ages of 2 1/2 and 6 were asked to make age category as well as relative age judgments of stimulus faces. Preschool-aged children were able to use variations in craniofacial profile shape, frontal face feature vertical placement, or facial wrinkling to identify the age category of a stimulus person. Children were also able to identify the older, but not the younger, of two faces on the basis of facial wrinkling, a finding consistent with previously demonstrated limitations in young children's use of relative age terms. The results were discussed in the context of research which reveals parallel effects of craniofacial shape and wrinkling on the age judgments of adults.     

维度的结合与分离对归类不确定性预测的影响

探讨在归类不确定的情境下目标与预测特征两个维度的结合或分离对被试特征预测的影响。共包括 3个实验 :实验 1在Murphy和Ross的研究的基础上进一步提高非靶类型中目标及预测特征的基本概率 ,考察被试的特征预测是否会受非靶类型信息的影响。实验 2探讨非靶类型的目标与预测特征结合与否是否会影响被试预测特征时对非靶类型信息的使用。实验 3探讨提高靶类型中目标与预测特征结合的比例是否影响被试对特征的预测。结果表明 :当非靶类型中目标与关键特征处于分离的状态时 ,被试在进行特征预测时没有利用非靶类型的信息 ,符合单类说的假设 ;而当非靶类型中目标与关键特征结合时 ,被试在进行特征预测时则会利用非靶类型的信息 ,符合Bayesian规则 ;靶类型中的目标与关键特征结合的比例提高 ,被试对特征预测的概率也随之提高。据此 ,本研究将目标与预测特征结合比例这个变量加入Bayesian规则的计算公式 ,对该预测模型进行了修正     

A Two-Stage Model of Category Construction

The current consensus is that most natural categories are not organized around strict definitions (a list of singly necessary and jointly sufficient features) but rather according to a family resemblance (FR) principle: Objects belong to the same category because they are similar to each other and dissimilar to objects in contrast categories. A number of computational models of category construction have been developed to provide an account of how and why people create FR categories (Anderson, 1990; Fisher, 1987). Surprisingly, however, only a few experiments on category construction or free sorting have been run and they suggest that people do not sort examples by the FR principle. We report several new experiments and a two-stage model for category construction. This model is contrasted with a variety of other models with respect to their ability to account for when FR sorting will and will not occur. The experiments serve to identify one basis for FR sorting and to support the two-stage model. The distinctive property of the two-stage model is that it assumes that people impose more structure than the examples support in the first stage and that the second stage adjusts for this difference between preferred and perceived structure. We speculate that people do not simply assimilate probabilistic structures but rather organize them in terms of discrete structures plus noise.     

Abstraction of prototypical information by adults and 10-month-old infants

The primary purpose of this study was to investigate whether preverbal infants, when presented with exemplars of an artificially constructed category, would abstract a prototypical representation of the category, and if so, whether this representation was formed by either "counting" or "averaging" the features that were varying among category members. Two experiments are reported. In Experiment 1, a set of stimuli was developed and tested for which it was demonstrated that adult subjects would readily abstruct either a modal or an average prototypical representation. The type of representation abstracted was found to be dependent on the discriminability of the feature values. In Experiment 2, 10-mo.-old infants were tested using a habituation paradigm with the stimuli developed in the first experiment. The results of this study indicated that the infants were also able to abstract the featural information that was varying among the exemplars of the category, and the infants formed an internal represenatation of the category by averaging feature values. Thus, the results clearly imply that infants are able to constructively process visual information and hence take a more active role in category formation than had been previously believed.     

Where to look first for an explanation of induction with uncertain categories

Recent research has examined how people predict unobserved features of an object when its category membership is ambiguous. The debate has focused on whether predictions are based solely on information from the most likely category, or whether information from other possible categories is also used. In the present experiment, we compared these category-based approaches with feature conjunction reasoning, where predictions are based on a comparison among exemplars (rather than categories) that share features with a target object. Reasoning strategies were assessed by examining patterns of feature prediction and by using an eye gaze measure of attention during induction. The main findings were (1) the majority of participants used feature conjunction rather than categorical strategies, (2) people predominantly gazed at the exemplars that were most similar to the target object, and (3) although people gazed most at the most probable category to which an object could belong, they also attended to other plausible category alternatives during induction. These findings question the extent to which category-based reasoning is used for induction when category membership is uncertain.     

Feature assignment in perception of auditory figure

Because the environment often includes multiple sounds that overlap in time, listeners must segregate a sound of interest (the auditory figure) from other co-occurring sounds (the unattended auditory ground). We conducted a series of experiments to clarify the principles governing the extraction of auditory figures. We distinguish between auditory "objects" (relatively punctate events, such as a dog's bark) and auditory "streams" (sounds involving a pattern over time, such as a galloping rhythm). In Experiments 1 and 2, on each trial 2 sounds-an object (a vowel) and a stream (a series of tones)-were presented with 1 target feature that could be perceptually grouped with either source. In each block of these experiments, listeners were required to attend to 1 of the 2 sounds, and report its perceived category. Across several experimental manipulations, listeners were more likely to allocate the feature to an impoverished object if the result of the grouping was a good, identifiable object. Perception of objects was quite sensitive to feature variation (noise masking), whereas perception of streams was more robust to feature variation. In Experiment 3, the number of sound sources competing for the feature was increased to 3. This produced a shift toward relying more on spatial cues than on the potential contribution of the feature to an object's perceptual quality. The results support a distinction between auditory objects and streams, and provide new information about the way that the auditory world is parsed.     

Convention in joint activity

Conventional behaviors develop from practice for regularly occurring problems of coordination within a community of actors. Reusing and extending conventional methods for coordinating behavior is the task of everyday reasoning .
The computational model presented in the paper details the emergence of convention in circumstances where there is no ruling body of knowledge developed by prior generations of actors within the community to guide behavior. The framework we assume combines social theories of cognition with human information processing models that have been developed within Cognitive Science. The model presented reflects both elements of the framework. Conventional behaviors are partially coded in the predisposition of participants in a joint activity to expect certain points of coordination to develop during the course of the activity. The expected points of coordination that are commonly assumed form a design for an activity . Because of uncertainty, interruptions, and numerous other opportunities to get off-track and out-of-synch, the participants must work jointly and continuously to achieve conventional coordination.
One feature of the model is that the community improves its performance despite the fact that individual actors reason independently about their experiences. Another important feature of the model is that the mechanisms for improving behavior are tied to the memory function of individual actors. A third important feature is that the social interaction among the participants simplifies and drives the everyday reasoning processes. An analysis of a large set of computational experiments supports the theoretical position that is developed regarding everyday reasoning and convention.     

Speeded induction under uncertainty: The influence of multiple categories and feature conjunctions

When people are uncertain about the category membership of an item (e.g., Is it a dog or a dingo?), research shows that they tend to rely only on the dominant or most likely category when making inductions (e.g., How likely is it to befriend me?). An exception has been reported using speeded induction judgments where participants appeared to use information from multiple categories to make inductions (Verde, Murphy, &; Ross, 2005). In two speeded induction studies, we found that participants tended to rely on the frequency with which features co-occurred when making feature predictions, independently of category membership. This pattern held whether categories were considered implicitly (Experiment 1) or explicitly (Experiment 2) prior to feature induction. The results converge with other recent work suggesting that people often rely on feature conjunction information, rather than category boundaries, when making inductions under uncertainty.     

Noncategorical approaches to feature prediction with uncertain categories

In four experiments, we investigated how people make feature predictions about objects whose category membership is uncertain. Artificial visual categories were presented and remained in view while a novel instance with a known feature, but uncertain category membership was presented. All four experiments showed that feature predictions about the test instance were most often based on feature correlations (referred to as feature conjunction reasoning). Experiment 1 showed that feature conjunction reasoning was generally preferred to category-based induction in a feature prediction task. Experiment 2 showed that people used all available exemplars to make feature conjunction predictions. Experiments 3 and 4 showed that the preference for predictions based on feature conjunction persisted even when category-level information was made more salient and inferences involving a larger number of categories were required. Little evidence of reasoning based on the consideration of multiple categories (e.g., Anderson, (Psychological Review, 98:409–429, 1991)) or the single, most probable category (e.g., Murphy & Ross, (Cognitive Psychology, 27:148–193, 1994)) was found.     

Classification in well-defined and ill-defined categories: evidence for common processing strategies

Early work in perceptual and conceptual categorization assumed that categories had criterial features and that category membership could be determined by logical rules for the combination of features. More recent theories have assumed that categories have an ill-defined structure and have prosposed probabilistic or global similarity models for the verification of category membership. In the experiments reported here, several models of categorization were compared, using one set of categories having criterial features and another set having an ill-defined structure. Schematic faces were used as exemplars in both cases. Because many models depend on distance in a multidimensional space for their predictions, in Experiment 1 a multidimensional scaling study was performed using the faces of both sets as stimuli, In Experiment 2, subjects learned the category membership of faces for the categories having criterial features. After learning, reaction times for category verification and typicality judgments were obtained. Subjects also judged the similarity of pairs of faces. Since these categories had characteristic as well as defining features, it was possible to test the predictions of the feature comparison model (Smith et al.), which asserts that reaction times and typicalities are affected by characteristic features. Only weak support for this model was obtained. Instead, it appeared that subjects developed logical rules for the classification of faces. A characteristic feature affected reaction times only when it was part of the rule system devised by the subject. The procedure for Experiment 3 was like that for Experiment 2, but with ill-defined rather than well-defined categories. The obtained reaction times had high correlations with some of the models for ill-defined categories. However, subjects' performance could best be described as one of feature testing based on a logical rule system for classification. These experiments indicate that whether or not categories have criterial features, subjects attempt to develop a set of feature tests that allow for exemplar classification. Previous evidence supporting probabilistic or similarity models may be interpreted as resulting from subjects' use of the most efficient rules for classification and the averaging of responses for subjects using different sets of rules.     

Task-relevant perceptual features can define categories in visual memory too

Although Konkle, Brady, Alvarez, and Oliva (2010, Journal of Experimental Psychology: General, 139(3), 558) claim that visual long-term memory (VLTM) is organized on underlying conceptual, not perceptual, information, visual memory results from visual search tasks are not well explained by this theory. We hypothesized that when viewing an object, any task-relevant visual information is critical to the organizational structure of VLTM. In two experiments, we examined the organization of VLTM by measuring the amount of retroactive interference created by objects possessing different combinations of task-relevant features. Based on task instructions, only the conceptual category was task relevant or both the conceptual category and a perceptual object feature were task relevant. Findings indicated that when made task relevant, perceptual object feature information, along with conceptual category information, could affect memory organization for objects in VLTM. However, when perceptual object feature information was task irrelevant, it did not contribute to memory organization; instead, memory defaulted to being organized around conceptual category information. These findings support the theory that a task-defined organizational structure is created in VLTM based on the relevance of particular object features and information.     

Overcoming default categorical bias in spatial memory

In the present study, we investigated whether a strong default categorical bias can be overcome in spatial memory by using alternative membership information. In three experiments, we tested location memory in a circular space while providing participants with an alternative categorization. We found that visual presentation of the boundaries of the alternative categories (Experiment 1) did not induce the use of the alternative categories in estimation. In contrast, visual cuing of the alternative category membership of a target (Experiment 2) and unique target feature information associated with each alternative category (Experiment 3) successfully led to the use of the alternative categories in estimation. Taken together, the results indicate that default categorical bias in spatial memory can be overcome when appropriate cues are provided. We discuss how these findings expand the category adjustment model (Huttenlocher, Hedges, & Duncan, 1991) in spatial memory by proposing a retrieval-based category adjustment (RCA) model.     

Linearity and morphological structure in derived words: evidence from category decision.

The present article investigates the influences of word-internal category information and linearity on the processing of morphologically complex words. Three sets of Finnish derived nouns consisting of a noun, adjective or verb stem, and a suffix were submitted to two on-line category decision tasks, where the subjects had to decide whether the word was a noun or a verb respectively. Both experiments show a stem-suffix category mismatch effect manifesting itself in prolonged reaction time latencies. Thus, the results indicate that the order of presentation and the word-internal structural information are intimately tied together also in the visual processing of morphological complex words.