Basic-level Concepts Research Articles

The hierarchical organization of semantic memory: Executive function in the processing of superordinate concepts

Research on the processing of objects at different hierarchical levels has suggested that understanding superordinate concepts (e.g. fruit), relative to basic level concepts (e.g. apple), requires greater semantic control demands. Yet, it is unclear which factors underlie this difference in executive processing. We built on previous research showing that superordinate concepts have less shared features among their members and therefore may involve higher semantic control requirements. To test this hypothesis, we developed an fMRI study in which we orthogonally manipulated feature sharedness (more shared vs. less shared) and concept level (superordinate vs. basic) in a sentence verification task. Sentences involving less shared features, relative to more shared features, significantly engaged the L lateral PFC. Importantly, sentences that included superordinate concepts, relative to those with basic level concepts, also revealed a stronger response in L lateral PFC, along with posterior temporal gyrus activation. There was also a significant interaction between feature sharedness and concept level in several PFC regions and L posterior temporal areas. The results suggest that relative to basic level concepts, processing superordinate concepts requires extra semantic control in L lateral PFC to coordinate information that is less shared by other members of the category level. These findings demonstrate that feature sharedness impacts the neural basis of semantic knowledge, and is a critical dimension in the processing of superordinate concepts.

Generating ontologies with basic level concepts from folksonomies

Abstract This paper deals with the problem of ontology generation. Ontology plays an important role in knowledge representation, and it is an artifact describing a certain reality with specific vocabulary. Recently many researchers have realized that folksonomy is a potential knowledge source for generating ontologies. Although some results have already been reported on generating ontologies from folksonomies, most of them do not consider what a more acceptable and applicable ontology for users should be, nor do they take human thinking into consideration. Cognitive psychologists find that most human knowledge is represented by basic level concepts which is a family of concepts frequently used by people in daily life. Taking cognitive psychology into consideration, we propose a method to generate ontologies with basic level concepts from folksonomies. Using Open Directory Project (ODP) as the benchmark, we demonstrate that the ontology generated by our method is reasonable and consistent with human thinking.

Conceptual Hierarchies in a Flat Attractor Network: Dynamics of Learning and Computations.

The structure of people's conceptual knowledge of concrete nouns has traditionally been viewed as hierarchical (Collins & Quillian, 1969). For example, superordinate concepts (vegetable) are assumed to reside at a higher level than basic-level concepts (carrot). A feature-based attractor network with a single layer of semantic features developed representations of both basic-level and superordinate concepts. No hierarchical structure was built into the network. In Experiment and Simulation 1, the graded structure of categories (typicality ratings) is accounted for by the flat attractor-network. Experiment and Simulation 2 show that, as with basic-level concepts, such a network predicts feature verification latencies for superordinate concepts (vegetable <is nutritious>). In Experiment and Simulation 3, counterintuitive results regarding the temporal dynamics of similarity in semantic priming are explained by the model. By treating both types of concepts the same in terms of representation, learning, and computations, the model provides new insights into semantic memory.

The general/specific breakdown of semantic memory and the nature of superordinate knowledge: Insights from superordinate and basic-level feature norms

The deterioration of semantic memory usually proceeds from more specific to more general superordinate categories, although rarer cases of superordinate knowledge impairment have also been reported. The nature of superordinate knowledge and the explanation of these two semantic impairments were evaluated from the analysis of superordinate and basic-level feature norms. The results show that, in comparison to basic-level concepts, superordinate concepts are not generally less informative and have similar feature distinctiveness and proportion of individual sensory features, but their features are less shared by their members. Results are in accord with explanations based on feature connection weights and/or concept confusability for the superordinate advantage cases. Results especially support an explanation for superordinate impairments in terms of higher semantic control requirements as related to features being less shared between concept members. Implications for patients with semantic impairments are also discussed.

The Interdisciplinary Perspective of Humane Intelligence: A Revisitation, Twenty Years Hence

Our current research involving intelligent interfaces for crisis managers, intelligence analysts, and police operations can be traced to a seminal position first published twenty years ago entitled Humane Intelligence. This perspective captures an adaptive systems philosophy through a wholistic blending of human and artificial intelligence designed to function within information surrounds. The original paper was foundational for prescribing how human factors might influence artificial intelligence, specifically within the domain of fighter pilots performing tactical operations. This paper revisits this theoretical approach to intelligent systems by reviewing principles, attributes, and ideas using a contemporary lens of human factors and ergonomics culture. After reviewing basic-level concepts the paper presents examples of how humane intelligence has both evolved and influenced several lines of research including concept mapping ontologies, fuzzy cognitive models, affective computing, scaled world technology, and human-agent architectures.

Semantic feature production norms for a large set of living and nonliving things

Semantic features have provided insight into numerous behavioral phenomena concerning concepts, categorization, and semantic memory in adults, children, and neuropsychological populations. Numerous theories and models in these areas are based on representations and computations involving semantic features. Consequently, empirically derived semantic feature production norms have played, and continue to play, a highly useful role in these domains. This article describes a set of feature norms collected from approximately 725 participants for 541 living (dog) and nonliving (chair) basic-level concepts, the largest such set of norms developed to date. This article describes the norms and numerous statistics associated with them. Our aim is to make these norms available to facilitate other research, while obviating the need to repeat the labor-intensive methods involved in collecting and analyzing such norms. The full set of norms may be downloaded from www.psychonomic.org/archive.

When Dogs Can Fly: Level of Abstraction as a Cue to Metaphorical Use of Verbs

How do people recognize metaphors? In nominal metaphors, such as "My lawyer is a shark," the metaphor vehicle "shark" refers to an abstract category of predatory creatures, not to the basic-level concept, the literal fish we call "shark." People can use the level of abstraction of the metaphor vehicle (shark) as a cue that the expression is intended metaphorically rather than literally (Glucksberg & Keysar, 1990). Can the metaphorical use of verbs, as in "he hopped on his bike and flew home" be recognized in the same way? We investigated whether the level of abstraction of a verb's referent provides a cue that the verb is used metaphorically rather than literally. We varied level of abstraction of verb use, and obtained judgments of metaphoricity as a function of abstraction level. As with nouns, verbs that are understood at a higher level of abstraction are rated as more metaphorical than when the same verbs could be interpreted at the basic (literal) level. Furthermore, this effect is graded: the higher the level of abstraction, the higher the rated metaphoricity. These findings suggest that people use level of abstraction as a cue to metaphoricity for both nominal and predicative metaphors.

Hard Words

How do children acquire the meaning of words? And why are words such as know harder for learners to acquire than words such as dog or jump? We suggest that the chief limiting factor in acquiring the vocabulary of natural languages consists not in overcoming conceptual difficulties with abstract word meanings but rather in mapping these meanings onto their corresponding lexical forms. This opening premise of our position, while controversial, is shared with some prior approaches. The present discussion moves forward from there to a detailed proposal for how the mapping problem for the lexicon is solved, as well as a presentation of experimental findings that support this account. We describe an overlapping series of steps through which novices move in representing the lexical forms and phrase structures of the exposure language, a probabilistic multiple-cue learning process known as syntactic bootstrapping. The machinery is set in motion by word-to-world pairing, a procedure available to novices from the onset, one that is efficient for a stock of lexical items (mostly nouns) that express concrete basic-level concepts. Armed with this foundational stock of "easy" words, learners achieve further lexical knowledge by an arm-over-arm process in which successively more sophisticated representations of linguistic structure are built. Lexical learning can thereby proceed by adding structure-to-world mapping methods to the earlier available machinery, enabling efficient learning of abstract items-the "hard" words. Thus acquisition of the lexicon and the clause-level syntax are interlocked throughout their course, rather than being distinct and separable parts of language learning. We concentrate detailed attention on two main questions. The first is how syntactic information, seemingly so limited, can affect word learning so pervasively. The second is how multiple sources of information converge to solve lexical learning problems for two types of verbs that pose principled obstacles for word-to-world mapping procedures. These types are perspective verbs (e.g., chase and flee) and credal verbs (e.g., think and know). As we discuss in closing, the outcome of the hypothesized learning procedure is a highly lexicalized grammar whose usefulness does not end with successful acquisition of the lexicon. Rather, these detailed and highly structured lexical representations serve the purposes of the incremental multiple-cue processing machinery by which people produce speech and parse the speech that they hear.

Basic-level nouns: first learned but misunderstood.

It is commonly believed that first-learned words correspond with first-learned categories (both described as 'basic level') leading to the belief that language acquisition is a reasonably good indicator of early cognition. However, toddlers often overextend their first words. Do these errors reflect their comprehension? Two experiments were conducted in order to examine two-year-olds' production and comprehension of basic-level terms. The results showed overextensions both in production (e.g. children labelled a rocket 'airplane') and comprehension (e.g. they pointed to a rocket when airplane was requested). One reason toddlers extend labels to a wider conceptual domain is because they have not clearly differentiated basic-level concepts from related conceptual categories.

Inductive Reasoning in Alzheimer's Disease

We evaluated knowledge of basic level and superordinate semantic relations and the role of cognitive resources during inductive reasoning in probable Alzheimer's disease (AD). Nineteen mildly demented AD patients and 17 healthy control subjects judged the truthfulness of arguments with a premise and a conclusion that contain familiar concepts coupled with “blank” predicates, such as “Spiders contain phosphatidylcholine; therefore all insects contain phosphatidylcholine.” Like healthy control subjects, AD patients were relatively insensitive to the typicality of the premise category when judging the strength of arguments with a conclusion containing a basic-level concept, but were relatively sensitive to typicality during judgments of arguments containing a superordinate in the conclusion. Moreover, AD patients resembled control subjects in judging arguments with an immediate superordinate in the conclusion compared to arguments with a distant superordinate. AD patients differed from control subjects because they could not take advantage of two premises in an argument containing basic-level concepts. We conclude that semantic knowledge is sufficiently preserved in AD to support inductive reasoning, but that limited cognitive resources may interfere with AD patients' ability to consider the entire spectrum of information available during semantic challenges.

Advancing Downward to the Basic Level

In 3 experiments, we studied whether infants and young children understand various basic-level conceptual distinctions in the domains of household artifacts, animals, and vehicles. Using small replicas, we modeled events such as washing dishes in a sink for children 14, 19, and 24 months old, and then gave them an exemplar from the same basic-level concept (another sink) and an exemplar of another concept from the same domain (bathtub). We measured which object they used to imitate the event. Fourteen-month-olds did not differentiate among basic-level categories in any of these domains, for example, washing dishes in both a tub and a sink, and putting both a rabbit and a bird in a nest. By 19 months, inappropriate behavior was greatly reduced for household artifacts and for vehicles, but not for animals. By 24 months, performance was mainly appropriate for all 3 domains. It was also shown that although 14-month-olds are not making many conceptual distinctions at the basic level, they are nevertheless beginning to make some broader conceptual distinctions among artifacts.

Lexical semantic and associative priming in Alzheimer's disease.

Semantic memory impairment was investigated in patients with probable Alzheimer's disease (AD) using a threshold oral word reading task to assess priming of different lexical relationships. Healthy elderly controls showed significant priming for associatively related nouns (tempest-teapot) and also for nouns semantically related either because both designate basic-level exemplars of a common superordinate category (cousin-nephew) or because the target names the superordinate category of the prime (daughter-relative). AD patients, in contrast, showed preserved priming of lexical associates but impaired priming of certain semantic relationships. They showed no priming between words designating coordinate exemplars within a category, despite preserved priming of the superordinate category label. Findings are consistent with the view that at least part of the semantic deficit in AD is due to disruption of semantic knowledge that affects relationships among basic-level concepts, more than the relationships between these concepts and their corresponding superordinate category of membership.

Lexical semantic and associative priming in Alzheimer's disease.

Semantic memory impairment was investigated in patients with probable Alzheimer's disease (AD) using a threshold oral word reading task to assess priming of different lexical relationships. Healthy elderly controls showed significant priming for associatively related nouns (tempest-teapot) and also for nouns semantically related either because both designate basic-level exemplars of a common superordinate category (cousin-nephew) or because the target names the superordinate category of the prime (daughter-relative). AD patients, in contrast, showed preserved priming of lexical associates but impaired priming of certain semantic relationships. They showed no priming between words designating coordinate exemplars within a category, despite preserved priming of the superordinate category label. Findings are consistent with the view that at least part of the semantic deficit in AD is due to disruption of semantic knowledge that affects relationships among basic-level concepts, more than the relationships between these concepts and their corresponding superordinate category of membership.

High level object recognition without an anterior inferior temporal lobe

Seven individuals with unilateral anterior inferior temporal (AIT) lobectomies performed two types of shape recognition tasks with line drawings of 3D objects briefly presented in either the left or the right visual field. In one task, subjects named familiar objects in a name priming paradigm. In the other task, subjects judged whether two objects, presented sequentially with an intervening mask, were the same or different in shape, disregarding differences in orientation of up to 60° in depth. They could not use names or basic level concepts to do the matching as the stimuli were either nonsense objects or, if familiar objects, were of same name–different-shaped exemplars on different trials. The disadvantage of presenting an image to the lobectomized hemisphere was negligible in both tasks. Two non-exclusive possibilities are suggested by this result: (a) Object recognition is completed posterior to AIT, likely at the temporal–occipital boundary, with no deleterious retrograde effects on object recognition from the AIT section, or (b) Callosal transfer of object information prior to AIT is completely efficient. These results, along with results of single unit recording and lesion experiments in the monkey, PET and MRI imaging in humans, and a plausibility argument based on the pattern of callosal connections suggest both are correct. Rather than mediating real-time object recognition, AIT may code representations for visual episodes and scenes.

The object in perspective

From the Conclusions The philosophy of language has indicated that there are serious omissions in the Cartesian theory of categorisation; notably, that members of a category are related by family resemblances and not necessarily by shared properties between members of categories. As observed by Lakoff, these findings are also in conflict with other popular positions held in the Artificial Intelligence community which view human reason as completely abstract, as well as assuming that every cognitive process is algorithmic in nature; that thought is purely a matter of symbol manipulation [9]. The former of these does not agree with findings of categorisation phenomena which imply that basic level concepts are characterised by basic-level perception and imaging capacities. The problem for the latter position is that the only way it can make sense of the understanding of two different conceptual systems is by means of the translation of one into another, or the translation of both into a third system. Understanding in terms of experience does not eliminate such organisational differences, so this is a problem which needs to be addressed. Self-organising activity will involve, minimally, the management of information, its representation both within a given organisation as well as across organisations and the effect of system evolution on the object. The object model, as currently based on the Cartesian view of categorisation, may not be sufficient for these requirements. Instead, it may need to accommodate what have been termed the holistic aspects of knowledge representation and categorisation. In any event, in order to define a set of requirements for self-organising architectures, we will need to bear in mind not only the object, but also what is implied by the object.

Metric invariance in object recognition: A review and further evidence.

Phenomenologically, human shape recognition appears to be invariant with changes of orientation in depth (up to parts occlusion), position in the visual field, and size. Recent versions of template theories (e.g., Ullman, 1989; Lowe, 1987) assume that these invariances are achieved through the application of transformations such as rotation, translation, and scaling of the image so that it can be matched metrically to a stored template. Presumably, such transformations would require time for their execution. We describe recent priming experiments in which the effects of a prior brief presentation of an image on its subsequent recognition are assessed. The results of these experiments indicate that the invariance is complete: The magnitude of visual priming (as distinct from name or basic level concept priming) is not affected by a change in position, size, orientation in depth, or the particular lines and vertices present in the image, as long as representations of the same components can be activated. An implemented seven layer neural network model (Hummel & Biederman, 1992) that captures these fundamental properties of human object recognition is described. Given a line drawing of an object, the model activates a viewpoint-invariant structural description of the object, specifying its parts and their interrelations. Visual priming is interpreted as a change in the connection weights for the activation of: a) cells, termed geon feature assemblies (GFAs), that conjoin the output of units that represent invariant, independent properties of a single geon and its relations (such as its type, aspect ratio, relations to other geons), or b) a change in the connection weights by which several GFAs activate a cell representing an object.

The concept of love viewed from a prototype perspective.

Even if superordinate concepts (such as fruit, vehicle, sport) are prototypically organized, basic-level concepts (such as apple, truck, hockey) might be classically defined in terms of individually necessary and jointly sufficient features. A series of 6 studies examined 1 basic-level concept in the domain of emotion, love, and found that it is better understood from a prototype than a classical perspective. The natural language concept of love has an internal structure and fuzzy borders: Maternal love, romantic love, affection, love of work, self-love, infatuation, and other subtypes of love can be reliably ordered from better to poorer examples of love. In turn, each subtype's goodness as an example of love (prototypicality) was found to predict various indices of its cognitive processing. Implications for a scientific definition and typology of love are discussed. Prototype theory has provided insights into concepts central to psychology, including behavioral act (Buss & Craik, 1983), personality trait (Cantor & Mischel, 1979), intelligence (Neisser, 1979), social situation (Cantor, Mischel, & Schwartz, 1982), and environmental setting (Tversky & Hemenway, 1983). The theory has inspired important new approaches to psychiatric diagnosis (Cantor, Smith, French, & Mezzich, 1980) and personality assessment (Broughton, 1984). The studies reported in this article are part of a larger project designed to explore the applicability of prototype theory to the domain of emotion concepts. We had two related purposes. The first was to use the tools of prototype theory to throw light on the elusive concept of love. We shall have something to say about both the definition of and typologies of love. The second purpose concerned prototype theory itself. Prior demonstrations of a prototype structure in concepts focused on superordinate concepts, such as fruit, vehicle, or emotion. Would the same results occur, we asked, for basic-level concepts, such as apple, truck, or love? We shall begin with the second issue.

Semantic encoding of pictures and words: Some neuropsychological observations

Abstract We describe a brain-damaged patient whose comprehension of written and spoken words is severely impaired; in general, MP can only determine the superordinate category of a word and is unable to obtain a more precise semantic description. Her understanding of pictures, in contrast, was found to be much more detailed. Further investigation of this ability, however, revealed definite constraints on the nature of the semantic description that the patient could extract from pictures. MP has access to knowledge of the basic-level concept for a picture (defined as the most inclusive member of a category for which a concrete image of the object as a whole can be formed), but appears unable to go beyond this level to compute subordinate information. In discussing the results, we evaluate possible differences in the way pictures and words are normally processed for meaning.