Roger W. Schvaneveldt

Functions of graphemic and phonemic codes in visual word-recognition

Previous investigators have argued that printed words are recognized directly from visual representations and/or phonological representations obtained through phonemic recoding. The present research tested these hypotheses by manipulating graphemic and phonemic relations within various pairs of letter strings. Ss in two experiments classified the pairs as words or nonwords. Reaction times and error rates were relatively small for word pairs (e.g., BRIBE-TRIBE) that were both graphemically, and phonemically similar. Graphemic similarity alone inhibited performance on other word pairs (e.g., COUCH-TOUCH). These and other results suggest that phonological representations play a significant role in visual word recognition and that there is a dependence between successive phonemic-encoding operations. An encoding-bias model is proposed to explain the data.

Lexical ambiguity, semantic context, and visual word recognition.

Some alternative hypotheses about the recognition of ambiguous words are considered. According to the selective-access hypothesis, prior semantic context biases people to access one meaning of an ambiguous word rather than another in lexical memory during recognition. In contrast, the nonselectiveaccess hypothesis states that all meanings of the word are accessed regardless of the context. We tested certain versions of these hypotheses by having students decide whether selected strings of letters were English words. The stimuli included test sequnces of three words in which the second word had two distinct possible meanings, whereas the first and third words were related to these meanings in various ways. When the first and third words were related to the same meaning of the ambiguous second word (e.g., SAVE-BANK-MONEY), the reaction time to recognize the third word decreased. But when the first and third words were related to different meanings of the second word (e.g., RIVER-BANK-MONEY), the reaction time for the third word was not reliably different from a control sequence with unrelated words. These and other data favor the selective-access hypothesis. Selective access to lexical memory is discussed in relation to models of word recognition.

Network Structures in Proximity Data

Publisher Summary This chapter discusses network representations and their relationship to proximity data. Proximity data are commonplace in the social and behavioral sciences. Networks have several properties that should be of value in representing the structure in proximity data. Networks reduce a large number of pairwise proximities to a smaller set of links. Compared to spatial scaling methods, networks focus on the closely related entities. Graph theory is the mathematical study of structures consisting of nodes with links connecting some pairs of nodes. In applications of networks, the nodes usually represent entities and the links represent pairwise relations among the entities. Because a set of nodes can be connected by links in many possible ways, a wide variety of structures can be represented by graphs. The methods used to obtain the proximity data used in the Pathfinder analyses are straightforward and analogous to methods employed to obtain data for cluster analysis

Measuring the structure of expertise

This report reviews work on defining and measuring conceptual structures of expert and novice fighter pilots. Individuals with widely varying expertise were tested. Cognitive structures were derived using multidimensional scaling (MDS) and link-weighted networks (Pathfinder). Experience differences among pilots were reflected in the conceptual structures. Detailed analyses of individual differences point to factors that distinguish experts and novices. Analysis of individual concepts identified areas of agreement and disagreement in the knowledge structures of experts and novices. Applications in selection, training and knowledge engineering are discussed.

Reflective Random Indexing and indirect inference: A scalable method for discovery of implicit connections

The discovery of implicit connections between terms that do not occur together in any scientific document underlies the model of literature-based knowledge discovery first proposed by Swanson. Corpus-derived statistical models of semantic distance such as Latent Semantic Analysis (LSA) have been evaluated previously as methods for the discovery of such implicit connections. However, LSA in particular is dependent on a computationally demanding method of dimension reduction as a means to obtain meaningful indirect inference, limiting its ability to scale to large text corpora. In this paper, we evaluate the ability of Random Indexing (RI), a scalable distributional model of word associations, to draw meaningful implicit relationships between terms in general and biomedical language. Proponents of this method have achieved comparable performance to LSA on several cognitive tasks while using a simpler and less computationally demanding method of dimension reduction than LSA employs. In this paper, we demonstrate that the original implementation of RI is ineffective at inferring meaningful indirect connections, and evaluate Reflective Random Indexing (RRI), an iterative variant of the method that is better able to perform indirect inference. RRI is shown to lead to more clearly related indirect connections and to outperform existing RI implementations in the prediction of future direct co-occurrence in the MEDLINE corpus.

What Is Learned From Artificial Grammars? Transfer Tests of Simple Association

Ss were trained on letter pairs or letter strings in an artificial grammar learning paradigm to determine the extent to which implicit learning is driven by simple associative knowledge. Learning on strings resulted in sensitivity to violations of grammaticality and in transfer to a changed letter set. Learning on letter pairs resulted in less sensitivity and no transfer. Discrepancies in performance were later reduced, but not eliminated, by equating the task demands of the conditions during learning. A direct test of associative knowledge showed that training on letter pairs resulted in knowledge of legal bigrams, but this knowledge was only weakly related to violation sensitivity. The experiments demonstrate that knowledge of isolated associations is sufficient to support some learning, but this knowledge cannot explain the more abstract knowledge that results from learning on complete exemplars.

The basis of transfer in artificial grammar learning

In two experiments, we examined the extent to which knowledge of sequential dependencies and/or patterns of repeating elements is used during transfer in artificial grammar learning. According to one view of transfer, learners abstract the grammar’s sequential dependencies and then learn a mapping to new vocabulary at test (Dienes, Altmann, & Gao, 1999). Elements that are repeated have no special status on this view, and so a logical prediction is that learners should transfer as well after exposure to a grammar without repetitions as after exposure to a grammar with them. On another view, repetition structure is the very basis of transfer (Brooks & Vokey, 1991; Mathews & Roussel, 1997). Learners were trained on grammars with or without repeating elements to test these competing views. Learners demonstrated considerable knowledge of sequential dependencies in their training vocabulary but did not use such knowledge to transfer to a new vocabulary. Transfer only occurred in the presence of repetition structure, demonstrating this to be the basis of transfer.

Graph theoretic foundations of pathfinder networks

Abstract This paper is primarily expository, relating elements of graph theory to a computational theory of psychological similarity (or dissimilarity). A class of networks called Pathfinder networks (PFNETs) is defined. PFNETs are derived from estimates of dissimilarity for pairs of entities. Thus, PFNETs can be used to reveal aspects of the structure inherent in a set of pairwise estimates of dissimilarity. In order to accommodate different assumptions about the nature of the measurement scale (i.e. ordinal, interval, ratio) underlying the data, the Minkowski r-metric (also known as the L norm) is adapted to computing distances in networks. PFNETs are derived from data by: (1) regarding the matrix of dissimilarities as a network adjacency matrix (the DATANET); (2) computing the distance matrix (or r-distance matrix using the Minkowski r-metric) of the DATANET and (3) reducing the DATANET by eliminating each arc that has weight greater than the r-distance between the nodes connected by the arc. PFNET properties of inclusion, relation to minimal spanning trees, and invariance under transformations of data are discussed.

Priority and Organization of Information Accessed by Pilots in Various Phases of Flight

In 1 project, 27 pilots rated the priority of information required for flight. These pilots were divided by flight experience into novices (65 - 820 hrs) and experienced pilots (1,600 to 17,000 hrs). Participants rated 29 information elements across 7 phases of flight. These data show the shifting priorities of information across phases of flight, and some clear differences in priority assignments appeared between the novices and the experienced pilots. In another project, 34 pilots, some the same as before, participated in the collection of relatedness data for 231 pairs of information elements. Pathfinder analysis and hierarchical clustering were conducted showing connections among these elements and grouping of the elements. Pilot experience had little influence on the form of the network of associations. The discussion explores the potential of these data for instrumentation layout and integration of cockpit information systems, datalink design, and development of flight instruction curricula.

A formal interface design methodology based on user knowledge

In this paper we propose a formal interface design methodology based on user knowledge. The general methodology consists of 1) obtaining distance estimates for pairs of system units (objects, actions, concepts), 2) transforming the distance estimates using scaling techniques (e.g., Pathfinder network analysis), and 3) organizing the system interface based on the scaling solution. Thus, the organization of the system is based on the cognitive models of users rather than the intuitions of designers. As an example, we discuss the application of our methodology to the design of a network-based indexing aid for the UNIX on-line documentation system (MAN).