It is widely held that there is a distinction between attentive and automatic cognitive processing. In research on attention using visual search tasks, the detection performance of human subjects in consistent mapping paradigms is generally regarded as indicating a shift, with practice, from serial, attentional, controlled processing to parallel, automatic processing, while detection performance in varied mapping paradigms is taken to indicate that processing remains under attentional control. This paper proposes a priority learning mechanism to model the effects of practice and the development of automaticity, in visual search tasks. A connectionist simulation model implements this learning algorithm. Five prominent features of visual search practice effects are simulated. These are: 1) in consistent mapping tasks, practice reduces processing time, particularly the slope of reaction times as a function of the number of comparisons; 2) in varied mapping tasks, there is no change in the slope of the reaction time function; 3) both the consistent and varied effects can occur concurrently; 4) reversing the target and distractor sets produces strong interference effects; and 5) the benefits of practice are a function of the degree of consistency.
Download article: PostScript. PDF.
Metrical phonology is a relatively successful theory that attempts to explain stress systems in language. This paper discusses a perceptron model of stress, pointing out interesting parallels between certain aspects of the model and the constructs and predictions of metrical theory. The distribution of learning times obtained from perceptron experiments corresponds with theoretical predictions of ``markedness.'' In addition, the weight patterns developed by perceptron learning bear a suggestive relationship to features of the linguistic analysis, particularly with regard to iteration and metrical feet. Our results suggest that simple statistical learning techniques have the potential to complement, and provide computational validation for, abstract theoretical investigations of linguistic domains.
Download article: PostScript. PDF.
We use connectionist modeling to develop an analysis of stress systems in terms of ease of learnability. In traditional linguistic analyses, learnability arguments determine default parameter settings based on the feasibilty of logically deducing correct settings from an initial state. Our approach provides an empirical alternative to such arguments. Based on perceptron learning experiments using data from nineteen human languages, we develop a novel characterization of stress patterns in terms of six parameters. These provide both a partial description of the stress pattern itself and a prediction of its learnability, without invoking abstract theoretical constructs such as metrical feet. This work demonstrates that machine learning methods can provide a fresh approach to understanding linguistic phenomena.
Download article: PostScript. PDF.
Linguistic categories play a key role in virtually every theory that has a bearing on human language. This paper presents a connectionist model of grammatical category formation and use, within the domain of the German nominal system. The model demonstrates (1) how categorical information can be created through co-occurrence learning; (2) how grammatical categorization and inflectional marking can be integrated in a single system; (3) how the use of co-occurrence information, semantic information and surface feature information can be usefully combined in a learning system; and (4) how a computational model can scale up toward simulating the full range of phenomena involved in an actual system of inflectional morphology. This is, to our knowledge, the first connectionist model to simultaneously address all these issues for a domain of language acquisition.
Download article: PostScript. PDF.
Findings in infant speech perception suggest that early phonological perceptions may be syllabic in nature, and that there is a loss of sensitivity to nonnative contrasts toward the end of the first year of life. We present a neural network model that simulates these two phenomena. In addition, the model and simulations (1) demonstrate how information about stress can be utilized in generating syllable-like perceptions; (2) provide a simple means of extracting static representations from a dynamic and co-articulated signal; and (3) indicate that the development of ``attractor'' states may be necessary in network models of these phenomena.
Download article: PostScript. PDF.
The paradigm of immediate serial recall (Baddeley, 1986) has been used extensively in investigation of working memory, but its relation to and implications for the nature of phonological processing have seldom been examined. We show that findings from this domain can be interpreted in two ways, and relate these two interpretations to a simple model of phonological processing. One interpretation emphasizes the availability of information from ``output'' phonological processing to ``input'' phonological processing, while the alternative account stipulates no such connections. On the basis of an experimental study designed to choose between the two accounts, we tentatively conclude that the interpretation suggesting output-input connectivity is supported. Establishment of this result would be of considerable interest, since it indicates that processes in language production can impact directly on processes in language perception.
Download article: PostScript. PDF.
This paper outlines a bottom-up research approach to studying lexical representation, emphasizing the development of infant phonological perception as a source of data that can illuminate the nature of phonological representations. It is proposed that neural network formalisms can provide a useful framework for thinking about these issues. We identify a key set of empirical results, understanding of which would yield considerable insight; this defines a modeling agenda for investigation of phonological representations. We report preliminary simulations that explore this agenda, exemplifying how network modeling techniques can contribute to understanding of these phenomena.
Download article: PostScript. PDF.
We question the widespread assumption that linguistic theory should guide the formulation of mechanistic accounts of human language processing. We develop a pseudo-linguistic theory for the domain of linguistic stress, based on observation of the learning behavior of a perceptron exposed to a variety of stress patterns. There are significant similarities between our analysis of perceptron stress learning and metrical phonology, the linguistic theory of human stress. Both approaches attempt to identify salient characteristics of the stress systems under examination without reference to the workings of the underlying processor. Our theory and computer simulations exhibit some strikingly suggestive correspondences with metrical theory. We show, however, that our high-level pseudo-linguistic account bears no causal relation to processing in the perceptron, and provides little insight into the nature of this processing. Because of the persuasive similarities between the nature of our theory and linguistic theorizing, we suggest that linguistic theory may be in much the same position. Contrary to the usual assumption, it may not provide useful guidance in attempts to identify processing mechanisms underlying human language.
Download article (Technical Report version): PostScript. PDF.
Results from the paradigm of immediate serial recall form the basis of the influential ``articulatory loop'' model of auditory-verbal short-term memory (Baddeley, 1986). Central to the development of these ideas have been results obtained in immediate serial recall under the condition of concurrent articulation. We re-examine the effects of concurrent articulation, and show that findings from immediate serial recall do not uniquely support the articulatory rehearsal hypothesis: the data can be accounted for by assuming a purely auditory rehearsal process. The question of whether the rehearsal process in fact has an ``articulatory'' component or is purely ``auditory'' has significance beyond the immediate domain of working memory, and makes contact with a number of important issues concerning phonological processing. We describe a series of experiments aimed at discriminating between the two hypotheses. Our results support an articulatory component in rehearsal, but also indicate that auditory interference plays a significant, but previously unrecognized, role in the concurrent articulation effect.
Recent behavioral evidence suggests that human vocabulary acquisition processes and verbal short-term memory abilities may be related (Gathercole & Baddeley, 1993). Investigation of this relationship has considerable significance for understanding of human language, of working memory, and of the relationship between short- and long-term memory systems. This paper presents a computational model of word learning, nonword repetition, and immediate serial recall. By providing an integrated account of these three abilities, the model provides a specification of how the mechanisms of immediate serial recall may be related to mechanisms of language processing more generally. Furthermore, the model provides fresh insight into the observed behavioral correlations between word learning and immediate serial recall. According to the model, these correlations can arise because of the common dependence of these two abilities on core phonological and semantic processing mechanisms. This contrasts with the explanation proposed in the working memory literature, viz., that word learning is dependent on verbal short-term memory (Gathercole, Willis, Emslie, & Baddeley, 1992). It is discussed how both explanations can be reconciled in terms of the present model.
Download article: PostScript. PDF.
In this paper, we explore the hypothesis that human vocabulary acquisition processes and verbal short-term memory abilities utilize a common cognitive and neural system. We begin by reviewing behavioral evidence for a shared set of processes. Next, we examine what the computational bases of such a shared system might be, and how vocabulary acquisition and verbal short-term memory might be related in mechanistic terms. We examine existing computational models of vocabulary acquisition and of verbal short-term memory, concluding that they fail to adequately relate these two domains. We then propose an alternative model which accounts not only for the relationship between word learning and verbal short-term memory, but also for a wide range of phenomena in verbal short-term memory. Furthermore, this new account provides a clear statement of the relationship between the proposed system, and mechanisms of language processing more generally. We then consider possible neural substrates for this cognitive system. We begin by reviewing what is known of the neural substrates of speech processing, and outline a conceptual framework within which a variety of seemingly contradictory neurophysiological and neuropsychological findings can be accommodated. The linkage of the shared system for vocabulary acquisition and verbal short-term memory to neural areas specifically involved in speech processing lends further support to our functional-level identification of the mechanisms of vocabulary acquisition and verbal short-term memory with those of language processing. The present work thus relates vocabulary acquisition and verbal short-term memory to each other and to speech processing, at a cognitive, computational, and neural level.
Download article: PDF.
In every spoken language, the words have two fundamental properties. First, they are temporal sequences; in the articulatory domain, a word is a sequence of gestures, and in the auditory domain, a sequence of sounds. Second, the relation between this serially ordered sequence and the word's meaning is arbitrary. Words with similar meaning do not typically have similar forms.
It is a truism that the nature of language has shaped the way that it is learned and processed. It is also generally agreed that the reverse is true; learning and processing constraints have profoundly influenced language. In this chapter, we start with the fact that words are serially ordered temporal sequences arbitrarily related to meaning, and ask how these properties reflect and are reflected in the processing system. Specifically, in the first section, we suggest that some aspects of vocabulary structure can be linked to the fact that words are sequences. Along with this, we will consider the production of words and introduce a simple recurrent network model that relates some speech error phenomena to the sequential nature of words. This account of sequence and speech errors embodies a view in which many aspects of language learning can be understood as proceeding via the gradual, experience-driven adjustment of connection weights between levels of representation. In the second section, we examine parallels between this general view of language learning, and implicit or procedural memory. In particular, we suggest that the systematic aspects of language learning are continuous with the mechanisms of procedural memory, and show that the simple recurrent network model that accounts for speech errors can also account for important phenomena in the domain of procedural memory. In the third section, we situate our speech error model in the cognitive system and situate our discussion of procedural memory more concretely in the context of language. We discuss how word forms can be acquired and associated with arbitrary meanings. We suggest that the seemingly unitary process of learning a new word can be fractionated into two components, which rely on two quite different memory systems and learning mechanisms. The discussion centers on the relation between the sequential and arbitrary aspects of words, and on the correspondence of these aspects of word learning with the procedural and declarative memory systems. Overall, we propose in this chapter that important aspects of language can be viewed as emerging from serial order and from procedural memory.
It seems an obvious truth that children are better language learners than adults. Children seem able to master a second language with ease, while adults are rarely successful at second language acquisition. Newport's (1990) Less is More hypothesis represented an attempt to explain these observations by invoking general cognitive mechanisms. This hypothesis takes as its starting point the observation that children exhibit reduced working memory capacity relative to adults and suggests that this reduction serves as a filter to aid children in deducing the structure of the language they are learning. We present two experiments testing a specific prediction that follows from the Less is More hypothesis, namely that adults will perform better on language learning tasks if their available working memory capacity is reduced. The experiments examined the learning of word boundaries and syntactic agreement, each with and without a concurrent cognitive load. The results of these experiments were contrary to the Less is More prediction, suggesting that other explanations must be found for the observed superior language learning performance of children over adults.
Download article: PDF.
Almost without exception, acquired language disorders resulting from focal brain injury are accompanied by impairments of verbal short-term memory (STM) and verbal learning. Moreover, disturbances in verbal STM are generally associated with language dysfunction, in some cases mild and not disruptive to most language activities. The co-occurrence of language and mnestic deficits in the aphasic population affords the opportunity to examine their relationships. Here we review evidence from aphasia supporting the view that verbal STM processes are inextricably linked with the lexical processing system (as opposed to drawing solely on an independent short-term memory system). These studies are discussed as a foundation for the development of a new model of word processing, verbal STM and learning. An inherent assumption of this model is that spreading activation processes which maintain activation of linguistic representations when generating a single word also contribute to the support of multiple word representations in verbal STM.
Purpose. To assess the effect of orthographic similarity neighborhood characteristics on pharmacistsÕ visual perception of typewritten drug names. Methods. Forty-five practicing pharmacists viewed 160 three-syllable drug names selected from the combined 1992-1996 databases of the U. S. National Ambulatory Medical Care Survey and the National Hospital Ambulatory Care Medical Survey. Twenty names each were selected at high and low levels of prescribing frequency (i.e., the log of the number of NAMCS/HAMCS prescriptions), neighborhood frequency (i.e., the average log prescribing frequencies of names within an edit distance of 3 from a given name), and neighborhood density (i.e., the number of names within an edit distance of 3 of a given name). Names were presented to pharmacists on a computer monitor for 3 seconds in a 36-point Times font. Names were masked by visual noise. The pharmacist's task was to correctly identify each name by typing in the correct spelling of the name. Logistic regression models were used to test hypotheses. Results. The probability of a perceptual error decreased significantly as prescribing frequency increased (b = -0.73, p < 0.0001). The probability of error increased significantly as neighborhood density increased (b = 0.06, p < 0.05). Neighborhood frequency had no effect on the probability of error. Conclusion. As opposed to rare names and names with many similar neighbors, frequently prescribed drug names and drug names with few similar neighbors are less likely to be visually misperceived by pharmacists, even when the names are substantially degraded by noise and brief exposure durations.
This article analyzes the relationship between skill learning and repetition priming, 2 implicit memory phenomena. A number of reports have suggested that skill learning and repetition priming can be dissociated from each other and are therefore based on different mechanisms. The authors present a theoretical analysis showing that previous results cannot be regarded as evidence of a processing dissociation between skill learning and repetition priming. The authors also present a specific single-mechanism computational model that simulates a specific experimental task and exhibits both skill learning and repetition priming, as well as a number of apparent dissociations between these measures. These theoretical and computational analyses provide complementary evidence that skill learning and repetition priming are aspects of a single underlying mechanism that has the characteristics of procedural memory.
Download article: PDF.
This presentation will argue that (1) implicit memory is based on statistical learning; (2) phonological learning of word forms is based on implicit memory, and therefore that (3) phonological learning of word forms is statistical learning. A theoretical analysis of two key implicit memory phenomena, skill learning and repetition priming, will focus on behavioral dissociations between them. First, it will be shown that the fact that skill learning but not repetition priming follows the power law of practice follows from the mathematical definitions of these constructs, and that this dissociation is therefore artifactual. Second, it will be shown that the presence or absence of correlations between these phenomena is also artifactual, and also follows from their definitions. Behavioral dissociations between these phenomena therefore cannot be regarded as evidence of a processing dissociation between them. A statistical learning based computational model will then be shown to account for specific empirical data, exhibiting a classic profile of skill learning and repetition priming, as well as a number of apparent dissociations between these phenomena. These theoretical and computational analyses provide complementary evidence that skill learning and repetition priming are aspects of a single underlying mechanism that supports implicit memory and has the characteristics of statistical learning. The hypothesis that phonological learning of word forms is based on implicit memory predicts that implicit memory tasks employing nonwords (i.e., novel phonological word forms) should yield a typical profile of skill learning and repetition priming. A typical multiple-repetition implicit memory task was devised in which participants were presented with nonwords. Some of the nonwords in each block of stimuli appeared only once during the experiment while other nonwords appeared in every block. Participants were simply required to repeat each stimulus as soon as it was presented. Performance functions were very similar to those in standard implcit memory tasks, exhibiting classic skill learning and repetition priming. These findings suggest that implicit memory plays a role in the learning of phonological forms, which in turn suggests a role for statistical mechanisms in phonological learning. Further evidence of the role of distributional statistics in phonological learning comes from a second manipulation in the study. If distributional statistics play a role in the learning of phonological word forms, then a word form's frequency-weighted neighborhood density (N) should impact repetition priming. To test this hypothesis, half of the nonword stimuli had a high N and half a low N. Neighborhood density was found to have a significant impact on learning of the nonwords. The effects of neighborhood density provide new evidence that phonological learning is affected by the distributional statistics of the environment. The presence of classic skill learning and repetition priming effects in nonword repetition provides complementary evidence regarding the nature of the underlying learning, suggesting it is based on implicit memory. The theoretical and computational analyses suggest that implicit memory is based on statistical learning mechanisms. Together the present results provide new evidence that learning novel phonological forms is based on statistical mechanisms.
Download article: PDF.
Two experiments examined whether the association between word learning, nonword repetition, and immediate serial recall observed in children also exists in normal adults. The experiments also introduce a novel paradigm for studying word learning. Experiment 1 studied the performance of 52 adults in nonword repetition, immediate serial recall, and word learning tasks, examining the correlation between these measures. The results indicate that the developmental relationships between all three abilities also exist in adults. Experiment 2 investigated the robustness of these results using different stimuli and a variant of the word learning task, and also examined performance in a visuospatial span task, to test an alternative account of the results of Experiment 1; the results from 58 adults provide further evidence that the developmental association between word learning, nonword repetition, and immediate serial recall extends into adulthood. The theoretical implications of these findings are discussed in terms of alternative models of the relationship between these abilities.
Download article: PDF.
Download article: PDF.
Download preprint: PDF.
Download article: PDF.
One of the most interesting results in the study of neighborhood effects on nonword repetition has been Vitevitch & Luce's (1998) finding that increased neighborhood density has a facilitatory effect on nonword repetition latency. However, analysis of the stimuli used by Vitevitch and Luce (1998) reveals a mean stimulus duration difference of 16 ms that may have contributed to the observed difference. Experiment 1 presents a replication of the basic response latency effect and Experiment 2 presents a reduction of the effect to nonsignificance after controlling for stimulus duration differences through stimulus extension and compression.
Download article: PDF.
We describe a set of pictorial and auditory stimuli that we have developed for use in word learning tasks in which the participant learns pairings of novel auditory sound patterns (``names'') with pictorial depictions of novel objects (``referents''). The pictorial referents are drawings of ``space aliens'', consisting of images that are variants of 144 different aliens. The auditory names are possible nonwords of English; the stimulus set consists of over 2,500 nonword stimuli recorded in a single voice, with controlled onsets, varying from one through seven syllables in length. The pictorial and nonword stimuli can also serve as independent stimulus sets for purposes other than word learning.
Download article: PDF.
Twelve experiments examined the effect of neighborhood density on repetition latency for nonwords. Previous reports have indicated that nonwords from high density neighborhoods are repeated with shorter latency than nonwords from low density neighborhoods (e.g., Vitevitch & Luce, 1998). Experiment 1 replicated these previously reported results; however, further analysis indicated an interaction of neighborhood density and stimulus duration in determining nonword repetition latency. Experiment 2 employed stimuli with reduced durational differences, finding that there was no effect of neighborhood density on repetition latency when stimulus duration was statistically controlled. Experiments 3 and 4 replicated these results with an alternative presentation regimen. Experiments 5-12 repeated these investigations with diferent stimulus sets, and obtained consistent effects of stimulus duration on repetition latency, and either no effect of neighborhood density or a latency advantage for low density rather than high density nonwords. The theoretical implications of these results for models of lexical processing are discussed.
Download article: PDF.
Download article: PDF.
A growing body of research has emphasized the linkage between performance in immediate serial recall of lists, nonword repetition, and word learning. Recently, it has been reported that primacy and recency effects are obtained in repetition of individual syllables within nonwords (Gupta, in press). Five experiments examined whether such within-nonword primacy and recency effects are attributable to common sequencing mechanisms that are shared with immediate list recall. Experiments 1 and 2 indicated that the primacy and recency effects are not simply due to greater morphological salience at the beginnings/endings of nonwords, and that the serial position effects generalize to different stimuli and across a variety of stimulus lengths. Experiment 3 indicated that the primacy and recency effects are similar to those obtained in list recall. Experiments 4 and 5 examined alternative hypotheses for the observed serial position effects, concluding that the alternative hypotheses fail to account for the obtained pattern of results. These results provide support for the common sequencing mechanisms hypothesis. The implications of these results are discussed in terms of the relationship between list recall and nonword repetition, and in terms of broader issues in word learning.
Download article: PDF.
Study of the phonological similarity effect (PSE) in immediate serial recall (ISR) has produced a conflicting body of results. Five experiments tested various theoretical ideas that together may help integrate these results. Experiments 1 and 2 tested alternative accounts that explain the effect of phonological similarity on item recall in terms of either feature overlap, linguistic structure, or serial order. In each experiment, participants' ISR was assessed for rhyming, alliterative, and similar nonrhyming/nonalliterative lists. The results were consistent with the predictions of the serial order account, with item recall being higher for rhyming than alliterative lists, and higher for alliterative than similar nonrhyming/nonalliterative lists. Experiments 3 and 4 showed that these item recall differences are reduced when list items repeat across lists. Experiment 5 employed rhyming and dissimilar one-syllable and two-syllable lists to demonstrate that recall for similar (rhyming) lists can be better than for dissimilar lists even in a typical ISR task using words, providing a direct reversal of the classic PSE. These and other previously published results are interpreted and integrated within a proposed theoretical framework that offers an account of the PSE.
Download preprint: PDF.
An unexceptionable view of word learning or vocabulary acquisition is that it entails learning a word form, a meaning, and the link between them (e.g., Saussure, 1916; Desrochers & Begg, 1987). However, this simple formulation encompasses a multitude of different abilities and subcomponents. As a result, the terms "vocabulary acquisition" and "word learning" have been used to mean a wide variety of different things. This article begins by outlining a simple functional framework for thinking about various types and aspects of word learning. It then applies this framework to characterizing several important streams of investigation within the word learning literature, in the belief that locating these investigations with respect to each other is a useful endeavor.
This page last updated October 24, 2005.