Models of word production

doi:10.1016/S1364-6613(99)01319-4

Trends in Cognitive Sciences

Volume 3, Issue 6, 1 June 1999, Pages 223-232

https://doi.org/10.1016/S1364-6613(99)01319-4 Get rights and content

Abstract

Research on spoken word production has been approached from two angles. In one research tradition, the analysis of spontaneous or induced speech errors led to models that can account for speech error distributions. In another tradition, the measurement of picture naming latencies led to chronometric models accounting for distributions of reaction times in word production. Both kinds of models are, however, dealing with the same underlying processes: (1) the speaker’s selection of a word that is semantically and syntactically appropriate; (2) the retrieval of the word’s phonological properties; (3) the rapid syllabification of the word in context; and (4) the preparation of the corresponding articulatory gestures. Models of both traditions explain these processes in terms of activation spreading through a localist, symbolic network. By and large, they share the main levels of representation: conceptual/semantic, syntactic, phonological and phonetic. They differ in various details, such as the amount of cascading and feedback in the network. These research traditions have begun to merge in recent years, leading to highly constructive experimentation. Currently, they are like two similar knives honing each other. A single pair of scissors is in the making.

Section snippets

Two kinds of model

All current models of word production are network models of some kind. In addition, they are, with one exception⁵, all ‘localist’, non-distributed models. That means that their nodes represent whole linguistic units, such as semantic features, syllables or phonological segments. Hence, they are all ‘symbolic’ models. Of the many models with ancestry in the speech error tradition6, 7, 8 only a few have been computer-implemented9, 10, 11. Among them, Dell’s two-step interactive activation model⁹

Conceptual preparation

The first step in accessing content words such as cat or select is the activation of a lexical concept, a concept for which you have a word or morpheme in your lexicon. Usually, such a concept is part of a larger message, but even in the simple case of naming a single object it is not trivial which lexical concept you should activate to refer to that object. It will depend on the discourse context whether it will be more effective for you to refer to a cat as cat, animal, siamese or anything

Lexical selection

In the chronometric tradition lexical selection has been studied with interference paradigms, in particular picture-word interference (see Box 1). The recurring finding has been that naming an object is slowed down when a distracter word is presented with the picture; the effect is stronger when the distracter word is semantically related to the target than when it is semantically unrelated and it is at maximum when picture and distracter word are presented simultaneously²⁷. The WEAVER model

Morpho-phonological encoding

When you are planning the sentence ‘they are selecting me’, you must retrieve from your lexicon the morpho-phonological codes for each of the selected words, among them the two morpheme-size codes select and ing (see Fig. 3), and compute their syllabification and accent structure in context (se-léc-ting). This naturally divides the process into ‘code retrieval’ and ‘prosodification’.

Phonetic encoding and articulation

As incremental prosodification proceeds, the resulting syllabic and larger prosodic structures should acquire phonetic shape. As a speaker you will incrementally prepare articulatory gestures for the syllables in their prosodic context. A core feature of the WEAVER model is the notion of a syllabary⁵¹. Statistics show that native speakers of English or Dutch do 80 percent of their talking with no more than about 500 different syllables¹⁸ (although these languages have many more than 10 000

Conclusion

There is still a long way to go before the two research traditions emerging from speech error analysis and from naming chronometry are fully reconciled. But there has been lively and highly constructive interaction, leading to a much improved understanding of the processes involved in lexical selection and phonological encoding. One unifying force has been computational modeling. Current implemented models share their major strata, they are localist and symbolic; they compute quite similar

Outstanding questions

•
How should error-based and chronometric models be further reconciled computationally and empirically?
•
What causes a speech error? Is it caused by occasional cascading or occasional feedback in a normally non-cascading, feed-forward system? Is it the product of noise in a normally cascading interactive system? Or is the origin of speech error something else entirely?
•
How does the word-production network relate to the word-perception network? How is self-monitoring realized in this combined system?
•

Acknowledgements

I gratefully acknowledge helpful commentary by Antje Meyer and by Gary Dell.

References (57)

G.S. Dell et al.
Structure and content in language production: a theory of frame constraints in phonological speech errors
Cognit. Sci.
(1993)
Stemberger, J.P. (1985) An interactive activation model of language production, in Progress in the Psychology of...
T. Berg
Linguistic Structure and Change
(1998)
W. La Heij et al.
Categorical interference and associative priming in picture naming
Br. J. Psychol.
(1990)
M. Bierwisch et al.
From lexical concepts to lexical items
Cognition
(1992)
A. Roelofs
A case for nondecomposition in conceptually driven word retrieval
J. Psycholinguist. Res.
(1997)
N. Martin
Phonological faciliation of semantic errors in normal and aphasic speakers
Lang. Cognit. Process.
(1996)
J.S. Bowers
Grossberg and colleagues solved the hyperonym problem over a decade ago
Behav. Brain Sci.
(1999)
M. Van Turennout et al.
Brain activity during speaking: from syntax to phonology in 40 milliseconds
Science
(1998)
G. Vigliocco, et al., Is ‘count’ and ‘mass’ information available when the noun is not?: an investigation of...

E. Marx, Gender processing in speech production: evidence from German, J. Psycholinguist. Res. ((in...

J.A. Goldsmith

Autosegmental and Metrical Phonology

(1990)

A. Roelofs et al.

Metrical structure in planning the production of spoken words

J. Exp. Psychol. Learn. Mem. Cognit.

(1997)

Indefrey, P. & Levelt, W.J.M. in The Cognitive Neurosciences (2nd edn) (Gazzaniga, M., ed.), MIT Press (in...

R.D. Kent et al.

Models of speech production

C.J. Price

The functional anatomy of word comprehension and production

Trends Cognit. Sci.

(1998)

G.A. Miller

The Science of Words

(1991)

A. Garnham

Slips of the tongue in the London–Lund corpus of spontaneous conversation

Linguistics

(1981)

W.J.M. Levelt

Speaking: From Intention to Articulation

(1989)

W.J.M. Levelt

Language production: a blueprint of the speaker

D.G. Mackay

The Organization of Perception and Action: A Theory for Language and other Cognitive Skills

(1987)

G.S. Dell

A spreading-activation theory of retrieval in sentence production

Psychol. Rev.

(1986)

T.A. Harley

Phonological activation of semantic competitors during lexical access in speech production

Lang. Cognit. Process.

(1993)

U. Schade et al.

The role of inhibition in a spreading-activation model of language production: II. Simulational perspective

J.

(1992)

Psycholinguist. Res.

(1992)

G.S. Dell

Lexical access in aphasic and non-aphasic speech

Psychol. Rev.

(1997)

M.F. Damian et al.

Semantic and phonological codes interact in single word production

J. Exp. Psychol. Learn. Mem. Cognit.

(1999)

G.W. Humphreys et al.

An interactive activation approach to object processing: effects of structural similarity, name frequency and task in normality and pathology

Memory

(1995)

M.O. Glaser et al.

Time course analysis of the Stroop phenomenon

J. Exp. Psychol. Hum. Percept. Perform.

(1982)

Cited by (423)

Speaking to remember: Model-based adaptive vocabulary learning using automatic speech recognition
2024, Computer Speech and Language
Memorizing vocabulary is a crucial aspect of learning a new language. While personalized learning- or intelligent tutoring systems can assist learners in memorizing vocabulary, the majority of such systems are limited to typing-based learning and do not allow for speech practice. Here, we aim to compare the efficiency of typing- and speech based vocabulary learning. Furthermore, we explore the possibilities of improving such speech-based learning using an adaptive algorithm based on a cognitive model of memory retrieval. We combined a response time-based algorithm for adaptive item scheduling that was originally developed for typing-based learning with automatic speech recognition technology and tested the system with 50 participants. We show that typing- and speech-based learning result in similar learning outcomes and that using a model-based, adaptive scheduling algorithm improves recall performance relative to traditional learning in both modalities, both immediately after learning and on follow-up tests. These results can inform the development of vocabulary learning applications that–unlike traditional systems–allow for speech-based input.
Linguistic features of stuttering during spontaneous speech
2023, Journal of Fluency Disorders
Previous work shows that linguistic features (e.g., word length, word frequency) impact the predictability of stuttering events. Most of this work has been conducted using reading tasks. Our study examined how linguistic features impact the predictability of stuttering events during spontaneous speech.
The data were sourced from the FluencyBank database and consisted of interviews with 35 adult stutterers (27,009 words). Three logistic regression mixed models were fit as the primary analyses: one model with four features (i.e., initial phoneme, grammatical function, word length, and word position within a sentence), a second model with six features (i.e., the features from the previous model plus word frequency and neighborhood density), and a third model with nine features (i.e., the features from the previous model plus bigram frequency, word concreteness, and typical age of word acquisition). We compared our models using the Area Under the Curve statistic.
The four-feature model revealed that initial phoneme, grammatical function, and word length were predictive of stuttering events. The six-feature model revealed that initial phoneme, word length, word frequency, and neighborhood density were predictive of stuttering events. The nine-feature model was not more predictive than the six-feature model.
Linguistic features that were previously found to be predictive of stuttering during reading were predictive of stuttering during spontaneous speech. The results indicate the influence of linguistic processes on the predictability of stuttering events such that words associated with increased planning demands (e.g., longer words, low frequency words) were more likely to be stuttered.
Infants aged 12 months use the gender feature in determiners to anticipate upcoming words: An eye-tracking study
2023, Journal of Child Language
Profiles of word-finding difficulties in school-aged children
2023, Journal of Child Language
An Information-Theoretic Account of Availability Effects in Language Production
2024, Topics in Cognitive Science
Automating intended target identification for paraphasias in discourse using a large language model
2023, Journal of Speech, Language, and Hearing Research

View all citing articles on Scopus

View full text

Trends in Cognitive Sciences

ReviewModels of word production

Abstract

Section snippets

Two kinds of model

Conceptual preparation

Lexical selection

Morpho-phonological encoding

Phonetic encoding and articulation

Conclusion

Outstanding questions

Acknowledgements

Cognit. Sci.

Br. J. Psychol.

Cognition

J. Psycholinguist. Res.

Lang. Cognit. Process.

Behav. Brain Sci.

Science

J. Exp. Psychol. Learn. Mem. Cognit.

Trends Cognit. Sci.

The Science of Words

Slips of the tongue in the London–Lund corpus of spontaneous conversation

Linguistics

Speaking: From Intention to Articulation

Language production: a blueprint of the speaker

The Organization of Perception and Action: A Theory for Language and other Cognitive Skills

A spreading-activation theory of retrieval in sentence production

Psychol. Rev.

Phonological activation of semantic competitors during lexical access in speech production

Lang. Cognit. Process.

The role of inhibition in a spreading-activation model of language production: II. Simulational perspective

J.

Psycholinguist. Res.

Lexical access in aphasic and non-aphasic speech

Psychol. Rev.

Semantic and phonological codes interact in single word production

J. Exp. Psychol. Learn. Mem. Cognit.

An interactive activation approach to object processing: effects of structural similarity, name frequency and task in normality and pathology

Memory

Time course analysis of the Stroop phenomenon

J. Exp. Psychol. Hum. Percept. Perform.

Review
Models of word production