TAKOUA Laib

AMINA Naceur

ABDERAOUF Bouchiha

MAROAN Tahar

Departement of English

Faculty of Letters and Foreign Languages

Larbi Tebessi University

Models of Spoken Word Production

(L1 production)

Professor: BASSMA Bougoffa

2022/2023
 2

Models of Spoken Word Production

(L1 production)

Outline:

Introduction

I. Logogen Model (1964, 1969)

II. Levelt Model (1999)

III. Meyer Standard Model of Word Form Encoding (2000)

IV. Interactive Activation Model (Dell, 1986; 1997)

V. Interactive Network Model of Caramazza (1997)

Conclusion

References

Introduction:

The process of spoken word production involves intricate cognitive mechanisms

that transform thoughts into linguistic expressions. Spoken word production models

aim to explain the cognitive processes involved in generating and articulating spoken

words. These models provide frameworks that describe the stages, components, and

interactions necessary for transforming conceptual information into linguistically

encoded representations that can be produced as speech. The study of spoken word

production has been a topic of interest in the field of psycholinguistics, focusing on

understanding how individuals retrieve and produce words in real-time during

language production. Over the years, several influential models have been proposed to

account for the complex nature of this process. This paper aims to provide a
 3

comprehensive analysis of some of these influential models: The Logogen Model

(1964, 1969), Levelt Model (1999), Meyer’s Standard Model of Word Form Encoding

(2000), Dell’s Interactive Activation Model (1986; 1997) and Caramazza’s Interactive

Network Model (1997; 1998) By examining these models, we can gain insights into

the underlying mechanisms and processes involved in L1 production.

Logogen Model (1964, 1969):

General Background: The Logogen Model, developed by James L. McClelland and

Jeffrey L. Elman, is a seminal model that provides a framework for understanding

word recognition and production. It posits the existence of logogens, which are

information-processing units responsible for word representations in the mental

lexicon. According to the model, logogens are activated by sensory input and

influence language production through various cognitive processes.

Stages: According to McClelland and Elman (1986) word production involve three

stages: input analysis, lexical search, and phonological encoding.

Input analysis stage: During the input analysis stage, sensory information is

processed, leading to the activation of relevant logogens.

The lexical search stage: involves the selection of the appropriate logogens, which

are then activated and compete for access to the phonological encoding stage.

Phonological encoding stage: here the selected word form is transformed into a

phonological representation for articulation.

Criticisms: The Logogen Model has made significant contributions to our

understanding of word production, emphasizing the importance of multiple stages and

competitive processes (McClelland & Elman, 1986) However; it has faced criticisms

regarding its limited consideration of contextual factors and the precise nature of
 4

logogens. Further research is required to refine and extend the Logogen Model's

theoretical foundations.(see figure 1)

Figure 1 logogen model

Levelt Model (1999):

General Background: The Levelt Model, proposed by Willem J. M. Levelt in 1999,

provides a comprehensive framework for understanding the process of spoken word

production. This model highlights the dynamic nature of language production and

emphasizes the interaction between different cognitive processes.

Stages: According to Levelt Model (1999), word production involves a series of

stages:

Conceptualization: During the conceptualization stage, the speaker forms the

intended message and develops a conceptual representation of what they want to

express. This stage involves activating relevant concepts and organizing them into a
 5

coherent message. The speaker considers the communicative goals, contextual

information, and the intended audience while constructing the message.

Lemma Selection: After conceptualization, the speaker moves to the lemma selection

stage. In this stage, potential word candidates (lemmas) that are semantically related

to the intended message are activated. The activation spreads through the mental

lexicon, retrieving words with similar meanings. Competition occurs among these

activated lemmas, and the most appropriate word is selected based on factors such as

semantic fit, syntactic structure, and contextual constraints.

Lexical Access: Once the lemma is selected, the lexical access stage begins. The

selected lemma is used to retrieve the corresponding word form from the mental

lexicon. This involves accessing stored phonological and syntactic information

associated with the lemma. The retrieved word form serves as the basis for further

processing.

Morphological Encoding: In cases where the selected word involves morphological

processes (e.g., inflection or derivation), a separate morphological encoding stage

occurs. This stage involves generating the appropriate morphological forms based on

the grammatical requirements of the language. Morphological rules and patterns are

applied to produce the desired word form.

Phonological Encoding: Following lexical access and potential morphological

encoding, the phonological encoding stage takes place. In this stage, the abstract

phonological representation of the word is transformed into a specific sequence of

speech sounds. This includes processes such as syllabification, phoneme selection,

and phoneme sequencing. The phonological representation is prepared for

articulation.
 6

Articulation: The final stage of the Levelt Model is articulation. In this stage, the

prepared phonological representation is sent to the articulatory system, which

coordinates the movements of the speech organs to produce the desired speech

sounds. The motor plans for speech production are executed, resulting in the physical

production of the spoken word.

The Levelt Model (1999) emphasizes the interactive nature of these stages, with

feedback loops enabling revisions and adjustments at each level. It provides a

comprehensive account of the cognitive processes involved in spoken word

production and has been influential in shaping our understanding of this complex

process.(see figure 2 &3)

Figure 2: Levelt's model of speech production. From (Levelt, 1989, reprinted

courtesy of The MIT Press)

Figure 3: The WEAVER++ model in outline (Levelt et al., 1999)

Criticism: Empirical studies have provided support for the Levelt Model (1999) by

demonstrating findings consistent with its predictions. For example, psycholinguistic

experiments have shown evidence of lemma selection and the cascading nature of

word form encoding (Levelt,1999). Despite its contributions, the Levelt Model (1999)

has also faced some criticisms. Some researchers have argued for the inclusion of

additional stages or the need for a more nuanced account of certain processes.

Ongoing research continues to refine and expand upon the Levelt Model to address

these concerns.

Meyer's Standard Model of Word-Form Encoding (2000):

General Background: Meyer's Standard Model of Word-Form Encoding, proposed

by Antje S. Meyer and Herbert Schriefers in 2000, focuses on the processes involved

in transforming the abstract lexical representation of a word into a specific word form

for production.

Stages: According to Meyer's Standard Model (2000), word-form encoding consists

of three main stages:

The lemma selection stage: the speaker retrieves the appropriate lemma from their

mental lexicon, which represents the abstract form of the word. The selected lemma

then activates its associated lexemes, which are specific word forms related to the

lemma.
 8

The lexeme selection stage: the most suitable lexeme is chosen based on factors such

as grammatical requirements and contextual constraints.

The phonological encoding stage: the chosen lexeme is transformed into its

phonological representation, which encompasses the sounds and syllables necessary

for articulation.

Contribution: Meyer's Standard Model (2000) emphasizes the role of phonological

encoding in word production and highlights the importance of lemma and lexeme

levels in the selection and retrieval of specific word forms. This model has provided

insights into the intricate processes involved in transforming abstract word

representations into phonological representations for speech production.

Applications: Experimental studies have supported the predictions of Meyer's

Standard Model (2000). For instance, research on picture-word interference has

demonstrated the facilitation of phonological encoding when there is a match between

the target word and the phonological properties of an interfering word (Meyer &

Schriefers, 2000).
 9

Figure 4 The Standard Model of Speech Production

Criticism: Critiques of Meyer's Standard Model (2000) have focused on the need for

further clarification and refinement of certain aspects, such as the specific

mechanisms involved in lexeme selection and the precise nature of the interaction

between different levels of representation.

Interactive Activation Model of Dell (1986, 1997):

General Background: The Interactive Activation Model, developed by Gary S. Dell,

is a computational model that aims to explain the processes involved in spoken word

production, specifically focusing on the activation and competition between

phonological representations of words.

The process: Dell's Interactive Activation Model (1986, 1997) posits that word

production involves a network of interconnected nodes representing phonological

units such as phonemes, syllables, and words. These nodes are organized in

hierarchical manner, with lower-level nodes representing smaller units and higher-

level nodes representing larger units. According to the model, the production of a

word begins with the activation of the corresponding higher-level word node. This

activation spreads to lower-level nodes, leading to the activation of phonemes and

syllables associated with the word. Importantly, activation also occurs bidirectionally,

allowing feedback from lower-level nodes to influence higher-level nodes. In the

Interactive Activation Model, competition plays a crucial role. When multiple word

nodes are activated, there is competitive inhibition, wherein the activation of one

word inhibits the activation of competing words. This competition helps to select the

most strongly activated word for production. The model also incorporates

phonological neighborhoods, which refer to words that are phonologically similar to

the target word. Activation from words in the neighborhood can spread to the target

word, influencing its activation and potential production errors.

Contribution: Empirical studies and computational simulations have provided

support for Dell's Interactive Activation Model (1986, 1997). Experimental findings

have shown patterns of phonological errors and the influence of phonological

neighborhood effects, consistent with the model's predictions.

Criticism: Critiques of the Interactive Activation Model highlight the need for further

refinement, such as the integration of other linguistic levels (e.g., syntax, semantics)

and the consideration of additional factors that influence word production.

Figure 5 Dell’s interactive model.

Interactive Network Model of Caramazza (1997, Caramazza & Miozzo, 1997,

1998):
 11

General Background: The Interactive Network Model, developed by Alfonso

Caramazza and Marcelo Miozzo during 1997- 1998 is a comprehensive model that

explains various aspects of spoken word production, including lexical selection,

morphological processing, and phonological encoding.

The Model’s basis: Caramazza's Interactive Network Model (1997, Caramazza &

Miozzo, 1997, 1998) posits the existence of multiple levels of processing involved in

word production, with interconnected networks representing different linguistic

components. These components include semantics, syntax, and phonology.

The process: In this model, word production begins with the activation of semantic

features associated with the intended concept. The activated semantic features then

propagate activation to lexical representations in the lexicon. This activation spreads

across different levels of representation, including syntactic and phonological units.

The Interactive Network Model emphasizes the role of competition and inhibition in

lexical selection. Words with similar semantic or phonological features compete for

activation, and the strongest competitor is selected for production. This competition is

influenced by factors such as word frequency and contextual constraints.Furthermore,

the model incorporates morphological processing, highlighting the role of

morphological units in word production. Morphological representations contribute to

both lexical access and the generation of phonological forms.

Applications: Empirical studies have provided support for the Interactive Network

Model (Caramazza, 1997; Caramazza & Miozzo, 1997, 1998). Experimental findings

have demonstrated effects such as semantic priming, morphological facilitation, and

phonological interference, which align with the predictions of the model.

Criticism: Critiques of the Interactive Network Model suggest the need for further

exploration of specific mechanisms, such as the integration of phonological and

morphological processes and the consideration of individual differences in word

production.

Speech production model of Caramazza 1996

Conclusion:

Taking everything into account, the previously discussed models have greatly

advanced our understanding of spoken word (L1) production. These models have

provided valuable insights into the intricate processes involved in generating and

articulating words, highlighting the role of multiple stages, competitive processes, and

interactive activation. Supported by empirical evidence, they have contributed to our

knowledge of how linguistic information is accessed, selected, and transformed

during word production. While these models have their limitations and room for
 13

further refinement, they serve as important foundations for future research and

continued advancements in our understanding of spoken language production.

References:

Caramazza, A. (1997). How many levels of processing are there in lexical access?

Cognitive Neuropsychology, 14(1), 177-208.

Caramazza, A., & Miozzo, M. (1997). The relation between syntactic and

phonological knowledge in lexical access: Evidence from the "tip-of-the-

tongue" phenomenon. Cognition, 64(3), 309-343.

Caramazza, A., & Miozzo, M. (1998). The relation between syntactic and

phonological knowledge in lexical access: Further evidence from the "tip-of-

the-tongue" phenomenon. Journal of Memory and Language, 39(2), 277-301.

Dell, G. S. (1986). A spreading-activation theory of retrieval in sentence production.

Psychological Review, 93(3), 283-321.

Dell, G. S. (1997). The structure of a model of word production. Behavioral and Brain

Sciences, 20(4), 627-628.

Levelt, W. J. (1999). Models of word production. Trends in Cognitive Sciences, 3(6),

223-232.
 14

Meyer, A. S., & Schriefers, H. (2000). Phonological facilitation in picture-word

interference experiments: Effects of stimulus onset asynchrony and types of

interfering stimuli. Journal of Experimental Psychology: Learning, Memory,

and Cognition, 26(5), 1142-1160.

McClelland, J. L., & Elman, J. L. (1986). The TRACE model of speech perception.

Cognitive Psychology, 18(1), 1-86.

McClelland, J. L., & Elman, J. L. (1991). The TRACE model of speech perception: I.

An expanded model. Psychological Review, 98(4), 523-568.

Paivio, A. (1969). Mental imagery in associative learning and memory. Psychological

Review, 76(3), 241-263.

Paivio, A. (1971). Imagery and Verbal Processes. Holt, Rinehart & Winston.

Paivio, A., & Csapo, K. (1969). Picture superiority in free recall: Imagery or dual

coding? Cognitive Psychology, 1(3), 218-247.