Models Of Language Production Research Articles

Parallel processing in language production

Following the Sixth International Workshop on Language Production (Edinburgh, UK, Sept., 2010), this special issue presents a collection of contributions concerned with a wide range of representational and processing components. In the present article, we review the evidence for parallel processing at different levels within the production system with the aim of identifying any generalisation or common characteristics that might underpin a robust model of language production. Our review synergises with the other articles of the special issue. After reviewing the literature, we conclude that the evidence for parallelism is stronger at some levels than at others and it is premature to take a strong stand for a unified principle that applies equally to all components of the production system. Following our review, we introduce the other articles represented within this special issue.

Nonfinite verbs and negotiating bilingualism in codeswitching: Implications for a language production model

This paper argues that a set of codeswitching data has implications for the nature of cognitive control in bilingualism and for models of language production in general. The data discussed are Embedded Language (EL) nonfinite verbs that occur in Matrix Language (ML) frames with appropriate ML inflectional morphology in some codeswitching (CS) corpora. Notably EL infinitives are involved, as inwo muconçevoirbe nuɖe. . . “they don't imagine that something . . .” (from Ewe–French CS). The main argument is that such nonfinite forms are selected because they only need checking at the lexical-conceptual level of abstract structure with the speaker's intended semantic-pragmatic meaning. That is, they do not project information about syntactic and argument structure that is included in the abstract structure of finite verbs. Nonfinite EL verbs occur because they better satisfy the speaker's intentions regarding semantic and pragmatic meaning than NL finite verbs. The employment of nonfinite EL verbs instead of EL finite verbs partially explains why codeswitching in general and such verb phrases in particular is perceived as fast and effortless. How one lexical entry (the EL nonfinite verb) can take on the morphosyntactic role of another one (the ML finite verb) implies flexibility in cognitive control at an abstract level. It also implies a certain malleability at an abstract level in the ML morphosyntactic frame that makes it possible to take in a nonfinite verb in a slot for a finite verb.

Reconciling phonological neighborhood effects in speech production through single trial analysis

A crucial step for understanding how lexical knowledge is represented is to describe the relative similarity of lexical items, and how it influences language processing. Previous studies of the effects of form similarity on word production have reported conflicting results, notably within and across languages. The aim of the present study was to clarify this empirical issue to provide specific constraints for theoretical models of language production. We investigated the role of phonological neighborhood density in a large-scale picture naming experiment using fine-grained statistical models. The results showed that increasing phonological neighborhood density has a detrimental effect on naming latencies, and re-analyses of independently obtained data sets provide supplementary evidence for this effect. Finally, we reviewed a large body of evidence concerning phonological neighborhood density effects in word production, and discussed the occurrence of facilitatory and inhibitory effects in accuracy measures. The overall pattern shows that phonological neighborhood generates two opposite forces, one facilitatory and one inhibitory. In cases where speech production is disrupted (e.g. certain aphasic symptoms), the facilitatory component may emerge, but inhibitory processes dominate in efficient naming by healthy speakers. These findings are difficult to accommodate in terms of monitoring processes, but can be explained within interactive activation accounts combining phonological facilitation and lexical competition.

The production of coerced expressions: Evidence from priming

• Four experiments investigated production of coerced and full-form sentences. • Participants showed priming for both coercion and constituent structure. • Repetition of the coercing verb and coerced action affected priming. • Coerced and full-form sentences use different semantic-to-syntactic mappings. We report four structural priming experiments investigating the syntactic and semantic processes involved in producing coerced and full-form sentences (e.g., The bricklayer began the wall vs. The bricklayer began building the wall ). Experiments 1 and 2 demonstrated priming for syntactic structure across sentences that involved the same coercing verb (e.g., began ). Experiment 1 (and the combined analysis of Experiments 1 and 2) further demonstrated priming for semantic structure when syntactic structure was controlled. Experiment 3 demonstrated repetition of coerced sentences when prime and target used the same coercing verb but not different coercing verbs. Experiment 4 demonstrated repetition of coerced sentences both when the prime and target involved the same (lexically unrealized) coerced action (e.g., building) and when they did not, although repetition was stronger when they did. We argue that speakers use distinct mappings from semantic to syntactic structure when producing coerced and full-form sentences, and propose an account of how a model of language production might incorporate these mappings.

Relative speed of processing affects interference in Stroop and picture–word interference paradigms: evidence from the distractor frequency effect

The Stroop and picture–word interference (PWI) paradigms play a pivotal role in theorising about cognitive processes in general, and language production in particular. A common assumption is that the same mechanisms underlie performance in these two paradigms. Despite this assumption there exist empirical discrepancies. Here we focused on providing a unifying account for the contrasting effects of distractor word frequency in the two paradigms. In four experiments, we addressed whether the contrast is due to inherent design differences between the two paradigms (i.e., grammatical class, stimulus display and relative speed of processing). The results showed that the distractor frequency effect is contingent upon the overall naming latencies in the paradigm, suggesting that the speed of processing target colours and pictures relative to distractor words affects performance in the two tasks. The implications of these results for various models of language production are discussed.

Is word-order similarity necessary for cross-linguistic structural priming?

This article presents two experiments employing two structural priming paradigms that investigated whether cross-linguistic syntactic priming occurred in Chinese and English passive sentences that differ in word order (production-to-production priming in Experiment 1 and comprehension-to-production priming in Experiment 2). Results revealed that cross-linguistic syntactic priming occurred in Chinese and English passive sentences, regardless of production of primes or comprehension of primes and language direction (L1–L2 or L2–L1). Our findings indicate that word-order similarity between languages is not necessary for cross-linguistic structural priming, supporting the view of a two-stage model of language production.

Phonologische Komplexität: Demisilben inVersprechern, Paraphasien und Neologismen

Abstract The following study addresses the question whether slips of the tongue(Frankfurter Versprecherkorpus) alter the phonological complexity of thetargeted demisyllables. The analysis is based on Clements (1990) phonologicaltheory, in which the markedness of demisyllables is calculated onthe basis of major class features and syllabic positions of the respectivephonemes. The analysis of the slips of the tongue showed no significantasymmetry. The erroneous as well as the intended demisyllables did notdiffer in their phonological complexity. Furthermore, slip repairs are insensitiveto phonological complexity. Even for phonological paraphasias phonologicalcomplexity is not a critical feature. In this respect they behavelike slips of the tongue. On the contrary, so-called abstruse neologismsproduced by aphasic patients are mostly constructed with unmarked demisyllables.To explain this difference, we refer to a modified language productionmodel by Buckingham (1990). In contrast to slips and paraphasias,a language processing component that is underused after languageacquisition is completed is activated during the production of abstruse neologismsdue to lexical blocking. This sublexical component, the so-calledrandom generator, fills the syllable positions in such a way that demisyllablesof low phonological complexity are generated.

Template Construction Grammar: From Visual Scene Description to Language Comprehension and Agrammatism

How does the language system coordinate with our visual system to yield flexible integration of linguistic, perceptual, and world-knowledge information when we communicate about the world we perceive? Schema theory is a computational framework that allows the simulation of perceptuo-motor coordination programs on the basis of known brain operating principles such as cooperative computation and distributed processing. We present first its application to a model of language production, SemRep/TCG, which combines a semantic representation of visual scenes (SemRep) with Template Construction Grammar (TCG) as a means to generate verbal descriptions of a scene from its associated SemRep graph. SemRep/TCG combines the neurocomputational framework of schema theory with the representational format of construction grammar in a model linking eye-tracking data to visual scene descriptions. We then offer a conceptual extension of TCG to include language comprehension and address data on the role of both world knowledge and grammatical semantics in the comprehension performances of agrammatic aphasic patients. This extension introduces a distinction between heavy and light semantics. The TCG model of language comprehension offers a computational framework to quantitatively analyze the distributed dynamics of language processes, focusing on the interactions between grammatical, world knowledge, and visual information. In particular, it reveals interesting implications for the understanding of the various patterns of comprehension performances of agrammatic aphasics measured using sentence-picture matching tasks. This new step in the life cycle of the model serves as a basis for exploring the specific challenges that neurolinguistic computational modeling poses to the neuroinformatics community.

The Proximate Phonological Unit of Chinese-English Bilinguals: Proficiency Matters

An essential step to create phonology according to the language production model by Levelt, Roelofs and Meyer is to assemble phonemes into a metrical frame. However, recently, it has been proposed that different languages may rely on different grain sizes of phonological units to construct phonology. For instance, it has been proposed that, instead of phonemes, Mandarin Chinese uses syllables and Japanese uses moras to fill the metrical frame. In this study, we used a masked priming-naming task to investigate how bilinguals assemble their phonology for each language when the two languages differ in grain size. Highly proficient Mandarin Chinese-English bilinguals showed a significant masked onset priming effect in English (L2), and a significant masked syllabic priming effect in Mandarin Chinese (L1). These results suggest that their proximate unit is phonemic in L2 (English), and that bilinguals may use different phonological units depending on the language that is being processed. Additionally, under some conditions, a significant sub-syllabic priming effect was observed even in Mandarin Chinese, which indicates that L2 phonology exerts influences on L1 target processing as a consequence of having a good command of English.

A modular approach to language production: Models and facts

Numerous cortical disorders affect language. We explore the connection between the observed language behavior and the underlying substrates by adopting a neurocomputational approach. To represent the observed trajectories of the discourse in patients with disorganized speech and in healthy participants, we design a graphical representation for the discourse as a trajectory that allows us to visualize and measure the degree of order in the discourse as a function of the disorder of the trajectories. Our work assumes that many of the properties of language production and comprehension can be understood in terms of the dynamics of modular networks of neural associative memories. Based upon this assumption, we connect three theoretical and empirical domains: (1) neural models of language processing and production, (2) statistical methods used in the construction of functional brain images, and (3) corpus linguistic tools, such as Latent Semantic Analysis (henceforth LSA), that are used to discover the topic organization of language. We show how the neurocomputational models intertwine with LSA and the mathematical basis of functional neuroimaging. Within this framework we describe the properties of a context-dependent neural model, based on matrix associative memories, that performs goal-oriented linguistic behavior. We link these matrix associative memory models with the mathematics that underlie functional neuroimaging techniques and present the “functional brain images” emerging from the model. This provides us with a completely “transparent box” with which to analyze the implication of some statistical images. Finally, we use these models to explore the possibility that functional synaptic disconnection can lead to an increase in connectivity between the representations of concepts that could explain some of the alterations in discourse displayed by patients with schizophrenia.

Cross-linguistic influence in non-native languages: explaining lexical transfer using language production models

The focus of this research is on the nature of lexical cross-linguistic influence (CLI) between non-native languages. Using oral interviews with 157 L1 Italian high-school students studying English and German as non-native languages, the project investigated which kinds of lexis appear to be more susceptible to transfer from German to English and discusses this is in the light of multilingual language production models. The results suggest that CLI is more likely to occur with content words and cognates. It also offers a suggestion that such non-native lexical CLI may be due to non-native languages being tagged as ‘foreign’ rather than as individual languages.

Principles of information organization in L2 use: Complex patterns of conceptual transfer

IRAL 51 (2013), 229–242 0019042X/2013/051-229 DOI 10.1515/iral-2013-0010 c ©Walter de Gruyter We will start the discussion of this special issue by placing the present domains of investigation within the broader context of the debate on languagespecificity in conceptualization processes. The debate has gained momentum in recent years, but central issues are still theoretically controversial and empirically inconsistent. The exact nature of the interrelation between language (in the sense of a specific language and specific linguistic structures) and the conceptual processes involved in language use is not clear. We can assume, on the basis of studies in this field and the focus on patterns of actual language usage, that the nature of this interrelation is highly dynamic in that it can differ across tasks, contexts, conditions and levels of complexity. The specific contribution of L2 studies to this debate is that it allows the comparison of processes of conceptualization which relate to the use of a specific language, to those of a different language, within an individual speaker. If cognitive processes during language tasks are somehow shaped by the categories of the respective mother tongue, speakers of a second language should – depending on the level of proficiency and the specific typological profile of the L1 (source language) and L2 (target language) – show divergent patterns from target standards with respect to the selection and organization of content for expression. Cook (1981) has already put the investigation of L2 users forward as a possible window on the language – cognition interrelation: “It would be highly useful if we could, so to speak, disengage the two processes of language and cognitive development and look at people whose level of thinking is out of step with their level of language (i.d second language learners)” (Cook 1981: 255; see also Bassetti and Cook 2011). When focusing on the dynamics of the interrelation between language and cognitive processing, it is useful to draw on a language productionmodel which spells out the different components and phases in the generation of a verbal representation of a given state of affairs. Levelt (1989) distinguishes three mod-

Apraxia of Speech: Concepts and Controversies

This article was written as an editorial to a collection of original articles on apraxia of speech (AOS) in which some of the more recent advancements in the understanding of this syndrome are discussed. It covers controversial issues concerning the theoretical foundations of AOS. Our approach was motivated by a change of perspective on motor speech that has taken place in neurobiology, neurolinguistics, phonology, and phonetics during the past few decades. The literature on AOS is reviewed from 3 different but overlapping perspectives-that is, a disconnection, a motor memory, and a fine motor skill perspective. Separate sections are devoted to the delimitations of AOS from oral facial apraxia, dysarthria, and phonological impairment. We conclude that many of the still unresolved conceptual issues about AOS arise from an underspecification of existing models of spoken language production. We suggest that phonological and motor impairments of sound production should be studied by an integrated approach.

Sound-sized segments are significant for Mandarin speakers

Do speakers of all languages use segmental speech sounds when they produce words? Existing models of language production generally assume a mental representation of individual segmental units, or phonemes, but the bulk of evidence comes from speakers of European languages in which the orthographic system codes explicitly for speech sounds. By contrast, in languages with nonalphabetical scripts, such as Mandarin Chinese, individual speech sounds are not orthographically represented, raising the possibility that speakers of these languages do not use phonemes as fundamental processing units. We used event-related potentials (ERPs) combined with behavioral measurement to investigate the role of phonemes in Mandarin production. Mandarin native speakers named colored line drawings of objects using color adjective-noun phrases; color and object name either shared the initial phoneme or were phonologically unrelated. Whereas naming latencies were unaffected by phoneme repetition, ERP responses were modulated from 200 ms after picture onset. Our ERP findings thus provide strong support for the claim that phonemic segments constitute fundamental units of phonological encoding even for speakers of languages that do not encode such units orthographically.

Naming and categorizing objects: Task differences modulate the polarity of semantic effects in the picture–word interference paradigm

The picture-word interference paradigm is a prominent tool for studying lexical retrieval during speech production. When participants name the pictures, interference from semantically related distractor words has regularly been shown. By contrast, when participants categorize the pictures, facilitation from semantically related distractors has typically been found. In the extant studies, however, differences in the task instructions (naming vs. categorizing) were confounded with the response level: While responses in naming were typically located at the basic level (e.g., "dog"), responses were located at the superordinate level in categorization (e.g., "animal"). The present study avoided this confound by having participants respond at the basic level in both naming and categorization, using the same pictures, distractors, and verbal responses. Our findings confirm the polarity reversal of the semantic effects--that is, semantic interference in naming, and semantic facilitation in categorization. These findings show that the polarity reversal of the semantic effect is indeed due to the different tasks and is not an artifact of the different response levels used in previous studies. Implications for current models of language production are discussed.

Factors determining semantic facilitation and interference in the cyclic naming paradigm.

The cyclic naming paradigm, in which participants are slower to name pictures blocked by semantic category than pictures in an unrelated context, offers a window into the dynamics of the mapping between lexical concepts and words. Here we provide evidence for the view that incremental adjustments to the connection weights from semantics to lexical items provides an elegant explanation of a range of observations within the cyclic naming paradigm. Our principal experimental manipulation is to vary the within-category semantic distance among items that must be named together in a block. In the first set of experiments we find that naming latencies are, if anything, faster for within-category semantically close blocks compared to within-category semantically far blocks, for the first presentation of items. This effect can be explained by the fact that there will be more spreading activation, and thus greater priming at the lexical level, for within-category semantically close blocks than within-category semantically far blocks. We test this explanation by inserting intervening filler items (geometric shapes), and show as predicted, that while intervening unrelated trials abolish short-lived semantic priming effects, the long-lag interference effect that is characteristic of this paradigm is unaffected. These data place new constraints on explanations of the cyclic naming effect, and related phenomena, within a model of language production.

Measuring Coarticulation in Spontaneous Speech: A Preliminary Report

UC Berkeley Phonology Lab Annual Report (2012) Measuring coarticulation in spontaneous speech: a preliminary report Melinda Fricke Keith Johnson University of California, Berkeley Introduction: why study spontaneous speech? Constantly improving processing capabilities have recently opened the door to a type of lan- guage data that was previously impossible to study: the phonetic properties of spontaneous speech. Within the past two decades in particular, studies of the acoustics of conversational speech have become more common, and more sophisticated. The present study seeks to further expand our understanding of natural speech processes by examining coarticulatory patterns in the spontaneously produced speech of adults and children. The study of spontaneous speech can shed considerable light on the workings of the language production system. In normal, everyday conversation, speakers are under quite different (and likely more stringent) pressures than in a laboratory setting. Laboratory speech is typically carefully controlled and often minimally creative; subjects often read words or sentences from a computer screen, repeating the same sentence or type of sentence many times, thus eliminating the usual need to go from concept to sentence construction to articulation. While laboratory studies have been invaluable for developing models of language production, the true testing ground for such models must be spontaneous speech, since it uniquely reveals how language production must proceed in the real world, in real time. The present study is concerned with coarticulatory patterns in particular. Coarticula- tion, the process by which any articulatory gesture affects adjacent articulatory gestures, can be anticipatory (as when knowledge of an upcoming gesture affects the realization of the gesture currently being executed) or perseverative (when an already initiated gesture carries over onto the articulatory realization of a following gesture). This study will be concerned with vowel-on-fricative coarticulation: the present research question is whether acoustic measurements can be identified that distinguish fricatives in a round vowel context from fricatives in a non-round vowel context in spontaneously produced speech. In addition to the main research question, the acoustic measurements under investigation will be ap-

A Critical Review of Levelt's Model of Language Production

A Critical Review of Levelt's Model of Language Production

Lexical activation in bilinguals’ speech production is dynamic: How language ambiguous words can affect cross-language activation

Is the bilingual language production system a dynamic system that can operate in different language activation states? Three experiments investigated to what extent cross-language phonological co-activation effects in language production are sensitive to the composition of the stimulus list. L1 Dutch–L2 English bilinguals decided whether or not a particular phoneme was part of the L2 English name of the picture. The phoneme was either part of the English name of the picture (/b/ or /t/ in bottle), the Dutch name of the picture (/f/ in fles [bottle], the cross-language condition), or was not part of either the English or Dutch names of the picture (/p/, the unrelated condition). In Experiment 1, we added a set of filler pictures with noncognate names in Dutch and English. In contrast, the filler pictures in Experiment 2 all had cognate names in Dutch and English. In Experiment 3 the fillers consisted of a mixture of pictures with cognate (25%) and noncognate (75%) names. Cross-language phonological co-activation appeared sensitive to the composition of the stimulus list: Phonological co-activation effects were observed in Experiments 2 and 3, but not in Experiment 1. The results indicate that the bilingual language production system dynamic and can operate in different modes, depending upon the composition of the stimulus list. We discuss implications for experimental paradigms used in the field of bilingual language production, and for current bilingual language production models.

Evidence for morphological composition at the form level in speech production

We investigated in two picture–word-interference experiments whether there is evidence for composition during compound production (Experiment 1). In Experiment 2, we tried to determine the level at which composition takes place. In Experiment 1, shared morphemes between distractor and target (HANDTASCHE, handbag) sped up naming regardless of category membership: semantically opaque (Plaudertasche, chatterbox) and semantically transparent distractors (Reisetasche, travelling bag) facilitated picture naming to a comparable degree. In Experiment 2, targets (BROTMESSER, bread knife) were combined with a simple word distractor (Kuchen, cake) categorically related to a target compound constituent but not related to the compound as a whole. This target was further paired with a categorically related compound distractor (Kochlöffel, wooden spoon). The simple word distractor was also paired with a categorically related compound constituent (BROT, bread). Whenever distractor (e.g., Kochlöffel, Kuchen) and target (BROTMESSER, BROT, respectively) were categorically related, semantic inhibition was observed. Distractors (e.g., Kuchen) related to only one compound constituent did not affect compound production. Taken together, our results indicate that during compound production a single lemma node is activated and that morphological composition takes place at the form level of representation. Current lexical selection mechanisms in language production models are not supported by these data.