Publications

Abstract

The order of acquisition of Dutch syllable types by first language learners is analyzed as following from an initial ranking and subsequent rerankings of constraints in an optimality theoretic grammar. Initially, structural constraints are all ranked above faithfulness constraints, leading to core syllable (CV) productions only. Subsequently, faithfulness gradually rises to the highest position in the ranking, allowing more and more marked syllable types to appear in production. Local conjunctions of Structural constraints allow for a more detailed analysis.

Abstract

Preparing words in speech production is normally a fast and accurate process. We generate them two or three per second in fluent conversation; and overtly naming a clear picture of an object can easily be initiated within 600 msec after picture onset. The underlying process, however, is exceedingly complex. The theory reviewed in this target article analyzes this process as staged and feedforward. After a first stage of conceptual preparation, word generation proceeds through lexical selection, morphological and phonological encoding, phonetic encoding, and articulation itself. In addition, the speaker exerts some degree of output control, by monitoring of self-produced internal and overt speech. The core of the theory, ranging from lexical selection to the initiation of phonetic encoding, is captured in a computational model, called WEAVER + +. Both the theory and the computational model have been developed in interaction with reaction time experiments, particularly in picture naming or related word production paradigms, with the aim of accounting. for the real-time processing in normal word production. A comprehensive review of theory, model, and experiments is presented. The model can handle some of the main observations in the domain of speech errors (the major empirical domain for most other theories of lexical access), and the theory opens new ways of approaching the cerebral organization of speech production by way of high-temporal-resolution imaging.

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.

Abstract

Making a self-repair in speech typically proceeds in three phases. The first phase involves the monitoring of one’s own speech and the interruption of the flow of speech when trouble is detected. From an analysis of 959 spontaneous self-repairs it appears that interrupting follows detection promptly, with the exception that correct words tend to be completed. Another finding is that detection of trouble improves towards the end of constituents. The second phase is characterized by hesitation, pausing, but especially the use of so-called editing terms. Which editing term is used depends on the nature of the speech trouble in a rather regular fashion: Speech errors induce other editing terms than words that are merely inappropriate, and trouble which is detected quickly by the speaker is preferably signalled by the use of ‘uh’. The third phase consists of making the repair proper The linguistic well-formedness of a repair is not dependent on the speaker’s respecting the integriv of constituents, but on the structural relation between original utterance and repair. A bi-conditional well-formedness rule links this relation to a corresponding relation between the conjuncts of a coordination. It is suggested that a similar relation holds also between question and answer. In all three cases the speaker respects certain Istructural commitments derived from an original utterance. It was finally shown that the editing term plus the first word of the repair proper almost always contain sufficient information for the listener to decide how the repair should be related to the original utterance. Speakers almost never produce misleading information in this respect. It is argued that speakers have little or no access to their speech production process; self-monitoring is probably based on parsing one’s own inner or overt speech.

Abstract

The topic of this paper is the way in which speakers order information in discourse. I will refer to this issue with the term "linearization", and will begin with two types of general remarks. The first one concerns the scope and relevance of the problem with reference to some existing literature. The second set of general remarks will be about the place of linearization in a theory of the speaker. The following, and main part of this paper, will be a summary report of research of linearization in a limited, but well-defined domain of discourse, namely the description of spatial networks.

Abstract

In their comment, Dell and O'Seaghdha (1991) adduced any effect on phonological probes for semantic alternatives to the activation of these probes in the lexical network. We argue that that interpretation is false and, in addition, that the model still cannot account for our data. Furthermore, and different from Dell and O'seaghda, we adduce semantic rebound to the lemma level, where it is so substantial that it should have shown up in our data. Finally, we question the function of feedback in a lexical network (other than eliciting speech errors) and discuss Dell's (1988) notion of a unified production-comprehension system.

Abstract

Spontaneous self-corrections in speech pose a communication problem; the speaker must make clear to the listener not only that the original Utterance was faulty, but where it was faulty and how the fault is to be corrected. Prosodic marking of corrections - making the prosody of the repair noticeably different from that of the original utterance - offers a resource which the speaker can exploit to provide the listener with such information. A corpus of more than 400 spontaneous speech repairs was analysed, and the prosodic characteristics compared with the syntactic and semantic characteristics of each repair. Prosodic marking showed no relationship at all with the syntactic characteristics of repairs. Instead, marking was associated with certain semantic factors: repairs were marked when the original utterance had been actually erroneous, rather than simply less appropriate than the repair; and repairs tended to be marked more often when the set of items encompassing the error and the repair was small rather than when it was large. These findings lend further weight to the characterization of accent as essentially semantic in function.

Abstract

In recent decades, psychologists have become increasingly interested in our ability to speak. This paper sketches the present theoretical perspective on this most complex skill of homo sapiens. The generation of fluent speech is based on the interaction of various processing components. These mechanisms are highly specialized, dedicated to performing specific subroutines, such as retrieving appropriate words, generating morpho-syntactic structure, computing the phonological target shape of syllables, words, phrases and whole utterances, and creating and executing articulatory programmes. As in any complex skill, there is a self-monitoring mechanism that checks the output. These component processes are targets of increasingly sophisticated experimental research, of which this paper presents a few salient examples.

Abstract

Speakers tend to repeat materials from previous talk. This tendency is experimentally established and manipulated in various question-answering situations. It is shown that a question's surface form can affect the format of the answer given, even if this form has little semantic or conversational consequence, as in the pair Q: (At) what time do you close. A: “(At)five o'clock.” Answerers tend to match the utterance to the prepositional (nonprepositional) form of the question. This “correspondence effect” may diminish or disappear when, following the question, additional verbal material is presented to the answerer. The experiments show that neither the articulatory buffer nor long-term memory is normally involved in this retention of recent speech. Retaining recent speech in working memory may fulfill a variety of functions for speaker and listener, among them the correct production and interpretation of surface anaphora. Reusing recent materials may, moreover, be more economical than regenerating speech anew from a semantic base, and thus contribute to fluency. But the realization of this strategy requires a production system in which linguistic formulation can take place relatively independent of, and parallel to, conceptual planning.

Abstract

PdR Press publications in cognition ; 1

Abstract

It has been hypothesised that human reasoning has a non-linguistic foundation, but is nevertheless influenced by the formal means available in a language. For example, Western logic is transparently related to European sentential connectives (e.g., and, if … then, or, not), some of which cannot be unambiguously expressed in other languages. The questionnaire explores reasoning tools and practices through investigating translation equivalents of English sentential connectives and collecting examples of “reasoned arguments”.

Abstract

Demonstratives are key items in understanding how a language constructs and interprets spatial relationships. They are also multi-functional, with applications to non-spatial deictic fields such as time, perception, person and discourse, and uses in anaphora and affect marking. This item consists of an overview of theoretical distinctions in demonstrative systems, followed by a set of practical queries and elicitation suggestions for demonstratives in “table top” space, wider spatial fields, and naturalistic data.

Abstract

The world’s languages encode a diverse range of topological relations. However, cross-linguistic investigation suggests that the relations IN, AT and ON are especially fundamental to the grammaticised expression of space. The purpose of this questionnaire is to collect information about adpositions, case markers, and spatial nominals that are involved in the expression of core IN/AT/ON meanings. The task explores the more general parts of a language’s topological system, with a view to testing certain hypotheses about the packaging of spatial concepts. The questionnaire consists of target translation sentences that focus on a number of dimensions including animacy, caused location and motion.

Abstract

Languages differ widely in terms of how they encode the fundamental concepts of location and position. For some languages, verbs have an important role to play in describing situations (e.g., whether a bottle is standing or lying on the table); for others, verbs are not used in describing location at all. This item outlines certain hypotheses concerning four “types” of languages: those that have verbless basic locatives; those that use a single verb; those that have several verbs available to express location; and those that use positional verbs. The document was originally published as an appendix to the 'Picture series for positional verbs' (https://doi.org/10.17617/2.2573831).

Abstract

In an earlier article (Levinson, 1987b), I raised the possibility that a Gricean theory of implicature might provide a systematic partial reduction of the Binding Conditions; the briefest of outlines is given in Section 2.1 below but the argumentation will be found in the earlier article. In this article I want, first, to show how that account might be further justified and extended, but then to introduce a radical alternative. This alternative uses the same pragmatic framework, but gives an account better adjusted to some languages. Finally, I shall attempt to show that both accounts can be combined by taking a diachronic perspective. The attraction of the combined account is that, suddenly, many facts about long-range reflexives and their associated logophoricity fall into place.

Abstract

This is the German translation of Stephen C. Levinson's »Pragmatics«.

Abstract

Several models of spoken word recognition postulate that recognition is achieved via a process of competition between lexical hypotheses. Competition not only provides a mechanism for isolated word recognition, it also assists in continuous speech recognition, since it offers a means of segmenting continuous input into individual words. We present statistics on the pattern of occurrence of words embedded in the polysyllabic words of the English vocabulary, showing that an overwhelming majority (84%) of polysyllables have shorter words embedded within them. Positional analyses show that these embeddings are most common at the onsets of the longer word. Although both phonological and syntactic constraints could rule out some embedded words, they do not remove the problem. Lexical competition provides a means of dealing with lexical embedding. It is also supported by a growing body of experimental evidence. We present results which indicate that competition operates both between word candidates that begin at the same point in the input and candidates that begin at different points (McQueen, Norris, & Cutler, 1994, Noms, McQueen, & Cutler, in press). We conclude that lexical competition is an essential component in models of continuous speech recognition.

Abstract

In 5 experiments, listeners heard words and nonwords, some cross-spliced so that they contained acoustic-phonetic mismatches. Performance was worse on mismatching than on matching items. Words cross-spliced with words and words cross-spliced with nonwords produced parallel results. However, in lexical decision and 1 of 3 phonetic decision experiments, performance on nonwords cross-spliced with words was poorer than on nonwords cross-spliced with nonwords. A gating study confirmed that there were misleading coarticulatory cues in the cross-spliced items; a sixth experiment showed that the earlier results were not due to interitem differences in the strength of these cues. Three models of phonetic decision making (the Race model, the TRACE model, and a postlexical model) did not explain the data. A new bottom-up model is outlined that accounts for the findings in terms of lexical involvement at a dedicated decision-making stage.

Abstract

Subjects named pictures while hearing distractor words that shared word-initial or word-final segments with the picture names or were unrelated to the picture names. The relative timing of distractor and picture presentation was varied. Compared with unrelated distractors, both types of related distractors facilitated picture naming under certain timing conditions. Begin-related distractors facilitated the naming responses if the shared segments began 150 ms before, at, or 150 ms after picture onset. By contrast, end-related distractors only facilitated the responses if the shared segments began at or 150 ms after picture onset. The results suggest that the phonological encoding of the beginning of a word is initiated before the encoding of its end.

Abstract

The production and interpretation of pronouns involves the identification of a mental referent and, in connected speech or text, a discourse antecedent. One of the few overt signals of the relationship between a pronoun and its antecedent is agreement in features such as number and grammatical gender. To examine how speakers create these signals, two experiments tested conceptual, lexical. and morphophonological accounts of pronoun production in Dutch. The experiments employed sentence completion and continuation tasks with materials containing noun phrases that conflicted or agreed in grammatical gender. The noun phrases served as the antecedents for demonstrative pronouns tin Experiment 1) and relative pronouns tin Experiment 2) that required gender marking. Gender errors were used to assess the nature of the processes that established the link between pronouns and antecedents. There were more gender errors when candidate antecedents conflicted in grammatical gender, counter to the predictions of a pure conceptual hypothesis. Gender marking on candidate antecedents did not change the magnitude of this interference effect, counter to the predictions of an overt-morphology hypothesis. Mirroring previous findings about pronoun comprehension, the results suggest that speakers of gender-marking languages call on specific linguistic information about antecedents in order to select pronouns and that the information consists of specifications of grammatical gender associated with the lemmas of words.

Abstract

Eight experiments were carried out investigating whether different parts of a syllable must be phonologically encoded in a specific order or whether they can be encoded in any order. A speech production task was used in which the subjects in each test trial had to utter one out of three or five response words as quickly as possible. In the so-called homogeneous condition these words were related in form, while in the heterogeneous condition they were unrelated in form. For monosyllabic response words shorter reaction times were obtained in the homogeneous than in the heterogeneous condition when the words had the same onset, but not when they had the same rhyme. Similarly, for disyllabic response words, the reaction times were shorter in the homogeneous than in the heterogeneous condition when the words shared only the onset of the first syllable, but not when they shared only its rhyme. Furthermore, a stronger facilitatory effect was observed when the words had the entire first syllable in common than when they only shared the onset, or the onset and the nucleus, but not the coda of the first syllable. These results suggest that syllables are phonologically encoded in two ordered steps, the first of which is dedicated to the onset and the second to the rhyme.

Abstract

A series of experiments was carried out investigating the time course of phonological encoding in language production, i.e., the question of whether all parts of the phonological form of a word are created in parallel, or whether they are created in a specific order. a speech production task was used in which the subjects in each test trial had to say one out of three or five response words as quickly as possible. In one condition, information was provided about part of the forms of the words to be uttered, in another condition this was not the case. The production of disyllabic words was speeded by information about their first syllable, but not by information about their second syllable. Experiments using trisyllabic words showed that a facilitatory effect could be obtained from information about the second syllable of the words, provided that the first syllable was also known. These findings suggest that the syllables of a word must be encoded strictly sequentially, according to their order in the word.

Abstract

Spoken utterances contain few reliable cues to word boundaries, but listeners nonetheless experience little difficulty identifying words in continuous speech. The authors present data and simulations that suggest that this ability is best accounted for by a model of spoken-word recognition combining competition between alternative lexical candidates and sensitivity to prosodic structure. In a word-spotting experiment, stress pattern effects emerged most clearly when there were many competing lexical candidates for part of the input. Thus, competition between simultaneously active word candidates can modulate the size of prosodic effects, which suggests that spoken-word recognition must be sensitive both to prosodic structure and to the effects of competition. A version of the Shortlist model ( D. G. Norris, 1994b) incorporating the Metrical Segmentation Strategy ( A. Cutler & D. Norris, 1988) accurately simulates the results using a lexicon of more than 25,000 words.

Abstract

Two recent studies (Coulson et al., 1998;Osterhout et al., 1996)examined the
relationship between the event-related brain potential (ERP) responses to linguistic syntactic anomalies (P600/SPS) and domain-general unexpected events (P300). Coulson et al. concluded that these responses are highly similar, whereas Osterhout et al. concluded that they are distinct. In this comment, we evaluate the relativemerits of these claims. We conclude that the available evidence indicates that the ERP response to syntactic anomalies is at least partially distinct from the ERP response to unexpected anomalies that do not involve a grammatical violation

Abstract

Japanese, British English and American English listeners were presented with spoken words in their native language, and asked to mark on a written transcript of each word the first natural division point in the word. The results showed clear and strong patterns of consensus, indicating that listeners have available to them conscious representations of within-word structure. Orthography did not play a strongly deciding role in the results. The patterns of response were at variance with results from on-line studies of speech segmentation, suggesting that the present task taps not those representations used in on-line listening, but levels of representation which may involve much richer knowledge of word-internal structure.

Abstract

Two experiments examined the processing of Tokyo Japanese pitchaccent distinctions by native speakers of Japanese from two accentlessvariety areas. In both experiments, listeners were presented with Tokyo Japanese speech materials used in an earlier study with Tokyo Japanese listeners, who clearly exploited the pitch-accent information in spokenword recognition. In the "rst experiment, listeners judged from which of two words, di!ering in accentual structure, isolated syllables had been extracted. Both new groups were, overall, as successful at this task as Tokyo Japanese speakers had been, but their response patterns differed from those of the Tokyo Japanese, for instance in that a bias towards H judgments in the Tokyo Japanese responses was weakened in the present groups' responses. In a second experiment, listeners heard word fragments and guessed what the words were; in this task, the speakers from accentless areas again performed significantly above chance, but their responses showed less sensitivity to the information in the input, and greater bias towards vocabulary distribution frequencies, than had been observed with the Tokyo Japanese listeners. The results suggest that experience with a local accentless dialect affects the processing of accent for word recognition in Tokyo Japanese, even for listeners with extensive exposure to Tokyo Japanese.

Abstract

"Event realisation" refers to the normal final state of the affected entity of an activity described by a verb. For example, the sentence John killed the mosquito entails that the mosquito is afterwards dead – this is the full realisation of a killing event. By contrast, a sentence such as John hit the mosquito does not entail the mosquito’s death (even though we might assume this to be a likely result). In using a certain verb, which features of event realisation are entailed and which are just likely? This questionnaire supports cross-linguistic exploration of event realisation for a range of event types.

Abstract

In the present PET study we explore some functional aspects of the interaction between attentional/control processes and learning/memory processes. The network of brain regions supporting recall of abstract designs were studied in a less practiced and in a well practiced state. The results indicate that automaticity, i.e., a decreased dependence on attentional and working memory resources, develops as a consequence of practice. This corresponds to the practice related decreases of activity in the prefrontal, anterior cingulate, and posterior parietal regions. In addition, the activity of the medial temporal regions decreased as a function of practice. This indicates an inverse relation between the strength of encoding and the activation of the MTL during retrieval. Furthermore, the pattern of practice related increases in the auditory, posterior insular-opercular extending into perisylvian supra marginal region, and the right mid occipito-temporal region, may reflect a lower degree of inhibitory attentional modulation of task irrelevant processing and more fully developed representations of the abstract designs, respectively. We also suggest that free recall is dependent on bilateral prefrontal processing, in particular non-automatic free recall. The present results cofirm previous functional neuroimaging studies of memory retrieval indicating that recall is subserved by a network of interacting brain regions. Furthermore, the results indicate that some components of the neural network subserving free recall may have a dynamic role and that there is a functional restructuring of the information processing networks during the learning process.

Abstract

Recent event-related FMRI studies indicate that the prefrontal (PFC) and the medial temporal lobe (MTL) regions are more active during effective encoding than during ineffective encoding. The within-subject design and the use of well-educated young college students in these studies makes it important to replicate these results in other study populations. In this PET study, we used an auditory word-pair association cued-recall paradigm and investigated a group of healthy upper middle-aged/older illiterate women. We observed a positive correlation between cued-recall success and the regional cerebral blood flow of the left inferior PFC (BA 47) and the MTLs. Specifically, we used the cuedrecall success as a covariate in a general linear model and the results confirmed that the left inferior PFC and the MTLare more active during effective encoding than during ineffective encoding. These effects were observed during encoding of both semantically and phonologically related word pairs, indicating that these effects are robust in the studied population, that is, reproducible within group. These results generalize the results of Brewer et al. (1998, Science 281, 1185– 1187) and Wagner et al. (1998, Science 281, 1188–1191) to an upper middle aged/older illiterate population. In addition, the present study indicates that effective relational encoding correlates positively with the activity of the anterior medial temporal lobe regions.

Abstract

A fundamental problem in the study of learning is that learning-related changes may be confounded by nonspecific time effects. There are several strategies for handling this problem. This problem may be of greater significance in functional magnetic resonance imaging (fMRI) compared to positron emission tomography (PET). Using the general linear model, we describe, compare, and discuss two approaches for separating learning-related from nonspecific time effects. The first approach makes assumptions on the general behavior of nonspecific effects and explicitly models these effects, i.e., nonspecific time effects are incorporated as a linear or nonlinear confounding covariate in the statistical model. The second strategy makes no a priori assumption concerning the form of nonspecific time effects, but implicitly controls for nonspecific effects using an interaction approach, i.e., learning effects are assessed with an interaction contrast. The two approaches depend on specific assumptions and have specific limitations. With certain experimental designs, both approaches may be used and the results compared, lending particular support to effects that are independent of the method used. A third and perhaps better approach that sometimes may be practically unfeasible is to use a completely temporally balanced experimental design. The choice of approach may be of particular importance when learning related effects are studied with fMRI.

Abstract

The objective of this study was to investigate the central processing of dynamic mechanical allodynia in patients with mononeuropathy. Regional cerebral bloodflow, as an indicator of neuronal activity, was measured with positron emission tomography. Paired comparisons were made between three different states; rest, allodynia during brushing the painful skin area, and brushing of the homologous contralateral area. Bilateral activations were observed in the primary somatosensory cortex (S1) and the secondary somatosensory cortex (S2) during allodynia compared to rest. The S1 activation contralateral to the site of the stimulus was more expressed during allodynia than during innocuous touch. Significant activations of the contralateral posterior parietal cortex, the periaqueductal gray (PAG), the thalamus bilaterally and motor areas were also observed in the allodynic state compared to both non-allodynic states. In the anterior cingulate cortex (ACC) there was only a suggested activation when the allodynic state was compared with the non-allodynic states. In order to account for the individual variability in the intensity of allodynia and ongoing spontaneous pain, rCBF was regressed on the individually reported pain intensity, and significant covariations were observed in the ACC and the right anterior insula. Significantly decreased regional blood flow was observed bilaterally in the medial and lateral temporal lobe as well as in the occipital and posterior cingulate cortices when the allodynic state was compared to the non-painful conditions. This finding is consistent with previous studies suggesting attentional modulation and a central coping strategy for known and expected painful stimuli. Involvement of the medial pain system has previously been reported in patients with mononeuropathy during ongoing spontaneous pain. This study reveals a bilateral activation of the lateral pain system as well as involvement of the medial pain system during dynamic mechanical allodynia in patients with mononeuropathy.

Abstract

This study investigated characteristics of motor cortex activation and response generation in Parkinson's disease with measures of electrocortical activity (lateralized readiness potential [LRP]), electromyographic activity (EMG), and isometric force in a noise-compatibility task. When presented with stimuli consisting of incompatible target and distracter elements asking for responses of opposite hands, patients were less able than control subjects to suppress activation of the motor cortex controlling the wrong response hand. This was manifested in the pattern of reaction times and in an incorrect lateralization of the LRP. Onset latency and rise time of the LRP did not differ between patients and control subjects, but EMG and response force developed more slowly in patients. Moreover, in patients but not in control subjects, the rate of development of EMG and response force decreased as reaction time increased. We hypothesize that this dissociation between electrocortical activity and peripheral measures in Parkinson's disease is the result of changes in motor cortex function that alter the relation between signal-related and movement-related neural activity in the motor cortex. In the LRP, this altered balance may obscure an abnormal development of movement-related neural activity.

Abstract

A patient who already had Wernick's aphasia due to a left temporal lobe lesion suffered a severe deterioration specifically of auditory language comprehension, subsequent to right temporal lobe infarction. A detailed comparison of his new condition with his language status before the second stroke revealed that the newly acquired deficit was limited to tasks related to auditory input. Further investigations demonstrated a speech perceptual disorder, which we analysed as due to deficits both at the level of general auditory processes and at the level of phonetic analysis. We discuss some arguments related to hemisphere specialization of phonetic processing and to the disconnection explanation of word deafness that support the hypothesis of word deafness being generally caused by mixed deficits.

Abstract

Speakers can use pronouns when their conceptual referents are accessible from the preceding discourse, as in 'The flower is red. It turns blue'. Theories of language production agree that in order to produce a noun semantic, syntactic, and phonological information must be accessed. However, little is known about lexical access to pronouns. In this paper, we propose a model of pronoun access in German. Since the forms of German pronouns depend on the grammatical gender of the nouns they replace, the model claims that speakers must access the syntactic representation of the replaced noun (its lemma) to select a pronoun. In two experiments using the lexical decision during naming paradigm [Levelt, W.J.M., Schriefers, H., Vorberg, D., Meyer, A.S., Pechmann, T., Havinga, J., 1991a. The time course of lexical access in speech production: a study of picture naming. Psychological Review 98, 122-142], we investigated whether lemma access automatically entails the activation of the corresponding word form or whether a word form is only activated when the noun itself is produced, but not when it is replaced by a pronoun. Experiment 1 showed that during pronoun production the phonological form of the replaced noun is activated. Experiment 2 demonstrated that this phonological activation was not a residual of the use of the noun in the preceding sentence. Thus, when a pronoun is produced, the lemma and the phonological form of the replaced noun become reactivated.

Abstract

According to certain theories of language production, lexical access to a content word consists of two independent and serially ordered stages. In the first, semantically driven stage, so-called lemmas are retrieved, i.e., lexical items that are specified with respect to syntactic and semantic properties, but not with respect to phonological characteristics. In the second stage, the corresponding wordforms, the so-called lexemes, are retrieved. This implies that the access to a content word involves an early stage of exclusively semantic activation and a later stage of exclusively phonological activation. This seriality assumption was tested experimentally, using a picture-word interference paradigm in which the interfering words were presented auditorily. The results show an interference effect of semantically related words on picture naming latencies at an early SOA (− 150 ms), and a facilitatory effect of phonologically related words at later SOAs (0 ms, + 150 ms). On the basis of these results it can be concluded that there is indeed a stage of lexical access to a content word where only its meaning is activated, followed by a stage where only its form is activated. These findings can be seen as empirical support for a two-stage model of lexical access, or, alternatively, as putting constraints on the parameters in a network model of lexical access, such as the model proposed by Dell and Reich.