Articulatory Planning Research Articles

Non-linguistic auditory processing in stuttering: Evidence from behavior and event-related brain potentials

Auditory processing deficits are hypothesized to play a role in the disorder of stuttering (e.g. Hall, J. W., & Jerger, J. (1978). Central auditory function in stutterers. Journal of Speech and Hearing Research, 21, 324–337). The current study focused on non-linguistic auditory processing without verbal responses to explore the relationship between behavior and neural activity in the absence of cognitive demands related to language processing and articulatory planning for speaking. A pure-tone, oddball paradigm was utilized to compare behavioral accuracy and reaction times for adults who stutter (AWS) and normally fluent speakers (NFS). Additionally, event-related potentials elicited by brief standard and target tones were compared for the two groups. Results revealed that, as a group, AWS tended to perform less accurately compared to the NFS and were slower to respond to target stimuli. However, inspection of individual data indicated that most of the AWS performed within the range of normally fluent speakers while a small subset of AWS were well outside the normal range. This subgroup of AWS also demonstrated early perceptual processes (as indexed by N100 and P200 amplitudes) indicative of reduced cortical representation of auditory input. The P300 mean amplitudes elicited in AWS tended to be reduced overall compared to those of the NFS, suggesting the possibility of weaker updates in working memory for representations of the target tone stimuli in AWS. Taken together, these findings point to the possibility that a subset of AWS exhibit non-linguistic auditory processing deficits related to altered cortical processing. Educational objectives : After reading this article, the reader will be able to: (1) summarize research findings of non-linguistic auditory processing in stuttering; (2) discuss the relationship between behavioral performance for auditory processing and the underlying event-related brain potentials; (3) discuss the importance of analyses of individual versus group data in stuttering; and (4) summarize how the findings of this study relate to a multifactorial model of stuttering.

Structural and functional abnormalities of the motor system in developmental stuttering.

Though stuttering is manifest in its motor characteristics, the cause of stuttering may not relate purely to impairments in the motor system as stuttering frequency is increased by linguistic factors, such as syntactic complexity and length of utterance, and decreased by changes in perception, such as masking or altering auditory feedback. Using functional and diffusion imaging, we examined brain structure and function in the motor and language areas in a group of young people who stutter. During speech production, irrespective of fluency or auditory feedback, the people who stuttered showed overactivity relative to controls in the anterior insula, cerebellum and midbrain bilaterally and underactivity in the ventral premotor, Rolandic opercular and sensorimotor cortex bilaterally and Heschl's gyrus on the left. These results are consistent with a recent meta-analysis of functional imaging studies in developmental stuttering. Two additional findings emerged from our study. First, we found overactivity in the midbrain, which was at the level of the substantia nigra and extended to the pedunculopontine nucleus, red nucleus and subthalamic nucleus. This overactivity is consistent with suggestions in previous studies of abnormal function of the basal ganglia or excessive dopamine in people who stutter. Second, we found underactivity of the cortical motor and premotor areas associated with articulation and speech production. Analysis of the diffusion data revealed that the integrity of the white matter underlying the underactive areas in ventral premotor cortex was reduced in people who stutter. The white matter tracts in this area via connections with posterior superior temporal and inferior parietal cortex provide a substrate for the integration of articulatory planning and sensory feedback, and via connections with primary motor cortex, a substrate for execution of articulatory movements. Our data support the conclusion that stuttering is a disorder related primarily to disruption in the cortical and subcortical neural systems supporting the selection, initiation and execution of motor sequences necessary for fluent speech production.

On the evolution of consonant harmony: the case of secondary articulation agreement

Consonant harmony involves long-distance featural assimilation, or agreement, of consonants across intervening segments. Current correspondence-based analyses of such sound patterns assume that they originate in the cognitive exigencies of articulatory planning, either synchronically, through the functional grounding of the constraints responsible, or diachronically, whereby processing factors incrementally shape the lexicon over time. This paper challenges the validity of this assumption as an all-purpose functional explanation for the full range of long-distance consonant agreement patterns by demonstrating that a variety of diachronic trajectories underlies their emergence and evolution. Focusing on the comparatively rare phenomenon ofsecondary articulation agreement, the evolutionary histories of three cases are examined: (labio)velarisation agreement in Pohnpeian (Oceanic), palatalisation agreement in Karaim (Turkic) and pharyngealisation agreement in Tsilhqot'in (Athabaskan). These histories provide explanations for a range of synchronic properties of the systems in question, some of which are problematic for restrictive typologies of consonant harmony.

Use of word length information in utterance planning

Griffin [Griffin, Z. M. (2003). A reversed length effect in coordinating the preparation and articulation of words in speaking. Psychonomic Bulletin & Review, 10, 603–609.] found that speakers naming object pairs spent more time before utterance onset looking at the second object when the first object name was short than when it was long. She proposed that this reversed length effect arose because the speakers’ decision when to initiate an utterance was based, in part, on their estimate of the spoken duration of the first object name and the time available during its articulation to plan the second object name. In Experiment 1 of the present study, participants named object pairs. They spent more time looking at the first object when its name was monosyllabic than when it was trisyllabic, and, as in Griffin’s study, the average gaze-speech lag (the time between the end of the gaze to the first object and onset of its name, which corresponds closely to the pre-speech inspection time for the second object) showed a reversed length effect. Experiments 2 and 3 showed that this effect was not due to a trade-off between the time speakers spent looking at the first and second object before speech onset. Experiment 4 yielded a reversed length effect when the second object was replaced by a symbol ( x or +), which the participants had to categorise. We propose a novel account of the reversed length effect, which links it to the incremental nature of phonological encoding and articulatory planning rather than the speaker’s estimate of the length of the first object name.

No Difference in Short-Term Memory Span Between Sign and Speech

The conclusions of Bavelier, Newport, Hall, Supalla, and Boutla (2006, this issue) are undermined by an inappropriate procedure for measuring articulation rate. They measured duration for reading a matrix of 200 letters (D. Bavelier, personal communication, May 17, 2006), which may involve a considerable load on visual attention, translation from print, and articulatory planning. This procedure may slow responses to a point where constraints on articulatory production cease to matter. With appropriate methods (see the footnote to Table 1), articulation rates are much faster than they found (cf. our Table 1 with their Table 1), which suggests their methods did indeed introduce processing difficulties. With our methods, the stimuli we used in our study (Wilson & Emmorey, 2006) produce equal articulation rates in American Sign Language (ASL) and English, t(3) = 0.57, d = 0.24, pTep = .64. In contrast, the English stimuli Bavelier et al. used are articulated substantially faster than the ASL stimuli in their previous study (Boutla, Supalla, Newport, & Bavelier, 2004), t(3) = 5.40, d = 1.20, prep = .96. We conclude that (a) the equal spans and equal WAIS (Wechsler Adult Intelligence Scale) scores for English and ASL that we reported (Table 1) cannot be explained by failure to match articulation rate, and (b) the span difference between English and ASL seen when comparing the results of Bavelier et al. and Boutla et al. (Table 1) can be explained by failure to match articulation rate. Bavelier et al. also compare span for English digits with span for the ASL letters they used in their earlier study, but growing evidence shows that digits are not comparable to other stimuli in short-term memory (e.g., Knops, Nuerk, Fimm, Vohn, & Willmes, 2006). Contra Bavelier et al., this lack of comparability cannot be explained by the fact that letters share more vowel sounds than numbers do. Instead, numbers have a privileged status in short-t rm memory, and cannot be treated as equivalent to other stimulus categories. In short, Bavelier et al. do not present any new comparison that cont dicts our findings. But have they nevertheless cast doubt on our findings? The stimuli Bavelier et al. used are articulated substantially faster than our English stimuli, *(3) = 4.07, d = 1.29,prep = .94, which suggests that articulation rate accounts for the difference in span between these two sets. But contrary to what Bavelier et al. imply, this does not show that our stimuli are more phonologically similar. Phonological similarity and articulation rate are separate factors. Might phonological similarity nevertheless contribute to the difference between their results and ours? Perhaps. Our stimulus set is indeed more phonologically similar, though not by much. (In our stimuli, vowel sounds are shared by F, S, and X; by B and V; and by H and K. In the stimuli used by Bavelier et al., vowel sounds are shared by M and S, by G and P, and by H and K.) But this misses the point. The real question is not how one set of English stimuli compares with another, but how English compares with ASL. Could the equal spans for English and ASL that we found (Wilson & Emmorey, 2006) be explained by failure to match phonological similarity? This is possible, but Bavelier et al. have provided no evidence that speaks to this question. Our data s ill stand as the best comparison to date between ASL and English. Finally, we address the fourto six-item limit on ASL span. We agree that speech can yield higher spans, but this is of little theoretical interest. Instead, these higher spans are attributable to the well-understood factor of articulation rate, for which speech holds an enormous advantage. We conclude that known factors account for differences in span, and that there is no residual difference when these factors are controlled. Address correspondence to Margaret Wilson, Department of Psychology, UCSC, Santa Cruz, CA 95064, e-mail: mlwilson@ucsc.edu.

The brain bases of reading late learned words: Evidence from functional MRI1

Age of acquisition (AoA) effects for reading have been documented in the literature. Recent studies have suggested that more effortful phonological retrieval is involved in picture naming for late learned words. Here we use functional magnetic resonance imaging to determine the neuroanatomical correlates of AoA effects in reading, and to explore whether AoA effects in the brain modulate activity in areas devoted to phonological processing. Participants were asked to covertly name blocks of early and late learned words while being scanned with functional MRI. Results revealed a relative increase in neural activity for late learned words in a set of brain areas involved in auditory-phonological processing (the planum temporale of the posterior-superior left temporal lobe) and articulatory motor planning (BA 44, putamen, and globus pallidus). These findings support the view that late learned words involve more effortful phonological retrieval, but in addition point to a second locus of the AoA effect in the articulatory component of the reading system.

Abstract Grammatical Processing of Nouns and Verbs in Broca's Area: Evidence from FMRI

The role of Broca's area in grammatical computation is unclear, because syntactic processing is often confounded with working memory, articulation, or semantic selection. Morphological processing potentially circumvents these problems. Using event-related functional magnetic resonance imaging (fMRI), we had 18 subjects silently inflect words or read them verbatim. Subtracting the activity pattern for reading from that for inflection, which indexes processes involved in inflection (holding constant lexical processing and articulatory planning) highlighted left Brodmann area (BA) 44/45 (Broca's area), BA 47, anterior insula, and medial supplementary motor area. Subtracting activity during zero inflection ( the hawk; they walk) from that during overt inflection ( the hawks; they walked), which highlights manipulation of phonological content, implicated subsets of the regions engaged by inflection as a whole. Subtracting activity during verbatim reading from activity during zero inflection (which highlights the manipulation of inflectional features) implicated distinct regions of BA 44, 47, and a premotor region (thereby tying these regions to grammatical features), but failed to implicate the insula or BA 45 (thereby tying these to articulation). These patterns were largely similar in nouns and verbs and in regular and irregular forms, suggesting these regions implement inflectional features cutting across word classes. Greater activity was observed for irregular than regular verbs in the anterior cingulate and supplementary motor area (SMA), possibly reflecting the blocking of regular or competing irregular candidates. The results confirm a role for Broca's area in abstract grammatical processing, and are interpreted in terms of a network of regions in left prefrontal cortex (PFC) that are recruited for processing abstract morphosyntactic features and overt morphophonological content.

Articulatory Planning Is Continuous and Sensitive to Informational Redundancy

This study investigates the relationship between word repetition, predictabil-ityfrom neighbouring words, and articulatory reduction in Dutch. For the sevenmost frequent words ending in the adjectival suffix -lijk, 40 occurrences were ran-domlyselected from a large database of face-to-face conversations. Analysis ofthe selected tokens showed that the degree of articulatory reduction (as measuredby duration and number of realized segments) was affected by repetition, pre-dictabilityfrom the previous word and predictability from the following word.Interestingly, not all of these effects were significant across morphemes and tar-getwords. Repetition effects were limited to suffixes, while effects of predictabil-ityfrom the previous word were restricted to the stems of two of the seven targetwords. Predictability from the following word affected the stems of all targetwords equally, but not all suffixes. The implications of these findings for models ofspeech production are discussed.

Articulatory planning in naming single word-kanji characters.

Yamazaki, et al. (1997) rejected word length measured by number of syllables as not significant in accounting for naming latencies for Japanese kanji characters. This study reexamined Yamazaki, et al.'s data and found that monosyllabic characters were named slower than two- and three-syllable characters. This finding is taken as suggesting the presence of an articulatory planning stage in speech production in Japanese.

The speaking brain: a tutorial introduction to fMRI experiments in the production of speech, prosody and syntax

In this study we will give an overview of the experimental work on the neuroanatomical correlates of language and speech production that we have done in recent years. First we will introduce the methodology of event-related functional magnetic neuro-imaging and the experimental paradigm that we employed. Then we will present and discuss the results of our experiments on (1) speech motor control, (2) articulatory complexity, (3) the neuroanatomical correlates of prosody, and (4) the neurocognitive substrates of syntactic processing. Experiments (1) and (2) show that the expected large motor speech network consisting of SMA, motor cortex and cerebellum is only active in planning and execution of simple articulatory movements. Increased articulatory complexity leads to more focused activation. Furthermore, we can show that only the execution of speech movements recruits the left anterior insula, while articulatory planning does not. The results of experiment (3) indicate that it is not the function of prosody (linguistic vs affective) that controls lateralization of prosodic processing, but that more general characteristics of the processing units like the size of the prosodic frame are responsible for the activation of different cortical regions. Finally, in experiment (4) we present first results on syntactic processing in speech production. Besides the expected activation of Broca's area we found activations in Wernicke's area and in the cerebellum. We have also found evidence for activations in other cortical areas, which are less often implicated in clinical studies on brain language correlations. The cognitive relevance of these areas and networks is still to be elucidated.

Nature of phonological delay in children with specific language impairment

This study investigated the nature of phonological delay in a group of children with specific language impairment. It was asked whether phonological errors in this group of children were generated by a slow but normal language learning process or whether they reflected a selective impairment in some representations that enhance normal acquisition and use of a language phonology. A group of 10 children with SLI (mean age = 5.1) was compared with three groups of normal children who were matched in age (age control group, mean age = 5.1), in sentence comprehension and recalling (grammar control group, mean age = 3.7), or who exhibited a phonological performance lower than the age average (group with low phonological performance, mean age = 4.4). The four groups of children were assessed in terms of: (1) responses to a mispronunciation detection task; and (2) error profiles with complex and simple syllabic structures. Performance on the mispronunciation detection task showed that the group with SLI could distinguish a target lexical item from acoustic non-word stimuli that were highly similar to it in terms of phonetic characteristics. An analysis of overall error rate at this task showed, however, that four children with SLI had a much lower performance than normal children of the same age, even when the auditory stimuli were tokens of the target word, or non-words that were phonetically different from the target. A difficulty in coordinating vocal actions in an articulatory plan accounted for error profiles with simple syllabic structures both for some children with SLI and normal children with phonological performance lower than the age average. A severe difficulty with representing complex syllabic structures was a homogeneous characteristic of the group with SLI and worked as the main indicator of impaired, rather than simply slow, phonological development.

Error-related brain activation during a Go/NoGo response inhibition task.

Inhibitory control and performance monitoring are critical executive functions of the human brain. Lesion and imaging studies have shown that the inferior frontal cortex plays an important role in inhibition of inappropriate response. In contrast, specific brain areas involved in error processing and their relation to those implicated in inhibitory control processes are unknown. In this study, we used a random effects model to investigate error-related brain activity associated with failure to inhibit response during a Go/NoGo task. Error-related brain activation was observed in the rostral aspect of the right anterior cingulate (BA 24/32) and adjoining medial prefrontal cortex, the left and right insular cortex and adjoining frontal operculum (BA 47) and left precuneus/posterior cingulate (BA 7/31/29). Brain activation related to response inhibition and competition was observed bilaterally in the dorsolateral prefrontal cortex (BA 9/46), pars triangularis region of the inferior frontal cortex (BA 45/47), premotor cortex (BA 6), inferior parietal lobule (BA 39), lingual gyrus and the caudate, as well as in the right dorsal anterior cingulate cortex (BA 24). These findings provide evidence for a distributed error processing system in the human brain that overlaps partially, but not completely, with brain regions involved in response inhibition and competition. In particular, the rostal anterior cingulate and posterior cingulate/precuneus as well as the left and right anterior insular cortex were activated only during error processing, but not during response competition, inhibition, selection, or execution. Our results also suggest that the brain regions involved in the error processing system overlap with brain areas implicated in the formulation and execution of articulatory plans.

Modality Effects and the Development of the Word Length Effect in Children

Two experiments investigated the development of the word length effect in children aged 4 to 10 years, comparing auditory and visual stimuli. The question addressed was whether word length effects emerged earlier with auditory presentation or visual presentation, or whether they emerged at the same age regardless of presentation modality. Results provided evidence that word length effects emerge earlier with visual than auditory presentation. The implication of our results is that with visual presentation, 4-year-olds engage in some form of verbalisation strategy that involves obtaining phonological representations of picture names and mapping them on to articulatory output plans. This strategy is clearly verbal in nature, but is not necessarily characterised as cumulative verbal rehearsal.

Brain regions involved in articulation

The left inferior frontal gyrus (Broca's area) is generally believed to be critical for the motor act of speech. A lesion-based analysis has, however, shown that the left anterior insula is necessary for accurate articulation. We used functional imaging in normal people to show the neural systems involved in speech during different speech tasks. 12 normal people underwent positron emission tomography with oxygen-15-labelled water as tracer. We measured cerebral activity while participants performed three different tasks: repetition of heard nouns at different rates; listening to single nouns at different rates; and anticipation of listening or repetition. We analysed the data with imaging software. Repetition of single words did not activate Broca's area but activity in three left-lateralised regions was seen: the anterior insula, a localised region in the lateral premotor cortex, and the posterior pallidum. The left anterior insula and lateral premotor cortex showed a conjunction of activity for hearing and articulation. In addition, articulation modulated the response to hearing words in the left dorsolateral temporal cortex, the physiological expression of the speaker's auditory attention being directed towards the stimuli and not his or her articulated responses. The formulation of an articulatory plan is a function of the left anterior insula and lateral premotor cortex and not of Broca's area. The left basal ganglia seem to be dominant for speech, although the axial muscles involved receive their motor output from both cerebral hemispheres.

Where Is the Length Effect? A Cross-Linguistic Study of Speech Production

Many models of speech production assume that one cannot begin to articulate a word before all its segmentalunits are inserted into the articulatory plan. Moreover, some of these models assume that segments are seriallyinserted from left to right. As a consequence, latencies to name words should increase with word length. In aseries of five experiments, however, we showed that the time to name a picture or retrieve a word associatedwith a symbol is not affected by the length of the word. Experiments 1 and 2 used French materials and participants, while Experiments 3, 4, and 5 were conducted with English materials and participants. These results are discussed in relation to current models of speech production and previous reports of length effects are reevaluated in light of these findings. We conclude that if words are encoded serially, then articulation can start before an entire phonological word has been encoded.

Phonological priming effects in stutterers

It is hypothesized that stuttering is caused by a perturbation of phonological encoding, i.e., the construction of the fully specified articulatory program on the basis of word form information stored in the mental lexicon. Nine stutterers and nine nonstutterers participated in a phonological priming experiment. In each trial, they were required to utter one word from a set of five as fast as possible upon visual presentation of a related cue word. In the so-called homogeneous conditions, the response words shared initial segments, either the initial consonant or the initial consonant and the subsequent vowel. In the heterogeneous conditions, the response words were phonemically unrelated. Nonstutterers had shorter speech onset latencies in the homogeneous conditions than in the heterogeneous conditions, and the difference was larger for the words sharing both consonant and vowel than for the words sharing the initial consonant only. In most stutterers, a reduction of speech onset occured only when the words shared both consonant and vowel. These results are taken to indicate that in stutterers the encoding of noninitial parts of syllables, particularly the (stressed) vowel, is delayed. The primary symptoms of stuttering—repetition or prolongation of syllable-initial segments—are argued to result from attempts at executing a syllable prior to the incorporation of correct vowel information in the articulatory plan.

Linguistic Processing of Words and Pictures

Theios and Amrhein (1989) proposed that longer picture-naming than wordreading times were due in part to differences in the time to access articulatory codes. Three studies of university students were designed to test that assumption. Latencies to articulate entire names were faster than latencies to articulate the name of the first letter of the referent both for pictures and words; however, the difference was significantly larger for pictures. The absence of the interaction when articulation was delayed suggests that the effect reflected articulatory planning rather than the articulation itself. Finally, the interaction occurred when subjects compared visual target letters or words with comparison words or pictures. The results support the assumption that pictures are more likely than words to activate a holistic lexical representation of their referents.

Dysarthria in Friedreich's ataxia: Timing of speech segments

Under certain conditions Friedreich's ataxia (FA) can be considered a model of afferent cerebellar dysfunction. In seven subjects suffering from FA acoustic analysis of sentence utterances, including durational measurements of syllables and intrasyllabic segments, was performed. The FA patients presented with prolongation of syllable, vowel and occlusion lengths. Furthermore, they showed less durational contrasts between stressed and unstressed syllables. These abnormalities may be explained by a slowing of articulatory movements. In contrast, the voice onset time (VOT) values and the variation coefficients of the segments considered were not consistently increased. With respect to interarticulatory coordination, thus, a specific timing deficit could not be disclosed in FA dysarthria. These results are discrepant to data obtained with syllable repetition tasks, as reported in the literature. Obviously, the motor control of sentence utterances differs from that of syllable repetitions. There is psycholinguistic evidence that the complex movement patterns underlying sentence utterances are controlled by a single articulatory plan. The cerebellum, therefore, might not be a prerequisite for the initiation of articulatory gestures within the framework of sentence utterances.

Intonation contours, prosodic structure and preboundary lengthening

The aim of the perception experiment reported here was to establish first whether English listeners expect longer preboundary syllable durations as the rank of the phonological boundary is higher; and, second, whether they expect longer preboundary syllable durations when the phonological boundary coincides with an intonation contour boundary. Judges were instructed to decide whether the duration of the pre boundary syllable was “all right”, “too long”, or “too short”. The stimuli were resynthesized utterances in which the durational characteristics of the boundary were varied. Although the results are not totally consistent, they suggest affirmative answers to both questions. The result for the second is particularly interesting, because the intonation contour boundary was not in all cases marked by a pitch movement on the experimental syllable itself. The perception of the intonation contour boundary can therefore take place on the basis of the listener’s interpretation of the utterance, together with the more global characteristics of the pitch contour. A third finding was that intonation contour boundaries elicited far more “call right” judgements than other boundaries. We tentatively suggest that this behaviour is the perceptual counterpart of the relaxation of articulatory control at the end of articulatory planning units.

The time course of lexical access in speech production: A study of picture naming.

Nijmegen University Nijmegen, The Netherlands Lexical access in object naming involves the activation of a set oflexical candidates, the selection of the appropriate (or target) item, and the phonological encoding of that item. Two views of lexical access in naming are compared. From one view, the 2-stage theory, phonological activation follows selection of the target item and is restricted to that item. From the other view, which is most explicit in activation-spreading theories, all activated lexical candidates are phonologically activated to some extent. A series of experiments is reported in which subjects performed acoustic lexical decision during object naming at different stimulus-onset asynchronies. The experiments show semantic activation of lexical candidates and phonological activation of the target item, but no phonological activation of other semantically activated items. This supports the 2-stage view. More- over, a mathematical model embodying the 2-stage view is fully compatible with the lexical deci- sion data obtained at different stimulus-onset asynchronies. One of a speaker's core skills is to lexicalize the concepts intended for expression. Lexicalization proceeds at a rate of two to three words per second in normal spontaneous speech, but doubling this rate is possible and not exceptional. The skill of lexicalizing a content word involves two components. The first one is to select the appropriate lexical item from among some tens of thousands of alternatives in the mental lexicon. The second one is to phonologically encode the selected item, that is, to retrieve its sound form, to create a phonological represen- tation for the item in its context, and to prepare its articulatory program. An extensive review of the literature on lexicalization can be found in Levelt (1989). This article addresses only one aspect of lexicalization, namely its time course. In particular, we examine whether the selection of an item and its phonologi- cal encoding can be considered to occur in two successive, non- overlapping stages. We acknowledge the invaluable contributions of John Nagengast and Johan Weustink, who programmed the computer-based experi- ments; ofGer Desserjer and Hans Fransen, who ran the experiments and assisted in data analysis; and of lnge Tarim, who provided graphi- cal assistance. We also acknowledge Gary Dell's and Picnic Zwitser- lood's detailed comments on an earlier version of this article, as well as the thorough comments of an anonymous reviewer. Herbert Schriefers is now at Freie Universit~it Berlin, Berlin, Federal Republic of Germany, and Thomas Pechmann is now at Universit~it des Saarlandes, Saarbriicken, Federal Republic of Germany. Correspondence concerning this article should be addressed to Wil- lem J. M. Levelt, Max-Planck-lnstitut for Psycholinguistik, Wundtlaan 1, NL-6525 XD Nijmegen, The Netherlands. 122 This is by no means a novel concept. One should rather say that it is the received view in the psycholinguistic literature (see especially Butterworth, 1980, 1989; Fromkin, 1971; Garrett, 1975, 1976, 1980; Kempen, 1977, 1978; Kempen & Huijbers, 1983; Levelt, 1983, 1989; Levelt & Maassen, 1981; Morton, 1969; Schriefers, Meyer, & Levelt, 1990). The first stage, lexical selection, makes available a semantically specified lexical item with its syntactic constraints. Kempen (1977, 1978) called this a lemma. Lemmas figure in grammatical encoding, specifically in the creation of syntactic frames. During the second stage, phonological encoding, phonological information is retrieved for each lemma. These phonological codes are used to create the articulatory plan for the utterance as a whole. Both Garrett (1976) and Kempen (1978), following Fry (1969), have stressed that the grammatical encoding and phonological encoding of an utterance normally run in parallel. Grammatical encoding, of which lexical selection is a proper part, is just slightly ahead of phonological encoding. The phonological encoding of a given item overlaps in time with the selection of a subsequent item. Only at the level of individual lexical items can one speak of successive stages. An item's semantic-syntactic makeup is accessed and used before its phonological makeup becomes available. Garrett (1975, 1976) argued for this separation of stages on the basis of speech error data. He distinguished between two classes of errors, word exchanges and sound exchanges, and could show that these classes differ in distributional properties. Word exchanges occur between phrases and involve words of the same syntactic category (as in this spring has a seat in it). Sound exchanges typically involve different category words in the same phrase (as in heft lemisphere). Word exchanges are