Elsevier

Brain and Language

Volume 107, Issue 2, November 2008, Pages 102-113
Brain and Language

The influence of syllable onset complexity and syllable frequency on speech motor control

https://doi.org/10.1016/j.bandl.2008.01.008Get rights and content

Abstract

Functional imaging studies have delineated a “minimal network for overt speech production”, encompassing mesiofrontal structures (supplementary motor area, anterior cingulate gyrus), bilateral pre- and postcentral convolutions, extending rostrally into posterior parts of the inferior frontal gyrus (IFG) of the language-dominant hemisphere, left anterior insula as well as bilateral components of the basal ganglia, the cerebellum, and the thalamus. In order to further elucidate the specific contribution of these cerebral regions to speech motor planning, subjects were asked to read aloud visually presented bisyllabic pseudowords during functional magnetic resonance imaging (fMRI). The test stimuli systematically varied in onset complexity (CCV versus CV) and frequency of occurrence (high-frequency, HF versus low-frequency, LF) of the initial syllable. A cognitive subtraction approach revealed a significant main effect of syllable onset complexity (CCV versus CV) at the level of left posterior IFG, left anterior insula, and both cerebellar hemispheres. Conceivably, these areas closely cooperate in the sequencing of subsyllabic aspects of the sound structure of verbal utterances. A significant main effect of syllable frequency (LF versus HF), by contrast, did not emerge. However, calculation of the time series of hemodynamic activation within the various cerebral structures engaged in speech motor control revealed this factor to enhance functional connectivity between Broca’s area and ipsilateral anterior insula.

Introduction

As a rule, recent models of spoken language production make a distinction between higher-order (“planning”) and lower-level (“execution”) aspects of speech motor control. For example, Levelt and coworkers (1999) postulated a computational stage, i.e., “phonetic encoding”, which transforms more abstract (“phonological”) word forms into motor programs, whereas a subsequent “articulator” translates these routines into the neural signals, “steering” the movements of the vocal tract. A similar hierarchy of control systems, usually, is assumed in the clinical literature, allowing for a separation of disorders of speech motor planning, i.e., apraxia of speech (AOS), and dysfunctions of speech motor execution, i.e., the dysarthrias (e.g., Duffy, 2005).

As a salient neuroanatomical characteristic, AOS represents a syndrome of the language-dominant hemisphere, almost exclusively bound to left-sided infarctions within the area of blood supply of the middle cerebral artery (Ackermann & Ziegler, in press). By contrast, unilateral supratentorial lesions rarely—if at all—give rise to severe and persisting dysarthria. Hence, unlike more basic motor execution processes, which are assumed to be organized in a bilateral fashion, the higher-order planning aspects of speech motor control show the same cerebral lateralization effects as the “core psycholinguistic language functions” (e.g., Caplan, 1987). The intra-hemispheric lesion site of AOS, however, is still a matter of dispute (for a review, see Ziegler, 2008). Based upon clinico-neuroanatomical correlation studies, the major hypotheses either focus upon posterior parts of the left inferior frontal gyrus (IFG; Broca, 1861, Hillis et al., 2004), ipsilateral anterior insular cortex (Dronkers, 1996), or the “face/mouth area” of the motor strip of the language-dominant hemisphere, including adjacent white matter structures as well as the anterior limb of the internal capsule (Fox et al., 2001, Schiff et al., 1983, Tanji et al., 2001).

Increasingly, functional imaging techniques now are applied within the domain of the speech sciences (Ackermann, Riecker, & Wildgruber, 2004b). The first systematic account of the cerebral circuitry bound to motor aspects of language production emerged as the by-product of a positron emission tomography (PET) investigation of lexical aspects of single-word processing (Petersen, Fox, Posner, Mintun, & Raichle, 1989). Subtraction of the hemodynamic responses to passive auditory/visual application of common nouns from the BOLD effects associated with overt repetition of such items was assumed to isolate the cerebral structures related to motor aspects of speech production. Besides supplementary motor area (SMA), sensorimotor cortex, and anterior–superior components of the cerebellum, an activation spot “buried” in the depth of the lateral sulcus could be detected. By contrast, Broca‘s area did not show any significant reactions. In line with several previous sporadic observations (see Ackermann et al., 2004b, for a review), a subsequent PET study confirmed these findings of left-hemisphere intrasylvian hemodynamic activity related to motor aspects of speech production and, more specifically, was able to assign this response to the anterior insula (Wise, Greene, Büchel, & Scott, 1999). Rostral intrasylvian cortex was assumed to represent the neural substrate of the planning stage of speech motor control, since a preceding clinico-neuroradiological correlation study had found cerebral lesions in AOS patients to be centered around the anterior insula (Dronkers, 1996).

As an alternative approach, a recent study of our group used the technique of functional connectivity analysis to further delineate the cerebral correlates of higher-order and lower-level aspects of speech motor control (Riecker et al., 2005). The exploration of the temporal dynamics of the cerebral hemodynamic activation patterns during syllable repetitions revealed the cortical and subcortical brain regions engaged in this task to be organized into two networks: high correlations emerged between (i) left SMA, left anterior insula, left dorsolateral frontal cortex, including Broca‘s area, and superior parts of the cerebellum, on the one hand, and (ii) sensorimotor cortex, thalamus, putamen/pallidum, left caudatum and inferior cerebellum, on the other. In consideration of the differential time course of the BOLD responses within these regions, the two networks were—tentatively—assigned to (i) the preparation/initiation and (ii) the execution of speech movements, respectively.

A straightforward functional imaging approach towards a separation of planning and executive components of the brain network of speech motor control should—most obviously—be based, first, on a computational description of the processing stages involved and, second, on experimental paradigms, derived from these theoretical suggestions. The most elaborated contemporary model of language production, put forward by Levelt and coworkers (1999), assumes phonetic planning to be centered around the retrieval and assembly of syllable-sized motor programs. Within this conceptual framework, the computational load of higher-order aspects of speech motor control must be expected to vary, in the first instance, with syllable frequency: Whereas the encoding of high-frequency (HF) syllables simply requires access to a “mental syllabary”, i.e., a store of highly automatized, pre-compiled, holistic motor routines representing entire syllable structures, items of a low frequency (LF) of occurrence must be assembled from smaller bits, e.g., from phonemes. As a consequence, the sub-syllabic route requires enhanced effort and, thus, poses higher temporal demands upon phonetic planning (Cholin, Levelt, & Schiller, 2006). In line with these suggestions, AOS subjects produce more errors in association with LF as compared to HF syllables (Aichert and Ziegler, 2004, Staiger and Ziegler, in press).

On the basis of Levelt’s theory of language production, experimental variation of syllable frequency should represent the most adequate probe of phonetic encoding operations. However, apart from a small number of studies of visual word processing (Carreiras, Mechelli, & Price, 2006), the available functional imaging studies of speech motor control have not strictly controlled for this factor. A recent fMRI investigation by Shuster and Lemieux (2005), for instance, assumed the computational load of phonetic encoding mechanisms to increase in parallel with word length. They found the contrast between four- and monosyllabic nouns not to be associated with any significant hemodynamic effects at the level of the left anterior insula and concluded that motor planning processes during language production do not engage intrasylvian cortex. However, the test materials used in this study were controlled for word but not for syllable frequency. It cannot be ruled out, therefore, that the monosyllabic items predominantly consisted of LF syllables while their longer cognates were composed of HF units. As a consequence, the expected impact of word length upon phonetic encoding processes might have been masked, and the observed distributional pattern of hemodynamic responses reflect unspecific effects of the longer test materials, e.g., increased articulatory effort.

Another recent model-driven study varied the two factors within-syllable complexity (simple versus complex syllable onset) and between-syllable complexity (re-iteration of the same token versus switching between three different items) in an orthogonal fashion to further elucidate the brain network bound to the preparation and the overt production of trisyllabic nonwords (Bohland & Guenther, 2006). An increase in sequential complexity of the test materials gave rise, among others, to enhanced hemodynamic activation of the left inferior frontal sulcus and left posterior parietal areas as well as bilateral responses of anterior insular and ventral motor/premotor areas, concomitant with a strong lateralization effect towards the language-dominant hemisphere. By contrast, sub-syllabic complexity, the second factor considered, yielded predominantly bilateral main effects, especially at the level of mesiofrontal and rostral intrasylvian cortex, the frontal operculum, and the cerebellum. However, this study also failed to systematically control for syllable frequency. As concerns the dimension within-syllable complexity, items with an initial consonant cluster are often less frequent than simple CV units. Variation of this factor, thus, might be confounded by syllable frequency. The concept of a mental syllabary, furthermore, assumes that onset complexity does not matter at the level of pre-compiled, holistic motor routines. Within the framework of that model, thus, complexity of syllable onset per se does not affect phonetic encoding processes. The theoretical status of the second dimension of the two-factorial experimental design considered by Bohland and Guenther (2006), i.e., complexity of syllable sequence, also raises questions. Levelt’s model of speech production does not account for an impact of this dimension upon phonetic encoding, since the process of compiling two or more successive syllables is assigned to the stage of motor implementation rather than motor programming (Levelt, Roelofs, & Meyer, 1999). Indeed, there is some evidence that the concatenation of two different syllables specifically challenges AOS patients (Deger & Ziegler, 2002), but a specific role of this operation in phonetic encoding has not been proved for normal speakers. Furthermore, repeated access to the same syllable three times in a row need not necessarily be less demanding than the encoding of three different successive syllables, since activation of the same unit might be hampered by refractory processes. In conclusion, the available functional imaging studies do not allow for a distinct separation of planning from execution processes within the domain of speech motor control and, hence, do not provide an unambiguous model-based account of the cerebral correlates of phonetic encoding in speech production.

In order to accomplish a more precise differentiation of the stage of phonetic planning from motor execution processes, as conceptualized in current psycholinguistic models, the test materials of the present fMRI study of overt speech production were systematically controlled for syllable frequency. As a second factor, complexity of syllable onset was varied in an orthogonal fashion to the former dimension. Onset complexity must be expected to have—within the framework of Levelt’s model—an impact upon phonetic encoding, since LF syllables have to be assembled from smaller constituents (Levelt et al., 1999). Hence, this psycholinguistic theory would predict that the neural network engaged in speech motor planning is sensitive to changes in syllable frequency and, within the domain of LF items, to changes in onset structure. A role of subsyllabic complexity, over and above the influence of syllable frequency, can also be expected on the basis of clinical data demonstrating that patients with phonetic encoding impairments are sensitive to the complexity of syllable constituents (e.g., Staiger & Ziegler, in press).

As a test of these predictions, bisyllabic nonwords, consistent with the phonotactic rules of German, were visually presented during fMRI measurements, and subjects were asked to read them as fast as possible. The test items were characterized by a trochaic pattern, i.e., the unmarked accent pattern of German, with the target units in the first (stressed) and a constant syllable in the second (unstressed) position. Besides a cognitive subtraction approach, functional connectivity analysis was applied to the obtained data set (Riecker et al., 2005), in order to identify brain structures characterized by a similar time course of hemodynamic activation.

Section snippets

Subjects

The present study recruited nine subjects (native speakers of German; age: 20–37 years, mean age = 28.4 years; 5 women), none of them reported a history of neurologic, psychiatric, or medical diseases. At clinical examination, hearing and visual capacities were found uncompromised. Informed consent had been obtained in line with the Declaration of Helsinki and the Institutional Review Board of the University of Ulm. All participants were right-handed according to the Edinburgh handedness scale

Behavioral data

The produced pseudowords of each subject were recorded during the fMRI experiment. After completion of the measurements, subjects did not report to have experienced any extra effort in association with the reading task. Furthermore, the performed analysis demonstrated an error rate of pseudoword reading of 2.83 ± 0.45% (mean/standard deviation).

Hemodynamic main effects

In order to select the VOI required for functional connectivity analysis, the hemodynamic main effects across all syllable categories and subjects were

Cerebral correlates of speech motor control

In line with preceding fMRI studies of our group, based on a syllable repetition paradigm (Riecker et al., 2005, Riecker et al., 2006), the observed pattern of hemodynamic main effects during the production of bisyllabic nonsense words across all stimulus categories encompasses the cerebral structures assumed to support, as inferred from clinical data, the generation of phonetic plans (fronto-opercular and intrasylvian cortex) as well as the execution of the respective vocal tract movement

Acknowledgment

This study was supported by the Deutsche Forschungsgemeinschaft (DFG AC 55/6-1). We thank Ingo Hertrich for excellent technical assistance.

References (51)

  • C.J. Price et al.

    Reading and reading disturbance

    Current Opinion in Neurobiology

    (2005)
  • A. Riecker et al.

    Articulatory/phonetic sequencing at the level of the anterior perisylvian cortex: A functional magnetic resonance imaging (fMRI) study

    Brain and Language

    (2000)
  • A. Riecker et al.

    The cerebral control of speech tempo: Opposite relationship between speaking rate and BOLD signal changes at striatal and cerebellar structures

    NeuroImage

    (2006)
  • A. Santi et al.

    Working memory and syntax interact in Broca’s area

    NeuroImage

    (2007)
  • L.I. Shuster et al.

    An fMRI investigation of covertly and overtly produced mono- and multisyllabic words

    Brain and Language

    (2005)
  • P. Sörös et al.

    Clustered functional MRI of overt speech production

    NeuroImage

    (2006)
  • K. Tanji et al.

    Pure anarthria with predominantly sequencing errors in phoneme articulation: A case report

    Cortex

    (2001)
  • R.J.S. Wise et al.

    Brain regions involved in articulation

    Lancet

    (1999)
  • W. Ziegler

    Apraxia of speech

  • W. Ziegler et al.

    The role of the left mesial frontal cortex in fluent speech: Evidence from a case of left supplementary motor area hemorrhage

    Neuropsychologia

    (1997)
  • H. Ackermann et al.

    Temporal organization of “internal speech” as a basis for cerebellar modulation of cognitive functions

    Behavioral and Cognitive Neuroscience Reviews

    (2004)
  • H. Ackermann et al.

    The contribution of the cerebellum to speech production and speech perception: Clinical and functional imaging data

    Cerebellum

    (2007)
  • H. Ackermann et al.

    Functional brain imaging of motor aspects of speech production

  • H. Ackermann et al.

    Akinetic mutism: A review of the literature

    Fortschritte der Neurologie und Psychiatrie

    (1995)
  • Ackermann, H., & Ziegler, W. (in press). Brain mechanisms underlying speech motor control. In W. J. Hardcastle & J....
  • Cited by (90)

    • The prevalence of apraxia of speech in chronic aphasia after stroke: A bayesian hierarchical analysis

      2022, Cortex
      Citation Excerpt :

      AOS is a condition that has always played an outstanding role in discussions about the neural organisation of language and speech, and it poses specific diagnostic and therapeutic challenges to clinicians treating persons with aphasia (PWA). As a sign of motor dysfunctions accompanying the linguistic impairment that underlies aphasia, the presence of speech apraxic symptoms provides an indication that the brain lesion must involve parts of the articulation planning network of the left hemisphere (Bohland & Guenther, 2006, pii; Papoutsi et al., 2009; Riecker et al., 2008, pii). Clinically, it suggests that therapeutic interventions based on motor rehabilitation principles need to be considered, either as a therapeutic focus or as an addendum to standard aphasia therapy (Ballard et al., 2015; McNeil et al., 2016).

    • Supplementary motor area aphasia revisited

      2020, Journal of Neurolinguistics
    • fMRI and acoustic analyses reveal neural correlates of gestural complexity and articulatory effort within bilateral inferior frontal gyrus during speech production

      2019, Neuropsychologia
      Citation Excerpt :

      Second, stimulus frequency was not controlled across conditions. This could be problematic because high-frequency syllables may be processed differently than low-frequency syllables (but see Brendel et al., 2011; Riecker et al., 2008 for counter evidence). For example, the Levelt model of speech production assumes that the most frequent syllables of a language are stored in a “mental syllabary” that mediates their translation into pre-compiled gestural scores for the articulators, while the motor programs of less frequent syllables are thought to be compiled from smaller sub-syllabic units (Levelt et al., 1999).

    View all citing articles on Scopus
    View full text