Preliminary results of a functional MRI study of brain activation patterns in stuttering and nonstuttering speakers during a lexical access task

https://doi.org/10.1016/j.jfludis.2003.08.002Get rights and content

Abstract

An fMRI study examining lexical access and lexical generation in nine nonstuttering and seven stuttering speakers is presented. Lexical access was examined during a word description task that was presented auditorily while subjects “silently” thought of the target words. Participants alternated between four 30-s rest blocks and four 30-s “active” blocks. Activation patterns were assessed utilizing a standard subtraction paradigm, where the activation during the rest blocks was subtracted from the activation during the active blocks. High levels of variability characterized activation patterns within both speaker groups. Group comparisons using random effects statistical analyses did not identify significant differences between the groups when corrected for multiple comparisons. Analyses were subsequently conducted by comparing the trends in the group activation patterns between the speaker groups using fixed (corrected) and random effects (uncorrected) analyses. Nonstuttering control speakers activated primarily left hemisphere cortical speech and language areas while the stuttering speakers appeared to produce more bilateral activation. Discussion of these results focuses on the specific within- and between-hemispheric activation patterns and possible interpretations of these patterns.

Educational objectives: The reader will learn about: (1) issues related to interpreting brain activation findings in stuttering speakers; (2) the role and neurological substrates of lexical access during speech production in nonstuttering and stuttering speakers; (3) the basics of functional MRI; and (4) the brain activation areas involved during a silent lexical retrieval task in nonstuttering and stuttering speakers.

Introduction

Numerous studies, utilizing a variety of techniques, support the notion that cerebral laterality differences exist between stuttering and nonstuttering speakers. For instance, increased right hemisphere activation during linguistic tasks in stuttering speakers has been demonstrated in dichotic listening experiments (e.g., Cimorell-Strong, Gilbert, & Frick, 1983; Curry & Gregory, 1969), EEG experiments (e.g., Boberg, Yeudall, Schopflocher, & Bo-Lassen, 1983; Moore, 1986, Moore & Haynes, 1980), and in early regional cerebral blood flow (rCBF) studies (e.g., Pool, Devous, Freeman, Watson, & Finitzo, 1991; Wood, Stump, McKeehan, Sheldon, & Proctor, 1980). However, reports of basic left/right laterality differences provide little explanation about speech and language planning in stuttering speakers. More importantly, the particular neural sites of activation (and deactivation) that might be implicated in stuttering need to be identified.

The findings from a series of positron emission tomography (PET) studies conducted over the past decade have led to more detailed descriptions of the neurological substrates of stuttering (e.g., Braun et al., 1997; De Nil, Kroll, Kapur, & Houle, 2000; Fox et al., 1996; Ingham, Fox, Ingham, & Zamarripa, 2000; Wu et al., 1995). A notable finding is that brain blood flow differences between stuttering and nonstuttering speakers during rest have not been reported (e.g., Braun et al., 1997, Ingham et al., 1996). Differences between the speaker groups tend to be greater when linguistic tasks are required. In addition, there is some evidence to suggest that progressively more demanding linguistic tasks show increasingly differing brain activation patterns between stuttering and nonstuttering speakers (e.g., Blomgren, McCormick, & Gneiting, 2002; De Nil et al., 2000).

A long-standing question in interpreting the findings of overactivations in the right hemispheres of stuttering speakers has been whether these findings are revealing the core problem, or whether the findings are merely by-products of stuttering. Ludlow (1999) points out that increased activity in the right motor regions in stuttering speakers could be related to stuttered-speech (speech presumably produced with high muscular effort). This potential stuttered-speech confound (i.e., right activation as a by-product) has just begun to be addressed. In a PET study of imagined stuttering, Ingham et al. (2000) reported that a task where participants simply imagined reading with stuttering (no overt speech) produced much of the same aberrant cerebral activations and deactivations as reading aloud tasks (with natural/overt stuttering). Also, imagined fluent reading produced similar activations as an induced fluency task (choral reading). Specifically, the imagined fluent reading task produced activations closer to those of “normal” speakers. While much more research needs to be conducted in this area of imagined stuttering versus imagined fluency, this is clearly a revealing line of research. At a minimum, it appears that overt speech may not be necessary to distinguish abnormal brain activation patterns in stuttering speakers compared to nonstuttering speakers.

Recent behavioral studies of linguistic processing have revealed subtle differences in the processing time required by stuttering adults when they are asked to make phonological, semantic and syntactic judgments about experimental stimuli (e.g., Bosshardt, 1993, Bosshardt, 1994; Prins, Main, & Wampler, 1997). For instance Bosshardt suggests that adult stuttering speakers demonstrate relatively long reaction times on tasks involved in phonological and lexical monitoring. Stuttering speakers took longer than nonstuttering speakers on tasks requiring them to monitor for rhymes and category membership of target words. Further, Prins et al. reported the stuttering speakers in their study displayed longer latencies than their control speakers on picture naming tasks. Prins et al. theorized that “slow lexicalization” (lexical access) in stuttering speakers could be implicated as a source of stuttering.

In an electrophysiological study, Blomgren et al. (2002) compared 10 stuttering speakers and 12 nonstuttering s peakers on the amplitudes and latencies of auditory P300 event-related potentials. The stuttering speakers displayed longer P300 latencies for linguistic stimuli compared to the nonstuttering speakers. Differences between the speaker groups on non-linguistic stimuli (tones) were not significant. These results suggest that stuttering speakers might process linguistic stimuli at a slower rate than nonstuttering speakers. Given these findings of delayed linguistic processing in stuttering speakers, further research examining the neurological substrates of lexical processing and lexical access appears warranted.

The role and neurological substrates of lexical access has garnered much interest in recent speech production literature (e.g., Fiez, 2001, Indefrey & Levelt, 2000; Levelt, Roelofs, & Meyer, 1999). Levelt et al. (1999) outlined an important theory of lexical access in speech production. Levelt et al. proposed that generation of words is a dual process. There is one system whose purpose is to select words from the mental dictionary (the “lexicon”). There is also a system that nearly simultaneously prepares the articulatory gestures for the selected words. Levelt et al., in their theory of lexical access, argue that there is a “somewhat fragile link between these systems. Each of these systems is itself staged. Hence, the theory views speech as a feedforward, staged process, ranging from conceptual preparation to the initiation of articulation” (p. 36).

In Levelt’s staged theory of lexical production, each stage has its own characteristic output representation. These stages minimally include:

  • 1.

    Lexical conceptualization, which includes the process leading up to the formation of a lexical concept, called conceptual preparation.

  • 2.

    Lexical selection, which involves retrieving a word from the lexicon. In normal speech we usually retrieve two to three words per second, so this is a rapidly occurring process. In fact, Levelt et al. point out that “this high-speed process is surprisingly robust; errors of lexical selection occur in the one per one thousand range” (p. 4).

  • 3.

    Morphophonological encoding and syllabification, which refers to jumping the rather nebulous gap from the “conceptual/syntactic” domain to the “phonological/articulatory” domain. This stage relates to the preparation of the required articulatory gestures, including the prosodic context in which the word will be eventually spoken. According to Levelt et al. (1999), activation of this stage requires three types of information related to word access: “the words’ morphological makeup, its metrical shape, and its segmental makeup” (p. 5).

  • 4.

    Phonetic encoding is the next stage, which refers to the “specification of the articulatory task.” Phonetic encoding includes planning the specific pattern of articulatory movements. It should be noted, however, that while conceptually this process appears to have solid face validity, our understanding of the specific phonetic encoding process is still woefully inadequate. There are numerous theories and explanations that have attempted to detail the phonetic encoding process, but these theories tend to be highly abstract (i.e., Dell, 1986, Houghton, 1990, Roelofs, 1996, Shattuck-Hufnagel, 1979, Vousden & Brown, 2000, Wickelgren, 1969). We do not know, for instance, whether the individual phonetic segments (phonemes) are planned segment by segment, or syllabically. Levelt et al. (1999) suggest that we program words syllabically, by accessing our “syllabary” (our syllable dictionary), yet solid empirical evidence supporting this notion is still lacking. Other issues to which we still do not understand are whether the individual phonetic segments (or syllables) are planned on a “simple” serially ordered, muscular/articulatory, activation basis, or whether movements are planned by a multi tiered command structure that issues gestalt-like speech system commands (i.e., combined respiratory, laryngeal, and various supralaryngeal movement patterns).

  • 5.

    Articulation is the final common output of this complex staged theory. In this stage, the phonetically encoded plan is put into action. The plan becomes the actual movements of the lungs, larynx, and supralaryngeal articulators. Coverage of speech motor theories is well beyond the scope of the present paper. Suffice to say that the speech motor control system, which governs articulation, involves the management of a series of extremely complex movement events.

  • 6.

    Feedback is the final stage of this lexical access theory. While it can be argued that external auditory (and/or tactile) feedback is not an absolute requisite component in the lexical access process per se, it may well be that we monitor our “internal speech” for errors before we actually articulate our words. In this case, then, we are referring to an internal, pre-articulation, feedback loop. The notion is that we monitor the accuracy of our speech plan productions during, or immediately after, the phonetic encoding stage (Levelt, 1983). In fact, this “final” step may be better placed between the phonetic encoding and articulation stages.

Recent theories examining pre-motor breakdowns as being centrally important in explaining stuttering have received much attention. In the early 1990’s, Postma & Kolk, 1990, Postma & Kolk, 1992, Postma & Kolk, 1993) developed the covert repair hypothesis as an explanation for stuttering. The covert repair hypothesis is based on the notion that stuttering results from “covert, prearticulatory repairing of speech programming errors on the ongoing speech” (Postma & Kolk, 1993, p. 472). This theory specifically implied stuttering resulted from a deficit in phonetic encoding and poor internal monitoring of the phonetic encoding process. Observable moments of stuttering were essentially (belated) attempts to repair errors sent downstream to the motor system. This theory was an outgrowth of Levelt’s (1983) monitoring and self-repair hypothesis, but also appears influenced by Dell’s (1986) spreading activation theory of speech production and Wingate’s (1988) language production view of stuttering. Specifically, Wingate contended “stuttering is a defect in the language production system, a defect that extends beyond the level of motor execution” (1988, p. 240). It may be hypothesized, then, that the “problem” may exist at pre-speech levels as early as morphological encoding, lexical selection (access), and perhaps even during lexical conceptualization.

There is ample evidence to conclude that tasks assessing lexical knowledge through various word finding tasks are left-lateralized in the vast majority of normal speakers (e.g., Cuenod et al., 1995; Grabowski, Damasio, & Damasio, 1998; McCarthy, Blamire, Rothman, Gruetter, & Shulman, 1993; Schlösser et al., 1998; Wood, Saling, Abbott, & Jackson, 2001). Lexical knowledge can be assessed through the use of a wide variety of lexical access or word generation tasks. Indefrey and Levelt (2000) provide an excellent overview of some of these tasks, which include picture naming, verb generation, generating words from a beginning letter, generating words based on a given category (noun generation), word repetition, word reading, and pseudoword reading. All of these tasks serve to tap into the core processes of conceptual preparation and subsequent mapping of this preparation onto a lexical concept (i.e., lexical selection). The major difference between the various types of tasks is in their “lead-in” process. The problem is that these lead-in processes (and the differences between them) are not well understood (Indefrey, 1997). In picture naming, the lead-in process includes some type of visual object recognition, which provides the internal percept of the object, which in turn leads to conceptual preparation. Many variables may affect this visual lead-in process, such as visual complexity, perspective, color versus black and white, and so forth. Verb generation is another lexical access task in which a subject is presented with a noun and immediately responds with a related verb (e.g., subject hears “volcano” and responds with “erupt”). For verb generation, the lead-in process includes either visual or auditory word recognition (depending on the stimulus), some type of visual imagery, long-term memory access of associated actions, followed by the process of retrieving associated action words. In other lead-in processes, such as generating a series of words beginning with a specific letter (orthographic lexical access tasks), the lead-in process may be quite diverse. Especially in English, there is substantial variance between letters and specific phonemes. For example, only five written vowel letters are used to represent the approximately 21 spoken vowel/diphthong phonemes in English. Therefore, the lead-in process during this task may involve phonemic strategies, visual letter imagery, visual word imagery, and/or other semantic processes.

The key component of both orthographic and noun generation lexical access tasks is that individuals are tapping into a finite reserve of lexical items. As more category items are retrieved, the “lexical pool” dwindles and the task becomes increasingly more complex (and more dependent on attentional and memory abilities). By contrast, single word generation tasks (e.g., picture naming, verb generation, word reading, etc.) involve a single word being generated (or repeated) in response to a single stimulus. Single word generation tasks, therefore, involve presenting a series of stimuli (auditory or visual), with a brief response time between presentations. Recently, the utility of lexical access and/or word generation tasks has been exploited as an aid to determine hemispheric dominance for language. Increasingly, lexical access tasks are being used as a component of testing batteries designed to determine hemispheric dominance for a variety of tasks in clinical populations (e.g., Heilbrun, Lee, & Alvord, 2001; Spreer et al., 2002).

It would appear that assessing lexical access in stuttering and nonstuttering speakers would be a useful means to compare language laterality differences and within-hemisphere differences between these speaker groups. The fact that lexical access tasks tend to be highly left biased, combined with the fact that overt speech production is not required, make this task a uniquely appropriate one in which to examine pre-speech language processes in stuttering speakers.

Much of our present knowledge of language processing in normal speakers, as well as the empirical evidence for the covert repair hypothesis of stuttering, and indeed most of Levelt’s theory, is based on reaction time research. There are certain advantages to reaction time research. Relatively simple testing paradigms and reasonably straightforward equipment requirements are obvious advantages. However, Müller (1999) points out that reaction time research has at least two disadvantages: (1) It is an indirect method of studying language generation processes. Many reaction time paradigms involve pressing buttons of one kind or another, thus, the responses being measured are actually longer than the cognitive processes to which are inferred. (2) Behavioral reaction time paradigms assume a level of cognitive awareness of the task on the part of the subject. A problem is that many cognitive processes cannot be examined behaviorally. Utilization of neurophysiological techniques (e.g., electroencephalography [EEG] or magnetoencephalography [MEG]) and hemodynamic techniques (e.g., positron emission tomography and functional magnetic resonance imaging [fMRI]) may be better suited to assess cognitive–language processes. The evaluation of language processing, including lexical access, is just beginning to be explored with these neuroimaging techniques.

Of the currently available imaging techniques, fMRI has become the dominant technique for studying the functional organization of the human brain during cognitive, perceptual, sensory and motor tasks (Menon, 2001). Compared to EEG and MEG, fMRI provides excellent spatial resolution, and compared to PET, fMRI is much less intrusive, not requiring the injection of radioactive isotopes. The fMRI technique works by detecting small changes in cerebral concentrations of deoxyhemoglobin (found in red blood cells). The mechanisms by which deoxyhemoglobin affects the MRI signal is termed blood oxygenation level dependence (BOLD). The BOLD effect really measures a combination of changes in blood flow, blood volume, and blood oxygenation within the capillaries and venous system of the brain (Lee, Duong, Yang, Iadecola, & Kim, 2001). The level of the hemodynamic response of a brain area is generally thought to be proportional to its level of neuronal activity. In other words, increases in blood flow in a particular area are related to increases in neuronal activity in that area. It is through this tenet that most researchers infer their BOLD findings to be indicative of regional brain activity. Functional MRI studies have typically utilized “block designs” whereby a number of stimuli items of the same type are presented in a block, and brain activation averaged over the duration of the block. The blocks may last from 20 s to several minutes. Subsequent blocks may include rest or a different stimulus type. More recently, event-related fMRI has been utilized. In event-related designs, the stimulus items can be presented in a pseudo-randomized fashion. Each design paradigm has certain advantages and disadvantages. Event-related designs tend to have lower detection power than blocked designs, but have several other advantages over the blocked design (Buckner et al., 1996; Pilgrim, Fadili, Fletcher, & Tyler, 2002). Primarily, event-related designs are time locked to stimuli presentation; therefore, measured activation may be more closely related to the experimental task or presented stimuli. Event-related designs are also less prone to imaging confounds from habituation effects and other task strategy effects. Conversely, blocked designs tend to have higher detection power than event-related designs and are generally considered to be the most efficient design (Friston, Zarahn, Josephs, Henson, & Dale, 1999). This conventional type of design is useful when conducting preliminary studies where low statistical power may be an issue.

As outlined above, the process of lexical access is complex. This complexity is echoed in the relatively numerous brain areas that appear to be activated during various lexical access tasks. The most salient areas identified through neuroimaging studies appear to be the left middle frontal gyrus (Brodmann’s areas 9 and 46) and the left inferior frontal gyrus (Brodmann’s areas 44 and 45) (Frith, Friston, & Liddle, 1991; Wood et al., 2001). These frontal areas loosely correspond to the classically defined Broca’s area. Activation has also been reported in the anterior cingulate gyrus, left inferior parietal lobule, left supplementary motor area, pre-motor cortex, anterior insula, and a variety of temporal lobe areas (Baker, Frith, & Dolan, 1997; Cuenod et al., 1995, Paulesu et al., 1997, Wise et al., 1991, Yetkin et al., 1996).

The purpose of this paper is to present preliminary fMRI data relating to activation areas involved in lexical access in a group of stuttering speakers and a group of nonstuttering speakers. In order to evaluate the neurological substrates of lexical access in stuttering and nonstuttering speakers, a unique, auditorily cued, lexical access task was utilized. The task involved word-description decoding. Participants were presented with descriptions of words (e.g., “something cold we eat on a cone”) and were required to retrieve and silently generate the target word (i.e., “ice-cream”) based on the description. The problem with many previously used lexical access tasks (i.e., orthographic lexical retrieval tasks and other verbal fluency tasks) is that they are highly dependent on controlled processing. As the stock of easily accessible items dwindles, the tasks become more and more complex and memory skills start to play significantly greater roles. In these tasks, attention and memory may serve uncontrolled modulating roles. Problems also exist with verb generation tasks in that they are highly dependent on additional non-linguistic processes such as visual imagery. This reliance on non-linguistic processes could also serve to confound any subsequent findings. Likewise, word repetition tasks and reading tasks seem to add many extraneous factors. In word repetition, the lead-in process involves word recognition, but not generation (lexical access) per se. For instance, individuals may repeat words they do not understand and nonwords, so the core process may be triggered at the downstream level of phonological encoding (Indefrey & Levelt, 2000). Word reading relies on visual word recognition, which in turn relies on several complicated occipital and parietal processes. From visual recognition, is appears possible to jump directly to phonological encoding, without necessarily going through a lexical conceptualization phase or lexical selection phase.

The word description task utilized in the present study appears to address some of the problematic issues inherent in previously used lexical access tasks. While the word description task is not without its own complications (specifically, activation of the dual processes of language comprehension and language production), it does appear to lessen the potential confounds of overly relying on memory skills, activating the visual/grapheme system, overly activating visual imagery centers, or overly taxing attentional processes. In addition, the word description task provides a consistent level of difficulty over the duration of the task.

In summary, the present study was designed to utilize fMRI to assess the neuro-biological substrates of lexical access in stuttering and nonstuttering speakers. A word description task was used to stimulate the lexical access process. We were particularly interested in examining between- and within-hemisphere patterns of activations in Broca’s area, primary motor cortex, supplementary motor cortex, insula, and auditory association areas.

Section snippets

Participants

Sixteen right-handed adults participated in this study. Seven of the participants stuttered (5 males, 2 females) and nine were normally fluent speakers (7 males, 2 females). The stuttering speakers ranged in age from 19 to 38 years (M=27.5, S.D.=7.9) and the nonstuttering speakers ranged in age from 20 to 40 years (M=29.6, S.D.=7.4). The stuttering speakers were recruited from stuttering treatment programs offered at the University of Utah. All scanning of the stuttering speakers took place

Results

The following results represent preliminary findings associated with a larger ongoing study. First, descriptive within-group activation results are reported for the nonstuttering and stuttering speakers. Second, descriptive group comparison results are presented. Group averages were based on a standard subtraction paradigm; for each group, the average activation patterns during the rest epochs were subtracted from the active epochs. The remaining activation areas were assumed to be reflective

Discussion

This study was conducted to examine and compare brain activation patterns during a lexical access task in stuttering and nonstuttering speakers. Functional MRI was utilized to assess the brain activation areas during the task in these two speaker groups. Activation patterns exhibited by the two speaker groups contained many similarities, but differences were also observed. However, high levels of activation variability within both speaker groups hampered appropriate between group statistical

References (52)

  • W. Moore

    Hemispheric alpha asymmetries of stutterers and nonstutterers for the recall and recognition of words and connected reading passages: Some relationships to severity of stuttering

    Journal of Fluency Disorders

    (1986)
  • L. Pilgrim et al.

    Overcoming confounds of stimulus blocking: An event-related fMRI design of semantic processing

    NeuroImage

    (2002)
  • A. Postma et al.

    Speech errors, disfluencies, and self-repairs of stutterers in accuracy conditions

    Journal of Fluency Disorders

    (1990)
  • A. Roelofs

    Serial order in planning the production of successive morphemes of a word

    Journal of Memory and Language

    (1996)
  • J. Vousden et al.

    Serial control of phonology in speech production: A hierarchical model

    Cognitive Psychology

    (2000)
  • A. Wood et al.

    A neurocognitive account of frontal lobe involvement in orthographic lexical retrieval: An fMRI study

    NeuroImage

    (2001)
  • F. Wood et al.

    Patterns of regional cerebral blood flow during attempted reading aloud by stutterers both on and off haloperidol medication evidence for inadequate left frontal activation during stuttering

    Brain and Language

    (1980)
  • S. Baker et al.

    The interaction between mood and cognitive function studies with PET

    Psychological Medicine

    (1997)
  • M. Blomgren et al.

    P300 ERPs in stutterers and nonstutterers: Stimulus and treatment effects [abstract]

    The ASHA Leader

    (2002)
  • H. Bosshardt

    Differences between stutterers’ and nonstutterers’ short-term recall and recognition performance

    Journal of Speech and Hearing Research

    (1993)
  • A. Braun et al.

    Altered patterns of cerebral activity during speech and language production in developmental stuttering. An H215O positron emission tomography study

    Brain

    (1997)
  • R. Buckner et al.

    Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging

    Proceedings of the National Academy of Science USA

    (1996)
  • C. Cuenod et al.

    Functional MRI during word generation using conventional equipment: A potential tool for language localization in the clinical environment

    Neurology

    (1995)
  • F. Curry et al.

    The performance of stutterers on dichotic listening tasks thought to reflect cerebral dominance

    Journal of Speech and Hearing Research

    (1969)
  • L. De Nil et al.

    A positron emission tomography study of silent and oral single word reading in stuttering and nonstuttering adults

    Journal of Speech, Language, and Hearing Research

    (2000)
  • G. Dell

    Spreading-activation theory of retrieval in sentence production

    Psychological Review

    (1986)
  • Cited by (0)

    View full text