A Theory of Stuttering
A Theory of Stuttering
In normal fluent speakers, the cortical activity related to speech processing is usually lateralized to the left hemisphere in almost all right-handers, but also in many left-handers. Therefore, the left brain hemisphere is referred to as the language-dominant hemisphere. Stuttering is typically associated with a lack of lateralization to the left hemisphere, i.e., activation is more symmetrical, which was suspected to be causal for the disorder by some researchers already in the twenties of the last century (see, e.g., Travis, 1978). However, in a recent MEG study with stuttering and non-stuttering preschool children, Sowman et al. (2014) did not find differences in the lateralization of brain activity between the groups in a picture-naming task – which does not support the hypothesis of a lack of lateralization being the cause of stuttering. In the framework of the theory presented in Chapter 2, the lack of lateralization in persistent stuttering can be explained in a new way, namely as a consequence of insufficient attention to the auditory channel during speech.
The idea the lateralization of speech processing may be related to attention is not new: Kinsbourne (1970) proposed an attention model of lateralization, assuming that “the simple act of anticipation of verbal stimuli may preferentially activate the left hemisphere” (cf. Lynn, 2010, p. 15). The attention model has been supported by several empirical studies not only of brain activation (see below), but also of cochlear activity as a function of attention: Markevych et al. (2011) found shared variance between central (cortical) and peripheral (cochlear) lateralization of the processing of consonant-vowel syllables, which was influenced by attention.
As already mentioned in Section 2.3, it was found in several fMRI studies that attention to the auditory channel strongly influenced the activation in auditory cortical areas. Sabri et al. (2008) found that some speech-related areas, particularly on the left hemisphere showed enhanced activation when verbal stimuli were presented – but only if the participant’s attention was directed to the auditory channel. Poeppel et al. (1996) found M100 responses (component of the event-related brain potentials ca. 100ms after stimulus onset) to be symmetric over both hemispheres when syllables were passively heard, but the response in the left temporal cortex increased and in the right temporal cortex decreased when the participants directed their attention to the syllables. The authors assumed that these lateralized cortical responses dependent on attention to verbal stimuli were concordant with hemispheric language dominance. In an fMRI study, Rämä et al. (2012) found greater activations in the right superior temporal gyrus when right-handed participants were distracted from aurally presented verbal stimuli, compared to a condition in which the participants directed their attention to the stimuli.
The lateralization of speech processing can even be influenced by whether attention is directed to the lexical or to the prosodic aspect of aurally presented sentences, Vingerhoets, Berckmoes, and Stroobant (2003) found activation more lateralized to the left hemisphere when participants paid attention to the lexical features of a sentence, but more symmetric with attention to prosody (however, the lateralization effect resulted from an increase on the right hemisphere in the prosodic condition). Similar results were obtained by Hugdahl et al.(2003): Attention to vowels or words was associated with a leftward asymmetry in brain activation, passive listening was not.
All these findings show the same relationship: Active listening to verbal stimuli is associated with greater activation in the language-dominant hemisphere – passive hearing or ignoring verbal stimuli is associated with a more symmetric activation bilaterally. However, all the experiments mentioned above were, first, conducted with normal fluent participants, and, second, the verbal stimulus was not the auditory feedback of one’s own speech.
A study with stutterers was conducted by Sato et al. (2011): Children and adults who stuttered as well as normal fluent controls listened to aurally presented verbal stimuli that contained either a phonemic or a prosodic contrast. In the controls, they found a clear left-hemispheric advantage for the phonemic contrast compared to the prosodic one. Children and adults who stuttered, by contrast, showed no significant difference between the two stimulus conditions – not a single subject who stuttered showed a left advantage in the phonemic contrast.
But does this result necessarily mean that the stutterers’ brains worked wrong? In the mentioned study, participants were neither asked to pay attention to the stimuli nor to attend particular features of the stimuli. Hence, the lack of left advantage in the phonemic contrast may result from a different habit of allocating attention. Possibly, stutterers tend to a more passive hearing, or they tend to pay attention more to the prosodic features of speech, or their attention is more attracted by the prosodic features even if a phonemic contrast is presented.
The study last mentioned was, indeed, conducted with stutterers, but the stimulus was not the auditory feedback of their own speech. However, we know from several brain imaging studies, that auditory association areas of the cortex are activated during speech in normal fluent individuals, but not or much weaker in stutterers (see Table 1 and the citations in Section 2.3). Therefore, I assume that, in normal fluent individuals, attention is allocated in a way that one’s own speech is perceived and processed properly, mainly in the speech dominant hemisphere. In stutterers, by contrast, auditory feedback of speech seems to be processed poorly, which is reflected in a lack of activation and, by this, in a lack of lateralization mainly on the temporal cortex.
However, a lack of lateralization has been observed not only on the temporal cortex, but also on the frontal cortex. The main cause might be that speech is not produced as automatically in stutterers as in normal fluent speakers. Stuttering is mainly experienced as a disorder of phonation or articulation, which probably are very automatic parts of speech production (read more). As a result of persistent stuttering, the affected person mostly loses the natural confidence that phonation and articulation run well automatically. Consequently, the person may try to control phonation or articulation by the will in order to avoid stuttering, or may substitute words with feared initial sounds, or may reformulate sentences internally to adjourn such words more to the end. All these maneuvers might be associated with activations on the right brain hemisphere, as was particularly found in the right SMA–basal ganglia–motor circuit and in the right frontal operculum (see, e.g., the meta analysis by Budde, Baron, and Fox, 2014)
The right frontal operculum (RFO) is the homologous cortex region of Broca’s area. In normal fluent speakers, RFO seems to be increasingly involved in linguistic processing when difficulty arises, i.e., when usual routines are not sufficient (Mitchell & Crow, 2005; Taylor & Regard, 2003); thus it is not surprising that this brain region was found to be overactivated during speech in stutterers. Particularly the search for synonyms and the internal reformulation of sentences may engage the right frontal language area (however, the right inferior frontal gyrus was found to be overactivated in stutterers also when verbal stimuli were only percecived – read more). The supplementary motor area (SMA) is mainly involved in the control of self-initiated, volitional movements. Generally and roughly said, a right shift in the frontal part of the brain might indicate a more volitional and creative behavior – also crisis handling is a kind of creative behavior – in contrast to the mainly automatic behavior that everyday speaking should be.
In sum, I think that the lack of lateralization to the language-dominant hemisphere in stuttering is caused by a misallocation of attention during speech: The lack of activation and, by that, of lateralization in the temporal cortex is caused by insufficient attention to auditory feedback, especially to its phonological and lexical aspect, and the lack of lateralization in the frontal cortex is caused by excessive attention to wording and/or by conscious, volitional control of phonation or articulation and, in general, by the fact that the person, more or less, is in a ‘crisis mode’ during stuttered speech.
Figure 11: Misallocation of attention and the lack of lateralization in stuttering. Note, that some secondary behaviors take much attention and, by that, reinforce the misallocation of attention during speech (see Fig. ,9), but I have omitted the revertive arrow for the sake of clarity.
It has been hypothesized by some researchers that the right-shift of brain activation observed in stutterers could be a compensation for structural and/or functional deficits in the left hemisphere – a not satisfactory compensation since stuttering still remains (e.g., Neumann et al., 2003) (read more). On this issue, the results of Chang et al. (2011) are interesting. They investigated structural and functional connectivity between several speech-related brain areas in adult stutterers and normal fluent controls: Between BA44 (part of Broca’s area) and pre-motor/motor areas, structural connectivity (i.e., fiber density) on the left hemisphere was lower in stutterers compared to controls, but this deficit was compensated for on the right hemisphere, where fiber density was increased in the stuttering group. Between BA44 and posterior superior temporal gyrus (pSTG), structural connectivity on the left hemisphere was also much lower in the stutterer group than in the controls (see Fig. 4 in Chang et al., 2011), but this deficit, probably concerning the integration of auditory information in speech control, was not compensated for on the right hemisphere.
Additionally to structural connectivity, also functional connectivity (i.e., temporarily correlated activity in two brain regions at resting state) between speech-related areas was examined in this study, and here, no deficit was found between BA44 and STG in the stutterers, as a group. If we interpret functional connectivity as the viability (standby, readiness to work) of the fiber connection between the areas of speech perception and speech control, and structural connectivity as an effect of utilization, i.e., of the frequency of activation (read more), then we can conclude: The connection between speech perception and speech control on the left brain hemisphere is quite able to work, but is less employed in the speech of stutterers – in contrast to normal fluent speakers, who regularly activate this connection during speech. I think, it is just as with the secondary auditory areas on the left hemisphere: they are activated during speech in normal fluent speakers, but not in stutterers (see Table 1). But the cause, in both cases, is not that those brain structures are damaged or maldeveloped, but that attention is allocated inappropriately with too little capacity remaining for the processing of auditory feedback.
A further empirical finding is noteworthy: In several studies, the structural integrity (fiber density) of white matter tracts in stutterers (as a group) was compared with that in normal fluent controls. Deficits in stutterers were found, among others, in fibers of the corpus callosum (Chang et al., 2015; Civier et al., 2015; Connally et al., 2014; Cykowski et al., 2010). Corpus callosum fibers connect homologous areas of both cortical hemispheres and regulate the division of labor between them with an area of the one side inhibiting the activity of the homologous area of the other side. Therefore, it is plausible to assume that corpus callosum fibers are crucial for the lateralization of speech processing, and one may hypothesize that the callosal fiber deficits found in stutterers are causal for the lack of language lateralization, or even for stuttering.
However, Szaflarski et al. (2006), examining 170 healthy right-handed children and adults aged between 5 and 67 years of age, found that language lateralization to the dominant hemisphere tendentially increases between the age of 5 and 20 years. This result suggests that language lateralization is not innate – just as language itself is not innate – but develops in the course of language acquisition, probably with the automatization of speech-language abilities. Further, Chang et al. (2015) compared the structural integrity of white matter in young children who stutter aged between 3 and 10 years, and controls. They, indeed, found structural deficits in several regions of the corpus callosum in the stutterer group, but the group differences increased with age, i.e., with the duration of stuttering, and there was hardly a group difference between the youngest children (see Chang et al., 2015, Fig. 2, lower row).
Putting the results of both the last mentioned studies together, they do not support the hypothesis that the structural deficits in fibers of the corpus callosum in stutterers are causal for the lack of language lateralization, or even for stuttering itself. I rather assume that both the structural deficits in callosal fibers and the lack of language lateralization are a result or concomitant of persistent stuttering. Indeed, I agree with Civier et al. (2015) who point to a disadvantageous recruitment of the right frontal cortex in speech production, but I do not believe that a reduction in interhemispheric inhibition (due to the fiber deficits?) is the cause.
Conversely, fibers might be weaker because they have less frequently been activated over time. As mentioned above, fiber maturation seems to depend on the frequency of activation. This activation, i.e., interhemispheric inhibition, is an unconscious, automatic process, that is, in turn, inhibited by the recruitment of the right motor and premotor cortex, that is, by the stutterer’s attempt to control articulation by the will (see above). The problem is that children, because of the stuttering, lose their original confidence in the automatic control of speech – the confidence that muscles automatically do what they should do in speaking. Hence, they mostly are in a mode of ‘crisis management’ when speaking, with the consequence that automatic speech control including the inhibition of the non-dominant speech/language areas cannot become well established and stabilized.
Speaking is often regarded as a conscious and voluntary behavior (a behavior controlled by the will), and consequently, a speech disorder is regarded as a disorder of a conscious and voluntary behavior. But what do we actually do consciously and voluntarily in spontaneous speech? We decide whether talking or not, and we choose the important content words that convey the message of a sentence (sometimes, however, one word leads to another in disputes). Function words are mostly inserted automatically. Likewise, words are strung together automatically in the correct syntactic order, and grammatical endings are attached. All these functions are not impaired in stuttering – neither the conscious and voluntary nor the ‘higher’ automatic functions. Only the very basic and probably most automatic functions, namely phonation and articulation, are impaired.
I think, it is important to understand that stuttering, in its core, is a disorder of a very automatic function. Speaking as a voluntary behavior, that is, the decision whether talking or not, and the conscious choice of words is impaired only by secondary symptoms of stuttering like fear and avoidance behavior.
(return)
In my view, saying the brain itself would compensate for deficits is problematic. There is, indeed, a plasticity that allows the brain to accommodate to changed requirements, but there is no supervisor in the brain who can assess if a part of the brain does not work well, and who can engage another part to take over the task. It is, instead, the person, who becomes aware of that something, e.g., speaking goes wrong. And it is the person, who tries to cope with the problem anyway, i.e., who changes his or her behavior, who controls articulation consciously, who plans utterances carefully before speaking, etc.
Naturally, changes in behavior entail changes in brain activation, and if they last over time, changes in brain activation result in structural changes: increase or decrease of gray or white matter volume or fiber density – effects demonstrated in some studies in recent years (Bengtsson et al., 2005; Draganski et al, 2004; Driemeyer et al., 2008; Keller & Just, 2009; Scholz et al., 2009).
That is important for the therapy of stuttering: Almost all therapies aim to change the client’s behavior in any way. If the new behavior is learned well and is frequently practiced in everyday talking, and by that, becomes automatic with time, it will not only result in a changed brain activity pattern, as was shown by Neumann et al. (2003) and Ingham et al. (2003), but also in structural changes in brain. I’m optimistic that research will provide evidence of such structural changes in the next years.
(return)
The structural integrity or density of nerve fiber tracts seems to be dependent on the frequency with which the fibers have been activated over time: Bengtsson et al. (2005) found a positive correlation between piano practicing since childhood and fiber tract structure in brain regions that are involved in piano playing. Two experiments showed that positive changes in white matter structure can be achieved by training over few weeks: Scholz et al. (2009) used juggling training with adults, Keller and Just (2008) applied reading exercises with 8-10-year-old poor readers – in both cases, training success was associated with enhanced structural integrity of fiber tracts that were involved in the respective task. Further, the biochemical relationship between activation and myelination was demonstrated in vitro on nerve fibers of mice.(Wake, Lee, & Fields, 2011; see also Fields, 2010; Petersen & Monk, 2015).
If frequent activation of fiber tracts over time results in advanced fiber maturation (myelination), then we can, conversely, assume that, if fibers have been rarely activated over years (which I assume to be the case in persistent stuttering for the fibers connecting left auditory cortex with left inferior frontal cortex), then their structural quality may be lower than normal. I will address this issue more extensively in Chapter 4.
(return)
Preibisch et al. (2003) found the RFO to be overactivated in stutterers not only during a reading task (that was completed fluently because of the masking effect of the noise in the MRI scanner), but also during a baseline task that consisted in passive viewing of meaningless signs. Further, in an auditory (speech and non-speech) perception task, Halag-Milo et al. (2016) found the right inferior frontal gyrus greater activated in adult stutters than in controls. Since also the left Heschl’s gyrus was overactivated in the stutterers. it seems as if they had to listen more intensively in order to detect the target sound in the presented speech / non-speech stimuli (the noise of the MRI scanner came in addition) – such tasks may be more difficult on average for stutterers because of slight anomalies in central auditory processing (see next Section). The results suggest that right frontal overactivation may be more than a compensation for stuttering, but may – like the lack of lateralization in the temporal cortex – reflect an anomaly in attention regulation and/or central auditory processing.
(return)