Thursday, September 27, 2018

The Main Interruption Rule: Is it obsolete?

The Main Interruption Rule (MIR) is an important component of my theory of stuttering. It describes a speaker’s behavior when he (or his/her brain) detects a speech error. MIR was proposed by Willem Levelt in his essay “Monitoring and self-repair in speech” (1983) and again in his book “Speaking. From intention to articulation” (Cambridge 1989). There, on page 478, he puts MIR as follows:

"Stop the flow of speech immediately upon detecting trouble."

The figure below shows why MIR is important in my theory of stuttering: I assume that the same interruptive mechanism is elicited when a real speech error is detected (A) and when an invalid error signal occurs because of insufficient processing of auditory feedback (B). In the latter case, the interruption results in stuttering when the speaker spontaneously tries to continue.
(By the way, the German version of this figure was inadvertently inserted in my English website for, I don’t know, probably some months. I’m sorry.)

A ‘real speech error’ in the above figure must be defined as a mispronounced or mistaken word, that is, as a phonological error or a classical ‘slip of the tongue’,like “a cuf of coffee”. Phonological errors are relatively rare in speaking in one’s native language. The invalid error signals assumed in my theory resemble phonological errors, i.e., they are invalid phonological error signals : The internal monitoring system ‘believes’ that a phoneme sequence was produced incorrectly or incompletely.

When I included MIR in my theory I did not check if there were any new empirical findings regarding the interruption of speech after error detection. Meanwhile I caught up on, and here are the results. I found two studies of self-monitoring and error repair during speech, both in the journal Cognition: Blackmer and Mitton (1991): “Theoriesof monitoring and the timing of repairs in spontaneous speech” and Seyfeddinipur,Kita, and Indefrey (2008): “Howspeakers interrupt themselves in managing problems in speaking:Evidence from self-repairs”

Blackmer and Mitton measured cut-off-to-repair intervals in spontaneous speech. They found many of these intervals to be very short (less than 100 ms, not rarely 0 ms), and they conclude that speakers often plan corrections while speaking and interrupt their speech not before this re-planning is finished, perhaps in order to avoid disfluent speech. Blackmer and Mitton deem such a behavior inconsistent with MIR which claims that speech, in a rule, is interrupted immediately after error detection.

Seyfeddinipur, Kita and Indefrey, however, reply: “The MIR can account for at least some zero ms cut-off-to-repair intervals if it is assumed that the interruption process itself takes time.” (838) During this reaction time, re-planning can already start and can, in simple cases, be finished at the moment of interruption. Therefore, MIR is quite consistent with short cut-off-to-repair intervals – especially if we not (as Levelt does) assume that every re-planning takes place in a separate “Formulator” module (see the discussion of Levelt’s model of speech production here in my website).

An example of a repair with a zero ms cut-off-to-repair interval was “willfiddily/ fully” (Blackmer & Mitton, p. 190). In this case, the speech motor program of ‘willfully’ was correctly selected and started, but then, for unknown reasons, articulation switched to another motor program (‘fiddly’). The repair is here not a reformulation, but rather a back-switch to the original motor program.

The basic idea in this consideration is: In spontaneous speech, formulation and articulation are one process that is guided by the intended message (which develops and changes in the course of speaking) and by routines (of syntax, grammar, word use). There is a complicated interaction between these factors, which accounts for the high number of corrections in spontaneous speech: Blackmer and Mitton found that speakers made repairs once every 4.8 s on average. ‘Repair’ is however broadly defined here and includes, e.g., reformulations in order to be more clear or more appropriate as well as word repetitions in order to get more time for speech planning. 

The problem: broad definition of the term 'error' 

In the Blackmer and Mitton study, only 3% of all errors were production-based errors, that is, mispronounced or mistaken words (the rest was plan-based, that is, consequence of uncertainty in conceptualization or formulation, e.g., editing terns and repetitions together represented 72.9% of all errors). In the context of my theory of stuttering, however, only production-based errors are relevant. As mentioned above, the invalid error signals assumed in my stuttering theory are invalid phonological errors. It is therefore only interesting for us how people (or their brains) respond to real phonological errors – and phonological errors are a subcategory of production-based errors. because a mispronounced word is a word produced, not an only planned one.

Seyfeddinipur, Kita, and Indefrey did not differentiate between production-based and other errors, but since they also used a broad definition of the term ‘repair’ (“a disfluency containing some indication of a speech suspension, e.g., a glottal stop, laryngalization, fillers, or silent pause greater than 200 ms, and a resumption in which there was a modification of the original delivery.” p. 839), the percentage of production-based errors might have been low in their material. It is hence not surprising that they came to the conclusion that “speakers interrupted themselves not at the moment they detected the problem but at the moment they were ready to produce the repair. Speakers preferred fluency over accuracy.” (837).

Astonishingly, they believe: “Speakers may interrupt themselves right away when they detect an expression with socially drastic consequences. By contrast, they may not interrupt themselves on detecting minor phonological errors such as a ‘cuf of coffee’.” (841). I don’t think so. When I feel I have said something inappropriate, I can decide whether or not I interrupt myself and make a correction. But in cases like “cuf of coffee” or in cases of grammatical errors I would automatically interrupt myself and make a repair. I think the worst case in speaking is the appearance someone does not master his native language.

Do I think so because I’m a stutterer? Vasic and Wijnen (2001; 2005) proposed a theory in which they assume a hyper-sensitivity of stutterers’ self-monitoring for normal speech disfluencies. I don’t share this view (and I think their data don’t really support their theory; read more here in my website).But I do believe that stutterers’ self-monitoring system is at least normally if not heightened sensitive to phonological errors (such as “cuf of coffee”) or phonological incompleteness. Someone who doesn’t detect such errors or who doesn’t respond to them with an interruption will, after my theory, hardly stutter because his/her brain will also hardly respond to invalid signals appearing like phonological errors.

Back to MIR: I think that both the studies discussed above are not very relevant in our context because of the the very broad definition of ‘error’ and the small percentage of production-based errors in the material (3% in Blackmer and Mitton, unknown but probably small in Seyfeddinipur et al.), thus these studies are not meaningful regarding the validity of MIR for production-based and especially for phonological speech errors. Let us now have a look at the data on the basis of which MIR was proposed. The figure below shows a diagram in Levelt (1983, p. 61). Levelt did not measure cut-off-to-repair times, but error-to -cut-off times. Note that the percentage of production-based errors in his material was 41.6%, thus these data are more meaningful for our concern.

Levelt measured the error-to-cut-off interval (delay of interruption) not in seconds but in syllables (horizontal axis). Vertically, the frequency in which the several delays occurred is shown. As one can see, immediate interruptions (0 syllables delay) were vastly most frequent.

Levelt’s results are confirmed by new data obtained by Martin Pfeiffer, published in his book “Selbstreparaturen im Deutschen. Syntaktischeund interaktionale Analyse” (Self-repairs in German. Syntactic and interactional analysis. Berlin 2015)*. Like Levelt, also Pfeiffer measured delays of interruption in syllables. Here the percentages of the several delays (Table7, p. 130):

0 syllables – 77.4%
1 syllable – 13.4%
2 syllables – 4.6%
3 syllables – 2.5%
4 syllables – 1.2%
5 syllables – 0.3%
6 syllables – 0.2%
7 syllables – 0,2%
8 syllables – 0.1%

Again, immediate interruptions were by far most frequent. Table 8, p. 132 in Pfeiffer's book shows that the delay of interruption does not depend on the type of repair (substitution, deletion, modifying insertion), which suggests that also the time needed for re-planning is not crucial for
the delay of interruption.

An important question in our context is that for a relationship between delay of interruption and type of error (phonological versus other errors). Table 72, p.281 in Martin Pfeiffer’s book shows delays of interruption depending on the linguistic level:

Pfeiffer writes: “There is a tendency to initiate phonological repairs especially quickly, whereas semantic and pragmatic repairs often are delayed.” (282, translated) So I think we can conclude that, at least for phonological errors, the Main Interruption Rule is not obsolete.

There is, however, a further problem: Blackmer and Mitton, Seyfeddinipur et al., and probably also Levelt believe that the interruption of speech after an error is the speaker’s decision, that is, a voluntary behavior. In my theory of stuttering, I assume that the interruption because of an invalid error signal is automatic and occurs independently from the speaker’s will. I will discuss this important issue in my next post in this blog.

Thursday, August 16, 2018

Speech sound discrimination and stuttering

Some weeks ago, a new research article about the speech perception ability of people who stutter was published in the Journal of Fluency Disorders: “Backward masking oftones and speech in people who do and do not stutter” by Shriva Basu, Robert S. Schlauch, and Javanthi Sasisekaran.

They replicated a backward masking experiment previously conducted by Peter Howell and colleagues (Howell et al., 2000; Howell & Williams, 2004; Howell, Davis, &Williams, 2006), in which the detection threshold for a short probe tone in a quiet background was compared with that in the presence of masking noise which was presented immediately after the tone. The decibel difference in the thresholds for the two conditions is the amount of “backward” masking which can be taken as a measure of accuracy and rapidity in central auditory processing.

Howell and colleagues found a greater amount of masking on average in children who stutter as compared with normal fluent children and also in children who persisted in stuttering as compared with those who recovered. There was, however, some overlap between the groups, thus Howell et al. (2006) concluded that an elevated backward masking threshold is sufficient but not necessary for stuttering to persist.

Different from Howell’s experiments, Basu and colleagues used adults (8 who stutter, 8 who don’t) as participants, and, in addition to a tone, vowel-consonant (VC) syllables as stimuli in a further experimental condition: 45 nonsense syllables, with 15 different consonants in three vowel contexts: /a/, /i/, or /u/. All stimuli were presented with and without backward masking. In the tone detection tasks, participants pushed a button when they perceived the tone, in the syllable recognition tasks, they repeated aloud what they heard.

The study confirmed the results obtained by Howell and colleagues: Stuttering participants, as a group, were significantly poorer than the control group in the conditions for tones and speech with backward masking immediately following the stimuli

Poor phoneme discrimination even without masking

The surprising result of the study, however, was that the stuttering group performed significantly poorer than the controls in the syllable recognition task even without masking, that is, in the control condition with quiet background – and there was no overlap across the two groups (see Fig. 2 in the paper).The authors propose two possible explanations for this finding: Either people who stutter have indistinct phonemic categories, or consonants were masked by vowels in the VC syllables.

I prefer the latter explanation. The first one would point to a pure speech perception deficit not related to the findings with non-speech stimuli. And if people who stutter had indistinct phonemic categories, one would expect them to have difficulty in distinct articulation, but this is usually not the case.

The second explanation, by contrast, is coherent with the results of the authors’ and Howell’s backward masking experiments using non-speech stimuli, and further with findings of other studies of auditory abilities of people who stutter, all suggesting a subtle deficit in central auditory processing (e.g., Chang et al., 2009; Hampton & Weber-Fox, 2008; Kikuchi et al., 2011; Prestes et al.,2016; Saltuklaroglu et al., 2017).

Kikuchiet al. (2011), who used click sounds as stimuli in an MEG study, found adults who stutter to have a less effective auditory gating on the left hemisphere, that is, the processing of redundant acoustic information is less suppressed. Evidence for reduced auditory gating comes also from Saltuklaroglu et al. (2017) in an EEG study. Especially these findings of reduced auditory gating are well consistent with the idea that short voiceless consonants are masked by adjacent vowels in auditory processing in people who stutter.

Interestingly, poor discrimination of short consonants that were combined with vowels was already found in an earlier study by Neef et al. (2012). Stimuli were two stop consonant-vowel (CV) continua: /ba/–/pa/ and /da/–/ta/, that is, /ba/ gradually step by step changed into /pa/ and /da/ into /ta/. Participants – adults who stutter and controls – were asked to decide for every step whether they heard /ba/ or /pa/, /ba/ or /pa/. The stuttering group performed weaker and less stable over time in this experiment.

Basu and colleagues used VC syllables as stimuli, Neef and colleagues used CV syllables. If masking by adjacent vowels caused the weaker perception of consonants in the stuttering groups in both studies, then the first case would be a case of forward masking. The second case would be one of backward masking consistent with the findings regarding backward masking of non-speech stimuli reported above.

Does attention modulate phoneme perception?

Basu and colleagues did not observe fatigue effects, participants responded with minimal delay in all trials. In the trials with speech stimuli, the need to aloud repeat the heard syllable may have helped maintaining the attentional focus. Neef and colleagues, who conducted five series of trials, found lower performance in the final series which was most pronounced in the stuttering group in the /da/–/ta/ continuum, and they assume that it could indicate reduced attention.

On the other hand, the performance in the stuttering group was similar to that of control participants in the third series (see Fig. 4 A and B in the paper). The authors wonder: “If one takes the optimal conditions in the laboratory into account, one sees that the problem is probably not an impossibility to perform but instead ready access to sufficient performance.” (285)

It is well known that attention modulates speech perception (see Section 2.3 in my website), but there is little knowledge about the impact of attention particularly on phoneme perception or -categorization. However, Mattys, Barden, and Samuel (2014) showed that (in normal fluent speakers) perceptual sensitivity for phonemes decreased almost linearly with the effort involved in a concurrent, distracting visual task. Likewise, Mattys and Palmer (2015) came to the conclusion that “cognitive load seems to disrupt sublexical encoding, possibly by impairing perceptual acuity at the auditory periphery.”

The deficits and abnormalities in the perception and processing of acoustic stimuli suggest at least a considerable subgroup of stutterers to have an auditory processing disorder (APD). Stavrinoset al. (2018) describe APD as “characterised by normal peripheral hearing, but abnormal processing of auditory information within the central auditory nervous system and by deficits in sound-in-noise discrimination”. Investigating the relationship between APD and attention deficits in children in a pilot study, they found that 20 of 27 children with APD demonstrated underlying attention deficits.

Implications for theory and (perhaps) for therapy

In my theory of stuttering (Attention Allocation Theory) I have assumed that insufficient attention to the auditory channel during speech results in poor processing of auditory feedback, which causes invalid error signals in an automatic monitoring system in the brain. Poor processing of auditory feedback means that parts of feedback information are not completely transmitted or not completely kept in working memory, so that the internal monitor “believes” that a word or phrase has not completely been produced when the next one is already starting.

The results of Basu et al. (2018) and of Neef et al. (2012) point to a further possibility for invalid error signals: Not only gaps in the stream of auditory feedback may occur, but also errors in phoneme categorization. If poor auditory processing of speech feedback leads to phonemes to be not detected or wrong categorized, then short voiceless consonants might be most probable candidates for such processing errors.

In Section 3.4 in my website (on consequences from the theory for therapy) I recommend listening to one’s own voice and words, particularly to the ends of words and to short, unstressed function words, and speaking in a powerful, sonorous voice in order to draw attention to the auditory channel and, in this way, to ameliorate the processing of auditory feedback. Given the new findings, it may also be helpful to attentively listen to short voiceless consonants. Doing so will automatically cause one to distinctly articulate them, which may prevent invalid error signals at these positions.

It is however not easy even for myself to follow my recommendations in some situations, especially when I need much attention for speech planning (e.g., when I’m engaged in explaining complicated things, or when a communication situation is ambiguous). As Mattys and colleagues showed, the perceptual sensitivity for phonemes decreases with growing cognitive load. It is therefore important to reduce the cognitive demands of speech planning whenever possible by pausing between clauses and between units of meaning. 


Sunday, July 29, 2018

Anomalous regulation of visual attention
in children who stutter

In my post from January, I mentioned a surprising finding in a study by Chang et al. (2018): They investigated intrinsic connectivity networks in children who do and do not stutter and found, among others, anomalous, mainly reduced functional connectivity within the visual network (VN) and even more reduced connectivity between VN and dorsal attention network as well as between VN and default mode network in children who stutter, as compared with their normal fluent peers (see Fig. 4 in the study). I took this finding as indicating a general deficit in the involvement of sensory input in the control of behavior. .

This view seems now to be confirmed by the results of a study from Finland, published online this week, by Johanna Piispala, Tuomo Stark, Eira Jansson-Verkasalo, and Mika Kallio: “Decreasedoccipital alpha oscillation in children who stutter during a visualGo/Nogo task.”

I still read only the Abstract which however provides sufficient information for our purpose. The researchers investigated the main oscillations of the brain in 7-9 year old children who stutter and and in age-matched typically developed children in order to discover potential differences related to attention and inhibitory control. EEG data were collected during a visual Go/Nogo task. Stuttering children showed reduced inhibition of the visual cortex and information processing in the absence of visual stimuli, which, so the authors conclude, “may be related to problems in attentional gating. […] Our findings support the view of stuttering as part of a wide-ranging brain dysfunction most likely involving also attentional and inhibitory networks.”

So much about the paper by Piispala and colleagues. The Attention Allocation Theory of stuttering proposed in my website describes a potential relation between attention regulation and a pathomechanism of stuttering, which can be summarized in the causal chain depicted below (it’s the same figure as in my website, but with some explanations).

The causal chain can be closed to a vicious cycle when a child, after having experienced many instances of stuttering, begins to expect this trouble whenever he or she starts talking: Expectation of stuttering, anticipatory struggle, and fear then strongly contribute to the misallocation of attention that has caused the disorder. ‘Disrupted feedback’ in the figure below means that sensory, mainly auditory feedback is not sufficiently processed, not completely transmitted or cached.

  For a long time I believed that this vicious cycle is the way in which stuttering becomes persistent, but this is probably not correct: In an diffusion tensor imaging study, Chow and Chang (2017) found differences in the brain between children who eventually recovered from stuttering and those who persisted (see my post from January). These differences were already present in very young children, thus they can hardly be consequences of stuttering. So I now assume that persistence or recovery are predetermined in most cases, either genetically or by early brain development prior to the onset of childhood stuttering.

The ‘vicious cycle’ might nevertheless exist, but influencing only the severity of stuttering including secondary behaviors. Getting out of that cycle, e.g., by desensitization might therefore be the first step of successful therapy.

Sunday, July 1, 2018

Shadow speech – normal and inverse

In the last months, I did not write here because I was working on a print version of my stuttering theory. This is now finished, and you can download the PDF (349 KB).

Today I want to discuss some aspects of a doctoral dissertation I found in the web some time ago:Inhibitionof stuttering from second speech signals: an evaluation of temporaland hierarchical aspects.” by Daniel J. Hudock (2012).

The studies reported in the dissertation examined, among others, the effect of shadowing on the speech of people who stutter. Shadow speech “is historically defined as the person who stutters lagging or shadowing behind a fluent speaker's utterance. Reductions under shadow speech typically range from 80-90%. “ (Hudock, 2012, Abstract).

First time, the effect of ‘inverse shadowing’ was investigated in this study: The person who stutters did not maintain the lag speaker position (as usual in shadowing), but was maintaining the lead speaker position. This condition is, in a manner, similar to delayed auditory feedback, but produced by a second speaker. Surprisingly, not only normal, but also inverse shadow speech reduced stuttering frequency approximately 80%.

What can we learn from this result? First, it clearly refutes the hypothesis that shadowing induces fluency by providing a correct pattern of speech in terms of rhythm, articulation, prosody, or what ever (this hypothesis was derived from the theory that stuttering is a learned incorrect manner of speaking). Inverse shadowing cannot provide any pattern of speaking, but it still reduces stuttering as well as normal shadowing.

It is further clear that shadow speech does not provide cues for syllable starts, as it has been assumed for choral speech and for speech paced by the beat of a metronome. By the way, also this assumption is wrong: If you each time wait for an external cue and then respond to it, you will never be in sync with the metronome or the co-speakers, but always too late because of the brain’s reaction time. Instead, you must capture the given pace or speech rate such that you can anticipate it. Then, you must adjust your own rhythm or rate and continuously monitor whether it is still in sync. Daniel Hudock points to the fact that, in choral reading, “two speakers may frequently alter speaker positions by speeding up, slowing down or emphasizing different word and sentence components in a constant dynamic fluctuation” (p. 89), which refutes the ‘cue hypothesis’ at least for choral speech.

Do mirror neurons reduce stuttering?

Back to shadow speech: How does it reduce stuttering? Daniel Hudock assumes that the activity of mirror neurons accounts for the similar reduction of stuttering during lead and lag speaker conditions in shadowing: “by perceiving second speech signals people who stutter immediately engage their mirror neuron systems, therefore inhibiting stuttering.” (p. 92). However, are mirror neurons not usually understood as a system that responses to an external sensory input, e.g., with a spontaneous imitation of a perceived behavior?

For example, Kalinowski and Saltuklaroglu (2003a) proposed “that the choral speech effect is a form direct imitation, a primitive and innate human capacity that is possibly mediated at the neuronal level by ‘mirror neurons’.[...] The engagement of these systems allows gestural sequences, including speech, to be fluently replicated. Choral speech and its permutations use the capacity for fluent imitation in people who stutter via a 'loose' gestural matching system in which gestures in the external signal possessing cues found in the intended utterance can serve as stuttering inhibitors.” (339)

Imitation however presupposes that the stutterer is maintaining the lag speaker position, which is the case in normal shadowing, but not (at least not always) in choral speech, and not in inverse shadowing. The activity of mirror neurons may therefore provide an explanation for the effect of normal shadowing, but not for the effect of choral speech and inverse shadowing. The hypothesis does further not account for fluency-inducing conditions in which no second speech signal is provided: speaking paced by a metronome, mouthing (pantomime speech), auditory masking.

By the way, Kalinowski and Saltuklarogly (2003b) equal ‘choral speech’ and ‘unison speech’ with ‘imitation speech’, and this (from my view) incorrect equation seems to be the basis for their theory that the engagement of mirror neurons induces fluency in choral speech and in its derivatives like delayed or frequency-altered auditory feedback. The fact that inverse shadowing has approximately the same effect as normal shadowing should be taken as a suggestion that even normal shadowing does not reduce stuttering by engaging mirror neurons.

The effect of inverse shadowing explained in the framework of the Attention Allocation Theory (AAT)

In my website (here) and in the PDF (p. 27), I explain the effect of normal shadowing in the framework of the AAT: The lag speaker is required to listen not only to the lead speaker, but also to his own speech in order to monitor whether he exactly follows. The lag speaker’s attention is drawn to the auditory channel and to auditory feedback, which improves the processing of auditory feedback (the core idea of the AAT is that stuttering results from invalid error signals due to insufficient processing of sensory, mainly auditory feedback, caused by a misallocation of attention during speech).

However, what is when the stutterer is not the lag, but the lead speaker in shadowing? Daniel Hudock points to the fact that this condition mimics delayed auditory feedback (DAF). In the framework of the AAT, I explain the DAF effect (and the effect of some other kinds of altered auditory feedback) in the following way: Altered auditory feedback sounds unfamiliar and odd, therefore it draws the speaker’s attention to the auditory channel, which improves the processing of auditory feedback.

This explanation is, particularly for DAF, supported by the results obtained by Foundas et al. (2013) and by Unger, Glück, and Cholewa(2012), who examined the effect of electronic devices which reduce stuttering by altered auditory feedback. In both studies, speech fluency was found to be significantly improved by the devices even in a control condition without DAF and without alteration of frequency (FAF): It might have been somewhat unfamiliar to hear one’s own voice not in the natural way, but through the device, and this drew the participants’ attention to the auditory channel. DAF and FAF seem to only increase this effect by making the feedback even more unfamiliar.
Can we therefore interpret the second speech signal in inverse shadowing as similar to an unfamiliar kind of auditory feedback – not only delayed, but also frequency-altered, as the lag speaker’s voice differs from the lead speaker’s voice – which draws the lead speaker’s attention to the auditory channel? I think we can. Interestingly, Daniel Hudock writes that “many participants in the current study self-reported how difficult it was to maintain their own speech productions while not being influenced by the second speaker.” (p. 90). What did they do to not being influenced? They perhaps focused on their own voice.

Therefore, I think the AAT is consistent with the finding that inverse shadowing reduces stuttering in approximately the same extent as normal shadowing does. In both, normal and inverse shadowing, the stutterer’s attention is drawn to the auditory channel, and he is required to listen to his own voice and speech. This improves the processing of auditory feedback and, by that, prevents invalid error signals in the monitoring system and resulting interruptions of speech flow.

The fluency-inducing effect of visual feedback

Finally, I want to briefly discuss another paper by Daniel Hudock and colleagues from 2011: “Stutteringinhibition via visual feedback at normal and fast speech rates.” The main finding of this study is that visual feedback of speech movements – participants viewing the lower portion of their face on a monitor – produced reductions in stuttering frequency ranging from 27% (without delay of feedback) to 62% (400ms feedback delay). Importantly, there was no significant main effect of speech rate or the interaction between speech rate and visual speech feedback; thus the reduction of stuttering cannot be explained as a result of slowed speech due to delayed feedback.

The AAT claims that stuttering results from insufficient processing of auditory feedback and/or of the sensory feedback of breathing – is this consistent with the finding that visual speech feedback reduces stuttering in such a degree?

First, we should consider that visual feedback cannot play any role in the control and self-monitoring of speech, simply because we usually get no visual speech feedback. It is therefore unlikely that visual feedback directly influenced the control of speech in the experiment. The effect may rather be indirect: The speaker’s attention is drawn to external perception in general and away from internal processes like speech planning, somatosensory feedback of articulation, or voluntary control of speech. This change in the allocation of attention may improve also the processing of auditory feedback.

Background of this hypothesis is the study conducted by Chang et al. (2018) who found an aberrant functional connectivity between and within intrinsic connectivity networks in the brain in children who stutter, among them between default mode network and dorsal and ventral attention network. They even found a reduced functional connectivity in the visual network, suggesting that external sensory information in general is not involved in the control of behavior in children who stutter in the same extend as in normal fluent children. The cause may be an imbalance in the attention system (which is suggested also by results of behavioral studies). Such an imbalance can be temporary corrected by powerful, e.g., unfamiliar external stimuli like visual speech feedback.

Monday, February 19, 2018

tDCS – a novel means for the treatment of stuttering

This post is about the article “Transcranial direct current stimulation over left inferior frontal cortex improves speech fluency in adults who stutter” by Jennifer Chesters, Riika Möttönen, and Kate E. Watkins, recently published online in the journal Brain – see here (free full text).

The authors applied transcranial direct current stimulation (tDCS) to the left inferior frontal cortex during speech production in combination with choral reading and metronome-timed speech. They found a significantly greater and more lasting effect of the fluency training combined with tDCS as compared with the same fluency training combined with sham stimulation.

The authors propose “that tDCS over the left inferior frontal cortex during the fluent mode of speaking facilitated plasticity of the frontal speech network and prolonged its normalized functioning, resulting in lasting improvements in fluency.” I completely agree with this explanation, and it is amazing that this could be shown after a fluency training of only five days with sessions of twenty minutes a day.

However, six weeks after intervention, the reduction of stuttering in conversation had decreased significantly also in the tDCS group. It is therefore, important to find out how the fluency training can be improved – it may not be sufficient to only extend the duration of training and tDCS. We should therefore ask what exactly happens during chorus reading or metronome-timed speech. What is the effect depending on, and how can a similar effect be achieved in everyday talking?

The role of the left inferior frontal gyrus

Chorus reading as well as metronome-timed speech were shown to transiently normalize the activation in the left inferior frontal gyrus (IFG, Broca’s area), associated with a reduction or elimination of stuttering. But not only the left IFG is under-activated during stuttered speech, the left auditory cortex (Wernicke’s area) is under-activated too, and also this is normalized by chorus reading and metronome-timed speech (see Table 1 on my website for an overview). There seems to be a relationship between left IFG activation and auditory activation during speech.

Fibers of the superior longitudinal fasciculus terminate in the IFG (see, e.g., Makris etal, 2005), and we can assume that the IFG is involved in the processing of auditory feedback information provided via that fiber tract. This processing, and by that, auditory-motor integration may be impaired when the left IFG is under-activated during stuttered speech. However, the immediate efficacy of chorus reading and metronome timing on speech fluency suggests that the left-hemispheric speech network is quite able to work well if certain requirements are met – which is obviously the case during chorus reading and metronome-timed speech. But what requirements are that?

Chorus reading and metronome-timed speech

Some researchers believe that the speaker gets clues for syllable starts, which helps stutterers because they have difficulty generating their own speech rhythm. But that is not plausible for at least two reasons. First, chorus reading and metronome-timed speech cannot function in the way that the speaker each time reacts to a signal from outside. If you, at each time, wait until you hear the beat of the metronome, or until you hear the co-speakers starting a syllable, and only then say the syllable yourself, then you will never be synchronous but always too late because of the reaction time. Instead, you are required to capture the given beat or pace so that you can predict and anticipate it. Then you must adjust your own pace to the given pace and continuously monitor whether you are still in sync with the metronome or the co-speakers. That is, you must attentively listen to both, the given pace and your own speech, in order to correct your pace if necessary.

Therefore, chorus reading and metronome-timed speech do not only make the speaker listen to the externally given beat or pace, but also listen to his/her own speech. In this way, these conditions improve the processing of auditory feedback (the processing of verbal input is attention-depending – see below) and by that, auditory-motor integration is improved, which results in fluent speech.

A second argument against the hypothesis people who stutter benefit from external cues for syllable starts is that they are quite able to generate their own rhythm, for example, in singing as well as in speaking accompanied by rhythmic arm/hand movements. This is confirmed by an experiment conducted by Howell and El-Yaniv (1987): Adults who stutter were reading a short story (1) normally, (2) while listening to the clicks of a metronome and (3) while listening to clicks occurring at the beginning of every syllable, triggered by the intensity of the speaker’s voice (i.e., it was the participant’s self-generated rhythm). The third condition reduced stuttering nearly as effectively as the second one: The mean number of disfluencies on average in the story was 20.25 in the normal condition, 0.6 with metronome, and 2.5 with click at syllable onset.

The background: attention allocation in motor behavior

Complex automatized sequential motor behavior, e.g., in manual working, sports, dancing, playing music, driving a car, etc., and also speaking requires the ongoing integration of sensory input, among them sensory feedback, in several modalities (visual, acoustic, tactile, kinesthetic) and with that the appropriate allocation of perceptual- and processing capacities for the respective behavior or task. I simply call this the allocation of attention, even if the person often is not aware of it, as it is an integral component of the behavior and was learned and automatized together with that motor ability.

Errors in automatized sequential behavior occur if the appropriate allocation of attention is disturbed or was not correctly learned. For example, attention may be distracted from sensory input, or overly focused on one component of the sensory input, e.g., on one sensory modality. In speaking, the first happens when the speaker is overly focused on the thoughts or emotions that shall be expressed, or on the fear of disfluency; the second happens when the speaker is too much focused on the feedback of articulatory movements (e.g. in the attempt to avoid stuttering) to the detriment of the auditory component and/or the proprioception of breathing.

I propose that developmental stuttering is caused by a misallocation of attention, that is, of perceptual- and processing capacity during speech. The misallocation may be due to several factors (see my last post in this blog), but chorus reading and metronome-timed speech seem to be tasks which compel the speaker to reallocate his/her attention, namely: to listen during speech – not only to the co-speaker or to the metronome, but also to his/her own speech.

Consequence for treatment
In chorus reading and in metronome-timed speech, the speaker is compelled to reallocate attention, but might mostly not become aware of this fact, especially not of the fact that he or she must listen not only to the co-speaker or the metronome, but also to his/her own speech. Thus the reallocation of attention is not maintained in everyday talking. One goal of therapy should therefore be to make clients aware of the necessity to reallocate attention during speech, and to practice listening to one’s own words in everyday situations.

Lateralization of the processing of verbal acoustic input is depending on attention (it is left-lateralized only during active listening; Poeppel et al.,1996; Rämä et al., 2012; Sabri et al., 2008), and particularly on attention to the lexical aspect of speech (attention to the prosodic or sound aspect draws processing to the right hemisphere; Hugdahl etal., 2003; Vingerhoets, Berckmoes, and Stroobant, 2003). What is true for the processing of external verbal input might also be true for the processing of auditory feedback. It is therefore important to listen not only to one’s own voice, but to one’s own words.

In this way, it should be possible to maintain a normal activation of the left-hemispheric speech network, and transcranial direct current stimulation seems to be a good means that can support this by promoting a change in brain structure.

Friday, January 26, 2018

Persistence versus recovery of stuttering

In this post, I want to sketch some ideas about the development of stuttering. Today stuttering is often called a neurodevelopmental disorder, which might be correct, however, the brain is plastic, especially in young children, and its development interacts with learning and with the development of behavioral habits and routines.

A part of behavior often overlooked is the allocation of attention, that is, of perceptual and processing capacity during motor tasks, including speaking. In my theory of stuttering, I assume that just this mostly unconscious aspect of behavioral control is the crucial causal factor. The figure below shows the hypothesized development from stuttering onset to recovery (persistent stuttering is not included in this figure). The colors mean: blue = developmental steps, red = causal factors, yellow = triggers or secondary negative factors, green = positive factors.

Transient developmental stuttering
Devekopment of transient stuttering and recovery

I think the cause of childhood stuttering, in most cases, is an imbalance in the development of sensorimotor integration, namely to the favor of action and to the detriment of perception and feedback processing. This becomes problematic in the change to connected speech and sentence production which requires a stronger involvement of auditory feedback and the feedback of breathing in speech control – the brain must perceive and keep in memory which constituents of a sentence have already been produced in order to correctly complete the sentence, and breathing times must be included in the speech sequence.

Of course children are not aware of changing behavior when they start forming sentences. Empirical findings suggesting such an imbalance in the attention system are
  • a high number of dopamine D2 receptors at age 2.5 to 3 (Alm 2004, Fig. 2; see also Rothmond et al. 2012, Fig. 3), associated with a tendency towards generally high motor activity;
  • lower fractional anisotropy (FA) in the left arcuate fasciculus, probably indicating delayed fiber maturation (see Fig. 1, Cluster 1 and 2 in Chow and Chang, in press – interestingly, these clusters of lower FA were much larger in the recovered group than in the persistent group). The structural deficits might be the manifestation (rather the result than the cause) of poor involvement of sensory feedback in motor control.
  • atypical network connectivity in children who stutter (persistent as well as recovered) compared to controls. See Fig. 5 in Chang et al (2018), especially the reduced connectivity between Dorsal Attention Network and Default Mode Network and the hyperconnectivity between Ventral Attention Network and Default Mode Network, indicating an imbalance in the control of attention. Interestingly, there was even found a reduced connectivity in the visual network in stuttering children (Fig. 4), suggesting a general deficit in the involvement of sensory input in the control of behavior.  
The idea that the onset of stuttering in most cases is related to the beginning of connected speech and sentence forming is not new. Bloodstein (2006) pointed to the facts that “early stuttering occurs only on the first word of a syntactic structure; stuttering does not appear to be influenced by word-related factors; early stuttering seldom occurs on one-word utterances; the earliest age at which stuttering is reported is 18 months, with the beginning of grammatical development; the age at which most onset of stuttering is reported, 2-5 years, coincides with the period during which children acquire syntax; considerable spontaneous recovery takes place at the time most children have mastered syntax; incipient stuttering is influenced by the length and grammatical complexity of utterances...”
A further suggestion to an imbalance in development as a causal factor comes from the observation that not a few children with early-onset stuttering show syntactic abilities and length of utterances well above the norm for their age (Watkins, Yairi, and Ambrose, 1999). Alm(2004) writes about that: “The group with early onset stuttering, who entered the study at age 2-3, showed syntactic abilities and length of utterances well above what was expected for their age. In fact, in some aspects the language abilities in this group were on a level with the norms for 2 years older children. This was true both for children who recovered and for children who persisted to stutter.” Too much attention (capacity) may be taken by sentence planning, and too little attention may remain for perception and feedback processing in these children.

Late-onset stuttering and the so called ‘psychogenic stuttering’ was included in the above figure because I think that cause and pathomechanism are the same as in developmental sttutering, but affected individuals seem to have no strong predisposition for stuttering, especially not for persistent stuttering. But in a person whose attention system is vulnerable, strongly negative emotions, distress, fear, or the aftermath of a trauma may result in a misallocation of attention during speech and by that in stuttering. Complete recovery is often reached in such cases by a supporting environment and/or by therapy, including psychotherapy that helps coping with traumatic experiences (see Table 1 in Chang et al.,2010).

Spontaneous recovery in general may be caused by a kind of unconscious learning effect: Children eventually learn to adapt their attentional allocation, that is, the allocation of their perceptual and processing capacity to the new demands of connected speech and sentence production. Such a learning effect manifests in brain structure: See, e.g., the upward trajectories of FA for the recovered group in Fig. 1 (Cluster 3 and 5), and Fig. 2 (Cluster 6) in Chow andChang (in press). I interpret those changes in brain structure as consequences of learning since several studies have shown that even a few weeks of practice (e.g., in reading, juggling) result in changes of gray or white matter structure (see Section 4.1 on my website).

When I assume that most stuttering children eventually learn to adapt their attentional allocation to the demands of connected speech. then this doesn’t mean they all learn to behave completely in the same way as children who have never stuttered do. The results depicted in Fig. 6 in Chang et al. (2008) suggest that the left brain hemisphere played a minor and the right one a greater role in speaking in the children who had recovered from stuttering, compared to the normal fluent controls. The cause could be that speech control in the recovered children, on average, was not as automatic (i.e., more volitional) as in normal fluent children. 

I don’t like to speak of compensation in the brain here, because there is no little manager sitting in the brain who notices a dysfunction and bids another part of the brain to sub. It is rather the child who becomes aware of a problem and tries to cope with it, and this behavior, over time, results in a structural change in the brain.

Persistent developmental stuttering

Development of persistent stuttering
If my above ideas about transient stuttering are true, then the question arises: Why does a minority of the children affected by stuttering not learn to adapt their attention to the demands of connected speech? Is there an additional causal factor impairing the allocation of attention in these children? In fact, there seems to exist such a factor: a subtle deficit in central auditory processing. Many empirical findings point in this direction (see Section 3.3.1 for an overview). Unfortunately, we have data about auditory processing in persistent stuttering only, but not in individuals who recovered. 

 Unfortunately, there are few data comparing auditory processing in those who persisted in stuttering and in those who recovered. The only study I know is by Howell, Davis, and Williams (2006) who compared children of both groups in a backward-masking task the performance or which is assumed to reflect the operation of central auditory processing, especially of temporal structure. They found an appropriately 10 decibel higher mean backward-masking threshold in the persistent group. The difference was statistically significant, but there was a high variability in the persistent group, thus the authors conclude that an auditory deficit may be sufficient, but not necessary, for the disorder to persist.
Further, Usler and Weber-Fox (2015) and Mohan and Weber (2015) found differences between children with persistent stuttering and those who recovered in the processing of auditorily presented verbal stimuli. Furthermore, Chow and Chang (in press) found a structural deficit (lower FA) in fibers of the splenium, i.e., the posterior part of the corpus callosum in children who eventually persisted in stuttering, but not in those who eventually recovered (see Fig. 1, Cluster 4 in the study). The affected fibers probably connect bilateral temporal regions (Kuvazeva, 2013), and lower FA may be related to a less effective labor division between hemispheres in auditory processing.

The difference in FA in the mentioned cluster is great already with the youngest children, and there is no much overlap between the persistent group, on one hand, and the recovered and control group, on the other hand. Therefore, the deficit can hardly be a consequence of stuttering. Interestingly, Chow, Liu, Bernstein Ratner, and Braun found a strong relation between the FA in the splenium and stuttering severity in adults who stutter (unpublished study; the results were presented at the 2014 ASHA Convention).

A further finding by Chow and Chang (in press) which distinguished the persistent from the recovered group of stuttering children is an initially higher FA value and an abnormal developmental trajectory in the thalamic radiation (see Fig. 2, Cluster 8, 9, and 10 in the study). These findings as well can hardly be explained as a consequence of stuttering. As the thalamus plays a central role in attention regulation, the findings may be related to an abnormal development of the attention system, possibly increasing the imbalance described above as the causal Factor 1. This assumption is supported by findings of anomalous, mainly decreased functional connectivity between Dorsal as well as Ventral Attention Network and Default Mode Network especially in children who eventually persisted in stuttering (Fig. 3 in Chang et al., 2018).

The hypothesis that transient and persistent developmental stuttering are in core the same disorder, and  that persistence is caused by an additional factor has already been proposed by Ambrose, Cox, and Yairi (1997) in a study of the genetic basis of persistence and recovery. They wrote: “ It was found that recovery or persistence is indeed transmitted, and further, that recovery does not appear to be a genetically milder form of stuttering, nor do the two types of stuttering appear to be genetically independent disorders. Data are most consistent with the hypothesis that persistent and recovered stuttering possess a common genetic etiology, and that persistence is, in part, due to additional genetic factors.”

The proportion of the two factors may differ between individuals, however, Factor 2 seems to be necessary for stuttering to become persistent. Factor 1 may be differently caused in persistent stuttering than in transient stuttering – as mentioned above, the clusters of reduced FA in the arcuate fasciculus (Cluster 1 and 2 in Fig. 1 in Chow and Chang, in press) were larger and the FA value in the anterior cluster was lower in the recovered than in the persistent group. However, in the persisten group, it was found an abnormal, stagnant or downward developmental trajectory of FA in the radiation of the thalamus (Fig. 2, Clusters 8, 9, 10). The thalamus is a part of the brain which plays a central role in the control of attention. We can assume that the male-to-female ratio in persistent stuttering has to do with Factor 1, namely with the fact that hyperactivity and impulsivity are more prevalent in males. 

Tuesday, January 2, 2018

Why are right frontal brain areas overly active?

In this post, I want to discuss the article “Structural connectivity of right frontal hyperactive areas scales with stuttering severity” by Nicole Neef and colleagues, recently published in the journal Brain – see here(free full text).

I start with the question of whether my theory is consistent with the results of the study. After that, I discuss the authors’ hypothesis that stuttering could be caused by a global response suppression mechanism. I use the following abbreviations: BG = basal ganglia, IFG = inferior frontal gyrus, MFG = middle frontal gyrus, SLF = superior longitudinal fasciculus, SMA = supplementary motor area, STG = superior temporal gyrus

Reduced fractional anisotropy in the SLF – related to deficient myelination?

In the left SLF/arcuate fasciculus of the stuttering participants, a weaker connectivity than in controls was found along the major diffusion direction of the fiber tracts. The authors conclude that this “favours the view that atypical structures are insufficiently myelinated or that the axonal packing is reduced therein”. This is consistent with my assumptions about the role of myelination in Section 4.1. However, I don’t believe that the structural deficits in the fiber tracts compromise signal transfer. 
If it was the case, and if that caused stuttering, then the disorder could hardly be as variable as it is. It could hardly be so much influenceable by situations, emotions, or anticipations; it could not suddenly be eliminated by conditions like chorus reading. We would expect stuttering to be a more invariable, only gradually changing disorder if it was immediately caused by a structural deficit.
However, there’s an alternative explanation: The fibers are able to work well, and their structural weakness is the result of reduced activation due to a habitual misallocation of attention, i.e., a misallocation of perceptual and processing capacity during speech, and perhaps during other automatic motor behavior (see Section 4.1). 

This view is supported by the evidence that fiber structure develops with training, i.e., depending on activation (e.g., Keller and Just, 2009; Scholz et al,2009), by the evidence of deficient attention regulation in stutterers (see 3.3.1), and by the fact that their auditory areas were often found to be deactivated during stuttered speech (see Table 1), suggesting that auditory feedback is insufficiently involved.

Compensation via uncinate fasciculus 

A negative correlation was found between the severity of stuttering and the connection strength (connection probability density) in the right uncinate fasciculus linking the frontal pole to higher-order auditory and multisensory areas of the STG. This finding is consistent with my assumption that an insufficient involvement of auditory feedback plays a crucial role in stuttering. But why is compensated for that deficit by the right uncinate fasciculus being additionally involved?

In Section 3.2, I report some findings suggesting that reduced auditory attention results in a shift of processing from the left more to the right brain hemisphere, and in Section 4.4, I develop a hypothesis based on the dual steam model, why semantically and syntactically correct speech can be produced with little auditory attention, reduced involvement of the SLF, reduced phonological feedback processing, depending more on the ventral stream part of which is uncinate fasciculus.

Is global response suppression the cause of stuttering?


Right posterior IFG and right MFG were found to be hyperactive during vocal imagination tasks (imagine speaking versus imagine humming a melody), and the connection strength in some fiber tracts originating from these hyperactive areas was positively correlated with stuttering severity. It is not clear whether these abnormalities are related to the cause of stuttering or to maladaptation or compensation, and the authors hypothesize “that stuttering might be caused by an overly active global response suppression mechanism mediated via the subthalamic nucleus-right IFG-basal ganglia hyperdirect pathway”.They assume that such a global response suppression mechanism “induces an unspecific broad inhibition” which “would hinder the smooth successive execution of appropriate motor actions”.

I think there are some arguments against this hypothesis. First, the subthalamic nucleus-right IFG-BG hyperdirect pathway (via which the global response suppression mechanism is mediated) seems to control voluntary behavior (after the model proposed by Goldberg, 1985). Likewise, right MFG and posterior IFG seem to control mainly voluntary behavior, namely “involved in the ability to apply executive control over actions”. In stuttering, by contrast, speech flow is disrupted against the person’s will.

Secondly, the question would arise why the assumed overactive global suppression mechanism only affects speaking, but not other motor behavior. If the hypothesis was true, it would mean that a neuronal network responsible for the smooth execution of all voluntary movement would completely fail many times a day – but strangely in speaking only.

A third argument against the above hypothesis is provided by cases in which persistent stuttering suddenly and completely disappeared after a lesion in the left hemisphere of the cerebellum (e.g., Bakheit, 2011; see also Section2.1) The crucial role the cerebellum plays in stuttering became clear in the study by Wymbs et al., 2013: They found little across-subject agreement of activated brain regions during stuttered speech – the only region which was overly active in all the four participants was the left cerebellum. Therefore, I think the left cerebellum is the likeliest candidate for the source of signals disrupting speech flow, not least because the cerebellum is involved in the response to errors in motor sequences (e.g, Zheng et al, 2013; see also footnotein Section 2.1).

What role may right SMA and BG play in stuttering?

If greater activation in the right IFG and MFG in the stuttering participants during the imagination tasks was related to the stop response, then perhaps because it was some more difficult for them to suddenly stop when the signal was displayed (I participated in the study myself, and I remember I needed some effort to stop when I just was in the flow in internally humming the melody). Such difficulty in inhibitory control seems to be rather typical of stutterers, as well as a tendency towards hyperactivity and impulsivity (see Section 3.3.1). However, I don’t believe that it immediately causes stuttering; it may rather be (i) a factor in the predisposition for stuttering and (ii) a factor influencing the severity of individual symptoms.

The first assumption – that an overly active SMA-BG circuit contributes to the predisposition for stuttering – is derived from the findings regarding a tendency towards hyperactivity, impulsivity, difficulty in attention regulation (shift of attention, dividing attention in dual tasks), and deficient inhibitory control. All this suggests an imbalance to the favor of voluntary, internally initiated, targeted action, but to the detriment of (non-targeted) perception and response, including the processing and involvement of sensory feedback – processes which are not voluntary but more unconscious and automatic. Likewise, the positive correlation between stuttering severity and the anatomical connections of the right frontal aslant tract linking the posterior IFG with SMA and preSMA may be related to a tendency towards too much control by the will

Background of the second assumption – that an overly active SMA-BG circuit contributes to the severity of symptoms – is the model that a stuttering event consists of two parts: (a) blockage of a motor program, (b) the speaker’s will to continue, that is, the ‘drive’ – the SMA-BG circuit starts the motor program again and again or keeps it active despite its execution is blocked (see also footnote in Section 2.1).

Since the SMA-BG circuit is responsible for voluntary behavior, its activity is influenceable by the will: In stuttering modification therapy, clients learn to give up the urge to speak when feeling a blockage. Overt symptoms or at least their severity can be reduced in this way, hard blocks, long prolongations, or often repeated iterations are avoided. Dopamine receptor blockers seem to act in a similar direction, as Alm (2004) concludes from the literature: the drug “exerts its main effect in reducing superfluous motor activation during stuttering, not in reducing the number of disruptions” (p. 337).

A third role of right frontal brain areas in stuttering must be mentioned for the sake of completeness: volitional control of speech, particularly deliberate speech planning in order to avoid or postpone words feared to be stuttered. Overactivation in the right frontal operculum, negatively correlated with stuttering severity but reduced after successful therapy, may be related to such compensatory behavioral habits.