A Theory of Stuttering
A Theory of Stuttering
In shadowing, the stutterer immediately copies what another person says, having only a vague (or no) idea of the words to be spoken until hearing them (see, e.g., Cherry & Sayers, 1956; Marland, 1957; Wingate, 1969b). That is, the other person (the “lead speaker”) and the stutterer speak concurrently most of the time, but not synchronously. The method proved very effective; for instance, Andrews et al. (1982) found a 95% reduction in stuttering on average. Shadowing was applied in stuttering treatment mainly in the 1960s (e.g., Kelham & McHale, 1966; Kondas, 1967; MacLaren, 1960; Shelton, 1975; but see also Rongna et al., 2018).
The time lag implies that stutterers get no cues for syllable starts in shadowing. The often astonishing fluency of stutterers during shadow speech was explained mainly in two ways. First, it was explained in the framework of a theory that regarded stuttering a learned wrong pattern of speaking. Fluency during shadowing was believed to result from the exact imitation of a correct speech pattern (e.g., Kelham & McHale, 1967; MacLaren, 1960). The second explanation was the distraction hypothesis. Cherry and Sayers (1956) believed that stutterers are distracted from their somehow distorted auditory feedback in shadowing, and Kelham and McHale (1966) even asserted that “shadowing prevents feedback from taking place at all” (p. 114).
The hypothesis that the imitation of a correct speech pattern eliminates stuttering during shadowing has been refuted by Hudock (2012), who examined the fluency-enhancing effect of two conditions: (1) the stutterer, as usual, shadows a normally fluent person’s speech; (2) the stutterer is the lead speaker whose words are shadowed by a normally fluent speaker. Stuttering frequency was reduced by approximately 80% in both conditions. The result indicates that the effect of shadowing does not depend on imitation (read more).
The hypothesis that stutterers are distracted from auditory feedback is as implausible for shadowing as it is for choral speech. The shadower needs to listen not only to the lead speaker, but also to his or her own speech to monitor whether a segment has been repeated completely so that the repetition of the next, already heard and “stored” segment can start. Similar to choral speech, the shadower’s attention is focused on both, the co-speaker and the auditory feedback of his or her own speech. The latter improves the processing and integration of auditory feedback and reduces stuttering.
Saltuklaroglu and Kalinowski (2011) investigated the “carry over effect” of shadowing and passive listening prior to normal speech production. Stutterers read aloud short text passages after either shadowing syllables or passively listening to the same syllables. Both shadowing and passive listening reduced stuttering in the ensuing readings. Shadowing as well as listening probably drew the participants’ attention to the auditory channel; that is, the attention system was modulated so that auditory feedback was better processed and integrated during the ensuing readings.
Slow speech as a means to reduce stuttering has been applied in treatment at least since the 19th century (see, e.g., Van Riper, 1973, for an overview). Empirical studies demonstrated that a sufficiently slow speech rate by way of syllable prolongation strongly reduces or eliminates stuttering. For instance, Perkins et al. (1979) found that stuttering was essentially eliminated when stutterers reduced their rate by approximately 75%. Perkins believed that lacking coordination of phonation, articulation, and respiration caused stuttering (see also Perkins et al., 1976). Hence, he assumed that, in prolonged speech, stutterers overcome that discoordination by adapting their rate to their limited motor control capacities.
However, this account is hardly consistent with the finding that speaking at a high rate (50% above normal or “as fast as possible”, respectively) did not result in an increase in stuttering frequency in stutterers (Ingham, Martin, & Kuhl, 1974; Kalinowski, Armson, & Stuart, 1995). Furthermore, the discoordination hypothesis is not well consistent with the finding that the fluency-enhancing effect of altered auditory feedback is not reduced by speaking at a high rate (Kalinowki et al. 1993, 1996).
Although the effect of other FCs, such as rhythmic speech, choral speech, or altered auditory feedback, does not depend on a slow rate, there is still some influence of rate. Hargrave et al. (1994) and MacLeod et al. (1995) found that a smaller reduction in stuttering by altered auditory feedback was achieved at fast speaking rates compared to normal rates, and Fransella and Beech (1965) found that slower rates in metronome-timed speech resulted in even less stuttering than faster rates. Nevertheless, the question arises as to why slow, prolonged speech reduces or even eliminates stuttering, although a high rate is obviously neither the cause of stuttering nor essential for fluent speech.
Studies of brain activity during slowed speech have provided some suggestions. In a neuroimaging study, Hashimoto and Sakai (2003) investigated the brain activation during slowed speech in normally fluent participants. From their findings, they conclude that slowed speech seems to be associated with a change in attention; the participants probably paid more attention to the task during slowed speech compared with speaking at their habitual rate. Kittilstved et al. (2018) investigated the effect of slowed speech on brain activity in normal speakers in an EEG study. Based on their data, they conclude that “prolonged speech was characterized by increased feedback due to the novelty of the task and/or increased monitoring” (p. 11).
The effect of slowed speech on brain activation in stutterers and normally fluent controls was investigated by De Nil et al. (2008). During speaking at a habitual rate, left auditory association areas were under-activated in stutterers as compared with controls, but this group difference disappeared during slowed speech. Furthermore, right auditory areas were greater activated during slowed speech than during habitual speech in the stuttering group. Together, the brain research results suggest that slowed, prolonged speech reduces stuttering similarly as rhythmic, choral, and shadow speech do, namely, by modulating the allocation of attention in favor of auditory feedback.
The effect on stuttering of slowed speech is strong, similar to that of choral reading, metronome-timed speech, or shadowing—FCs that require active listening, not only to co-speakers or a given rhythm, but also to the auditory feedback of one’s speech. This is not the case in slow, prolonged speech, but there is a mechanism that closely couples prolonged speech to auditory feedback: Kalveram (1983) found that the duration of long-phonated syllables and, with that, the start of the next following syllable is controlled based on auditory feedback. He called this mechanism “audio-phonatory coupling” (see also Jäncke, 1989, 1991; Kalveram & Jäncke, 1989; Natke 1999).
Audio-phonatory coupling means that the auditory feedback of the start of a long-spoken syllable provides the information by which the speaker determines the duration of that syllable. Audio-phonatory coupling is only possible with long syllables because the processing of auditory feedback takes too much time for controlling the duration of short syllables in this way.
Audio-phonatory coupling implies that the production of long syllables requires the involvement of auditory feedback, but the production of short syllables does not. If all syllables of an utterance are produced as long syllables, as in slow, prolonged speech, then auditory feedback is permanently involved, such that speech control cannot decouple from it. This explains the great reduction in stuttering during slow, prolonged speech (see also Section 2.9).
It has been known for a long time that almost all stutterers do not stutter when singing. Wiechmann and Richter (1966) reported that 31 out of 1582 German school-age children (approximately 2%) stuttered during singing. Andrews et al. (1982) found stuttering to be reduced by 98% on average during singing compared to speaking, with 0% stuttering in two out of three participants and 4% in one participant.
The fluency-inducing effect of singing has sometimes been regarded as an effect of rhythm (e.g., Johnson & Rosen, 1937), but Wingate (1969b) pointed to the fact that the correspondence between words, melody, and beat is often not very close in songs. He therefore suggested that changes in vocalization, which are actually an integral part of singing, make stuttering disappear. Healey, Mallard, and Adams (1976) concluded from their findings that a reduced rate and the familiarity of the melody and lyrics may contribute to the fluency-inducing effect of singing. Colcord and Adams (1979) examined the change in vocalization during singing and concluded that the extension of voicing duration may partially or solely be responsible for the elimination of stuttering.
Glover et al. (1996) came to different conclusions. They asked stutterers to read and to sing prose passages at a normal and at a fast rate. Stuttering was reduced by 75% on average in singing as compared with reading, and there was no difference in stuttering frequency with rate. In the singing conditions, participants spontaneously created idiosyncratic melodies or melody-like structures, and their singing was not always very “musical”; nevertheless, they experienced a great enhancement of fluency. The authors conclude that the fluency-inducing effect of singing can neither be attributed to memorized material nor to an imposed rhythm, and also not to a reduced rate.
Stager, Jeffries, and Braun (2003) searched for a common feature of metronome-timed rhythmic speech and singing that presumably causes their fluency-inducing effect on stutterers. They found an increase in voicing duration during singing, but not during rhythmic speech, and they concluded that extension of voicing duration cannot be the common mechanism searched for. Then, they examined the brain activity in stutterers during rhythmic speech and singing and found greater activations in the auditory association areas in both FCs as compared with two disfluency-evoking conditions. Stager and colleagues conclude that rhythmic speech as well as singing may reduce stuttering “by enabling more efficient use of auditory information” (p. 333).
Singing requires the integration of auditory feedback mechanisms with the vocal motor system (Zarate & Zatorre, 2008). Therefore, I propose that, when singing a familiar song, attention is automatically drawn to one’s own voice to monitor whether the intonation is according to the melody. Listening to one’s voice might be necessary even more when improvising a melody, as in the above-mentioned experiment by Glover et al. (1996). An additional fluency-inducing factor is that, in song, syllables (more precisely, their phonated parts) are often prolonged, and their duration must be timed exactly according to the melody. This is controlled by audio-phonatory coupling (see above in Section 3.6).
A special variant of speech shadowing is simultaneous interpreting. It is, so to speak, shadowing in another language. I personally know two stutterers who can translate simultaneously, one of them from Russian into German, the other between Turkish/Kurdish and German. The latter, a woman, has worked as a professional interpreter for many years. Both do stutter in normal communication, but are fluent when simultaneously interpreting. This makes it clear that the fluency during shadowing is not based on imitation.
(return)