In adolescents and adults whose stuttering has become persistent, symptoms do typically not occur immediately at sentence onset, i.e., on the first syllable of the first word. The cause might be that, over time, almost all stutterers learn to plan sentences incrementally, thus not all planning effort is required at sentence onset. However, stuttering still mostly occurs more in the initial portion than at the end of clauses and sentences (read more), because also with incremental planning, more attention to planning is needed in the initial part of a clause or a sentence than at the end (read more). But different from early childhood stuttering, in persistent suttering, sentence planning is no longer the main cause for the misallocation of attention during speech. Caruso et al. (1994) have shown that stutterers produce more disfluencies under cognitive stress, and cognitive stress comes about by high demands in sentence planning, but also by the attempt to avoid or to cover stuttering.The following thesis should appear plausible prima facie:
In persistent stuttering, the main cause for a misallocation of attention during speech is no longer sentence planning, but rather stuttering itself: It is the fear of stuttering, the attempt to avoid or to control stuttering, and the stress as a consequence thereof which occupies the speaker’s attention.
Some other factors come in addition: Stuttering frequently occurs on the salient words of a sentence, namely on long words as well as on words with high information load, i.e., on unexpected or rarely occurring words ((read more)). Also this pattern might be caused by a misallocation of attention: Put the case that the speaker, abuzz with what he or she wants to tell, directs attention only to the intended message and to the salient words so that attention runs ahead to these words. As a result, too little attention to the auditory channel may remain at the preceding ‘unimportant’ word (e.g., a short, unstressed function word). Consequently, this word (or its end) is not detected by the monitoring system, and the monitor blocks the speaking program of the subsequent salient word attention was prematurely focused on. That means: If stressed words, long words, or words with high information load are more likely stuttered, then it is not due to special features of those words, but because the speaker’s attention is more likely focused on them.
As already mentioned, fear or emotional stress during speech can demand much attention: When a child is required to choose words carefully in order to avoid misunderstanding or punishment, when he or she anxiously watches the listeners’ countenance and response, then much attention is needed for that, and too little attention (perceptual- and processing capacity) may remain for auditory feedback. This accounts for cases in which the onset of stuttering is apparently triggered by psychical factors. But a similar situation arises after a child has become aware of his/her stuttering: Then, talking is often angst-burdened, and attention is automatically focused on the threat, which interferes the processing of non-threat-related information (Compton, 2003; Ferneyhough et al., 2013; Koster et al., 2004).
In many cases, stutterers direct much attention to the avoidance or control of stuttering, or to the listener’s response to symptoms. In the attempt to avoid stuttering, words are chosen carefully, words with feared initial sounds are substituted, and/or much attention is directed to articulatory or breathing movements and their control by the will. All these factors contribute to the misallocation of attention during speech. A vicious cycle, as depicted in Figure 9, develops and is reinforced by various secondary behaviors. These reinforcing factors may crucially determine the severity of stuttering.
Figure 9: The vicious cycle of persistent stuttering and the impact of some secondary behaviors. Note that the causal chain (see Fig. 8) is closed to a cycle by the awareness of the disorder and the expectation of stuttering, and that attention allocation is the main interface via which secondary behaviors reinforce the vicious cycle.
A correlation between the expectation and the occurrence of stuttering was already described in the Anticipatory Struggle Hypothesis (Bloodstein, 1958; 1975): In reading, a word is more likely stuttered if the person anticipates stuttering on this word. In fact, Knott, Johnson, and Webster (1937) found that 96 % of anticipations of stuttering were followed by stuttering, and almost 94 % of all stuttering events occurred on words on which stuttering had been anticipated. These results were broadly confirmed in later studies (see Bloodstein & Bernstein Ratner, 2008, for an overview) (read more about how anticipations of stuttering arise).
However, I think it is not an increased muscular tension due to the anticipation of struggle, as Bloodstein believed, that results in stuttering – it may, at most, lead to more severe symptoms. But the expectation of struggle draws the speaker’s attention prematurely to the ‘dangerous’ word – perhaps, the speaker would substitute or postpone this word in self-formulated speech (which was impossible in the reading tasks). Or the speaker tries to prepare the articulation of the feared word onset – by means of ‘abnormal preparatory sets’, as Van Riper (1973) assumed. Therefore, it might be the misallocation of attention due to an anticipation of stuttering that results in stuttering just at the expected position.
It must be emphasized that stuttering occurs also without any expectation, namely at the onset of childhood stuttering – it is the main weakness of the Anticipatory Struggle Hypothesis being unable to convincingly explain the outbreak of stuttering – as well as very often in persistent stuttering, in spontaneous speech more than in reading. The crucial cause of the disorder is that attention is too much detracted from hearing one’s own speech and/or from the self-perception of breathing. Anticipatory struggle can cause a misallocation of attention, but the misallocation can also be caused by focusing too much on the content (the message) or on careful formulation (sentence planning), or by worry about the listeners’response. – Now, we can explain the distribution of stuttering events in connected speech more universally:
A word is stuttered more likely if it attracts the speaker’s attention, whether because of its content or its length, or because stuttering is expected on this word. In all these cases, it is not actually the features of the word which triggers stuttering, but the fact that the speaker’s attention is prematurely focused on this word, with the effect that the completion of the preceding speech unit is not detected by the monitoring system.
The same may be true for syllables: In English as well as in German language, first syllables of words very often are stressed, and if the word is stuttered, then usually the first, i.e., the stressed syllable is stuttered. If another syllable than the first is stuttered, then usually, a stressed syllables is affected (Brown, 1938; Hahn, 1942; Natke et al., 2002). This so called ‘stress effect’ was also found in preschool-aged stuttering children (Natke et al., 2001; 2004a). Further syllables with feared initial sounds may more likely be stuttered. In all these cases, the speaker’s attention is prematurely shifts the stressed syllable or to the syllable starting with a feared sound, thus the end of the preceding unstressed syllable is not detected by the monitor. More about stuttering on syllables within words in Section 2.1, see here.
It is a weakness of all theories that ascribe stuttering to a breakdown or a loss of control that they have difficulty explaining the variability of stuttering: Why should speech control break down in some situations, but not in others? The ‘demands and capacities’ approach does not provide satisfactory explanation: Not uncommonly, one and the same person is fluent when speaking in front of a large audience, but stutters when talking with friends. I think that speech control never breaks down in stuttering. In the contrary, the control system behaves in its normal manner and interrupts speech flow when it has detected a mismatch between expectation and feedback. However, a situation, e.g., soliloquizing, talking to a baby or a pet versus talking to a teacher or a superior, or to a large audience – and not least the speaker’s attitude towards the situation – influence the allocation of attention. Again, attention allocation might be the crucial ‘interface’, via which speech situations can intensify or inhibit the pathomechanism of stuttering.
Finally, an argument alleged by Postma and Kolk (1992) must be mentioned: They regarded the finding that stutterers have no difficulty detecting their own speech errors as evidence against all theories that claim distortions of auditory feedback being the cause of stuttering. However, in the proposed theory, it is assumed that only the ends of speech units, especially of inconspicuous, unstressed, ‘unimportant’ words, and the ends of unstressed syllables within words are not detected by the monitoring system. Slips of the tongue affecting only these positions are rare, at least with talking in one’s own native language and with normal language ability. Analyzing a collection of 474 German speech errors (Wiese, 2010), I have found only 28 errors that meet those criteria (ca. 5,9 %), among them a number of grammatical errors which cast doubt on their originators’ language ability. That means: A temporal overlap of the feedback disruptions assumed here with common speech errors is not very likely, therefore, the present theory does not predict differences between stutterers and nonstutterers in the detection and repair of speech errors. (read more).
The presentation of the theory, in the narrower sense, ends here. In the following chapters, some specific issues and empirical findings are discussed in the light of the proposed theory.
In reading, the first word of a sentence is, on average, more frequently stuttered also in older children and adults (Brown, 1945). However, Griggs and Still (1979) published individual results of a reading task, and as can be seen there, only three of their six subjects (four school-aged children and two adults) stuttered more frequently on the first word, and the bias was strong in only two subjects. In spontaneous speech, stuttering on the 2., 3., or 4. word of a sentence is very typical in older children and adults, but not stuttering on the first word (Heina, 1955, reported by St. Louis, 1979; Hannah & Gardner,1968; Lanyon, 1969). Koopmans, Slis, and Rietveld (1991) found 13.0 % words on the first position and 15.6 % words on the second position of clauses stuttered on average in the spontaneous speech of adult stutterers.
I often listened to stuttered speech in self-help groups and paid attention to if stuttering occurred on the first word of sentences – it was very rarely the case. On the one hand, the first word of an English or German sentence often is a small function word – an article, a pronoun, a preposition, etc. – and suchlike words are less often stuttered than content words. On the other hand: If a sentence starts with a content word, some stutterers developed the habit of beginning with a ‘starter word’. This behavior, indeed, facilitates speaking at the onset of the sentence, but because the speaker’s attention is already directed to the following content word in the anticipation of trouble, this content word is probably stuttered then.
Tornick and Bloodstein (1976) conducted an experiment that, in my view, shows how important attention is for the issue of whether stuttering occurs at sentence onset or not: “Fourteen stutterers read, in random order, 20 long and 20 short sentences. The long sentences were constructed by means of additions to the short ones: for example, She learned to swim and She learned to swim in the clear water of the lake. Only the words that the pairs of sentences had in common were compared for occurrence of stuttering. Significantly more stuttering was found on the same words when they served as the initial segments of long sentences than when they stood alone as short sentences.” (quoted from the abstract). Only 35 % of the stuttering events in the long sentences occurred at the onset. The authors assumed that this was the case because of motor planning, or anticipated motor complexity, however, I would simply say that attention was often prematurely directed to the second part of the long sentences.
Interestingly, Levelt (1983, 1995) found that speech errors occurring at the end of a phrase are more likely detected and repaired than speech errors occurring at the beginning of a phrase – and he assumed that, at the beginning of a phrase, more attention is needed for planning and less attention for monitoring remains. I think, both, better detection of speech errors aa well as reduced stuttering (due to fewer invalid error signals) in the back portions of syntactic units is attributable to the same cause: More attention for auditory feedback is available at the end than at the beginning of syntactic units. That does not contradict the assumption that attention is detracted from auditory feedback at the ends of speech units: Slips of the tongue rarely happen at this position (see last footnote).
Long words are more probably stuttered by adults (Brown & Moren, 1942; Quarrington, Conway, & Siegel, 1962 Schlesinger, Melkman. & Levi, 1966; Silverman & Williams, 1967; Soderberg, 1966; Taylor, 1966; Wingate, 1967) and children (Williams, Silverman, & Kools, 1969).
Information load is the predictability of a word in a context: the lower the predictability, the higher the information load. Words with high information load are more frequently stuttered (Quarrington, 1965; Schlesinger et al., 1967; Soderberg, 1967). Rarely occurring words are more probably stuttered than frequently occurring words (Ronson, 1976; Schlesinger, Melkman, & Levi, 1966; Wingate, 1967); however, this effect overlaps with that of information load. See also Bloodstein & Bernstein Ratner (2008) or St. Louis (1979) for an overview.
Recently, Garcia-Barrera and Davidow (2015) have proposed a comprehensive theory describing how anticipations of stuttering arise. They discuss three possibilities. The first two possibilities start from the assumption that the monitoring system plays a central role in this process, because stutterers learn by the self-monitoring of speech, i.e., by hearing, via the external feedback loop, which words or sounds are frequently stuttered. I think that’s wrong. From my own experience as a stutterer, I know that a blockage of speech flow is always firstly perceived by proprioception. The experience of one’s own stuttering is not to hear oneself stutter, but to feel the disability to get the word out. Hence, the perception of one’s own stuttering is completely different from the perception of a self-produced error (slip of the tongue) and has nothing to do with the self-monitoring of speech.
The third possibility they discussed is independent from the self-monitoring of speech, and I deem it the true theory: The repeated experience of stuttering on a certain word or initial sound results in a ‘somatic marker’ (Damasio, 1994; 1995) by which it is coupled with this word or sound. I think the somatic marker is coupled with the speaking program: Every time when the person is going to speak the word or the initial sound, and when its speaking program is activated, the somatic experience that was coupled with the stuttering moments is co-activated.
Most speech errors happening to people despite good language ability seem to be due to confusions of words that start with the same or very similar initial sounds and/or similar first syllables, hence most errors occur at the beginning of words. Therefore, tongue twisters (or jaw breakers) are usually built in the way that many words having the same or similar initial sounds and/or syllables are strung together. Here some examples: