تاثیر پیش بینی تشخیص اشتباه کلمه بر احتمال لکنت زبان

This study investigates whether the experience of stuttering can result from the speaker's anticipation of his words being misrecognized. Twelve adults who stutter (AWS) repeated single words into what appeared to be an automatic speech-recognition system. Following each iteration of each word, participants provided a self-rating of whether they stuttered on it and the computer then provided feedback implying its correct or incorrect recognition of it. Each word was repeated four times. Unbeknown to participants, ‘Correct’ and ‘Incorrect’ recognition of words by the system was pre-determined and bore no relation to the actual quality of participants’ iterations of those words. For words uttered in the ‘Correct recognition’ condition, the likelihood of AWS self-reporting stuttering on a word diminished across iterations, whereas for words in the ‘Incorrect recognition’ condition it remained static. On the basis of the findings it is argued that: (a) in AWS, the anticipation that a word will be misrecognized increases the relative likelihood of stuttering on that word in the future; and (b) this effect is independent of the degree of difficulty inherent in the formulation and motor execution of the word itself, although it may interact with it. Mechanisms that can account for these findings and yet are also congruent with the wider range of evidence from psycholinguistic and speech motor control domains are discussed. It is concluded that stuttered disfluencies may best be explained as resulting from the inappropriate functioning of covert repair and/or variable release threshold mechanisms in response to the anticipation of communication failure.

People who stutter (PWS) do not stutter all the time. Rather, stuttering moments are more likely to occur on specific words, with specific conversation partners and in specific speaking situations, such as talking over the telephone, before groups, etc. (Bloodstein & Bernstein Ratner, 2008, chap. 10). The exact pattern of their occurrence may, however, vary considerably from one PWS to another, and a different pattern is found in young children who stutter compared to older children and adults (Bloodstein, 2001, Bloodstein and Grossman, 1981, Dworzynski et al., 2004 and Howell et al., 1999). In young children who stutter (CWS), stuttering is most likely to occur on utterances that are linguistically or motorically complex (Bernstein Ratner, 1997, Bloodstein and Grossman, 1981, Logan and Conture, 1997 and Yairi and Ambrose, 2005), in line with the view that the language or speech production systems of young CWS are not yet sufficiently developed to enable them to fluently produce utterances with an age-appropriate level of complexity (e.g., Bernstein Ratner, 1997 and Conture et al., 2004). In older children and adults who stutter (AWS) evidence of impaired language production or speech motor control is more equivocal. Although experimental studies have found that, compared to controls, AWS tend to have slower speech-onset latencies (e.g., Burger and Wijnen, 1999, Lieshout et al., 1996, Sasisekaran and De Nil, 2006 and Tsiamtsiouris and Cairns, 2009), these could simply reflect speakers’ attempts to adapt to the disorder. AWS have been found to make more phonological-encoding and word-order errors, in both inner and overt speech (Brocklehurst & Corley, 2011) and show more variability in fine motor coordination (e.g., Kleinow and Smith, 2000, Loucks et al., 2007 and Max et al., 2003). However, in all such studies there is a large degree of overlap between the stuttering and control participant groups. Thus, it seems likely that, in AWS, stuttering events may sometimes occur even in the absence of any significant ongoing underlying impairment in language formulation or speech motor control (Conture et al., 2004). In the current study we explore the extent to which stuttering-like disfluencies can be precipitated on specific words independently of any formulation or articulation difficulty that production of those words might entail. We describe an experiment designed to test whether the likelihood of stuttering increases when participants produce specific words which they have been led to believe will be difficult (for a speech-recognition system) to recognize. To put the study into perspective, we begin with an overview of two very different theoretical perspectives on the causes of moments of stuttering: Stuttering as a symptom of adaptation to underlying formulation or production impairment, as exemplified by the Covert Repair and EXPLAN Hypotheses (Howell and Au-Yeung, 2002, Kolk and Postma, 1997 and Postma and Kolk, 1993), and stuttering as an anticipatory struggle response, as exemplified by the Anticipatory Struggle Hypothesis (Bloodstein, 1958 and Bloodstein, 1975). 1.1. Stuttering as a symptom of adaptation to underlying impairment Findings from brain imaging research suggest that, as a group, PWS have both structural and functional weaknesses in areas of the brain associated with syllable planning and production (see Watkins, Smith, Davis, & Howell, 2008, for a review). The accumulation of such evidence has stimulated the development of a number of hypotheses that posit that PWS have underlying language or speech production deficits and that stuttered disfluencies arise as the unintended side-effects of their attempts to adapt to those deficits (Civier et al., 2010, Howell and Au-Yeung, 2002, Max et al., 2004, Postma and Kolk, 1993 and Vasić and Wijnen, 2005). Most commonly the adaptations that lead to stuttering are believed to involve overburdened ‘covert error repair’ or ‘restart’ mechanisms which, under more normal conditions, serve to regulate the flow of speech and ensure that it is relatively free of errors, thus helping the speaker to make himself understood and maintain his conversation turn during times of language-formulation difficulty. Perhaps the best known of these hypotheses is the Covert Repair Hypothesis (CRH: Kolk and Postma, 1997 and Postma and Kolk, 1993) which is predicated on the view that speakers audit their inner speech to check their planned utterances for encoding errors (Levelt, 1983 and Levelt, 1989). Because speech planning takes place somewhat in advance of motor execution, if an error is detected in inner speech, the speaker may have time to stop and reformulate the plan, and thus repair the error before starting to speak. The CRH accounts for the different symptoms of stuttering (whole and part-word repetitions, prolongations and blocks) by postulating that these are the overt symptoms of covert repairs that have been only partially successful because there was insufficient time to repair the error. Thus if cancellation of the erroneous speech plan occurs just as the first phoneme is about to be uttered, a silent pause, or ‘block’, may result while the speaker reformulates it, whereas if cancellation occurs after the first phoneme, syllable or word has already been uttered, a (phoneme, syllable or word) repetition may result, and if this happens several times in a row, then multiple repetitions may occur. More recently a similar mechanism, involving error detection and ‘motor resets’, has been postulated to operate at the level of speech motor control (Civier et al., 2010 and Max et al., 2004), and an alternative, threshold-based mechanism whereby stuttered disfluencies arise in response to speakers’ attempts to execute speech-plans which are simply incomplete or insufficiently activated, rather than containing actual errors, has been posited – in the EXPLAN hypothesis (Howell, 2003, Howell, 2011 and Howell and Au-Yeung, 2002). Such mechanisms provide plausible explanations for the variety of stuttering-like disfluencies that occur in both PWS as well as in normally fluent speakers. They also provide compelling explanations for why the likelihood of stuttering tends to decrease on subsequent iterations of previously spoken words (the ‘adaptation effect’; Brutten and Dancer, 1980 and Johnson and Knott, 1937);1 why PWS are particularly likely to stutter on word onsets; why the likelihood of stuttering occurring on a word is strongly influenced by its grammatical function (Bloodstein, 2006 and Howell and Sackin, 2001), length, position in the sentence, frequency and predictability (Brown, 1937, Brown, 1945 and Newman and Bernstein Ratner, 2007); for why stuttering is more common on utterances that are longer and/or more complex (Logan and Conture, 1995, Logan and Conture, 1997 and Newman and Bernstein Ratner, 2007); and for why young children whose language and articulation skills lag behind those of their peers may be more likely to stutter (Bernstein Ratner, 1997; however, cf. Nippold, 1990 and Nippold, 2001). However, adaptation hypotheses, such as the CRH, EXPLAN and their motor-control equivalents, are less successful at accounting for other observations in relation to the distributions of stuttering events in older children and adults. In particular, the Covert Repair Hypothesis fails to account for the lack of any discernible correlation between the frequency with which AWS produce inner-speech errors and their stuttering severity in everyday speaking situations (Brocklehurst & Corley, 2011), and more generally, adaptation hypotheses fail to account for why older children and adults frequently stutter on isolated, commonly occurring single words; why they have particular difficulty uttering their names; why they are influenced so strongly by the characteristics of the listener and the overall dynamics of the speaking situation, and in particular, why they appear to be able to speak complex utterances perfectly fluently when there is no listener present; and why some speakers with severe language or speech production disorders do not stutter. (see Bloodstein & Bernstein Ratner, 2008, chap. 10 for an extensive review of such observations). Thus it appears that, although compensatory responses to underlying difficulties in language production or speech motor control may plausibly account for the stuttering-like disfluencies of young children, they cannot fully account for the persistence of stuttering in older children and adults. An alternative possibility investigated in the current study is that, in adults, stuttering-like disfluencies occur as a side-effect of compensatory responses (of one type or another) to the anticipation of difficulty, and such anticipation may stem from memories of having experienced difficulty speaking or communicating in similar situations in the past (cf. Conture et al., 2004). 1.2. Stuttering as an anticipatory struggle response The term ‘anticipatory struggle’ was first used by Bloodstein in the 1950s to describe a broad category of hypotheses, all of which share the idea that PWS believe that speaking is difficult and this belief in some way interferes with the smooth running of the processes that underpin fluent speech (see Bloodstein & Bernstein Ratner, 2008, chap. 2, for a review). Anticipatory struggle hypotheses have proposed a variety of mechanisms to account for how the anticipation of stuttering can lead to the production of stuttering-like disfluencies, including ‘approach-avoidance conflict’ (Sheehan, 1953); abnormal ‘preparatory sets’ (Van Riper, 1973), and ‘tension and fragmentation’ (Bloodstein, 1975). Central to Bloodstein's own (1975) ‘Anticipatory Struggle Hypothesis’ is the notion that the primary symptoms of stuttering (repetitions, prolongations and blocks) are essentially tensions and fragmentations in speech, which arise in response to stimuli representative of past speech failure, and which originally arose in response to the experience of difficulty with speech, language, and/or communication in early childhood. Tension and fragmentation are regarded as the symptoms of “trying too hard,” and “taking the activity apart to do it piece by piece” (Bloodstein, 1975, p. 4) that characteristically occur when an individual wishes to execute a complex motor activity and yet doubts that he will be successful. By conceptualizing stuttering in this way, Bloodstein's (1975) Anticipatory Struggle Hypothesis provides a plausible explanation for why stuttering is more likely to occur in children with impaired or delayed development of linguistic skills and/or speech motor control. Importantly, unlike adaptation hypotheses, Bloodstein's hypothesis does not attribute stuttering directly to the speaker's attempts to overcome current instances of production difficulty. Instead, it posits that stuttering arises in response to the belief that, in particular situations, particular sounds or words will be difficult to speak. Thus it allows for the possibility that the exact nature of the impairment or delay that underlies that belief may differ from child to child, and in some individuals, the impairment or delay that originally caused the belief to become established may no longer be present. Thus, Bloodstein's hypothesis also provides a parsimonious explanation for how stuttering may persist even after any language or speech impairment/delay has resolved, by postulating that a vicious circle is established whereby the anticipation itself precipitates the struggle that was anticipated. Further, because it identifies stuttering as a disorder of communication in which the responses of the listener are every bit as important as the speech of the speaker, it provides a seemingly parsimonious explanation for a range of common observations in relation to stuttering, including why PWS rarely have difficulty speaking to themselves or when they do not care what the listener thinks of them or what they say; and conversely, why they may find it so much more difficult to speak fluently to certain people, about certain topics and in certain social situations (Bloodstein, 1949, Bloodstein, 1950a and Bloodstein, 1950b). However, despite its appeal, Bloodstein's (1975) Anticipatory Struggle Hypothesis has two important weaknesses. Firstly, ‘tension and fragmentation’ is not well specified and fails to provide an adequate explanation for precisely why stuttering-like disfluencies manifest in the variety of ways that they do (as repetitions, prolongations or blocks). In comparison, the more recent psycholinguistic hypotheses, outlined above, are much more successful. And secondly, although the notion of anticipatory struggle provides a parsimonious explanation for the observational data and self-reports regarding the moments when stuttering occurs, it has proved particularly difficult to test experimentally. 1.3. The current study The current study constitutes an experimental investigation of the influence of anticipation on the likelihood of stuttering. Specifically, it investigates whether the experience of stuttering can be precipitated on specific words by instilling, in the speaker, the anticipation that those words will not be recognized by the recipient. Our experiment was loosely based on a paradigm developed by Hansen (1955), originally designed to test the effects of different valences of audience response on stuttering severity. In Hansen's experiment, participants who stutter performed a variety of reading and photograph description tasks in front of an audience ranging from 12 to 25 people. The lighting was turned down so participants could not see the audience's faces. Positive or negative audience feedback was delivered to the speakers indirectly, by means of a series of green and red lights and corresponding counters, located on a table in front of the speaker. The speaker was led to believe that feedback was controlled by the audience, whereas in reality, it was manipulated by the experimenter. Hansen found that, although overall there was a general decrease in stuttering over the duration of the experiment, the rate of decrease was greater where positive feedback was delivered than where negative feedback was delivered. These trends became noticeable after a short time lag, and were most noticeable during spontaneous speech when it was easier for the speaker to focus on the feedback. In our experiment, instead of speaking to an audience, participants who stutter spoke single words into what they believed was speech-recognition software on a computer, and received automatic online feedback indicating whether or not those words had been correctly recognized. We designed the paradigm in this way because we specifically wanted to investigate the effect of anticipation of word misrecognition (rather than anticipation of a negative listener response). To avoid any possibility that participants’ performances might be affected by the fear of negative evaluation by potential listeners or over-hearers, participants provided their own self-reports of stuttering and were led to believe that they were not being recorded, that nobody was listening to them or able to hear them speak, and that the speech-recognition process was entirely automatic. As in Hansen's (1955) experiment, feedback was, in reality, predetermined, and bore no relationship to the accuracy or fluency with which participants spoke. Participants were prompted to utter each target word four times, receiving feedback after each attempt. Across the four iterations, the feedback consistently indicated either correct or incorrect recognition of the target word. Thus, participants could predict with increasing confidence whether or not the remaining iterations of the target word were likely to be correctly or incorrectly recognized by the software. We hypothesized that, due to the ‘adaptation effect’ (Brutten and Dancer, 1980 and Johnson and Knott, 1937), there would be an underlying trend for self-reports of stuttering to decrease across iterations. If stuttered disfluencies result solely from an underlying language or speech production impairment, this reduction would be unaffected by whether the software apparently failed, or succeeded, to recognize each word spoken, since lexical difficulty was held constant across conditions. Evidence that the word-recognition feedback appearing on the computer screen affected participants’ performance on subsequent iterations of the same word would, however, implicate an additional process. If that process is related to the anticipation of a struggle to articulate words sufficiently well for them to be recognized, then regardless of whether or not participants who stuttered had underlying production deficits, participants should be relatively more likely to produce stuttering-like disfluencies in the condition where the software apparently failed to recognize their productions of a particular word.

Adults who stutter are more likely to stutter on single words when the speaking circumstances lead them to anticipate communication failure on that word. This effect is independent of the degree of difficulty inherent in the production of the word itself, although it may interact with it. This finding is incompatible with hypotheses that posit that stuttering in adults occurs as a direct result of language or speech production difficulty alone. The findings are most consistent with the hypothesis that persistent stuttering is characterized by inappropriate functioning of covert repair and/or variable release threshold mechanisms that, under more normal circumstances, may serve to ensure the speaker achieves a high level of phonetic accuracy in situations where he believes he is likely to be misheard or misunderstood.