انگیختگی آواشناسی در بزرگسالانی که لکنت زبان دارند

The purpose of this study was to compare the speed of phonological encoding between adults who stutter (AWS) and adults who do not stutter (ANS). Fifteen male AWS and 15 age- and gender-matched ANS participated in the study. Speech onset latency was obtained for both groups and stuttering frequency was calculated for AWS during three phonological priming tasks: (1) heterogeneous, during which the participants’ single-word verbal responses differed phonemically; (2) C-homogeneous, during which the participants’ response words shared the initial consonant; and (3) CV-homogeneous, during which the participants’ response words shared the initial consonant and vowel. Response words containing the same C and CV patterns in the two homogeneous conditions served as phonological primes for one another, while the response words in the heterogeneous condition did not. During each task, the participants produced a verbal response after being visually presented with a semantically related cue word, with cue-response pairs being learned beforehand. The data showed that AWS had significantly longer speech onset latency when compared to ANS in all priming conditions, priming had a facilitating effect on word retrieval for both groups, and there was no significant change in stuttering frequency across the conditions for AWS. This suggests that phonological encoding may play no role, or only a minor role, in stuttering. Educational objectives: The reader will be able to: (1) describe previous research paradigms that have been used to assess phonological encoding in adults and children who stutter; (2) explain performance similarities and differences between adults who do and do not stutter during various phonological priming conditions; (3) compare the present findings to past research that examined the relationship between phonological encoding and stuttering.

The American Speech-Language-Hearing Association (ASHA, 1999) provides a detailed and widely accepted definition of stuttering, which identifies physiological, psychological, motor, linguistic, and/or auditory factors as possibly contributing to its occurrence. Despite extensive research, the impact of these factors on stuttering has yet to be fully explored and understood. Although many explanations have been proposed in relation to its causes, none of the existing theories provides a full account of all aspects of the disorder, which leaves its underlying mechanisms unknown. Various theories of stuttering relate this disorder to deficits in linguistic processes, such as semantic encoding (e.g., Bosshardt et al., 2002 and Hennessey et al., 2008), grammatical encoding (e.g., Prins, Main, & Wampler, 1997), and phonological encoding (PE) (e.g., Burger and Wijnen, 1999, Hennessey et al., 2008 and Sasisekaran and De Nil, 2006). The present study investigated a possible relationship between PE and stuttering, and is therefore concerned with one of the linguistic theories, the Covert Repair Hypothesis (CRH) (Postma & Kolk, 1993). Postma and Kolk (1993) used Levelt's (1989) model of speech production and Dell's (1986) connectionist model of PE in postulating the CRH, which suggested that stuttering is caused by a disturbance in the formulating component of the speech production system. More specifically, Postma and Kolk (1993) stated that individuals who stutter are slower in PE than individuals who do not stutter. The hypothesis suggests that phonetic plans cannot be completed and sent to the articulatory buffer as quickly as in fluent speakers. This delay is detected by the internal feedback loop as an error that needs correction, and, as a consequence, covert repairs occur. At the same time, the speech articulator overtly compensates for this delay by manifesting different stuttering symptoms including repetitions, audible prolongations, and inaudible prolongations (blocks). If the error is found in the coda of a syllable (e.g., /t/ in cat), PE will continue until the correct phoneme is chosen, which will result in overt repetitions of the onset (i.e., initial sound) and nucleus (i.e., vowel) of the syllable (/kæ-kæ-kæ/). If there is an error in the sound following a continuant (e.g., /Λ/ in luck), the continuant is audibly prolonged until the phoneme is encoded. Errors in the onset of words or syllables will result in inaudible prolongations. Speech flow will be interrupted for repairs, but the speaker may begin moving the articulators and building up muscle tension, which is recognized as a block. In any of these events, overt speech will contain the primary symptoms of stuttering. Postma and Kolk (1993) view these as adaptation symptoms that represent the articulator's attempts to cope with the impaired system, but which creates an asynchrony between the phonological and the motor systems. To support this notion, the fluency enhancing effect of a slower speech rate on stuttering is used, which, according to Postma and Kolk (1993), comes from the greater amount of time that is provided for PE. By decreasing the speech rate, the articulator works more slowly and is synchronized with the phonological encoder. There is a growing body of literature that has examined PE in individuals who stutter. Through use of various methodologies, PE has been investigated in both adults who stutter (AWS) and children who stutter (CWS), with each producing variable findings. 1.1. Studies on PE in stuttering Studies that focused on PE in stuttering have used a variety of paradigms, including phonological priming, nonword repetition, phonological (rhyme and phoneme) monitoring, as well as several others. Phonological priming is the paradigm most frequently used to examine the relationship between PE and stuttered speech. Wijnen and Boers (1994) and Burger and Wijnen (1999) used a phonological priming task, which was developed by Meyer, 1990 and Meyer, 1991. This task requires that participants produce a target word after being visually presented with a semantically related cue word, with cue-target word pairs being learned beforehand. Two conditions are created: the homogeneous condition, in which the target words share either the onset (C-prime) or the onset and nucleus (CV-prime), and the heterogeneous condition, in which the target words are phonologically unrelated, that is, they do not share an onset. It is assumed that phonologically related words will prime one another and that the priming effect will increase as the number of shared phonemes increases (Wijnen & Boers, 1994). Therefore, speech onset latencies (SOL) in homogeneous conditions should be shorter than in heterogeneous conditions, and SOL should be shorter in CV-prime when compared to C-prime. Since the CRH suggested that PE is slower in individuals who stutter, and stuttering usually occurs at the beginning of a word (on the first syllable), the predictions were that (a) a C-prime would not decrease SOL in individuals who stutter because the PE of the remaining segments in the syllable could be slower, and (b) a CV-prime should alleviate the PE problem, at least with regard to the initial syllable, which would result in reduced SOL (Wijnen & Boers, 1994). Whether phonological priming is associated with PE or phonetic planning or a combination of both is debatable. However, as postulated by the CRH, stuttering occurs due to the covert repair of errors caused by slower PE but which remain undetected until the phonetic plan is inspected by the internal feedback loop. It may be inferred that, regardless of which of the two processes are facilitated by priming, the presumed culprit of the problem (i.e., the delay in the formation of the phonetic plan) should be eradicated. Wijnen and Boers (1994) reported that adults who do not stutter (ANS) produced significantly different SOL in all three conditions, while AWS produced significantly shorter SOL in the CV- but not in the C-homogeneous condition when compared to the heterogeneous condition. This means to say that the two groups responded differently to phonological priming, as confirmed by a significant group by condition interaction, which was interpreted as supportive of the CRH. Burger and Wijnen (1999), however, found that the priming effect was the same for both groups: the participants produced shorter SOL in homogeneous than heterogeneous conditions, with the CV-homogeneous condition priming significantly better than the C-homogeneous condition. Although AWS were generally slower than ANS in their responses, since the two groups responded to priming in the same way, the findings were interpreted as not supportive of the CRH. Additional priming studies have used picture naming tasks with phonological primes presented auditorily (Byrd et al., 2007, Hennessey et al., 2008 and Melnick et al., 2003). Melnick et al. (2003) measured the speech response times of CWS and children who do not stutter (CNS) during three types of phonological priming: no prime, related prime, and unrelated prime. In the no prime condition, children were asked to name pictures as quickly as possible following their presentation. In the related prime and unrelated prime conditions, responses were auditorily primed by the initial syllable of the target word and a syllable unrelated to the target response. In both groups, the related prime condition produced significantly shorter response times than the no prime condition. A lack of significant between-group differences in any of the conditions and the absence of significant interactions did not provide support for the CRH. Byrd et al. (2007) investigated the effects of neutral, holistic, and incremental auditory priming on picture naming in three- and five-year-old CWS and CNS. During the neutral auditory, incremental, and holistic priming conditions, the participants were presented with the following cues: a pure tone prior to picture presentation, the onset of the target word prior to picture presentation, and the entire word without the onset prior to picture presentation. To compare the holistic and incremental conditions, the neutral SOL was subtracted from the holistic SOL and from the incremental SOL. Three-year-old children were faster in picture naming during the holistic than during the incremental priming condition. While CNS became faster during the incremental when compared to the holistic priming condition as they reached the age of five years, CWS did not. Byrd et al. (2007) suggested that PE development may be delayed in CWS. Hennessey et al. (2008) used semantically related, phonologically related, and unrelated auditory primes during picture naming to investigate semantic and PE in AWS. They also measured response times during word-nonword comparison tasks to examine phonetic planning in AWS. The findings resembled those previously reported by Burger and Wijnen (1999): overall AWS responded slower than their controls, but there was no significant group by condition interaction. This suggested that they did not significantly differ from ANS in semantic, phonological, and phonetic encoding skills, thereby providing no evidence of a causal relationship between these processes and stuttering. Nonword repetition involves the presentation of novel phonological sequences, which are then temporarily stored in the memory and recalled upon request. This task is often used for assessing the phonological working memory (e.g., Anderson, Wagovich, & Hall, 2006), and, as such, it has also been utilized to explore PE in stuttering. Hakim and Bernstein Ratner (2004), Anderson et al. (2006), Bakhtiar, Ali, and Sadegh (2007), and Anderson and Wagovich (2010) used nonword repetition to study the phonological abilities of CWS who were matched with CNS on various standardized measures of language, articulation, intelligence, and/or short-term memory. Hakim and Bernstein Ratner (2004) reported that CWS were less accurate than CNS in two-, three-, four-, and five-syllable nonword repetition and that they made more phoneme errors than CNS; however, the difference reached significance only for three-syllable stimuli. Similarly, Anderson et al. (2006) found significantly more incorrect repetitions of two- and three-syllable nonwords and more phoneme errors in CWS than in CNS, with Anderson and Wagovich (2010) confirming the former. These outcomes were attributed to a phonological processing deficit (Hakim & Bernstein Ratner, 2004) and weaker phonological working memory (Anderson et al., 2006). Contrary to these studies, Bakhtiar et al. (2007) did not find significant differences between CWS and CNS during repetition of two-, three-, four-, and five-syllable nonwords. More recently, Smith, Sadagopan, Walsh, and Weber-Fox (2010) examined nonword repetition abilities in adults. Along with analyzing behavioral accuracy scores for nonwords produced in isolation and in a carrier phrase, kinematic data regarding the consistency of articulatory coordination were investigated as well. Behaviorally, AWS did not differ from ANS. However, AWS had higher lip aperture variability than ANS, and this difference was more pronounced for longer, phonologically more complex nonwords. In addition, even when this variability decreased with practice, it did not reach the stability levels of ANS. Therefore, while the behavioral data did not provide evidence of a phonological deficit in AWS, the kinematic findings did. Rhyme monitoring relies primarily on the PE of visually or auditorily presented cue-target word pairs. The processing of phonological information is a prerequisite for comparing word endings and for making rhyme decisions. Bosshardt and Fransen (1996) and Bosshardt et al. (2002) used rhyme and category recognition tasks to measure the speed of phonological and semantic encoding in AWS. AWS were slower than ANS in retrieving semantic but not phonological information when making category and rhyme decisions for target words embedded in silently read sentences (Bosshardt & Fransen, 1996). Furthermore, AWS did not differ from ANS in the accuracy and speed of PE across the conditions when making rhyme decisions for word pairs with and without sentence generation as a concurrent task. These findings suggested that AWS do not have PE deficits (Bosshardt et al., 2002). In addition, Weber-Fox, Spencer, Spruill, and Smith (2004) examined the phonological processes of AWS by recording their event-related brain potentials, response accuracy, and manual response times during a silent rhyme judgment task. Stimuli were visually presented cue-target word pairs, which were either congruent (similar in orthography and rhyming or different in orthography and not rhyming) or incongruent (similar in orthography, but not rhyming or different in orthography, but similar in rhyming). The latter condition required more cognitive processing than the former. Results showed that AWS had significantly slower response times than ANS in the incongruent condition. No other between-group differences were observed. The authors concluded that it is the greater difficulty in cognitive processing and not phonological processing that may be an underlying component of stuttering. Similar to rhyme monitoring, phoneme monitoring requires the PE of presented stimuli. Sasisekaran and De Nil (2006) and Sasisekaran, De Nil, Smyth, and Johnson (2006) used a silent phoneme monitoring task, during which the participants noted the presence or the absence of a certain phoneme in the names of presented pictures. In both studies, AWS were significantly slower in phoneme monitoring than ANS. No significant between-group differences were found for phoneme monitoring of auditorily presented words (Sasisekaran & De Nil, 2006) or for simple manual response speed, auditory monitoring, and overt picture naming (Sasisekaran et al., 2006). These findings were interpreted as indicative of delayed PE in AWS. In addition to the above described methodologies, other paradigms have been used to investigate PE in stuttering. These include the analysis of dysfluency clusters in spontaneous speech (LaSalle & Conture, 1995), the analysis of covertly and overtly produced tongue-twisters (Brocklehurst & Corley, 2011), repetition priming of early and late acquired words (Anderson, 2008), and picture naming of phonologically dense and sparse words (Arnold et al., 2005, Bernstein Ratner et al., 2009 and Newman and Bernstein Ratner, 2007). LaSalle and Conture (1995) examined the frequency, type, and possible causes of stuttering moments in between- and within-word dysfluency clusters in the spontaneous speech of CWS and CNS. The purpose of their study was to investigate whether stuttering moments result from covert repairs of errors that have been detected in the phonetic plan. Based on the location of stuttering within a dysfluency cluster, inferences were made about the type of repair (overt or covert) by which it was precipitated. The analysis of the position of stuttering within dysfluency clusters suggested that stuttering moments were just as likely to occur on any word in the cluster as they were on either overt or covert errors (i.e., overt or covert repairs), which was interpreted as lending support to the CRH (LaSalle & Conture, 1995). Brocklehurst and Corley (2011) investigated PE in AWS by analyzing the error count during covert and overt tongue-twister productions. AWS produced significantly more errors than ANS in both conditions, but there was no correlation between the number of covert errors and the number of stuttering moments. Therefore, while a PE deficiency may be one of the many features of stuttering, there was no evidence of PE errors directly precipitating overt stuttering. Anderson (2008) measured the effect of repetition priming and word acquisition age on picture naming speed in preschool CWS. Both variables presumably influence the connection between semantic and phonological word representations. CWS and CNS could not be distinguished based on their naming latencies. However, repetition priming of late acquired words significantly improved naming accuracy in CWS. This effect was not detected in CNS, who were at or near ceiling performance, possibly due to having stronger semantic-phonological connections than CWS. Arnold et al. (2005) used phonologically dense and phonologically sparse words in a picture naming task to compare PE between preschool CWS and CNS. Phonologically dense words have many phonological neighbors (i.e., words phonologically different in only one phoneme), while phonologically sparse words have few phonological neighbors. The results indicated that preschool children produce phonologically dense words slower and less accurately than phonologically sparse words. There were no significant differences between the two groups, which suggested that CWS and CNS do not differ in their PE abilities. The effect of phonological neighborhood density on the speed and accuracy of picture naming was further investigated in CWS by Bernstein Ratner et al. (2009) and in AWS by Newman and Bernstein Ratner (2007). Similar to Arnold et al.’s (2005) findings, no significant between-group differences were found for either children or adults, suggesting that a causal relationship between PE and stuttering is unlikely. 1.2. The present research The previous research has attempted to examine whether slower and/or deficient PE may be one of the underlying mechanisms of stuttering. The inconsistency of reported findings is not a unique occurrence: investigations of any component involved in speech-language production have yet to produce uniform results. While the heterogeneity of the stuttering population (e.g., differences in stuttering onset, symptomatology, severity) may contribute to such varied findings, methodological differences among research studies might well be an additional problem. Because different research paradigms and participants of varying age ranges have been used to assess a speech-language component of interest, adequate comparisons of obtained results are not always possible. Despite or perhaps because of the contradictory reports, the possibility of a relationship between PE and stuttering cannot be ruled out and warrants further investigation. The main goal of the present study was to obtain additional data regarding PE in AWS and to examine the predictions of the CRH with this population. For this purpose, the phonological priming task described in Wijnen and Boers (1994) and Burger and Wijnen (1999) was modified (see Section 2.2) and used. The reasons for choosing this particular paradigm were three-fold. First, it has been suggested that deficient or, as specified by the CRH, slower PE underlies stuttering, and some evidence supporting this notion is available (e.g., Sasisekaran and De Nil, 2006 and Wijnen and Boers, 1994). A phonological priming task that facilitates word encoding by providing initial phoneme(s) should result in SOL that decreases as the length of the provided phoneme string increases. If PE is slower or somehow deficient in stuttering, the pattern of SOL changes in AWS should differ from the pattern of SOL changes in ANS when both types of participants are under the same phonological priming conditions (e.g., Burger and Wijnen, 1999 and Hennessey et al., 2008). Second, if stuttering dysfluencies occur due to slower PE and subsequent covert repairs (CRH, Postma & Kolk, 1993), phonological priming conditions, especially those that provide a longer phoneme string, should decrease stuttering frequency in AWS (CRH, Postma & Kolk, 1993). And third, by using an already existing paradigm it was possible to compare the present findings to the data reported by earlier, methodologically similar studies (Burger and Wijnen, 1999, Hennessey et al., 2008 and Wijnen and Boers, 1994). To summarize, the phonological priming task used in the present study allows for both response speed and stuttering frequency analyses as well as direct comparison with earlier studies, while the other previously used paradigms do not. For example, all earlier methodologies would have enabled collection of data related to response speed under various conditions; however, some designs (a) did not use overt speech, thus eliminating the possibility of stuttering frequency analysis (e.g., Bosshardt and Fransen, 1996, Sasisekaran and De Nil, 2006, Sasisekaran et al., 2006 and Weber-Fox et al., 2004), (b) assessed CWS (e.g., Arnold et al., 2005, Bakhtiar et al., 2007 and Byrd et al., 2007), thus eliminating the possibility of comparing the present data to earlier data obtained for AWS, and/or c) used auditorily presented primes (e.g., Hennessey et al., 2008 and Melnick et al., 2003), thus introducing a possible confounding effect of central auditory processing deficits on stuttering (e.g., Blood, 1996, Blood and Blood, 1984 and Cimorell-Strong et al., 1983). Considering the postulates of the CRH and the existing research, it was hypothesized that if slower or deficient PE underlies stuttering, (a) AWS would have longer SOL than ANS in the heterogeneous and C-homogeneous conditions, but not in the CV-homogeneous condition, (b) the priming effects of the C- and CV-homogeneous conditions would be different for AWS when compared to ANS, and (c) stuttering frequency would decrease in the CV-homogeneous, but not the C-homogeneous condition for AWS. These findings would not only confirm that PE may be deficient in AWS, but they would lend additional support to the CRH.

The three main findings of this investigation can be summarized as follows: (a) AWS produced significantly longer SOL values than ANS overall, (b) the phonological priming effect on SOL was similar for AWS and ANS, and (c) stuttering frequency did not decrease as the length of the provided phoneme sequence increased. When these findings are viewed in relation to the CRH, it can be concluded that the present data provide no evidence to support the hypothesis. When compared to earlier studies that used phonological priming to examine PE in AWS, the present findings support Burger and Wijnen (1999) and Hennessey et al. (2008), but not Wijnen and Boers (1994).