Phonological memory in young children who stutter

Highlights

• Inconsistent nonword repetition performance reported in the literature for CWS.

• Several factors can influence performance on nonword repetition tasks.

• Groups were carefully matched on language, sex, and SES.

• CWS performed significantly less well on nonword repetition tasks than CWNS.

Abstract

This study investigated phonological memory in 5- and 6-year old children who stutter. Participants were 11 children who stutter matched on general language abilities, maternal education level, and sex to 11 typically fluent children. Participants completed norm-referenced nonword repetition and digit span tasks, as well as measures of expressive and receptive vocabulary and articulation. The nonword repetition task included stimuli that ranged from 1 to 7 syllables, while the digit naming task contained number strings containing 2–10 digits. Standardized tests of vocabulary and articulation abilities were tested as well. Groups were comparable on measures expressive vocabulary, receptive vocabulary, and articulation. Despite the fact that the majority of participants scored within typical limits, young children who stutter still performed significantly less well than children who do not stutter on the nonword repetition task. No between-group differences were revealed in the digit naming task. Typically fluent children demonstrated strong correlations between phonological memory tasks and language measures, while children who stutter did not. These findings indicate that young children who stutter may have sub-clinical differences in nonword repetition.

Introduction

Several theories implicate motoric, temperamental, and linguistic differences that may contribute to the disruption of the forward flow of speech in stuttering (Bloodstein & Bernstein Ratner, 2008). In particular, theories involving psycholinguistic abilities suggest that a breakdown or delay may occur during the process of phonological encoding, or the retrieval and construction of the phonological segments of words. According to these theories, breakdowns or delays at the level of phonological encoding may then result in disfluent speech (Howell & Au-Yeung, 2002; Perkins, Kent, & Curlee, 1991; Postma & Kolk, 1993). Many researchers propose that the construction of phonological segments during phonological encoding requires the use of phonological memory, or the ability to maintain phonological and auditory information for short-term retrieval while the entirety of the phonological code is constructed (Acheson and MacDonald, 2009, Alt and Plante, 2006, Bajaj, 2007; Haberlandt, Thomas, Lawrence, & Krohn, 2005). Several authors have also suggested that phonological memory abilities are lower in young children who stutter than in young children who do not stutter (Anderson & Wagovich, 2010; Anderson, Wagovich, & Hall, 2006; Spencer & Weber-Fox, 2014), although the findings are inconsistent (Bakhtiar, Ali, & Sadegh, 2009; Smith, Goffman, Sasisekaran, & Weber-Fox, 2012). A better understanding of phonological memory abilities in young children who stutter will allow for determination of underlying cognitive mechanisms that may be affected in children who stutter.

A prominent model of working memory by Baddeley, 2000, Baddeley, 2003 proposes a four-component memory system that consists of a supervisory component (central executive) and three subservient systems (visuospatial sketchpad, phonological loop, and episodic buffer). The central executive mediates attention and directs resources to the subservient systems that operate as relatively passive stores of information. The visuospatial sketchpad stores visual and spatial information, while the phonological loop stores auditory and speech-based information. The phonological loop is comprised of two additional components: a phonological store and an articulatory rehearsal mechanism. The phonological store temporarily maintains auditory information for short-term retrieval, but is subject to rapid decay after approximately 2 seconds. The content in the phonological loop can be refreshed via silent or overt articulatory rehearsal to allow the phonological code to be recycled and maintained for longer periods of time (Baddeley, 2000, Baddeley and Larsen, 2007). The episodic buffer, a recently added component to the model, provides a link to long-term memory stores (e.g., the lexicon) and integrates the visual and auditory information from the other subservient systems regardless of the input mechanism (Baddeley, 2000). Some researchers argue that access to long-term stores may also help refresh the content in the phonological store by accessing the phonological code found in the lexicon (Dell, Schwartz, Martin, Saffran, & Gagnon, 1997; Hoffman, Jefferies, Ehsan, Jones & Lambon Ralph, 2009; Martin & Gupta, 2004; Martin, Lesch, & Bartha, 1999; Patterson, Graham, & Hodges, 1994; Thorn, Gathercole, & Frankish, 2005). That is, access to the phonological code in pre-existing lexical entries may be used along with silent or overt articulatory rehearsal to help refresh and maintain the content held in the phonological store. This can occur even when attempting to remember nonword stimuli (Coady & Aslin, 2004). Phonological working memory requires input from several aspects of Baddeley’s model, including access to long-term memory stores via the episodic buffer and attentional control via the central executive. Although differences in attentional control have been identified in children who stutter that could influence the processing of the central executive (Anderson, Pellowski, Conture, & Kelly, 2003; Anderson & Wagovich, 2010; Embrechts, Ebben, Franke, & van de Poel, 2000; Karrass et al., 2006), the focus of the current study explores whether inefficient or disrupted phonological memory may lead to difficulty in the maintenance of the phonological code for subsequent use in speech and language planning, thereby contributing to stuttering (e.g., Bajaj, 2007).

Nonword repetition tasks essentially measure the quality of the phonological representations held in working memory. That is, how well a person can maintain and access novel phonological code (i.e., nonwords) from the phonological store (Archibald and Gathercole, 2007, Dollaghan and Campbell, 1998, Gathercole and Baddeley, 1993). Typical tasks of nonword repetition require a participant to listen to and perceive the acoustic signal of the nonword and repeat it back exactly as it was heard. After hearing the nonword, a novel phonological and articulatory plan is assembled while articulatory rehearsal refreshes the signal continuously in the phonological store until the nonword stimuli can be repeated. During this process, the episodic buffer can also access phonological information from pre-existing lexical entries to help refresh decaying phonological code of the nonword during articulatory rehearsal. The more phonological characteristics a nonword shares with a real word (i.e., “word-like” nonwords) the more the lexicon can help support nonword repetition, particularly in young children (Coady and Aslin, 2004, Gathercole, 2006, Gathercole, 2007). Once the nonword is repeated by the participant, it is scored as correct or incorrect. A percentage of correct phonemes can also be calculated (e.g., Anderson et al., 2006; Anderson & Wagovitch, 2010; Hakim & Bernstein Ratner, 2004).

Digit span tasks are also frequently used in phonological memory research and can be used to measure the capacity of a person’s phonological working memory (Jones & Macken, 2015). Capacity is a measure of how much phonological information can be held and accessed from the phonological store before the signal decays beyond retrieval (e.g., Conway, Cowan, Bunting, Therriault, & Minkoff, 2002). Digit span tasks use numbers or other “closed set” stimuli (i.e., stimuli with a limited number of items in a set, such as letters or numbers) that are presented in series of increasing lengths. The participant perceives the auditory signal of the stimuli, stores and rehearses the signal in the phonological loop, and then repeats back what was heard in the exact order it was given.

Phonological memory can be assessed in children as young as 2 when using a modified nonword repetition task (Hoff, Core, & Bridges, 2008; Torrington Eaton, Newman, Bernstein Ratner, & Rowe, 2015). Studies show that this skill continues to develop until approximately age 10 (Chiat, 2006; Gathercole, Service, Hitch, Adams, & Martin, 1999; Snowling & Hulme, 1994). Clear developmental differences exist in children’s phonological memory abilities, with younger children possessing more limited skills than older school-age children. Although matching participants by chronological age is a fairly common practice, empirical evidence suggests that several additional factors influence performance on phonological memory tasks (Dollaghan, Biber, & Campbell, 1995). For example, a strong reciprocal relationship exists between general language and phonological memory abilities that is particularly pronounced in young children 4–6 years old (Gathercole, 2006, Gathercole, 2007). Children with strong language skills tend to have strong phonological memory skills, while children with less robust but still typically developing language systems tend to have poorer phonological memory skills (Gathercole, Service et al., 1999). Socioeconomic status (SES) is another factor that is highly correlated with phonological memory ability (Dollaghan et al., 1999; Engel, Santos, & Gathercole, 2008; McDowell, Lonigan, & Goldstein, 2007). Children from lower SES backgrounds tend to perform less well on nonword repetition tasks than children from higher SES backgrounds. The influence of sex differences in performance on digit naming tasks, a measure of phonological memory capacity, is also evident. Female children tend to perform better than male children on digit naming tasks. These differences are evident in children five years of age yet disappear for older children and adults (Lynn & Irwing, 2008).

Evidence suggests that the linguistic differences reported for children who stutter can be characterized as subtle and “sub-clinical” (Hakim & Bernstein Ratner, 2004; Hall, Wagovich, & Bernstein Ratner, 2007; Ntourou, Conture, & Lipsey, 2011; Pelczarski & Yaruss, 2014; cf. Nippold, 2012), indicating that children who stutter are not presumed to exhibit clinically identifiable disorders in their language and phonological processing abilities. The subtlety of these reported differences require researchers to pay close attention to several participant and task factors that may mask subtle yet potentially meaningful differences in the phonological memory skills of children who stutter.

In addition to the individual factors that influence phonological memory discussed above (general language ability, SES, and sex), the characteristics of the nonword stimuli themselves can influence performance (Cholin, Levelt, & Schiller, 2006; Dollaghan and Campbell, 1998, Dollaghan and Campbell, 2003; Moore, Tompkins, & Dollaghan, 2010; Vitevitch & Luce, 2005). Modifying phonotactic probability (the likelihood of occurrence of a phonological sequence), phonemic complexity (early vs. late developing phonemes; single consonants vs. consonant clusters), the length of the nonword stimuli (short vs. long), and lexical similarity (more or less “word-like”) can make a nonword repetition task easier or more challenging for participants to complete. Accuracy of nonword repetition increases in relationship to the number of features a nonword shares with existing lexical entities (Coady & Aslin, 2004). Thus, nonwords that are less “word-like” (using infrequently heard phonological segments with low phonotactic probability), more phonologically complex (containing later-developing phonemes or consonant clusters), and longer (in number of segments or syllables) are more difficult to produce accurately than more “word-like” nonwords with higher phonotactic probability that are phonologically simple and shorter in length (e.g., Bowey, 2001, Coady and Aslin, 2004, Gathercole, 1995, Gathercole, 2006).

The Comprehensive Test of Phonological Processing (CTOPP) is a standardized test that contains subtests designed to measure phonological memory by manipulating these characteristics of nonwords. The stimuli used in the nonword repetition task of the CTOPP contain less “word-like” stimuli, later-developing phonemes, and gradual increases in length (from one-syllable to seven-syllable nonwords) throughout the task. The result is that the CTOPP contains stimuli that are more challenging than nonword repetition tasks that do not consider these factors. It is particularly important to consider these characteristics of nonwords when examining children of different ages to ensure that task difficulty is not too advanced for younger children yet still advanced enough to allow for a differentiation of subtly different phonological abilities. When these factors are not controlled in the stimuli, results are more difficult to interpret, because any findings could be explained by the characteristics of the stimuli rather than actual differences in children’s phonological abilities.

Several recent studies have provided empirical evidence supporting the claim that phonological memory in children who stutter is significantly different from that seen in children who do not stutter (Anderson and Wagovich, 2010, Anderson et al., 2006; Hakim & Bernstein Ratner, 2004; Oyoun, El Dessouky, Sahar, & Fawzy, 2010; Spencer & Weber-Fox, 2014). Still, the evidence is not conclusive, for other studies have found no significant differences (Bakhtiar et al., 2009, Smith et al., 2012). These discrepancies may be due in part to differences in the age of the participants and the tasks used, as well as general language development factors that can greatly influence performance.

Hakim and Bernstein Ratner (2004) administered the Children’s Nonword Repetition task (CNRep; Gathercole, Willis, Baddeley, & Emslie, 1994) to children ages 4–8 years. They reported generally lower scores for children who stutter overall, with a significant between-group difference at the 3-syllable level. The authors concluded that there were no between-group differences for four- and five-syllable nonwords due to floor effects (i.e., children in both groups found the task too difficult, so error levels were high for all participants). Anderson et al. (2006) also administered the CNRep to 12 children who stutter between the ages of 3;0 and 5;2 matched by age, sex, and SES to 12 children who do not stutter. Children who stutter produced significantly fewer correct productions of two- and three-syllable nonwords than their nonstuttering peers and demonstrated nearly twice as many phoneme errors in 3-syllable nonwords as compared to children who do not stutter. No between-group difference was reported for four- or five-syllable stimuli. Anderson and Wagovich (2010) also reported similar performance on the CNRep in preschool children when investigating 9 children who stutter and 14 children who do not stutter, ages 3;6 to 5;2, with significant differences reported for two- and three-syllable stimuli. The authors also reported significantly more errors overall for children who stutter. Oyoun et al. (2010) investigated nonword repetition in children who stutter ages 5–13. They reported significant differences between children who stutter and children who do not stutter on 2- and 3- syllable nonword stimuli and in a visual memory picture-number task, but they found no differences on other measures of working memory (i.e., digit and letter span tasks).

Spencer and Weber-Fox (2014) conducted a prospective longitudinal study on children aged 3;9 to 5; 8 that explored speech and language factors that may contribute to the persistence or recovery of childhood stuttering. Several standardized tests of speech and language were administered including the Test of Auditory Comprehension of Language, (TACL-3; Carrow-Woolfolk, 1999), Structured Photographic Expressive Language Test (SPELT-3; Dawson, Stout, & Eyer, 2003), Bankson-Bernthal Test of Phonology (BBTOP; Bankson & Bernthal, 1990), Nonword Repetition Task (NRT, Dollaghan & Campbell, 1998), and Test of Auditory Perceptual Skills − Revised (TAPS-R; Gardner, 1985). The authors reported significantly reduced performance on the NRT and BBTOP for the children who persisted in stuttering as compared to the children who later recovered and suggested that articulation and nonword repetition abilities may be helpful in identifying young children who are at greater risk for continuing to stuttering. Authors also reported no significant differences in a digit span task between children who persisted in stuttering and children who do not stutter.

Unlike the findings reported above that found that children who stutter performed less well on nonword repetition tasks than children who do not stutter, some studies did not reveal differences in performance on nonword repetition tasks. Bakhtiar et al. (2009) investigated the phonological memory skills of 5- to 8-year old children who stutter utilizing 2- and 3-syllable nonwords. Lower overall scores for children who stutter were reported; however, no significant differences were found. Use of only 2- and 3- syllable nonwords in the older children may have resulted in a ceiling effect. Thus, the study may not have been able to identify potential differences in phonological memory that become apparent under a larger cognitive processing load. Smith et al. (2012) did not report any differences between 4- and 5-year old children who do and do not stutter with typical language abilities for a number of phonological memory measures that included auditory digit- and word-span tasks, the NRT, and kinematic measures. Only children who stutter who also exhibited concomitant speech or language disorders performed significantly less well than the children with typical language skills on the tasks.

In summary, all but two studies (Bakhtiar et al., 2009, Smith et al., 2012) have reported lower scores for children who stutter, although not all of the differences reached statistical significance, and the patterns of difference vary across the studies. Altogether, the studies discussed above suggest that differences in phonological memory ability exist; however, the evidence is not conclusive. Although the majority of the participants in the studies discussed above included young children (e.g., 4–6) as well children older than 6 years of age, the results of the studies with older children may not be directly compared to studies with only young children who stutter. Additionally, discrepancies in the literature may be due to the influences of specific factors described above (sex, SES, and language abilities), as not every study controlled for all these variables. Further study using carefully matched pairs of children who stutter and children who do not stutter, as well as sufficiently challenging stimuli, is needed to determine if phonological memory is a factor that contributes to stuttering. The current study was designed to address this need by answering the following research questions:

• Are the phonological memory (nonword repetition and digit span) skills of children who stutter different from children who do not stutter?

• Do children who stutter demonstrate the expected strong relationship (Coady & Evans, 2008) between phonological memory and other language measures (i.e. articulation abilities, expressive/receptive vocabulary)?