Abstract
Speech sound disorders (SSDs) are characterized by delays in the accurate production of age-appropriate speech sounds (Lewis et al., 2015). Accurate speech production relies on emerging skills in multiple domains—perceptual, cognitive, linguistic, and motoric. An impairment in any one of these domains could result in abnormal speech sound development. For approximately 3.9% of children, abnormal speech sound production persists past the age of 8 years (Lewis et al., 2015; Wren, McLeod, White, Miller, & Roulstone, 2013). The causal mechanisms associated with persistent SSD (P-SSD) are unknown (Munson, Baylis, Krause, & Yim, 2010). There is empirical data that show working memory (WM) has a critical role on speech learning in young children (Adams & Gathercole, 1995; Couture & McCauley, 2000).
In models of WM, there is a cognitive function in which auditory and/or visual information are temporarily stored and manipulated (Baddeley, Gathercole, & Papagno, 1998; Gathercole & Baddeley, 1990; Gathercole, Willis, Baddeley, & Emslie, 1994). WM contributes to one’s reading, word learning, acquiring language, mathematical processing, and reasoning (Gathercole, Alloway, Willis, & Adams, 2006). For speech production, WM is necessary to accurately and consistently store sounds and readily and appropriately retrieve them (Oakhill & Kyle, 2000). If the WM storage and retrieval system is deficient, it is likely to manifest as a phonological deficit (e.g., SSD, dyslexia, or both).
Baddeley and Hitch (1974) proposed a model of WM with three empirically supported components: phonological loop, visual–spatial sketchpad, and central executive. The phonological loop serves as a store for auditory information, (e.g., speech). The visual–spatial sketchpad stores visually presented information (e.g., pictures). The primary responsibility of the central executive component is to allocate attention resources to either the phonological loop or visual–spatial sketchpad (Baddeley, 1992). The phonological loop consists of two subcomponents—a phonological store and an articulatory rehearsal mechanism. The phonological store is where information is organized based on similar features. The articulatory rehearsal mechanism allows information in the phonological store to be refreshed through subvocal articulation to avoid decay (Baddeley, 2007). The authors of this article suggest that it is possible that for children with P-SSD, the articulatory rehearsal mechanism is negatively affected by their difficulty producing speech. As they get older and their SSD persists, the lack of maturation of the speech production system contributes negatively to their acquisition of strong phonological representations. This can negatively affect a child’s ability to achieve academic success with reading, writing, and spelling (Sutherland & Gillon, 2005).
The goal of this study was to examine WM in school-age children with a P-SSD compared to their typically developing peers. The authors used Baddeley’s Working Memory model (Baddeley & Hitch, 1974) to examine the three primary components of WM—phonological loop, visual–spatial sketchpad, and central executive—in older children with a P-SSD compared to their age-matched typically developing peers.
The Study
Participants
Participants were 40 children in second- through fifth-grade (aged 7.5–11.8 years). The sample of 40 was composed of two groups: participants with P-SSD and participants who were typically developing. Children with SSD were not required to be currently receiving treatment. This was primarily because many older children with SSD can be dismissed from treatment before their disorder is completely resolved (Raitano, Pennington, Tunick, Boada, & Shriberg, 2004) or they no longer qualify for services because they have not made adequate progress (Preston & Edwards, 2007).
Experimental Tasks
Three experimental tasks were administered to represent each of the three constructs of WM within the Baddeley and Hitch (1974) model: phonological loop, visual–spatial sketchpad, and central executive skills.
Phonological loop task
Nonword repetition has been used extensively in research as a metric of phonological WM. In this study, phonological WM was measured by a serial recall nonword repetition task administered on a laptop. The stimuli consisted of consonant–vowel–consonant nonwords that followed the phonotactics of the English language, but carried no meaning.
Participants listened to nonword lists increasing in length: 4 one-nonword lists, 4 two-nonword lists, 4 three-nonword lists, and 4 four-nonword lists. After listening to each list, the participants saw a smiley face on the screen, indicating that they were to repeat the nonwords back in serial order. Percentage of consonants correct (PCC) was calculated for the words recalled in accurate serial order. Words that were not recalled in serial order were not counted in the PCC or the analyses. Care was taken to control for the speech sound production errors in the P-SSD group. For each child, their consistent pattern of speech errors was determined (e.g., /θ/ substituted for /s/ in all contexts), and those specific errors were marked as correct. That is, for the child who repeated the nonword /hαs/ as /hαθ/, the production was scored as correct if produced in the correct serial order (e.g., PCC would be calculated as 100%—two correct consonants produced out of two opportunities). This procedure greatly reduced the possibility that production errors were solely responsible for reduced phonological memory in the group with a P-SSD.
Visual–spatial sketchpad task
Visual–spatial skills were tested using the Spatial Relations subtest from the Woodcock–Johnson III Tests of Cognitive Abilities (Woodcock, McGrew, & Mater, 2001). This subtest is reported to measure “manipulation of visual images in space” in which the participant is required to manipulate objects “in the imagination of the ‘mind’s eye’” (p. 7). The Spatial Relations task required participants to examine four pieces of a puzzle and decide which two or three combined to form the intended complete shape. The task increases in complexity as the shapes are rotated and become more similar. Participants were required to hold visual representations of the small shapes in WM while examining how those shapes may combine differentially to form the larger complete shape.
Central executive task
Central executive function was assessed using a stop signal inhibition task (Gray, Hogan, Alt, Cowan, & Greene, 2011–2016). The task was set within a child-friendly “pirate game” in which monsters invaded the pirate’s island. Participants saw a variety of monsters appear singularly upon the computer screen. Participants were instructed to press the space bar on the computer each time they saw a monster flash on the screen (i.e., go trial) unless they heard an auditory signal at the same time they saw the monster (i.e., stop trial). On the stop trials, the stop signal was presented simultaneously with the visual stimulus. Participants were required to inhibit the natural response to press the space bar. This task is designed to measure inhibitory control, which is a function of the central executive portion of WM.
Results
Performance decreased significantly for all children as the length of nonwords increased.
In reading the nonwords, children with a P-SSD performed less well than children who were typically developing, albeit in the same pattern.
The groups significantly differed when they were to recall two and four nonwords in serial order.
There was a significant and positive relation between nonword repetition and nonverbal intelligence.
The children with a P-SSD did not differ significantly from the typical children on the visual–spatial and central executive tasks.
Discussion
The children with P-SSD in the study had poor phonological WM, but the relation between phonological WM ability and the presence of a P-SSD was mediated by nonverbal intelligence scores. Two recent studies have corroborated the finding that children with P-SSD often have lower nonverbal intelligence (Cabbage, Farquharson, & Hogan, 2015; Lewis et al., 2015). The authors of this project propose three possible explanations for the complexities seen in the present study:
Weak phonological WM
Although all children in the present study had normal nonverbal intelligence scores, it appeared that children who had a P-SSD paired with a low average nonverbal intelligence are likely to exhibit weak phonological WM skills. This is clinically relevant because it is not commonplace to test nonverbal intelligence in children with any form of SSD. This study supports the need to consider the influence of nonverbal intelligence in children with P-SSDs.
A connection between speech production ability and WM ability is clear when considering the role of the articulatory rehearsal mechanism. Articulation rate contributes to one’s ability to retain information in the phonological loop by keeping phonological information available for recall via the articulatory rehearsal mechanism (Baddeley, 2007). Consequently, poor articulation ability may contribute to poor performance on a phonological loop task. A child with a SSD would rely on poorly specified phonological representations to produce sounds in words, resulting in the SSD (Sutherland & Gillon, 2005).
The role of low average nonverbal intelligence may be explained through the process of redintegration which is a process by which linguistic knowledge is used to correct errors (Baddeley, 2007). The redintegration concept has been used to explain why repeating nonwords is more difficult than repeating real words and why longer words are more difficult to recall than shorter words. A child with a SSD who has normal or high nonverbal intelligence may use redintegration to bootstrap into phonological memory skills.
Deficits in establishing motoric representations within memory
Baddeley (2003) suggested that the process of setting up speech motor plans may contribute to the use of the articulatory rehearsal. Children with SSD exhibit weaker oral and fine motor skills compared to typically developing peers. In addition, children in these studies exhibit low normal scores on language and cognition measures. It is possible that motoric skills and linguistic skills interact in a way that either complements speech production, or works in a negative cycle to attenuate speech production skills.
Binding of linguistic and motoric abilities
Baddeley (2000) modified his model to include a fourth component—the episodic buffer. The episodic buffer is conceptualized as a limited capacity space in which information from various sources is bound together for temporary use or manipulation. For speech production, the episodic buffer offers a space to integrate phonological and linguistic representations with motor representations. It is possible that children with P-SSD do not have obvious poor linguistic or motoric skills, but instead have poor binding. If this is true, this may explain why children with P-SSDs often have normal, albeit low average, linguistic and motor skills and, importantly, provide support for a cognitive deficit.
Implications
Results of this study indicate that children with P-SSDs present with complex linguistic and cognitive deficits. These results have substantial importance with respect to assessment.
First, it is not commonplace to test nonverbal intelligence (and sometimes language) in children with SSDs. These results suggest that this knowledge of a child’s nonverbal intelligence may inform prognosis for children with SSDs. That is, it is plausible that younger children at risk for P-SSDs have low nonverbal intelligence and that this may serve as a red flag for earlier intervening services.
Second, WM appears to be a sensitive measure of phonological skills. As such, nonword repetition may be a helpful early screener to determine risk of persistence in children with SSDs.