INT J LANG COMMUN DISORD, JANUARY–FEBRUARY VOL.
51, NO. 1, 61–73
Research Report The relationship between phonological short-term memory, receptive vocabulary, and fast mapping in children with specific language impairment Emily Jackson, Suze Leitao and Mary Claessen School of Psychology and Speech Pathology, Curtin University, Perth, WA, Australia
(Received September 2014; accepted April 2015) Abstract Background: Children with specific language impairment (SLI) often experience word-learning difficulties, which are suggested to originate in the early stage of word learning: fast mapping. Some previous research indicates significantly poorer fast mapping capabilities in children with SLI compared with typically developing (TD) counterparts, with a range of methodological factors impacting on the consistency of this finding. Research has explored key issues that might underlie fast mapping difficulties in children with SLI, with strong theoretical support but little empirical evidence for the role of phonological short-term memory (STM). Additionally, further research is required to explore the influence of receptive vocabulary on fast mapping capabilities. Understanding the factors associated with fast mapping difficulties that are experienced by children with SLI may lead to greater theoretically driven word-learning intervention. Aims: To investigate whether children with SLI demonstrate significant difficulties with fast mapping, and to explore the related factors. It was hypothesized that children with SLI would score significantly lower on a fast mapping production task compared with TD children, and that phonological STM and receptive vocabulary would significantly predict fast mapping production scores in both groups of children. Methods & Procedures: Twenty-three children with SLI (mean = 64.39 months, SD = 4.10 months) and 26 TD children (mean = 65.92 months, SD = 2.98) were recruited from specialist language and mainstream schools. All participants took part in a unique, interactive fast-mapping task whereby nine novel objects with non-word labels were presented and production accuracy was assessed. A non-word repetition test and the Peabody Picture Vocabulary Test—Fourth Edition (PPVT-IV) were also administered as measures of phonological STM capacity and receptive vocabulary, respectively. Outcomes & Results: Results of the fast-mapping task indicated that children with SLI had significantly poorer fast mapping production scores than TD children. Scores from the non-word repetition task were also significantly lower for the SLI group, revealing reduced phonological STM capacity. Phonological STM capacity and receptive vocabulary emerged as significant predictors of fast mapping performance when the group data were combined in a multiple regression analysis. Conclusions & Implications: These results suggest that the word-learning difficulties experienced by children with SLI may originate at the fast mapping stage, and that phonological STM and receptive vocabulary significantly predict fast mapping ability. These findings contribute to the theoretical understanding of word-learning difficulties in children with SLI and may inform lexical learning intervention. Keywords: working memory, vocabulary, fast mapping, specific language impairment (SLI).
What this paper adds? Children with SLI demonstrate difficulty with vocabulary and language development. In an attempt to understand the possible causes of poor lexical development, research has explored the fast mapping capabilities of children with SLI. However, there is inconsistent evidence that children with SLI have markedly reduced fast mapping capabilities compared with TD children. Furthermore, there is a wealth of research highlighting impaired phonological STM in
Address correspondence to: Suze Leitˆao, School of Psychology and Speech Pathology, Curtin University, Kent Street, Bentley, Perth, WA 6102, Australia; e-mail: S.leitˆ[email protected]
International Journal of Language & Communication Disorders C 2015 Royal College of Speech and Language Therapists ISSN 1368-2822 print/ISSN 1460-6984 online DOI: 10.1111/1460-6984.12185
Emily Jackson et al. children with SLI and strong theoretical ties between this working memory component and fast mapping capabilities, yet little evidence exists that phonological STM contributes to fast mapping difficulty in children with SLI. The present study employed a unique fast-mapping task (that addressed a number of methodological concerns in previous research) and found that children with SLI demonstrated significantly poorer fast mapping production skills than TD children. Additionally, this study provided empirical support for the theoretical suggestions that the deficit in phonological STM experienced by children with SLI is significantly related to their fast mapping difficulties. Receptive vocabulary was also found to be a factor significantly associated with fast mapping. These findings have implications for our understanding of SLI and for developing theoretically sound word-learning interventions.
Introduction Specific language impairment (SLI) describes language that is impaired despite normal hearing and average non-verbal cognitive abilities (Leonard 1998). Children with SLI may exhibit difficulties with receptive and expressive language development at several levels (including phonological, lexical, semantic, syntactic and pragmatic), with lexical breakdown suggested to originate at the initial stages of the word-learning process (Gray and Brinkley 2011, Kan and Windsor 2010). Understanding the mechanisms associated with poor word learning in children with SLI will lead to a greater theoretical appreciation of vocabulary development and may inform more focused intervention approaches (Alt 2011). Word learning The mapping theory proposed by Chiat (2001) can be applied to the stages of word learning, which involve input processing of sound-based (phonological) and meaning-based (semantic) information and the ability to map between the two. Fast mapping is considered the first stage in the process of laying down lexical representations, whereby the child initially hears the word, stores its phonological form and begins to map semantic features to that form. Slow mapping subsequently occurs over an extended period where the child hears and refines the phonological and semantic representations of each word as it recurs in different contexts (Chiat 2001). Mapping theory posits that this process (referred to as gradual abstraction) allows progressive establishment of a word’s precise meaning and phonological form within long-term memory, and also its syntactic relation to other words (Chiat and Roy 2008). This suggests that if a child ineffectively fast maps a new word there will be implications for slow mapping success (Chiat and Roy 2008). That is, if the phonological form is not sufficiently robust it may not be effectively recognized as it recurs in different contexts (Gray 2006). This is likely to result in poor refinement of the phonological, semantic and syntactic functions of the word throughout slow mapping, with implications for restricted vocabulary and poor comprehension and production of language (Chiat 2001).
Fast mapping in SLI Research suggests that children with SLI exhibit significant difficulties with fast mapping in comparison with typically developing (TD) children. Rice and colleagues conducted a series of studies utilizing a quick incidental learning (QUIL) paradigm whereby novel words were presented within a video-watching task and participants (aged 5) were given no prompting to attend to stimuli (Rice et al. 1992, 1994). For conditions with few exposures to the novel label (i.e., three), children with SLI performed significantly lower than TD children on post-viewing comprehension tasks, however mapping was comparable in conditions with more exposures (i.e., 10; Rice et al. 1994). This finding (consistent with Skipp et al. 2002) indicates that providing more exposures to new words may start the slow mapping process (whereby links to long-term memory are made), which may facilitate performance on comprehension tasks (Chiat 2001). Alt et al. (2004) provide further evidence of poor fast mapping in children with SLI (aged 4;0–6;5) compared with age-matched TD children. Three exposures to 12 novel objects and 12 novel actions with nonword lexical labels were provided during a computer task, and each child’s comprehension of the semantic features was tested as well as their ability to correctly recognize each novel phonological label (Alt et al. 2004). Results showed that the SLI group recognized significantly fewer phonological labels than age-matched TD children, indicating reduced ability to retain a stable phonological form for the novel words (Alt and Plante 2006, Alt et al. 2004). Children with SLI also demonstrated significantly reduced accuracy when answering comprehension questions regarding the semantic features of novel words. This is likely to exacerbate their difficulty when acquiring new vocabulary and lead to reduced understanding of the meanings conveyed by newly learned lexical labels (Alt et al. 2004). Gray (2004) also found fast mapping difficulties in children with SLI (aged 4;0–5;11) compared with TD age-matched peers. In this study, participants were presented with four nouns (which were low frequency real English words). For each item, one model was provided, and then a comprehension probe and a production probe were administered. The results indicated that
STM, receptive vocabulary and fast mapping in children with SLI children with SLI had significantly poorer comprehension of fast mapped nouns, however they were not significantly different to TD peers for production of the novel word labels (Gray 2004), potentially reflecting methodological concerns (Kan and Windsor 2010). For instance, phonotactic probability (the frequency with which phoneme segments of the novel word occurs within English words) was not controlled in that real English words were used for the ‘novel word’ stimuli. This potentially introduced a bias whereby fast mapping of the phonological code and hence, initial production, was facilitated by knowledge of high-frequency phoneme sequences (Gray and Brinkley 2011, Munson et al. 2005b). Nevertheless, performance on fast mapping comprehension and production tasks significantly predicted the number of words children with SLI learned in a subsequent word-learning task (Gray 2004). This demonstrates the key role of fast mapping in long-term lexical acquisition and highlights the importance of exploring factors contributing to fast mapping difficulties in order to provide effective intervention (Kan and Windsor 2010). In contrast to studies demonstrating poor fast mapping in SLI, Gray (2006) found that, in a subset of children aged 3–6, those with SLI demonstrated significantly lower comprehension and production of fast mapped labels compared with TD children only at age 5 (Gray 2006). The lack of significant group differences at ages 3, 4 and 6 might reflect that six exposures to the novel words were provided, potentially demonstrating that the children with SLI had increased opportunity to form stable representations as part of the slow mapping process, thus allowing them to perform comparably with TD children (Chiat 2001, Gray 2006). Further, Gray (2006) suggests that significantly lower fast mapping skills in children with SLI at age 5 was due to this group of children presenting with more severe language difficulties than other age groups and that this indirectly influenced their fast mapping ability. With further regard to production, non-significant group differences may have resulted from SLI and TD participants at all ages performing close to basal level, likely due to the fact that all four target objects were presented before administration of the production probes, rather than testing production in between presentation of new items (Gray 2006). These inconsistent findings highlight the need for further research to consider the number of items presented and the timing of assessment probes following presentation. Exploring phonological STM and non-word repetition in SLI A key issue that emerges from the fast mapping literature pertains to the reasons why children with SLI might fast
map less efficiently than TD peers (Gray 2006, Kan and Windsor 2010). One potential explanation is the notion of reduced capacity of phonological short-term memory (STM), which is the ‘storage’ component of the phonological loop mechanism in Baddeley’s (2003) model of working memory. Phonological STM is suggested to store memory traces of sound-based information temporarily, while the second component of the phonological loop—the sub-vocal rehearsal mechanism—refreshes this information. Phonological STM holds most interest to fast mapping investigations, with initial encoding of novel phonological representations suggested to occur within this temporary storage buffer (Alt 2011, Baddeley 2003). Furthermore, the rehearsal mechanism purportedly maintains these representations whilst semantic information is encoded, resulting in fast mapped formmeaning associations (Archibald and Gathercole 2006). In line with suggestions that fast mapping requires input processing of sound-based information in phonological STM, fast mapping difficulties for children with SLI might be attributable to deficient phonological STM capacity (Alt 2011, Montgomery et al. 2010). Preliminary support for this suggestion emerges from consistently poor performance in children with SLI on non-word repetition tasks (a robust measure of phonological STM capacity; Archibald and Gathercole 2006). For instance, a meta-analysis showed that across 23 nonword repetition studies, children with SLI performed 1.27 SD (standard deviation) below TD children, indicating a significant difference (Graf Estes et al. 2007). Additionally, a word-length effect for non-word repetition has been demonstrated whereby repetition accuracy in children with SLI was comparable with TD children for shorter non-words (one and two syllables) but significantly lower for longer non-words (e.g., three and four syllables; Graf Estes et al. 2007, Jones et al. 2010). This indicates that the ability to encode phonological information within phonological STM is subject to reduced capacity limitations in SLI and that repetition of longer non-words is sufficiently sensitive to reveal this deficit (Jones et al. 2010). This provides support for utilizing non-word repetition as a measure of phonological STM. While fast mapping does involve some additional processes to those required to complete non-word repetition tasks (including processing and encoding semantic information and refining the output phonological code), the core processes of the two tasks are similar (i.e. hearing and processing a phonological code, encoding an output phonological representation, and articulating the word). It is therefore expected that patterns of difficulty with non-word repetition for children with SLI would extend to fast mapping given the similar processing requirements of these two tasks and the involvement that phonological STM plays in encoding new phonological forms (Baddeley 2003).
Emily Jackson et al. Relation between phonological STM and fast mapping in SLI
Supporting the suggestion that phonological STM is one of the factors underlying fast mapping, previous research has found robust associations between phonological STM and word learning in TD children (e.g., Bowey 2001). However, few studies have specifically explored the association between phonological STM and fast mapping in children with SLI. One of the few studies found significant correlations (in pre-primary SLI and TD groups) between non-word repetition scores and fast mapping of semantic features and phonological labels (Alt and Plante 2006). This, and a similar finding by Gray (2004) provides preliminary support for previous theoretical suggestions that the ability to hold speech information in phonological STM facilitates form-meaning mapping (Alt and Plante 2006, Baddeley 2003). More recently, Alt (2011) investigated the relationship between phonological STM and fast mapping by exploring the effect of word length on fast mapping. Groups of SLI and TD children (aged 7–8) were given two presentations of two- and four-syllable novel words on a computer-based task. The results indicated that children with SLI fast mapped with comparable accuracy with TD children for two-syllable novel words but with significantly less accuracy for four-syllable stimuli, implicating deficient phonological STM capacity in poor mapping of longer words (Alt 2011). However, no nonword repetition measure was included. Given that difficulties in phonological STM in SLI are mainly evident in longer non-words, indicating problems in storage capacity, future research is required including an explicit measure of phonological STM so that it may be more clearly explored as a factor underlying fast mapping. An investigation into the direct association between nonword repetition and fast mapping may provide more robust evidence that reduced phonological STM capacity underlies poor fast mapping in children with SLI (Montgomery and Windsor 2007). In contrast, Gray (2006) specifically explored the association between phonological STM (measured by non-word repetition and digit span tasks) and fast mapping in children with SLI (aged 3–6) and found no significant relationship. Whilst this finding indicates that phonological STM might not be a key underlying process in fast mapping, Gray (2006) suggested that the experimental conditions of the fast-mapping task potentially masked an association. In particular, the novel word stimuli were short (two syllables). This may reflect a word length effect in that the phonological STM deficits in children with SLI are revealed through the use of longer stimuli (Graf Estes et al. 2007, Jones et al. 2010). In addition to such methodological issues,
the general lack of studies exploring the non-word repetition–fast mapping association in children with SLI warrants the need for further research. Relation between receptive vocabulary, phonological STM and fast mapping In addition to the issue of phonological STM, vocabulary knowledge has been posited to play a role in the acquisition of new words (Baddeley 2003). Baddeley’s (2003) model of working memory posits a reciprocal relationship between phonological STM (i.e., the mechanism suggested to be closely involved in fast mapping) and long-term memory (i.e., vocabulary knowledge), which becomes increasingly reciprocal after the age of 5 (Baddeley 2003). That said, a rich and varied vocabulary would assist new word learning, as the learner has prior lexical knowledge from which to draw to support learning of new phonological and semantic forms. Further, good phonological STM facilitates vocabulary development given its role in temporarily storing new phonological information. The role played by existing vocabulary in new vocabulary acquisition (including fast mapping) is preliminarily supported by evidence of a significant relationship between vocabulary and non-word repetition (as a measure of phonological STM; Baddeley 2003). For instance, Munson et al. (2005) found a significant association in children with SLI between non-word repetition and expressive and receptive vocabulary. Further, they found that children with smaller vocabularies (i.e. those with SLI) were more influenced by phonotactic probability, in that they more accurately repeated words with high compared with low phonotactic probability. This suggested that children with less robust phonological representations in stored vocabulary displayed more difficulty with the task of making generalizations effectively to create new phonological representations (as required when repeating non-words; Munson et al. 2005b). Given the processing similarities between nonword repetition and fast mapping, vocabulary is suggested to similarly impact upon fast mapping. That is, children with restricted vocabularies might experience difficulties abstracting specific phonological forms from the input signal and establishing production routines for novel words ‘because they do not have as many stored sub-lexical patterns from which to draw’ (Gray 2006: 957). There is equivocal evidence to support this suggestion, with Rice and colleagues finding no association between receptive vocabulary and fast mapping (QUIL) comprehension scores (Rice et al. 1992, 1994). However, in these studies children were sampled based on lower vocabulary performance, hence limiting the variability among scores required to yield a significant
STM, receptive vocabulary and fast mapping in children with SLI correlation (Gray 2006). Furthermore, Gray (2003) found no significant correlation between receptive vocabulary scores and fast mapping production for SLI or TD groups. Conversely, Alt et al. (2004) found significant correlations between receptive vocabulary and recognition of fast mapped lexical labels for both nouns and verbs in children with SLI. Similarly, Gray (2004) found that receptive vocabulary performance accounted for a significant 21% and 26% of the variance in fast mapping production and comprehension, respectively. However, Gray (2004) examined the SLI and TD groups within a pooled dataset in the regression analysis, indicating that TD scores may have ‘influenced the overall relationship’ (p. 966). This warrants the need for further investigation within separate groups. The findings of significant correlations between receptive vocabulary and fast mapping might indicate that children with larger vocabularies have greater existing lexical knowledge from which to draw in order to facilitate the establishment of new phonological and semantic representations (Alt et al. 2004, Gray 2006). This would additionally contribute to our understanding of the suggested reciprocal relationship between long-term memory (i.e., vocabulary) and the processing of new phonological information in STM (Baddeley 2003). Finally, as children with SLI tend to display smaller vocabularies than their TD peers (Montgomery et al. 2010), it is important to explore lexical knowledge as a potential underlying issue contributing to poor fast mapping in children with SLI. Present study As fast mapping is arguably a vital stage in the development of vocabulary (Chiat and Roy 2008), the present study explored whether children with SLI fast map phonological labels with significantly less accuracy than TD children. We explored this question using a novel fast-mapping task carefully designed to overcome methodological concerns of previous research (i.e., too many exposures, uncontrolled phonotactic probability, length of novel words, assessment measure and timing of assessment probes). Therefore, we chose to provide three exposures to each novel word as this aligns with the literature’s view that up to three presentations captures the initial process of form-meaning mapping, and four or more exposures may provide opportunity for the slow mapping process to begin (Alt and Plante 2006, Chiat and Roy 2008). We also chose to select novel words with low phonotactic probability, as the study aimed to accurately capture each child’s ability to accurately encode an entirely unfamiliar phonological form, without prior lexical knowledge potentially biasing the word-learning process (Munson et al. 2005b). Further, multisyllabic novel words (i.e. ranging from two to four syllables)
were selected as it is suggested that phonological STM capacity limitations are revealed when children with SLI are presented with longer non-words (Alt 2011, Jones et al. 2010). An additional consideration for the present study was the selection of a production task to measure fast mapping accuracy. In comprehension and recognition tasks, the child’s ability to accurately fast map the novel phonological form may be less sensitively captured as, even with a partial or inaccurate phonological representation of the new word, a child may have enough information to be able to comprehend or recognize the word (Kan and Windsor 2010). Instead, production requires sufficient storage of the phonological information to establish an output representation, therefore this assessment measure was chosen as it was thought to reflect more closely the child’s ability to learn ‘more fine-grained knowledge of the phonemic structure of non-words’ (Alt 2011: 181). Finally, we were careful to probe for production accuracy following the presentations of each individual word so to avoid floor effects seen in previous studies where all novel stimuli were presented before production was assessed (Gray 2006). These methodological considerations informed the first hypothesis:
r Children with SLI will score significantly lower than TD children on production of fast mapped phonological labels. Further, to contribute to the theoretical understanding of fast mapping difficulties in SLI, we explored research questions regarding whether children with SLI have significantly poorer non-word repetition than TD children, and if phonological STM and receptive vocabulary predict fast mapping capabilities in these groups. These questions informed the second, third and fourth hypotheses:
r Children with SLI will score significantly lower on a non-word repetition task than TD children,
r Phonological STM (measured by non-word repetition) will significantly predict fast mapping performance in SLI and TD groups, r Receptive vocabulary will significantly predict fast mapping performance in SLI and TD groups. A greater appreciation of fast mapping capabilities in children with SLI, in addition to factors underlying performance, may lead to greater theoretical understanding of lexical learning and therefore allow the development of more effective, theoretically sound word-learning intervention (Alt 2011, Chiat 2001, Kan and Windsor 2010).
Emily Jackson et al.
Table 1. Means and standard deviations on the CELF-P2 and Raven’s CPM SLI
CELF-P2 CLSa Raven’s CPMb
Notes: Standard scores provided. b Raw scores provided. CI = confidence interval; CLS = core language score.
Method Participants Forty-nine pre-primary children were recruited for this study. All participants were monolingual English speaking; had no history of hearing difficulties; and had age-appropriate behavioural, pragmatic and articulatory skills, as determined by their classroom teachers. Twenty-three children with SLI (18 males, 5 females, mean = 64.39 months, SD = 4.10 months, age range = 5;0–5;11) were recruited from a language development centre (LDC). To qualify for an LDC, students must have a primary language impairment, average or above-average non-verbal cognitive skills, and appropriate adaptive behaviour abilities. Each child met the criteria for LDC placement, confirmed in this study by further testing of poor language skills and average non-verbal cognition. That is, children were required to obtain a core language score of 85 or less (1 SD below the mean) on the Clinical Evaluation of Language Fundamentals Preschool—Second Edition (CELF-P2; Wiig et al. 2006), and a raw score of 14 or more on the Raven’s Coloured Progressive Matrices (CPM; Raven 2003). Raven (2003) specified that a raw score of 14 is the expected average score for pre-primary children. Twenty-six TD children (10 males, 16 females, mean = 65.92 months, SD = 2.98 months, age range = 5;0–5;11) were recruited from a metropolitan primary school. TD participants were required to have a core language score of above 85 and a raw score of 14 or more on the Raven’s CPM (see table 1 for participant selection scores). A Welch’s t-test confirmed that SLI and TD groups did not differ significantly in age. However, gender matching was not possible. Consistent with the frequently observed pattern of gender asymmetry in language impairment, there were a greater number of males than females enrolled at the LDC. While we attempted to take this into account by selecting more males with typical language, consent to participate was greater among female than male children in the mainstream school. The sampling method involved teachers at the schools identifying potential participants and distribut-
ing consent forms. At the LDC and metropolitan school respectively, 70% and 58% of the sample approached consented to participate. Participants were assessed individually in a quiet, familiar room at their school. Measures Participant selection The core language score (including three subtests: sentence structure, word structure and expressive vocabulary) provided a measure of general language ability, and the Raven’s CPM measured non-verbal cognitive ability. Data collection The Nonword Repetition Test (NRT; Dollaghan and Campbell 1998) was used to measure phonological STM capacity. This assessment involved repetition of 16 non-words (ranging from one to four syllables, with four non-words at each length), and each syllable did not correspond to an English word (see Dollaghan and Campbell 1998 for additional assessment information). Following phonetic transcription and pronunciation guidelines outlined by Dollaghan and Campbell (1998) the non-words were recorded on an Echo Pulse Pen by an English-speaking female and then presented to each participant using a Macbook Air laptop computer. Per cent phonemes correct (PPC) scores were used to compare groups on non-word repetition and to explore phonological STM as a fast mapping predictor (see ‘Procedure’ for further scoring details). The Peabody Picture Vocabulary Test—Fourth Edition (PPVT-IV; Dunn and Dunn 2007) measured receptive vocabulary. Standard scores were used to explore receptive vocabulary as a fast mapping predictor. The fast-mapping task involved presentation of nine novel alien objects (hand constructed with clay) in an interactive activity where production accuracy was measured. The aliens had non-word phonological labels, the selection of which involved consideration of phonotactic probability and articulatory complexity. First, using the N-Watch algorithm (Davis 2005) the phonotactic probabilities of a randomized set of 227 real words (two-, three- and four-syllable lengths) were calculated. This yielded a range of phonotactic probabilities expected from a random sample of real words. For each syllable length, the mean and SD of the phonotactic probabilities were calculated. Twenty-nine non-words were then selected from previous non-word repetition and fast mapping studies (Gathercole et al. 1994, Gupta 2003, Munson et al. 2005a) and were processed using the N-Watch algorithm (Davis 2005). Non-words with phonotactic probabilities that fell below 1 SD below the mean of
STM, receptive vocabulary and fast mapping in children with SLI Table 2. Sets of non-word labels for fast-mapping task Syllable length
Two-syllable /poʊdɔd/ /doʊnug/ /jugɔɪn/ /boʊgib/ Three-syllable /kɜdəwəmb/ /gɔnəpek/ /fɪkətæmp/ /tʌgnədit/ Four-syllable /jelæntɪfɜ/ /gufeʃɜgʊs/ /jɔfeʃɜged/
the real word samples were considered to have low phonotactic probability, and from this selection three non-words at each syllable length (plus an additional two- and three-syllable non-word for training) were chosen as stimuli for the fast-mapping task (table 2). Twoand three-syllable stimuli were pronounced with stress on the first syllable. One four-syllable item (/jelæntɪfɜ/) was pronounced with emphasis on the second syllable, and the third syllable was emphasized in the remaining four-syllable items. This difference in syllable emphasis is due to the stimuli being selected from different studies with differing pronunciation guidelines (Gathercole et al. 1994, Gupta 2003, Munson et al. 2005a). Finally, non-words were selected if they contained phonemes considered to generally develop before age 5. Some consonant clusters were included in the non-words. Procedure Pilot testing was conducted with four TD pre-primary children in their own homes. All participant selection and data collection measures were administered. The fast mapping procedure was refined, including novel object presentation and timing and administration of production probes. When more than one novel word was presented before probing for production, the children provided either no response or random guesses. As this would likely result in many participants across both groups performing at basal level (e.g., as seen in Gray 2006), it was determined that production probes would be administered following the presentations of each individual item. Participant selection Participant selection measures were administered in two 15–20-min sessions on the same day: the CELF-P2 in the first session and the Raven’s CPM in the second. All participants had no teacher-reported articulation difficulties, confirmed by examiner observations during testing. Data collection A week following participant selection, data collection assessments were administered in two 20-min sessions
on the same day. The NRT (administered according to Dollaghan and Campbell 1998) and PPVT-IV (administered according to guidelines in Dunn and Dunn 2007) were conducted within the first session and the fast-mapping task within the second. For the non-word repetition and fast-mapping tasks, responses were audio recorded on an Echo Pulse Pen and later transcribed phonetically and scored using the PPC measure. The use of any age-appropriate articulation patterns (e.g., stopping /v/ to /b/) were not scored as errors. Both vowels and consonants were scored to calculate PPC. Additionally, a research assistant re-scored all NRT and fast mapping recordings and a Spearman’s rho correlation of .96 indicated high association and therefore high reliability between the scorers. The fast-mapping task included a training and experimental phase. For the training phase, an interactive activity was conducted to construct a simple ‘farm’ scene using coloured blocks and block toys. A handconstructed rocket was introduced and participants were told that there were aliens inside who wanted to look around the farm, and that the aliens had strange names to which they needed to pay attention. Participants were instructed to ‘land’ the rocket on the farm and then pick out one alien. Participants could not see the aliens inside the rocket, ensuring random presentation of items within the training set. Three verbal presentations of the non-word label (using simple instructional or commenting phrases) were provided within an interactive activity using the farm scene (e.g., ‘You picked out /poʊdɔd/ . . . Put /poʊdɔd/ on the horse . . . /Poʊdɔd/ likes the horse!’). Dialogue throughout presentations was allowed so long as the labels were said three times within 60 s. Presentation phrases depended on participant choices within the activity, however semantic attributes of the aliens were not emphasized. After 10–15 s following the third verbal presentation, a production probe was administered: ‘There are more aliens to get out, but we need to put that one away first. What’s its name?’ Participants were given positive feedback for correct responses and neutral feedback for incorrect responses. Prompts to ‘try any of the sounds in the alien’s name’ were provided if no response was given. If participants created names for the aliens (e.g., ‘tractor man’) prompting was provided to ‘listen carefully to the alien names when they get out from the rocket’. The same procedure was followed with the second training item. For the fast mapping experimental phase, the procedure closely followed that of the training phase. Set 1 of the aliens was presented using the farm scene, but participants changed the blocks to a park scene for Set 2 and to a backyard scene for Set 3. Each set was presented using the training phase procedure, however,
Emily Jackson et al.
Table 3. Descriptive statistics for NRT, PPVT-IV and fast mapping for SLI and TD groups SLI Measure PPVT-IVa NRTb FMc
N 23 23 23
Mean 90.83 65.59 33.36
Table 4. Statistics for missing data analysis using Little’s Missing Completely at Random Test Missing Data
TD SD 7.13 8.91 15.01
N 25 26 25
Mean 109.35 87.46 69.29
SD 10.61 4.55 13.18
Notes: a Standard scores provided. b PPC scores provided. c PPC scores provided. FM = fast mapping.
non-word labels were presented within phrases suited to the relevant scene (e.g., ‘Put /jugɔɪn/ on the swing’ was used for the park scene). Neutral feedback was given for all responses to production probes. Research design A quasi-experimental between groups research design was employed. This design was chosen as the study aimed to compare two independent samples of children and participants could not be randomly assigned to groups. Results Descriptive statistics Descriptive statistics for receptive vocabulary, non-word repetition and fast mapping scores are reported in table 3. Two TD subjects were removed from analyses as their high non-verbal cognitive skills and receptive vocabulary scores met criteria for extreme outliers (±3 z-scores). Missing data One missing data point was identified (NRT score in the SLI group). Little’s Missing Completely at Random (MCAR) test was used to determine the pattern of missingness and the data point was found to be MCAR, indicating that the missing value occurred entirely at random and is not due to effects from the variables of interest in this study. Expectation maximization was used to estimate the value of the missing NRT score (see table 4 for associated statistics). Fast mapping production and non-word repetition: SLI versus TD A one-way analysis of covariance (ANCOVA) was used to explore the hypothesis that children with SLI would score significantly lower on production of fast mapped phonological labels. As there was an observed difference between the average Raven’s CPM scores between the groups (table 1), a covariate was included to partial out
Note: NRT = non-word repetition score; EM = estimated maximization statistics; MCAR = Little’s Missing Completely at Random Test for the extent and assessment of missingness.
the effects of participants’ non-verbal cognitive scores on fast mapping ability. Assumptions of normality, linearity, homogeneity of variances and homogeneity of regression slopes were not violated. The ANCOVA indicated that after accounting for non-verbal cognitive scores, there was a statistically significant difference between the SLI and TD groups on fast mapping production, F(1, 45) = 58.68, p < .001, η2 = .57, indicating a large effect size. Post-hoc testing revealed that the participants in the SLI group (mean = 33.36 PPC, SD = 15.01 PPC) had significantly lower fast mapping production scores than the TD group (mean = 69.29 PPC, SD = 13.18 PPC). An ANCOVA was also employed to explore the hypothesis that children with SLI would score significantly lower on non-word repetition than TD children. Raven’s CPM scores were included as a covariate to partial out the effects of participants’ non-verbal cognitive scores on non-word repetition performance. Assumptions of normality, linearity, homogeneity of variances and homogeneity of regression slopes were not violated. The ANCOVA indicated that after accounting for non-verbal cognitive skills, there was a statistically significant difference in non-word repetition scores between the SLI and TD groups, F(1, 45) = 86.58, p < .001, η2 = .66, indicating a large effect size. Post-hoc testing indicated that the participants in the SLI group (mean = 65.59 PPC, SD = 8.91 PPC) had significantly lower non-word repetition scores than the TD group (mean = 87.46 PPC, SD = 4.55 PPC). Relationship between phonological STM, receptive vocabulary and fast mapping In order to test the third and fourth hypotheses, correlations between phonological STM, receptive vocabulary and fast mapping were explored. Additionally, correlations between fast mapping, age and non-verbal cognition scores were examined, as word learning may be related to age and general cognitive ability (Kan and Windsor 2010). Spearman’s rho correlations were employed where normality was violated; however, all other assumptions were satisfied. When SLI and TD groups were examined separately, fast mapping production correlated non-significantly with age, non-verbal cognition, phonological STM and receptive vocabulary,
STM, receptive vocabulary and fast mapping in children with SLI Table 5. Correlations (r) between fast mapping and predictor variables in SLI and TD groups Correlations (r)
FM: age FM: non-verbal cognitive skills FM: phonological STM FM: receptive vocabulary
SLI + TD
.25 .11 .24 .32
.37 .10 .26 .05
.33a .39a .75b .66b
Table 6. Unstandardized (B) and standardized (β) regression coefficients and squared semi-partial correlations (sr2 ) for predictors of fast mapping in the multiple regression model SLI + TD
Age Non-verbal cognitive skills Phonological STM Receptive vocabulary
semi-partial correlations for predictors in the regression model). While age and non-verbal cognition were included for the purposes of the regression analysis, the relationship between these factors and fast mapping will not receive further attention as it goes beyond the scope of this paper. Discussion
Notes: Non-parametric correlations (Spearman’s rho) are shown in bold face. FM = fast mapping. a p < .05, two-tailed. b p < .001, two-tailed.
B [95% CI]
1.60 [0.38, 2.82] 1.35 [–0.05, 2.74] 0.50 [0.02, 0.98] 1.06 [0.56, 1.55]
0.60 0.69 0.24 0.25
0.25 0.18 0.28 0.52
.14a .08 .10a .31b
Notes: CI = confidence interval; SE = standard error. a p < .05, two-tailed. b p < .001, two-tailed.
potentially reflecting lack of power due to small sample sizes (table 5). However, when SLI and TD groups were included in a pooled dataset, all correlations were significant, therefore age, non-verbal cognition, non-word repetition and receptive vocabulary were included as predictors in the regression analysis. To estimate the proportion of variance in fast mapping production (in the pooled dataset) that can be accounted for by age, non-verbal cognitive skills, phonological STM and receptive vocabulary, a standard multiple regression analysis (MRA) was performed. The following assumptions were met: normality, linearity and homoscedasticity of residuals; multivariate outliers and multicolinearity. Mild-to-moderate violations of normality were identified but were not concerning given the robust nature of MRA. Furthermore, two univariate outliers (SLI and TD receptive vocabulary) were identified via inspection of box plots but were not eliminated as they were judged to be valid data and were not extreme scores (i.e., ±3 z-scores). The MRA revealed that, in combination, the predictor variables accounted for a significant 71% of variance in fast mapping production, R2 = .71, adjusted R2 = .68, F(4, 42) = 25.68, p < .001. Examined separately, age, phonological STM and receptive vocabulary uniquely accounted for a significant proportion of the variability in fast mapping (see table 6 for unstandardized and standardized regression coefficients and squared
The purpose of the present study was to compare children with SLI and TD children on fast mapping production and to explore phonological STM and receptive vocabulary as potential mechanisms relating to fast mapping performance. All hypotheses were supported in this study. Fast mapping: SLI versus TD In line with the first hypothesis we found that children with SLI produced phonological forms of fast mapped words with significantly less accuracy than TD children. Further, these significant group differences were found after correcting for non-verbal cognitive skills, indicating that the problems children with SLI experienced with fast mapping are more likely related to language and phonological STM abilities, rather than other cognitive skills. This aligns with the literature’s view that SLI describes children who have language impairment in the presence of normal non-verbal cognition (Leonard 1998). Explanations for why we found a significant group difference in fast mapping production performance when previous evidence was equivocal/did not might lie in differing fast mapping methodologies. First, in line with mapping theory’s explanation of fast mapping (Chiat 2001) we provided minimal (three) presentations of the novel words and found significant group differences, supporting previous studies that also presented the novel stimuli three times (e.g., Alt et al. 2004). In contrast, some previous studies provided more exposures (e.g., six, 10) and found non-significant group differences (Gray 2006, Rice et al. 1994, Skipp et al. 2002). According to mapping theory, providing many exposures to new words facilitates gradual abstraction of form-meaning associations into long-term memory, suggesting that some previous ‘fast mapping’ tasks more accurately reflected a stage of slow mapping. The involvement of slow mapping may have assisted children with SLI to perform comparably with TD children by facilitating establishment of more stable novel word mappings (Chiat and Roy 2008). Our finding of significantly lower fast mapping production accuracy in children with SLI may therefore contrast previous findings as our task
70 more purely reflected fast mapping capabilities (Gray 2006, Rice et al. 1994, Skipp et al. 2002). Secondly, our task design was less demanding than previous studies where non-significant group differences in production were found. Gray (2006) presented all four novel items before administering each probe, resulting in both SLI and TD groups having very low production scores, thus yielding non-significant group differences. We conducted pilot testing and found a similar pattern of low performance when presenting more than one novel item at a time. We therefore presented one item at a time before administering each probe, which resulted in production scores more accurately capturing a range of capabilities, as the task was not exceedingly difficult (Kan and Windsor 2010). Thirdly, the use of novel words with low phonotactic probability might explain why our findings contrast previous research. For instance, Gray (2004) used unfamiliar English words as stimuli, likely resulting in high phonotactic probability (Munson et al. 2005b). Children with SLI and TD children might fast map more effectively when phonotactic probability is high (Alt and Plante 2006), indicating that the non-significant group differences found by Gray (2004) potentially resulted from biased fast mapping performance whereby children drew upon knowledge of high-frequency phoneme sequences to facilitate mapping. By using low phonotactic probability stimuli we gained a more accurate understanding of the between-group differences by reducing the potential for fast mapping to be influenced by previous lexical knowledge. Given that the presentation of novel words with high and/or variable phonotactic probabilities may more accurately reflect real-life learning situations, the focus of future research could be to compare fast mapping capabilities using novel words with differing levels of phonotactic probability (Munson et al. 2005b). In summary, this finding of significantly lower fast mapping production accuracy in children with SLI supports the suggestion that word-learning difficulties for children with SLI likely originate when fast mapping the phonological form (Kan and Windsor 2010). A key theoretical implication of this finding is its alignment with mapping theory’s proposal that poor phonological specification during fast mapping has a cascading effect on slow mapping, potentially explaining reduced vocabulary growth in children with SLI (Montgomery et al. 2010). This suggestion has been previously supported by evidence that poor fast mapping predicts disrupted long-term word learning in children with SLI (Gray 2003), therefore clinical implications arise for providing intervention targeted at strengthening new phonological forms. For instance, providing clear exemplars and contrasting similar-sounding words might support more accurate phonological encoding, leading to more
Emily Jackson et al. effective slow mapping to refine phonological, semantic, and syntactic elements for successful language comprehension and production (Chiat 2001, Gray and Brinkley 2011). Future research should explore effective clinical strategies for supporting fast mapping in children with SLI. To inform further theoretically sound word-learning intervention, future research might investigate encoding of specific semantic features using a similar task given that children with SLI previously demonstrated additional difficulty with this aspect of fast mapping (Alt and Plante 2006). Another consideration for further research is the effect of word class on fast mapping. We utilized proper nouns, essentially making this a study of ‘name learning’. That said, it would be important for future studies to replicate the findings with content nouns. Investigations into verb learning should also be conducted given that previous studies showed SLI and TD groups experience greater difficulty learning verbs, potentially due to the syntactic complexity (Alt et al. 2004). Additionally, different findings might emerge when using comprehension and recognition outcome measures given that production involves additional processes (e.g., articulation, establishment and refinement of an output representation) that may not impact on performance on comprehension and recognition tasks (Kan and Windsor 2010). Future research might therefore compare fast mapping of different word classes and explore the impact of different outcome measures in order to further inform goals and strategies for wordlearning intervention.
Relation between phonological STM and fast mapping In support of the second hypothesis and consistent with previous research, children with SLI scored significantly lower than TD children on non-word repetition, indicating reduced phonological STM capacity (Graf Estes et al. 2007, Jones et al. 2010). Significant group differences were found after partialling out non-verbal cognitive ability scores, which is unsurprising as non-word repetition involves verbal abilities (Baddeley 2003). As the phonological STM element of working memory is argued to be a key underlying process in fast mapping (Baddeley 2003), it was explored in the regression analysis and found to be a small (10%) but significant predictor of fast mapping production, therefore supporting our third hypothesis. Contrary to the findings of Gray (2006), our data showed that children with reduced phonological STM capabilities displayed more difficulty creating stable phonological representations of novel words (Archibald and Gathercole 2006, Montgomery et al. 2010).
STM, receptive vocabulary and fast mapping in children with SLI This finding aligns with previous theoretical reports that impaired phonological STM creates capacity limitations during lexical learning given its role in phonological encoding of novel word forms (Archibald and Gathercole 2006). Despite the strong theoretical link between phonological STM and fast mapping, there has been little previous empirical evidence to support this relationship (Alt and Plante 2006, Gray 2004). For instance, the current finding of a significant predictive relationship between phonological STM and fast mapping is in contrast to the findings of Gray (2006). A potential explanation for this is that we presented a range of novel word lengths for fast mapping whereas Gray (2006) used only two-syllable stimuli. This indicates that our fast-mapping task placed more demand on phonological STM, thereby yielding greater variability in scores that allowed an association to be identified. Alternately, our finding that phonological STM predicted fast mapping may be in contrast to previous studies due to a key limitation in the present study. We found non-significant correlations in separate SLI and TD groups between fast mapping and phonological STM (and other predictor variables), likely due to small sample sizes. Predictors of fast mapping were therefore explored in a pooled dataset, indicating that this finding of phonological STM as a significant fast mapping predictor might be an artefact of a greater distribution of scores created by combining groups that differed in language ability (Gray 2006). Further research replicating this study with larger sample sizes is required to determine if this finding is generalizable to separate SLI and TD groups. Despite this limitation, given the sparse previous evidence indicating a significant relationship between phonological STM and fast mapping in children with SLI (Alt and Plante 2006, Gray 2004), this finding provides support for the theory that phonological STM is a critical theoretical component of fast mapping. Poor fast mapping can have deleterious effects on vocabulary growth, and vice versa: poor vocabulary has a negative impact on fast mapping ability, causing a cycle of poor performance. Therefore to support long-term vocabulary growth, clinical implications arise for supporting phonological STM during fast mapping, such as reducing task demands (e.g., teaching shorter novel words when a child is learning a new concept before introducing longer novel words, and teaching sets of phonetically different words rather than similar-sounding words; Alt 2011). Further research should explore strategies to support phonological STM during word learning for children with SLI, both in the clinic and the classroom. A natural extension of this research is to explore fast mapping across different novel word lengths given the previous evidence of a differential effect of length on non-word repetition accuracy in SLI and TD groups,
and emerging evidence that this effect extends to fast mapping performance (Alt 2011, Jones et al. 2010). This may further inform strategies to support phonological STM in SLI when teaching new words of varying length (Alt 2011). It is important also to note that issues with phonological memory have been shown to be more characteristic of dyslexia than SLI (Catts et al. 2005). This highlights the importance for future research to further explore the link between phonological STM, language learning and literacy development in children with SLI and dyslexia. Relationship between receptive vocabulary, phonological STM and fast mapping In addition to phonological STM, receptive vocabulary was a significant predictor (31%) of fast mapping production accuracy in the combined groups, supporting our fourth hypothesis and echoing previous research demonstrating a robust relationship between receptive vocabulary and non-word repetition (e.g., Munson et al. 2005a). Given the similar linguistic processes and phonological STM demands for non-word repetition and our fast-mapping task (required to hear, encode and say the stimuli, one item at a time), this relationship between receptive vocabulary and fast mapping is unsurprising (Baddeley 2003). Furthermore, this finding is consistent with previous theoretical reports that children with larger receptive vocabularies have greater extant sub-lexical patterns to draw upon, thus facilitating encoding of novel phonological forms and establishment of new motor programs for production (Gray 2006). Further theoretical implications are that this finding lends support to the proposal of a reciprocal relationship between phonological STM and long-term memory (e.g., receptive vocabulary) after age 5, with good phonological STM facilitating fast mapping and therefore long-term word learning, and a rich vocabulary assisting fast mapping in phonological STM (Baddeley 2003). Because we examined SLI and TD in combination, further research is required to confirm that receptive vocabulary predicts fast mapping in children with SLI, which will provide further insight into the relationship between short- and long-term memory and the potential impact on word learning (Baddeley 2003). With regard to existing literature this finding contrasts the results of some previous studies, however these studies utilized low receptive vocabulary as an inclusion criterion for SLI (Rice et al. 1992, 1994). Therefore, the lack of potential variability in receptive vocabulary scores in each group, and the fact that the fast mapping– receptive vocabulary relationship was examined within separate groups, might explain why no significant association was identified (Gray 2006). Instead, we
72 selected SLI and TD groups based on low versus average language ability across a number of domains, one of which was expressive vocabulary. Given that expressive and receptive vocabularies are separate but highly related knowledge bases, it was likely that receptive vocabulary scores across the two groups were greatly distributed, which allowed a relationship to be identified (especially when the data were pooled in the regression analysis). This pattern of results is consistent with earlier work by Gray (2004), who also found a significant association between receptive vocabulary and fast mapping with SLI and TD groups combined, potentially reflecting increased variability created by pooling groups in the regression. Again, this warrants the need for future research to investigate the influence of receptive vocabulary on fast mapping in separate SLI and TD groups (Gray 2004). Overall, the regression analysis indicated that receptive vocabulary uniquely accounted for the greatest amount of variance in fast mapping, followed by age, non-word repetition and non-verbal cognitive ability. We might have expected phonological STM to be a stronger predictor of fast mapping given their robust theoretical link (Baddeley 2003). However, difficulties observed with fast mapping in this study may have occurred due to other issues, such as problems with attention to input, processing speed, encoding, storing and refining the phonological code, activating and retrieving the representation for output, and articulation (Montgomery et al. 2010). Therefore, these other factors may have contributed to the smaller than expected relationship. Future research should consider the influence of additional cognitive constraints (e.g., executive functions) underlying fast mapping difficulties in children with SLI. This may further our understanding of processes contributing to poor word learning in SLI (Montgomery et al. 2010). Some final limitations to consider relate to participant selection. First, while in this study we did not aim to include TD children only with above average language skills, the TD children available for participation showed average or above average performance on the core language score on the CELF-P2 (Wiig et al. 2006). Future research should aim to include TD children with varied language abilities (including low average children), as they may display different patterns of fast mapping capabilities than their higher language-level peers. Secondly, selecting participants with a raw score of 14 or more on the Raven’s CPM (Raven 2003) resulted in all participants having average to above average nonverbal cognitive abilities. Future research should consider the use of a measure of non-verbal cognition with more sensitive scoring procedures, and include children with low average abilities as they may also demonstrate different fast mapping capabilities. Finally, for greater
Emily Jackson et al. clinical applicability, future research should also gender match participants (which was not possible in this study) and explore word-learning capabilities across different genders. Summary This research contributes to the literature by providing evidence (within a carefully designed task) that preprimary children with SLI fast map novel phonological labels with significantly less accuracy than TD peers. This lends support to the argument that disruption to the initial stage of lexical learning may explain poor vocabulary development—and, by extension, poor language learning—in children with SLI (Chiat 2001). Furthermore, the results highlight the involvement of phonological STM and receptive vocabulary in fast mapping, contributing to our understanding of the reciprocal nature of the relationship between working memory and long-term vocabulary knowledge. Finally, this research has a number of clinical implications for supporting the development of vocabulary, which will encourage further research questions and lead to the development of theoretically sound word-learning intervention (Alt 2011). Acknowledgements This research was supported by a grant (#RAF-2013–08) from the Standard Research Allocation Fund Grants Schemes through the School of Psychology and Speech Pathology, Curtin University. The authors also thank the schools and participants involved in this study. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this paper.
References ALT, M., 2011, Phonological working memory impairments in children with specific language impairment: where does the problem lie? Journal of Communication Disorders, 44, 173–185. ALT, M. and PLANTE, E., 2006, Factors that influence lexical and semantic fast mapping of young children with specific language impairment. Journal of Speech, Language and Hearing Research, 49, 941–954. ALT, M., PLANTE, E. and CREUSERE, M., 2004, Semantic features in fast-mapping: performance of preschoolers with specific language impairment versus preschoolers with normal language. Journal of Speech, Language and Hearing Research, 47, 407–420. ARCHIBALD, L. M. D. and GATHERCOLE, S. E., 2006, Short-term and working memory in specific language impairment. International Journal of Language and Communication Disorders, 41, 675–693. BADDELEY, A., 2003, Working memory and language: an overview. Journal of Communication Disorders, 36, 189–208. BOWEY, J. A., 2001, Nonword repetition and young children’s receptive vocabulary: a longitudinal study. Applied Psycholinguistics, 22, 441–469.
STM, receptive vocabulary and fast mapping in children with SLI CATTS, H. W., ADOLF, S. M., HOGAN, T. P. and WEISME, S. E., 2005, Are specific language impairment and dyslexia distinct disoder? Journal of Speech, Language and Hearing Research, 48, 1378–1399. CHIAT, S., 2001, Mapping theories of developmental language impairment: premises, predictions and evidence. Language and Cognitive Processes, 16, 113–142. CHIAT, S. and ROY, P., 2008, Early phonological and sociocognitive skills as predictors of later language and social communication outcomes. Journal of Child Psychology and Psychiatry, 49, 635– 645. DAVIS, C. J., 2005, N-Watch: a program for deriving neighborhood size and other psycholinguistic statistics. Behaviour Research Methods, 37, 65–70. DOLLAGHAN, C. and CAMPBELL, T. F., 1998, Nonword repetition and child language impairment. Journal of Speech, Language and Hearing Research, 41, 1136–1146. DUNN, D. M. and DUNN, D. M., 2007, Peabody Picture Vocabulary Test, Fourth Edition (Minneapolis, MN: Pearson). GATHERCOLE, S. E., WILLIS, C. S., BADDELEY, A. D. and EMSLIE, H., 1994, The children’s test of nonword repetition: a test of phonological working memory. Memory, 2, 103– 127. GRAF ESTES, K., EVANS, J. L. and ELSE-QUEST, N. M., 2007, Differences in the nonword repetition performance of children with and without specific language impairment: a meta-analysis. Journal of Speech, Language and Hearing Research, 50, 177– 195. GRAY, S., 2003, Word-learning by preschoolers with specific language impairment: what predicts success? Journal of Speech, Language and Hearing Research, 46, 56–67. GRAY, S., 2004Word learning by preschoolers with specific language impairment: predictors and poor learners. Journal of Speech, Language and Hearing Research, 47, 1117– 1132. GRAY, S., 2006, The relationship between phonological memory, receptive vocabulary, and fast mapping in young children with specific language impairment. Journal of Speech and Hearing Research, 49, 955–969. GRAY, S. and BRINKLEY, S., 2011, Fast mapping and word learning by preschoolers with specific language impairment in a supported learning context: effect of encoding cues, phonotactic probability, and object familiarity. Journal of Speech, Language and Hearing Research, 54, 870–884. GUPTA, P., 2003, Examining the relationship between word learning, nonword repetition, and immediate serial recall in adults. Quarterly Journal of Experimental Psychology, 56A, 1213– 1236.
JONES, G., TAMBURELLI, M., WATSON, S. E., GOBET, F. and PINE, J. M., 2010, Lexicality and frequency in specific language impairment: accuracy and error data from two nonword repetition tests. Journal of Speech, Language and Hearing Research, 53, 1642–1655. KAN, P. F. and WINDSOR, J., 2010, Word learning in children with primary language impairment: a meta-analysis. Journal of Speech, Language and Hearing Research, 53, 739–756. LEONARD, L. B., 1998, Children with Specific Language Impairment (Cambridge, MA: MIT Press). MONTGOMERY, J. W., MAGIMAIRAJ, B. M. and FINNEY, M. C., 2010, Working memory and specific language impairment: an update on the relation and perspectives on assessment and treatment. American Journal of Speech–Language Pathology, 19, 78–94. MONTGOMERY, J. W. and WINDSOR, J. 2007, Examining the language performances of children with and without specific language impairment: contributions of phonological short-term memory and speed of processing. Journal of Speech, Language and Hearing Research, 50, 778–797. MUNSON, B., EDWARDS, J. and BECKMAN, M. E., 2005a, Relationships between nonword repetition accuracy and other measures of linguistic development in children with phonological disorders. Journal of Speech, Language and Hearing Research, 48, 61–78. MUNSON, B., KURTZ, B. A. and WINDSOR, J., 2005b, The influence of vocabulary size, phonotactic probability, and wordlikeness on nonword repetitions of children with and without specific language impairment. Journal of Speech, Language and Hearing Research, 48, 1033–1047. RAVEN, J. C., 2003, Raven’s Coloured Progressive Matrices (San Antonio, TX: Pearson). RICE, M. L., BUHR, J. C. and OETTING, J. B., 1992, Specificlanguage-impaired children’s quick incidental learning of words: the effect of a pause. Journal of Speech, Language and Hearing Research, 35, 1040–1048. RICE, M. L., OETTING, J. B., MARQUIS, J., BODE, J. and PAE, S., 1994, Frequency of input effects on word comprehension of children with specific language impairment. Journal of Speech and Hearing Research, 37, 106–122. SKIPP, A., WINDFUHR, K. L. and CONTI-RAMSDEN, G., 2002, Children’s grammatical categories of verb and noun: a comparative look at children with specific language impairment (SLI) and normal language (NL). International Journal of Language and Communication Disorders, 37, 253–271. WIIG, E. H., SECORD, W. A. and SEMEL, E., 2006, Clinical Evaluation of Language Fundamentals Preschool—Second Edition (Sydney, NSW: NCS Pearson).