Cognitive and linguistic profiles of children with SLI

Several classification systems have so far tried to account for the heterogeneity of children diagnosed with SLI (Bishop, 1997; Botting & Conti-Ramsden, 2004; Gérard, 1993; Parisse & Maillart, 2009; Rapin & Allen, 1988; van Weerdenburg et al. 2006, see 1.6.2). With regard to these investigations, several critical comments can be made. First, some of the assessment batteries appear too ‘limited’, neglecting some of the language (e.g. measurements of discourse) or the cognitive domains (e.g. information processing or working memory abilities). Second, populations of each investigation depend on the specific definition each author grants to the pathology and therefore all investigations do not concern the same populations. Third, methodologies also differ, some of them are empirical investigations without appropriate statistical analyses, some of them are based on clinical judgments whereas others are based on test scores. In the context of a significant heterogeneity and different operational definitions, we conducted a study (chapter 3) to characterize the cognitive and linguistic profiles of children diagnosed with SLI at a Reference Centre for Learning Disorders in France (Zourou, Magnan, Ecalle & Gonzalez-Monge, 2009). The basic purpose of this study was to determine 1/ the linguistic profiles of children diagnosed with SLI and 2/ if different linguistic profiles are associated with different cognitive profiles (based on scores at the WISC-IV battery).

We examined 208 medical files of children assessed and diagnosed at the Reference Center for Learning Disorders of Lyon in France between 2007 and 2009. We created a database of their scores on all test batteries used by staff professionals²³. Children whose primary diagnostic was other than SLI (e.g. dyspraxia) and children whose assessment were incomplete or were not assessed with the WISC-IV battery (Wechsler, 2005) but with some previous battery (e.g. WISC-III) were excluded. Data analysis concerns 40 medical files of children diagnosed with SLI. Measures included in data analysis were the assessment of global intelligence (WISC-IV, Wechsler 2005) and productive and receptive vocabulary abilities (tests ELO, Khomsi, 2001; or N-EEL, Chevrié-Muller & Plaza, 2001), as there were the only measures in which the entire population was assessed.

Overall, performances of the population on the language measures (Figure 2) were characterized by great heterogeneity both on productive (z-scores from -2.42 to 0.57) and receptive vocabulary (z-scores from -1.96 to 0.82). Performances on the WISC-IV battery were also characterized by great heterogeneity. However, z-scores on all four indices (composite scores) of the WISC-IV, verbal comprehension (VCI), perceptual reasoning (PRI), working memory (WMI) and processing speed (PSI), were negative suggesting low performances on these four domains. On the WMI, we obtained weak z-scores (M=-1.48, SD=0.83) for the majority of the population (Figure 3). This observation strengthens the idea already suggested by various authors that working memory (WM) could be used as a valuable clinical marker of SLI. The correlation matrix revealed low coefficients between z-scores on vocabulary (productive and receptive) and the four indices of WISC-IV, suggesting a relative independency between performances in language and intelligence (Table 3). To examine the heterogeneity of the population, we performed a cluster analysis (K-means clustering) on the two language parameters (receptive and productive vocabulary). Three groups of children emerged; cluster A presenting a mixed type disorder with important difficulties on both productive and receptive vocabulary, cluster B presenting an expressive type disorder and cluster C with normal scores on the vocabulary measures (Figure 4). Repeated ANOVA’s on the four indices and the ten core subtest of the WISC-IV revealed, however, no Group effect. In other words, the three clusters (A, B and C) did not present significant differences at the WISC-IV assessment. A more qualitative analysis, however, revealed the weak aspects of every cluster on the ten core subtests suggesting some associated cognitive difficulties. In fact, except for poor performances on Digit Span and Letter-Number Sequencing observed in all three clusters, cluster A performed poorly also on Block Design suggesting motor coordination difficulties and cluster B on Coding suggesting processing speed difficulties.

This investigation has been a first attempt of an ‘ecological’ classification study based on language and general intelligence measures of children consulting for SLI France. Factors taken into account were the severity and the profile (receptive vs productive vs mixed) of language difficulties and the presence of associated cognitive difficulties. Despite, its limits (limited number of participants and of language factors finally used in the statistical analysis) several conclusions can be drawn of this study. The first one is the heterogeneity of the language deficits both in profile and in severity (some children presented difficulties on both productive and receptive vocabulary and some obtained normal scores on both of these measures) observed in children with SLI. The second is that children with SLI present associated cognitive difficulties in phonological WM (VWM) but also in motor coordination and in speed of processing. An interesting finding is that children presented poor performances on measures of phonological WM and this independently of their specific linguistic profile (cluster A, B or C). In the light of the above, we evoke two things. The first is the need of a new classification investigation taking into account both the dynamic linguistic profiles of children with SLI (see for example, Botting & Conti-Ramsden, 2004, 1.6.2) and their cognitive profile (see for example, Parisse & Maillart, 2009, 1.6.2). In fact, ‘certain combinations of linguistic and non-linguistic impairments could be used as more reliable markers for differentiated diagnoses or to improve predictions concerning the children’s development’ (Parisse & Maillart, ib., p.112). The second is the importance of a twofold assessment of both language and global intelligence abilities. However, we believe that traditional psychometric assessment, which only reflects the child’s current performances (static level of achievement), needs to be reinforced by additional dynamic assessment procedures revealing the child’s learning potential (Krassowski & Plant, 1997; Hasson & Botting, 2010).