JCPSLP Vol 15 No 2 2013

monolingual English-speaking peers on norm-referenced vocabulary tests, possibly due to limited experience with the testing format and/or limited exposure to the vocabulary items. Therefore, SLPs working with CALD clients are instead recommended to use comprehensive assessments incorporating a variety of data sources to describe the children’s language functioning, including interviews, structured observation, criterion-referenced assessments, language samples, and dynamic assessment (SPA, 2009; White & Jin, 2011). Decades of research have revealed the value of language sample analysis (LSA) in describing children’s language abilities and there is a general consensus that LSA should be part of SLPs’ regular assessment protocol (e.g., Westerveld, 2011). When completing LSA, the SLP takes a sample of the child’s language use in a functional context, such as telling a story or conversing with a parent. Language samples are typically recorded, transcribed, and analysed using measures that are intended to generalise to other contexts and serve as a general indicator of a child’s expressive language ability (Miller, Andriacchi, Nockerts, Westerveld, & Gillon, 2012). In addition, multiple measures of language proficiency can be derived from a single sample, providing a rich description of a child’s relative strengths and weaknesses. For example, Westerveld and Gillon (2010) documented three distinct linguistic domains that can be acquired from a single oral narrative language sample (content, grammatical ability, and grammatical complexity) and used to develop profiles of children’s narrative ability. Most research on LSA has been completed with monolingual English-speaking children. Results from this research indicate LSA is effective in distinguishing children with language impairment from their typically developing peers (e.g., Heilmann, Miller, & Nockerts, 2010). Naturalistic and descriptive criterion-referenced assessments, such as LSA, tend to be more appropriate and less biased for CALD populations than their norm- referenced counterparts (White & Jin, 2011). By using LSA, clinicians can minimise some of the format biases observed in norm-referenced language assessments. While most norm-referenced tests require children to perform decontextualised language tasks on demand, language sampling procedures simply require children to engage in naturalistic discourse. Typical language sample contexts, such as conversations and narratives, are present in some form across cultures and should be familiar to most clients (e.g., Bliss & McCabe, 2008). There are many examples of successfully using LSA procedures to assess the oral language skills of CALD populations (e.g., Miller, Heilmann, Nockerts, Iglesias, Fabiano, & Francis, 2006; Ooi & Wong, 2012; Thordardottir, Rothenberg, Rivard, & Naves, 2006). However, most of these studies were completed where the language was spoken by the majority of the population, such as Chinese in Malaysia (Ooi & Wong, 2012), or by a large percentage of the population, such as Spanish in the United States (Miller et al., 2006) and French in Canada (Thordardottir et al., 2006). While these studies provide general support for the use of LSA with CALD children, the results do not fully generalise to the Asia-Pacific region and its wide range of languages and cultures, where clinicians are less likely to Language sample analysis: A powerful tool within the comprehensive assessment

know the language of the client or have strong knowledge of the child’s culture. Our goal is to present a framework for using LSA within the comprehensive assessment of CALD children in Australia, which is summarised in Box 1. Using interview data to plan for collecting language samples The first step in the comprehensive evaluation of CALD children is to consult with those familiar with the child and his/her culture, including family members, teachers, and cultural informants. By using interviews, SLPs have the opportunity to summarise concerns about the child, developmental milestones, and family history (Kohnert, 2010; Restrepo, 1998). Restrepo (1998) demonstrated the critical role of parent report in identifying the presence or absence of a true impairment by testing a group of predominantly Spanish-speaking CALD children using a battery of norm-referenced tests, language sample measures, and parent report. By comparing the sensitivity and specificity of various combinations of tests, Restrepo identified that parent report coupled with a language sample measure was most effective in identifying children with true language impairment. The SLP can also use information from the interview to estimate the child’s relative proficiency in L1 and L2, which will assist in planning the types of direct assessments to administer (Kohnert, 2010). Knowing a child’s relative proficiency across L1 and L2 assists the SLP in interpreting the assessment data. If a child has limited proficiency in a language, we would still expect age-level performance in the dominant language if there is no impairment. If, on the other hand, low performance is observed in both the dominant and non-dominant language, the child would have a true impairment. Gutiérrez-Clellen and Kreiter (2003) developed a questionnaire that document children’s first and second language proficiency and is highly correlated with direct measures of language proficiency. The SLP can further use the interview to identify the child’s familiarity with the planned testing procedures. More accurate performance is observed in CALD children when they complete assessments in familiar formats. For example, Fagundes, Haynes, Haak, and Moran (1998) compared the performance of mainstream children (Caucasian) and CALD children (African-American) from the United States who completed a norm-referenced language test in two formats: standard procedures with line drawings (familiar to Caucasian but unfamiliar to African-American) and adapted procedures embedded into thematic activities (familiar to both Caucasian and African- American). Fagundes et al. (1998) found that there were significant differences between the groups for the standard procedures but no significant differences for the modified thematic procedures, suggesting that the African-American children’s performance was more accurately captured with the familiar task. A variety of methods for eliciting language samples have been described in the literature, including play-based samples, conversations, interviews, narratives, and expository samples (see Westerveld, 2011). During the interview, the SLP can identify the most appropriate conversational partners for the child or determine the child’s experience retelling stories. Collecting and analysing language samples in English For all clients with at least a basic level of proficiency in English, we recommend analysing English samples for use

88

JCPSLP Volume 15, Number 2 2013

Journal of Clinical Practice in Speech-Language Pathology