JCPSLP Vol 15 No 2 2013

Table 1. Performance of a child with language impairment pre- and post-intervention compared to a database of speakers with typical language development Time 1 Time 2 Current age: 7;0 Current age: 7;3 Database: NZ Retell Database: NZ Retell 87 database participants 60 database participants Time 1 Time 2 Transcript length Score ± SD Score ± SD Total utterances 10 –0.92 12 –0.78 Total words 97 –0.36 98 –0.5 Elapsed time 1.77 0.15 2.20 0.61 Syntax/morphology MLU in words 7.70 0.83 7.17 0.42 MLU in morphemes 8.00 0.66 7.33 0.10 Semantics Number different words 44 –0.58 49 –0.48 Number total words 77 –0.61 86 –0.56 Mazes & abandoned utterances Number maze words 21 0.85 13 0.11 % maze words 21%* 1.55 13% 0.55 Verbal fluency & rate Words/minute 54.91 –0.74 44.55* –1.12 Within-utterance pauses 3** 4.57 2** 3.59 Between-utterance pauses 1 0.35 0 –0.55 Omissions & error codes Omitted words 0 –0.36 0 –0.30 Omitted bound morphemes 0 –0.15 0 –0.18 Word-level errors 6** 3.82 2 0.68 Utterance-level errors 3** 5.01 1* 1.67 * 1 SD below age-matched peers ** 2 SD below/above age-matched peers Note: Time 1 summarises baseline performance and Time 2 summarises data after an intensive intervention. “Score” columns summarise children’s performance and “± SD” represent the difference (in standard deviations) between the child and mean scores of children in the database. NZ retell was elicited using the Westerveld and Gillon language sampling protocol (http://www.griffith.edu.au/health/school- rehabilitation-sciences/staff/dr-marleen-westerveld/language-sampling-and-other-resources).

mainstream Americans or mainstream New Zealanders), and length of the sample. While there is not a database specific to Australian speakers, Westerveld and Heilmann (2012) documented that measures from American and New Zealand samples were not significantly different; we expect that there would also be minimal differences when comparing those databases to samples from mainstream monolingual Australian children. Measures from the child’s sample can then be compared to the normative comparison group and tracked over time. Table 1 shows an example with a 7-year-old child with a language impairment who completed two separate narrative retells. In the first columns (Time 1), it is evident that most aspects of the child’s productive language were in the low-normal range compared to his age-matched peers. His lowest scores were associated with word-level and utterance-level errors. Further examination of the language sample revealed that he had significant difficulty with pronouns, past tense, and prepositions, which were addressed in an intensive intervention. The Time 2 columns summarise measures from a second narrative retell that was collected three months later. After the intervention, a marked reduction was observed in both word-level and utterance-level errors.

Language Transcripts (SALT; Miller, Gillon, & Westerveld, 2012), Computerized Language Analysis (CLAN; MacWhinney & Snow, 1985), and Parrot Easy Language Sample Analysis (www.parrotsoftware.com). Differences between programs relate to the usability of the software, availability of customer support, and fee for use. After entering a transcribed and coded transcript, the software programs automatically and accurately generate multiple measures to describe children’s language skills, including measures of linguistic form (e.g., mean length of utterance, use of obligatory morphemes), content (e.g., number of different words), and use (e.g., percentage of words in mazes). Software programs typically summarise children’s language sample measures in a chart that can be inserted into a clinical report and archived in the child’s file (see Table 1 for an example). Each software program has unique features that facilitate interpretation of the language sample data. With the SALT software, for example, SLPs have the opportunity to compare their client’s performance with typical speakers in one of the multiple databases. Clinicians can customise their comparisons based on the type of sample collected (e.g., conversation, narrative retells), population (e.g.,

91

JCPSLP Volume 15, Number 2 2013

www.speechpathologyaustralia.org.au