JCPSLP VOL 15 No 1 March 2013

start of the subsequent session, preschoolers engaged only in sentence build-up (“Put it all together”), where the subject+verb+object was combined to produce a grammatically correct sentence (sentence point). Sentence- breakdown was not required once the 80% criterion was achieved. Random observations (20% of sessions) by observers using a checklist of critical intervention elements (e.g., the intervention procedure, identification of the interventionist, the session number, group type, length of the session, and the techniques used during the session) revealed that intervention fidelity was maintained 100% of the time. See Appendix for a sample intervention routine. For computer-assisted intervention, expressive grammar training was completed using My Sentence Builder . This program is embedded within a syntactic slot-filler approach with visual representations for semantic and grammatical elements provided using picture support. Using a drill-play approach with modelling and repetition, preschooler–SLP dyads moved from screen-to-screen, selecting components during sentence-breakdown. The SLP took preschoolers to the sentence-creation screen first and told them they would be “making up” things about boys and girls. The dyad progressed through the subject, verb, object selection screens to choose the subject, verb, object for placement in the sentence-box at the bottom of the screen. Preschoolers were then asked to “put it all together”. Following a correct production, preschoolers were taken to the sentence-selection and animation-production screens. The slow, deliberate, and sequential selection used with computer-based visual representations was the key intervention difference between computer-assisted and table-top intervention. For table-top intervention, preschoolers engaged in clinician–client dyads for the same multi-step intervention procedures, this time using typical table-top materials (e.g., books, felt, or paper dollhouse objects) to demonstrate the semantic elements within the same drill-play activities. Emphatic stress was included to increase the salience of sentence components (grammatical and semantic) in contrast to the computer-based syntactic slot-filler approach. This technique involved the SLP verbally stressing sentence components during the multi- step procedure. In comparison to computer-assisted intervention, consistent visual support demonstrating grammatical elements was not provided. Instead, visual support was provided using table-top materials for semantic elements. Preschoolers in the waitlist control group did not receive expressive grammar intervention from the SLP during the study. At the end of the study, intervention was offered. Design and analysis A pre-post-follow-up design was employed. The secondary analysis of the Washington et al. (2011) data was completed using two mixed model multivariate analyses of variance (MANOVAs) with pre-set alphas ( p < .05). Effect sizes “an estimate of the effect of intervention” (Portney & Watkins, 2009, p. 373), represented by eta squared ( N 2 ) and partial eta squared ( N p 2 ), were also reported. The first MANOVA compared the three groups (computer-assisted, table-top, no intervention: between-subjects factor) for DSS and MLU change scores (dependent variables) for the 3-month gain1, 3-month gain2 and 6-month gain (i.e., gain period, within-subjects factor). The second MANOVA compared the three groups for DSS per cent error rates (number of incorrect attempts for personal pronoun, main verb and number of utterances not awarded a sentence

grammatical categories established by Lee (1974) that indicate grammar development and complexity in young children (i.e., indefinite pronouns/noun modifier, personal pronouns, main verbs, secondary verbs, negatives, conjunctions, interrogative reversals, wh-questions). Growth beyond the 10th percentile (pre-test performance) to “within normal limits” represented clinically meaningful growth (i.e., acceleration) in spontaneous grammar skills. Since participants were 3- to 4-years of age ( M = 4;4), grammatical performance was compared to the DSS point growth expected over a 6-month period for typical 4-year-olds (Lee & Canter, 1971). The 50th percentile was used as the expectation for normal developmental change, representing a 0.76 point-gain for this age group (Lee & Canter, 1971). Clinically meaningful DSS growth was established at or greater than the 0.76 DSS point-gain for change between pre-intervention to post-intervention (3-month gain1), post-intervention to 3 months post-intervention (3-month gain2), and pre- intervention to 3 months post-intervention (6-month gain). To establish DSS scoring reliability, 10% of language samples were randomly chosen for analysis by graduate SLP students. Inter-rater reliability for DSS, including point- to-point comparisons for word transcription, appropriate DSS sentences, category, and scoring was 91.3%, 97.2%, 90.8%, and 90.8%, respectively. Preschoolers’ mean length of utterance (MLU; Brown, 1973; Miller, 1981) was also calculated for the same language samples. MLU is another useful measure of grammatical morphology that offers information about use of morphemes and developmental change, but is limited in capturing changes in grammatical complexity (Goffman & Leonard, 2000). Due to this limitation, (a) MLU change scores were calculated to determine if gains in use of morphemes co-occurred with gains in grammatical complexity as measured by the DSS and (b) the 0.76 criterion applied to the DSS change scores was not applied to the MLU change scores. At pre-test, a univariate ANOVA revealed non significant between-group differences in MLU ( p = .140). An analysis of DSS per cent error rates for number of incorrect attempts for the personal pronoun , main verb , and number of utterances not awarded a sentence point , representing targeted grammatical categories, was also completed. There were no significant differences on per cent error rates between the three groups at pre- intervention for personal pronoun ( p = .501), main verb ( p = .072), and sentence point ( p = .081). Errors in these categories decrease with time for typically developing preschoolers (Lee, 1974) and show improvement with intervention for children with language impairment (Fey, Cleave, Long, & Hughes, 1993; Lee, Koenigsknecht, & Mulgern, 1975). Intervention Preschoolers receiving intervention in the Washington et al. (2011) study participated in 20-minute sessions once weekly for 10 weeks with an SLP, typical of the intensity and frequency of intervention services under the government-funded initiative. The following procedure was used: (a) a 2-to-7-minute practice block introduced the routine; (b) sentence-breakdown was used to individually elicit sentence components (subject-noun phrase, verb, object-noun phrase) utilising the following questions, “Who do you want to play with?” (subject), “What is s/he doing?” (verb), and “What does s/he want to play with?” (object); (c) sessions followed the same procedure until 80% accuracy over two consecutive sessions was achieved; and (d) at the

9

JCPSLP Volume 15, Number 1 2013

www.speechpathologyaustralia.org.au