TO THE EDITOR

Comments on Tyler, A. A., & Lewis, K. E. (2005). Relationships among consistency/variability and other phonological measures over time. Topics in Language Disorders, 25(3), 243-253.

Tyler and Lewis' study investigated the relationship between selected phonological measures, including measures of variability/consistency, in children with speech and language disorders. They evaluated response to intervention delivered twice weekly for 24 weeks. Both language and phonological structures were targeted in therapy. The article is important because it reflects increasing awareness that children with developmental phonological disorders can have very different surface error patterns. The article also emphasised the need for research to explore the implications of different speech error patterns for clinical practice.

Tyler and Lewis drew two important implications for assessment and intervention. First, they proposed that assessment procedures for severity (Percentage Consonants Correct [PCC]; Shriberg & Kwiatkowski, 1982), variability/consistency (Error Consistency Index [ECI]; Tyler, 2002), and whole-word complexity (Phonological Mean Length of Utterance [PMLU]; Ingram & Ingram, 2001) measure the same construct. They suggested that PCC is the "measure of choice" for reasons of familiarity and validity.

Tyler and Lewis' evidence, drawn from the measures they used, does not, however, discriminate between phoneme variability/consistency and whole-word consistency.

* PCC measures the accuracy of individual phonemes irrespective of lexical item or phonetic context so that [tat] [pat] and [at] for cat are all considered to be equally in error.

* ECI measures the total number of substitutions a child makes in each word position for each of the 23 consonant phonemes. This measures consonant phoneme variability. If the word slide were produced consistently as [gI:aI] and sun were produced consistently as [d^n], the ECI would classify /s/as being produced inconsistently. In contrast, whole-word ECI measures the total number of substitutions a child makes in each word position for each of the 23 consonant phonemes. This measures consonant phoneme variability. If the word slide were produced consistently as [gI:aI] and sun were produced consistently as [d^n], the ECI would classify /s/ as being produced inconsistently. In contrast, whole-word consistency measures a child's production of a lexical item so neither slide nor sun would be considered inconsistent if the child produced the items the same way each time. A whole-word consistency measure examines different productions of the same word, in terms of consonants, vowels, and syllables. For example, in our data, a 7-year-old boy's production of slide as [gI:ajkI:] [diw^] [j[alpha]] was inconsistent with respect not only to consonants but also vowels and syllable structure (Dodd, Holm, Crosbie, & McIntosh, in press). These examples indicate that the ECI has limitations as a measure of consistency. It does not capture consistency of whole-word production.

* PMLU is a measure of whole-word complexity calculated by assigning one point for each phoneme represented and an additional point for each accurate phoneme in a sample. This measure, like PCC, is a measure of severity.

Tyler and Lewis found high positive correlations between these three measures. This is not surprising given that all three measures assess accuracy at the phoneme level. In contrast, our recent research (Crosbie, Holm, & Dodd, 2005) involved two groups of children who were matched for age and severity (using PCC) but who differed significantly on a measure of whole-word consistency. The children with inconsistent speech disorder had a mean PCC of 44.6 (SD = 11.4) and a mean inconsistency score of 51.4 (SD = 8) on the Diagnostic Evaluation of Articulation and Phonology (DEAP; Dodd, Hua, Crosbie, Holm, & Ozanne, 2002). The children with consistent speech disorder had a mean PCC of 49.6 (SD = 13.5) and a mean inconsistency score of 28.5 (SD = 8.4). Severity and consistency of production are not necessarily correlated when whole-word consistency of production is measured.

Systematic experimental comparisons of children who make inconsistent errors (defined by measures of whole-word consistency) with those who make consistent errors have been done (Dodd, Holm, Crosbie, & McCormack, 2005). Using a psycholinguistic assessment approach, the two groups of children had specific profiles of strengths and weaknesses. The findings indicate that different surface speech error patterns reflect different underlying deficits that cannot be identified by a gross measure of severity like PCC. Thus, we take issue with Tyler and Lewis' first clinical implication; if variability is measured as the tendency to repeat individual words inconsistently, it is not at all equivalent to PCC and other measures of speech sound production.

Tyler and Lewis' second implication addressed intervention: children with different surface speech error patterns represent "two ends of a continuum" of disorder that respond equally well to behavioral interventions. The evidence upon which this implication is based is problematic for a number of reasons:

* Group Selection: The two groups were selected in terms of their performance on the ECI measure, resulting in a very large difference between the two groups in severity as measured by the PCC. Although qualitative rather than quantitative measures of improvement were reported, the difference in severity still confounds the comparison. While the groups reflect differences in phoneme variability as measured by the ECI, they may not reflect two different groups in terms of their consistency of word production. In this sense, they are very different from the children with inconsistent speech disorder as described by Dodd (1995, 2005) and Dodd and Bradford (2000). Consequently, Tyler and Lewis' intervention study did not assess whether children with inconsistent speech disorder characterized by whole-word inconsistency respond to treatment using the same behavioral interventions as children who make consistent errors. Systematic investigation of the efficacy of different intervention approaches for consistent and inconsistent phonological disorders (based on whole-word consistency) indicates that children respond differentially to an intervention approach (Crosbie, Pine, Holm, & Dodd, in press; Dodd & Bradford, 2000; Holm, Crosbie, & Dodd, 2005). One study that directly compared two intervention approaches found that children who made whole-word inconsistent errors responded better to a core vocabulary approach than to a phonological contrast approach. Their whole-word consistency and accuracy of production significantly increased. In contrast, children with whole-word consistent errors responded better to the phonological contrast approach (Crosbie et al., 2005).

* Characteristics of participants: The data reported by Tyler and Lewis were generated from a study designed to examine intervention outcomes for children with co-occurring speech and language difficulties. This is an important clinical population; however, it is difficult to differentiate the impact of language ability. All of our previous studies, and most studies of phonological disorder (e.g., Forrest & Elbert, 2001; Gierut, 2001), have specifically excluded children who have co-occurring receptive language problems to avoid this confound. It is difficult to draw implications about phonology when the population studied may not be representative of children who have an isolated speech disorder. Furthermore, the preintervention capabilities of the two groups are not described beyond PCC and ECI. It is important to know whether children in the two groups were matched for receptive and expressive language ability.

* Content and scheduling of intervention: It is not clear that the children in the two groups compared (i.e., high and low ECI score) received identical intervention. Tyler and Lewis (Tyler, Lewis, & Welch, 2003) emphasised that different intervention sequences did not lead to significant differences in phonological change. This may not be true, however, for the smaller groups who were selected on the characteristics of their phonological disorder (ECI). Furthermore, it is not clear whether the sequence of therapy received was balanced between the groups.

Tyler and Lewis argued that it is important that the measures used to describe surface speech error patterns capture different aspects of the phonological system. We propose that measuring consistency of individual phoneme production across words (as performed by Tyler & Lewis) provides different information from a measure of whole-word consistency of production when the same words are sampled more than once in the same linguistic context (as performed by Dodd & colleagues). Tyler and Lewis' intervention study was originally designed to answer a question about the effect of different sequences of therapy focusing on syntax and phonology. Consequently, the post hoc analyses seem confounded by a number of factors that were not controlled from the outset. It seems imprudent then to draw such strong clinical implications about assessment and intervention for children who make inconsistent speech errors. Neither of Tyler and Lewis' clinical implications is wholly justified by the evidence they present.

Barbara Dodd

Alison Holm

Sharon Crosbie

Perinatal Research Centre The University of Queensland, Australia

REFERENCES