RESPONSE

To Dodd, Holm, and Crosbie regardingTyler, A. A., & Lewis, K. E. (2005). Relationships among consistency/variability and other phonological measures over time. Topics in Language Disorders, 25, 243-254.

In their letter to the editor, Dodd, Holm, and Crosbie take issue primarily with the measurement of consistency/variability used by Tyler and Lewis (2005). First, the conclusion Dodd et al. draw that the error consistency index (ECI) does not capture whole-word consistency, or intraword variability, as it has been called in the developmental phonological literature (Ferguson & Farwell, 1975), is entirely correct. The measure was not meant to capture consistency of repeated productions of a lexical item, but rather the consistency/variability of errors for individual target phonemes, across contexts, for the phonological system as a whole. Thus, it is important to distinguish the way in which consistency/variability is viewed and measured.

Currently, there are numerous and equally differing perspectives on ways to measure what is called consistency, inconsistency, and variability for children with speech sound disorders (Betz & Stoel-Gammon, 2005). Dodd et al.'s letter simply highlights these different perspectives regarding the definition and measurement of consistency/variability. Although it is important to elicit repeated productions of the same word, it is just as important to consider errors in the production of a single phoneme across words, as well as broader measures of error type and frequency. Measures involving error substitutions or changes in the production of a particular phoneme across words are quite common among the array of measures proposed (Betz & Stoel-Gammon, 2005; Ingram, 2002; Isermann, 2001; Shriberg, Allen, McSweeny, & Wilson, 2001). How one defines variability, operationalizes its measurement, and selects participants demonstrating the construct will necessarily influence findings involving measurement of the variable over time, as well as in response to treatment. The measure used by Tyler and Lewis was intended to capture the consistency/variability of phoneme errors; as such it is not unexpected that it was correlated to Percentage Consonants Correct (PCC), while a measure of a different type of variability, intraword variability, would not be correlated to PCC. A more important implication from the different measures of (in)consistency in the work of Dodd and colleagues and Tyler and Lewis is that we do not yet know how different measures of consistency/variability relate to one another

Second, Dodd et al. suggest that the data do not justify drawing the clinical implication that the groups investigated by Tyler and Lewis responded equally well to the same intervention. Dodd et al. note that the group exhibiting variable error substitutes was not reflective of their group with "inconsistent speech disorder" (i.e., whole-word inconsistency), which is precisely the case. The participants in the Tyler and Lewis study were specifically selected for their below-normal abilities in both speech and language. This group is representative of a large subgroup of the preschool population with speech disorder (Shriberg, Lewis, Tomblin, McSweeny, Karlsson, & Scheer, 2005). It is apparent that not only are the measures different in the work of Dodd and colleagues but the participant populations differ as well. Participants in the study by Crosbie, Holm, and Dodd (2005) averaged 6 years of age, did not have expressive language deficits, and were selected for their whole-word consistency or inconsistency. The Tyler and Lewis participants averaged 4 years 2 months, had expressive language deficits, and were grouped by their error consistency across all target phonemes sampled representatively. Participants were randomly assigned to the interventions that differed only in how their similar goals were cycled over a 24-week period, as noted both in the Tyler and Lewis article and in other sources (Tyler, Lewis, Haskill, & Tolbert, 2002, 2003). Dodd et al. appear to be suggesting that systematic group differences in language abilities and/or assignment to treatment sequence may have been responsible for the groups' similar response to treatment: For this to be the case one would have to imagine these systematic differences were sufficient to contravene the finding of similar response to treatment across the independent variable of consistency/variability. This seems unlikely. However, we did fail to clarify that there were no group differences in language ability and that treatment sequences were balanced across the two groups. Thus, the analyses are not confounded by methodological flaws; rather the consistency/variability measure used to group participants, as well as the participant characteristics, are entirely different in the work referred to by Dodd and colleagues as compared to that of Tyler and colleagues.

We agree that different subgroups may need different interventions. Our data, however, clearly showed that when we compared two groups differing in error consistency and severity, they responded similarly to a speech--language intervention. Quantitative outcome measures of PCC change and increased consistency for treatment targets (both quantitative measures) were remarkably similar, wholly justifying the conclusions for this participant group. These results cannot be extended to children with only a phonological disorder who have been grouped by their whole-word inconsistency. We make no claims that we have identified a unique subgroup or that the participants had an inconsistent disorder as defined by Dodd and colleagues. In fact, identifying the relevant and unique subgroups within the heterogeneous population of children with speech sound disorder and testing different interventions for them remains an area of fertile research.

REFERENCES