Extending the Application of Tense and Agreement Measures: A Reply to Rispoli and Hadley (2018)

Abstract

Purpose

In this reply, we respond to comments on our article “Tracking the Growth of Tense and Agreement in Children With Specific Language Impairment: Differences Between Measures of Accuracy, Diversity, and Productivity.”

Conclusion

The finite verb morphology composite can be disproportionately affected by frequently occurring grammatical forms produced through direct activation. This assumption was one of the reasons we wished to compare this measure to the tense marker total and the tense/agreement productivity score. The latter two measures provide valuable developmental information that is not available from the finite morphology composite. Yet, the finite verb morphology composite shows good diagnostic accuracy and an interpretable pattern of growth and is relatively stable across different sample sizes.

We appreciate the opportunity to respond to the comments provided by Rispoli and Hadley (2018). Although we do not see the goals of our original article (Leonard, Haebig, Deevy, & Brown, 2017) in quite the same way as Rispoli and Hadley do, we find ourselves in agreement with most of the points they raise. The purpose of our study was to examine the growth reflected in three measures of tense and agreement (T/A) morphology during a period when children with specific language impairment (SLI) were in treatment. One measure was the finite verb morphology composite (FVMC); the other two were the tense marker total (TMT) and tense/agreement productivity (TAP) score. The latter two were developed in Rispoli and Hadley's laboratory for use with children between 2 and 3 years of age. Prior to our study, each measure had been supported by considerable evidence, including extensions to the preschool age (e.g., Gladfelter & Leonard, 2013). However, little was known about their pattern of growth in children with SLI—despite the fact that T/A morphology is one of the hallmarks of this disorder. In our view, the three measures represented an excellent combination because they measure quite different things and, therefore, were likely to show differences in their patterns of growth.

Rispoli and Hadley (2018) offer a number of constructive criticisms that we see as falling into three general areas. One is that they see little value in measures such as the FVMC before children show good evidence of sentence diversity. A second is that they disagree with the way (they believe) we interpreted the contribution of direct activation to the children's early use of T/A forms. The third criticism revolved around the way we advocated for control of sample size when interpreting children's TMT and TAP scores. Some more specific points were also raised that we will try to incorporate within these broader criticisms.

Limitations of the FVMC

We heartily endorse Rispoli and Hadley's (2018) point that the FVMC is heavily influenced by T/A productions that might have arisen from direct activation rather than from grammatical encoding. This characteristic was one of the chief reasons we wanted to include measures such as the TMT and TAP score—because these latter measures are far less likely to be influenced by directly activated forms. The first quote from our article that Rispoli and Hadley provides in their letter to the editor (Leonard et al., 2017, p. 3591, “Because the FVMC….”) makes clear that this distinction was an important element of our thinking.

However, we disagree that measures such as the FVMC have questionable value early on. They may constitute flawed measures of T/A diversity and productivity (an assumption that helped to motivate our study), but they nevertheless change with development and—crucial for our purposes—show good sensitivity and specificity in distinguishing children with SLI from their typically developing peers. This measure is not just any measure “supersized to encompass a set of grammatically related morphemes” (Rispoli & Hadley, 2018, pp. 1–2), for these related morphemes deal with tense and agreement. Noun-related composite measures, for example, do not possess the diagnostic accuracy seen in the FVMC (e.g., Bedore & Leonard, 1998). Of course, measures with good sensitivity and specificity can serve to identify children at risk for language impairment without being direct measures of a child's actual language. Nonword repetition measures represent one such example. However, the fact that the FVMC focuses specifically on morphemes of tense/agreement is very likely to be a key factor in its clinical utility.

We were struck by the very similar growth curves of the FVMC at two different sample sizes (100 utterances and the full samples). We are less dubious than Rispoli and Hadley (2018) about the deceleration seen in the quadratic component because it made sense given that growth could slow once children had acquired the more frequent forms produced through direct activation, leaving much of remaining development in the hands of grammatical encoding. Once grammatical encoding became the dominant form of T/A use, more growth would be expected (had we followed the children further), though, as shown by Rice, Wexler, and Hershberger (1998), subsequent growth would likely be slow in these children, and marked by additional periods of deceleration. Finally, we do not characterize the deceleration period in our data as a “plateau.” The means at the final time period were numerically higher than the means from one month earlier, suggesting that growth was continuing, albeit at a slower rate.

Direct Activation and the Role of Frequency

Rispoli and Hadley (2018) interpreted the second quote from our article (Leonard et al., 2017, p. 3598, “Note that although contractions…”) as arguing that frequency independent of direct activation was responsible for the early growth seen in the children's FVMCs. This was not intended on our part. We agree with Rispoli and Hadley that forms can be produced at an early point through direct activation precisely because they are so frequent. The point we evidently did not state clearly enough was that, in principle, a form ordinarily produced by a child through direct activation might not be used frequently by that child in a particular language sample and therefore would not have the impact on the FVMC that we observed in our language samples. In fact, we believe that forms of this type (e.g., that's, it's) that we observed were produced through direct activation. Our wording could have been better to make this point.

Language Sample Size

A third concern was our suggestion that comparisons across children using the TMT and TAP scores should use a consistent spontaneous language sample size. Rispoli and Hadley (2018) employed samples that were 60 min in duration in their work with younger children. We do not believe that a consistent duration would solve the issue, as children differ in how much they say in any time frame, resulting in samples from some children being larger than samples from other children. However, we do agree strongly with their point that samples should be obtained in a consistent manner and in a way that allows for a wide range of T/A forms in diverse sentences. Our samples were not planned with the TMT and TAP score in mind, so our samples did not have all of the positive features prescribed by Rispoli and Hadley. On the other hand, we were encouraged by how consistently these measures showed growth over the study period, even when all samples were of the same size. Ours was not a “call to standardize” the TMT and TAP score on 100-utterance samples; we used that sample size because it fit our available data and represented a size that might be employed in a clinical setting.

Rispoli and Hadley (2018) expressed a laudable concern that their measures be used in a thoughtful manner. Their original intent with these measures was to “quantify developmental differences in the onset of tense and agreement marking” (Rispoli & Hadley, 2018, p. 2). Yet, with appropriate care, these measures can be applied successfully to slightly later points in development. In our view, the continued developmental progression shown by the TMT and TAP scores well into the preschool years coupled with their diagnostic accuracy make them an especially important contribution to the field and an ideal focus for future clinical research.