Reliability of the auditory brainstem responses to speech over one year in school-age children: A reply to Drs. McFarland and Cacace

In Hornickel, Knowles, and Kraus (Hornickel et al., 2012), we provide evidence that the speech-evoked auditory brainstem response (speech-ABR) in typically-developing eight to thirteen year-old children is stable over the course of one year. Our comprehensive study of the reliability of multiple speech-ABR measures yielded reliabilities ranging from 0.11 to 0.82 (reprinted in Table 1 column 1). While reliabilities higher than 0.7 are preferred for clinical diagnosis (Nunnally, 1959), suitable reliability for group analyses is 0.6 (Salvia and Ysseldyke, 2004). Drs. McFarland and Cacace have raised concerns that the reliabilities reported in Hornickel et al. (2012) were low which, given the large number of measures reported, increased the likelihood of false alarms if used diagnostically. While we understand their concerns, our intent was to be inclusive in our reporting of reliability and we acknowledge that when the situation demands it, such as in clinical use, one would choose a small number of measures with the highest reliability. The reliability of the speech-ABR will probably never reach the levels seen for click-evoked brainstem responses for a variety of reasons, including the comparatively richer, spectrotemporallydynamic information present in speech compared to clicks and the experience-dependent plasticity of the speech-ABR. However, it is also likely that both the age of the subjects and the retest interval contributed to test–retest variance in Hornickel et al. (2012). The data were collected over a one year test–retest interval, in contrast to the two week interval typical for diagnostic behavioral assessments (McGrew and Woodcock, 2001; Salvia and Ysseldyke, 2004; Torgensen et al., 1999; Wagner et al., 1999). As may be expected, reliability of behavioral measures decreases as the retest interval increases (McGrew and Woodcock, 2001). Previous studies of speech-ABRs used test–retest intervals only a fewmonths in length (Russo et al., 2004; Song et al., 2011). Moreover, ABRs of children are known to be more variable than adults both between and within subjects (Lauter and Oyler, 1992), and click-evoked reliability decreases with longer test–retest intervals (Tusa et al., 1994). Thus, a longer test–retest interval, a younger subject population, and a complex stimulation all contribute to test–retest variability. Due to our modest sample size (n 1⁄4 26), we elected to assess reliability using Spearman’s correlations. While we agree with Drs. McFarland and Cacace that rank ordering the data for reliability estimates might be more appropriate for rank ordered clinical measures, Spearman’s correlations are the more conservative estimate. Reliabilities are largely the same when calculated as Pearson’s correlations (compare Table 1, columns 1 and 2). In fact, a greater number of speech-ABR measures show significant