Auditory frequency discrimination in developmental dyslexia: A meta-analysis

Abstract

Auditory frequency discrimination has been used as an index of sensory processing in developmental language disorders such as dyslexia, where group differences have often been interpreted as evidence for a basic deficit in auditory processing that underpins and constrains individual variability in the development of phonological skills. Here, we conducted a meta-analysis to evaluate the cumulative evidence for group differences in frequency discrimination and to explore the impact of some potential moderator variables that could contribute to variability in effect-size estimations across studies. Our analyses revealed mean effect sizes for group differences on frequency discrimination tasks on the order of three-quarters of a standard deviation, but in the presence of substantial inter-study variability in their magnitude. Moderator variable analyses indicated that factors related both to participant variability on behavioural and cognitive variables associated with the dyslexia phenotype, and to variability in the task design, contributed to differences in the magnitude of effect size across studies. The apparently complex pattern of results was compounded by the lack of concurrent, standardised metrics of cognitive and reading component skills across the constituent studies. Differences on sensory processing tasks are often reported in studies of developmental disorders, but these need to be more carefully interpreted in the context of non-sensory factors, which may explain significant inter- and intra-group variance in the dependent measure of interest.

Keywords: auditory; developmental dyslexia; frequency discrimination; meta-analysis; phonological awareness; reading.

Figure 1

Illustration of the frequency discrimination task types used by the studies included in the meta‐analyses. For each task, the plot depicts the arrangement of tones over time and frequency, with brackets indicating the intervals, that is, response options, typically available to the participant. Typical task designs were as follows: (a) 2‐AFC tasks where participants were asked to identify which of two tones was higher in pitch. (b) Single‐interval tasks where participants were asked to report whether a pair of tones were the same or different in pitch. (c) AXB tasks involve the presentation of three tones, and this design was most typically used in 2‐AFC tasks (as illustrated), with one “reference” tone which never changed in the second position. Participants identified which of the other two tones flanking the reference differed in pitch. A similar 3‐tone design was also used in 3‐AFC tasks where any one of the three tones could differ. (d) 2‐AFC tasks involving two sequences of two tones, which were either identical or, in the target interval, had one tone with a different pitch. (e) Similar to task (d) but with a longer sequence of five tones in each interval

Figure 2

The Hedges' g and associated SE (±1) for each study, ranked and numbered as in Table 1

Figure 3

(a) Covariance between effect‐size for frequency discrimination and effect‐size for phoneme deletion (n = 8). Negative values for phonemic deletion indicate poorer ability on this measure for the dyslexia group compared to the controls. (b) Covariance between effect‐size for frequency discrimination and effect‐size for non‐word reading (n = 21). Negative values for non‐word reading indicate poorer ability on this measure by the dyslexia sample compared to controls. (c) Vertical lines reflect the difference in percent correct on non‐word reading between groups of dyslexics and controls from the same study