Analysis of phone-errors in Reading Disabled children

Jiachen Tu

Permalink

https://hdl.handle.net/2142/114918 Copy

Description

Title

Analysis of phone-errors in Reading Disabled children

Author(s)

Jiachen Tu

Issue Date

2021-08-13

Keyword(s)

• phone
• errors
• reading
• diability
• literacy
• disabled children

Language

en

Abstract

Objectives: Reading disability (RD) is a key obstacle in the development of literacy (Hanford, 2018; Wong, 2011; Torgesen, 2004). Studies show that 15-20% of grade-school students have RD, and that this has lifelong consequences for the individual and the wider community. Based on two experimental tasks (SCO and NSCM), the current study examines a key potential source of RD in young children (8-12 years old), namely that due to on deficits in phone-level perception. Our findings on RD phone perception loss is predictable, in our view, based on Harvey Fletcher’s venerable 1921 theory of speech perception (Allen, 2005b,a). The Syllable-Confusion Oddball (SCO) procedure is an unsupervised 3-interval forced choice (3-IFC) closed-set task, to determine which of more than 20 phones have perceptual errors. The Nonsense Syllable Confusion Matrix (NSCM) procedure is a supervised 1-interval open set task, where the subject hears one or more of 20 consonant vowel (CV), and orally reports back what they heard. SCO and NSCM complement each other, providing related information. Design: In the SCO task a sequence of three random nonsense syllable sounds, as spoken by three different talkers (either all CVs or all VCs), is presented to normal-reading control (RC) subjects and reading impaired (RD) subjects, ages 8-12 yrs. Two of the syllables contained the same CV, while the third (the oddball) is different, either in its C or V. The subject’s task was to identify the oddball sound. The NSCM task complements the SCO task by measuring the detailed map of an phone confusions, as a confusion count matrix or as a directed graph. In each trial, listeners hear a single CV or VC, and are instructed to orally repeat the syllable they heard. Two transcribers coded the verbal report. More than ten children, having fully-documented RD, normal hearing, and normal language function, completed the two tasks. Their performance was compared to that of six control (RC) children. On average 1,500 trials were performed on each child, over a two-week period, for both the RC (20-40 trials per syllable) and RD groups (30-40 trials per syllable), for both tasks. The subjects were balanced for gender and ethnicity. Results: The study showed that the proportion of errors was between 3 to 5 times greater for RD listeners (3-50% error) compared with the RC listeners (10% error). The RC subjects show a greatly reduced intra-confusion variance and a similar inter-confusion patterns, allowing for the useful definition of a fictions ‘average-normal’ (inter-subject), which meaningfully characterized the RC confusions. This suggests that increasing the number of RC subjects would simply increase the false-positive (random) errors, without changing the makeup of the average-normal subject (Singh and Allen, 2012). Unlike the RC, the RD subjects were highly idiosyncratic (large individual differences in their confusion pattern errors). Conclusions: It was clear from these data that increasing the number of RD subjects would simply add more idiosyncratic subjects. Given the rather high RD error entropy (Singh and Allen, 2012, Fig. 1), it seems unlikely that patterns of RD error would emerge. Perhaps more important is, that the specific errors from an individual indicate a program of treatment, targeted at those sounds having the largest error. We conclude that RD children have a significant idiosyncratic (inter-confusion) phone-level speech perception problem, captured in the confusion patterns. With the confusion matrix information, it should be possible to generate specific diagnostic feedback to improve phone recognition. What needs to be shown in a follow up study, is that diagnostic feedback improves reading performance

Graduation Semester

2022-10-12T16:29:18-05:00

Type of Resource

text

Owning Collections