Peter Flipsen

Measuring the intelligibility of conversational speech in children

Conversational speech is the most socially‐valid context for evaluating speech intelligibility, but it is not routinely examined. This may be because it is difficult to reliably count the number of words in the unintelligible portions of the sample. In this study four different approaches to dealing with this problem are examined. Each is based on the assumption that it is possible to perceive syllables in unintelligible strings even when the target words are unknown; these unintelligible syllables can then be used to estimate the number of unintelligible words in these samples using at least four different approaches. Preliminary data are presented for each of the four approaches based on conversational speech from two convenience samples including 320 children with normal (or normalized) speech and 202 children with speech delay. Differences among the four approaches are discussed.

Acoustic phenotypes for speech-genetics studies: toward an acoustic marker for residual /s/ distortions

A companion paper addresses the need for phenotype markers for speech-genetics studies and provides reference data for US English rhotics that can be used for phenotype research. The present paper uses these reference data to derive and test an acoustic marker to discriminate the residual /з/ distortions of adolescents with two speech disorder histories. One speech disorder history includes significant speech delay; the other history is a speech disorder limited to only speech sound distortions of /r/, /з/ and/or /e/. The first subtype of speech delay is posited to be genetically transmitted, whereas the origins of the second subtype are posited to be associated with shared and non-shared environmental variance. Speech samples from 84 9 to 17-year-old speakers were divided into four groups based on speech history and speech errors at assessment. Group 1 children had prior speech delay and residual rhotic distortions, Group 2 children had only prior and residual rhotic distortions, and children in the two control groups had normal or normalized speech. Statistically significant logistic regression models indicated that an acoustic marker successfully discriminated residual derhotacized /з/ tokens produced by speakers in Group 1 from residual derhotacized /з/ tokens produced by speakers in Group 2. The marker was a z score less than 6.0 for Formant 2 subtracted from Formant 3 (i.e. zF3-F2<6.0) as measured at the constriction interval for /з/ targets. Sensitivity (percentage of correctly identified derhotacized /з / tokens from Group 1 speakers) for the acoustic marker was 85%. Specificity (percentage of correctly rejected derhotacized /϶/ tokens from Group 2 speakers) was 79%. Discussion considers methodological, phonological, and genetic perspectives that might account for the articulatory differences in the residual /з/ distortions of adolescents with the two different speech histories.

Acoustic phenotypes for speech-genetics studies: reference data for residual /з/ distortions

An eventual genetic account of at least one subtype of child speech disorders may require the use of acoustic markers to phenotype children and family members. Acoustic markers have the potential for sensitivity and specificity that are not available using auditory-perceptual procedures such as phonetic transcription. Prior reports on the use of acoustics in speech-genetics research have described methods to complete large-scale acoustic analyses, addressed relevant technical and linguistic sampling issues, and provided an acoustic reference database for /s/ production in adolescents. The present paper addresses additional methodological issues and demonstrates how reference data might be used for speech-genetics studies of another class of frequently persisting speech errors - distortions of the English rhotics /r/ and /з/. Findings support the need to adjust acoustic reference data by age and gender, and to subgroup reference data by rhotic phoneme and phonetic context. We describe a z score procedure t...An eventual genetic account of at least one subtype of child speech disorders may require the use of acoustic markers to phenotype children and family members. Acoustic markers have the potential for sensitivity and specificity that are not available using auditory-perceptual procedures such as phonetic transcription. Prior reports on the use of acoustics in speech-genetics research have described methods to complete large-scale acoustic analyses, addressed relevant technical and linguistic sampling issues, and provided an acoustic reference database for /s/ production in adolescents. The present paper addresses additional methodological issues and demonstrates how reference data might be used for speech-genetics studies of another class of frequently persisting speech errors - distortions of the English rhotics /r/ and /з/. Findings support the need to adjust acoustic reference data by age and gender, and to subgroup reference data by rhotic phoneme and phonetic context. We describe a z score procedure that accommodates these methodological needs. A companion paper uses these data to develop an acoustic phenotype marker of residual /з/ distortions for speech-genetics research.

A diagnostic marker for speech delay associated with otitis media with effusion: the intelligibility-speech gap.

The goal of this study was to determine if notably reduced intelligibility is a potential diagnostic marker for children with speech delay and histories of early recurrent otitis media with effusion (SD‐OME). Intelligibility was assessed in one 5–10 minute conversational speech sample from each of 281 speakers. The OME histories of 148 of these children with normal speech acquisition were described in two prior reports. OME histories of 85 additional children with speech delay were obtained from case history reports. For both groups, the children with positive OME (OME+) histories had significantly lower intelligibility scores but significantly higher speech production scores than children with negative OME (OME−) histories. Findings for a diagnostic marker to discriminate speech delayed children with OME+ versus OME− histories were promising, considering that the data were obtained retrospectively and did not include audiological information characterizing childrens concurrent fluctuant hearing loss. The formula for the diagnostic marker, termed the Intelligibility‐Speech Gap, was identified by a machine learning routine. Diagnostic accuracy findings for the marker were as follows: positive predictive value=74%, negative predictive value=86%, sensitivity=79%, specificity=83%, positive likelihood ratio=4.6 and negative likelihood ratio=0.25. Discussion considers speech processing perspectives on the source of the intelligibility‐speech gap in children with suspected SD‐OME, and methodological perspectives on its development as a diagnostic marker of one etiological subtype of speech delay.

Syllables per word in typical and delayed speech acquisition

A number of authors have presented data on the word length (measured in syllables) in the spontaneous speech of children across the developmental period. These data suggest a developmental trend of increasing length with age. The current study sought to examine this possibility in more detail. Conversational speech data from 320 children with normal (or normalized) speech confirmed that the number of syllables per word in conversational speech increases significantly from age 3–8 years. Data from the conversational speech of 202 children with speech delay however showed no such trend. Reasons for the differences between the two groups are discussed.

Emergence and Prevalence of Persistent and Residual Speech Errors

This report reviews the existing literature on the emergence and prevalence of speech sound errors in older children and adults. Findings from several different sources suggest that 1 to 2% of the young adult population overall may present with these errors. Up to 75% of these errors may resolve on their own between the end of the developmental period (i.e., age 9 years) and the end of high school, though speech therapy services may still be justified to reduce any negative social consequences of these errors. At least two different sources of these errors are postulated. Residual speech errors (the more common of the two types) appear to arise as a leftovers from a much earlier speech delay and were originally omission or substitution errors but migrated closer to normal to become distortions. Persistent speech errors, on the other hand, appear to be distortions from an early age that reflect long-instantiated habits. Whether this distinction points to different responses to particular interventions, or whether either type is more or less likely to resolve spontaneously is not clear at this time.

Effectiveness of Systematic Articulation Training Program Accessing Computers (SATPAC) Approach to Remediate Dentalized and Interdental/S,Z/: A Preliminary Study:

Traditional methods for treating speech distortion errors in older school-age children have tended to yield mixed success. The current study was a preliminary evaluation of an alternative approach called the Systematic Articulation Training Program Accessing Computers (SATPAC), which was tested for the remediation of /s/ and /z/. Procedures involved a sequence of phonetic placement and/or oral-motor placement cues as needed to establish the targets, followed by concentrated drill structured around a facilitating context nonsense word and then advanced to more natural contexts. Participants were 18 children aged 6 years, 9 months to 11 years, 10 months. Treatment involved once per week, individual, 10-min. sessions with an experienced speech-language pathologist. Group A (n = 9) received 15 weeks of treatment, while treatment was delayed for Group B (n = 9). Then the groups were reversed. During period one, Group A (treated) significantly improved their accuracy of /s, z/ in spontaneous speech, while Group B (untreated) showed no significant change. During period two, Group B improved significantly when treatment was applied. The majority of the participants retained proficiency two years later.

Psychometric Characteristics of Single-Word Tests of Children's Speech Sound Production.

PURPOSE Our understanding of test construction has improved since the now-classic review by McCauley and Swisher (1984). The current review article examines the psychometric characteristics of current single-word tests of speech sound production in an attempt to determine whether our tests have improved since then. It also provides a resource that clinicians may use to help them make test selection decisions for their particular client populations. METHOD Ten tests published since 1990 were reviewed to determine whether they met the 10 criteria set out by McCauley and Swisher (1984), as well as 7 additional criteria. RESULTS All of the tests reviewed met at least 3 of McCauley and Swishers (1984) original criteria, and 9 of 10 tests met at least 5 of them. Most of the tests met some of the additional criteria as well. CONCLUSIONS The state of the art for single-word tests of speech sound production in children appears to have improved in the last 30 years. There remains, however, room for improvement.

Predicting the Future: A Case Study in Prognostication

Purpose Clinicians are regularly asked to make long-term prognoses. The aim of the current report was to present one systematic approach to doing so. A case example from a malpractice case involving a child fitted with a cochlear implant was presented. Implantation occurred at age 17 months (activation 1 month later), but due to a procedural error, the implant was not functional for 19 months. The problem was ultimately rectified, but the legal case hinged largely on whether the child would be able to make up for the lost time. Method A review of the literature on long-term outcomes in children with cochlear implants was conducted. Using 4 studies measuring outcomes 7-10 years later, outcomes were compared between children implanted at age 17-18 months and those implanted at age 36-37 months. Results Analysis suggested no potential impact on nonverbal cognitive skills. However, analysis in the areas of speech perception, word comprehension, speech intelligibility, and reading suggested that after 7-10 years, this child would potentially continue to be approximately 1-2 years behind where she might otherwise have been. Conclusions This case illustrated the possibility of deriving a long-term prognosis using a systematic examination of the existing outcomes literature. Such an approach is consistent with our mandate to engage in evidence-based practice.