Intelligibility refers to the ‘understandability’ of speech, the match between the intention of the speaker and the response of the listener, and the ability to use speech to communicate effectively in everyday situations. It is the most immediate criterion by which a child’s communicative attempts are judged. Clearly it is a very important concept, but one which is often difficult to measure and address in intervention for children with speech difficulties.
It is known that greater intelligibility is associated with increased chronological age. By three years of age, a child’s spontaneous speech should be at least 50% intelligible to unfamiliar adults. By four years of age, a child’s spontaneous speech should be intelligible to unfamiliar adults, even though some articulation and phonological differences are likely to be present. Gordon-Brannan and Hodson (2000) carried out an investigation of pre-school children’s intelligibility by comparing the scores obtained for a range of severity and intelligibility measures. The children ranged in phonological proficiency from those with adult-like speech to those with severe difficulties. They suggested that any child above the age of four years with a speech intelligibility score of less than 66% (i.e. less than two thirds of utterances understood by unfamiliar listeners) should be considered a candidate for intervention.
Three approaches typically used for measuring intelligibility are:
- Word identification tasks: The tasks require a listener, or a group of listeners, to write down what the child says. An open response format is most commonly used with the listeners instructed to write down what they think the child is saying. In some cases closed-set tasks are used, where listeners are given a range of multiple-choice alternatives from which to select their responses. Scoring procedures vary, but typically involve sentences being scored on the number of key words correct or by the total number of words correctly identified.
- Listener-rating scales: Listeners are required to make judgements about the child’s intelligibility using a rating scale. Interval scaling requires the listener to assign a number to each recorded stimulus – most commonly 5, 7 and 9 point scales where, for example, 1=completely unintelligible, 9=completely intelligible. Direct magnitude estimation requires an estimate – typically a percentage – of parts of a sentence which are understood, e.g. 100% would indicate that a listener understood the entire sentence, whereas 50% would suggest that they understood only about half of the words.
- Formal assessments: There are some published intelligibility assessments which make use of one or both of the methods described above. The Children’s Speech Intelligibility Measure (CSIM, Wilcox and Morris, 1999) has been devised for the assessment of single word intelligibility. Fifty words are randomly selected from given sets of words and the child repeats these onto an audiotape. Two to three listeners are then required to listen to the tape, either transcribing the words or using a multiple-choice format.
Intelligibility levels are frequently used in making clinical decisions. For this reason measurements need to be accurate and reliable. Such measures are also important in order for clinicians to evaluate any changes occurring in a child’s intelligibility as a result of their treatment since:
“Intelligible speech is the long term goal for most intervention approaches for children with speech disorders.”(Dodd and Bradford, 2000, p.191)
However, intelligibility measures should move beyond mere descriptions of intelligibility and attempt also to seek explanations for intelligibility deficits. For example, two children with apraxia of speech may obtain identical scores on a single word intelligibility test, but qualitatively their speech may sound very different. Analysis of error matrices may reveal different strengths and weaknesses. A comprehensive model of intelligibility needs to account for these differences, which can then be used in treating speech disorders and monitoring progress.
As a starting point, it is helpful to conceptualise speech intelligibility scores in terms of three major variables: (1) the characteristics of the speaker (e.g. the child’s age and nature of his/her speech difficulties), (2) the speech material used (e.g. single words or spontaneous sentences) and its mode of transmission (e.g. audio or video recordings vs. live presentations), and (3) the characteristics of the listeners who perform the evaluations (e.g. are they familiar or unfamiliar with the child?). Any work on intelligibility needs to quantify parameters in each of these areas as well as considering interaction between them. For example, in terms of speaker factors we know that both segmental (i.e. individual sounds) and supra-segmental factors (e.g. rate, intonation and rhythm of speech) influence intelligibility. Speech sound production seems to be a key factor in influencing intelligibility, but it is not likely to account for more than 50% of what makes an individual understandable (De Bodt, Hernandez-Diaz and Van De Heyning, 2002). Segmental characteristics strongly associated with reduced intelligibility include omission of word initial phonemes, voicing errors, errors of consonant clusters, consonant substitutions and unidentifiable distortions – all of which may be observed in children with apraxia.
Much of the research into supra-segmental features has focused on adults with acquired speech disorders rather than children. We know that supra-segmental factors have an important contribution to make in the speech of adults with dysarthria. For example, an excessively fast or slow rate may reduce intelligibility; atypical stress patterns may also reduce intelligibility since listeners use stress as an automatic language processing strategy and have certain expectations. Speech perception does not simply involve recording a sequence of sounds as the speech signal enters the auditory system, but attending preferentially to certain aspects of the signal. These strategies are well-practised and natural, having been shaped by the predictable structure of the incoming signal. Listening to the unusual speech signal produced by someone with a speech disorder challenges normal perception abilities, making it an effortful process. Since children with apraxia of speech often have difficulties with prosody, it is likely that supra-segmental factors will also have a very important role to play in their intelligibility – and thus their intervention.
It seems clear that the interaction of segmental and supra-segmental factors is an essential aspect of understanding speech intelligibility. Yorkston, Dowden and Beukelman (1992) suggest that it is precisely this sort of information that is needed to move towards “an intelligibility-based model of intervention.” In order to help individual children with apraxia of speech, it is important for clinicians to know which factors – both segmental and suprasegmental – have the greatest effect on intelligibility so that these can be prioritised in intervention. Such factors will vary from child to child, but careful investigation of intelligibility on an individual basis will better enable us to support individual children, enhance our theoretical understanding of intelligibility, as well as adding to the armoury of outcomes measures that allows us to ensure interventions are socially and functionally relevant.
De Bodt, M., Hernandez-Diaz, H., & Van De Heyning, P. (2002). Intelligibility as a linear combination of dimensions in dysarthric speech. Journal of Communication Disorders, 35, 283-292.
Dodd, B., & Bradford, A. (2000). A comparison of three therapy methods for children with different types of developmental phonological disorder. International Journal of Language and Communication Disorders, 35, 189-209.
Gordon-Brannan, M., & Hodson, B. (2000). Intelligibility/Severity Measurements of Prekindergarten Children’s Speech. American Journal of Speech-Language Pathology, 9, 141-150.
Wilcox, K., & Morris, S. (1999). Children’s Speech Intelligibility Measure (CSIM).USA: Psychological Corporation.
Yorkston, K., Dowden, P., & Beukelman, D. (1992). Intelligibility measurement as a tool in the clinical management of dyarthric speakers. In R. Kent (Ed.), Intelligibility in Speech Disorders: Theory, measurement and management. Amsterdam: John Benjamins.
[Dr Michelle Pascoe is a registered Speech and Language Therapist in the United Kingdom. She is currently an Interdisciplinary ESRC (Economic and Social Research Council) / MRC (Medical Research Council) Fellow based at the Department of Human Communication Sciences at the University of Sheffield, U.K. Michelle’s current research focuses on intelligibility as a clinical outcomes measure for children with speech difficulties, and is carried out in collaboration with Professor Bill Wells, Professor Pam Enderby and Professor Joy Stackhouse from the University of Sheffield.]