A cloze test uses missing ______ to assess language ability.
How can audiologists measure ______ to help assess listening difficulties?
Language ability and listening difficulties
While it seems like stepping onto someone else’s turf, there is a strong need for audiologists to be able to measure the language ability of any clients seeking an assessment for auditory processing disorders (APD). The reason for this lies in the complex connections between listening difficulties (LiD), APD and language ability.
People (children or adults) who are referred and assessed for APD typically don’t complain of having an APD. Rather, they (or their parents or teacher) complain that they are having trouble understanding speech, especially when there is background noise. That is, they are having difficulty listening. However, an LiD can have a variety of causes. It could be hearing loss (whether or not visible in the audiogram), an APD (of which there are several types), a cognitive deficit (especially memory or attention), or a language deficit. Further, for each person, more than one of these deficits may be contributing to the LiD.
As if that was not sufficiently complex, some of these underlying deficits may be the cause of others. For example, protracted or repeated middle ear infections during the first five years of life may have left a lasting spatial processing disorder (one type of APD) that affects the child’s ability to understand speech in noisy environments, thereby slowing the rate of language acquisition. In turn, that deficit in language ability then makes it harder to fill in the gaps when listening to speech in noise.
Alternatively, delayed language, instead of being a consequence of some auditory processing disorder might instead be the dominant, or only, cause of the LiD. Whenever there is sufficient noise or reverberation, some words are masked, or only partially heard, preventing their direct recognition. Someone with sufficient language ability will use all the surrounding words in the sentence that were understood, plus knowledge of the topic being discussed, to try and identify the word that was not itself understood. When successful, complete understanding of the sentence is achieved. If the person has a deficit in their language ability relative to age peers, inferring the missing word(s) may not be possible. The result is an LiD that is most apparent, or only apparent, when there is sufficient background noise and / or reverberation.
"AudiCloze plays each sentence auditorily... one word has been completely removed and replaced by a whistle... The listener’s task is to infer the missing word"
Several researchers have commented on the similarity of the abilities of people diagnosed by audiologists as having APD to those diagnosed by speech pathologists as having developmental language delay [1,2]. Given that a deficit in language can be a consequence of APD, or a cause of LiD, this similarity of symptoms is not surprising.
For an audiologist to determine the dominant cause of a reported listening difficulty, it is therefore highly desirable for them to know something about the child’s overall language ability. In the many cases where the child has not already been assessed for language, it would be helpful if the audiologist could directly measure it. To meet this need, a new computer-based test called AudiCloze has been developed.
Measuring language
Cloze tests, in which one or more words in a sentence are omitted, and must be guessed by the client on the basis of context, are well established. They provide an overall estimate of the ability to use language to infer meaning. They are mostly administered in written form, but this requires the person being tested to be able to read proficiently. Also, the neural processes likely differ from when speech is heard, as the written information can be re-read if needed, reducing demands on memory.
AudiCloze therefore plays each sentence auditorily. The sentence is in quiet and at a comfortable level, so is extremely easy to hear and understand. However, in every sentence, one word has been completely removed and replaced by a whistle of the same length to mark its position. For example, in the sentence ‘He went down and she went [whistle]’, the listener’s task is to infer the word that has been removed (which, in this case, was ‘up’), or to state some other word that also makes the whole sentence meaningful. In this case ‘too’ would be a reasonable response, indicating that the person had sufficient understanding of the sentence to come up with a meaningful alternative.
The sentences created for AudiCloze use words familiar to young children (based on the statistics of British Broadcasting Corporation broadcasts to children). All sentences have high internal context, achieved by combining words that are semantically related in some way. That way, those with good language ability have the greatest chance of accurately guessing the missing word.
Relation to speech understanding in noise
To test the impact of language on speech understanding in noise, language ability measured with AudiCloze was compared to speech understanding in noise measured with the Test of Listening Difficulties – Universal (ToLD-U). ToLD-U is designed to be a realistic simulation of understanding speech in noisy, reverberant places, like many classrooms. The test material are sentences of five to 12 words, spoken in a casual manner, using words familiar to young children. The background noise comes from six child talkers, processed to sound like they are at different distances and directions from the listener. The target speaker, an adult female, sounds like she is directly ahead. The test measure is the signal-to-noise ratio at which 75% of the words in each sentence are understood.
Figure 1: Speech understanding in noise measured with ToLD-U versus language ability measured with AudiCloze. Circles represent data measured on young adults [3] and triangles represent data measured on children aged six to 12 years [4]. Data points in red are for native speakers of English and those in blue are for speakers of English as a second language.
Figure 1 shows speech understanding versus language ability for children and adults drawn from the general population. All scores are expressed as z-scores, which indicate how many standard deviations an individual falls above or below the age-expected population mean. For every one-unit decrease in language ability (as measured by AudiCloze) there is a 0.5-unit decrease in the ability to understand speech in noise and reverberation (as measured by ToLD-U). Note that this result is not saying that language ability is all there is to speech understanding in noise, just that it is an important cause to consider, along with any type of auditory processing disorder.
It is our hope that AudiCloze becomes a routine part of the test battery whenever a child, or even adult, is assessed by an audiologist to determine the cause of listening difficulties that cannot be explained by elevated hearing thresholds. Normative data exists from age six through to adult, and results are automatically corrected for age and presented as z-scores.
References
1. Dawes P, Bishop D. Auditory processing disorder in relation to developmental disorders of language, communication and attention: a review and critique. Int J Lang Commun Disord 2009;44(4):440–65.
2. de Wit E, van Dijk P, Hanekamp S, et al. Same or Different: The Overlap Between Children With Auditory Processing Disorders and Children With Other Developmental Disorders: A Systematic Review. Ear Hear 2018;39(1):1–19.
3. Luengtaweekul P. The impact of rise time discrimination, language ability, memory, and attention on speech recognition in noise. M.Res. thesis, Macquarie University 2023 (LiD).
4. Zhou X, Dillon H, Tomlin D, et al. The impact of age, sex, language ability, and ESL status on children’s recognition of nonwords and sentences in noise and reverberation as measured by the LiSN-R and ToLD-U. (Submitted).
Declaration of competing interests: AudiCloze has been released commercially by Sound Scouts HQ Pty Ltd. AudiCloze has been licensed by Macquarie University to Sound Scouts, and Macquarie pays a share of any royalties received from Sound Scouts to Harvey Dillon as an employee of Macquarie. Harvey Dillon also consults casually to Sound Scouts.
