Discussion

The purpose of the present study was to investigate the spontaneous speech intelligibility of seven-year-old normally hearing (NH) children and their early implanted peers with a cochlear implant (CI). For this purpose, one hundred and five adult listeners transcribed short speech samples that originated from a picture cued storytelling task.

Intelligibility of children with CI in comparison to NH children

In this study, intelligibility was measured by means of entropy values. These values express the degree of (dis-)agreement between the transcriptions of the listeners/transcribers. If listeners all agreed on the transcription of an utterance, this suggested that the utterance was perfectly intelligible and vice versa. The results showed that the agreement was significantly higher for NH children than for children with CI. In other words, NH children’s intelligibility appears to be higher than that of children with CI. For both groups, there was an effect of chronological age, which indicated that younger children reached lower intelligibility scores, whereas older children’s intelligibility improved. This result is in line with other studies in which the factor chronological age was found to correlate with language outcomes such as intelligibility (Boons et al., 2013; Chin et al., 2003; Flipsen & Colvard, 2006). Also, the intelligibility of both groups did not reach a plateau, as can be deduced from the lack of a quadratic effect of age. However, it is unknown if and when children reach maximal intelligibility. In line with Miller (2013: 606) who noted “that even ‘healthy’

speakers do not achieve 100% intelligibility”, we argue that transcriptions of spontaneous speech rarely reach maximal intelligibility, even in normally hearing adults. Ideally, a longitudinal follow-up of well-matched NH and CI children should further investigate this matter.

It is also noteworthy that, despite the fact that the intelligibility scores of NH and CI children differed significantly, there was a large amount of overlap between the scores of these groups. Three out of four children with CI reached intelligibility scores well within 2 SDs of the mean

result is in line with an increasing number of studies that show that early implanted children are catching up with their NH peers after a few years of device experience (Boons et al., 2013; Geers & Nicholas, 2013; Habib et al., 2010; Nicholas & Geers, 2007; Wie, 2010). Since the children in our sample were matched on several parameters, it is difficult to determine why some children’s intelligibility is lower. However, on scrutinizing the individual characteristics of the admittedly limited number of participating children, it is striking that two of the four children with the lowest intelligibility scores (i.e. CI1 and CI2) were implanted bilaterally quite late, at the respective ages of 6;3 years and 5;10 years old. Moreover, the hearing loss of CI12 – who also had an intelligibility score below the boundary of 2 SDs of the mean score of NH children – was caused by a CMV infection during the mother’s pregnancy, which has been suggested to lead to less favourable speech and language outcomes than for example a genetic connexin 26 associated hearing loss (Ramirez Inscoe & Nikolopoulos, 2004).

Transcription of spontaneous speech: entropy

Previously, spontaneous speech has primarily been investigated by means of rating scales. Since the intelligibility of early implanted children who have several years of device use is generally fairly high, the traditional rating scales do not provide enough nuance for judging the speech of these children, since they soon reach ceiling scores. Transcriptions offer a more detailed approach. In this study, the transcriptions were analysed in terms of the degree of agreement between listeners.

The agreement between transcribers was expressed as entropy scores. However, entropy is not only useful in spontaneous speech studies.

Entropy is also a suitable measure in transcription tasks of read or imitated speech, especially to shed light on the degree of (un)intelligibility of speech samples. Usually, studies focussing on read or imitated speech calculate the intelligibility by calculating the percentage of correctly identified words (relative to the total number of words). Incorrectly identified words are not taken into account. This obscures to a considerable extend the degree of intelligibility of the read or imitated speech: the intelligibility score remains the same whether or not an incorrectly transcribed word is rendered in exactly the same way by the transcribers. For instance, suppose that a particular word is transcribed correctly in 50% of the cases. What does the remaining 50% of the transcriptions consist of? Possibly the remaining 50% of the transcriptions contains exactly the same word so that there are only two variants in the transcriptions (e.g., the correct “frog” and an incorrect “frogs”). But it is also conceivable that all the incorrect transcriptions are different words (e.g., “frogs”, “fox”, “fog”, etc.). In both cases, the percentage correct is 50%, but the entropy score will be markedly different in both cases. In the case in which there are only two different forms in the transcriptions, the entropy score is still fairly small. However, in the case of the second scenario, the entropy score is greatly affected by the number of different or even unique transcribed words (see for instance the rightmost column in Table 2), and the entropy score will be fairly elevated. Hence, these different scenarios are reflected in differences of the entropy score.

Perspectives for future research

In the present study entropy was used as a measure of the intelligibility of children’s speech. Since no other study approached speech intelligibility of children with CI in a similar way, the intelligibility scores of the present study cannot be compared with those reported in other research. Therefore, there is a need for more studies using entropy in order to expand the empirical basis.

Assessing intelligibility of spontaneous speech with transcriptions offers several additional advantages that reach beyond the general intelligibility scores obtained from for example rating scales. These advantages have been explored in studies using read or imitated speech, but the method proposed in this study enables to expand them to spontaneous speech. For example, the present approach enables a post hoc qualitative analysis of the transcriptions, and enables to relate these aspects to the child’s intelligibility. Further in-depth research should investigate whether there are specific linguistic or speech aspects that vary in the transcriptions of spontaneous speech and whether they can explain a change in the child’s intelligibility. For example, does the degree of (dis-)agreement between the listeners increase with certain phonemes, certain word types or with the length of the utterance? In other words, what are the linguistic determinants of higher entropy?

A preliminary qualitative investigation of our data revealed several errors at the segmental level. For example, the transcriptions showed voicing discrepancies (e.g., boom [bo:m] ‘tree’ – pomp [pɔmp] ‘pump’) and deviances on the place of articulation between the transcriptions of listeners (e.g., hen [ɦɛn] ‘them’ – hem [ɦɛm] ‘him’; gaat [ɣa:t] ‘goes’ – had

[ɦɑt] ‘has’). Listeners identified different vowels (including diphthongs) at the same position (e.g., bijen [bɛjən] ‘bees’ – buien [bœyən] ‘shower’) as well as consonants (e.g., was [wɑs] ‘was’ – valt [vɑlt] ‘fell’). In our sample, it was also remarkable that word endings were often deviant (e.g., kikker

‘frog’ – kikkers ‘frogs’; schoen ‘shoe’ – schoenen ‘shoes’; sta ‘stand’ – staat

‘stands’) and that function words exhibited more chaos than content words. These results are in line with studies on inflectional morphology that find that children with CI lag behind on their NH peers (Hammer, 2010; Laaha et al., 2015; Szagun, 2000).

In addition, a post-hoc examination of the transcriptions reveals that the entropy calculations can further be refined. In the present study, in computing the entropy score each deviance in the transcriptions was equally weighed. In other words, each difference equally increased the word entropy score. However, some deviances in the transcriptions are fairly small (e.g., kikker ‘frog’ vs. kikkers ‘frogs’), whereas others can really be considered as mismatches (e.g., jongen ‘boy’ vs. hond ‘dog’). Further research is needed for finding fruitful avenues to refine the measure by taking into account the (linguistic) distance between different transcriptions. For example, the orthographic transcriptions of the listeners could be converted to and aligned on a phonemic level.

Subsequently, the entropy calculations could take into account the phonological distance of the different alignments.

Another significant aspect of this study is the effect of chronological age. Despite the fact that only children around the age of seven participated, there is a main effect of chronological age. Since the sample

as: the older children perform better than the younger children.

Surprisingly, this effect of chronological age is also found for NH children, whereas the current literature states that NH children reach full intelligibility at age four (Chin & Tsai, 2001; Weiss, 1982). This benchmark for NH children’s intelligibility is often implicitly accepted in other studies:

although other studies suggest that the group of NH children is not homogeneous either (Chin et al., 2003; Leonard et al., 1980), many studies do not include a NH control group (Chin et al., 2001; Flipsen & Colvard, 2006; Peng et al., 2004; Svirsky et al., 2000b). Hence, at this point a longitudinal follow-up is called for in order to confirm that children’s speech intelligibility still continues to improve up to and after age seven.