Passage comprehension performance in children with cochlear implants and/or hearing aids: the effects of voice quality and multi-talker babble noise in relation to executive function

Abstract

Purpose: Speech signal degradation such as a voice disorder presented in quiet or in combination with multi-talker babble noise could affect listening comprehension in children with hearing impairment. This study aims to investigate the effects of voice quality and multi-talker babble noise on passage comprehension in children with using cochlear implants (CIs) and/or hearing aids (HAs). It also aims to examine what role executive functioning has for passage comprehension in listening conditions with degraded signals (voice quality and multi-talker babble noise) in children using CI/HA. Methods: Twenty-three children (10 boys and 13 girls; mean age 9 years) using CI and/or HA were tested for passage comprehension in four listening conditions: a typical voice or a (hoarse) dysphonic, voice presented in quiet or in multi-talker babble noise. Results: The results show that the dysphonic voice did not affect passage comprehension in quiet or in noise. Multi-talker babble noise decreased passage comprehension compared to performance in quiet. No interactions with executive function were found. Conclusions: In conclusion, children with CI/HA seem to struggle with comprehension in poor sound environments, which in turn may reduce learning opportunities at school.

Keywords:

• Children
• executive function
• dysphonic voice
• multi-talker babble noise

Introduction

Speech signals can be degraded by environmental influences during transmission such as the presence of background noise or by source degradation caused by speech and voice disorders [1]. Even with proper listening devices, the presence of a hearing loss can be considered to cause a signal degradation which is inherent to the individual [1]. Several effects have been proposed to be the outcomes of these degradations: failure to recognize speech, attention reduction, and suboptimal use of the memory capacity which all contribute to adverse conditions for perceptual learning [1]. Here, it is assumed that speech processing is an automatic and rapid process when the listening conditions are optimal. This automatic and rapid process is implicit and relies on multimodal input. The input is connected to the individual phonological representations which in turn are used for matching the speech sounds to their long-term representations in the semantic memory. This matching is the basis for both identification of lexical representations and their meaning [2–4]. However, the matching is impeded and becomes imperfect by speech signal degradation caused by, for example, the presence of noise, a dysphonic voice, or a hearing loss. To be able to identify a match, explicit processing is required and it requires the explicit use of cognitive resources [2–4]. When the explicit speech processing demands cognitive resources to be successful, less resources remain available for comprehension. This means that the cost of the explicit processing of a deviant voice quality or a background noise could be manifested as poorer performance. The present study examines the effects of voice quality and multi-talker babble noise on passage comprehension in children with cochlear implant (CI) and/or hearing aid (HA). In addition, it examines what role executive functioning has for passage comprehension in listening conditions with degraded signals in these children. Executive function is the ability to plan, organize, solve problems and reason and is top-down processes used to concentrate, attend and suppress automatic behavior or responses that are contra productive for an individual [5].

Voice disorders such as a dysphonic voice are reported more often by teachers than by the overall population [6–8]. Dysphonic voice is defined as a voice perceived as having increased pressure (hyperfunction), breathiness, roughness and instability compared to a typical voice. The higher prevalence seems to be related to speaking in noise. The dysphonic voice quality is the result of functional load. In comparison with a typical voice quality, these changes are not seldom perceived as a voice that is instable, breathy, rough, or hyperfunctional (pressed) [9]. A dysphonic voice seems to influence performance in language comprehension tasks for listeners with normal hearing [10–15]. Sometimes effects of the dysphonic voice are found when measuring dynamic aspects of performance by process measures such as reaction times, number of self-corrections, etc., but not necessarily on accuracy scores. For example, the dysphonic voice quality did not influence normally hearing children’s overall accuracy in a sentence-based forced-choice language comprehension task in the auditory modality. However, the dysphonic voice decreased accuracy in the most difficult tasks and it increased the number of self-corrections made on the easier tasks [15]. von Lochow et al. [16] examined the effect of the dysphonic voice on normal-hearing children’s performance on a passage comprehension task in the auditory modality, a task more similar to tasks encountered in the children’s everyday life than a sentence comprehension task. They found a relationship between voice quality and executive functioning of the individual child: children with better executive functioning were more influenced by the effect of the dysphonic voice than children with poorer executive functioning but only at a certain intermediate level of task complexity. This corroborates findings by Lyberg Åhlander et al. [15] although a sentence–based task was used in that study. Knowledge is, however, sparse about how children’s perceptual and cognitive capacity is related to comprehension and recall of auditory information when the signal is degraded. This will be further explored in the present paper.

Children with normal hearing seem more susceptible to the adverse effects of background noise than adults [17–19]. The adverse effects seem to be accentuated when the noise signal consists of discernable (auditory) competing speakers [20,21]. This implies that the adverse effect of background noise increases when masking is informational. We refer to informational masking as a listening condition when an audible speech signal and an audible speech masker (distractor) occur at the same time but spectral and temporal separation between the speech signal and the speech masker is not possible as both are speech signals [22]. In children’s everyday life, listening situations with informational masking are common and cause degradation of speech signals [1]. In listening comprehension tasks, children’s content knowledge, cognitive and linguistic abilities are related to the efficacy of the integration of the speech signal to long-term representations (e.g. lexical representations and memory) which affects their task performance [23–29]. For example, children with normal hearing may use cognitive resources to a higher extent when performing sentence-based listening tasks in background noise leaving less resources available for the task at hand [23]. Sullivan et al. [29] examined passage comprehension in quiet and in multi-talker babble noise (signal-to-noise ratio, SNR, –5 dB) for children with normal hearing. They found that passage comprehension was poorer in noise than in quiet. In addition, they found that passage comprehension showed a stronger relationship with working memory capacity in noise in comparison to quiet. von Lochow et al. [16] examined passage comprehension in children with normal hearing and found significantly poorer performance in listening conditions with one and four competing speakers (SNR 5 dB) compared to performance in quiet. In addition, they found that better executive functioning was associated to less susceptibility to the adverse effects of the competing speakers. These previous findings are consistent with the assumptions by Rönnberg et al. [30] that explicit processing caused by the adverse listening condition requires more cognitive resources which in turn affects task performance.

However, these previous studies tested children with normal hearing. It is most likely that a hearing loss provides signal degradation also with the use of HA and/or CI [31–34]. Few studies have tested these assumptions on pediatric populations and empirical data is, to our knowledge, lacking on children with CI/HA. Thus, effects of voice quality in quiet and in multi-talker babble noise are still unexplored for children, whose listening skills may be limited due to a hearing impairment.

When studying the effect of signal degradation in children, it is important to select both tasks and listening conditions that are similar to those that the children meet in their daily life if the ambition is to approach ecological validity in the study. A narrative task such as an auditory passage comprehension task resembles every day learning activities in its requirements on listening, understanding, comprehension, and remembering. Furthermore, these narrative listening tasks seldom occur in quiet in the classroom but rather in the presence of competing sounds such as background noise or competing speakers [35]. Teachers have to make themselves heard in noisy backgrounds, which strain and inevitably change their voice quality in different ways. Therefore, the aim of the present study is to investigate the effects of voice quality and multi-talker babble noise on passage comprehension in children with using CI/HA. Furthermore, a second aim of the present study is to examine what role executive functioning has for passage comprehension in listening conditions with degraded signals (voice quality and multi-talker babble noise) in children using CI/HA.

Methods

Participants

Twenty-four children were recruited either through their schools (n = 14) or at the summer camp held by the Swedish Organisation for Children with Cochlear Implants or Hearing Aids (Barnplantorna) (n = 10). The inclusion criteria were regular use of HA, CI or a combination of both, age between 6 and 13 years old (72–156 months), and proficiency in Swedish (not necessarily native speakers). All 24 children met these criteria and were tested in their school or at the camp. The children attended schools throughout Sweden. During testing, one child did not complete a number of tests and was therefore excluded from analysis. Thus, the final sample for this group consisted of 23 children (10 boys and 13 girls).

The mean age was 09:03 (years:months) ranging from 06:03 to 13:00. Twelve children were CI users (11 bilateral and one unilateral). Eight children were HA users (six bilateral and two unilateral). Three children were bimodal users (one CI in one ear and HA in the other). Information on individual fittings was not available. Twenty children attended typical schools and were integrated in classes with normal hearing peers. Three children attended schools for children with hearing impairment. Eleven children required special services at school. Eight children were currently in contact with a speech-language pathologist (SLP) and 17 children had previously been in contact with an SLP. These numbers may seem high, but SLP contact is mandatory in Sweden for children with hearing impairment. Four children were not native Swedish speakers and all four were multilingual. The other 19 children were native Swedish speakers. All children were spoken language users and most attended mainstream schools.

The same examiner tested all children in a quiet room either at their school or at the summer camp. The study was conducted in accordance to the Helsinki Declaration and with the approval of the Regional Ethics Committee in Lund, Sweden (application 2014/408).

Passage comprehension task

A passage comprehension test in the auditory modality was collected from a subtest of the Swedish version of the Clinical Evaluation of Language Fundamentals – Fourth Edition (CELF-4) [36]. Passages from this subtest were used to test passage comprehension with a typical and a dysphonic voice presented in quiet and in background noise. Several factors influence individual performance on CELF-4 passages such as language, cognitive and social skills. Originally, the passage comprehension test includes two exercise passages and six test passages. In the present study, one exercise passage and only four test passages were presented to the children. These passages have different subject content and durations (approximately 40–55 s/passage). The task was to listen to a passage and thereafter answer five content questions about the passage. An example of a passage content is a story about a reading contest at a school. An example of a question is “What were the school children supposed to do?”.

Preparations of the listening conditions

The two exercise and the four selected test passages were recorded twice by the same female speaker (49 years, one of the authors): once with a typical voice and once with a provoked dysphonic voice. The dysphonic voice was provoked using a vocally loading procedure, according to the method of Whitling et al. [37]. This was done by reading aloud for 20 min in a background noise presented at 85 dB SPL (Leq, i.e. the equivalent continuous sound level) from a laptop computer via a Fostex SPA 12 loudspeaker (Fostex Corporation, Tokyo, Japan) in a sound proof booth. The background noise presentation level was verified with a Brüel and Kjaer integrating-averaging sound level meter (type 2240) using an integration time of 60 s. The background noise consisted of the International Speech Test Signal (ISTS) by Holube et al. [38]. This signal consists of concatenated speech that cannot be interpreted semantically. The speaker was instructed to read aloud and make herself heard over the background noise at all times. This set-up forced the speaker to read aloud at approximately 90–95 dB SPL. After this vocally loading procedure, the recording of the dysphonic voice was made immediately and without any rest.

The following procedures were used to record both the typical and the dysphonic voice. During the recordings, a multi-talker babble noise was presented at 55 dB SPL (Leq verified as in the vocally loading procedure) from a laptop computer connected through the same Fostex SPA 12 loudspeaker. Making herself heard over this noise, the speaker read the five CELF-4 passages. This procedure generated speech levels of about 60–65 dB SPL. The level of the speech signals was monitored with a head-mounted microphone MKE 2, no 09_1 (Sennheiser, en-de.sennheiser.com/). Also here, the multi-talker babble noise consisted of the ISTS, but with the modification that it had been both time-shifted and duplicated eight times [39] to achieve a less modulated noise signal without changing the spectral content. The CELF-4 passages were recorded using a Lectret HE-747 microphone connected to a Zoom H2 (Zoom Corporation, Tokyo, Japan) with a 44.1 kHz/16-bit sampling frequency. After the completion of the recordings, Adobe Audition (version 6; Adobe Systems, San Jose, CA) was used to normalize the separate passages to an equal average root-mean-square in dB. Normalization was made after removing pauses and other silent sections and these pauses and sections were added again after the normalization procedure.

The voice quality authenticity of both the typical and dysphonic voices was judged by three SLPs with long clinical experience with voice disorders. The SLPs assessed the recorded passages for pressure (hyperfunction), breathiness, roughness and instability, and degree of voice disorder using an analysis software, Visual sort and rate – Visor [40]. In their assessment, the degree of voice disorder was rated for each parameter on a visual analogue scale ranging from 0 (no occurrence) to 10 (maximum occurrence). On this scale, scores equal to or higher than 5 is considered as pathological. The SPLs assessed the recordings together and made their final judgements in consensus. In this analysis, hyperfunction was judged as 1 for the typical voice and 7 for the dysphonic voice. All other parameters were judged as 0. The overall degree of voice disorder was judged as 0 for the typical voice and as 4 for the dysphonic voice.

In the present study, two types of background listening conditions were used, quiet and four competing speakers [16]. This multi-talker babble noise consisted of the combined individual recordings of four girls (aged 9–11 years). Each girl was recorded individually in a soundproof booth while reading separate chapters from an age appropriate book [41]. The recordings were made using a Zoom H2 (Zoom Corporation, Tokyo, Japan; 44.1 kHz/16-bit sampling frequency) with a portable microphone JTS (UT16HW) and a JTS (US800ID) receiver. While recording, the girls read their chapter several times. After their conclusion, the following procedures were made offline to the recordings in Adobe Audition (version 6; Adobe Systems, San Jose, CA). Initially, each girl’s best recording was identified. These best recordings were then normalized to an equal average root-mean-square dB after removing pauses and other silent sections. After adding the pauses and the other silent sections again, the normalized recordings were added together to generate the four speaker background listening condition. This multi-talker babble noise was added to the recordings of the CELF-4 passages at 10 dB lower than the target speech signal (the voice reading the CELF-4 passages) thus resulting in an SNR of 10 dB. A 10 dB SNR was selected to improve target signal audibility while providing a challenging listening condition for the listeners. In the listening conditions with the multi-talker babble noise, the target signal was preceded and followed by the competing speaker signal during 1 s.

In the end, two listening conditions in quiet were generated – one with the typical voice and one with the dysphonic voice – and two listening conditions with multi-talker babble noise – also one with the typical voice and one with the dysphonic voice. Thus, all four listening conditions were generated for each CELF-4 test passage to allow for balancing between children in order to avoid order and fatigue effects.

Procedures for the passage comprehension task

The children were verbally instructed to listen to four passages presented one at the time. They were also instructed that sometimes the passages would be presented in quiet and sometimes with competing speakers in the background. After listening to each passage, the children were required to answer five questions asked by the examiner about the passage content. Initially, the children listened to the exercise passage to practice and to allow for them to familiarize themselves with the task. Following this, the children listened to the four passages in the four different listening conditions. Using a Latin squares design both presentation order and passage content were counterbalanced across the children to avoid order and fatigue effects. The speech signal was presented at 70 dB SPL in all listening conditions.

Executive function

The Elithorn Mazes (EMs) were used to assess executive functions such as organization, planning skills, response inhibition, and processing in children (Wechsler Intelligence Scale for Children – Fourth Edition; [42]). They have previously been used in studies examining the relationship between executive function and the effect of voice quality in children [16,43,44]. EM is intended for use on children aged 8–16 years old. The task is completed on paper using a pen. Verbal instructions were given stating that the task for the children was to draw a path out of the maze by crossing a predefined number of dots. Three exercise mazes were completed by each child prior to the test. Seven mazes with increasing difficulty were presented. Each correctly completed maze was assigned a score of eight. Thus, the EM total score was therefore 56 for all seven mazes. In the present study, we used the maximum number of completed mazes as our measure of executive functioning (referred to as EM_level in the following sections).

Overall procedures

The same predetermined test order was used for all children throughout the testing. The session begun with the CELF-4 passages followed by EM. The CELF-4 was presented from a laptop computer connected to a Bose Companion 2 loudspeaker. The loudspeaker was adjusted to the level of the child’s head, located at 0° azimuth, and placed at a distance of 1 m from the child. Using this set-up, the presentation levels were verified with a Brüel and Kjaer integrating-averaging sound level meter (type 2240) using a 1000 Hz tone with an average root-mean-square (dB) equal to the speech signals. The individual test session duration lasted for approximately slightly more than 1 h. Each child was tested alone in a quiet room.

Results

The results are reported as passage comprehension performance (CELF-4 responses) and EM_level. The analysis is first made including all children. In a second step to reduce the heterogeneity of the sample, the analysis is made using only children using bilateral CI and HA. An alpha level of 0.05 was considered statistically significant. Parametric statistics were used.

A repeated measures ANOVA was conducted using all included children to test the effect of listening condition (typical voice in quiet, dysphonic voice in quiet, typical voice in multi-talker babble noise, and the dysphonic voice in multi-talker babble noise) on passage comprehension performance. Voice quality (typical/dysphonic) and Background listening condition (quiet/noise) were entered as within-subject variables. Age and EM_level were added as covariates. The mean and standard deviations are shown in Figure 1, where a higher score indicates better passage comprehension performance. No interaction effects were seen. No significant effect was seen for voice quality (Wilk’s Lambda = 0.998, F[1, 20] = 0.034, p = .857, η² = 0.008). A significant effect was seen for background listening condition (Wilk’s Lambda = 0.779, F[1, 20] = 5.659, p = .027, η² = 0.221). These findings show that voice quality does not affect passage comprehension. Passage comprehension decreases significantly in multi-talker babble noise in comparison with quiet. No effects of age or executive function were seen on these results.

Figure 1. Average passage comprehension performance in four different listening conditions. A higher score indicates better performance. The maximum score is five. Error bars denote standard deviations (N = 23).

A second repeated measures ANOVA was conducted using only children with bilateral CI (n = 11) and HA (n = 6) to test the effect of listening condition (typical voice in quiet, dysphonic voice in quiet, typical voice in multi-talker babble noise, and the dysphonic voice in multi-talker babble noise) and listening devices (CI or HA) on passage comprehension performance. Voice quality (typical/dysphonic) and background listening condition (quiet/noise) were entered as within-subject variables. Type of listening device (CI/HA) was entered as between-subject variable. Age and EM_level were added as covariates. The mean passage comprehension performance and standard deviations are shown in Figure 2, where a higher score indicates better performance. No interaction effects were seen. No significant effect was seen for voice quality (Wilk’s Lambda = 0.959, F[1, 13] = 0.562, p = .467, partial η² = 0.041) and a small effect size was seen [45]. No significant effect was seen for background listening condition (Wilk’s Lambda = 0.808, F[1, 13] = 3.094, p = .102, partial η² = 0.192), but a large effect size was seen. No significant between-subjects effect was seen (F[1, 13] = 21.949, p = .164, partial η² = 0.143). A medium effect size was seen. These findings show that voice quality has a small effect on passage comprehension. Multi-talker babble noise has a large effect on passage comprehension. Listening devices have a medium effect on passage comprehension indicating slightly poorer performance for children with CI. No effects of age or executive function were seen on these results.

Figure 2. Average passage comprehension performance in four different listening conditions for children using bilateral CI (n = 11) and HA (n = 6). A higher score indicates better performance. The maximum score is five. Error bars denote standard deviations.

In summary, the findings suggest that voice quality does not influence passage comprehension in children using CI/HA while background noise influences passage comprehension negatively. However, the type of listening device seems to influence the effect of background noise although not to a large extent. Passage comprehension was not influenced by age or executive functioning.

Discussion

The effect of voice quality on passage comprehension

The present findings showed that passage comprehension performance in quiet and in multi-talker babble noise was not significantly influenced by speaker’s voice quality in children with CI/HA. Previous studies on children with normal hearing indicate that listening to a dysphonic voice in quiet affects language comprehension [10–14]. Lyberg Åhlander et al. [14] used a sentence-based forced-choice task (TROG-2) and demonstrated, as in the present study, that a dysphonic voice did not affect overall language comprehension performance. However, they found that performance deteriorated particularly in more difficult sentences for the dysphonic voice in comparison to a typical voice. The present findings are also in line with the findings by von Lochow et al. [16] who used the same passage comprehension task as in the present study. They found that a dysphonic voice did not influence passage comprehension directly in quiet. Furthermore, von Lochow et al. [16] found that in noise children with better executive functioning were more influenced by the effect of the dysphonic voice than children with poorer executive functioning. This association was only found when listening to the dysphonic voice presented with one competing speaker. von Lochow et al. [16] did not find this association when the children listened to the dysphonic voice presented in quiet or with four competing speakers. This indicates that the effect of the dysphonic voice was only discernable at a certain intermediate level of task demand. As proposed by Lyberg Åhlander et al. [13], the discrepancy between the present findings and those of von Lochow et al. [16] may be context dependent: differences in level of task demand and differences in factors inherent to the individual. What we see is thus a complex pattern where a more sophisticated model is needed such as the Framework for Understanding Effortful Listening (FUEL; [46]).

The present findings could also suggest that the children with CH/HA perceive not only the dysphonic voice but also the typical voice as degraded. Listening devices are not able to restore normal hearing [31–34]. Furthermore, signal processing and limitations in the devices may generate a degraded speech signal. Thus, it may not have been possible to discern any differences between the typical and the dysphonic voice in quiet and in multi-talker babble noise. Despite that the SLP’s judged voice order grade was four of 10 for the dysphonic voice, it is possible that a more degraded speech signal may have been able to elicit an effect. Future studies are required.

The effect of multi-talker babble noise on passage comprehension

At present, there seems to be no previous study examining passage comprehension in quiet and in multi-talker babble noise for children with CI/HA. Sullivan et al. [29] tested passage comprehension (with an assumingly typical voice) in multi-talker babble noise (four speakers) in children with normal hearing and found that performance deteriorated in noise. However, Sullivan et al. [29] used a negative SNR (SNR –5 dB), which may have influenced the audibility of the speech signal. von Lochow et al. [16] found significantly deteriorated passage comprehension performance in multi-talker babble noise at a less favorable SNR (SNR +5 dB) compared to listening in quiet for children with normal hearing. The present findings that performance for children with CI/HA on a passage comprehension task with a typical voice in multi-talker babble noise (SNR +10 dB) was poorer than performance in quiet are similar to these previous findings on children with normal hearing.

Notably, when comparing the present data with those of von Lochow et al. [16] it can be seen that the overall level of passage comprehension performance is substantially poorer for the children with CI/HA at the more favourable SNR than those for children with normal hearing: the average performance in quiet for children with CI/HA is about one score lower. Furthermore, the detrimental effect of the multi-talker babble noise is approximately twice as large for children with CI/HA than for children with normal hearing despite the fact that the latter children were on average about two years younger than the present sample. This indicates that the present passage comprehension task is generally more challenging for children with CI/HA than children with normal hearing. Narrative tasks such as the present passage comprehension task are common in schools today. This suggests, since these children struggle with comprehension, that the learning opportunities may become more limited for children with CI/HA. In addition, the performance was not to a large extent influenced by the type of listening devices used. The present passage comprehension task is very similar to narrative tasks that these children encounter in their schools. The present data give an insight into how these children actually perform when they are tested with their listening devices such as they are used in their everyday life. Thus, when not compensated for e.g. poorer audibility how do these children perform? This knowledge is important since it demonstrates the gap between what we aim to achieve when fitting the listening devices and what the actual outcomes are. The findings suggest that these children struggle with comprehension in their everyday life. Therefore, there is a high possibility that these children will have poorer prerequisites for learning and academic success in the long run.

Passage comprehension and executive function

The findings from the present study do not support the notion that degraded speech signals require explicit processing relying on the use of more cognitive resources than implicit processing [1–4]. However, executive function was assessed in an offline manner in the present study. It is possible that online measures would provide more conclusive information. Also, it is possible that the task level difficulty was too great to discern a relationship. The children may be struggling too much with the task engaging all available cognitive and linguistic resources available. The mediating effect of executive function demonstrated earlier may therefore not have been possible to demonstrate in the present study.

Study limitations and future directions

In the present study, data were collected from 23 children with CI/HA. The study group is heterogeneous and it is possible that a larger, more homogeneous sample could have provided more salient findings. However, when examining only children with bilateral CI and bilateral HA, a similar trend was seen as for the whole sample. Furthermore, we did not have the opportunity to control for the accuracy of the fitting of their listening devices and audiological baseline data (e.g. age at fitting/implant, audibility of speech signals) which may have influenced the present findings. Although a within-subject design was utilized that compares each child with him-/herself reduces this possible influence, future studies need to directly collect and assess this information. In future studies, nonword discrimination [47] could be used to assess and control for potential audibility issues. Furthermore, it is also possible that the fitting of different signal processing algorithms in the listening devices have influenced the present findings despite the within-subject design. The indications from the present analysis show that there is a medium effect of listening devices on passage comprehension. Therefore, in future studies, it would be valuable to test the effect of voice quality and background noise in children using similar devices with similar signal processing in a larger sample.

Despite previous use of the present tests in research and at the clinic, another passage comprehension task could have better fitted the present study design as prior experience and knowledge about the passage content can have influenced the outcomes on that task (i.e. as the content of the present passage may have been more familiar and known to some children). In the present passage comprehension task (CELF-4), the child listened to a short story and then answered questions on its explicit as well as implicit content which also required the child to make inferences. The passage comprehension represents language processing at a challenging meta-linguistic level [26,28]. To respond correctly in the present passage comprehension task, the child is required to both understand the content linguistically and integrate linguistic knowledge with previous experiences and knowledge about the world [27]. These are factors that we did not control for in the present study, but they would be possible to bypass by using passages with more fictive content in future studies as sometimes have been utilized in adults [48,49]. On the other hand, the findings in present study were obtained with a passage comprehension task that resembles children’s every day learning activities.

We found no influence of executive functioning on the children’s performance in the passage comprehension task. The executive functioning measure used (EM_level) assess general executive functions such as organization, planning skills, inhibition, and processing. It is possible that finer examination of executive control could provide additional information on what influence the effect of listening condition in passage comprehension in children with CI/HA. Examples of more specific executive functions that should be assessed in future studies are working memory, inhibitory control and cognitive flexibility [50]. In addition, as suggested by Brännström and coworkers [44], listening to a dysphonic voice seems to require explicit processing. Therefore, perceived listening effort may increase when listening to a dysphonic voice. Based on this, future studies could examine the effect of the dysphonic voice in relation to perceived listening effort in children with CI/HA.

With the design of the present study, we aimed at a high ecological validity, coming as close as possible to the real classroom situation. Hence, based on the findings, some clinical implications can be made. The noisy classroom poses challenges for both children and teacher. The possibility to benefit from teaching for children with CI/HA might be considerably reduced compared to the possibilities for children with normal hearing, due to challenges from a noisy learning environment – and from those posed by a dysphonic teacher voice. For the teacher, dysphonia is a common effect of talking over noise, affecting teaching and classroom communication as well as well-being. Hence, to raise teachers’ and schools’ awareness of the detrimental effects of a noisy learning and teaching environment seem crucial.

To sum up, ecologically valid experiments like this do offer more questions than answers. It has become obvious that our current explanations may be far too simplistic. Task demands, cognitive capacity and noise quality influence children’s performance. These interacting internal and external factors need to be fitted into a single framework describing children’s listening in difficult sound environments. FUEL provides a framework for understanding effortful listening in adults with hearing impairment [46] and can be described as a dynamic theoretical approach: people have different “grit” when listening in adverse listening conditions and motivation may change over time. It would be valuable to try and explain listening effort in maturing children within a FUEL for children that accounts for ongoing development of linguistic, socio-pragmatic and executive functions. Therefore, future studies will require coherent methods which targets both internal and external factors.

Conclusions

Passage comprehension in children with CI/HA seems not to be not influenced by a dysphonic voice quality, but multi-talker babble noise decreases passage comprehension performance. Overall, passage comprehension appears considerably lower for children with CI/HA than children with normal hearing. Children with CI/HA seem to struggle with comprehension which in turn may reduce learning opportunities at school.

Acknowledgements

The authors like to extend their gratitude to the participating children, their parents, and the Swedish Organisation for Children with Cochlear Implants or Hearing Aids (Barnplantorna).

Disclosure statement

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article. Viveka Lyberg-Åhlander is an associate editor with the journal.

Additional information

Funding

This study was partially financed by the Hearing Foundation (Hörselfonden – Hörselskadades Riksförbund) and the Linneaus’ Environment Cognition, Communication and Learning at Lund University.

Notes on contributors

K. Jonas Brännström

K. Jonas Brännström, PhD in audiology and reg. audiologist, associate professor in audiology at Department of Clinical Science in Lund, Logopedics, Phoniatrics and Audiology, Lund University, Sweden.

Heike von Lochow

Heike von Lochow, MSc in audiology and reg. audiologist, worked as a research assistant at the same department.

Viveka Lyberg Åhlander

Viveka Lyberg-Åhlander, PhD and reg. speech pathologist, associate professor in speech and language pathology at the same department.

Birgitta Sahlén

Birgitta Sahlén, PhD and reg. speech pathologist, professor in speech and language pathology at the same department.