Conversational Fluency of Children Who Use Cochlear Implants

Abstract

Purpose 

Children who have significant hearing loss often experience difficulty in engaging in everyday conversations. They may spend an inordinate amount of time in communication breakdown or in silence. The purpose of the present investigation was to assess the oral conversation fluency of young cochlear implant users using both objective and subjective measuring procedures. Conversational fluency relates to how smoothly a conversation unfolds. The children’s performance was compared with that of children who have normal hearing, and related to other measures of communication skill and other child variables.

Design 

One hundred eighty-one cochlear implant users and 24 children with normal hearing engaged in conversations with a clinician, using an oral mode of communication. Audio-video tapes of the conversations were analyzed to yield the following measures: percent of time the child and clinician spent trying to repair a breakdown in communication, percent of time the two spent sitting in silence, and the ratio between the amount of time the child spoke and the amount of time the clinician spoke. In addition, for the cochlear implant users, 10 judges viewed excerpts of the tapes and gave their impressions and reactions to the children and their conversations, using five-point rating scales. The objective and subjective measures were related to the children’s speech intelligibility, mean length utterance, and speech recognition in both an auditory-only and auditory-plus-vision condition.

Results 

The cochlear implant users spent significantly more time in communication breakdown and in silence than the children with normal hearing, and children who are in an educational placement that emphasizes oral communication spent less time in breakdown than children who are in educational placement that use both speech and sign. Speech intelligibility and receptive language were the best predictors of communication breakdown. Judges perceived children who spent less time in communication breakdown more favorably than children who spent more time, and reacted to them more positively. Finally, there was a high correlation between the subjective measures of the conversation and the objective measures.

Conclusions 

Educational programs should provide explicit instruction to promote conversational fluency. The results show a need for communication therapy, and suggest that the children will be perceived and reacted to more favorably if they learn how to manage communication difficulties. The results also support the need for extensive speech-language therapy for these children. The objective measures employed in this investigation appear to be a valid and useful means for assessing conversational fluency.

In Brief

One frequent consequence of profound deafness is some impairment in oral conversational interactions. Three characteristics of a videotaped oral conversation between a deaf child and an unfamiliar adult were measured: time spent repairing communication breakdown, time spent in silence, and the ratio of speaking times between the two conversational partners. Even after 4 to 6 years of cochlear implant use, deaf children still experienced poor conversational fluency when compared with children with normal hearing. They spent significantly more time in silence and in communication breakdown. Deaf children who had the most intelligible speech and the best receptive language were those with the most fluent conversations. Children whose educational program emphasized speech were more fluent than children whose program emphasized sign language. Normal-hearing adults who viewed videotaped segments of the conversations rated fluent communicators more favorably than nonfluent communicators.

Much of the substance of human relationships is established by our conversational interactions—by what we say, how we say it, and how we listen. We engage in conversation for a myriad of reasons. For example, we converse so to share ideas, relate experiences, express needs, influence others, and establish intimacy. In recent years, sociolinguists have developed measures that assess how smoothly conversation unfolds. Collectively, these measures indicate conversational fluency (^{Erber, 1996}). In previous work we have demonstrated that many adults who have hearing loss often experience poor conversational fluency (^{Tye-Murray & Witt, 1996;Tye-Murray, Witt, Schum, & Sobaski, 1995}).

When conversational fluency is poor, numerous communication breakdowns occur, and many interchanges may be required before they are resolved. A communication breakdown is an instance in which one person says something and another person does not recognize the message. Poor conversational fluency is also signaled by the presence of prolonged silences, and one conversational partner may speak too much or too little. In a conversation that has poor fluency, the conversational partners are often limited in the topics they discuss, and how they talk about them. For instance, topics might be limited to concrete objects in the room (e.g., “That’s a pretty shirt you have on today.”) and might be discussed with common vocabulary (“shirt” as opposed to “halter top”). On the other hand, when conversation fluency is high, need for clarification is minimal, participants have ample opportunity to speak, and the conversation has few prolonged silent intervals. Topic selection and concomitant discussion appear not to be limited by the communication and conversational skills of the participants (e.g., “I read in today’s paper that the mayor of New York is seeking a city-wide tax on hotel suites.”). Table 1 presents excerpts from a conversation with poor and a conversation with high conversational fluency for illustration.

TABLE 1: Excerpts from conversations having high conversational fluency and low conversational fluency

Receipt of a cochlear implant enhances a deaf child’s acquisition of speech perception, speech production, and language competence (^{Geers & Brenner, 2003;Geers, Nicholas, & Sedey, 2003;Tobey, Geers, Brenner, Altuna, & Gabbet, 2003}). These skills likely enhance children’s oral conversational fluency in real-world conversational situations, although to date, conversational fluency per se has not been examined for young cochlear implant users.

Most research about oral conversational fluency that has focused on children who have significant hearing loss has concerned children’s use of specific kinds of repair strategies and communication breakdown management (e.g., ^{Caissie & Wilson, 1995;Ciocci & Baran, 1998;Givens & Greenfeld, 1982;Most, 2002}), and children who use hearing aids. The research suggests that children who are hard-of-hearing and deaf tend to revise utterances that their communication partners do not understand, and that they often rely on non-linguistic responses when seeking clarification of their partner’s messages. In contrast, children with normal hearing are more likely to repeat than revise their utterances, and more likely to use linguistic responses when seeking clarification. Although one might surmise that these differences in managing communication breakdowns between children who have hearing loss and children who have normal hearing may differentially affect conversational fluency, the issue has not been addressed directly by research.

It is possible that other factors besides the obvious ones of speech production, language, and speech perception might affect how well a child can engage in conversations. For example, some children may have developed behaviors to avoid the occurrence of communication breakdowns (say, they either dominate or withdraw from speaking turns), some may be especially effective in repairing breakdowns, with “effective” meaning that little time is devoted to the repair process. These strategies might promote high conversational fluency.

Similarly, other qualities of the child besides those we measure with our classic speech-language-hearing test batteries might also interact with fluency, such as intelligence, working memory, or social adjustment. For example, a particularly bright child may be adroit at repairing communication breakdowns or in avoiding them altogether. A child with a limited working memory may have difficulty in attending to a lengthy utterance, and may therefore experience excessive communication breakdowns, perhaps asking, “What was that again?” A child who has poor social adjustment may adopt a strategy of remaining silent during conversation.

Our previous research about conversational fluency has primarily focused on adults who have hearing loss, and adult’s use of communication strategies. For example, ^{Tye-Murray et al. (1995)} videotaped adult implant users conversing with both a familiar and then an unfamiliar communication partner. The conversations were transcribed (a labor-intensive and time-consuming process) and analyzed. The goal of this study was to evaluate adult implant users’ use of repair strategies and document the kinds of conversational behaviors they used to circumvent communication difficulties. We found that some people’s tact is to take longer speaking turns so the conversational partner has little opportunity to speak; thus, the implant user is rarely in a position to not recognize the partner’s spoken utterance. Conversely, another tact is for an implant user to take fewer and shorter speaking turns, so the partner is forced to bare the onus of maintaining the conversation. With this tact, if communication breakdown occurs, the implant user might get away with bluffing and pretending to understand.

In a similar study, ^{Tye-Murray and Witt (1996)} utilized a mean length turn (MLT) measure and an MLT ratio. In this study, mean length turn was computed by counting the number of words spoken on average per speaking turn. The MLT ratio was the ratio of the average MLT of the cochlear implant user and the average MLT of the communication partner. When the ratio approached one, it was assumed that speaking time during the conversation was shared, and that conversational fluency was high (see ^{Caissie & Rockwell, 1993}, for a similar example of assessing conversational fluency in adults who have hearing loss).

Four goals motivated the present investigation. First, we set out to describe objectively the conversational fluency of a large group of young cochlear implant users. To this end, we utilized the DYALOG software program (^{Erber, 1996}). Videotapes of conversation were collected. The software was then used to assess: 1) time spent in repairing communication breakdown; 2) time spent in silence; and 3) sharing of speaking time (mean length turn ratio). These latter two measures (2 and 3) were novel applications for the software package and allowed us to streamline the analysis process.

The second goal was to assess whether children who use cochlear implants approach the conversational fluency of children who have normal hearing. We were interested in determining whether cochlear implantation allows children to engage in conversation as easily as children who have normal hearing. Although traditional measures of communication abilities (such as speech intelligibility and speech recognition scores) provide important information to teachers and clinicians, it is in the ecologically relevant context of everyday conversations that the true value of cochlear implants may be assessed. How well can children carry on a conversation with an unfamiliar adult (using only spoken communication), a task that they many must engage in on an everyday basis?

The third goal was to account for the variability in performance. A priori, we assumed that the conversational fluency of 181 deaf children would vary greatly. We set out to identify those variables that might best predict fluency. Thus, if a child experiences difficulties when engaging in conversation, we might have direction for remediation.

Finally, we compared subjective data with our objective measures to obtain information about the construct validity of our measurement procedures. Construct validity is the extent to which our objective DYALOG measures capture the construct of conversational fluency.

Methods

Participants

Characteristics of the 181 children with hearing loss who participated in this study are described in detail elsewhere (^{Geers & Brenner, 2003}). Participants were 8 or 9 yr old at the time of testing. All were deafened before 3 yr of age and were implanted by 5½ yr of age (most under age 5). Although most of the children were reportedly deaf from birth, almost one fourth of them had some known etiology of deafness after birth.

Most of the children with cochlear implants (83%) were enrolled in mainstream classes with hearing children for at least part of each school day at the time of data collection. Children were about equally divided between oral (N = 98) and simultaneous communication (N = 83) (SC) modes at the time of testing. Ratings of classroom communication mode that were provided by the child’s parents included three levels of SC programs (1-an emphasis on sign language; 2-an equal speech and sign emphasis; and 3-an emphasis on speech) and three levels of Oral programs (4-cued speech; 5-auditory/oral; and 6-auditory/verbal). Average communication mode ratings for the preimplant and for each of the 4 yr after implantation were used to summarize this variable. A total of 89 children obtained average mode ratings between 1.0 and 3.9, indicating predominant placement in total communication classrooms. The remaining 92 children obtained average mode scores between 4.0 and 6.0, indicating predominant placement in oral classrooms.

Twenty-four children who have normal hearing (NH) from a local private elementary school were included for assessing the degree to which communication skills exhibited by the children with implants approached normal levels. They were of the same chronological age as the children with hearing loss and of approximately equal gender distribution.

Procedure

The children with cochlear implants completed a battery of tests designed to assess their language, speech intelligibility, speech recognition skills, cognitive abilities, and their social adjustment. The following measures were selected from the battery to examine their contribution to oral conversational fluency:

Speech Intelligibility

• Children imitated a clinician speaking each of the 36 ^{McGarr (1983)} sentences. Children’s responses were recorded on a DAT recorder. Individual sentences were computer edited and stored in wave files for presentation to normal-hearing adult subjects, who served as judges of speech intelligibility. Judges were asked to write down as much of the sentence as they could understand. Three judges provided responses for each sentence, for each child (see ^{Tobey et al., 2003}).

Auditory-Only Speech Perception

• The Bamford-Kowal-Bench (BKB) Sentence Test (^{Bamford & Wilson, 1979}) was administered from an auditory-only recording at 70 dB SPL, and the percent of keywords correctly imitated by the child was obtained.

Audio-Visual Speech Perception

• The Children’s Audiovisual Speech Enhancement Test (CAVET) (Tye-Murray & Geers, 1997) was administered from videotape in an auditory-plus-vision mode at 70 dB SPL. Children viewed the talker’s head and shoulders while she spoke the test items and repeated the word using their preferred communication modality.

Receptive Language

• The Test for Auditory Comprehension of Language-Revised (TACL-R) (^{Carrow, 1985}), a measure of syntax comprehension, was administered to all children using Simultaneous Communication (i.e., speech and signed English) regardless of the child’s regular mode of communication. The child pointed to the picture that best represented the sentence or language structure presented. Scores were expressed as age-equivalent scores compared with normally hearing age-mates.

Expressive Language

• Spoken language competence was measured in a videotaped 20-minute conversation with an examiner who did not use sign language. The child’s spoken words were transcribed orthographically and verified by trained teachers of the deaf. Counts were made of the number of words per utterance (WD/UTT), which was used as one estimate of expressive language level (see ^{Geers et al., 2003}).

Psycho-Social Adjustment

• Parent(s) completed the Meadow-Kendall Social-Emotional Assessment Inventory for Deaf and Hearing Impaired Students (^{Meadow, 1983}) for their child. Although this inventory was originally intended by the test authors to be completed by the child’s classroom teacher, it was administered in this case to the child’s parent(s), as teachers were not available to provide judgments.

Working Memory

• The Digit Span subtest of the Wechsler Intelligence Scale for Children–III (^{Wechsler, 1991}) was administered in the child’s preferred communication mode. Standard procedures from the test manual were used to determine the longest series of digits the child could repeat (forward span) and the longest series of digits the child could repeat in reverse order (backward span).

Conversational Fluency

• Each child was videotaped in an oral interview with an unfamiliar female adult (see ^{Geers et al., 2003}). An oral conversational fluency rating was obtained from the last 10 minutes of the conversation. During this portion of the conversation, sets of open-ended questions were used to stimulate conversation (e.g., What do you like to do in the summer/winter? What is your favorite movie/TV show?). The examiner was instructed to engage in a “normal” conversational interchange in which both parties contribute to the dialog. She was also instructed to make an effort to understand the child and to insure that the child understood her. The examiner spoke with out using sign language. Three different examiners, all teachers of the deaf, conducted the interviews; two of them spoke with the children with cochlear implants and another spoke with the children with normal hearing.

Each videotaped conversation was viewed by the same research assistant, who was trained in use of the DYALOG conversational analysis procedure (^{Erber, 1996}). In performing the DYALOG analysis, the research assistant watched each video tape three separate times, sitting before a computer keyboard. On the first viewing, the research assistant pressed the space bar each time the child spoke, and held it down for the duration of the child’s utterance. For the second viewing, she recorded the utterances of the conversational partner in a similar manner. Finally, during the third viewing, when the research assistant was well familiarized with the content of the conversation, she pressed the space bar at the onset of each communication breakdown and released it at the offset. From the computer records, the following measures were computed for each conversation: the percent of time the child talked during the conversation (Child MLT), the percent of time the examiner talked (Examiner MLT), the percent of time spent in silence, and the percent of time spent in communication breakdown. In addition, we computed the MLT ratio between child and examiner talk-time.

For a subset of 47 children, 26 SC and 21 Oral, we also obtained subjective measures from a panel of judges, using the following procedures. A 3-minute sample was excerpted from the center of each videotaped conversation. Ten judges, who were graduate students, viewed each sample and completed three separate rating scales after viewing each one. The first scale assessed judges’ impressions of the child, and asked them to indicate on scales from one to five whether the child was sociable-unsociable, self-sufficient-helpless, and cooperative-uncooperative. The second scale assessed judges’ reactions to a child, and included three items: “If I were talking to the child in this conversation, I would feel successful-unsuccessful; relaxed-anxious; motivated-unmotivated.” These two scales are similar to scales developed for a similar purpose by ^{Gagne, Stelmacovich, and Yovetich (1991)}. Finally, the third scale assessed the judges’ subjective impression of the conversations themselves. Judges expressed their agreement on a scale from one to five, where one indicates high agreement and five indicates high disagreement, indicating that “The child talked too much,” “There were awkward pauses,” and “There was a meaningful exchange of information.”

Results

Conversational Fluency of the Three Groups of Children

Table 2 summarizes the conversational fluency measures for the three groups of children. On average, the NH children spent virtually no time engaged in communication breakdowns (0.4% of their conversations, on average) whereas the SC children spent almost a quarter of their conversational time engaged in repairing communication breakdowns, almost twice as more time as did the Oral group (21% versus 9% of their conversations, on average). A 1-way analysis of variance revealed a significant main effect (F (2, 202) = 23.36, p < 0.0001). Post-test Bonferroni/Dunn tests indicated that the NH children spent significantly less time in communication breakdowns than either the Oral children (p < 0.02) or the SC children (p < 0.0001), and the Oral children spent significantly less time in breakdowns than did the SC children (p < 0.0001).

TABLE 2: Conversational fluency measures for the three subject groups, SC (simultaneous communication, N = 89 children), Oral (N = 92 children), and NH (normal-hearing, N = 24 children).

The NH children also spent less time sitting in silence during their conversations (5% of conversation time, on average) than did either the Oral children (15%) or the SC children (18%). A 1-way analysis of variance revealed a significant effect of group type ((2, 202) F = 24.3, p < 0.0001). The NH children spent less time in silence than either the Oral children (p < 0.0001) or the SC children (p < 0.0001), and there was not a statistically significant difference between the two implant groups (p > 0.01).

Finally, Table 2 presents a summary of the MLTs (i.e., percentage of speaking time) of the children and their conversational partners in the three subject groups. Interestingly, the SC children and their conversational partners demonstrated equal MLT ratios (on average 40:42, or 0.95), whereas the Oral children and the NH children tended to take longer speaking turns than did their partners (48:37 or 1.3 and 57:37 or 1.5, respectively).

Relationships between Speech Recognition and Speech Production/Language Skills and the Conversational Measures

Table 3 summarizes the relationship between the cochlear implant users’ measures of speech recognition (The BKB Sentence Test and The CAVET), speech production (The McGarr Sentence Test), and expressive language (MLU) and four indices of their conversational performance with their conversational partner: percent time the child and conversational partner spent sitting in silence, percent time the child and partner spent in communication breakdown, percent time a child talked during the course of the conversation, and percent time the conversational partner talked. Pearson correlations were computed to assess whether the production, recognition, and language measures related to the conversational measures.

TABLE 3: Pearson correlations between the The CAVET Sentence Test (a measure of audiovisual speech recognition), The BKB Test (a measure of auditory-only speech recognition), the McGarr Test (a measure of speech intelligibility), and MLU (a measure of language) and two indices of conversational fluency, % time spent in silence and % time spent in communication breakdown

Performance on the CAVET and the BKB Sentence Test was predictive of both the percent of time spent in communication breakdown and the amount of time spent in silence, with poorer performance being linked with increased amounts of communication breakdown and silence. Similarly, children who had better speech intelligibility and longer MLUs appeared to spend less of their conversational time in silence and less time in communication breakdown. Children with good speech intelligibility and longer MLUs tended to take longer speaking turns, whereas the conversational partner took shorter turns.

Accounting for Communication Breakdown

We next addressed which variables were most predictive of communication breakdown. In addition to MLU, the McGarr Sentence Test, the BKB Sentence Test, and the CAVET, we entered the TACL Age Equivalent Score, the Meadow Kendall Social Adjustment Score, the WISC Similarities score, and the WISC digit span score into a stepwise regression analyses. Together, these measures accounted for 68% of the variance in communication breakdown with the following measures making a significant contribution. Performance on the McGarr Sentence Test was highly predictive of communication breakdown, accounting for 60% of the variance. An additional 4% of the variance was predicted by the TACL, and then additional 2% by the CAVET and finally, an additional 2% explained by gender (with girls spending less time in communication breakdown on average, than boys). The remaining predictor variables did not account for the additional variance.

Subjective Measures of Conversational Fluency and Relation to Objective Measures

For the subgroup of 47 children who use cochlear implants, the judges’ impressions of a child (i.e., child is sociable-unsociable, self-sufficient-helpless, cooperative-uncooperative) averaged 2.0 (SD = 0.8) on the five point rating scale, with a score of 1 corresponding to the highest agreement with the right side of each of the three scales. Pearson correlations revealed a significant correlation between the amount of time spent in communication breakdown and judges’ impressions (r = 0.81, p < 0.01), with longer percentages associated with more unfavorable impressions. Correlations between judges’ impressions and percent time the child talked during a conversation and percent time spent in silence were not significant.

Judges’ reactions to a child (i.e., “I would feel successful-unsuccessful, relaxed-anxious, motivated-unmotivated”) averaged 2.3 (SD = 1.0), with a score of 1 reflecting the successful, a score of 5 representing “unsuccessful,” and so forth. Again, there was a significant correlation between judges’ reactions and time spent in communication breakdown (r = 0.84, p < 0.001). The more time spent in communication breakdown, the less favorable the reaction. In addition, there was a significant negative correlation between the percentage of time that child talked and judges’ reactions (r = −0.44, p < 0.01), with children who spoke little being less favorably perceived than children who spoke more. Time spent in silence was not related to judges’ reactions.

There appeared to be a strong relationship between the subjective and objective measures of the conversation itself. The correlation between the questionnaire item, “The child talked too much,” was significantly correlated with the percentage of time the child talked (r = 0.62, p < 0.01). On average, judges assigned this item a value of 3.9 (SD = 0.7), where a score of 1 = I strongly agree and a score of 5 = “I strongly disagree.” The correlation between the questionnaire item, “There were awkward pauses,” and the percentage of time spent in silence was also significant (r = 0.43, p < 0.01). On average, the judges assigned this item a score of 3.6 (SD = 0.7). The questionnaire item, “There was a meaningful exchange of information” was correlated with the percentage of time spent in communication breakdown (r = 0.88, p < 0.001). On average, judges expressed their agreement on this last item as a score of 2.3 (SD = 0.9).

Discussion

The results suggest that children who have used cochlear implants for 4 or 5 yr still experience poor conversational fluency when compared with children with normal hearing. They spend more time in communication breakdown and more time in silence. Although judges might find that they exchange meaningful information during their conversations, on average, they still do not achieve the highest ratings on a five-point scale on a questionnaire item that assesses agreement for the statement, “There was a meaningful exchange of information.” Scores for this item average 2.3 on a five-point scale.

Interestingly, the analyses revealed that some of the child variables most amenable to intervention best account for the variance in communication breakdown durations. The scores for the McGarr Sentence Test, an index of speech intelligibility, and the TACL, a score reflecting receptive language, accounted for 66% of the variance of the time spent in communication breakdown. One interpretation of this finding is that by focusing attention on improving children’s speech intelligibility and expanding their receptive language, we might improve their conversational fluency. This might be one justification for expanding speech/language therapy services for this group.

The decreased conversational fluency of implant users, compared with their normal-hearing peers, also suggests the need for explicit instruction in managing communication difficulties and for enhancing their conversational fluency. A few programs are available for this purpose. For instance, ^{Tye-Murray (1992)} presents a communication therapy program that includes instruction in good listening behaviors, requesting message clarifications, and correcting environmental listening problems and inappropriate talker behaviors. ^{Elfenbein (1994)} has developed a five-step program that helps children develop communication repair strategies. These kinds of instruction might be a valuable inclusion in the academic curricula, with special emphasis on the use of repair strategies.

If through instruction we can improve children’s ability to manage their communication breakdowns, we might yield an additional benefit besides improved conversational fluency. The present results showed that judges perceive children who spend less time in communication breakdowns more favorably than they perceive children who spend more time in communication breakdowns. They have both better impressions of the child and better reactions to their conversations. These findings suggest that communication therapy might enhance their social stature among those persons with whom they converse by minimizing their breakdowns.

The Oral children spent less time in communication breakdown than did the SC children. This finding might reflect overall better speech intelligibility in this group (see ^{Tobey et al., 2003}), especially because the speech intelligibility measure was the best predictor of communication breakdown. This finding suggests that children whose educational program emphasizes speech and auditory skill development might best be able to use their speech skills to converse with someone who does not know sign language.

The measures of conversational fluency used in this investigation appear to have good validity. There was a strong relationship between some of the objective measures of conversational fluency and the judges’ subjective impressions. These relationships relate to 1) the amount of time the child talked and impressions of whether the child talked too much; and 2) the amount of time spent in communication breakdown and the impressions of whether meaningful communication occurred. These relationships serve to validate the DYALOG procedure as a useful means to measure conversational fluency. In addition, Pearson correlations showed that the percent time spent in communication breakdown and percent time spent in silence were correlated with measures of speech intelligibility, language, auditory speech perception, and audiovisual speech perception. One would predict that children who had better expressive and receptive communication skills would also have better conversational fluency, and this was indeed the case.

Finally, as noted in the introduction, conversational fluency is sometimes considered to be high when conversational partners take about equal speaking turns. In this investigation, the normal-hearing children (and for the present study, they represented the gold standard by which the cochlear implant users were compared) took longer speaking turns than did the examiner, 57 seconds per turn on average versus 37 seconds per turn (MLT ratio = 1.5). The present finding might relate to the nature of the data collection procedure. The examiner was provided with a list of open-ended questions to stimulate conversation. It might be that when the child had good speech intelligibility and language, he or she could answer the questions in length, thereby increasing speaking time. When a child had poor speech intelligibility and language, and poor speech recognition skills, he or she might have had less to say in response to the questions and/or may not have comprehended the examiner’s stimulating remarks.

In summary, the present results suggest that children who use cochlear implants would benefit from receiving communication therapy. Many exhibit deficits in conversational fluency, and those who do are perceived less favorably as conversational partners than those who do not. In addition, we might promote extensive speech-language therapy for these children as the results showed that children with better intelligibility and better receptive language also tend to experience fewer communication breakdowns.

Acknowledgments:

This study was supported by Grant No. DC03100 from the National Institute on Deafness and other Communication Disorders (NIDCD) of the National Institutes of Health to Central Institute for the Deaf (CID). Conversational samples were elicited from the children with cochlear implants by Julia Biedenstein and LaShawn Cole and from the hearing children by Megan McAvit. DYALOG analysis of the videotaped conversations was conducted by Elizabeth Mauze. Christine Brenner conducted the statistical analysis of the data.