The Relationship of Audibility and the Development of Canonical Babbling in Young Children With Hearing Impairment

Children with hearing impairment.

Thirteen children (six males and seven females) with hearing impairment from various cities in Iowa participated in this study. The children began the longitudinal study at an average age of 13.6 months (range 4.5–19 months) and completed the study at an average age of 31.8 months (range 25–37.5 months). Initially, all participants with hearing impairment were HA users. Seven of the children were fit with HAs by 12 months of age and six were fit with HAs by 18 months of age. All participants with hearing impairment were considered potential CI candidates based on their poorer ear pure-tone averages (PTAs). All participants received a CI during the course of this study at an average age of 17 months (range 12–23 months). Better ear PTAs (average of thresholds at 500, 1000, and 2000 Hz) based on behavioral audiological evaluation and physiological testing ranged from 68 dBHL to no response at the limits of the audiometer. All children’s cognitive abilities appeared to be “within normal limits” in accordance with the Bayley scales of infant development—II’s “Mental” subscale.13 (Bayley, 1993). All children had no known visual abnormalities or motor delays. American English was the language spoken in each child’s home. In addition, all children utilized a system of total communication in the home that involved the use of sign language in American English word order. See Table 1 for demographic information on individual study participants.

Children with normal hearing.

Two children with normal hearing thresholds (both males) participated in this study. These children were selected from a group of normal hearing children being followed as part of a larger project. They were selected for inclusion in this study because they were the first two to reach 12 months of age by the end of data collection. The children began the study at 3 and 5 months of age and completed the study at 12 months of age. The children’s cognitive abilities appeared to be “within normal limits” in accordance with the Bayley scales of infant development—II’s “Mental” subscale.13 (Bayley, 1993). Neither had known visual abnormalities or motor delays. American English was the language spoken in each child’s home.

Data from these normal hearing children were included in the current study to illustrate that the data collection and analysis procedures were able to capture the emergence of canonical babble in the time frame indicated in the literature for children with normal hearing. These children were not intended to serve as a control group, yet they do represent one end of the continuum of audibility.

Vocalization coding.

Digital audio tape recordings from each participant’s sessions were coded using the !OBSERVE data collection software (Sander, 2001). The software allowed the user to assign keys of the computer keyboard to vocalization types of interest. During coding, a key was pressed that corresponded to the vocalization type heard. A second-by-second record of coding and a frequency count of each vocalization type were generated. Vocalizations were coded into one of the three types: (a) Canonical, (b) Word, and (c) Other. Vocalization types were based on the stages of infant vocal development of Oller (1980) and Stark (1980) and were coded once per breath group during a first pass of coding. A breath group was defined as a vocalization or grouping of vocalizations demarcated by a period of quiet of at least 2 s. Definitions used to code vocalization type as well as coding rules for use during the first pass of coding are included in the Appendix A.

A second pass of coding was conducted in order to determine the number of syllables produced during each session. A CSR was calculated for each session of each participant until the criterion of a CSR of at least 0.15 was established. A CSR was calculated by counting the number of canonical syllables in a session and dividing by the total number of syllables in that session (Oller, 2000; Oller & Lynch, 1992; Steffens et al., 1992). See Appendix B for detailed instructions for conducting the syllable count.

A second ratio, the CBUR, was calculated for all sessions of each participant. The CBUR was defined by this investigator as the number of canonical utterances (at the level of the breath group) divided by the total number of utterances (at the level of the breath group). The CBUR is a ratio to compare the quantity of canonical utterances with the child's overall utterance production.

Using independent trained listeners, interobserver agreement was obtained for 30% of each participant's sessions during the HA phase as well as 30% of each participant's sessions during the CI phase. For the participants with normal hearing, interobserver agreement was obtained for 30% of each participant's sessions. Agreement was calculated separately for each vocalization type. Agreement sessions were randomly selected for each participant. An agreement was defined as the same type and number of vocalization scored by both listeners within ±3 s of each other. A disagreement was defined as one listener scoring a vocalization that the other listener either did not score or disagreed regarding type or number. Agreement percentages were calculated based on 6-s interval-by-interval comparisons of the listeners’ records, in which the smaller number of responses in each interval was divided by the larger number of responses. These fractions were then summed across all intervals, divided by the total number of intervals in the session, and multiplied by 100 to obtain the percent agreement between the two listeners. See Table 2 for mean agreement scores and ranges for each participant.

An independent observer also redetermined the session/date of entry of each child into the canonical babbling stage. The observer did this by returning to the raw data and independently recounting syllables and recalculating the CSR for each session until the criterion (CSR = 0.15) for entry into the canonical babbling stage was met. For all 15 participants, there was 100% agreement regarding the date of entry into the canonical babbling stage.

Audibility.

In order to quantify audibility of speech for each child, the Speech Intelligibility Index (SII) was calculated. An SII is based on the ANSI S3.5-1997 and is highly correlated with the intelligibility of speech under a variety of adverse listening conditions. The SII is a measure designed to quantify the proportion of the speech signal that is audible to the listener, with enhanced weighting for the frequency regions of speech that carry the most information. The SII ranges from 0.0 to 1.0. An SII of 0.0 is consistent with none of the speech signal being audible, whereas an SII of 1.0 is consistent with normal hearing. In order to calculate an SII, the following information is required: (a) long-term average speech spectrum, (b) insertion gain for speech, and (c) thresholds in dBHL. Extrapolation and interpolation are used if necessary information is excluded.

In this article, the 1/3 octave band levels of the “hip position” from the SHARP program (Omaha, NE) (Stelmachowicz, Lewis, Kalberer, & Creutz, 1994) were used to calculate SII. All calculations of SII were based on the standard band importance function for standard speech. The hip position from the SHARP program was used to gain a more complete understanding of the audibility provided by a young child’s HAs in a typical listening situation (i.e., while being carried by the caregiver).

Insertion gain was also required in order to calculate SII. Insertion gain in 1/3 octave bands was calculated by taking each child’s measured real ear–aided response (REAR) and subtracting a norm-based real ear–unaided response (REUR) as well as the input level of the signal used to obtain an REAR. Real-ear measures for each child were obtained from the University’s Hearing Aid Center as well as from local providers, when available. The Hearing Aid Center verified HA fit using real-ear measures and the Desired Sensation Level fitting strategy (Moodie, Seewald, & Sinclair, 1994; Seewald, 1995). The Hearing Aid Center used a 65 dBSPL pure-tone sweep as the input signal to obtain REAR. REUR for calculation of the SII was obtained by using the age norms established by Kruger (1987) for infants and young children. In summary, the following equation was used to obtain insertion gain in 1/3 octave bands:

where IG = insertion gain, X_n = frequency of interest, REAR = real ear–aided response, REUR = real ear–unaided response, and IS = input signal.

The last requirement for calculation of SII was unaided threshold information. Unaided threshold information was obtained from chart review. The best most recent audiometric data were used for calculation of SII during HA use. See Table 3 for the SIIs calculated for each child during HA use. SII was calculated during CI use by inserting sound field behavioral thresholds, obtained with the child wearing the implant, in place of unaided threshold information. SII was calculated for the same speech spectrum as mentioned above. Insertion gain was not entered for the CI users. See Table 3 for the SIIs calculated for each child during CI use.

Participant age.

Chronological age was calculated for each child for each session and represented the child's actual age in months. All participants were born within 1 month of their due dates and, therefore, chronological age represented the child's age from date of birth. Hearing age was also calculated for each child for each session and represented (a) the difference in months between the date the child was fit with HAs and the session date, provided that best ear-aided audibility was equivalent to an SII of at least .35 with the “hip” spectrum or (b) if best ear-aided audibility was less than 0.35 with the “hip” spectrum, hearing age represented the difference in months between the date the child's CI was stimulated and the session date. An SII criterion of .35 for hearing age onset during HA use was chosen because after the data were examined (Figure 2), this value represented the cutoff between children who failed to develop canonical babble during HA use and children who developed canonical babble while wearing HAs.

Individual data: children with hearing impairment.

Two patterns of vocal development were observed for the participants with hearing impairment. In general, the first pattern consisted of the emergence of canonical babble during HA use with SIIs ≥.35. Five participants fit this pattern. The results of Pattern 1 are displayed in Figures 3 and 4. Axel, Norman, and Cathy (Figure 3) began canonical babble at 3.5, 9.0, and 6.5 months hearing age, respectively. Levi and Jane (Figure 4) began canonical babble at 7.0 and 1.0 months hearing age, respectively. Four of the five children with this pattern of vocalization development began first oral-word production during the course of this study. First oral words for these four children were documented at an average chronological age of 24.5 months (range 22–30.25 months). Finally, it should be noted that once children achieved a CBUR of ≥0.15, subsequent CBURs generally did not fall below this value.

In general, the second pattern consisted of the emergence of canonical babble during CI use. These participants’ SIIs were ≤0.31 during HA use and did not exceed .31 until after cochlear implantation. The results of Pattern 2 are displayed in Figures 5–7. Carrie, Sheila, and Robert (Figure 5) began canonical babble at 12.75, 4.5, and 8 months hearing age, respectively. Mary, Sandra, and Karl (Figure 6) began canonical babble at 4.75, 3, and 13.25 months hearing age, respectively. Ben and Kyra (Figure 7) began canonical babble at 6 and 12 months hearing age, respectively. Five of the eight children with this pattern of vocalization development began first oral-word production during the course of this study. First oral words for these five children were documented at an average chronological age of 27.9 months (range 21.75–37.5 months). Finally, it should be noted once children achieved a CBUR of ≥0.15, subsequent CBURs did not fall below this value.

Individual data: children with normal hearing.

Figure 8 shows the results of the two children with normal hearing. Data from these normal hearing children were included in the current study to illustrate that the data collection and analysis procedures were able to capture the emergence of canonical babble in the time frame indicated in the literature for children with normal hearing.

Mark's results are displayed in the top panel of Figure 8. Mark began canonical babble at 8 months hearing age. Larry's results are displayed in the bottom panel of Figure 8. Larry began canonical babble at 9 months hearing age. Occurrence of first word production was not documented for Mark or Larry during the course of this study. Finally, it should be noted that once a CBUR ≥ 0.15 was achieved, the trend for subsequent CBURs was increasing. Exceptions to this trend occurred, for both children, in sessions immediately following onset of canonical babble.

Contributors to age of onset of canonical babble.

In those children who began canonical babble during HA use (Figures 3 and 4), it was noted that four of the five had at least 3.5 months of HA use prior to the onset of canonical babble and all had SIIs of at least .35. A multiple linear regression analysis was run to examine the contributions of these two independent variables (SII and length of HA use) to the outcome measure (age of onset of canonical babble). Two children were not included in this analysis because detailed HA information was unavailable. Children with normal hearing were included in this analysis and assigned a length of HA use of “0” months. From this analysis of 13 children, the correlation coefficient was r = .743, R² = 0.552. Thus, we can say that the model, or linear combination of SII and length of HA use, accounts for 55% of the variation in age of onset of canonical babble. Furthermore, of the two variables in the model, only SII was a statistically significant predictor of age of onset of canonical babble (beta = −29.832, p = .006). Length of HA use had little impact on the model (beta = −0.0963, p = .807,). In fact, in a model with only SII as a predictor variable, the R² is 0.54. Thus, adding length of HA use only increases the R² on age of onset of canonical babble by 0.01%. It should be noted that the statistical analysis was repeated with the length of HA use as “12” months for children with normal hearing, and the results did not change.

Determination of audibility and changes in audibility over time.

It is possible and likely that audibility changed over time during HA use and also during CI use. Insertion gain provided by HA use was a variable required for calculation of the SII. It was sometimes the case that children had poorly fitting ear molds, and therefore, the volume setting of their HAs was reduced to prevent constant feedback. Reducing volume also results in reduced insertion gain and, therefore, a lower actual SII than predicted. In addition, each participant's audibility was determined by the best aided ear's SII. It is possible that a participant's poorer aided ear may contribute to audibility in some manner. Audibility of the speech signal may also have been reduced during an episode of otitis media. Episodes of otitis media were not followed for effect on unaided behavioral threshold; however, it is acknowledged that children with otitis media may have reduced thresholds. Finally, audibility may change depending on the type of message conveyed to children during daily routines. Audibility was calculated using the 1/3 octave band importance function for standard speech with the hip spectrum. It is likely that throughout the day, a child would receive different types of speech input in a variety of relative orientations to the speaker, both of which would affect audibility. Therefore, each participant's audibility during HA use may have been higher or lower than reported due to numerous factors.

Factors that may have affected audibility during CI use involve the process of tuning the CI. In theory, once a CI is provided, a child has access to the entire speech signal, with audibility determined primarily by the sensitivity setting of the microphone (Henry, McKay, McDermott, & Clark, 2000). In practice, a child's MAP (threshold and comfort levels of the CI) evolves over time. Many visits are required during the first year of CI use in order to fine tune a child's MAP. Manipulations in electrode threshold and comfort levels are common during the first year. In addition, depending on the microphone sensitivity, the child may have more or less audibility provided. A second factor that may affect audibility is the type of signal conveyed by the CI. The CI conveys speech as an electrical signal, and the HA conveys speech as an acoustic signal. Therefore, an SII with the CI may not equal the same SII during HA use. A third factor that may affect audibility during CI use is the use of a HA in the nonimplanted ear. Several study participants continued to wear a HA in the nonimplanted ear (Cathy, Jane, Levi, and Norman). Interestingly, the participants who did so were the same participants who met criterion for canonical babble onset during HA use and were the participants with the greatest audibility during HA use (SII ≥ 0.35). For children who wore both a HA and a CI, it is possible that the audibility provided was different from that predicted solely from CI use (Ching, Psarros, Hill, & Smither, 2002). Finally, as with HA users, it is likely that throughout the day, the child may receive different types of speech inputs in a variety of relative orientations to the speaker, both of which will affect audibility. Therefore, each participant's audibility during CI use may have been higher or lower than reported due to numerous factors.

The manner in which age of onset of canonical babble was determined.

Although the results of this study support that age of onset of canonical babble decreased as audibility of the speech signal increased, actual age of onset of canonical babble was unknown for study participants with hearing impairment. Because children were evaluated approximately once every 2 months, the age of onset reported was a conservative estimate and represented the oldest possible age of onset of canonical babble for each participant (i.e., canonical babble onset could have occurred at any point between the last visit during which canonical babble was not apparent and the first visit during which canonical babble was apparent).

Age of identification of hearing loss and subsequent HA fitting.

Age of identification of hearing loss was an additional factor that potentially affected the outcome of this study. This variable determines when the HA fitting takes place and when early intervention services begin. Hearing losses experienced by the participants of this study were identified between 3 and 15 months of age.

Age of early intervention as well as frequency and intensity of intervention.

A final factor that may have affected the outcome of this study was the amount of early intervention provided to the participants. It was routinely noticed that following cochlear implantation, children received more frequent and more intensive early intervention services. Furthermore, it was not uncommon for a child to shift from having an early interventionist visit the home once per month during HA use to receiving in home therapy weekly during CI use. The amount and intensity of early intervention services varied across participants during HA and CI use. Although no formal evaluation of early intervention services was conducted, they clearly differed among participants and potentially contributed to the study findings. Early identification of hearing loss and subsequent early intervention have been shown to be major predictors of future speech and language development (Moeller, 2000; Yoshinaga-Itano, Sedey, Coulter, & Mehl, 1998). Therefore, it is likely that age of identification and early intervention did play a role in the outcome of this study.

Limitations.

There are at least four limitations to this article. First, the children with hearing loss who participated in the study were all diagnosed with severe–profound, sensorineural hearing impairment and had no apparent cognitive or motor disabilities. Thus, applicability of the present results to other degrees and types of hearing impairment and to children with cognitive and motor disabilities is limited. Specifically, for some children with motor “speech” impairments, canonical babbling may not be an accurate indicator of benefit from HA use (Kishon-Rabin et al., 2005). It is safe to infer, however, that children with hearing impairment that is not as severe would benefit similarly from strategies that maximize audibility. Second, age of enrollment prevented a detailed within-participant comparison of vocal development between the HA and CI phases. That is, children who were enrolled at older ages had fewer vocalization samples prior to receiving a CI. One potential benefit of the range of age of enrollment is that it provided varying lengths of time spent within the HA phase and the CI phase and thus showed that criterion for audibility and not chronological age accounted for vocal development. This effect is demonstrated by the single subject graphs displayed in a nonconcurrent multiple baseline format. Third, the number of sessions conducted during the HA and CI phases for some participants can be seen as a limitation. Specifically, for four participants, too few data points were obtained during the HA phase to determine a trend in the data. However, taken collectively, the direction in the overall rate of vocalizations was similar across all participants. Finally, the session length (number of minutes of data collection for any given session) varied per visit and across participants as described in the Method section. This variation could account for a systematic difference in the results, in that longer sessions might result in greater CBURs or vice versa. To investigate this issue, a Pearson Product Moment Correlation was calculated using the number of minutes of each session and the CBUR calculated for that session. The resulting correlation showed no significant relationship with a small correlation coefficient of .168.