Skip to main content
NCBI home page
Journal of Speech, Language, and Hearing Research : JSLHR logoLink to Journal of Speech, Language, and Hearing Research : JSLHR
. 2015 Jun;58(3):987–1000. doi: 10.1044/2015_JSLHR-H-14-0135

Are Young Children With Cochlear Implants Sensitive to the Statistics of Words in the Ambient Spoken Language?

Ling-Yu Guo 1,✉, Karla K McGregor 2, Linda J Spencer 3
PMCID: PMC4490103  NIHMSID: NIHMS694277  PMID: 25677929

Abstract

Purpose

The purpose of this study was to determine whether children with cochlear implants (CIs) are sensitive to statistical characteristics of words in the ambient spoken language, whether that sensitivity changes in expected ways as their spoken lexicon grows, and whether that sensitivity varies with unilateral or bilateral implantation.

Method

We analyzed archival data collected from the parents of 36 children who received cochlear implantation (20 unilateral, 16 bilateral) before 24 months of age. The parents reported their children's word productions 12 months after implantation using the MacArthur Communicative Development Inventories: Words and Sentences (Fenson et al., 1993). We computed the number of words, out of 292 possible monosyllabic nouns, verbs, and adjectives, that each child was reported to say and calculated the average phonotactic probability, neighborhood density, and word frequency of the reported words.

Results

Spoken vocabulary size positively correlated with average phonotactic probability and negatively correlated with average neighborhood density, but only in children with bilateral CIs.

Conclusion

At 12 months postimplantation, children with bilateral CIs demonstrate sensitivity to statistical characteristics of words in the ambient spoken language akin to that reported for children with normal hearing during the early stages of lexical development. Children with unilateral CIs do not.

Children with cochlear implants (CIs) tend to demonstrate more limited spoken vocabulary skills than children with normal hearing of the same chronological age on standardized tests (Geers, Moog, Biedenstein, Brenner, & Hayes, 2009; Johnson & Goswami, 2010) and in spontaneous speech (Geers, Nicholas, & Sedey, 2003). For instance, only about 50% of 153 five- and six-year-old children with CIs in Geers et al. (2009) had achieved age-appropriate performance on standardized tests of receptive and expressive vocabulary. The delayed lexical development in children with CIs may result from the deprivation of auditory input before implantation and/or slower-than-typical growth rates of vocabulary afterward (Connor, Craig, Raudenbush, Heavner, & Zwolan, 2006; but see Hayes, Geers, Treiman, & Moog, 2009). Previous studies have identified a variety of predictors for the rate of lexical development in children with CIs, such as age of implantation, speech perception skills, and parental education levels (Connor, Hieber, Arts, & Zwolan, 2000; Geers & Nicholas, 2013). Although these studies are informative, they do not reveal much about the underlying mechanisms or processes that could influence word learning in children with CIs.

An emerging line of research focuses on the roles of cognitive mechanisms and word-learning processes in lexical outcomes of children with CIs. This research indicates that lexical deficits in children with CIs may be associated with limited verbal working memory span (Conway, Pisoni, Anaya, Karpicke, & Henning, 2011), slow processing speed (Grieco-Calub, Saffran, & Litovsky, 2009), and/or impaired executive function (Fagan, Pisoni, Horn, & Dillon, 2007). In addition, as compared to children with normal hearing, children with CIs show reduced attention to sound sequences of words (Houston, Pisoni, Kirk, Ying, & Miyamoto, 2003), delayed linking of spoken words to their visual referents (i.e., fast mapping; Houston, Stewart, Moberly, Hollich, & Miyamoto, 2012), and reduced ability to maintain lexical representations over time (Walker & McGregor, 2013). Taken together, these studies suggest that children with CIs demonstrate deficits in word-learning processes, which could result from inefficient cognitive mechanisms.

In the current study, we asked whether the deficient word-learning processes in children with CIs manifest as atypical statistical patterns in the lexicon. Specifically, we asked whether children with CIs are influenced by phonotactic probability, neighborhood density, and word frequency in early lexical acquisition. Children with normal hearing tend to learn words with rare sound sequences, dense neighborhoods (i.e., words with many similar sounding words; see below), and high (token) frequency earlier than those with common sound sequences, sparse neighborhoods, or low frequency (Goodman, Dale, & Li, 2008; Stokes, 2010; Storkel, 2004, 2009). To demonstrate these patterns, children must perceive the segmental components and their sequences in words and track statistical distributions over words. Given that children with CIs receive a degraded, electrical signal instead of an acoustic signal (Wilson, 2006) and demonstrate impaired implicit learning of sequential information (Conway, Karpicke, et al., 2011), they may be insensitive to the regularities conveyed by phonotactic probability, neighborhood density, and word frequency in early lexical acquisition. To test this hypothesis, the current study examined phonotactic probability, neighborhood density, and frequency of words that were produced by children who received implants before 24 months of age and who had 12 months of experience in using CIs. We also compared children with unilateral CIs to those with bilateral CIs to determine whether both groups demonstrate sensitivity to word statistics in the process of lexical acquisition similar to children with normal hearing. In what follows, we first review the roles of phonotactic probability, neighborhood density, and word frequency in lexical development in children with normal hearing and in children with CIs. We also review studies that compare the language outcomes in children with unilateral CIs and in those with bilateral CIs, and then we lay out the scope of the present study.

The Emerging Lexicon of Children With Normal Hearing

Phonotactic probability refers to “the likelihood of occurrence of individual sounds and sound sequences” in words (Hoover, Storkel, & Hogan, 2010, p. 101). For instance, the phonotactic probability of the word dog can be obtained by computing the likelihood of the occurrence of /d/ at the first position, /ɔ/ at the second position, and /ɡ/ at the third position and by computing the likelihood of occurrence of /dɔ/ at the first position and /ɔɡ/ at the second position. A word that is composed of common sounds and sound sequences is one with high phonotactic probability (e.g., cat /kӕt/), whereas a word that is composed of rare sounds and sound sequences is one with low phonotactic probability (e.g., those /ðoz/). Phonotactic probability is considered a property of the phonological representation of words because it indexes the likelihood of phonemes and phoneme sequences of words (Storkel, 2009).

Unlike phonotactic probability, neighborhood density is considered a property of the lexical representation of a word because it quantifies whole-word similarity between that word and other words. The neighborhood density of a given word is operationally defined as the number of words that differ from this word by one phoneme substitution, addition, or deletion (Luce & Pisoni, 1998). For instance, the lexical neighbors of the word seek may include, but are not limited to, peek, speak, and see. A word that has many lexical neighbors (e.g., sit) resides in a dense neighborhood, whereas a word that has few lexical neighbors (e.g., these) resides in a sparse neighborhood (Storkel, 2004). Notice that there is necessarily a correlation between phonotactic probability and neighborhood density. If a given sequence of sounds is common in the language, it will be high in phonotactic probability and, often, high in neighborhood density. Nevertheless, these constructs also exhibit a degree of independence (Vitevitch, Armbrüster, & Chu, 2004). For example, bug is relatively low in phonotactic probability, but it has many neighbors (e.g., but, bum, bag, big, tug, rug). Thus, phonotactic probability and neighborhood density can have independent effects on lexical acquisition (e.g., Hoover et al., 2010; Storkel & Lee, 2011) as well as other processes (for a review, see Storkel, Armbrüster, & Hogan, 2006).

Word frequency is operationally defined as the number of occurrences of a given word in a corpus of adult-directed speech (adult-to-adult speech; e.g., Stokes, 2010), adult-directed written language (e.g., Storkel, 2009), child-directed speech (adult-to-child speech; e.g., Goodman et al., 2008), or children's spontaneous speech (e.g., Storkel, 2004). Word frequency is orthogonal to both phonotactic probability and neighborhood density. For instance, the word those is low in phonotactic probability and neighborhood density but is used with great frequency in English.

Several studies have examined how phonotactic probability, neighborhood density, and word frequency may affect the age of acquisition of early words using the normative data from MacArthur–Bates Communicative Development Inventories (CDI) provided by its creators (Fenson et al., 1993, 2007). Storkel (2004) investigated the extent to which the age of acquisition of nouns in the CDI normative data (Fenson et al., 1993) is associated with neighborhood density and word frequency. Nouns from the infant version—CDI: Words and Gestures (CDI:WG)—and the toddler version—CDI: Words and Sentences (CDI:WS)—were combined. Age of acquisition of a given noun was operationally defined as the earliest age when at least 50% of the children in the normative sample were reported to say the word. Neighborhood density of each noun was computed using the Hoosier Mental Lexicon database (Nusbaum, Pisoni, & Davis, 1984). Word frequency of each noun was determined using a corpus of children's spontaneous speech from Moe, Hopkins, and Rush (1982), which reflects the expressive vocabulary of first-graders. The results showed that, as age of acquisition of nouns increased, both neighborhood density and word frequency decreased. That is, earlier acquired nouns tended to have higher frequency and neighborhood density than later acquired nouns. Only as the child amasses a larger vocabulary does he or she often map words that are less frequent or words from sparser neighborhoods.

Using different corpora, Storkel (2009) extended her inquiry to include consideration of phonotactic probability. Specifically, she asked the extent to which the age of acquisition of nouns on the CDI:WS form (Fenson et al., 1993) can be accounted for by phonotactic probability, neighborhood density, and word frequency. She did replicate the finding that earlier acquired words tend to come from denser neighborhoods than later acquired words. Somewhat counterintuitively, she also found that early acquired words tended to have lower phonotactic probability than later acquired words. Here the partially orthogonal relationship between neighborhood density and phonotactic probability is revealed in disparate influences on the child's developing lexicon. Furthermore, these disparate patterns can be explained: Words with rare sound sequences are more likely to be recognized as new words and, therefore, to trigger word learning than words with common sound sequences (Storkel, 2009; Storkel, Maekawa, & Hoover, 2010). On the other hand, adults tend to hyperarticulate words from dense neighborhoods to distinguish them from similar sounding words (Scarborough, 2012). Thus, words from dense neighborhoods are perceptually more salient and, consequently, easier to learn (Stokes, 2010).

In contrast to Storkel (2004), there was no relationship between word frequency and age of acquisition in Storkel (2009); however, word frequency in Storkel (2009) was derived from a corpus of adult-directed written language (Kučera & Francis, 1967) instead of a corpus of children's spontaneous speech or child-directed speech, and this might account for the difference. In fact, Goodman et al. (2008), like Storkel (2004), reported a negative correlation between word frequency in child-directed speech and age of acquisition in the expressive lexicon. Moreover, they reported this relationship to disappear when a corpus of adult-directed written language (i.e., Kučera & Francis, 1967) was used to estimate word frequency. Based on Goodman et al. (2008), we conclude that the relationship between word frequency and age of acquisition is best examined with estimates of word frequency that capture input directed to children (i.e., child-directed speech) as compared to estimates from adult-directed language.

Instead of examining age of acquisition of words from the CDI normative sample, Stokes (2010) collected her own sample. She explored children's sensitivity to statistical information in the ambient language by looking at the average neighborhood density and word frequency of words produced by children with small or large vocabulary sizes. Based on previous studies (Goodman et al., 2008; Storkel, 2004), Stokes hypothesized that children's expressive vocabulary size should negatively correlate with mean word frequency and neighborhood density of the words that they produce. To test her hypotheses, Stokes asked the parents of 232 British English-speaking children between the ages of 24 and 30 months to complete a British adaptation of the CDI:WS form (Klee & Harrison, 2001). Vocabulary size of each child was computed by tallying the number of monosyllabic nouns, verbs, and adjectives on the British CDI:WS form that a child produced. The average neighborhood density and word frequency of the lexical items that a child produced were calculated. Neighborhood density and word frequency in this study were estimated using a corpus of adult-directed speech established by De Cara and Goswami (2002). The results showed that children's vocabulary size negatively correlated with neighborhood density, as predicted, but positively correlated with word frequency. That is, as vocabulary size increased, neighborhood density decreased, and word frequency increased. Those findings were replicated in French-speaking children (Stokes, Kern, & dos Santos, 2012) and in Danish-speaking children (Stokes, Bleses, Basbøll, & Lambertsen, 2012).

Taken together, these results suggest that young children with normal hearing are sensitive to the statistics of the ambient language they hear. This sensitivity is manifested as changes in the statistics of the child's own lexicon over developmental time. Words with low phonotactic probability tend to be acquired earlier than those with high phonotactic probability (Storkel, 2009; Storkel et al., 2010). Density also varies over development. Words with high neighborhood density tend to be acquired earlier than those with low neighborhood density (Stokes, 2010; Storkel, 2009; Storkel et al., 2010). The effect of word frequency has been more difficult to determine. Whereas some studies found that word frequency was negatively correlated with age of acquisition (Goodman et al., 2008), others found that the correlation was positive or was nonsignificant (Storkel, 2009). The discrepancy between studies may have occurred because different types of corpora (e.g., adult-directed or child-directed corpora) were used to estimate word frequency. Given that studies reporting negative correlations (i.e., earlier age of acquisition associated with highly frequent words) employed child-directed rather than adult-directed corpora for determination of frequency, we view this relationship as the more likely. Laboratory studies in which young children are taught novel words reveal that frequent exposure to a given word makes for a stronger representation of that word in lexical memory (Riches, Tomasello, & Conti-Ramsden, 2005) and that learners typically need more exposures to add a word to the expressive lexicon than to the receptive lexicon (Gordon & McGregor, 2014).

The Emerging Lexicon in Children With CIs

To the best of our knowledge, no published studies have determined the roles of phonotactic probability, neighborhood density, or word frequency in the emerging lexicon in children with CIs; however, two studies have examined the validity of the CDI in documenting emerging language skills in this population. Stallings, Gao, and Svirsky (2000) examined the extent to which the number of words produced on the CDI:WS (Fenson et al., 1993) correlated with the Peabody Picture Vocabulary Test–III (PPVT-III; Dunn & Dunn, 1997) and with the Reynell Development Language Scales (RDLS; Reynell & Huntley, 1985) in 18 children who received cochlear implantation between the ages of 23 and 56 months at 12 months postimplantation. At 12 months postimplantation, the number of words produced on the CDI:WS ranged from four to 667 (M = 364.94, SD = 185.83). The number of words produced on the CDI:WS significantly correlated with the raw scores of the PPVT-III (r = .64) and with the raw scores of the receptive (r = .88) and expressive (r = .87) language scales of the RDLS in children with CIs. Thal, DesJardin, and Eisenberg (2007) further confirmed the validity of CDI:WS (Fenson et al., 2007) using the RDLS and language sample measures (e.g., word types) in 24 children who were prelingually deafened and who had 3 to 60 months of experience in using CIs.

Language Outcomes in Children With Unilateral Versus Bilateral CIs

There has been a debate regarding whether bilateral CIs should be the standard for children who are prelingually deafened (Papsin & Gordon, 2008). Presumably, bilateral CIs should generate additional benefits over uniltateral CIs because of binaural processing (Sarant, Harris, Bennet, & Bant, 2014). In a review of 29 studies on pediatric CI users, Johnston, Durieux-Smith, Angus, O'Connor, and Fitzpatrick (2009) concluded that children using bilateral CIs tend to show better performance in sound localization and speech recognition in noise than those with unilateral CIs.

The advantage of bilateral CIs over unilateral CIs on language acquisition, however, remains unclear given mixed findings from different studies (Boons et al., 2012; Lammers, Venekamp, Grolman, & van der Heijden, 2014; Niparko et al., 2010). Nittrouer and Chapman (2009) compared the language outcomes in children with unilateral CIs (but no hearing aids in the other ear), those with unilateral CIs plus hearing aids, and those with bilateral CIs using a standardized expressive vocabulary test and language sample measures (i.e., mean utterance length and use of pronouns). The age of first implantation was similar across the three groups, with a mean of approximately 18 months. Among those with bilateral CIs, seven were simultaneously implanted, and 19 were sequentially implanted (mean age of second implantation = 32.3 months). All children were tested at 42 months of age (±1 month). Thus, the mean length of device use across children at the time of testing was approximately 24 months. Overall, there were no group differences in the expressive vocabulary test or in the language sample measures. Because most of the children with bilateral CIs were implanted sequentially, Nittrouer and Chapman (2009) reasoned that the lack of significant differences between the unilateral and the bilateral groups may have occurred because children with bilateral CIs did not have sufficient experience with bilateral processing. To address this issue, the authors conducted another analysis in which only children with bilateral CIs who had at least 12 months of experience using two CIs were included for comparisons. Yet, the results remained the same. Nittrouer and Chapman (2009), therefore, concluded that there was no trustworthy evidence supporting the practice of bilateral cochlear implantations for children who are prelingually deafened at early ages. This finding was consistent with a large-scale longitudinal study by Niparko et al. (2010), which did not find the advantage of bilateral CIs over unilateral CIs on the developmental trajectories of receptive or expressive language.

Sarant et al. (2014) compared language outcomes in 5- and 8-year-old children with unilateral or bilateral CIs using the Peabody Picture Vocabulary Test–Fourth Edition (PPVT-4; Dunn & Dunn, 2007), the Preschool Language Scale–Fourth Edition (PLS-4; Zimmerman, Steiner, & Pond, 2002), and the Clinical Evaluation of Language Fundamentals–Fourth Edition (CELF-4; Semel, Wiig, & Secord, 2003). The PPVT-4 was administered for both age groups. The PLS-4 and CELF-4 were administered for 5- and 8-year-olds, respectively. The mean age of first implantation was approximately 24 months for the unilateral group and 16 months for the bilateral group. The mean age of the second CIs for the bilateral group was approximately 35 months; therefore, these children were sequentially implanted. The mean length of device use was 62 months for the bilateral group and 55 months for the unilateral group. When child characteristics (e.g., length of device use) and family characteristics (e.g., parent education) were controlled, children with bilateral CIs outperformed those with unilateral CIs on the PPVT-4 and the CELF-4, but not on the PLS-4. Thus, Sarant et al. (2014) concluded that bilateral CIs do benefit vocabulary and general language acquisition (at least as manifested on one measure) over unilateral CIs. This finding was consistent with Boons et al. (2012), who found that children with bilateral CIs performed significantly better than those with unilateral CIs on receptive and expressive language tests at 3 years postimplantation.

In summary, whereas Sarant et al. (2014) found a bilateral advantage for lexical acquisition, Nittrouer and Chapman (2009) did not. It could be that the advantage emerges over time; after all, the participants in Sarant et al.'s study were older. Here, we set this issue aside. We aimed to examine potential benefits of bilateral implantation as manifested in a more normal process, rather than product, of early word learning. Whether or not it is the case that the two groups differ in vocabulary size, our greater interest was whether the pattern of relationships between the number of words in the spoken lexicon and the statistical characteristics of the vocabulary words in the spoken lexicon reveal normal processes of word learning. It is possible that bilateral processing may reduce the listening effort (Hughes & Galvin, 2013), which allows children with bilateral CIs to allocate more attention and/or other cognitive resources to the process of lexical acquisition (Houston et al., 2003). Thus, children with bilateral CIs may process the sound sequence of words in a greater detail, and hence demonstrate stronger sensitivity to word statistics, than children with unilateral CIs.

The Present Study

To determine the statistical characteristics of early vocabulary in children with CIs, we evaluated the relationship between vocabulary size and phonotactic probability, neighborhood density, and word frequency in children who received CIs before 24 months of age and who had 12 months of experience in using CIs. We also evaluated whether the relationship between vocabulary size and vocabulary characteristics would vary in children with uniltateral CIs and in those with simultaneous, bilateral CIs. To these ends, archival data of the CDI:WS forms (Fenson et al., 1993) were obtained from a longitudinal project (see below) for analysis.

The specific questions addressed were as follows: (a) To what extent does vocabulary size correlate with phonotactic probability, neighborhood density, and word frequency in children with CIs as a whole? (b) Do the patterns of correlations between vocabulary size and phonotactic probability, neighborhood density, and word frequency differ in children with CIs who received unilateral cochlear implantation and in those who received bilateral cochlear implantation? Based on the studies on children with normal hearing, we predicted that vocabulary size should positively correlate with phonotactic probability and negatively correlate with neighborhood and word frequency in children with CIs despite the constraints in speech perception and statistical learning. Finally, we predicted these correlations might be stronger for children with bilateral CIs than for those with unilateral CIs because of the advantage of bilateral auditory processing (Johnston et al., 2009; Sarant et al., 2014).

Method

The current investigation used archival data from a longitudinal project of prelingually deafened children who received cochlear implantation (Dunn et al., 2014). The longitudinal project was approved by the Institutional Review Board at the University of Iowa. All children were recruited when they came to the University of Iowa Hospitals and Clinics for cochlear implantation. The research team obtained the consent from the parents who were willing to enroll their children in the longitudinal project. In the project protocol, the children were seen approximately five times during the first year after implantation and annually thereafter for device setting and follow-up and for data collection regarding speech, language, and hearing development. One part of the protocol required the parent to complete the CDI:WS form 1 year postimplantation to document the child's lexical and grammatical development. The parents of 48 children with CIs completed the CDI:WS form. To be included for the present study, children had to receive cochlear implantation before 24 months of age. This decision was made to reduce the effect of age of implantation. In addition, because the current population of pediatric CI recipients typically receives cochlear implantation at or before 24 months of age, this criterion makes our findings more applicable to the current population of CI users. Twelve children were excluded for not meeting this criterion.

Participants

Thirty-six children (20 boys, 16 girls) who received cochlear implantation before 24 months of age (M = 15.42 months, range = 10–23 months) were included in the current study. They received implants between the years of 1998 and 2010. Among these children, 16 received bilateral implants, and 20 received unilateral implants. The children with bilateral CIs all received implants simultaneously; therefore, potential confounds presented by differing time intervals between the first and second CIs were eliminated (Boons et al., 2012). The use of a hearing aid for the unimplanted ear was recommended for all of the children with unilateral CIs; four children with unilateral CIs followed this recommendation.

According to parent reports on the Minnesota Child Development Inventory (Ireton & Thwing, 1974), none of the children presented with frank signs of cognitive, motor, or social-psychological deficits. The children were from the midwestern United States—31 from Iowa; two from Missouri; and one each from Illinois, North Dakota, and Wisconsin—and all were speakers of standard American English. At the time of data collection, the mean chronological age of the participants was 27.6 months (SD = 3.10 months, range = 22–35 months). Socioeconomic status was based on maternal education, with 17% (6/36) having a postcollege degree, 50% (18/36) having a college degree, 30% (11/36) having a high school diploma, and 3% (1/36) taking some high-school courses.

All children received speech-language services from the early intervention program provided by the state at the time of data collection. They were enrolled in a total communication program (n = 24) or in an oral communication program (n = 12) for aural habilitation according to parent reports. The parents filled out an Educational and Special Services Information form developed by the team of the longitudinal project before implantation and at the 12-month postimplantation visit. On this form, the parents completed a communication section that indicated whether their children used sign language, and if so, whether they used American Sign Language, Signed English, or Signing Exact English. In addition, they completed a section that indicated whether their children enrolled in a program that included the instruction of sign languages. Children who were reported to enroll in an intervention program that included the instruction or use of sign languages were coded as being in a total communication program. In contrast, children who were reported to enroll in an intervention program that did not involve the instruction or use of sign languages were coded as being in an oral communication program. All children who were enrolled in a total communication program, except one child (CI-15), learned/used Signed English or Signed Exact English at home and during speech-language services. Child CI-15 was exposed to some American Sign Language during speech-language services, although the primary sign language he learned/used was Signed English. No children changed the intervention program (i.e., total communication vs. oral communication) between the preimplantation visit and the 12-month postimplantation visit.

The total communication program typically included weekly home visits by an early childhood specialist who provided a combination of sign language instruction, information on targeting speech and language development, and information on child development in general. Weekly visits lasted between 1 and 3 hr. The oral communication program was similar to the total communication program except that no sign language instruction was included.

Because the present study compared the lexicons of children with unilateral or bilateral CIs, Table 1 presents background information by group. The unilateral and the bilateral groups did not differ significantly in preimplantation hearing levels or years of maternal education, Fs < 1.74, ps > .20, ηp2 < .053. In addition, the unilateral and the bilateral groups did not differ significantly in the distribution of gender (χ2 = 0.01, p = .94) or communication mode (χ2 = 1.41, p = .23). Children with unilateral CIs were implanted primarily between the years of 1998 and 2004, whereas those with bilateral CIs were implanted primarily between 2006 and 2010. Although children with bilateral CIs tended to be implanted more recently and use newer models of speech processors than those with unilateral CIs, processing strategies were similar across the groups. For children with unilateral CIs, 85% (17/20) used the advanced combination encoder (ACE), 10% (2/20) used spectral peak coding, and 5% (1/20) used Advanced Bionics high resolution. For those with bilateral CIs, 81% (13/16) used ACE for both ears, and 13% (2/16) used high ACE (Hi-ACE) for both ears. One child (6%) with bilateral CIs used ACE for one ear and Hi-ACE for the other ear. Thus, the majority of the children in each group used the same processing strategy (i.e., ACE). Children with unilateral CIs, on average, were implanted 2 months later and consequently were 2 months older at the time of testing than those with bilateral CIs, F(1, 34) = 4.36, p = .04, ηp2 = .11. It should be noted that children's chronological age at the time of testing was their age of implantation plus 12 months. Thus, taking either age of implantation or chronological age into account for statistical analyses would yield the same results. For this reason, we controlled or partialled out only children's chronological age, instead of both age of implantation and chronological age, in the statistical analyses whenever appropriate.

Table 1.

Mean (standard deviation) of background variables by number of CIs.

Subjects Pre-CI hearing (dB HL)a Age of implantation (months) Chronological age (months)b Maternal education (years) Communication mode
All children (20 B, 16 G) 114.69 (9.54) 15.42 (3.14) 27.42 (3.14) 16.03 (2.35) 24 TC, 12 OC
Unilateral (11 B, 9 G) 114.05 (9.91) 16.35 (3.65) 28.35 (3.65) 15.60 (2.47) 15 TC, 5 OC
Bilateral (9 B, 7 G) 115.50 (9.32) 14.25 (1.88) 26.25 (1.88) 16.69 (2.06) 9 TC, 7 OC
Open in a new tab

Note. The number within parentheses indicates the standard deviation.

CI = cochlear implant; HL = hearing loss; B = boys; G = girls; TC = total communication; OC = oral communication.

a

Pre-CI hearing (dB HL) = mean pure tone average across 500, 1000, 2000, and 4000 Hz in dB HL for the better ear before cochlear implantation.

b

Chronological age at the time of participation in months.

Materials

At 12 months postimplantation, a parent of each child completed the CDI:WS, a parent report for children between the ages of 16 and 30 months. The CDI:WS form includes 680 words organized by categories (e.g., animals, food and drink, action words, and descriptive words) and other items related to grammatical development (e.g., utterance complexity). Parents were instructed to endorse each specific word and grammatical form they had observed their child to say, not sign, because the original purpose of using the CDI:WS in the longitudinal project was to document children's development of spoken language. For the purpose of the present study, we analyzed only the word endorsements.

Data Reduction, Coding, and Analysis

To compare the results to those of typically developing children in Stokes (2010), we followed her methodology and included only monosyllabic nouns, verbs, and adjectives in the analysis. Monosyllabic, instead of multisyllabic, words were included because it was difficult, although possible, to compute lexical neighbors for multisyllabic words, for example, the word television (Amano, 1996; Stokes, 2010). In addition, only nouns, verbs, and adjectives were included because words in other syntactic categories may show different acquisition patterns. For instance, Goodman et al. (2008) found that prepositions tended to have higher token frequency, but tended to be acquired later, than nouns, verbs, or adjectives. Including syntactic categories such as prepositions in the study might potentially make the frequency effect difficult to interpret. Questions words were also not included because some question words can be classified into different syntactic categories, depending on how they are used. For instance, the word what can be a pronoun (e.g., What is it?) or an adjective (e.g., What book is it?) in a sentence. The frequency of these words could not be determined unambiguously given the multiple syntactic categories (Goodman et al., 2008). Connecting words (e.g., so, then) were not included in the analysis for the same reason.

Thus, the following word categories on the CDI:WS form were excluded: sound effects and animal sounds, pronouns, question words, prepositions and locations, quantifiers and articles, helping verbs, and connecting words. Within the categories that were included, words that were not nouns, verbs, or adjectives (e.g., high) or were multisyllabic (e.g., water, pretend, empty) were excluded. Among the remaining 310 words, we found that nine words were listed in two word categories, including can, clean, drink, dry, fish, slide, swing, watch, and work. Following Goodman et al. (2008), we further excluded these words for analysis because the two meanings of the same word could not be distinguished computationally in the Child Language Data Exchange System (CHILDES; MacWhinney, 2000) corpora (see below), and thus the word frequency of these words could not be determined unambiguously. In addition, the word eye was excluded for analysis because the present study computed biphone frequency and this word had only one phoneme. The remaining words were 168 nouns, 88 verbs, and 36 adjectives, for a total of 292 words.

Coding phonotactic probability. The phonotactic probability of each word in the current study was determined using the online calculator developed by Storkel and Hoover (2010) and the associated child corpus. The child corpus consists of 6,412 words from the databases of Kolson (1960) and Moe, Hopkins, and Rush (1982). Two types of phonotactic probability were computed using the online calculator: positional segment sum and biphone sum. To compute the positional segment sum of a word (e.g., sit), the positional segment frequency of each component phoneme (e.g., /s/, /ɪ/, /t/) was first computed. The positional segment frequency of a target phoneme (e.g., /t/ in sit) was computed by the log frequency of words that had the target phoneme at a specific position (e.g., number of words that had /t/ at the third position) in the corpus divided by the log frequency of words that contained any sounds at the same position (e.g., number of words that had any phonemes at the third position). The positional segment frequency of each phoneme was then tallied as positional segment sum. Similarly, the biphone sum was computed by summing up the biphone frequency of each sound sequence in the target word (e.g., /sɪ/ and /ɪt/). The biphone frequency of a target phoneme sequence (e.g., /sɪ/ in sit) was computed by the log frequency of words that had the target phoneme sequence at a specific position (e.g., number of words that had /sɪ/ at the first position) in the corpus divided by the log frequency of words that contained any sounds at the same position (e.g., number of words that had any phoneme sequences at the first position).

Coding neighborhood density. Following Stokes (2010) and Storkel (2004), we used an adult corpus, that is, the Hoosier Mental Lexicon (Nusbaum et al., 1984), to estimate the neighborhood density of the vocabulary items. Recall that words with dense neighborhoods facilitate early lexical acquisition more than those with sparse neighborhoods because words with dense neighborhoods tend to be hyperarticulated and hence perceptually more salient in the input (Scarborough, 2012). It seemed to us that an adult corpus would be more appropriate to reflect this density effect than a child corpus because it is the adult's lexicon, not the child's lexicon, that leads to the hyperarticulation in the input.

The Hoosier Mental Lexicon consists of 19,290 words from the New Merriam-Webster Pocket Dictionary (1964). The online calculator developed by Storkel and Hoover (2010) also allowed the estimation of neighborhood density of words using the Hoosier Mental Lexicon. In the online calculator, the neighborhood density value of a word is defined as the number of words that have one-phoneme difference (i.e., substitution, deletion, or addition) from the target word (Luce & Pisoni, 1998). For instance, pit, it, and spit can all be counted as lexical neighbors of the word sit. The word sit has 36 neighbors in total. To be particularly cautious, we also computed the neighborhood density of the target words using the child corpus in Storkel and Hoover (2010). We found that neighborhood density estimated from the adult corpus and from the child corpus were highly correlated (r = .93, p < .001) and yielded the same results for the analyses in the current investigation. Thus, we reported only the results based on the adult corpus below.

Coding word frequency. The word frequency of each lexical item was estimated using a database of child-directed speech processed by Li, Burgess, and Lund (2000) from the CHILDES database (MacWhinney, 2000). We chose to use child-directed speech instead of adult-directed speech (e.g., Kučera & Francis, 1967) because Goodman et al. (2008) showed that word frequency derived from child-directed speech accounted for the age of acquisition of early words better than that from adult-directed speech. The database of Li et al. (2000) consists of 28 corpora from the CHILDES database: Bates, Belfast, Bernstein, Bliss, Bloom70, Bloom73, Brown, Clark, Cornell, Demetras1, Demestras2, Fletcher, Gathercole, Hall, Higginson, Howe, Kuczaj, Macboys (MacWhinney), Peters, Post, Sachs, Snow, Suppes, Valian, van Houton, van Kleek, Warren, and Wells. The ages of the children in these corpora were between 7 and 89 months. The majority (i.e., approximately 75%) of the transcripts involved children who were 48 months old or younger, which was close to the chronological ages of children in the present study. The total number of lexical items in the database of Li et al. (2000) has approximately 24,000 word types and 2.6 million word tokens. It should be noted that all inflected forms of a word are counted as separate types. For instance, wait, waited, waits, and waiting are counted as four word types. The frequency of each word type is available in the database.

The log frequency of each of the 292 words on the CDI form was estimated by summing up the raw frequency of the base form and the inflected forms (i.e., lemma frequency) of a given word in the Li et al. (2000) database and then converting the sum into a logarithmic scale with a base of 10. For instance, the log frequency of WAIT (3.44) on the CDI form was determined by summing up the raw frequency of wait (2,466), waited (24), waits (seven), and waiting (237) and then converting the sum (2,734) into a logarithmic scale. We chose to use log frequency, instead of raw frequency, to eliminate the skewness of distribution of raw frequency (Newman, Sawusch, & Luce, 1997). To be particularly cautious, we also estimated the word frequency using only the base, uninflected form (i.e., base word frequency) in the database. We found that the lemma frequency and base word frequency were highly correlated (r = .99, p < .001) and generated the same results for the analyses in the current investigation. Thus, we reported only the results based on lemma frequency below.

Analysis. Vocabulary scores were determined on the basis of the full form (680 words) of the CDI:WS. Vocabulary size was then computed by summing up the number of words that a child was reported to produce on the reduced form (292 words). We then computed the statistical characteristics of the vocabulary, that is, the average phonotactic probability, neighborhood density, and frequency across the words reported on the reduced form for each child. Pearson product-moment correlations were used to examine the relations between children's vocabulary size and vocabulary statistical characteristics. Because we used these multiple correlations to test preplanned hypotheses, we did not apply a Bonferroni correction. This decision was also consistent with previous studies on children with speech-language disorders (e.g., Conway, Karpicke, et al., 2011; Kover & Weismer, 2014).

Results

Preliminary Analyses of Vocabulary Scores

All Children With CIs. Table 2 presents the mean CDI:WS raw score and z score based on the full form for all children. The mean raw score was 193.69 (SD = 156.21, range = 16–602) for all children. We also compared the children's performance to their same-aged hearing peers in the normative sample in the CDI:WS. Given that the normative sample included children between 16 and 30 months, children with CIs in the current study who were older than 30 months (six children) were compared to 30-month-olds. The mean and standard deviation of each age group in the normative sample were used to compute the z score for each child. The resulting z scores ranged from −3.69 to 1.13. As a group, 22 (61%) of the 36 children with CIs scored more than 1.25 SDs below the mean (i.e., approximately the 10th percentile). It should be noted that even though six children with CIs were compared to children who were younger in chronological age (i.e., 30 months), five of them still scored more than 1.25 SDs below the mean. In addition, we further compared the vocabulary scores of children with CIs to that of children with normal hearing who had the same hearing age (i.e., 12-month-olds) using the normative data from the infant version of CDI (i.e., CDI:WG). Children with CIs were all at least 1.08 SDs above the mean (i.e. approximately the 86th percentile) when they were compared to children with normal hearing who had the same hearing age.

Table 2.

Descriptive statistics of spoken vocabulary score/size, phonotactic probability, neighborhood density value, and word frequency by number of cochlear implants.

Variables M SD Range
Spoken vocabulary score (out of 680 words)
 All children (N = 36) 193.69 156.21 16 to 602
 Unilateral (n = 20) 194.55 145.52 23 to 593
 Bilateral (n = 16) 192.63 173.53 16 to 602
Spoken vocabulary z score (out of 680 words)
 All children −1.37 1.00 −3.69 to 1.13
 Unilateral −1.54 1.10 −3.69 to 1.13
 Bilateral −1.15 0.83 −2.12 to 0.97
Spoken vocabulary size (out of 292 words)
 All children 68.83 60.91 7 to 252
 Unilateral 70.50 59.36 9 to 252
 Bilateral 66.75 64.69 7 to 221
Phonotactic probability segmenta
 All children .180 .017 .147 to .237
 Unilateral .182 .019 .161 to .237
 Bilateral .177 .014 .147 to .205
Phonotactic probability biphoneb
 All children .012 .002 .007 to .017
 Unilateral .012 .002 .008 to .017
 Bilateral .011 .002 .007 to .014
Neighborhood density
 All children 12.93 1.20 11.11 to 17.22
 Unilateral 12.76 1.17 11.11 to 17.22
 Bilateral 13.12 1.49 11.64 to 15.69
Word frequency (log)
 All children 2.58 0.09 2.40 to 2.89
 Unilateral 2.59 0.09 2.46 to 2.89
 Bilateral 2.57 0.09 2.40 to 2.76
Open in a new tab
a

Phonotactic probability segment = positional segment sum.

b

Phonotactic probability biphone = biphone sum.

We wondered what factors might contribute to the large range of vocabulary scores demonstrated by the children with CIs. Recall that children in the present study received CIs between the ages of 10 to 23 months. Thus, the chronological age of these children at the time of testing could range from 22 to 35 months because the data were collected at the visit of 12 months postimplantation. Because vocabulary scores increase with chronological age in children with normal hearing (Fenson et al., 1993), we speculated that the wide range of vocabulary scores in the present study may partly be explained by children's chronological age. Linear regression showed that 13% of the variability in the vocabulary score was accounted for by the children's chronological age (R2 = .13, p = .03). Children with CIs who were older at the time of testing tended to have higher vocabulary scores. Linear regressions further indicated that neither preimplantation hearing levels nor years of maternal education significantly accounted for the variance of vocabulary scores (R2 < .01, ps > .26). One-way analysis of variance indicated that boys (M = 167.60, SD = 149.68) and girls (M = 226.31, SD = 162.84) did not differ significantly in vocabulary scores, F(1, 34) = 1.27, p = .27, ηp2 = .036. For these reasons, we did not further consider the factors of preimplantation hearing levels, years of maternal education, or child's gender in the analyses. Instead, we took children's chronological age (and therefore age of implantation) into consideration whenever appropriate.

Children With Unilateral or Bilateral CIs. Above we reported a wide range of vocabulary scores at 12 months postimplantation. A portion of that range was accounted for by differences in age at implantation/current chronological age. As we turn to the comparison of statistical patterns in children with unilateral and bilateral CIs, we take the preliminary step of determining whether the difference in number of CIs might also contribute to the variability in vocabulary scores. A one-way analysis of covariance (ANCOVA) with group (unilateral, bilateral) as the between-subjects factor and chronological age as the covariate revealed no statistically significant differences, F(1, 33) = 0.59, p = .45, ηp2 = .017.

Table 2 also shows descriptive statistics of the vocabulary size based on the 292-word list, phonotactic probability (positional segment sum, biphone sum), neighborhood density, and word frequency by number of CIs. One-way ANCOVAs were conducted to compare the differences between the unilateral and the bilateral groups on those variables, with children's chronological age treated as covariates. The results showed that children with unilateral CIs and children with bilateral CIs did not differ significantly in vocabulary size, F(1, 33) = 0.49, p = .49, ηp2 = .014. In addition, the mean phonotactic probability, neighborhood density, or word frequency of vocabulary did not differ significantly across groups, Fs < 1.88, p > .18, ηp2 < .05.

Relations Between Vocabulary Size and Lexical Characteristics

All Children With CIs. Table 3 presents the correlation coefficients between vocabulary size and phonotactic probability, neighborhood density, and word frequency, with and without children's chronological age being partialled out. After partialling out children's chronological age, vocabulary size was positively correlated with phonotactic probability in biphone sum (p = .03) as predicted. However, neither the predicted positive relationship between neighborhood density and vocabulary size nor the predicted negative relationship between word frequency and vocabulary size was obtained. Could the mix of children with unilateral and bilateral CIs in this analysis have masked these predicted patterns? For this we turn to the final comparison.

Table 3.

Correlations between spoken vocabulary size and phonotactic probability, neighborhood density, and word frequency.

Subjects Phonotactic probability: Segment sum Phonotactic probability: Biphone sum Neighborhood density Word frequency
Age not partialled out
 All children .25 .46** −0.39** −0.29*
 Unilateral .08 .29 −0.24 −0.28
 Bilateral .54* .67** −0.58* −0.31
Age partialled out
#x2003;All children .14 .37* −0.25 −0.18
 Unilateral .01 .21 −0.12 −0.19
 Bilateral .40 .65** −0.47* 0.14
Open in a new tab
*

Significant at .05 level, one-tailed.

**

Significant at .01 level, one-tailed.

Children With Unilateral or Bilateral CIs. For children with unilateral CIs, vocabulary size did not correlate significantly with any of the phonological or lexical characteristics after partialling out age (ps > .20). For children with bilateral CIs, after partialling out age, vocabulary size remained positively correlated with phonotactic probability in biphone sum (p = .004) and negatively correlated with neighborhood density (p = .04), as predicted. There was no relationship between vocabulary size and word frequency or phonotactic probability in segment sum.

Discussion

One year after implantation, children with CIs were highly variable in spoken vocabulary scores, with z scores ranging from −3.69 to 1.13. This variability, which is consistent with previous studies (Stallings et al., 2000; Thal et al., 2007), was partly accounted for by children's chronological age. Despite having their CIs for identical periods of time, older children tended to have larger spoken vocabularies than younger children. Over half of the group (n = 22) demonstrated scores more than 1.25 SDs below the mean for their chronological age, and this too is consistent with previous reports of vocabulary limitations among children with CIs (Geers et al., 2009). Children with unilateral CIs or bilateral CIs did not differ significantly in spoken vocabulary scores, again a finding consistent with previous literature (Niparko et al, 2010; Nittrouer & Chapman, 2009).

The purposes of the present study were to determine whether vocabulary size correlates with phonotactic probability, neighborhood density, and word frequency in children with CIs and, if so, whether the patterns of correlations differ in children with unilateral CIs and in those with bilateral CIs. Contrary to prediction, vocabulary size was not significantly correlated with word frequency either in children with unilateral CIs or in those with bilateral CIs. Although we used a corpus of child-directed speech to estimate word frequency, we did not find the negative relationship between word frequency and vocabulary size that was reported for children with normal hearing in Goodman et al. (2008). Perhaps the relationship between word frequency and vocabulary size is elusive because frequency is a highly individual construct. Words frequently addressed to (and perceived by) children with normal hearing may differ from those frequently addressed to children with CIs. Future studies in which frequency is determined on the basis of recordings of parental language addressed to child study participants rather than published corpora would be a valuable step toward testing this hypothesis. We do not, at this point, conclude that children with CIs are unable to track frequency statistics. That conclusion does not seem valid given findings that they can indeed track the statistics of sublexical and lexical patterns, at least in cases of where implantation is bilateral. We explore these findings below.

Word Characteristics and the Early Lexicon in Children With CIs

In this study, we found that vocabulary size positively correlated with biphone sum measures of phonotactic probability, but only in children with bilateral CIs. Children with bilateral CIs who had larger vocabulary size were more likely to have learned words with common sound sequences than those who had smaller vocabulary size. The same pattern is found among children with normal hearing (Storkel, 2009; Storkel & Lee, 2011). Storkel (2009) suggests that words with rare sound sequences facilitate early lexical acquisition because these words are more likely to be recognized as novel words and hence trigger word-learning processes than are words with common sound sequences.

We also found that among children with bilateral CIs, and only among those with bilateral CIs, earlier words tended to be from dense neighborhoods, whereas later words were from sparse neighborhoods as is evidenced in the negative correlation between productive vocabulary size and neighborhood density. This finding is consistent with studies of children with normal hearing (Stokes, 2010; Storkel, 2004, 2009). Stokes (2010) suggests that words from dense neighborhoods tend to be acquired earlier than those from sparse neighborhoods because adults tend to hyperarticulate words from dense neighborhoods to avoid confusions, which consequently makes words from dense neighborhoods more salient. Because of the bilateral advantage on speech perception (Grieco-Calub & Litovsky, 2012; Hughes & Galvin, 2013; Lovett, Kitterick, Hewitt, & Summerfield, 2010), children with bilateral CIs may be able to benefit from these hyperarticulated patterns more than children with unilateral CIs.

In summary, despite late and reduced access to an acoustic signal, children with bilateral CIs demonstrated robust influences of word-level statistical patterns on their early lexicons. Words that have rare sound sequences or dense neighborhoods characterize the early lexicons of children with bilateral CIs, just as they characterize the lexicons of children with normal hearing (Goodman et al., 2008; Stokes, 2010; Storkel, 2004, 2009). Children with unilateral CIs, however, did not show the sensitivity to word characteristics in their early lexicons.

The difference between children with unilateral CIs and those with bilateral CIs in the sensitivity to word statistics cannot be attributed to preimplant hearing levels or maternal education levels because the groups did not differ on these variables. They also did not differ in processing strategies; the majority of children in each group used the processing strategy of ACE. However, they did differ in the models of their speech processors. There is no published evidence showing that different processors with the same processing strategy would lead to significant differences in speech and language outcomes (Wilson, 2006). There is also no published evidence consistently showing that different processing strategies from different makes at the same time period (e.g., ACE in Cochlear devices, Hi-Res in Advance Bionics devices) result in significant differences in outcomes. Therefore, although we acknowledge this confound, we think it is unlikely that it can account for the statistical sensitivity demonstrated by the bilateral group.

Although the distribution of communication mode did not differ significantly between groups, the bilateral group included more children using oral communication (i.e., 44% or 7/16) than the unilateral group (i.e., 25% or 5/20). One could argue that the statistical sensitivity demonstrated by the bilateral group may have resulted from the inclusion of more children using oral communication. We would like to point out that the findings on the outcomes of communication modes have been mixed (e.g., Connor et al., 2000; Kirk, Miyamoto, Ying, Perdew, & Zuganelis, 2000). Thus, there is no consistent evidence suggesting that children using oral communication would be more sensitive to the statistics of words than those using total communication or vice versa. To be particularly cautious, we conducted a post hoc analysis to see whether the relation between vocabulary size and word characteristics would vary with communication mode. Children using oral communication or total communication showed the same trend: After age was partialled out, spoken vocabulary size was not correlated with any of the word characteristics in either group. Therefore, the difference in the sensitivity to word statistics between children with unilateral and bilateral CIs is unlikely to be due to the proportion of children using oral/total communication in each group.

Thus, it seems that the number of CIs that children receive does make a difference in language acquisition as early as 12 months after implantation. Children with bilateral CIs demonstrate sensitivity to word statistics in early lexical acquisition in a manner similar to children with normal hearing, whereas those with unilateral CIs do not. The current study demonstrates a bilateral advantage on the process of early lexical acquisition. Why should this be? The statistical characteristics of spoken words are conveyed acoustically. If binaural input enhances perception of acoustic information (Grieco-Calub & Litovsky, 2012; Lovett et al., 2010), children with bilateral CIs may process the sound sequence of words in greater detail and thus demonstrate stronger sensitivity to word statistics than children with unilateral CIs. Moreover, binaural processing may reduce listening effort (Hughes & Galvin, 2013), which allows children with bilateral CIs to allocate more attention and/or other cognitive resources to the process of lexical acquisition (Houston et al., 2003; Just & Carpenter, 1992). In fact, it has been suggested that early auditory experience can affect the development of general cognitive processes that are not even specific to audition or spoken language (Conway, Pisoni, & Kronenberger, 2009; Houston et al., 2012). We cannot determine which of these reasons, if any, apply on the basis of the current data set, but we find it compelling that children with bilateral CIs demonstrate a more normal process of lexical acquisition.

However, if children with bilateral CIs are more sensitive to the statistical patterns of words than those with unilateral CIs, why did the bilateral and unilateral groups show no differences in the overall vocabulary scores? Why does a more normal process not lead to a more normal product? Recall that Sarant et al. (2014) found that children with bilateral CIs who, on average, had 3.64 years of bilateral CI experience demonstrated higher receptive vocabulary skills than those with unilateral CIs. Similarly, Boons et al. (2012) reported that children with bilateral CIs demonstrated higher expressive vocabulary skills than those with unilateral CIs at 3 years postimplantation. It is possible that the stronger sensitivity to word statistics in children with bilateral CIs may not be manifested in vocabulary scores at 1 year postimplantation. However, this sensitivity may gradually generate cascade effects such that a larger vocabulary is evident by 3 years postimplantation. Longitudinal studies that investigate vocabulary size and growth rate of vocabulary in children with unilateral or bilateral CIs beyond 1 year postimplantation are needed to address this issue.

Limitations and Future Directions

By taking advantage of an extant database from a genuine clinical setting, we faced limitations that must be considered. First, we had a small number of children with CIs for the unilateral group (n = 20) and for the bilateral group (n = 16). One concern is that the nonsignificant correlations between vocabulary size and word characteristics in the unilateral group may have resulted from insufficient statistical power. Although we are not able to rule out this possibility, we would like to point out that the bilateral group had fewer participants than the unilateral group, and yet some of the correlations between vocabulary size and word characteristics in the bilateral group were significant. Thus, the small number of participants may not fully explain the null findings in the unilateral group.

Second, the data are correlational not causal and cross-sectional not longitudinal, and thus change over developmental time was inferred by comparing children who had amassed smaller and larger vocabularies. In addition, only four (20%) of the 20 children with unilateral CIs in the present study used hearing aids for the unimplanted ear, limiting generalization of these findings to children who wear both a CI and a hearing aid. Also, we have acknowledged some confounds between the bilateral and unilateral CI groups. These include differences in the model of speech processor and in the proportion of children experiencing one or the other communication training mode. Because bilateral implantation became more common over time, it is also the case that the children with bilateral CIs were implanted later than those with unilateral CIs. It is difficult to know what advances might be associated with that gap in time, but one possibility is that the children with bilateral CIs had more experienced clinicians. To fully understand statistical patterns associated with word learning in children who have CIs, a longitudinal sample that extends over a longer period of time postimplantation would be useful as would a training study wherein causal effects could be determined via manipulation of phonotactic probability, neighborhood density, and word frequency. Ideally both would be designed prospectively so that confounds could be eliminated.

Clinical Implications

Sensitivity to statistical information in spoken language facilitates acquisition of sounds, words, morphemes, and sentence structures (Maye, Werker, & Gerken, 2002; Misyak & Christiansen, 2012; Saffran, Aslin, & Newport, 1996). In a series of studies, Conway and colleagues (Conway, Pisoni, et al., 2011; Conway, Karpicke, et al., 2011) indicate that children with CIs show poorer statistical learning skills than age-matched children with normal hearing and that statistical learning skills predict language levels in children with CIs. However, most of the children with CIs in these studies received unilateral cochlear implantation. For instance, 23 (92%) out of 25 children in Conway, Pisoni, et al. (2011) used unilateral CIs. Although we did not directly test children's statistical learning skills, the findings of the present study suggest that bilateral CIs may facilitate the development of sensitivity to statistical information in the ambient input as early as 12 months postimplantation. If future studies further support the current findings, this would mean that bilateral cochlear implantation not only benefits binaural processing but also (implicit) statistical learning.

Conclusion

Despite late and atypical access to the spoken signal, children with bilateral CIs are sensitive to statistical characteristics of words in the ambient spoken language. Like children with normal hearing, their early lexicons include many words that have many similar sounding neighbors but that consist of rare phonotactic sequences. Lexical growth is characterized by the addition of words with fewer neighbors and more common phonotactic sequences. This finding adds to a small body of literature on the benefits of binaural implantation. Replication and extension of this finding is of utmost importance.

Acknowledgments

This research is supported in part by National Institute on Deafness and Other Communication Disorders Grant 5 P50 DC00242 (awarded to Bruce Gantz), and by the New Century Scholar Research Grant from the American Speech-Language-Hearing Foundation (awarded to Ling-Yu Guo). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Deafness and Other Communication Disorders or the American Speech-Language-Hearing Foundation. We are grateful to the children who participated as well as to their parents who filled in the parent reports. We also thank J. Bruce Tomblin for granting the access to the archival data, Camille Dunn for retrieving the archival data from the database, Kayla Kuehlewind and Nicole Triscuit for processing the data, and Holly Storkel for providing advice on data coding.

Funding Statement

This research is supported in part by National Institute on Deafness and Other Communication Disorders Grant 5 P50 DC00242 (awarded to Bruce Gantz), and by the New Century Scholar Research Grant from the American Speech-Language-Hearing Foundation (awarded to Ling-Yu Guo). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Deafness and Other Communication Disorders or the American Speech-Language-Hearing Foundation.

References

  1. Amano S. (1996). Some observations on neighborhood statistics of spoken English words. Research on Spoken Language Processing Progress Report, 21, 439–453. [Google Scholar]
  2. Boons T., Brokx J. L., Frijns J. H., Peeraer L., Phillips B., Vermeulen A., … van Wieringen A. (2012). Effect of pediatric bilateral cochlear implantation on language development. Archives of Pediatrics and Adolescent Medicine, 166(1), 28–34. doi:10.1001/archpediatrics.2011.748 [DOI] [PubMed] [Google Scholar]
  3. Connor C. M., Craig H. K., Raudenbush S. W., Heavner K., & Zwolan T. A. (2006). The age at which young deaf children receive cochlear implants and their vocabulary and speech-production growth: Is there an added value for early implantation? Ear and Hearing, 27(6), 628–644. doi:10.1097/01.aud.0000240640.59205.42 [DOI] [PubMed] [Google Scholar]
  4. Connor C. M., Hieber S., Arts H. A., & Zwolan T. A. (2000). Speech, vocabulary, and the education of children using cochlear implants: Oral or total communication? Journal of Speech, Language, and Hearing Research, 43, 1185–1204. [DOI] [PubMed] [Google Scholar]
  5. Conway C. M., Karpicke J., Anaya E. M., Henning S. C., Kronenberger W. G., & Pisoni D. B. (2011). Nonverbal cognition in deaf children following cochlear implantation: Motor sequencing disturbances mediate language delays. Developmental Neuropsychology, 36(2), 237–254. doi:10.1080/87565641.2010.549869 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Conway C. M., Pisoni D. B., Anaya E. M., Karpicke J., & Henning S. C. (2011). Implicit sequence learning in deaf children with cochlear implants. Developmental Science, 14(1), 69–82. doi:10.1111/j.1467-7687.2010.00960.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Conway C. M., Pisoni D. B., & Kronenberger W. G. (2009). The importance of sound for cognitive sequencing abilities: The auditory scaffolding hypothesis. Current Directions in Psychological Science, 18(5), 275–279. doi:10.1111/j.1467-8721.2009.01651.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. De Cara B., & Goswami U. (2002). Similarity relations among spoken words: The special status of rimes in English. Behavior Research Methods, Instruments, & Computers, 34(3), 416–423. doi:10.3758/bf03195470 [DOI] [PubMed] [Google Scholar]
  9. Dunn C. C., Walker E. A., Oleson J., Kenworthy M., Van Voorst T., Tomblin J. B., … Gantz B. J. (2014). Longitudinal speech perception and language performance in pediatric cochlear implant users: The effect of age at implantation. Ear and Hearing, 35(2), 148–160. doi:10.1097/AUD.0b013e3182a4a8f0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dunn L. M., & Dunn D. M. (1997). Peabody Picture Vocabulary Test–III. Circle Pines, MN: AGS. [Google Scholar]
  11. Dunn L. M., & Dunn D. M. (2007). Peabody Picture Vocabulary Test–Fourth Edition. San Antonio, TX: Pearson. [Google Scholar]
  12. Fagan M. K., Pisoni D. B., Horn D. L., & Dillon C. M. (2007). Neuropsychological correlates of vocabulary, reading, and working memory in deaf children with cochlear implants. Journal of Deaf Studies and Deaf Education, 12(4), 461–471. doi:10.1093/deafed/enm023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fenson L., Dale P., Reznick S., Thal D., Bates E., & Hartung J. (1993). The MacArthur Communicative Development Inventories. San Diego, CA: Singular. [Google Scholar]
  14. Fenson L., Marchman V., Thal D., Dale P., Reznick S., & Bates E. (2007). The MacArthur-Bates Communicative Development Inventories: User's guide and technical manual (2nd ed.). Baltimore, MD: Brookes. [Google Scholar]
  15. Geers A. E., Moog J. S., Biedenstein J., Brenner C., & Hayes H. (2009). Spoken language scores of children using cochlear implants compared to hearing age-mates at school entry. Journal of Deaf Studies and Deaf Education, 14(3), 371–385. doi:10.1093/deafed/enn046 [DOI] [PubMed] [Google Scholar]
  16. Geers A. E., & Nicholas J. G. (2013). Enduring advantages of early cochlear implantation for spoken language development. Journal of Speech, Language, and Hearing Research, 56, 643–655. doi:10.1044/1092-4388(2012/11-0347) [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Geers A. E., Nicholas J. G., & Sedey A. L. (2003). Language skills of children with early cochlear implantation. Ear and Hearing, 24(1), 46S–58S. doi:10.1097/01.AUD.0000051689.57380.1B [DOI] [PubMed] [Google Scholar]
  18. Goodman J., Dale P., & Li P. (2008). Does frequency count? Parental input and the acquisition of vocabulary. Journal of Child Language, 35(3), 515–531. doi:10.1017/S0305000907008641 [DOI] [PubMed] [Google Scholar]
  19. Gordon K. R., & McGregor K. K. (2014). A spatially supported forced-choice recognition test reveals children's long-term memory for newly learned word forms. Frontiers in Psychology, 5, 1–12. doi:10.3389/fpsyg.2014.00164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Grieco-Calub T. M., & Litovsky R. Y. (2012). Spatial acuity in 2-to-3-year-old children with normal acoustic hearing, unilateral cochlear implants, and bilateral cochlear implants. Ear and Hearing, 33, 561–572. doi:10.1097/AUD.0b013e31824c7801 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Grieco-Calub T. M., Saffran J. R., & Litovsky R. Y. (2009). Spoken word recognition in toddlers who use cochlear implants. Journal of Speech, Language, and Hearing Research, 52, 1390–1400. doi:10.1044/1092-4388(2009/08-0154) [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Hayes H., Geers A. E., Treiman R., & Moog J. S. (2009). Receptive vocabulary development in deaf children with cochlear implants: Achievement in an intensive auditory-oral educational setting. Ear and Hearing, 30(1), 128–135. doi:10.1097/AUD.0b013e3181926524 [DOI] [PubMed] [Google Scholar]
  23. Hoover J. R., Storkel H. L., & Hogan T. P. (2010). A cross-sectional comparison of the effects of phonotactic probability and neighborhood density on word learning by preschool children. Journal of Memory and Language, 63(1), 100–116. doi:10.1016/j.jml.2010.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Houston D. M., Pisoni D. B., Kirk K. I., Ying E. A., & Miyamoto R. T. (2003). Speech perception skills of deaf infants following cochlear implantation: A first report. International Journal of Pediatric Otorhinolaryngology, 67(5), 479–495. doi:10.1016/S0165-5876(03)00005-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Houston D. M., Stewart J., Moberly A., Hollich G., & Miyamoto R. T. (2012). Word learning in deaf children with cochlear implants: Effects of early auditory experience. Developmental Science, 15(3), 448–461. doi:10.1111/j.1467-7687.2012.01140.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Hughes K. C., & Galvin K. L. (2013). Measuring listening effort expended by adolescents and young adults with unilateral or bilateral cochlear implants or normal hearing. Cochlear Implants International, 14(3), 121–129. doi:10.1179/1754762812Y.0000000009 [DOI] [PubMed] [Google Scholar]
  27. Ireton H., & Thwing E. (1974). Manual for the Minnesota Child Development Inventory. Minneapolis, MN: Behavior Science Systems. [Google Scholar]
  28. Johnson C., & Goswami U. (2010). Phonological awareness, vocabulary, and reading in deaf children with cochlear implants. Journal of Speech, Language, and Hearing Research, 53, 237–261. doi:10.1044/1092-4388(2009/08-0139) [DOI] [PubMed] [Google Scholar]
  29. Johnston J. C., Durieux-Smith A., Angus D., O'Connor A., & Fitzpatrick E. (2009). Bilateral paediatric cochlear implants: A critical review. International Journal of Audiology, 48(9), 601–617. doi:10.1080/14992020802665967 [DOI] [PubMed] [Google Scholar]
  30. Just M. A., & Carpenter P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99(1), 122–149. doi:10.1037/0033-295x.99.1.122 [DOI] [PubMed] [Google Scholar]
  31. Kirk K. I., Miyamoto R. T., Ying E. A., Perdew A. E., & Zuganelis H. (2000). Cochlear implantation in young children: Effects of age at implantation and communication mode. Volta Review, 102(4), 127–144. [Google Scholar]
  32. Klee T., & Harrison C. (2001, July). CDI words and sentences: Validity and preliminary norms for British English. Paper presented at Child Language Seminar, University of Hertfordshire, England. [Google Scholar]
  33. Kolson C. (1960). The vocabulary of kindergarten children (Unpublished doctoral dissertation). University of Pittsburgh, PA. [Google Scholar]
  34. Kover S. T., & Weismer S. E. (2014). Lexical characteristics of expressive vocabulary in toddlers with autism spectrum disorder. Journal of Speech, Language, and Hearing Research, 57, 1428–1441. doi:10.1044/2014_jslhr-l-13-0006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Kučera H., & Francis W. (1967). Computational analysis of present-day American English. Providence, RI: Brown University. [Google Scholar]
  36. Lammers M. J. W., Venekamp R. P., Grolman W., & van der Heijden G. J. M. G. (2014). Bilateral cochlear implantation in children and the impact of the inter-implant interval. The Laryngoscope, 124(4), 993–999. doi:10.1002/lary.24395 [DOI] [PubMed] [Google Scholar]
  37. Li P., Burgess C., & Lund K. (2000). The acquisition of word meaning through global lexical co-occurrences. In Clark E. V. (Ed.), Proceedings of the Thirtieth Annual Child Language Research Forum (pp. 166–178). Stanford, CA: Center for the Study of Language and Information. [Google Scholar]
  38. Lovett R. E. S., Kitterick P. T., Hewitt C. E., & Summerfield A. Q. (2010). Bilateral or unilateral cochlear implantation for deaf children: An observational study. Archives of Disease in Childhood, 95(2), 107–112. doi:10.1136/adc.2009.160325 [DOI] [PubMed] [Google Scholar]
  39. Luce P. A., & Pisoni D. B. (1998). Recognizing spoken words: The neighborhood activation model. Ear and Hearing, 19(1), 1–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. MacWhinney B. (2000). The CHILDES project: Tools for analyzing talk (3rd ed.). Mahwah, NJ: Erlbaum. [Google Scholar]
  41. Maye J., Werker J. F., & Gerken L. (2002). Infant sensitivity to distributional information can affect phonetic discrimination. Cognition, 82(3), B101–B111. doi:10.1016/S0010-0277(01)00157-3 [DOI] [PubMed] [Google Scholar]
  42. Misyak J. B., & Christiansen M. H. (2012). Statistical learning and language: An individual differences study. Language Learning, 62(1), 302–331. doi:10.1111/j.1467-9922.2010.00626.x [Google Scholar]
  43. Moe A., Hopkins C., & Rush R. T. (1982). The vocabulary of first-grade children. Springfield, IL: Charles C Thomas. [Google Scholar]
  44. Newman R. S., Sawusch J. R., & Luce P. A. (1997). Lexical neighborhood effects in phonetic processing. Journal of Experimental Psychology: Human Perception and Performance, 23(3), 873–889. doi:10.1037/0096-1523.23.3.873 [DOI] [PubMed] [Google Scholar]
  45. Niparko J. K., Tobey E. A., Thal D. J., Eisenberg L. S., Wang N.-Y., Quittner A. L., & Fink N. E. (2010). Spoken language development in children following cochlear implantation. The Journal of the American Medical Association, 303(15), 1498–1506. doi:10.1001/jama.2010.451 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Nittrouer S., & Chapman C. (2009). The effects of bilateral electric and bimodal electric—Acoustic stimulation on language development. Trends in Amplification, 13(3), 190–205. doi:10.1177/1084713809346160 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Nusbaum H., Pisoni D. B., & Davis C. (1984). Sizing up the Hoosier Mental Lexicon: Measuring the familiarity of 20,000 words (Research on Spoken Perception Progress Report no. 10). Bloomington: Indiana University, Speech Research Laboratory.
  48. Papsin B. C., & Gordon K. A. (2008). Bilateral cochlear implants should be the standard for children with bilateral sensorineural deafness. Current Opinion in Otolaryngology & Head and Neck Surgery, 16(1), 69–74. doi:10.1097/MOO.1090b1013e3282f1095e1097c [DOI] [PubMed] [Google Scholar]
  49. Reynell J., & Huntley M. (1985). Reynell Developmental Language Scales (2nd ed.). Windsor, England: Nfer Nelson. [Google Scholar]
  50. Riches N. G., Tomasello M., & Conti-Ramsden G. (2005). Verb learning in children with SLI: Frequency and spacing effects. Journal of Speech, Language, and Hearing Research, 48(6), 1397–1411. doi:10.1044/1092-4388(2005/097) [DOI] [PubMed] [Google Scholar]
  51. Saffran J. R., Aslin R. N., & Newport E. L. (1996, December 13). Statistical learning by 8-month-old infants. Science, 274, 1926–1928. [DOI] [PubMed] [Google Scholar]
  52. Sarant J., Harris D., Bennet L., & Bant S. (2014). Bilateral versus unilateral cochlear implants in children: A study of spoken language outcomes. Ear and Hearing, 35(4), 396–409. doi:10.1097/aud.0000000000000022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Scarborough R. (2012). Lexical similarity and speech production: Neighborhoods for nonwords. Lingua, 122(2), 164–176. doi:10.1016/j.lingua.2011.06.006 [Google Scholar]
  54. Semel E., Wiig E., & Secord W. (2003). Clinical Evaluation of Language Fundamentals–Fourth Edition (CELF-4). San Antonio, TX: Pearson. [Google Scholar]
  55. Stallings L. M., Gao S., & Svirsky M. A. (2000). Assessing the language abilities of pediatric cochlear implant users across a broad range of ages and performance abilities. Volta Review, 102(4), 215–235. [Google Scholar]
  56. Stokes S. F. (2010). Neighborhood density and word frequency predict vocabulary size in toddlers. Journal of Speech, Language, and Hearing Research, 53, 670–683. doi:10.1044/1092-4388(2009/08-0254) [DOI] [PubMed] [Google Scholar]
  57. Stokes S. F., Bleses D., Basbøll H., & Lambertsen C. (2012). Statistical learning in emerging lexicons: The case of Danish. Journal of Speech, Language, and Hearing Research, 55, 1265–1273. doi:10.1044/1092-4388(2012/10-0291) [DOI] [PubMed] [Google Scholar]
  58. Stokes S. F., Kern S., & dos Santos C. (2012). Extended statistical learning as an account for slow vocabulary growth. Journal of Child Language, 39(1), 105–129. doi:10.1017/S0305000911000031 [DOI] [PubMed] [Google Scholar]
  59. Storkel H. L. (2004). Do children acquire dense neighborhoods? An investigation of similarity neighborhoods in lexical acquisition. Applied Psycholinguistics, 25(2), 201–221. doi:10.1017/S0142716404001109 [Google Scholar]
  60. Storkel H. L. (2009). Developmental differences in the effects of phonological, lexical and semantic variables on word learning by infants. Journal of Child Language, 36(2), 291–321. doi:10.1017/s030500090800891x [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Storkel H. L., Armbrüster J., & Hogan T. P. (2006). Differentiating phonotactic probability and neighborhood density in adult word learning. Journal of Speech, Language, and Hearing Research, 49, 1175–1192. doi:10.1044/1092-4388(2006/085) [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Storkel H. L., & Hoover J. (2010). An online calculator to compute phonotactic probability and neighborhood density on the basis of child corpora of spoken American English. Behavior Research Methods, 42(2), 497–506. doi:10.3758/brm.42.2.497 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Storkel H. L., & Lee S.-Y. (2011). The independent effects of phonotactic probability and neighborhood density on lexical acquisition by preschool children. Language and Cognitive Processes, 26(2), 191–211. doi:10.1080/01690961003787609 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Storkel H. L., Maekawa J., & Hoover J. R. (2010). Differentiating the effects of phonotactic probability and neighborhood density on vocabulary comprehension and production: A comparison of preschool children with versus without phonological delays. Journal of Speech, Language, and Hearing Research, 53, 933–949. doi:10.1044/1092-4388(2009/09-0075) [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Thal D., DesJardin J., & Eisenberg L. (2007). Validity of the MacArthure–Bates Communicative Development Inventories in children with cochlear implants. American Journal of Speech-Language Pathology, 16, 54–64. doi:10.1044/1058-0360(2007/007) [DOI] [PubMed] [Google Scholar]
  66. Vitevitch M. S., Armbrüster J., & Chu S. (2004). Sublexical and lexical representations in speech production: Effects of phonotactic probability and onset density. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30(2), 514–529. doi:10.1037/0278-7393.30.2.514 [DOI] [PubMed] [Google Scholar]
  67. Walker E. A., & McGregor K. K. (2013). Word learning processes in children with cochlear implants. Journal of Speech, Language, and Hearing Research, 56, 375–387. doi:10.1044/1092-4388(2012/11-0343) [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Wilson B. (2006). Speech processing strategies. In Cooper H., & Craddock L. (Eds.), Cochlear implants: A practical guide (2nd ed., pp. 21–69). Philadelphia, PA: Wiley. [Google Scholar]
  69. Zimmerman I., Steiner V., & Pond R. (2002). Preschool Language Scale–Fourth Edition. San Antonio, TX: Pearson. [Google Scholar]

Articles from Journal of Speech, Language, and Hearing Research : JSLHR are provided here courtesy of American Speech-Language-Hearing Association

RESOURCES