Corina et al state, “Sound scientific research evaluating the effects of an intervention is predicated on random selection or assignment to treatment and control groups. . . . The Geers et al study violates this basic tenet of research design.”

Although we agree that randomly assigning families at the time of diagnosis to intervention programs that differ in their use of sign language would be an efficient method to address sign language benefits, this approach is impractical and unethical for a variety of reasons. A decision regarding a child’s communication mode arises from parental choice based on information available at the time the child becomes a candidate for cochlear implantation. In the current observational cohort study, we examined more prospectively collected preimplant demographic characteristics than researchers in any previous study on this topic in the literature, introducing unprecedented transparency into the interpretation of group comparability at baseline.

Hall et al state, “. . .although the authors quantified exposure to manual communication (including ASL), they did not measure proficiency in ASL. Children who have not acquired the grammar of ASL are not predicted to benefit from it as a foundation for subsequent mastery of English.” Caselli et al state, “Effects of ‘sign language exposure’ may have been carried by participants who used an artificial signing system, received late exposure relative to the critical period of language acquisition, had only 1 ASL model, and families with limited to no ASL proficiency.”

From study onset, parental and clinician goals for participants in the Childhood Development after Cochlear Implantation (CDaCI) study have focused on spoken language development, and that outcome drove the research questions in this paper. We sought to examine long-term consequences of early communication mode decisions made for deaf children receiving CIs by parents who have typical hearing.

In our study, we include data from a representative sample of early-implanted children reared in real-world situations across the United States, in which the majority of hearing parents typically lack proficiency in ASL and, therefore, cannot provide a language-rich environment in both ASL and spoken English. Differences in syntax, phonology, and spatial information between ASL and spoken language1,2  limit simultaneous use of both languages. Therefore, conscientious and proficient use of ASL would detract from the amount of time spent stimulating and reinforcing spoken language development and could influence the structure of spoken language. Therefore, most hearing parents choose to employ an “artificial” sign language to accompany spoken English.

We do not aim to examine the effects of proficient ASL exposure on outcomes in pediatric CI users. Rather, we investigated whether more focused spoken language exposure, free from distraction of manual signs, would provide a more attainable, broad-based strategy for verbal language development.

Hall et al state, “. . .the best interpretation of this study is that families self-select their method of communication as a result of their child’s development in English. Under this view, the use of manual communication is a consequence of limited progress in English, not a cause. . . . Parents of deaf children who are not progressing with their cochlear implant (CI) may be more likely to begin (or continue) signing with their child. This would imply that poor oral outcomes encourage the use of signing rather than the use of signing limiting oral outcomes.”

Baseline measures (Table 1) and communication mode decisions were made in infancy (ie, before the child was 38 months old). Children in the 3 groups exhibited similar spoken language at that time (actually, children in the long-term sign exposure group had the highest average spoken vocabulary size preimplant), making “self-selection” to use sign on the basis of limited progress in spoken English questionable. Average baseline auditory perception ratings do not differ significantly among groups. Preimplant-aided pure-tone average thresholds differed by less than 5 dB across groups. Average parent ratings of auditory-based behaviors on the Infant-Toddler Meaningful Auditory Integration Scale (IT-MAIS) represent limited auditory skills with hearing aids before implantation in all groups (mean scores of 9.8 and 5.8 in the no sign group and the long-term sign groups, respectively) out of a maximum score of 40.

Corina et al state, “These data indicate the sign exposure groups were experiencing profound difficulty with auditory perception and speech identification. It will surprise no one that children with better auditory perception skills and speech identification skills at the outset will be the children who end up with the best auditory language scores.”

The 3 participant groups did not statistically significantly differ on auditory perception or spoken language skills at study onset. Furthermore, baseline auditory perception scores were not significantly correlated with any outcome measures (r values ranged from –0.02 to 0.19). Table 2 shows emerging differences over time in speech perception abilities, measured via an index based on a hierarchy of tests. Two possible interpretations can explain the widening of the gap between performance of the no sign and sign-exposed groups over time: (1) parents of children who do not receive adequate auditory benefit from their CI continue or begin manual communication efforts to interact more efficiently with their child, or (2) children reared in families using sign develop auditory perception skills more slowly than those reared in nonsigning families. Our results appear to support the second explanation because all 3 groups of children exhibited improved speech perception scores over time. Signing parents typically initiated sign language use before their child received a CI, and none of the no sign families began to sign later in response to poor auditory or spoken English progress. Parents in the long-term sign language group reportedly decreased their use of sign over time (from 63% of the day at baseline to 29% at 36 months postimplant). The 27 families in the short-term sign group stopped using sign during the first 2 years of CI use, presumably because it was no longer necessary for communication, although their outcomes did not reach the high levels achieved by the no sign group. Nevertheless, we cannot state unequivocally whether parents’ use of sign was in response to their child’s auditory development or whether children’s auditory speech perception was negatively affected by parental sign use.

Hall et al state, “. . .observed differences might be attributable to other demographic factors that likely affected initial inclusion, attrition over time, and/or performance on the assessments (eg, socioeconomic status, etiology of deafness). Because sign language use may covary with these additional factors, the reported effects might well disappear if these factors were controlled. . . . To satisfactorily demonstrate that sign language exposure harms spoken language development, the authors must demonstrate the following: (1) all baseline measures were equivalent, (2) groups were not self-selected, and (3) participant attrition was not systematic.”

It is important to recognize the large variability in all of these characteristics, which explains why differences in average scores cannot be interpreted without recognizing the substantial overlap in distributions. Although statistical analysis failed to indicate significant differences in baseline metrics, some averages favor 1 group and some averages favor another group (eg, higher maternal education and family income appear to favor the no sign exposure group, whereas the long-term sign exposure group exhibits later onset of profound deafness, higher average IQ, and greater exposure to parent-infant intervention before cochlear implantation). Correcting for all of these variables in the analysis with multiple outcomes and time points would have added an unnecessary level of complexity to the article, so presumed equivalence was based on failure to find statistically significant group differences. Furthermore, some preimplant characteristics (eg, auditory speech perception) were measured with rating scales that are, at best, ordinal in nature, limiting their usefulness in correction analyses.

However, as a result of receiving several different comments related to potential for baseline group differences, although not statistically significant, to be driving the statistically significant group differences in later outcomes, we have run regression models adjusting for implant activation age, IT-MAIS score, enrollment in a parent-infant program, maternal education, and household income at baseline. It should be noted that because of substantial correlations among some variables, reported predictors are subject to suppression effects. Results of these regression models are presented here in Table A. Even after adjusting for the mentioned characteristics, the groups exposed to sign language performed significantly poorer than participants not exposed to sign language on spoken language (Comprehensive Assessment of Spoken Language) assessed in the late elementary years and on speech intelligibility.

Table A

Adjusted Linear Regression Results Comparing the Sign Exposure Groups on Language, Reading, and Speech Intelligibility Outcomes

CharacteristicEarly Elementary Spoken Language βEarly Elementary Reading Comprehension βLate Elementary Spoken Language βLate Elementary Reading Comprehension βSpeech Intelligibility β, %
Sign language exposure group (1 no sign, 2 short-term sign, 3 long-term sign) −2.7 .9 −6.9* −2.5 −9.6** 
Activation age (1 y increase) −12.3*** −10.4** −12.0** −7.8** −3.6 
Auditory perception (1-point increase on IT-MAIS) .5 .2 .6 .5* .1 
Parent-infant program (yes versus no) .8 .3 7.5 5.0 4.1 
Maternal education (college graduate versus not) 4.9 4.3 7.7 5.3 6.0 
Household income (low versus not) −10.0* −3.4 −5.1 −2.4 3.9 
CharacteristicEarly Elementary Spoken Language βEarly Elementary Reading Comprehension βLate Elementary Spoken Language βLate Elementary Reading Comprehension βSpeech Intelligibility β, %
Sign language exposure group (1 no sign, 2 short-term sign, 3 long-term sign) −2.7 .9 −6.9* −2.5 −9.6** 
Activation age (1 y increase) −12.3*** −10.4** −12.0** −7.8** −3.6 
Auditory perception (1-point increase on IT-MAIS) .5 .2 .6 .5* .1 
Parent-infant program (yes versus no) .8 .3 7.5 5.0 4.1 
Maternal education (college graduate versus not) 4.9 4.3 7.7 5.3 6.0 
Household income (low versus not) −10.0* −3.4 −5.1 −2.4 3.9 

*P < .05; ** P < .01; *** P < .001.

Martin et al state, “The 40 children who met eligibility criteria but were excluded because of a lack of follow-up data may have influenced the outcomes. Families experiencing poor progress with their child’s CI may stop their follow-up appointments, for instance.”

As was stated in the article, 40 children who met the implant age criteria were excluded because of a lack of sufficient follow-up data, many of whom moved away from the implant center and were unable to return for follow-up visits. Similar rates of sign language use at baseline were reported for these 40 children (55% sign, 45% no sign) as for the sample with complete data (60% sign, 40% no sign). There was no significant tendency for participant attrition to be associated with either signing or nonsigning families preimplant.

Caselli et al state, “Although this study was designed for the authors to look narrowly at English-based outcomes, the authors overinterpret the results as evidence against the assertion that a natural sign language can be beneficial for deaf children. Although English proficiency is certainly 1 route to success, it is not a necessary condition for it.”

Our study should not be considered an indictment on the use of ASL, but rather it adds new information to the literature on this age-old controversy. One of the most consistent findings of CDaCI has been the large variability in outcomes that has persisted throughout the 15 years of data collection. Researchers from the many reports from this study have examined this variability without focusing exclusively on the “star performers.” In fact, in 1 recent study, Barnard et al reported specifically on those children who had not progressed in their auditory skill development (ie, did not achieve open-set speech recognition) after 5 years of experience with the CI.3  Without adequate access to speech, visual language becomes essential, whether it be speechreading, cued speech, simultaneous communication, or ASL. This was the case for the children in the Barnard et al study. As has been mentioned earlier in this response, the overarching aim of CDaCI has been to study the development of spoken language after cochlear implantation, which injects a certain bias into the research design. Each family needs to make their own informed decision about what works best for their unique situation. One of the most important variables in any parent-child relationship, regardless of auditory status, is the presence of effective and efficient communication. Parents need to keep open lines of communication between parent and child and to provide a language-rich environment to enhance that child’s listening and language development.

For more comments on the article “Early Sign Language Exposure and Cochlear Implantation Benefits,” go to http://pediatrics.aappublications.org/content/140/1/e20163489.comments.

1
Emmorey
K
.
Language, Cognition, and the Brain: Insights From Sign Language Research
.
Mahway, NJ
:
Lawrence Erlbaum Associates
;
2002
2
Emmorey
K
,
Borinstein
HB
,
Thompson
R
.
Bimodal bilingualism: code-blending between spoken English and American Sign Language
. In:
Cohen
J
,
McAlister
KT
,
Rolstad
K
, and
MacSwan
J
, eds.
ISB4: Proceedings of the 4th International Symposium on Bilingualism
.
Somerville, MA
:
Cascadilla Press
;
2005
:
663
673
3
Barnard
JM
,
Fisher
LM
,
Johnson
KC
, et al
;
CDaCI Investigative Team
.
A prospective, longitudinal study of US children unable to achieve open-set speech recognition 5 years after cochlear implantation
.
Otol Neurotol
.
2015
;
36
(
6
):
985
992

Competing Interests

CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.