Symposium
Chair: Jennifer Ganger
Discussant: Karin Stromswold
This symposium brings together work in an area that until recently wasalmost completely unexplored: the genetics of language in infancy. In eachof these papers, the traditional twin method is employed to examine bothgenetic and environmental contributions to language development. Besidesbeing among the first attempts to apply behavior genetics to language atsuch a young age, the papers in this symposium are unique in that theyincorporate research on three complementary topics, all of which areessential to the complete understanding of language development. These are:normal vocabulary and syntactic development, abnormal development, and therelationship between language and cognition.The four papers in this symposium are in stunning agreement on normalvocabulary development. From four laboratories, four populations, threecountries, and two languages, all the researchers conclude that normalvocabulary development in infancy (measured either by parent checklist,parent journal, or experimenter-administered test of verbal ability) is moreinfluenced by common environment than by heredity, although heredity doesplay a non-negligible role (accounting for between 10 and 33% of thevariance).However, the papers herein go beyond this basic finding in three ways.First, the genetics of language delay are considered in two papers, usingtwo different populations (the United Kingdom and French-speaking Canada).While normal vocabulary development has only moderate heritability, delayeddevelopment-that is, development in children confined to the lowest 5% ofthe population in terms of vocabulary size-showed a much higher heritability(48-79%). Furthermore, persistent language delay (or lack thereof) betweenthe ages of 2 and 3 is also characterized by large genetic influence.The third paper goes beyond vocabulary size and general verbal knowledge tolook specifically at the onset of combinatorial syntactic ability byexamining the first use of productive multiple-word combinations. The onsetof these combinations was found to have higher heritability than the onsetof single words, suggesting a qualitative difference between learningisolated words and putting them together into sentences.Finally, in the fourth paper, the relationship between language andcognition is considered. At 18 months, some genetic variance is found to beunique to language, but there is substantial genetic variance shared byverbal and non-verbal abilities as well. At 24 months, however, a timewhen linguistic ability is blossoming, a new genetic factor unique tolanguage emerges.
Details of individual items:
paper
Language development is characterized not only by substantial variability at each age, but by variability in developmental trajectory. Both the sources and the consequences of this variation are not well understood. In the first part of this paper, we examine vocabulary development in a sample of 2,410 twin pairs from the Twins Early Development Study (TEDS), with particularfocus on the later development of children who are delayed ( lowest 5%) at 2. A 100 item vocabulary checklist and a grammar scale, both based on the MacArthur Communicative Development Inventory, were used at each age.Genetic analyses of vocabulary at 3 largely replicated the previously reported findings at 2: modest individual differences heritability (25%),but very substantial group heritability for the lowest scorers (.79). Also,as before, a comparison of analyses with different criteria confirms thatthe 5% criterion captures a significant qualitative distinction. Although the overall phenotypic correlation in vocabulary from 2 to 3 is .61, this leaves substantial variation in growth pattern. Only 45% of the 2 year probands are still in the lowest 5% at 3, and the majority of the others are not even in the near-clinical range (6th-15th percentile). Neither parentaleducational attainment nor a measure of nonverbal ability (the PARCA) contribute substantially to the prediction from 2 to 3. In contrast, longitudinal DF analyses, an application of bivariate DF analysis, provide evidence for a strong genetic component to the stability, or lack thereof,in scores from 2 to 3. These results, taken with the outcome of concordance analyses for persistent (delayed at 2 and at 3) and transient (delayed at 2 but not at 3) delay, suggest that early language delay as indexed by vocabulary scores in the lowest 5% at age 2 includes genetically distinct categories for persistent and transient delay. Language development is also characterized by variation in profile acrosscomponents in many cases. The second focus of this paper is the relation between vocabulary and grammatical development, which is at the core of many current controversies in linguistics and psycholinguistics. Behavioral genetic techniques can illuminate the question of whether the strongempirical correlation that has been observed in early development simply reflects common effects of facilitative environments, or whether commonunderlying mechanisms, influenced by the same genes, are responsible for the correlation. We explored this issue in a sample of 2,898 2-year old twin pairs. Moderate heritabilities were found for both vocabulary (25%) and grammar (39%). As in previous studies, the two aspects of language development were substantially correlated (r.66). Behavioral genetic modeling of the relation between vocabulary and grammar produced an estimated value of .61 for the genetic correlation rG, a measure of the overlap of the genetic effects that contribute to the two aspects of language development. Thus there is little evidence for a dissociation of lexical and grammatical development at this stage.
paper
There is growing interest in the etiology of early language acquisition,both for individual differences in language development and for languagedelays. It remains unclear whether the factors influencing languagedevelopment are the same as those associated with delays in the lowest 5% ofthe population. Recent research using twin methodology has looked at geneticand environmental influences on rate of early language acquisition (Dale etal., in press; Ganger et al., 1998; Reznick, Corley and Robinson, 1997) andon expressive vocabulary delays in young children (Dale et al, 1998).Results from these studies show that genetic influences explain the majorityof the variance for expressive vocabulary delay while they remain modest forvocabulary development in general. These striking results for the delayedchildren have never been replicated. Furthermore, there is no evidence thatresults from the individual differences studies replicate for non-Englishspeaking samples. This study uses twin methodology with French-speaking 18 month oldtwins toassess genetic and environmental influences on individual differences andfor delayed children in expressive vocabulary and receptive language. Thesample for this study is a subset of 448 twins recruited for the QuebecNew-Born Twin Study. At 18 months, twins were assessed separately via parentreports on the Spontaneous Speech and Receptive Language Record. Anexpressive vocabulary score was computed based on a 77 word scale(alpha3D0.97). An additional score for receptive language was computed addingup every correctly executed AB command BB out of 12 items (alpha3D0.72). Meanscore for the expressive vocabulary scale is 18.7 (15.5 SD) and mean scorefor the receptive language scale is 9.8 (2.2 SD) with no significant sexdifferences on either scale. The delayed sample is selected based on thebottom 5% of the distribution for each scale. Estimates of genetic and environmental influences are assessed usingstructural equation modeling of MZ and DZ covariance matrices. Parameterestimates for the language delay groups are assessed via DF regression(DeFries and Fulker, 1985, 1988).Results1. Estimates of genetic and environmental influences for expressivevocabulary and receptive language for the whole sample are similar : geneticcomponents account for 21% (7%-37 CI) of the variance in expressivevocabulary and 14% (0% -31% CI) in receptive language; shared environmentcomponents account for the majority of the variance with 65% (50%-72%) forexpressive vocabulary and 67% (50%-80%) for receptive language.2. Estimates for the delayed groups reveal a different picture. For thereceptive language scale, estimates for the delayed group are notsignificantly different from the parameters estimated for the whole sample.46or expressive vocabulary however, heritability for the delayed groupincreases up to 48% and the shared environment portion of the variancedecreases significantly. This suggests that genetic factors may play a moreimportant role in expressive vocabulary delay than they do in vocabularydevelopment in the general population. The differential etiology for delaydoes not however seem to apply to all aspects of language acquisition.Results will be discussed in light of published research on English-speaking children.
paper
Individual differences in language development have been a focus of childlanguage research for decades. Although it has traditionally been assumedthat the environment plays a primary role in determining such differences,researchers have recently begun to investigate the contribution of heredityto early language development as well (Reznick, Corley, & Robinson, 1997;Dale, Simonoff, Bishop, Eley, Oliver, Price, Purcell, Stevenson, & Plomin,1998; Dale, Dionne, Eley, & Plomin, 1999; Ganger, Pinker, Chawla, & Baker,1999). This study uses twins to examine the heritability of two aspects ofearly language development: vocabulary and syntax, comparing them in twoways.Expressive vocabulary development was studied in 57 identical and 35(same-sex) fraternal twin pairs by using parent journals which began at thefirst word whenever possible and continued to about 100 words. Twins werecompared on the age at which they reached two early milestones of vocabularydevelopment: 25th word and the 50th word. Heritability accounted for onlyabout 10% of the variance in age at reaching the 25th word, while commonenvironment accounted for close to 90%. Similarly, at 50 words, about 20%of the variance was due to heritability, and 70% due common environment.Early grammatical development was studied by assessing twins' ages at theirfirst productive word combination. 32 identical and 23 fraternal twin pairswere compared on this measure. In contrast to vocabulary development,behavior genetic analysis revealed that common environment had little or noinfluence on the acquisition of word combinations, while heritabilityaccounted for over 80% of the variance. The difference in heritabilitybetween the age at early vocabulary milestones and the age at first wordsuggests that while early word learning has only a small influence fromgenetic factors, early grammatical development has a larger one.As an additional test of genetic influence on lexical versus grammaticaldevelopment, the percentage of both nouns and verbs in the twins' first 50words was compared (other categories were discarded for this analysis).Since all verbs have inherent syntax in their argument structure while mostearly nouns do not, comparing these two proportions provides another way toexamine the lexical-grammatical distinction. The proportion of nouns inthese early vocabularies showed no heritability (less than 5%) and largecommon environment (approximately 80%), while the proportion of verbs showeda somewhat larger heritability (about 35%) and similar common environment(also about 35%). Although this difference is not as great as that foundbetween isolated words and word combinations, it is suggestive of the samekind of difference in lexical versus grammatical development.In sum, although some genetic influence was uncovered in both vocabulary andsyntax, the difference in the influence of heritability on grammaticaldevelopment seems to be greater than that on purely lexical development.This finding meshes well with what we know about normal language: words arearbitrary sound-meaning pairs that must be learned from the environment,while grammatical combinations must be generated by the child rather thanimitated.
paper
Multivariate behavioral genetic methods can be used to detect relations between language and cognition with respect to the sources of their genetic and environmental influences. J. S. Reznick, R. Corley, and J. Robinson (1997), for instance, found genetic influences on language at 20 months that were not related to nonverbal cognition. At 24 months, however, a single genetic factor influenced both language and nonverbal cognition. Relations across ages and abilities (e.g., relations between nonverbal cognition at 20 months and language at 24 months) were not examined. The present study modeled the genetic and environmental structure of influences on language and nonverbal cognition in a sample of twins at 18 and 24 months of age. Overlaps between the two abilities within an age and across ages were examined. The sample consisted of 220 twin pairs from the Louisville Twin Study. Each twin was assessed at 18 and 24 months of age using the Bayley Scales of Infant Development. Language scores (LS) and nonverbal scores (NVS) were drawn from the Bayley items. There were significant heritable components (indicated by h2) to 18-month NVS (h2 3D .54), 18-month LS (h2 3D .33), 24-month LS (h2 3D .16); but not 24-month NVS (h2 3D .12). Shared environment accounted for a significant amount of variability (indicated by c2) in 18-month LS (c2 3D .49), 24-month NVS (c2 3D .50), and 24-month LS (c2 3D .67); but not 18-month NVS (c2 3D .10). Multivariate analysis of the NVS and LS covariance structures was used todetermine the amount of overlap in influences on the two abilities. At 18 months there were genetic influences that were unique to LS and influences that were common to LS and NVS. At 24 months a new genetic factor that accounted for variance in only LS appeared. Across ages, the factor that influenced 18-month NVS also influenced 24-month NVS. Similarly the factor that influenced 18-month LS influenced 24-month LS. Interestingly, there was no evidence for cross-age cross-ability influences. That is, genetic influences on 18-month NVS did not influence 24-month LS, and influences on18-month LS did not influence 24-month NVS. At 18 months there were small influences of shared environment that were unique to each ability, but there was also significant overlap in the influences on NVS and LS. Factors influencing 18-month LS and NVS also influenced 24-month scores, but no new factors appeared at 24 months. Nonshared environment tended to be restricted to specific abilities and occurred only within an age. These findings indicate that as early as 18 months there is evidence for a differentiation of language and nonverbal ability with respect to genetic influences, but there is also genetic variance that is shared by the two abilities. At 24 months, coinciding with the emergence of more advanced skills, a new genetic factor that is restricted to language appears. With respect to shared environment, there are some influences that are restricted to a specific ability. However, the results suggest that there are also shared-environmental influences that operate across abilities and ages.