The Development of Language Skills Among Children Exposed to Crack-Cocaine In Utero

The present study examines the language development among cocaine-exposed and non-exposed 24-month old inner-city children. Natural language data was collected from two groups of 17 mother-child dyads during a 5-minute semi-structured play session. Both groups of dyads were drawn from a large-scale on-going longitudinal study that has followed children and mothers since the second trimester of pregnancy. Both Exposed and Control dyads were recruited from the same inner-city population and matched on socio-economic status. Although few group differences achieved statistical significance, Exposed dyads showed consistent lower mean scores on most measures of basic language complexity than did Control dyads. A morphological analysis also suggested that Exposed children produced less complex speech. Further research with larger samples and longer speech samples is needed. Malakoff, Mayes & Underhill; WPA 4/12/96 2 INTRODUCTION In the late 1980's and early 1990's, increasing media coverage was devoted to the "problem" of "crack babies, " as the media has termed the children born to women who used cocaine (usually crack-cocaine) in pregnancy. Coverage and concern focused, in particular, on the increased levels of behavioral and language problems with which schools were soon to be faced. However, an examination of the empirical research from this era reveals very little empirical evidence to back up these dire predictions of devastating developmental effects attributable directly to prenatal cocaine-exposure. In particular, there has been little systematic longitudinal study of children exposed to cocaine in utero. The first longitudinal findings of the effects of prenatal cocaine exposure on cognitive abilities at 24 months were published in 1992 by Chasnoff and his colleagues (Chasnoff, Griffith, Freier & Murray, 1992). Results from the 36-month follow-up was published in 1994 (Griffith, Azuma & Chasnoff, 1994). The exposed cohort did not differ from their unexposed peers on global measures of intellectual ability at either age. However there were some indications of group differences in cognitive ability. At 24 months, a significantly greater proportion of exposed children scored more than two standard deviations below the mean on the Mental Scale of the Bayley Developmental Index. At 36 months, cocaine was shown to have a significant impact on the verbal reasoning subscale. The small empirical literature on developmental outcome is mixed. In general, these findings are difficult to interpret and compare due to a number of methodological problems or differences: mixed-drug exposure, small sample sizes; wide age groupings. A recent longitudinal study of 35 drug-exposed children in Amsterdam (van Baar & de Graaf, 1994) found Malakoff, Mayes & Underhill; WPA 4/12/96 3 cognitive and language delays from 3.5 to 5.5 years. However, this study included children exposed to heroin and methadone, as well as cocaine. Rotholtz, Snyder & Peters (1995) and Hawley et. al. (1995) both found no effect of cocaine on language development. However, both studies included children from two to five years of age, had small sample sizes for this range, and relied on maternal report or medical histories. Despite the lack of conclusive evidence associating prenatal cocaine exposure with later language delays, there are nonetheless reasons for concern. Maternal cocaine-use has been associated with a wide range of negative environmental and caregiving factors that can impact on developmental outcome. These include extreme poverty, domestic violence, homelesseness, maternal depression, and dysfunctional parenting. Furthermore, recent research on language acquisition is increasingly pointing to the importance of the nature and quality of the caregiver linguistic input during infancy and early childhood. Thus, in addition to the possible neurological sequelae of prenatal cocaine exposure, environmental factors associated with cocaine indicate that prenatal cocaine-exposure may be a serious risk factor for developmental outcome. A study I carried out in collaboration with Linda Mayes and Richard Schottenfeld at Yale University (Malakoff, Mayes, & Schottenfeld, 1994) underscored the importance of considering the postnatal environment when examining the long term effects of prenatal cocaine exposure on outcome. We assessed language comprehension and production of 21 preschool children of cocaine-using mothers who were attending a daycare program attached to a drug-treatment program for pregnant women. Thus all of the preschoolers were currently living with mothers who were Malakoff, Mayes & Underhill; WPA 4/12/96 4 using cocaine. In addition, about half of our sample had also been exposed to cocaine in utero, based upon maternal report and, for about 80% of the sample, urine toxicology screens at birth. We thus had two groups of preschoolsers: those who had been exposed to maternal cocaine-use both prenatally and in the postnatal environment, and those who had only experienced the environment associated with maternal cocaine-use after bir th. The results from our study indicated that 1) there was a great degree of heterogeneity in language outcome within both groups of preschoolers 2) the vast majority of the children showed serious language delays, although there were rare but notable exceptions. 3) the two groups did not differ on any measures of language development, nor were there any group trends. THE PRESENT STUDY The present study is part of a larger longitudinal study of the effects of prenatal cocaine exposure on outcome. The goals of the present study were to assess and compare the language abilities of two cohorts of 24-month olds who had been followed since the second trimester of pregnanacy: a cohort with a history of prenatal cocaine exposure and a control cohort matched on socio-economic status. METHODS SUBJECTS Subjects were 34 mother-child dyads who were participating in a large multi-year longitudinal study of the developmental effects of prenatal cocaine exposure currently being conducted at the Yale Child Malakoff, Mayes & Underhill; WPA 4/12/96 5 Study Center. Seventeen Control dyads (7 boys, 10 girls) who had participated in the 24-month follow-up were matched on age with seventeen Exposed dyads (9 boys, 8 girls). Exposed and Control dyads were recruited from the same inner-city population when the mother was in the second trimester of pregnancy. Children in the Exposed group were exposed to cocaine in utero while those in the Control group were not exposed to cocaine at any point in their lives. Two children in the Exposed dyads were accompanied by their grandmothers; the remaining children were accompanied by their mothers. The mean age of the toddlers was 24 months, 5 days (SD = 15.32 days), with a range from 23 to 25 months (701 to 762 days). Ages were calculated using a corrected date of birth for four subjects born at 36 weeks or less. Three Exposed dyads were Caucasian; all other dyads were African-American. Mothers in both groups were young, with a mean age of 26. Only 65 percent of the Control mothers and 53 percent of the Exposed mothers had completed highschool. Although no mothers were chronic alcohol users, 53 percent of the Exposed mothers used alcohol in pregnancy, while 29 percent of the Control mothers did. Finally, 19 percent of the Control mothers and 33 percent of the Exposed mothers reported smoking marijuana during pregnancy. Additional background data are presented in Table 1. PROCEDURE Natural language data was collected from a five to eigth minute semi-structured play session between mother and child, these sessions being part of a more extensive 24-month follow-up visit that were Malakoff, Mayes & Underhill; WPA 4/12/96 6 conducted at the Yale Child Study Center. The play sessions were conducted in a small room with a one-way mirror to permit observation and videotaping. No experimenter was present in the room during the play session. The room contained a given set of toys that included a teddy-bear, a tea-set, blocks, a blanket, a train, two picture books, a ball and a telephone. Mothers were simply instructed to play with their children as they normally would. Each videotaped session was transcribed by two independent transcribers, and discrepancies were resolved by a three-person team who viewed the entire videotape. Two members of this team had not previously transcribed the session. The CHILDES language analysis system (MacWhinney, 1995) was used to code and analyze the transcribed sessions. Transcriptions were coded for five broad morphological categories (nouns, verbs, modifiers, communicators, and functors). Language measures were obtained using the MLU, MLT, and FREQ analyses. Variable Definitions • Utterance: Phrase or sentence. • Turn: Complete exchange by one speaker. • MLU: Mean number of morphemes per utterance. • TTR: Ratio of different words/number of words. Ignores variation in tense and number. • Different Words: Total number of different words used over 4 minutes. • Communicators: Yes, no, thank-you, and non-words that are used to sustain an interaction. • Modifiers: Adjectives and adverbs. Malakoff, Mayes & Underhill; WPA 4/12/96 7 • Functors: Prepositions, pronouns, determiners, quantifiers, whwords, etc. RESULTS The mean length of the play session for both groups was 5.7 minutes (SD = 1.0), with range of 4.27 to 8. 27 minutes. In order to control for differences in session length, we carried out analyses only on the first four minutes of the play session. Although there were few statistically significant group differences, the analyses revealed striking trends differentiating the two samples. These trends were apparent for both the language production of the mothers and that of the toddlers. Basic Language Measures The variables used to assess basic language complexity were: number of words, mean words per utterance (MWU), mean words per turn (MWT), mean length of utterance (MLU), number of different words (DW), and type-token ratio (TTR). The first four variable capture the complexity of the production in terms of its length. Because MLU counts number of morphemes, rather than number of words, it may not accurately reflect the langauge complexity anong this population because of dialectical differences. We thus also included MWU and MWT, which are not as sensitive to these difference. The latter two variables, TTR and DW, capture teh diversity of the vocabulary produced. Both measures count only root words; that is noun pluralisation and verb conjugation are ignored. Because TTR is sensitive to the total number of words produced, the total number of difference words produced over four minutes was also included in the analyses. Malakoff, Mayes & Underhill; WPA 4/12/96 8 Both Exposed and Control dyads showed great heterogeneity on all language measures (see Table 2). Mean MLU for both groups was at the lower end of the range reported for children of this age. Although there were no significant group difference among the toddlers, Figures 1 and 2 reveal a striking trend suggesting that children and their caregivers were producing less complex language than that found among the Control group. Mean Z-scores on all measures of child language complexity were below the grand mean for the Exposed , while theControl group consistently was above the grand mean. However, none of these group differences achieved statistical significance. Caregivers' language production revealed a similar pattern of results, with the exception of TTR. The caregivers' language production differed significantly on MWU (t= 2.20; p < .05), while the mean difference on MLU was close to significance (t= 1.84; p <.10). TTR showed the reverse pattern, with Control mothers showing a lower ratio of different words to total words than Exposed mothers. Control mothers also produced more words than Control mothers Morphological Analysis Findings from the morphological analysis were consistent with the previous analyses. Figure 3 shows that, as compared to their Control peers, children in the Exposed dyads relied more heavily oncommunicators, and were less likely to use functors, modifiers, and verbs than their peers in the Control group. Although only the mean difference for percentage of functors used achieved significance (t = 2.36; p<.05) the overall trend suggests that toddlers in the Exposed group were producing less complex language than their Control peers.. Malakoff, Mayes & Underhill; WPA 4/12/96 9 There were few differences between the groups in the composition of the mother's language. The largest categories of words used by all mothers were functors (a catch-all category for "little words") and verbs , reflecting, most likely, the degree to which commands were used by both groups. Modifiers (adverbs and adjectives) were infrequently used by both groups. Type of play We have begun examining language complexity separately for linguistic interaction centering on the book (bookplay) and play with other objects (non-bookplay) . I just briefly touch on these findings as they are still preliminary. As expected, we found significant differences in the nature of the language used in these two types of play: linguistic interaction centering on the book was less rich than that centering on other objects. Less language was produced, overall, during bookplay (controling for time), and the language produced was less diverse. Our preliminary analyses indicate that the trends suggesting a greater linguistic complexity among Control dyads that we found for the overall play session are also apparent when the two types of play are examined separately. That is, during both bookplay and non-bookplay, the linguistic interaction between Exposed dyads is less complex than that between Control dyads. IN CONCLUSION Although few of the group differences achieved statistical significance, there are striking and consistent trends suggesting that the language used in play interaction is less complex between the Exposed dyads than between the Control dyads. In addition, the results suggest Malakoff, Mayes & Underhill; WPA 4/12/96 1 0 that the children in Exposed dyads show less morphological development than their Control peers.. Given large individual differences within both samples these findings must still be treated as suggestive. We are currently expanding our sample and beginning the transcription and analysis of 30-month fol low-up. Malakoff, Mayes & Underhill; WPA 4/12/96 1 1