An extensive literature on cognitive dysfunction in children with fetal alcohol syndrome (FAS) exists. Numerous researchers have consistently found intellectual deficits in children with FAS, the average IQ of these children being in the borderline range (70–79) (Mattson et al., 1997; Streissguth et al., 1990). On tests of academic achievement, children exposed to alcohol prenatally tend to earn lower scores on arithmetic tests than on other tests (Streissguth et al., 1994). There is increasing evidence that alcohol-exposed children are deficient in different components of attention and executive control functioning (Coles et al., 1997, Kodituwakku et al., 1995). Other researchers have obtained evidence for impairments in information processing (Jacobson, 1998), aspects of number processing (Kopera-Frye et al., 1996), visual-spatial reasoning (Carmichael Olson et al., 1998), visual memory (Uecker and Nadel, 1996), verbal learning and memory (Mattson et al., 1996), language (Abkarian, 1992), and motor functions (Roebuck et al., 1998 a,b). Behavioral and emotional difficulties in these children have also been documented (Steinhausen and Spohr, 1998). Thus, researchers have found relative weaknesses in most areas of cognitive functioning, as well as in emotional functioning, in children with substantial prenatal alcohol exposure.

Notwithstanding notable advances in delineating neurobehavioral deficits in children with FAS, a neurocognitive profile that characterizes this syndrome has not yet been determined. This is not surprising because cognitive dysfunction associated with a teratogen comprises interactive effects of multiple factors (Fraser, 1977). There is considerable animal and human literature that indicates the variability of teratogenic effects as a result of interaction between a teratogen, the genotype it is acting upon, and environmental variables (Fraser, 1965; Kalter, 1965). O’Connor et al. (1993) found that prenatal alcohol exposure had a negative impact on infant affect which, in turn, influenced the elicitation of environmental stimulation needed for cognitive development. A host of variables that are known to impact negatively on cognitive and emotional development of the child are often found in the environment of a drinking mother (Coles, 1995). These include low socioeconomic status (Aylward, 1992), maternal depression (Field, 1998), low parental education, violence, and social disruption.

Cultural influences on central nervous system dysfunction of a child prenatally exposed to alcohol are ill-understood. There is a vast literature on cross-cultural differences in cognitive development (Kagan and Klein, 1973; Rogoff, 1986). Cultural differences have been documented in planning and problem solving (Strohschneider and Güss, 1998), mental rotation (Núñez et al., 1998), interpretation of pictures (Liddel, 1997), memory (Kagan and Klein, 1973), perceptual skills (Segall et al., 1957), intellectual functioning (Price-Williams, 1962), linguistic competencies (Werner, 1979), and motor development (Super, 1976). Accordingly, it is reasonable to expect some variability in cognitive functioning in children with FAS as a function of the cultural context.

Although FAS has been reported in nonwestern nations (Tanka et al., 1981), most neurodevelopmental studies of alcohol-exposed children have been conducted in North America and Europe. Thus, the study of neurocognitive functioning in children prenatally exposed to alcohol from nonwestern cultures may contribute considerably to our understanding of the interactions between alcohol, genetic factors, and environment.

This article reports on the results of a study aimed at characterizing general cognitive functioning in a group of children with FAS from a community in the Western Cape Province of South Africa. The study was carried out as part of a large collaborative research project involving South African and US investigators (May et al., in press). Several factors led to the selection of this site. First, high prevalence rates of FAS among children in this community have long been suspected by clinicians. Croxford and Viljoen (1999) found heavy drinking practices among women of the region who attended prenatal clinics. Second, the majority of mothers who had given birth to a child with FAS were available for interviews. Third, the majority of children with FAS live with their biological mothers. Fourth, this Afrikaans-speaking community, which was made up predominantly of people of mixed ancestry (referred to as ‘colored’ in South Africa), was culturally homogenous. Fifth, and finally, the community was very supportive of the project.

In view of the exploratory nature of the study, the primary hypotheses were broad. It was expected that both alcohol exposure and adverse postnatal environmental conditions would impact on cognitive performance of the children with FAS in the community. However, no specific hypotheses regarding interactions due to cultural and genetic factors were formulated.

Participants

Sixty-eight children in grade 1 classes, 34 diagnosed with FAS and 34 controls, participated. The FAS group comprised 20 boys and 14 girls, who ranged in age from 6.33 years to 8.00 years with a mean of 6.99 years. Comprehensive maternal interview data were available only for 25 of these children, because five mothers had died and the others had moved out of the area. Twenty-three children with FAS were living with their biological mothers. Predominantly farm laborers, these mothers had received limited formal education (see Table 1) and were economically disadvantaged [mean family income = 128.17 rands per week (18 US dollars)]. Alcohol use or abuse was highly prevalent in their homes, with the reported maternal current consumption of about 12.9 drinks a week (about 183 g), about five times that of control mothers in this community (Viljoen et al., unpublished data, 2000).

---

Table 1. Group Comparisons on Participant Characteristics and Demographic Variables

---

Identification of the FAS Group

As described by May et al. (in press), the identification of the FAS group was accomplished in two phases. In the first phase, 406 children in grade 1 classes were examined fully to establish local growth parameters and identify alcohol related features. In the second phase, only those children who fell below the 10th percentile in head circumference or height and weight (N = 220) were referred for a complete morphology evaluation. Each child was evaluated by two teams of dysmorphologists, who were blind to the child’s background and cognitive functioning. Fifty-three children received a preliminary diagnosis of FAS based on morphology examinations conducted in the above two phases. Forty-nine controls with no evidence of growth retardation were selected from the same community. The children who were 8 years of age or younger were given the Griffiths Mental Development Scales (Griffiths, 1984) and those who were older than age 8, the Senior South African Individual Scales (Van Eerden, 1992). Information on prenatal alcohol exposure of these children was obtained through an extensive maternal interview [see May et al. (in press)] for information on the questions regarding the quantity and frequency of exposure]. The test administrator was unaware of the dysmorphology data and the child’s alcohol exposure history. Furthermore, the maternal interviewer was blind to the dysmorphology and developmental test results. Forty-six children received a final diagnosis of FAS, based on the dysmorphology data, exposure information, and developmental test results. Included in the present study were 34 children with FAS who took the Griffiths Mental Development Scales.

Control Group

Thirty-four controls were selected from the sample of 49 children described above. The mothers of these controls did not admit to moderate or heavy drinking during pregnancy. The two groups were matched for age, sex, first language, family income, and where possible, school. As Table 1 shows, the control group had significantly higher measures of height (mean = 115.05 cm), weight (mean = 20.96 kg), and head circumference (mean = 50.90 mm) than the FAS group.

Developmental Assessment

As noted above, the Griffiths Mental Development Scales was used to assess cognitive and motor development of the participants. This test is designed to measure the developmental progression in six areas: Locomotor, Personal-Social, Hearing and Speech, Eye and Hand Co-ordination, Performance, and Practical Reasoning. The first five scales contain items pertaining to the child’s development from infancy to middle childhood (age 8) and the last scale begins at the third year of life. Because each of these subscales is designed to measure a separate sequence of events, a subscale can be used alone. They are also scaled differently, with means ranging from 99.79 to 100.46 and standard deviations from 15.58 to 17.43. Norms for the Griffiths Scales, which were originally standardized in the UK, are available for some South African population groups.

Subscales

Locomotor: The Locomotor scale measures gross-motor skills ranging from pushing with feet and holding head erect in infancy to jumping and skipping in middle childhood.

Personal-Social: This subscale measures adaptive skills such as stating the first and last names, dressing self, tying shoe laces, and having special play mates.

Hearing and Speech: Considered to be the most intellectual on the test, this subscale measures verbal ability.

Eye and Hand Coordination: This subscale contains items related to fine motor coordination such as drawing geometric figures and threading beads.

Performance: The purpose of this subscale is to measure skills in manipulation and speed of performance. It contains timed tasks involving pattern construction.

Practical Reasoning: This subscale measures a range of skills that involve practical reasoning, such as learning numerical concepts related to money and time and spatial concepts related to orientation in space.

Home Environment

Three measures of the participant’s home environment were obtained: quality of caregiving, maternal drinking, and maternal education. After home visits, the maternal interviewer rated the quality of caregiving at home on a 5-point scale: (1). children’s needs attended to, positive parent-child interactions; (2). some parent-child interactions and no history of violence; (3). some evidence of child neglect and alcoholism; (4). history of domestic violence, family discord, and neglect; (5). history of violence, neglect, and physical abuse, and children removed from home because of these adverse conditions.

Data Analysis

Group means for subject characteristics and demographic variables were compared with the independent t test. To test the effect of alcohol exposure on cognitive-motor skills, we performed a multivariate analysis of covariance (MANCOVA) on the Griffiths subscales covarying out maternal education. The decision to perform MANCOVA was dictated by two observations. First, intercorrelations for the Griffiths scores ranged from moderate to high (0.37 to 0.77), suggesting that the subscales were partially redundant (Table 2). Second, there was a significant group difference in maternal education (Table 1), which correlated with the Griffiths Mental Development Scales (Table 2). Accordingly, testing of a group effect on the subscales of the Griffiths required adjustment for the difference in maternal education. The main effect of the home environment was not tested because the assessment of the quality of caregiving was judged to be unreliable. There was a high correlation between the quality of caregiving scores and maternal drinking (r = 0.56, p < 0.0001), which raised the question of whether the presence of alcoholism in the family negatively biased the ratings of the home environment.

---

Table 2. Intercorrelations for the Griffiths Subscales and Maternal Education

---

A significant multivariate effect was followed up with univariate analyses of covariance (ANCOVAs), again adjusting for maternal education. One major limitation of the ANCOVAs concerned their failure to take the intercorrelations among the dependent measures into consideration. Therefore, we supplemented ANCOVAs with two other methods that considered the relatedness of the dependent measures in the assessment of the relative contributions of these measures to the overall group effect (Weinfurt, 1998). Recommended by Huberty and colleagues (Huberty, 1984; Huberty and Smith, 1982), the first method involved the use of the F-to-remove values from a stepwise discriminant function analysis. We forced all the subscales and the covariate in to the SPSS-X discriminant function analysis and obtained the F-to-remove values from the last step of the stepwise procedure. The results of this method were confirmed by a conceptually comparable procedure recommended by Wilkinson (1975). The latter procedure involved performing a series of MANCOVAs leaving out one dependent variable from the analysis at a time. The change in the multivariate F value resulting from the removal of a variable was used as an index of the contribution of that variable to the overall effect.

The variance-covariance matrices of the two groups were homogenous, as indicated by a nonsignificant Box’s M test result (F = 1.262, p = 0.188). Maternal education was determined to be an appropriate covariate, because it showed a linear relationship with dependent measures (Table 2) and because it satisfied the assumption of the homogeneity of regression. The tenability of the homogeneity of regression assumption was indicated by a nonsignificant group by maternal education interaction [ F(6,47) = 0.375, p = 0.89].

Adjusted group means are presented in Table 3. A MANCOVA showed that the two groups significantly differed on the combined Griffiths scales after adjusting for differences in maternal education [ F(6,48) = 5.81, p < 0.001]. Inspection of the adjusted means revealed that the FAS group performed worse than the control group on most subscales.

---

Table 3. Adjusted Means (M), Standard Errors (SE), and Analysis of Covariance Results for the Griffiths Subscales

---

As is shown in Table 4, the Speech and Hearing subscale contributed most to the separation of the two groups, as reflected by the highest F-to-remove value (6.669) and the highest reduction in the overall F (0.75) as a result of removing it. An ANCOVA showed that the group difference in this subscale was highly significant after controlling for maternal education [F(1,53) = 21.518, p < 0.0001]. It is also shown in Table 3 that the ANCOVAs revealed highly significant effects on the Practical Reasoning, Performance, and Eye-Hand subscales, the F values being 17.52, 16.58, and 14.83 respectively. However, the F-to-remove values associated with these three subscales were low, suggesting that they were partially redundant (Table 4). In other words, any of these variables could be removed from the model without compromising it after entering the Speech and Hearing subscale.

---

Table 4. Ordering of Dependent Variables in Terms of Their Relative Contribution to the Multivariate Effect

---

The results showed that the Personal and Social subscale marginally contributed to the overall effect (F-to-remove value = 4.747), despite a nonsignificant univariate effect [F(1,53) = 0.941]. However, the Motor subscale did not contribute to the overall effect, as indicated by a low F-to-remove value (0.094). Furthermore, there was no significant group effect on gross motor skills [ F(1,53) = 1.98], both groups performing in the average range (Table 3).

In view of the finding that the Practical Reasoning, Eye Hand, and Performance subscales became redundant after the Speech and Hearing subscale was entered in to the analysis, we created a composite variable by combining these four scores. We performed a MANCOVA of the combination of the composite score, Personal-Social, and Motor subscales adjusting for the difference in maternal education. This analysis showed a highly significant group effect on the combined dependent measures [ F(3,51) = 11.37, p < 0.0001]. A comparison of relative contributions of the dependent variables revealed that the composite variable primarily accounted for the separation of the two groups. While the Personal-Social subscale accounted for some variance, the Motor subscale did not contribute to the overall effect. The F-to-remove values for the composite score, the Personal-Social and Motor subscales were respectively as follows: 30.68, 4.20, and 0.04.

As noted above, it has been well documented that children with FAS perform less competently than controls on higher-order cognitive tasks such as those measuring intellectual ability and attention. A number of researchers have obtained evidence that children with FAS are impaired at practical reasoning, especially numerical reasoning. Streissguth and colleagues (1994) reported that alcohol-exposed children showed, among other things, academic problems, particularly in arithmetic. Coles et al. (1991) found that alcohol-exposed children exhibited relative weaknesses in sequential processing and preacademic skills in mathematics. There are numerous reports that alcohol-affected children perform less proficiently than controls on a range of nonverbal tests called performance tests, such as construction of designs with blocks. Alcohol-affected children tend to earn a lower Performance IQ than their age- and gender-matched controls (Mattson et al., 1997). It has also been established that individuals with prenatal alcohol exposure display fine motor skills problems (Barr et al., 1990).

Studies of behavioral outcome in language in children with prenatal alcohol exposure have, however, produced inconsistent results. Some researchers have reported that children with prenatal alcohol exposure perform less efficiently than controls on language tests, including those measuring naming (Mattson and Riley, 1998), word comprehension (Conry, 1990), grammatical and semantic abilities (Becker et al., 1990), and pragmatics (Abkarian et al., 1992).

In the current study, the FAS group was markedly deficient in language skills in comparison with the control group. Some researchers have, however, failed to find a significant effect of prenatal alcohol exposure on language development in the affected child (Greene et al., 1990). In this study, Greene and his colleagues assessed language development in a large cohort of alcohol exposed children at 1, 2, and 3 years of age. As a group, however, these children displayed only mild physical effects of fetal alcohol exposure, the cohort including only one child with the full syndrome. This child was found to have significant language impairments. In the current study, on the other hand, all the children in the alcohol-exposed group were diagnosed with FAS. Thus, the group difference in the severity of alcohol-related damage seems to be the primary explanation for the discrepant outcome in language observed in the latter two studies.

Adaptive skills, as assessed by the Griffiths Mental Development Scales, only marginally contributed to the multivariate effect in the presence of cognitive variables. Thus, while adaptive skills were relatively independent of cognitive competencies, the former were less discriminating between the two groups than the latter. This probably explains why there is a lack of consensus among researchers with regard to the significance of adaptive skills as a behavioral measure of alcohol related brain damage. Coles and colleagues (1991) found that alcohol-affected children were comparable to controls in adaptive skills after adjusting for current maternal drinking. Kelly et al. (2000), on the other hand, have shown the comparability of evidence from animal and human research to support the hypothesis that social deficits in alcohol affected individuals are largely a result of alcohol insult rather than an environmental effect. A number of other researchers have reported social deficits in individuals with prenatal alcohol exposure (Thomas et al., 1998; Streissguth et al., 1991). Even though children with prenatal alcohol exposure may fail to acquire complex personal-social skills (e.g., understanding subtle social cues), they appear to be capable of learning relatively simple adaptive skills. Given that the control group in the present study were also from impoverished social environments they probably had not acquired advanced skills for the age group, thus diminishing the group difference. Furthermore, because the children with FAS in this study were not treated as “disabled” by teachers and parents, they were probably not deprived of opportunities to acquire basic adaptive skills.

The lack of a significant difference between the FAS and control groups in gross motor behavior is at variance with the findings of some human (Kyllerman et al., 1985) and numerous animal studies (Hannigan and Riley, 1989). Using computerized dynamic posturegraphy, Roebuck and colleagues (1998 a) have demonstrated that children exposed to alcohol prenatally had more difficulty than controls in maintaining postural balance when somatosensory and visual feedback were inaccurate. In a subsequent study, Roebuck et al.(1998b) obtained electromygraphic evidence suggesting a possible association between the balance deficits in alcohol-exposed children and a central processing dysfunction. Using animal models, researchers have demonstrated that perinatal alcohol exposure leads to a range of motor deficits including ataxia, and gait dysfunction (Hannigan and Riley, 1989; Meyer et al., 1990). It is probable that the Locomotor subscale of the Griffiths is not sufficiently sensitive to detect subtle deficits in gross motor behavior in alcohol-affected, school-aged children. The assessment of gross motor skills with this test involves observing motor behaviors (e.g., jumping, running) performed by the child to command. Chandler et al. (1996), who used methods comparable to those used in the current study, also failed to find motor deficits in a group of 3 year olds exposed to alcohol and marijuana prenatally.

It is important to note the limitations of the present data set. First, as noted above, the Griffiths Mental Development Scales were not sufficiently sensitive to detect subtle deficits in some areas of functioning. Furthermore, subscales on the test were too broad to allow a detailed analysis of functions. It has been documented that children with neurodevelopmental disorders succeed on some tasks using the processes that are different from those used by children without developing neurodevelopmental disorders (Karmiloff-Smith, 2000). Uncovering such differences will require the use of narrow band tests designed to probe specific processes (Bellugi et al., 1994). Second, the sample was relatively small and thus there was clearly limited power to detect certain possible effects. Third, we did not adequately assess environmental variables pertinent to cognitive function. For example, there is a growing literature showing that language input is an important contributor to the child’s language development (Huttenlocher, 1998). Accordingly, an accurate assessment of language impairment resulting from prenatal alcohol exposure will require adjusting for possible group differences in maternal verbal input during mother-child interactions.

Despite these limitations, the current study provides the first description of cognitive-motor functioning in a group of reliably diagnosed children with FAS in a nonwestern community. The findings suggest that the pattern of cognitive-motor deficits in this group of children is, by and large, commensurate with that reported in the literature. Accordingly, it seems reasonable to conclude that neurobehavioral findings on prenatal alcohol exposure can be generalized across ethnic-cultural boundaries.

ACKNOWLEDGMENTS

We gratefully acknowledge the assistance of the principals of the participating schools and the Western Cape Department of Education. The authors also thank Jon M. Aase, Kenneth L. Jones, and Luther Robinson for performing clinical morphology evaluations; Lesley Brooke for facilitating testing of the children; Julie Croxford for conducting maternal interviews; J. Philip Gossage for compiling maternal data; Harold Delaney for giving advice on data analysis; Barbara Laughton for performing developmental evaluations; and Faye Calhoun and Ken Warren for their support and encouragement.