Minority Students in Special and Gifted Education (2002)

As Chapter 2 suggests, one can observe variation in the proportion of students from different ethnic groups assigned to special education and gifted and talented programs without knowing whether there are too many or too few members of any racial/ethnic group in any given category. To answer such a question, one would have to understand the source of the disproportion.

The committee considered three potential explanations, which are not mutually exclusive and which may well operate in tandem:

1. By the time they reach school age, children differ in the cognitive and behavioral characteristics that are related to placement in special education and gifted and talented programs. These differences may be distributed disproportionately among children in different racial/ethnic groups.

2. Schools may have an independent influence on the academic success and behavioral problems of students that varies with the racial/ethnic composition of students in the school, or with the race or ethnicity of the individual student.

3. Standards (or the implementation of standards) for referral and assessment of students for special education and gifted and talented programs may be biased, or they may be applied differentially across racial/ethnic groups to produce disproportion.

In this part we focus on the first explanation, asking whether characteristics that predict achievement and behavior problems differ across racial/ ethnic groups. To do so, we ask what is known about factors that significantly contribute to variation in cognitive and behavioral function. Because such a review could itself span volumes, we focus in Chapter 3 on factors for which a research base is available to suggest both that the factor is significant in cognitive and behavioral development and that prevalence differs by race or ethnicity.

In Chapter 4 we review what is known from a now-extensive research base about early intervention programs and their potential to improve cognitive and behavioral outcomes for children at risk. We focus particularly on the more limited evidence available regarding the impact of early intervention on the placement of children in special education programs once they have entered school. Our early childhood recommendations appear at the end of Chapter 4.

Research in a variety of biological and social sciences in the past few decades has brought about substantial change in earlier understandings of the contributors to cognitive and behavioral function. In classic works by Galton (1869) and Burt et al. (1934), differences in intelligence were attributed to heredity, emphasizing a perception of the child as constitutionally separate from the environment. In the social sciences, however, a series of landmark studies in the 1930s and 1940s of infants and young children reared in institutions drew attention to the environmental and contextual contributors to child development (Ramey and Sackett, 2000). The research that ensued using animal models (Sackett et al., 1999), the study of children who experienced deprivation in institutional settings, and the proactive early intervention efforts in the 1960s collectively provided compelling evidence that early experience matters a great deal.

While genetic and physiological factors continue to play a central role in the understanding of cognitive and behavioral performance, the perception of the child as constitutionally separate from the environment no longer holds. Understanding the development of child behavior increasingly has required a focus on aspects of the environment that serve as moderators of performance (Sameroff, 1993; Ceci et al., 1997). The analytic lenses and

methods of different social sciences have focused attention on different correlates of achievement and behavior. Economics has focused on the role of family income and the education (or human capital) of parents; sociology looks more at the community, school, and family structure; and psychology focuses on the interactions among family members and other important individuals to understand social, emotional, and cognitive development. In seminal work that launched a line of research in social ecology, Bronfenbrenner (1979) suggested that the development of the child needs to be viewed as influenced by all of these factors. The current scientific task is to catalog and describe the relevant contributions of these dynamic components through time.

As the tools of the social sciences have become more powerful, so have those for studying the brain. We have come to understand that biological and environmental factors are not completely separate parts of the picture (Shore, 1997; Wahlsten and Gottlieb, 1997; Bidell and Fischer, 1997; Hunt, 1997). They combine as two pigments in a single paint, together determining a color that neither alone could create. Genetic and health influences themselves are no longer seen as purely biological (National Research Council [NRC], 2000b). Genetic expression is now understood not as a fixed and predetermined influence, but as a probabilistic propensity responsive in some degree to environmental influence (Plomin, 1997; Sameroff, 2000). Researchers can observe in animal studies and, to a more limited extent, in human studies that environmental experiences change the very physiology of the brain: encoding new experiences fosters new brain growth (Greenough and Black, 1992; Black and Greenough, 1986).

Contemporary genetics suggests further that the gene-environment dynamic is not one in which each has a distinct but separate role to play, nor that environment determines whether a gene does or does not exert the influence of its predetermined code. Rather, the function of the genetic system is itself context dependent (Bidell and Fischer, 1997). A dramatic instance is the case of a parasitic wasp that lays its eggs in two different hosts, a butterfly or a fly. Offspring that develop in the butterfly host have wings, but those that develop in the fly host do not, despite an identical genetic code (Gottlieb, 1992; Bidell and Fischer, 1997). While a substantial body of research has demonstrated the importance of genetics in explaining variation in cognitive and behavioral performance (Bouchard, 1997; Hunt, 1997), it is clear that genetic variation cannot be understood separately from context.

Figure 3-1 presents one schema that explicitly acknowledges the dynamic, reciprocal interplay between biology and experience (Ramey and Ramey, 2000). In this model, cognitive, social, and emotional development is an outgrowth of the transactions between children and the significant others in their environment. But a myriad of factors—biological, social,

economic, and cultural—influence the behaviors of both the child and the adults engaged in those interactions.

Below we review the current knowledge base regarding early influences on cognition and behavior by looking first at research regarding the biological influences on early development and then the research on environmental (social, emotional, economic) influences. The artificial nature of the dichotomy between biological and environmental influences is perhaps most evident when we discuss the role of poverty under the social and environmental context of development. Each of the biological factors discussed is found to vary with poverty status as well. Increasingly, research suggests that the biological and social worlds must be seen as tightly intertwined if the goal is to understand the cognitive and behavioral outcomes for children and the potential roles for social intervention (McLoyd and Lozoff, 2001; Ramey and Ramey, 1998). Despite the contemporary understanding of their inseparablity, the research enterprises regarding biological and social contributors have for the most part been conducted independently and from different disciplinary research traditions. We therefore look at each piece individually, after which we turn to their interactions.

Our focus in this chapter is, of necessity, on early harms and risk factors that impair normal development, as well as interventions that can diminish the impact of those risk factors. Our limited attention to issues regarding accelerated development reflects the research base, and the research base in turn reflects research opportunities (NRC, 2000c). Much of what we have learned about the developing brain, for example, we have learned because an abnormal event (premature birth, trauma, fetal alcohol syndrome) has occurred to call attention to the phenomenon. The group for study is clearly defined, and the contrasting case between the normal and the abnormal circumstance is clear. The group of high achievers is not so easily defined by an event. Moreover, the social policies designed to address the needs of disadvantaged children provide opportunities for research on the effects of physical and environmental risk, and of its amelioration, on development. No similar scaled, sustained research effort has been undertaken to better understand high achievement.

Nonetheless, the complex of factors that influence student achievement is likely to do so across the entire distribution. In Figure 3-2, achievement is plotted as a normal distribution, with the “main population” representing a hypothetical circumstance of a general population in which students differ in achievement because they themselves differ and because their environments differ within an average or low-risk range. The diagonal area of the distribution represents a hypothetical group of students who might require additional supports (special education at the lower end, gifted education at the upper end) when teaching targets students at the mean. We focus in this chapter on circumstances that diminish achievement—or shift the location

FIGURE 3-2 Idealized representation depicting displacement of subgroups with regard to main population on any variable that is normally distributed.

SOURCE: Case, Griffin, and Kelly (1999). Reprinted with permission.

of the curve back, as in the “subpopulation” for those developing in high-risk environments. This shift simultaneously increases the number of children with special needs at the lower end and decreases the number of high achievers who may be identified as gifted at the upper end. In a sense then, this chapter is about both groups, although those cases at the left tail of the distribution have been studied more because of their distinguishing characteristics than those in the right tail.

The importance of the early years of life to development is incontrovertible (Ramey et al., 2000; Ramey and Ramey, 1999; NRC, 2000a). The unparalleled pace of brain growth and the development of fundamental cognitive, emotional, social, and motor processes make the period from conception through infancy one of exceptional opportunity and vulnerability (McLoyd and Lozoff, 2001). While the plasticity of the brain appears to extend well into adolescence, with growth in some areas of the brain as late as the third decade of life (NRC, 2000a), children who experience biological insults and stressors early in life are at greater risk for long-term developmental problems (McLoyd and Lozoff, 2001). Deprivation in the extreme can produce functional mental retardation and aberrant social and emotional behavior in animals born healthy and with good genetic endowment (Ramey and Ramey, 1999). In humans, mild mental retardation with

no documented biomedical cause has been observed at elevated levels among very poor families (Garber, 1988).

For any individual child, genetic and experiential information come together in a process that organizes the brain to function. An NRC report on the science of early childhood development lists environmental factors that play a significant role in modulating prenatal and early postnatal brain development (see NRC, 2000a:199). The list, although not exhaustive, includes factors selected on the basis of clinical importance, the availability of basic research on brain effects, and/or the existence of relevant clinical studies (Table 3-1). In this report, we focus on a subset of these factors, which research suggests are implicated in differential developmental outcomes for children by race: premature birth (adequate gestation), fetal alcohol and nicotine exposure, and micronutrient deficiency, and exposure to lead. We do not suggest that these factors are uniquely important to healthy development. Other critical factors, such as the role of iodine in cognitive development, are not considered here because in this country they are unlikely to contribute to current developmental differences, since effective prevention measures have eliminated the iodine deficiency problem for children of all races (Stanbury, 1998).

In each year in the past decade, between 7 and 8 percent of babies were born at weights below 2,500 grams. The vast majority of low-birthweight

children have normal outcomes. As a group, however, low-birthweight babies have higher rates of neurodevelopmental and behavioral problems (Hack et al., 1995; McLoyd and Lozoff, 2001). They are more likely to have lower IQ, cerebral palsy, less emotional maturity and social competence, and attentional difficulties (National Research Council, 2000a). A recent study of siblings found that those born weighing less than 5.5 pounds were almost four times less likely to graduate from high school by age 19 than their normal-birthweight siblings—15.2 percent of low-birthweight siblings, compared with 57.5 percent of normal-birthweight siblings graduated on time (Conley and Bennett, 2000).

The neurocognitive differences that are observed with low birthweight are more pronounced the lower the weight (Breslau et al., 1996). Similarly, the child’s general developmental status and intelligence scores decrease with reductions in gestational age (Saigal et al., 1991).¹ At the borders of viability (22-24 weeks) where mortality is high, neurological damage to babies who survive is often sustained (Allen et al., 1993). But even lower-risk preterm babies (27-34 weeks) sometimes show cognitive lags compared with their full-term counterparts (de Haan et al., 2000).

Damage from premature birth arises in part due to the interruption of the normal process of brain development in utero, including the expected intrauterine stimuli and nutrients important for growth (NRC, 2000a). Recent research suggests that even when preterm infants have benign neonatal courses, they show poorer performance on elicited imitation tasks at 18 months (de Haan et al., 2000). But premature birth also increases the probability that infants will experience pathological events that directly injure the brain. Intracranial hemorrhage, for example, occurs in approximately 20 percent of 28- to 34-week infants and 60 percent of infants born between 24 and 28 weeks. The hemorrhage tends to be more severe at lower gestational ages, resulting in a higher likelihood of a major disability. Even with less severe hemorrhages, however, the risk of minor disabilities— including behavior problems, attention problems, and memory deficits— rises (Lowe and Papile, 1990; Ross et al., 1996; National Research Council 2000a; McLoyd and Lozoff, 2001).

In the United States, low birthweight is more common among blacks than any other racial/ethnic group (McLoyd and Lozoff, 2001; David and Collins, 1997; Foster, 1997) (see Table 3-2). Blacks are about twice as likely as whites to be born at low birthweights (see Figure 3-3), even controlling for socioeconomic status (Conley and Bennett, 2000; Foster, 1997). Interestingly, the incidence of low birthweight for babies of African-born

black women more closely resembles that of U.S.-born whites than of U.S.-born blacks (David and Collins, 1997). Among whites there is a strong association between maternal education and low birthweight (National Center for Health Statistics, 1998; Guyer et al., 1997). While this is true of blacks as well, the rate for black mothers who have 16 or more years of education is still above that of whites with less than a high school education.

The link between income and the incidence of low birthweight has been well established (McLoyd and Lozoff, 2001; NRC, 2000a; Kiely et al., 1994). This relationship persists even when the mother’s educational attainment, sex, birth order, and race/ethnicity are controlled (Conley and Bennett, 2000). In a recent provocative study, however, income lost its significance when parental birthweight status was controlled. The probability of having a low-birthweight child increased fourfold if the mother herself had low birthweight, and sixfold if the father had low birthweight (Conley and Bennett, 2000). This is a single study, however, and has not been replicated to our knowledge. At the same time that this study questioned the role of income in predicting the incidence of low birthweight, it found that an income-to-needs ratio of the family during the child’s first five years was a significant predictor of the effect of low birthweight on timely high school graduation.

The incidence of low birthweight declined in the 1970s and early 1980s but has risen 10 percent since then—from a low of 6.7 in 1984 to 7.6 in 1998. Much of this is due to the increase in the odds of survival for low-birthweight babies due to increases in medical technologies (Seelman and Sweeney, 1995) and to a rise in multiple-birth rates among white women. The rate has declined overall for black mothers but has remained stable (at about 3 percent) for very small babies of 1,500 grams or less (McLoyd and Lozoff, 2001).

Several interventions have been shown to reduce the incidence of low birthweight: prenatal care, maternal nutrition and adequate weight gain during pregnancy, control of hypertension, and avoidance of long work hours and excessive physical exertion toward the end of pregnancy (Luke et al., 1995; McLoyd and Lozoff, 2001). Interventions focused on improving outcomes for low-birthweight babies have also demonstrated some effectiveness. These range from changes in the care these infants receive in neonatal intensive care units (Als, 1997; Hernandez-Reif and Field, 2000) to the Infant Health and Development Program, which provided comprehensive services to the infants and their families for several months after discharge (see Box 3-1). Additional stimulation of low-birthweight babies can reduce the cognitive impact, especially for the heavier babies in families with lower socioeconomic status (Hack et al., 1995; Ramey et al., 1992).

FIGURE 3-3 Percentage of infants born of low birthweight by race and Hispanic origin, 1980-1998.

Adapted from Centers for Disease Control and Prevention (2000) and National Center for Health Statistics (1998).

Maternal alcohol consumption during pregnancy can impair the physical and mental development of the fetus, although the vulnerability of individual fetuses varies for reasons that are not yet entirely understood (NRC, 1996). In its most serious form, fetal alcohol syndrome (FAS) causes craniofacial changes, growth retardation, and central nervous system impairment, including mental retardation and/or hyperactivity (NRC, 1996). Even among children who do not have FAS, however, moderate to heavy drinking during pregnancy has been associated with growth deficits and developmental lags (Streissguth et al., 1996).

National data on the effects of alcohol on fetuses are limited. Indeed, the potentially serious effects of alcohol have been recognized only in the past 30 years. Data collected in 1988 in the National Maternal and Infant Health Survey (Faden et al., 1997) suggest that heavy alcohol consumption during pregnancy (six or more drinks per week) is confined to a relatively small segment of the maternal population. But that rate is considerably higher for American Indian/Alaskan Native women (2.2 percent) and black women (1.2 percent), than for white (0.4 percent), Hispanic (0.3 percent), or Asian/Pacific Islander women (0.7 percent) (see Figure 3-4). The incidence of FAS births is approximately 10 times higher among blacks than among whites (Abel, 1995). No national data are available for other racial/ ethnic groups; however, a surveillance project in four communities (Duimstra et al., 1993) estimated that the rate may be 30 to 40 times higher

for American Indians/Alaskan Natives than for whites (McLoyd and Lozoff, 2001).

While the reported number of women who drink during pregnancy has declined since the mid-1980s (Serdula et al., 1991), the overall change was driven by a decrease in light drinking (Hankin et al., 1993). In 1995 the Centers for Disease Control and Prevention (CDC) found the incidence of drinking at a level that put the fetus at risk for neurobiological damage was at 4.5 percent (Ebrahim et al., 1998). No data are available on differences by race/ethnicity over time.

FIGURE 3-4 Alcohol consumption after finding out about pregnancy: Expectant mothers in the United States, 1988.

SOURCE: Faden, Graubard, and Dufour (1997). Reprinted with permission.

Alcohol is not alone in its harmful effects on a developing fetus. There is a substantial body of literature to suggest that nicotine has a detrimental impact (Levin and Slotkin, 1998), including increasing the probability of low birthweight (Aronson et al., 1993; Morrison et al., 1993) with the consequences described above. Long-term effects of maternal smoking during pregnancy on later child behavior, controlling for birthweight and other confounding effects, have been found in many studies (Williams et al., 1998; Weitzman et al., 1992; Fergusson et al., 1993), although some have found the effects to be substantial (Williams et al., 1998) and others small (McGee and Stanton, 1994). Mild attentional (Denson et al., 1975; Fried, 1992; Landesman-Dwyer and Emanuel, 1979; Picone et al., 1982a, b; Jacobson et al., 1984) and cognitive effects (Fergusson et al., 1993; Hardy and Mellits, 1972; Lefkowitz, 1981; Naeye and Peters, 1984; Keeping et al., 1989; Butler and Goldstein, 1973; Dunn and McBurney, 1977; Rantakallio, 1983; Gueguen et al., 1995) have been found as well. At 5 and 6 years of age, children exposed to tobacco prenatally had lower receptive language scores and poorer performance on memory tasks (Fried et al., 1992 a, b). Most effects occur at higher exposures (20 or more cigarettes a day) (Williams et al., 1998; Levin and Slotkin, 1998).

Because maternal smoking may be correlated with other maternal conditions and behaviors related to child outcomes, the causal connection between tobacco and those outcomes is difficult to establish with certainty, although some studies have been done on large-scale longitudinal data that allow for control of a great many confounding factors (Williams et al., 1998). As with lead exposure, research using animals allows for fuller experimental control. Such research confirms that prenatal exposure to nicotine is itself related to adverse consequences, including damage to the central nervous system (see Levin and Slotkin, 1998, for a review). Adverse effects on cognition and behavior are not as robust as the physiological effects. Levin and Slotkin hypothesize that redundancy in neural systems allows for the use of alternative pathways in order to compensate for damage. If this is the case, then higher levels of complexity should uncover the difference between exposed and unexposed rats. As with animal study of the effects of lead exposure, higher levels of complexity did reveal lower performance in exposed rats (Levin et al., 1996; Levin and Slotkin, 1998).

Cigarette smoking is substantially higher among American Indian/Alaskan Native pregnant women than among any other racial/ethnic group. For Asians, blacks, and Hispanics, smoking rates during pregnancy are below that of whites (see Table 3-3). Several studies have found, however, that the biochemical measurement of serum cotinine, the primary metabolite of nicotine, is higher for non-Hispanic blacks than for non-Hispanic whites at the same exposure level (Caraballo et al., 1998; Clark et al., 1996; English et al., 1994; Wagenknecht et al., 1990; Pattishall et al., 1985). Serum cotinine is a widely used indicator of tobacco use and environmental tobacco exposure. In light of this finding, it is particularly encouraging that between 1983 and 1998 the number of young black women who smoke fell from almost 28 percent to under 10 percent—far below the rate for their white counterparts.

Exposure of a fetus to cocaine has been of increasing concern in the past 15 years as usage rates have risen. Careful research is complicated, however, because the illegal status of the drugs affects sampling, and because cocaine use is often accompanied by the use of other drugs and by alcohol and tobacco use. The independent contribution of the cocaine is thus difficult to determine (Msall et al., 1998). A recent attempt at a meta-analysis of the research on cocaine use concludes that available studies are sufficiently flawed to make any conclusions from them questionable (Lester

et al., 1998). Animal studies of cocaine exposure at very high levels show effects on growth, but behavioral and cognitive consequences have not yet been established (Paule, 1998).

Children who are seriously malnourished tend to have low IQs (Stein and Kassab, 1970; Winick et al., 1975; Zeskind and Ramey, 1978, 1981). Malnourishment, however, is generally coincident with other stressors— including poverty, poor schooling, and neglect—that make it difficult to identify the impact of malnutrition alone (Sigman and Whaley, 1998). Moreover, malnutrition has been found in some studies to affect motivational and emotional responsiveness (Galler et al., 1983; Sigman and Whaley, 1998), suggesting that the effect on cognition may be mediated, at least in part, through reduced attention and interaction.

A few studies that have controlled for parental socioeconomic status have found positive associations between nutritional supplementation and IQ. One such study with Kenyan children (Sigman et al., 1989) found positive correlations with animal protein and fat intake. Several studies using random assignment experimental designs with pregnant women thought to be at risk found vitamin and mineral supplementation during pregnancy increased the child’s IQ at age 1 (Rush et al., 1980) and age 4 (Harrel et al., 1955) compared with control children (Eysenck and Schoenthaler, 1997).

Vitamins and minerals in the diet play an important role in both physical and mental well-being (Essman, 1987). An association between nutritional supplementation and IQ scores has been found (Dean and Morgenthaler, 1990; Dean et al., 1993), as has an association between supplementation and behavior (Schoenthaler, 1991). One of the strongest claims for the impact of micronutrients is Lynn’s (1990) argument that increases in the mean IQ of the population over time (Flynn, 1987) can be explained largely by improved nutrition. While some support Lynn’s view of the importance of nutrition with caution regarding the ability to specifically isolate its contribution (Sigman and Whaley, 1998), others accept as incontrovertible the role of nutrition in cognitive development but caution that continued rises in IQ in countries like the United States and the Netherlands since 1970 are not likely to be explained by nutrition, suggesting other explanatory variables are important as well (Martorell, 1998).

Iron deficiency is one of the most common single-nutrient disorders (McLoyd and Lozoff, 2001). Its consequences are wide-ranging, including compromised cognitive and social development, short attention span, and impaired learning capacity (Viteri, 1998; Lozoff et al., 2000). The effects of iron deficiency interact with other developmental stressors because it increases the absorption of lead and impairs absorption of fat. There is considerable evidence that malnutrition and altered iron transport contribute to the detrimental effects of prenatal alcohol exposure (McLoyd and Lozoff, 2001). Iron deficiency in pregnant women is associated with poorer birth outcomes, including low birthweight (Viteri, 1998).

Iron affects cognition and behavior through its impact on brain structure and function. It plays a role in both myelin formation and in the operation of neurotransmitters. Roncagliolo et al. (1998) report direct evidence of its adverse effect on brain development in human infants.

Children with iron deficiency anemia during infancy have poorer scores on measures of behavior and development (Nokes et al., 1998). Of particular importance, the effects of early deficiencies extend well beyond early childhood. Even a full course of iron treatment does not appear to reverse the impact on mental or motor test scores or remediate behavior differences in most infants (Nokes et al., 1998), early school-age children (Lozoff et al.,

1991), or adolescents (Lozoff et al., 1997). The persistent consequences of iron deficiency long after it has been eliminated are not yet fully understood (Lozoff et al., 2000). One plausible explanation is that iron deficiency is correlated with other parent and home characteristics that affect development. Research by Lozoff et al. (2000) controlling for an array of such characteristics continued to find a substantial effect more than 10 years after iron deficiency therapy. In a longitudinal sample of 191 children who had been tested for iron deficiency during infancy and treated if found deficient, the outcomes on a variety of behavioral dimensions (see Figure 3-5 and on cognitive dimensions (see Figure 3-6) continued to differ for the 48 children who had chronic, severe iron deficiency in infancy. A greater proportion of the iron-deficient group had repeated a grade (26 vs. 12 percent, p = .04), and more of the iron-deficient group had been referred for special education services or tutoring (21 vs. 7 percent; p = .02), although at the time of the study there was no significant difference in the proportion receiving such services.

There are marked differences in the incidence of iron deficiency among racial/ethnic groups in the United States (Ogden, 1998). While iron deficiency among infants has been on the decline due to iron-fortified formula and baby cereal as well as to an increase in breast-feeding (McLoyd and Lozoff, 2001), the rate of decline has been substantially greater for whites than for blacks and Hispanics. About 5 percent of poor black and Mexican American children still suffer from iron deficiency anemia, about twice the rate for whites, and iron deficiency with or without anemia affects many more children in all racial/ethnic groups. As Figure 3-7 indicates, income is correlated with iron deficiency for all race groups. Since larger percentages of the minority groups fall below 185 percent of poverty, however, the proportions of minority children with iron deficiency are considerably higher than that of whites.

While the association of iron with variation in cognition and behavior appears to be pronounced, other nutritional influences on performance, particularly from vitamin supplementation, have been claimed as well. Eysenck and Schoenthaler (1997) provide a careful review of this literature, as do Sigman and Whaley (1998). Several conclusions can be drawn from their work that are highly relevant to our present concern:

1. Inadequate levels of vitamins and minerals in the bloodstream reduce a child’s IQ, and supplementation of the child’s standard diet can raise nonverbal IQ significantly.

2. A consistent effect of supplementation on young infants is on motor skills (Pollitt et al., 1994). Infant motor skills are predictive of later cognitive abilities among children in developing countries (Sigman and Whaley, 1998).

3. The younger the child, the greater the effects of supplementation. There is little effect beyond the teenage years.

4. Approximately 20 percent of children in the United States respond to supplementation with IQ increases of 9+ points over test-retest increases in a placebo group. However, no effects are found for children with adequate levels of vitamins and minerals in their diets. The concentration of effects is likely to be greatest among disadvantaged children.

5. Effects of micronutrient supplementation have been demonstrated to continue for one year and may last longer.

One provocative natural experiment of the effect of dietary changes on academic performance took place in New York City public schools in the late 1970s and early 1980s. Schoenthaler et al. (1986a, b) analyzed the results of dietary modifications in the foods supplied to the schools. In school years 1979-1980, 1980-1981, and 1982-1983 there was a gradual elimination of synthetic colors, synthetic flavors, and selected preservatives. High-sucrose foods were gradually eliminated. When Schoenthaler and colleagues compared the student percentile rankings on the California Achievement Test, the results were striking (see Figure 3-8). The average ranking in the 41st percentile in the three years before the changes rose to 47th, 51st, and 55th in each of the three change years. In 1981-1982, when no new changes were introduced, the scores remained flat. Gains were largest for students doing worst academically. In 1979, 12.4 percent of students were performing two or more grades below level. At the end of 1983, that rate had dropped to 4.9 percent. While the precise nutritional change was not measured in this study, the authors argue that the foods eliminated tend to be low in the ratio of essential nutrients to calories, thus increasing the proportion of available foods with a higher ratio of nutrients to calories (Eysenck and Schoenthaler, 1997). The claim, however plausible, was not tested.

Lead, a common element in the earth’s crust, becomes harmful to humans only when it is bioavailable: that is, when it is ingested in paint chips or dust that contain lead, taken into the lungs via pollution from leaded gasoline, absorbed through foods that have been stored in lead soldered cans or ceramics (Rice, 1998), or consumed in drinking water that has flowed through lead-soldered pipes (NRC, 1993). Lead is both carried in the bloodstream and stored in bone and soft tissue. The fetal months and early childhood years of rapid bone and tissue growth therefore constitute a particularly vulnerable period for lead exposure.

Childhood lead poisoning was recognized only in the past century, a period that was marked by dramatic shifts in lead exposure. Widespread exposure to lead first rose, particularly with the addition of lead to gaso-

FIGURE 3-8 National ranking of 803 New York City public schools before and after diet changes.

SOURCE: Schoenthaler, Doraz, and Wakefield (1986a). Reprinted with permission.

line, in the 1920s (Elias et al., 1975). The latter half of the century was marked by a sharp decline in exposure as zinc and titanium oxide replaced lead in paint in the 1950s (Needleman, 2000). As consciousness of childhood lead poisoning grew, lead in paint was banned entirely in 1978 and was removed from gasoline in 1986. Blood lead levels responded. The average for young children in the United States and other industrialized countries has decreased dramatically from 15 mg/µ in the late 1970s to 4 mg/µ or less² currently (Rice, 1998).

The average decline in lead load in the last few decades, however, has not been shared evenly. From a 1991-1994 survey by the CDC of children ages 1-5, the U.S. Department of Health and Human Services estimated that about 4.4 percent of children in that age group had harmful levels of lead in their blood. However, more than 8 percent of children who participated in federal health care programs for low-income and uninsured families, including Medicaid, the Health Center Program,³ and the Special

Supplemental Nutrition Program for Women, Infants, and Children (WIC), had harmful lead levels (U.S. General Accounting Office, 1999).

Children in inner-city neighborhoods with older housing stocks tend to have higher lead exposure levels. And while children from all income and racial/ethnic groups live in houses built before the 1950s when lead in paint was common, children living in older, poorer, inner-city neighborhoods where maintenance of the housing stock is more limited are more likely to be exposed to lead from deteriorating paint (Centers for Disease Control and Prevention, 2000). The level of lead exposure is substantially higher for blacks than for whites, but in both race groups there is a dramatically higher incidence among children from low-income families: more than twice the incidence among low-income whites, and almost five times the incidence among low-income blacks (see Table 3-4). Mexican Americans have substantially higher incidence than do whites, but a rate that is approximately one-third that for blacks (Table 3-5).

At the same time that federal protections were reducing lead exposure, epidemiological research in this country and abroad was pointing to adverse effects from lead exposure at ever lower levels. Until the early 1970s, the acceptable concentration of blood lead in the United States was 60 mg/µ in children and 80 mg/µ in adults (NRC, 1993). Acceptable concentrations were lowered several times, until in 1990 the Science Advisory Board of the U.S. Environmental Protection Agency identified a blood lead concentration of 10 mg/µ as the maximum safe level for young children. CDC lowered its guideline to the same level, and the National Research Council concurred with the selection of 10 mg/µ as the concentration of concern in children in 1993 (NRC, 1993).

Lead levels at or above 10 mg/µ have been associated with a variety of adverse effects in infants, children, and pregnant women. We focus here on those associated with school performance. Research findings regarding the effect of lead on IQ have been somewhat controversial (Ernhart et al., 1993; Needleman, 1993). Most, though not all, studies find such an effect. Meta-analyses of both cross-sectional and longitudinal studies of lead on IQ conclude that there is a decline of 2-3 points when blood lead rises from 10 to 20 mg/µ (Rice, 1998). Perhaps more important for our purposes, numerous studies point to a relationship between lead and a variety of behaviors closely related to school success and the probability of being referred for special or gifted education, including impairment of attentional processes, impulsivity and hyperactivity, difficulty in changing response strategy, problems in social adjustment, and poor school performance more generally (Rice, 1998).

In a study of 2,000 1st and 2nd grade children in Boston, for example, teachers’ ratings of children on measures of distractibility, lack of persistence, dependence, impulsivity, and ability to follow instructions rose in a dose-dependent fashion with the lead levels measured in the children’s deciduous teeth (Needleman et al., 1979). Separate studies using measures of blood lead (Yule et al., 1984) and hair lead (Tuthill, 1996) concentrations on these same behaviors found similar dose-dependent responses. Graphic display of the striking results of these three studies appear in Figure 3-9. Other studies from New Zealand (Fergusson et al., 1988c; Silva et al., 1988), Mexico (Munoz et al., 1993), Yugoslavia (Wasserman et al., 1995), and the United States (Leviton et al., 1993) found similar adverse effects on behaviors related to social and academic success in the classroom. Several

additional studies found an effect of lead on classroom behavior measurement scales (Yule et al., 1981; Yule and Lansdown, 1981) and on measures of internalizing behavior (like anxiety or withdrawal) and externalizing behavior (aggression, overreaction) (Sciarillo et al., 1992; Needleman et al., 1996; Needleman, 2000; Bellinger et al., 1994b; Rice, 1998).

Reaction time and flexible use of strategies—both characteristics associated with high achievement in school—were tested in several studies. Needleman et al. (1979) found longer reaction times in a simple task for children with higher dentine lead levels. These findings were replicated in a study done in London (Hunter et al., 1985) using blood lead concentration levels. Results of the two studies were later combined (blood lead levels were known for many of the children in the Needleman et al. study), showing an orderly dose-effect relationship between blood lead and reaction time. These findings were replicated in studies of Greek children (Hatzakis et al., 1987), German children (Winneke et al., 1983; Winneke and Kraemer, 1984), and a cohort of 1879 multiethnic European children (Winneke et al., 1990).

Two studies of strategy use employed the Wisconsin Car Sorting Test, a test of abstract thinking, sustained attention, and ability to change response strategy as needed. Students with higher blood lead levels performed more poorly at age 10 (Stiles and Bellinger, 1993), perseverating in an old strategy even when a new one was required. A cohort of 79 19- and 20-year-olds showed an ability to select and respond to critical information and to shift focus adaptively that declined with increases in dentine lead levels (Bellinger et al., 1994a). These findings are consistent with those from a robust body of experimental research on animals exposed to low body burdens of lead found frequently in children (Winneke et al., 1977; Carson et al., 1974; Rice, 1998).

Several studies have looked at measures of students’ school achievement directly. A study in Scotland (Fulton et al., 1987) found lead-related deficits in numeracy and literacy skills. The New Zealand study by Fergusson et al. (1988a, b, c) found deficits in reading, math, spelling, and handwriting. Similarly, Yule found deficits in school performance, including spelling and reading, and Leviton et al. (1993) found those deficits in girls but not boys in a Boston study.

Other measures of school outcome have been studied as well. A study in Denmark (Lyngbye et al., 1990) found an increased need for special education among 1st graders as a function of increased lead levels. Bellinger et al. (1984) found that the need for remedial education and the incidence of grade retention by 6th grade were associated with dentine lead levels of students measured in 1st grade. And a follow-up investigation of children studied by Needleman in 1976 found in young adulthood a dose-dependent

increase in reading disability (Needleman et al., 1990) and failure to complete high school associated with lead level.

When a child is identified with an elevated lead level, any treatment that does not eliminate the exposure is inadequate (Etzel and Balk, eds., 1999). Current federal policy requires that all state Medicaid programs cover a one-time environmental investigation to determine the lead source and necessary case management services.⁴ But less than half of state Medicaid agencies reported covering these services in 1999 (Centers for Disease Control and Prevention, 2000).

If the first principle of intervention is to identify the source and limit exposure, it stands to reason that this effort should be undertaken before the child is initially exposed if the likely source of exposure can be targeted effectively. Because effective efforts to remove lead from paint, gasoline, drinking water, and food cans have largely eliminated new sources of toxic lead, substantial inroads into reducing the number of children with high lead levels will require limiting exposure to existing lead paint from the older housing stock, particularly in low-income neighborhoods. Lead abatement has been supported by several federal task forces (U.S. President’s Task Force, 2000; Centers for Disease Control and Prevention, 1991, 2000).

In 1991, CDC argued in favor of lead abatement and estimated the cost of the effort at $32 billion—about half of the estimated benefits (at a 3 percent discount rate). In 2000, a President’s Task Force on Environmental Health Risks and Safety Risks to Children again recommended elimination of lead from the housing stock by 2010. The technology for doing so has improved and become less expensive over the past decade. The task force estimated the cost and benefits of both a lead abatement effort and a more modest effort at interim control of exposure. They concluded that, in the long run, removal of lead through an abatement program is less expensive, although the stability of that result depends on the discount rate. While the cost of abatement was estimated at a total of $20.7 billion compared with $2.3 billion for interim controls, the net quantifiable benefits of abatement at discount rates at or near 3 percent were substantially larger.⁵

In the United States, racial/ethnic identification and poverty status are closely tied. While this is true for adults, it is even more so for children (see Figure 3-10). Decades of data collection and analysis have firmly established the strength and consistency of associations between socioeconomic status and cognitive, educational, emotional, occupational, and health outcomes (NRC, 2001b; Duncan and Brooks-Gunn, 1997a; Blank, 1994; Keating and Hertzman, 1999; Gottfried, 1984; Neisser et al., 1996; Stipek and Ryan, 1997).

A quarter of a century ago, a study by Broman et al. (1975) looked at the effects of 169 biomedical and behavioral variables during infancy on intellectual performance at age 4 in a sample of 26,760 children. Only 11 of the variables were social or family behavioral factors, but two of these— socioeconomic status (SES) and mother’s education—were the most predictive of all the variables (Sameroff, 1993). The relationship between family socioeconomic status and school failure and behavior problems in children appears in other countries as well, including Britain, Finland, and Sweden (Pagani et al., 1997), although the gradients are not always as steep as in the United States (Case et al., 1999).

More recent research has taken a more refined look at poverty status, including the severity, duration, and timing of poverty (Brooks-Gunn and

FIGURE 3-10 Children under 18 living in poverty: 1980-1997.

SOURCE: U.S. Bureau of the Census (1999b).

Duncan, 1997; Brooks-Gunn et al., 1999; Smith et al., 1997; Duncan and Brooks-Gunn, 1997a). A study by Smith et al. (1997) found that in two very different samples, the effect of poverty on cognitive ability (as measured by IQ, verbal ability, and achievement tests) varied dramatically depending on the severity of poverty. This has direct implications for minority children, since black children are four times as likely, and Hispanic children three times as likely as white children to live in families with income under 50 percent of the poverty threshold (see Table 3-6). A change of one unit in the family income-to-needs ratio in the Smith et al. study was associated with a 3.0 to 3.7 point increase in the child’s score on the various cognitive assessments. A study by Brooks-Gunn et al. (1999) found similarly striking results. Graphs of income-to-needs ratios plotted against standardized IQ scores and Peabody Individual Achievement Test (PIAT) math scores appear in Figures 3-11 and 3-12. The math scores indicate that the depth of poverty (or the level of affluence) matters, and the results for ages 7-8 compared with ages 8-9 suggest that the magnitude of the effect increases over time. The stronger effect of poverty on cognitive scores with age is found by Smith et al. (1997) as well.

Not surprisingly, duration of poverty matters as well. The study by Smith et al. (1997) found that children who lived in persistently poor families scored on average 6-9 points lower on cognitive assessments, while those whose poverty was transient scored 4-5 points lower.

Does the timing of poverty matter? A review of the effect of the timing of poverty on child outcomes suggests that the income gradient is operating throughout the first two decades of life (Duncan and Brooks-Gunn, 1997b). But the effects of income on cognitive performance and school achievement appear to be particularly strong in the early years (Brooks-Gunn et al., 1999). In a study of the effects of poverty on completed schooling, much more powerful effects of income between birth and age 5 were found than at other points in childhood (Axinn et al., 1997). Since poverty is negatively correlated with school readiness on a variety of dimensions (National Center for Education Statistics, 2000, 2001), and low readiness is associated with grade failure, school disengagement, and school dropout (Barnett, 1995; Brooks-Gunn et al., 1993; Guo et al., 1996; Ramey and Ramey, 1994; Schweinhart and Weikart, 1997), this finding is not surprising.

That socioeconomic status—particularly income and mother’s education—matters is beyond dispute. By itself, however, it tells us very little. More recent research has focused on understanding the ways in which poverty status and these outcomes may be linked (Sameroff, 2000; Duncan and Brooks Gunn, 1997a; Ramey et al., 1998).

Why are child outcomes worse in families with low SES? An answer to this question requires more than the establishment of correlations; it requires an understanding of the supports for child development and the ways in which these supports are compromised in low-SES family circumstances.

Two recent NRC reports synthesized research on the development of young children. From Neurons to Neighborhoods: The Science of Early Childhood Development (NRC, 2000c) focuses on the period from birth to kindergarten entry, and Eager to Learn: Educating Our Preschoolers (NRC, 2001b) focuses on children ages 2-5. Both volumes emphasize the interconnectivity of cognitive, motor, and social-emotional development. And

FIGURE 3-11 Income-to-needs ratios and child cognitive ability: Deep poverty and IQ scores, age 5, IHDP data set.

SOURCE: Brooks-Gunn, Duncan, and Britto (1999). Reprinted with permission of The Guilford Press, NY.

both argue that despite the enormous complexity of early development, one thing is abundantly clear: the weight of successful development in the early years falls most heavily on the child’s relationships with primary adult caregivers.

Children, themselves tremendously diverse in the individual characteristics they bring into the world, develop in family and community contexts that vary widely. The committees that produced these reports were largely in agreement that despite this diversity, all children appear to require certain things from early abiding relationships in order to flourish:

1. a reliable, supporting relationship that establishes a sense of security and safety,

2. an affectionate relationship that supports the development of self-esteem,

3. responsiveness of the adult to the child that strengthens the child’s sense of self-efficacy, and

4. support for the growth of new capabilities that are within the child’s reach, including reciprocal interactions that promote language development and the ability to resolve conflicts cooperatively and respectfully.

FIGURE 3-12 Income-to-needs and child cognitive ability: Deep poverty and math ability (PIAT-Math), NLSY-CS data set.

SOURCE: Brooks-Gunn, Duncan, and Britto (1999). Reprinted with permission of The Guilford Press, NY.

“In these ways, relationships shape the development of self-awareness, social competence, conscience, emotional growth and emotion regulation, learning and cognitive growth, and a variety of other foundational developmental accomplishments” (NRC, 2000c:265).

Each family and child has particular supports and stressors—from within the family and without—that affect the quality and quantity of the interactions among family members that are so critical to development. Poverty and maternal education can affect these supports in a number of ways, including maternal depression, differential knowledge and beliefs that shape parent-child interactions, resources available to access quality child care and other educational materials and resources, and exposure to stressful events. While the reciprocal interactions between the child and parent are the “engines that actually drive the outcome,” parental knowledge and other resources influence the effectiveness of the process (Ceci et al., 1997). Moreover, poverty is highly correlated with single-parent status, decreasing the parental attention available to the child.

Numerous studies in the 1960s detected a strong association between the quality of a child’s home environment—indexed by dimensions such as responsivity and sensitivity of the mother to her child, the amount and level of language stimulation, direct teaching and parenting styles—and children’s intellectual and problem-solving competencies (Hunt, 1961; Vygotsky, 1962; Hess and Shipman, 1965; Bee et al., 1969). Over the next four decades, hundreds of additional studies have affirmed this strong association (reviews by Maccoby and Martin, 1983; Huston et al., 1994; Cowan et al., 1994). When Bee and her associates investigated early predictors of IQ and language development, they found that mother-infant interaction was one of the best predictors at every age tested, as good as actual child performance (Bee et al., 1982).

While the association between parenting style and cognitive development has been confirmed in a substantial body of literature, the shared genetic endowment of parents and children is the competing explanation for that association (Scarr, 1997). One study (Riksen-Walraven, 1978) of 100 Dutch mothers’ interactions with their 9-month-old babies using a highly unusual experimental design did find that different styles of parenting cause differential cognitive development in children measured by exploratory behavior and speed of learning in a contingency task (see Box 3-2).

Poverty, especially persistent poverty, is strongly correlated with less optimal home environments (Garrett et al., 1994). Some studies have attributed as much as half of the gap in achievement test scores in preschool-age children and a third of the gap in school-age children to differences in the home learning environments of high-income and low-income children (Smith et al., 1997).

The effects of poverty on the home environment may be manifested in parenting practices. Findings from a large number of longitudinal studies accord in demonstrating strong and negative effects of social and economic hardships on parenting practices in the families of young children. In a study of young boys in grades K-2, Bank et al. (1993) found that social disadvantage predicted harsh parental discipline, which in turn predicted aggressive child behavior. In another study by Conger et al. (1997), harsh parenting and parental financial conflict mediated the relationship between both marital instability and poverty on child behavior and academic problems. A study by Repetti and Wood (1997) found that on days when mothers experienced increased stress at work, they responded by being more irritable and withdrawn in interactions with their children.

Poor parenting practices and a negative home learning environment may result in conduct problems. Studies have reported that 7 to 25 percent

of preschool children meet the diagnostic criteria for what is called oppositional defiant disorder, with the highest rates found in low-income welfare families (Offord et al., 1986, 1987). During the preschool years, powerful antecedents of emotional and behavior problems are found in the interaction of children, their siblings and peers, and their parents in the home setting. In particular, coercive, irritable, and ineffective discipline and other parenting behaviors have been consistently implicated in the development of conduct problems throughout childhood (Patterson et al., 1992; Reid and Eddy, 1997). There is also abundant and consistent evidence that the early development of conduct problems is strongly predictive of behavioral problems in kindergarten, elementary school, and beyond (Patterson et al., 1992; Reid and Eddy, 1997; Ensminger et al., 1983; Goldstein et al., 1980; Walker et al., 1987).

Language development in the early years is particularly important to later school success in reading and acquiring content knowledge (Snow and Paez, in press). The best single predictor of reading success is vocabulary size (Anderson and Nagy, 1992). Substantial differences between the vocabulary size of children in low-income families and those in middle-class

families have been well documented, as has the connection between the vocabulary of the child and the vocabulary used by the parent (NRC, 1998; Hart and Risley, 1992, 1995; Davidson, 1993). Higher-SES mothers have been found to talk to children more, sustain conversation longer, respond in a more contingent fashion to their children’s speech, and elicit more response from the child (Hoff-Ginsberg and Tardif, 1995; Hart and Risley, 1995; Hoff-Ginsberg, 1991).

Other research has reported an association between parents’ income and education level and their interactions with their children in ways that are relevant to mainstream schooling as well, including prompting infants to respond to books and pictures, and asking questions that require labeling and organizing knowledge into categories (Schieffelin and Ochs, 1983). They are also more likely to provide access to materials, time, and adult support for exploratory play that the child is encouraged to initiate (Bradley et al., 1994). Garrett et al. (1994) found that as the income-to-needs ratio rose, so did the quality of the home environment.

The National Center for Education Statistics is collecting longitudinal data on a nationally representative sample of children as they enter kindergarten and following them through 5th grade. The survey, called the Early Childhood Longitudinal Study (ECLS) also collects data on characteristics of the child’s family and home environment. Among families with more than 100 children’s books in the home, whites were represented at five times the rate of blacks or Hispanics, and among those with fewest books the proportions reversed. Asians were similar to other minority groups. Welfare status and a primary language other than English in the home were also associated with having fewer books and recordings (U.S. Department of Education, 1998).

An estimated 1 in 10 women with young children experiences depression (Dickstein et al., 1998; Gelfand et al., 1996). Estimated rates for mothers living in poverty, however, range from 13 to 28 percent (Danziger et al., 2000; Lennon et al., 1998; Moore et al., 1995; Olson and Pavetti, 1996). In two large samples of poor women in work and training programs, over 40 percent were found to have clinically significant depressive symptoms (Quint et al., 1997; U.S. Department of Health and Human Services, 1995). The higher prevalence is postulated to arise from the stress and loss of control that accompany persistent economic pressures (Brody et al., 1995; Brody and Flor, 1997; Caplovitz, 1979; Conger et al., 1992; Dressler, 1985; Kessler et al., 1987; McLoyd et al., 1994). While severe income constraints can serve as a catalyst to depression, it should be noted that

most of the mothers in poor families are not depressed (Edin and Lein, 1997; Brody and Flor, 1997; NRC, 2000a).

Maternal depression has been consistently implicated in reducing the quality of parenting and disruptions in the emotional relationship between parent and child (NRC, 2000a). Particularly relevant to the development of children’s emotional and behavioral problems, depressed mothers are less likely to be consistent with their children (McLoyd, 1997). They are more likely to withdraw and to respond with less emotion and energy and, when they do engage, they are more likely to do so in an intrusive or hostile manner (Brody and Forehand, 1986; Brody et al., 1994; Frankel and Harmon, 1996; Patterson, 1986; Tronick and Weinberg, 1997; Zeanah et al., 1997). Infants of depressed mothers are more likely to withdraw as well and show reduced levels of activity and dysphoria (Cummings and Davies, 1994, 1999; Dawson et al., 1992; Frankel and Harmon, 1996; Murray and Cooper, 1997; Seifer et al., 1996; van Ijzendoorn et al., 1992).

In a study by Bettes (1988), maternal depression was associated with linguistic as well as emotional development; 10 of 36 mothers studied were rated as depressed. Tape recordings indicated that when babies cooed, the nondepressed mothers quickly responded, whereas depressed mothers had a greater latency and their vocal patterns were not tied to their children’s vocal output. At 3 to 4 months, there were no differences in the vocalization patterns of the infants, but after 6 to 9 months, the babies of depressed mothers vocalized much less.

As an isolated risk factor, maternal depression may have relatively little impact on development (Rutter, 1979; Cummings and Davies, 1994; Seifer et al., 1996; Zeanah et al., 1997). But since prevalence rates are much higher for mothers living in poverty, depression is often combined with other risks. As a group, children with depressed mothers are at higher risk of emotional and behavior problems, and these in turn are associated with difficulties in school, aggression, poor peer relationships, and reduced ability to exercise self-control (Campbell et al., 1995; Cummings and Davies, 1994; Dawson and Ashman, 2000; Zeanah et al., 1997). These children also have higher incidence of psychopathology themselves (Cummings and Davies, 1994; Downey and Coyne, 1990; Zeanah et al., 1997).

Because young children are far more likely to spend a significant amount of time in child care today than at any time in the past, a great deal of attention has been devoted in recent years to understanding the consequences of that care. The two NRC reports mentioned above review extensive literatures in this regard (NRC, 2000c, 2001b). The conclusions of relevance for our purposes are rather obvious: the consequences of child

care depend largely on the quality of that care, and the characteristics of quality in child care are the same as those in home care. At its core, the quality of child care depends on the quality of the interactions between the caregiver and the child. Characteristics of those interactions that benefit the child—security, affection, responsiveness, and support for emerging abilities—are the same as those with a parent.

And as in relationships with parents, secure attachments to child care providers are associated with adaptive social development (Howes et al., 1992; Oppenheim et al., 1988; Peisner-Feinberg et al., 2000; Pianta and Nimetz, 1991; Sroufe et al., 1983), more competent interactions with adults, and more sophisticated play with peers (Howes and Smith, 1995; Howes et al., 1998, 1994), effects that last into the school years (Howes, 2000). And as with the home environment, quality interactions in child care have been positively associated with cognitive and linguistic development (Burchinal et al., 1996; Galinsky et al., 1994; National Institute of Child Health and Human Development, 1999; Peisner-Feinberg and Burchinal, 1997; Peisner-Feinberg et al., 2000).

Central to determining the quality of child care are the characteristics of the caregivers: their education, early childhood training, and attitudes about their job and the children in their charge. And the ability to carry out their work well is positively influenced by small child-adult ratios and small group size (NRC, 2000c, 2001b). Clearly, creating quality in a child care program is directly related to program cost.

Efforts to assess child care quality in the United States have concluded that from 10 to 20 percent of arrangements fall below minimal standards of adequacy (Cost, Quality, and Outcomes Study Team, 1995; Galinsky et al., 1994; Helburn, 1995; Whitebook et al., 1990). These settings are characterized by “caregivers who more often ignore than respond to young children’s bids for attention and affection, a dearth of age-appropriate or educational toys, and children who spend much of their time wandering aimlessly around, unengaged with adults, other children, or materials” (NRC, 2000c:320). At the other extreme, fewer than 20 percent of toddlers and preschoolers were in settings considered to be of high quality (National Institute of Child Health and Human Development Early Child Care Research Network, 1996).

The gains from quality child care are often greatest for children from low-SES families (Peisner-Feinberg and Burchinal, 1997), but the higher cost of quality care means access for these families is restricted without government subsidy or provision of services. In the private marketplace, children from poorer, more stressed homes receive lower-quality care than other children (Howes and Olenick, 1986; National Institute of Child Health and Human Development, 1997b; Phillips et al., 1994). Families with low incomes spend a substantially higher proportion of that income

on child care but are nonetheless priced out of higher cost forms of care in many areas of the country (U.S. Department of Health and Human Services, 1999; Giannarelli and Barsimantov, 2000).

There is an exception to the rule that quality of child care is directly related to income level. Poor families who receive subsidies for child care or access to early intervention programs like Head Start often receive care that is of higher quality than that obtainable by low-income families that are not eligible for these supports. Head Start programs (discussed below) are characterized by a relatively compressed range of quality. While there are few very high-quality Head Start centers, none is characterized by the substandard care found in the private child care market (Administration of Children, Youth, and Families, 2001).

As with other risk factors, it is important to maintain a broad perspective: poor-quality child care is not deterministic. A strong attachment relationship with a parent, and the benefits that ensue, appear in large measure to protect children from the negative effects of poor-quality child care (National Institute of Child Health and Human Development, 1997a; Roggman et al., 1994; Symons, 1998). However, the limited financial and human capital resources that are predictive of low-quality care often are accompanied by other stressors. Several studies have found that when young children are exposed to risk factors at home and are in poor-quality child care, they are also more likely to experience insensitive mothering (Belsky et al., 1996; Clarke et al., 1997; Tresch et al., 1988). It is worth noting as well that mothers in the National Institute of Child Health and Human Development study living at or near the poverty line whose children were in full-time, high-quality child care were more responsive and affectionate with their infants than low-income mothers raising their children at home or in lower-quality care (National Institute of Child Health and Human Development, 1997c). And other studies have found child care to be a protective factor for infants and children living in poverty (Caughy et al., 1994) or with depressed mothers (Cohn et al., 1986, 1991).

Since Bronfenbrenner published his influential article proposing an “ecological” model of development in 1979, substantial empirical research has examined the effect of a combination of risk factors that together determine a child’s experience. Many of the factors described above had been shown to have a significant impact on development, but individually any one factor could explain only a small portion of the variation in outcome. While poverty or low birthweight has a measurable impact on average, clearly some children with those characteristics do well. The very notion of risk suggests an uncertain or probabilistic outcome.

In the same year that Bronfenbrenner published his article, Michael Rutter looked at risk factors that helped explain child psychiatric disorders. He included in his list of variables severe marital discord, low social status, overcrowding or large family size, paternal criminality, maternal psychiatric disorder, and admission into the care of the local authority (Rutter, 1979). At the time, Rutter described his results as “interesting and surprising.” Children with any single risk factor were no more likely to have a psychiatric disorder than were children with no risk factors (see Figure 3-13 ). But when any two stresses occurred together, the risk went up fourfold, and with four stresses, tenfold.

The past two decades have witnessed research efforts to replicate and extend the multiple risk model to look at a wider array of both risks and outcomes. The results are no longer surprising; it is now quite widely accepted that the number of risk factors that children face is more important than the impact of any single factor (Sameroff, 2000; Williams et al., 1990; Fergusson et al., 1994). Indeed, it has been argued that the challenge posed by adversity may in fact be a necessary condition for life’s achievements (Bandura, 1997; Lewis, 1997; Sameroff, 2000). Yet as the number of stresses increases, the chance for a positive outcome drops off precipitously. This can be seen quite dramatically in the Rochester Longitudinal Study of a group of children from the prenatal period through age 18 from a socially heterogeneous set of families (Sameroff, 2000).

The study measured the impact of risk factors at age 4 on both cognitive outcome (measured by the Weschler Preschool and Primary Scale of

FIGURE 3-13 Multiplicity of risk factors and child psychiatric disorder.

SOURCE: Rutter (1979). Reprinted with permission.

Intelligence verbal intelligence score) and mental health outcomes (measured by the Rochester Adaptive Behavior Inventory). The risk factors considered were maternal mental illness; high maternal anxiety; rigidity in the attitudes, beliefs, and values of mothers regarding the child’s development; few positive maternal interactions with the child during infancy; less maternal education than high school; head of household in an unskilled occupation; disadvantaged minority status; single parenthood; stressful life events; and large family size. While each variable had a statistically significant negative impact by itself, no single variable was able to predict much of the variation. The total number of risk factors, however, was a powerful predictor. On the intelligence test, children with no environmental risks scored more than 30 points higher than children with eight or nine risk factors. No preschoolers in the zero-risk category had an IQ below 85, but 26 percent of those in the high-risk group did. And 4-year-olds with five or more risk factors were 12.3 times as likely to have clinical mental health symptoms as those with fewer risks (Sameroff, 2000).

Child development theory in recent years has incorporated the notion that children not only react to their environment, but also help create it as their behavior elicits responses from those around them (NRC, 2000c, 2001b). In an effort to determine the role played by the characteristics a child brings—including temperament, perinatal physical condition, interactive behaviors, and competence in motor behaviors and regulatory abilities—children were assessed during the first 12 months on a variety of development and behavior scales.⁶ The children were assessed again at age 4 on both social-emotional competence (mental health) and on IQ. Infant competence scores were rendered insignificant compared with environmental risk. “High competent infants in high risk environments did worse as 4-year-olds than low competent infants in low risk environments . . . individual characteristics were not able to overcome the effects of environmental adversity. If one wants to predict the developmental course for infants, attention to the accumulation of environmental risk factors would be the best strategy” (Sameroff, 2000:26-27) (see Figure 3-14).

Data from the National Center for Education Statistics on children entering kindergarten demonstrate how striking are the accumulated differences in knowledge and skill development across SES groups by the time

children reach the schoolhouse door. The survey collects data on emergent literacy and numeracy skills and content knowledge. It also collects teacher and parent ratings on children’s social skills (National Center for Education Statistics, 2000).

Table 3-7 displays differences by family and child characteristics in the skills that have been established to be prerequisites to learning to read: knowing that print reads left to right, knowing where to go when a line of print ends, and knowing where a story ends. Without a regression analysis, the independent effects of poverty, race, maternal education, marital status, and a primary language other than English cannot be disentangled. The simple correlations, however, are pronounced for each characteristic. At the extremes, 47 percent of white children with a mother who graduated from high school had all three skills, while only 11 percent of black children with mothers who did not graduate from high school had all three. The same pattern can be found for prereading skill level in letter recognition, beginning and ending sound identification, and identifying words by sight or in context and for early mathematics skills, including number and shape recognition, relative size comparison, ordinal sequencing, the ability to add, subtract, multiply, or divide small numbers.

Finally, social and emotional skills differ by SES as well. While these skills are of value in and of themselves, for the purposes of this report their relationship to later academic achievement and behavior is noteworthy (Swartz and Walker, 1984). While the Early Childhood Longitudinal Study collects data on a variety of measures, we focus here on self-regulatory and motivation characteristics and problem behaviors as rated by the teacher. Teacher ratings may incorporate bias (discussed in Chapter 5), but these are the ratings that are likely to influence special education placement.

Teachers see differences between boys and girls in the ability to attend, with only 58 percent of boys rated as being able to attend often, compared with 74 percent of girls. They also rate white and Asian children as better able to attend and as more persistent than black and Hispanic children. Children’s ratings on all attributes rise with parents’ education levels, and children in two-parent families are rated higher on average than those in one-parent families (see Table 3-8).

With respect to problem behaviors, the number of children who argue or fight with others is relatively small; most children can get along in the classroom. However, the differences by race are substantial, with Asians rated as exhibiting few problem behaviors and black children exhibiting the highest rate (see Table 3-9). Hispanic and white children receive similar ratings.

FIGURE 3-14 Relation of infant competence to competence and IQ scores at 4 years of age, controlling for early multiple environmental risk scores.

SOURCE: Sameroff (2000). Reprinted with permission of John Wiley & Sons.

The early disparities in school readiness are of marked significance because they portend future achievement gaps. While gaps in the particular skills of letter and number identification close during the kindergarten year,

children who enter kindergarten with those skills already in place have made strides in other areas that take them beyond early skill development. Figure 3-15 shows the gains in reading scores over the course of the kindergarten year by maternal education level. While all children gained significantly over the course of the year, the gap did not narrow, even though

FIGURE 3-15 First-time kindergartners’ reading mean scale scores, by mother’s education: Fall 1998 and Spring 1999.

SOURCE: U.S. Department of Education, National Center for Education Statistics, Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999.

almost all children had acquired letter recognition and print awareness skills (West et al., 2001). As children move through the school years, those who read well read more (Stanovich, 1986) and therefore acquire a larger knowledge base.

A similar pattern occurs in mathematics: children from low-SES groups acquire the same knowledge as those from higher-SES groups, but they acquire it later (West et al., 2001). Griffin et al. (1994) found that low-income 5- to 6-year-olds performed like middle-income 3- to 4-year-olds on a test of early math skills.

The implications of the lag apply not only to special education, but also to gifted education. At the upper end of the achievement distribution in the literacy domain are children who can recognize words by sight or can add and subtract in the spring of the kindergarten year. Figures 3-16 and 3-17 plot the percentage of such children by the number of risk characteristics present, including less maternal education than high school, family receiving welfare or food stamps, single-parent household, and primary language other than English. While about 1 in 5 children in families with none of those risk factors has mastered these skills, the representation of children with two or more risk factors in that category is very low.

Disparities in school readiness are also manifested in the development of peer and student-teacher relationships. We know from research on the development of behavior and emotional problems that young children who already exhibit aggressive, disruptive behaviors when they enter school are

FIGURE 3-16 Percentage of first-time kindergartners recognizing the words by sight, by number of risk factors: Fall 1998 and Spring 1999.

SOURCE: U.S. Department of Education, National Center for Education Statistics, Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999.

often not equipped with the necessary skills to develop healthy peer and adult relationships later on (Goldstein et al., 1980; Patterson, 1986; Patterson et al., 1992; Walker et al., 1987). We also know that aggressive and violent boys differ from less aggressive boys on measures of interpersonal problem solving, with the scores of aggressive and violent boys dem-

FIGURE 3-17 Percentage of first-time kindergartners adding and subtracting, by number of risk factors: Fall 1998 and Spring 1999.

SOURCE: U.S. Department of Education, National Center for Education Statistics, Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999.

onstrating significantly poorer skills (Lochman and Dodge, 1994). This inability to appropriately solve problems, coupled with the use of coercive behaviors, makes it extremely difficult for antisocial students to attend, concentrate, and learn the basic academic skills necessary to function in school. These learning skill deficits, which often develop before school entry, cause students to have trouble moving successfully through the curriculum, because they usually need additional time and assistance to help them achieve mastery (Fuchs et al., 1993; Walker et al., 1995; Gleason et al., 1991).

The weight of the evidence reviewed above suggests that in order to have an education system in which non-Asian minority students (and disadvantaged students more generally) are not represented in disproportionately high numbers among those at the low end of the achievement distribution and in disproportionately low numbers at the high end of that distribution, efforts to support the cognitive, social, and emotional development of those children in the years before they arrive at kindergarten are critical. This is not to say that early experience sets a child on an unalterable course. We know, for example, that some schools do far better than others at promoting achievement among high-risk children (discussed in Chapter 5 ). Yet when children are exposed to many risk factors early on, promoting school success will be a much more difficult task for both the child and the school.