Relations between Infant Irritability and Maternal
in Low_Income Families
Much of the research regarding the relation between difficult infant temperament and maternal responsiveness has been guided by an interactionist view of child development, which postulates that influence between parent and child behaviors is bidirectional (Bell, 1968; Crockenberg, 1986). However, the nature of this mutual relation is not clear. Theoreticians generally agree that two equally acceptable hypotheses exist (Bates, 1980; Thomas, Chess, & Birch, 1968). First, it is possible that these maternal and infant characteristics are mutually influential because intense, frequent, and long bouts of fussing and crying inhibit maternal ability to respond to infant needs, resulting in less responsive parenting and potentially greater infant difficulty. Alternatively, infant difficulty may elicit increased maternal responsiveness, thereby possibly reducing later infant difficulty. These viewpoints are not necessarily antithetical. According to the "goodness_of_fit" perspective, proposed by Thomas, Chess, Birch, Hertzig, and Korn (1963), both of these relations might exist. The nature of the relation may depend on moderating factors within the dyad, such as the infant's age or gender (Crockenberg, 1986). In accord with the ecological view of child development (Bronfenbrenner, 1977), moderating factors external to the dyad, such as level of maternal social support, might also influence the relation between difficulty and responsiveness (Crockenberg, 1986). By integrating and applying the three perspectives (interactional, goodness_of_fit, and ecological), it is plausible that relations between infant difficulty and maternal responsiveness could be negative, positive, or nonexistent depending on factors within or outside of the dyad.
Studies of relations between infant difficulty and maternal responsiveness have produced mixed results. Investigators who have examined this relation in isolation from contextual factors have found infant difficulty to be associated with both negative (Campbell, 1979; Crockenberg & McCluskey, 1986; Lee & Bates, 1985; Milliones, 1978; Spangler, 1990) and positive (Bates, Olson, Pettit, & Bayles, 1982; Crockenberg & Acredolo, 1983; Crockenberg & Smith, 1982) maternal behaviors or have found inconsistent relations (Peter_Martin & Wachs, 1984). Consideration of contextual factors suggests that mixed results are due in part to methodological differences between studies, including measurement strategies, family SES, infant age, level of maternal social support, or infant gender. For example, varied strategies for measuring infant difficulty may account for mixed results. Seifer, Schiller, Sameroff, Resnick, and Riordan (1996) and Coffman, Levitt, Guacci, and Silver (1992) found relations between infant temperament and maternal responsiveness only when temperament was measured observationally and not when it was measured by maternal report. Furthermore, negative relations between difficulty and responsive behavior have been demonstrated in low SES samples (Linn & Horowitz, 1983; Milliones, 1978; Schuler, Black, & Starr, 1995; van den Boom & Hoeksma, 1994), while results from middle_class or mixed samples have varied. Similarly, studies examining older infants 1 year of age or older generally have produced negative relations between difficulty and responsiveness (Coffman et al., 1992; Lee & Bates, 1985; Maccoby, Snow, & Jacklin, 1984; Spangler, 1990), while studies of younger infants have demonstrated positive relations (Bates et al., 1982; Crockenberg & Acredolo, 1983; Crockenberg & Smith, 1982).
Low levels of social support may also influence the relation between maternal responsiveness and infant difficulty by inhibiting positive maternal behaviors (Adamkos, Ryan, Ullman, Pascoe, Diaz, & Chessare, 1986; Crnic, Greenberg, Ragozin, Robinson, & Basham, 1983; Weinraub & Wolf, 1983), or might moderate the relation by interacting with difficulty in the prediction of maternal responsiveness (Crockenberg, 1981; Crockenberg & McCluskey, 1986). In Crockenberg's studies, only mothers of "difficult" infants who experienced low social support showed lower levels of responsivity and sensitivity. Low social support was not related to the behavior of mothers of "easy" infants. Therefore, it seems equally plausible that these contextual factors contribute to the relation between responsiveness and difficulty through their negative influence on maternal responsiveness or that they moderate the relation by differentially influencing mothers of difficult or easy babies.
The possible role of gender in the relation between difficulty and responsiveness is less clear. Gender is one dimension of individual variation through which infants influence maternal behavior (i.e. one type of child effect), and gender_stereotyped attitudes about how girls and boys should behave differentially influence behavior toward them (Maccoby & Jacklin, 1974). However, even if it is assumed that gender stereotypes are widely held, predictions vary about consequent maternal behavior in response to infant difficulty. For example, it could be that mothers increase responsiveness to difficult girls because they see them as more dependent or sensitive than boys, or because they have greater empathy in a same_sex relationship. Mothers may decrease responsiveness to difficult boys because they feel boys should learn to be tough and independent. It may also be that other traits or behaviors that vary by gender, such as sociability or activity, influence maternal behavior toward equally difficult girls and boys (Maccoby et al., 1984). The little empirical literature regarding relations between infant gender and the difficulty_responsiveness relation is mixed (Coffman et al., 1992; Crockenberg & Smith, 1982; Klein, 1984; Maccoby et al., 1984), although it could be inferred from these studies that middle_class mothers are more likely to respond negatively to difficult boys than they are to difficult girls. Neither the theoretical or empirical literature suggests clear hypotheses about gender moderation of the relation between infant difficulty and maternal responsiveness. Therefore, analyses at this point are necessarily hypothesis_generating instead of hypothesis_testing (Bates, 1980).
In sum, evidence regarding relations between infant difficulty and maternal responsiveness is mixed. Despite the plausible role of contextual factors, including infant gender and maternal social support, little is known about whether and how they may exert influence. Presently, the authors aimed to extend the current knowledge base concerning relations between difficulty and responsiveness by having a sufficiently large sample size to conduct analyses by gender, assessing the role of maternal social support, independently measuring infant difficulty and maternal responsiveness, measuring infant difficulty separately through observation and maternal report, and by employing a low_income sample so as to potentially extend the generalizability of findings to an understudied population.
Based on previous research with low_income mothers and their infants during the second year of life, it was hypothesized that the concurrent and longitudinal relations between infant difficulty and maternal responsiveness would be negative (Coffman et al., 1992; Lee & Bates, 1985; Linn & Horowitz, 1983; Maccoby et al., 1984; Milliones, 1978; Schuler et al., 1995; Spangler, 1990). It was also predicted that like Coffman et al. (1992) and Seifer et al. (1996), stronger results would be obtained when infant difficulty was measured through observation versus maternal report. Although gender effects between difficulty and responsiveness were difficult to predict based upon current theoretical and empirical information (Crockenberg, 1986; Crockenberg & Smith, 1982; Klein, 1984; Maccoby & Jacklin, 1974; Maccoby et al., 1984), it seemed likely that the negative relation expected in the total sample would be carried primarily by the boys, so that relations between difficulty and responsiveness would be more negative for boys than for girls. Based primarily on the work of Crnic and colleagues (1983, 1984, 1986), it was predicted that maternal social support would contribute to the difficulty_responsiveness relation through its positive association with concurrently measured maternal responsiveness. Additionally, the possibility that social support moderates the relation between responsiveness and difficulty by creating or exacerbating a negative association was investigated (Crockenberg, 1981, 1986).
As part of two larger studies of development in children from low_income families, participants were recruited from the Women, Infants, and Children (WIC) Nutritional Supplement clinics throughout the Pittsburgh, PA metropolitan area between 1989 and 1992. The WIC program supplies nutritional services and food supplements to low_income women with children under age 6. Data from each study were originally analyzed separately. Because results were generally consistent across studies, data from each sample were then combined to prevent redundancy and increase power.
Mothers of infants approaching 12 months were asked to participate. Two hundred and thirty_five mother_infant pairs participated in the 12_month lab visit; approximately half were mother_son and half were mother_daughter dyads. Of these, 22 dyads were lost to attrition, leaving 213 dyads who participated in the 18_month assessment. Descriptive demographic statistics are presented in Table 1. No significant demographic differences were found when dyads retained versus dyads lost at 18 months were compared. In addition to subject attrition, some data were lost to procedural errors and equipment failure. Valid cases for variables ranged from 202 to 233.
Insert Table 1 about here
Videotaped laboratory assessments were conducted at 12 and 18 months. Mothers were reimbursed after each visit. Assessment tasks were selected to vary in stress level so that mother and infant behavior could be observed across a spectrum of contexts (e.g. free play vs. the Strange Situation). Assessments began with a brief warm_up period during which the infant played with toys and the assessment activities were explained to the mother. The videotaped tasks at 12 months began with a 15_minute free play period in which the infant played with a standardized set of toys while the mother and examiner sat nearby completing questionnaires. The free play was followed by the High Chair task (Martin, 1981) in which the infant was seated in a high chair, facing away from mother, for three minutes. The mother was asked to complete questionnaires but was allowed to care for her infant in whatever way was comfortable for her, without removing him/her from the chair. After three minutes, the mother set the infant on the floor and continued to work on the questionnaires. During the following three_minute No Toys task (Smith & Pederson, 1988) there were no toys in the room and the mother was allowed again to care for her infant in whatever way was comfortable. The assessment activities continued with a series of Teaching tasks (Martin, 1981) in which the mother was asked to help her infant complete four challenging toy tasks, each for three minutes. These interactive tasks were followed by a ten_minute break. The videotaped assessment concluded with the Strange Situation (Ainsworth & Wittig, 1969), a series of seven three_minute separation and reunion episodes involving the infant, the mother, and an unfamiliar examiner.
The 18_month lab assessment was similar to the 12_month assessment. However, after the free play period, a five_minute Clean Up task (Matas, Arend, & Sroufe, 1978) was included in which the mother was asked to get her infant to put the toys into a basket. Additionally, the High Chair task was not administered at 18 months and three instead of four Teaching tasks were administered. Consequently, a slightly different order for the structured tasks was employed, as follows: 1) free play, 2) Clean Up task, 3) No Toys task, 4) three Teaching tasks, 5) break, 6) Strange Situation.
Task order for subjects from the second study varied slightly. For those subjects, the 12_month videotaped lab tasks were ordered as follows: 1) Strange Situation, 2) break, 3) free play, 2) High Chair task, 3) No Toys task, 4) four Teaching tasks. At 18 months the tasks were ordered as follows: 1) free play, 2) Clean Up task, 3) No Toys task, 4) High Chair task, 5) three Teaching tasks, 6) break, 7) Strange Situation., 8) five_minute free play.
Maternal Responsiveness Maternal responsiveness was measured using an adaptation of the No Toys task (Smith & Pederson, 1988) from which videotaped observations of appropriate and contingent maternal behavior in response to infant behavior, or lack thereof, were coded. This measure was intended to assess general level of unresponsiveness rather than a particular style of unresponsiveness. Smith and Pederson (1988) found this task to successfully discriminate attachment classifications of infants of middle_class mothers, and Vondra, Shaw, and Kevenides (1995) found the same in a subset of the subjects used presently. Additionally, scores from this measure showed small to moderate significant positive relations with the Emotional and Verbal Responsivity, Involvement with the Child, and Avoidance of Restriction and Punishment subscales of the HOME Inventory (Caldwell & Bradley, 1984) at 15 and 24 months taken on these same subjects.
During the No Toys task, infants were free to move about the sparsely furnished room without toys to play with. Mothers were asked to complete questionnaires but were told they could care for their infants in whatever way was comfortable for them. Five maternal behaviors (look, vocalize, touch, hold, or none of these behaviors) and five infant behaviors (look, vocalize, touch, fuss/cry, or none of these behaviors) were coded molecularly from videotape on a continual basis throughout the three_minute task. Each tape was viewed separately for maternal behaviors and for infant behaviors. Interrater reliability of behavior codes in the form of percent agreement was calculated on 18 cases for subjects from the first study and ranged from 62% to 100%, with a mean of 83%. For subjects from the second study, interrater reliability ranged from 72% to 91%, with a mean of 81%.
These codes provided the basis for scoring mothers as appropriate, appropriate when required (i.e. in response to infant touching, fussing, crying, or looking and vocalizing), insufficient, or intrusive during each of the three_second intervals. A scoring key which lists all possible combinations of mother and infant behaviors in each maternal behavior category was used. For example, if an infant were to look and vocalize during a three_second interval, mother was required to look, vocalize, or look and vocalize to her infant within six seconds (two intervals) to be scored appropriate. Less (e.g. continuing to work on questionnaires without responding) was scored as insufficient and more (e.g. holding) was scored as intrusive. To ensure responsive behaviors were not scored as intrusive, the category of appropriate when not required existed so that "excessive" behaviors that were actually responsive, rather than truly intrusive, would be scored as appropriate . Mothers received a separate score for each of the four behavior categories. For analyses, scores for appropriate and appropriate when required were combined so that maternal behavior in response to more and less demanding, or even undemanding infant behavior would be included in this measure of maternal responsiveness. A mother whose infant made few bids for attention could receive a high responsiveness score if she too initiated little interaction and therefore matched her infant's behavior. Furthermore, because doing as little as looking at mother was considered an infant bid for attention, very few infants initiated so infrequently that their mothers' tendencies toward responsive behavior could not be assessed.
Infant Irritability Infant difficulty was conceptualized and measured as irritability because it is central to most or all conceptualizations of difficult infant temperament (Bates, 1980; Crockenberg, 1986) and it was measured observationally in the lab and through maternal report. For the observed rating, one molecular and three global ratings of infant irritability were made from approximately 70_minute_long videotapes of infants at both 12 and 18 months. As described above, in these videotapes infants participated in activities that varied in level of stress. This way irritability could be assessed in a variety of situations and the likelihood of detecting meaningful differences between infants could be maximized. For the molecular rating, coders recorded the amount of time the infant spent fussing and crying on a stopwatch as they reviewed each videotape. Percent of time spent fussing and crying was calculated by dividing seconds of fussing and crying by total seconds in which the infant was observable. The same coders made global ratings of the amount and intensity of the infant's fussing and crying based on five_point scales after watching each entire videotape. Coders also made a global rating of infant difficulty on a five_point scale which reflected the observer's judgment of how temperamentally difficult the infant was overall compared to other infants his/her age, based primarily on the amount and intensity of the infant's fussing and crying. Interrater agreement for all components of the observed irritability measure, in the form of weighted kappas, was calculated across 22 randomly selected cases (12 subjects from the first study and 10 from the second) and ranged from .77 to .96, with a mean of .87. Scores for percent of time spent fussing and crying, global amount and global intensity of fussing and crying, and global difficulty were standardized and summed into an observed irritability score for each 12_ and 18_month assessment, with reliability coefficient alphas of .84 and .91, respectively.
Maternal report of infant irritability was obtained using the difficulty factor of the Infant Characteristics Questionnaire (ICQ, Bates, Freeland, & Lounsbury, 1979). As with the observed ratings of irritability, items from the ICQ difficulty factor primarily reflect the frequency, duration, and intensity of the infant's fussing and crying, and were administered at both the 12_ and 18_month visits, producing a score for each age. The difficulty factor has been found to have adequate test_retest reliability and convergence with other parent_reported and observed measures of infant difficulty (Bates et al., 1979; Hubert, Wachs, Peters_Martin, & Gandour, 1982). The measure has been normed on infants 6_, 13_, and 24_months_old, and the questionnaire used varies slightly across age. In this study, the 12_month score was derived from the 13_month questionnaire and norms, and the 18_month score was derived from the 24_month questionnaire and norms.
A multimethod assessment approach for the measurement of temperament traits, such as irritability, has been recommended in which multiple informants are used in multiple settings (Bates, 1980; Crockenberg, 1986). However, analyses were run separately using each measure for two reasons. First, correlations between maternal and observer report obtained in these and in other studies are low (Seifer, Sameroff, Barrett, & Krafchuk, 1994). Presently, the correlation between observed and maternal report of infant irritability at 12 months was .19 (p < .01), and at 18 months was .25 (p < .001). It is becoming increasingly clear that maternal and observer reports of infant temperament provide different, but possibly equally important, information about infants' behavioral styles (Seifer et al., 1996). Second, it is important that ratings of infant traits and maternal behavior be made independently when both are being investigated (Crockenberg, 1986), thus the separate use of the observed irritability measure.
Maternal Social Support Mother's perception of social support was measured using the 19_item General Life Satisfaction Index (GLS, Crnic, Greenberg, Ragozin, Robinson, & Basham, 1983) which assesses support and satisfaction with support in areas of the mother's life including intimate relationships, friendships, and neighborhood and community involvement. Scores representing total amount of support and total satisfaction with support were analyzed separately. Responses on the questionnaire among a sample of mothers of infants have demonstrated substantial concurrent relations to perceived social support as assessed on other self_report measures (Crnic et al., 1983).
Results will be presented in three stages: 1) descriptive data regarding independent and dependent variables, 2) data regarding the relation between infant irritability and maternal responsiveness with emphasis on differences between measurement strategies and gender, 3) data regarding the relation between maternal social support and maternal responsiveness, and the possible moderating role of maternal social support on any irritability _ responsiveness relations. All analyses were originally run separately for subjects from each of the two studies. Because results were generally consistent across those samples, data from all subjects was combined for the analyses presented below.
Descriptive data regarding measures of irritability, responsiveness, and maternal social support are presented in Table 2. Gender differences were examined using a MANOVA procedure, in which gender was the independent variable and measures of irritability, responsiveness, and maternal social support were the dependent variables. Scores by gender are presented in the Table notes only when they are significantly different from each other. There was a nonsignificant trend for girls' and boys' ICQ scores at 18_months to be different (F 1,169 = 3.88, p < .05), with girls being rated as more difficult. No other statistically significant gender differences were found.
Insert Table 2 about here
Levels of variables across time were compared using a series of two_tailed t_tests of means at 12 and 18 months in which the significance level was set at .01 to adjust for the number of t_tests computed (12). In the total sample, the mothers' ratings of infant irritability decreased from 12 to 18 months (t 1, 208 = 14.59, p < .001). However, scores on the components of the observed irritability measure increased over time. Because the observed irritability measure at each age was a cumulation of standardized scores, variation in mean scores across time would not be expected. Instead, mean differences on the one molecular and three global measures comprising the irritability scores were compared. In the total sample, all global measures increased over time (amount fuss/cry: t 1, 182 = 3.34, p < .001; intensity fuss/cry: t 1, 182 = 3.23,p < .001; difficulty: t 1, 182 = 2.7, p < .01). The pattern of changes in boys was the same as that in the total sample, whereas for girls only the molecular score for percent of time spent fussing and crying increased significantly over time. No changes across time were detected in measures of maternal responsiveness or maternal social support in the total sample or when t_tests were computed separately by gender.
Individual rankings of infant irritability were moderately stable from 12 to 18 months. In the total sample, maternal reports of irritability at 12 and 18 months were correlated .52 (p < .001). Twelve_ and 18_month observed ratings of irritability were correlated .27 (p < .01). Scores for maternal responsiveness at 12 and 18 months were weakly related (r = .14, p < .05). However, maternal social support was found to be quite stable from 12 to 18 months (r = .57, p < .001). Mothers and observers ratings of temperament were also correlated .19 (p < .01) at 12 months and .25 (p < .001) at 18 months. Similar patterns were found when correlations were computed separately by gender.
Irritability _ Responsiveness Relations
Relations between infant irritability and maternal responsiveness in the total sample are presented in Figures 1 and 2 (sorry, figures not available on this website). Figure 1 contains relations computed using observed irritability scores, and Figure 2 contains relations computed using ICQ scores.
Insert Figures 1 and 2 about here (sorry, Figures can't be accessed)
Concurrent relations at 12 months, those between irritability at 12 and 18 months, and those between responsiveness at 12 and 18 months were examined using zero_order correlations. Concurrent relations at 18 months were examined using zero_order correlations presented in parentheses, and partial correlations in which relations with 12_month measures were partialled out. Longitudinal relations between irritability and responsiveness were investigated using zero_order correlations presented in parentheses, and partial correlations in which relations with previous and concurrent infant and maternal variables were partialled out. For example, irritability at 12 months and responsiveness at 18 months were partialled out of the correlation between responsiveness at 12 months and irritability at 18 months. Given the number of correlations computed (12), the p value required for significance was set at .01.
As reported in the descriptive data section above, measures of infant irritability across time were moderately correlated. Measures of maternal responsiveness across time were weakly correlated. In three of four cases, relations between irritability and responsiveness when measured concurrently were negative and were significant or approached significance. At 12 months, as seen in Figure 1, the correlation between observed irritability and maternal responsiveness was _.39 (p < .001). At 18 months this correlation was _.44 (p < .001). As seen in Figure 2, the correlation between irritability and maternal responsiveness computed using ICQ scores at 12 months was _.10 (p < .10 and therefore nonsignificant) and at 18 months was _.17 (p< .05 and therefore only approached significance). Additionally, no significant longitudinal relations between maternal responsiveness and either measure of infant irritability were found.
Measurement Strategies. Results varied significantly depending on the method by which infant irritability was measured. The correlations in Figures 1 and 2 were compared using two_tailed t_tests for within_sample correlations employing a common variable. When relations between 12_month maternal responsiveness and each 12_month measure of infant irritability were compared the difference between correlations was highly significant (t1,204 = 3.6, p < .01) with a larger correlation computed using the observed measure of irritability (r = _.39 vs. r = _.10). Observed irritability also produced a larger correlation with maternal responsiveness at 18 months (r = _.44) than did the ICQ (r = _.17), and the difference between the correlations was highly significant (t1,183 = 3.51, p < .01). Despite these notable differences in concurrent correlations, measurement strategies did not produce differences in longitudinal correlations. Additionally, using a two_tailed Fisher's r_to_z transformation, the correlation between ICQ scores at 12 and 18 months was found to be larger than that between observed irritability measures at 12 and 18 months (z = 2.92, p < .01).
Gender. The correlations described above were computed separately by gender and produced results similar to those in the total sample. Differences between correlations for boys and girls were tested using two_tailed Fisher r_to_z transformations and none was significant.
Maternal Social Support
Relations between maternal social support and the responsiveness_irritability relation were examined first by computing zero_order correlations between the separate GLS scores and maternal responsiveness. The significant correlation using GLS Satisfaction at 12 months was small and in the opposite direction of that predicted (r = _.16, p < .05). The 12_month correlation using GLS Amount was nonsignificant (_.04). At 18 months the correlations were also nonsignificant (_.02 with GLS Amount, and .03 with GLS Satisfaction). When these correlations were computed separately by gender, results were similar to those in the total sample.
To test the hypothesis that maternal social support moderates the relation between maternal responsiveness and infant irritability, four hierarchical regressions were computed in the total sample. For these analyses, irritability was measured through observation only so that measures of maternal social support and infant irritability were not derived from the same source. In the first equation, maternal responsiveness at 12 months was regressed on 12_month observed irritability scores, 12_month GLS Amount, and their interaction term. In the second equation, maternal responsiveness at 12 months was regressed on 12_month observed irritability scores, 12_month GLS Satisfaction, and their interaction term. The third and fourth equations were identical to these but used 18_month measures.
At 12 months, the overall equation using GLS Amount was significant (F3, 178 = 14.12, p< .001), and the interaction between infant irritability and the amount of maternal social support added significant variance to the prediction of maternal responsiveness (F change = 6.85, p < .001, R2 change = .03). When this equation was computed separately for girls and boys, only for boys was the interaction term significant. Examination of the plotted interaction revealed main effects for infant irritability and for GLS Amount, with higher levels of irritability and higher levels of maternal support associated with lower levels of maternal responsiveness. The interaction was revealed in a greater discrepancy between maternal responsiveness scores across high and low levels of GLS Amount when irritability levels were low. Also at 12 months, the overall equation using GLS Satisfaction was significant (F3, 178 = 13.28, p < .001), but the interaction term between infant irritability and mother's satisfaction with her support did not add significant variance to the prediction of maternal responsiveness. This was true in the total sample and when equations were computed separately by gender.
At 18 months the overall equation using GLS Amount was significant (F3, 186 = 19.25, p< .001), and there was a nonsignificant trend for the interaction term to add variance beyond that accounted for by infant irritability and amount of maternal social support alone (F change = 3.27,p = .07, R2 change = .01). Like the 12_month equation, examination of the plotted interaction revealed a main effect for infant irritability, with higher irritability associated with lower levels of maternal responsiveness. However, unlike the 12_month equation, lower amounts of social support were associated with higher levels of maternal responsiveness only when infant irritability was also high. When irritability was low, lower amounts of social support were associated with slightly lower levels of responsiveness. When these equations were computed separately by gender, patterns similar to those in the total sample were found only for girls. In the fourth equation, which was also significant (F3, 186 = 18.31, p < .001), the interaction term between infant irritability and mother's satisfaction with her social support did not add significant variance to the prediction of maternal responsiveness. This was true in the total sample and when equations were computed separately by gender.
Relations between infant irritability and maternal responsiveness were investigated in this study of low_income mother_infant pairs, with particular emphasis on whether results varied as a function of measurement strategies, gender, or amount of maternal social support or satisfaction with that support. Concurrent relations between irritability and responsiveness were, in general, negative and were larger when irritability was measured observationally rather than by maternal report. Correlations were similar across gender. No longitudinal relations were found between irritability and responsiveness. Maternal social support was weakly related to responsiveness and to responses to infant irritability. Furthermore, the nature of these relations was unexpected in that higher amounts of social support were generally associated with lower levels of responsiveness.
A number of results from analyses of descriptive data warrant comment. Mothers rated girls as more difficult than boys at 18 months. Given that this finding is inconsistent with reports in the literature of no differences between girls' and boys' temperamental difficulty (e.g. Prior, Smart, Sanson, & Oberklaid, 1993), and the fact that this pattern was not found at 12 months, it seems the finding, although interesting, was sample specific and therefore not robust.
Maternally_reported levels of irritability decreased significantly from 12 to 18 months, but observed levels of irritability increased. Given that mothers were reporting about irritability and not about other traits and behaviors that might contribute to a perception of difficulty (e.g. activity or noncompliance), it makes sense that levels would decrease over time because infants likely become better emotionally regulated with age. However, ICQ items generally ask mothers at each age to rate their children compared to their beliefs about the "average" child, not compared to younger infants or their children at a younger age. If they compared their 12_month_olds to an "average" 12_month_old and 18_month_olds to an "average" 18_month_old, the ICQ means would not have decreased over time because at each age some children would by definition be more and some would be less irritable than "average". Mothers' ratings may be only somewhat sensitive to normative developmental changes. Mothers did rate their children as less irritable at 18 months than they did at 12 months, but the question "less irritable than what or whom?" remains. The question "more irritable than what or whom" applies to the observer ratings, which increased over time. Although well_trained observers are exposed to many children and should be sensitive to what children of different ages are typically like, their ratings are not necessarily more accurate than mothers' ratings at capturing a child's ranking on a particular trait or behavior.
The correlations between maternal and observer reports of irritability were modest but statistically significant. These findings are consistent with correlations in the literature ranging from .20 to .40 between temperament measured via mothers versus observers. It seems that these measures tap somewhat different constructs, both of which may be scientifically important (Bates, 1980; Seifer et al., 1994). As suggested above, one is not argued to be "better" than the other and the data here do not address that issue. Instead it is suggested that the source of information be considered carefully when results regarding measures of infant temperament are interpreted.
Maternally_reported irritability scores across age were moderately correlated and these correlations were larger than those obtained using the observed measure, but the latter showed significant correlations as well. Although this difference may suggest superiority of the maternal report measure, it is also possible that the difference is due to reporter bias. At each age, different observers rated each videotape, but the same mothers rated their infants. Given concern about relations between maternal characteristics and their ratings of infant temperament (Daniels, Plomin, & Greenhalgh, 1984; Matheny, Wilson, & Thoben, 1987; Sameroff, Seifer, & Elias, 1982; Vaughn and colleagues, 1980, 1981, 1987), it seems that the consistency of ICQ scores at 12 and 18 months may reflect maternal predispositions to view their children similarly across time. We do not mean to suggest a weakness in mothers' ratings of infant temperament, but instead argue again for careful consideration of the source of temperament information when results are interpreted. Additionally, the observed irritability measure may also have been biased or at least unrepresentative in its own way. Despite the range of laboratory situations across which irritability was assessed, it is not possible to replicate the conditions of the home environment where most of the maternal behaviors relevant to infant development occur.
Scores for maternal responsiveness at 12 and 18 months were weakly correlated, which is likely due to the brevity of the measure but may reflect real change in individual rankings of responsiveness across time. This may be especially true of this low_income sample. Typically, the significant stressors experienced by low_income families wax and wane and may be related to changes in maternal behavior over time. Of the few studies that empirically investigate the stability of maternal responsive behaviors throughout infancy, at least one (Crockenberg & McCluskey, 1986) also reported a nonsignificant correlation between maternal responsiveness at 3 months and sensitivity at 12 months, despite their use of more extensive assessments of maternal behavior.
As hypothesized, on a particular day during the second year of life, mothers of irritable infants were less responsive than mothers of "easy" infants, regardless of whether mothers or observers rated infant irritability. However, it is not possible to tell from concurrent correlations whether unresponsive mothers produced increasingly irritable children, or whether irritable children depleted maternal ability or will to respond appropriately. These questions were tested by examining longitudinal relations between irritability and responsiveness and no significant correlations were found. Theoretically it seems unlikely that mother and infant behaviors were not mutually influential over time, and these results are inconsistent with findings in the literature of negative longitudinal relations between responsive maternal behavior and aspects of difficult infant temperament (Campbell, 1979; Maccoby et al., 1984; Seifer et al., 1996; Spangler, 1990, van den Boom & Hoeksma, 1994). Differences in the specific conceptualization and measurement of responsiveness and difficulty might account for our discrepant findings. But it is also possible that our brief measure of maternal responsiveness was not sensitive enough to capture the longitudinal relations that have been demonstrated by others. It may also be that other aspects of infant temperament (e.g. activity or sociability) not measured in this study are more clearly related to responsive maternal behavior than our measure of negative emotionality.
Despite theory and empirical observations from other researchers that inspired a search for gender differences in mother_infant interaction, only one minor difference was found. At 12 months, amount of maternal social support moderated irritability_responsiveness relations for boys only, and at 18 months, the same held true for girls only. Apparently beliefs mothers have, if any, about differences between irritable girl and boy infants are not powerful enough to influence their behavior. This seems to be true at least in the context of many potentially more important influences on maternal behavior, such as maternal personality or education, infant factors including parity or other temperament traits, or even the lab setting.
Results regarding the use of observer versus maternal report of irritability showed larger correlations using observer ratings of irritability, as was hypothesized based on reports of similar results (Coffman et al., 1992; Seifer et al., 1996). However, it is not clear why observer ratings of irritability would better predict maternal behavior than her own opinion. Seifer et al. (1996, p. 22) suggested that compared to maternal appraisal, "directly observed infant behavior . . . may relate more to the specifics of how infants react to various situations." The ICQ may instead be capturing some aspect of the mother or the mother_infant relationship that is less relevant to specific maternal responsive behaviors (Belsky, Hsieh, & Crnic, 1998; Earls, 1981).
It is also possible that a measurement issue accounts in part for the larger correlations obtained using observer reports of irritability. Specifically, maternal responsiveness was measured during a three_minute task that overlapped with an hour_long observational assessment of infant irritability so that our judgment of maternal responsiveness might have depended in part on the infant's level of irritability during the assessment. This is unlikely for two reasons: 1) the observed infant irritability score was based on many more minutes than those also used to judge maternal responsiveness, and 2) our maternal responsiveness score included behavior that was both appropriate when a response was required and appropriate when no response was required so that a mother could be described as more or less responsive regardless of whether her infant was irritable during those three minutes.
The relations found between maternal social support and responsive behavior were unexpected but generally weak. At 12 months, lower amounts of social support were directly related to higher levels of maternal responsiveness, but satisfaction with support was unrelated. No direct relations between amount of support or satisfaction were obtained at 18 months. At 12 months the amount of maternal social support showed a slightly stronger relation with responsiveness when irritability was low, not high. And again the association was in the reverse direction of that predicted. There was a trend for the same finding to be significant at 18 months. Significant results were generated only with scores reflecting mother's total amount of social support, not her satisfaction with that support. It has been suggested that types of social support are differentially related to outcome (Crnic and colleagues, 1983, 1984, 1986). It seems that differences are obtained when amount of support versus satisfaction with that support are considered separately as well.
A possible explanation for the unexpected findings is that our measure of maternal responsiveness may have been too brief or restricting to produce meaningful relations with measures of maternal social support. And it is equally plausible that the generally weak relations are attributable to the restricted socio_economic status (SES) of our subjects. In most studies reporting relations between maternal social support and responsiveness (Crnic and colleagues, 1983, 1984, 1986; Crockenberg, 1981), subject SES was not restricted which may have increased the chance for significant correlations between maternal behavior and social support, which is likely related to SES. Perhaps predictions about relations in our low_income sample should not have been based on findings from studies using subjects of varied SES. In fact, Denham and Moser (1994) sampled only "upper_middle SES" subjects and failed to find relations between maternal sensitivity and a cumulative measure of maternal social support and stress. Although these findings do not help to clarify our own unexpected ones, together they suggest that the relations between maternal social support and responsiveness within SES may not be the same as those found across SES. Also, the unexpected relations between lower social support and higher responsiveness and the gender differences at each age may well be sample_specific and therefore non_robust phenomena. Most importantly, our findings regarding maternal social support were weak and should not be "over_interpreted".
Strengths of this paper include the use of a longitudinal design, a sample large enough to conduct analyses by gender, two methods for measuring infant irritability, and an understudied low_income sample. Additionally, data were originally analyzed separately from two similar studies, with the second study replicating findings from the first. However, these studies also had notable limitations that may prevent definitive conclusions from being drawn. First and foremost, the validity of our measure of maternal responsiveness could be questioned because it was a brief, somewhat restrictive, lab measure and scores showed little stability from 12 to 18 months. Although some results confirming our hypotheses were obtained using this measure, the weakness of relations both between responsiveness and maternal social support and longitudinally between responsiveness and infant irritability was surprising and may be due to the limitations of the maternal responsiveness measure.
Second, although the use of both maternal and observer report of irritability is considered a strength of this study, our strategies for measuring irritability are not without criticism. Each measurement strategy depended on data from a single time point at each age. Irritability perhaps would have been more reliably measured through the use of aggregated assessments over days or weeks at each age, and through the use of home observations in addition to the lab observations (Belsky et al., 1998; Seifer et al., 1994).
In summary, relations between infant irritability and maternal responsiveness do vary as a function of measurement strategies. Consequently, mixed findings in the literature thus far are likely due in part to methodological differences between studies, including how irritability was measured. Findings contained in this paper confirm that irritable infants from low_income families tend to experience less responsive mothering during the second year of life, which helps to explain mixed findings thus far between samples that have varied regarding infant age and family SES. Gender was not related and maternal social support was only weakly related to irritability_responsiveness relations in this sample. Although investigation of gender and maternal social support did not help to clarify previous mixed findings, it is possible that research using other moderators not investigated here, such as maternal personality variables or other infant temperament traits, will.
Adamkos, H., Ryan, K., Ullman, D. G., Pascoe, J., Diaz, R., & Chessare, J. (1986). Maternal social support as a predictor of mother_child stress and stimulation. Child Abuse and Neglect, 10, 463_470.
Ainsworth, M. S., & Wittig, B. A. (1969). Attachment and exploratory behavior of one_year_ olds in a Strange Situation. In B. M. Foss (Ed.), Determinants of infant behavior, Vol. 4. (pp. 111_136). London: Methuen.
Bates, J. E. (1980). The concept of difficult temperament. Merrill_Palmer Quarterly, 26, 299_319.
Bates, J. E., Freeland, C. B., & Lounsbury, M. L. (1979). Measurement of infant difficultness. Child Development, 50, 794_803.
Bates, J., E., Olson, S. L., Pettit, G. S., & Bayles, K. (1982). Dimensions of individuality in the mother_infant relationship at six months of age. Child Development, 53, 446_461.
Bell, R. Q. (1968). A reinterpretation of the direction of effects in studies of socialization. Psychological Review, 75, 81_95.
Belsky, J., Hsieh, K., & Crnic, K. (1998). Mothering, fathering, and infant negativity as antecedents of boys externalizing problems and inhibition at age 3: Differential susceptibility to rearing experience? Development and Psychopathology, 10, 301_320.
Bronfenbrenner, U. (1977). Toward an experimental ecology of human development. American Psychologist, 32, 513_531.
Caldwell, B. M., & Bradley, R. H. (1984). Home Observation for Measurement of the Environment. Little Rock: University of Arkansas at Little Rock.
Campbell, S. B. (1979). Mother_infant interaction as a function of maternal ratings of temperament. Child Psychiatry and Human Development, 10, 67_76.
Coffman, S., Levitt, M. J., Guacci, N., & Silver, M. (1992). Temperament and interactive effects: Mothers and infants in a teaching situation. Issues in Comprehensive Pediatric Nursing, 15, 169_182.
Crnic, K. A., Greenberg, M. T., Ragozin, A. S., Robinson, N. M., & Basham, R. B. (1983). Effects of stress and social support on mothers and premature and full_term infants. Child Development, 54, 209_217.
Crnic, K. A., Greenberg, M. T., Robinson, N. M., & Ragozin, A. S. (1984). Maternal stress and social support: Effects on the mother_infant relationship from birth to eighteen months. American Journal of Orthopsychiatry, 54, 224_235.
Crnic, K. A., Greenberg, M. T., & Slough, N. M. (1986). Early stress and social support influences on mothers' and high_risk infants' functioning in late infancy. Infant Mental Health Journal, 7, 19_33.
Crockenberg, S. B. (1981). Infant irritability, mother responsiveness, and social support influences on the security of infant_mother attachment. Child Development, 52, 857_865.
Crockenberg, S. B. (1986). Are temperamental differences in babies associated with predictable differences in care giving? In J. V. Lerner and R. M. Lerner (Eds.), Temperament and social interaction during infancy and childhood. New directions for child development, 31 (pp. 53_73). San Francisco: Jossey_Bass.
Crockenberg, S. B., & Acredolo, C. (1983). Infant temperament ratings: A function of infants, of mothers, or both? Infant Behavior and Development, 6, 61_72.
Crockenberg, S. B., & McCluskey, K. (1986). Change in maternal behavior during the baby's first year of life. Child Development, 57, 746_753
Crockenberg, S. B., & Smith, P. (1982). Antecedents of mother_infant interaction and infant irritability in the first three months of life. Infant Behavior and Development, 5, 105_119.
Daniels, D., Plomin, R., & Greenhalgh, J. (1984). Correlates of difficult temperament in infancy. Child Development, 55, 1184_1194.
Denham, S. A., & Moser, M. H. (1994). Mothers' attachment to infants: Relations with infant temperament, stress, and responsive maternal behavior. Early Child Development and Care, 98, 1_6.
Earls, F. (1981). Temperament characteristics and behavior problems in three_year_old children. Journal of Nervous and Mental Disease, 169, 367_373.
Hubert, N. C., Wachs, T. D., Peters_Martin, P., & Gandour, M. J. (1982). The study of temperament: Measurement and conceptual issues. Child Development, 53, 571_600.
Klein, P. S. (1984). Behavior of Israeli mothers toward infants in relation to infants' perceived temperament. Child Development, 55, 1212_1218.
Lee, C. L., & Bates, J. E. (1985). Mother_child interaction and age two years and perceived difficult temperament. Child Development, 56, 1314_1325.
Linn, P. L., & Horowitz, F. D. (1983). The relationship between infant individual differences and mother_infant interaction during the neonatal period. Infant Behavior and Development, 6, 415_427.
Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of sex differences. Stanford, CA: Stanford University Press.
Maccoby, E. E., Snow., M. E., & Jacklin, C. N. (1984). Children's dispositions and mother_ child interaction at 12 and 18 months: A short_term longitudinal study. Developmental Psychology, 20, 459_472.
Martin, J. (1981). A longitudinal study of the consequences of early mother_infant interaction: a microanalytic approach. Monographs of the Society for Research in Child Development, 46 (3, Serial No. 190).
Matas, L., Arend, R. A., & Sroufe, L. A. (1978). Continuity of adaption in the second year: The relationship between quality of attachment and later competence. Child Development, 49, 547_556.
Matheny, A. P., Wilson, R. S., & Thoben, A. S. (1987). Home and mother: Relations with infant temperament. Developmental Psychology, 23, 323_331.
Milliones, J. (1978). Relationship between perceived child temperament and maternal behaviors. Child Development, 49, 1255_1257.
Peters_Martin, P., & Wachs, T. D. (1984). A longitudinal study of temperament and its correlates in the first 12 months of life. Infant Behavior and Development, 7, 285_298.
Prior, M., Smart, D., Sanson, A., & Oberklaid, F. (1993). Sex differences in psychological adjustment from infancy to 8 years. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 291_304.
Sameroff, A. J., Seifer, R., & Elias, P. K. (1982). Sociocultural variability in infant temperament ratings. Child Development, 53, 164_173.
Schuler, M. E., Black, M. M., & Starr, R. H. (1995). Determinants of mother_infant interaction: Effects of prenatal drug exposure, social support, and infant temperament. Journal of Clinical Child Psychology, 24, 397_405.
Seifer, R., Sameroff, A. J., Barrett, L. C., & Krafchuk, E. (1994). Infant temperament measure by multiple observations and maternal report. Child Development, 65, 1478_1490.
Seifer, R., Schiller, M., Sameroff, A. J., Resnick, S., & Riordan, K. (1996). Attachment, maternal sensitivity, and infant temperament during the first year of life. Developmental Psychology, 32, 12_25.
Smith, P. B., & Pederson, D. R. (1988). Maternal sensitivity and patterns of infant_mother attachment. Child Development, 59, 1097_1101.
Spangler, G. (1990). Mother, child, and situational correlates of toddler's social competence. Infant Behavior and Development, 13, 405_419.
Thomas, A., Chess, S., & Birch, H. (1968). Temperament and behavior disorders in children. New York: New York University Press.
Thomas, A., Chess, S., Birch, H. G., Hertzig, M. R., & Korn, S. (1963). Behavioral individuality in childhood. New York: New York University Press.
van den Boom, D. C., & Hoeksma, J. B. (1994). The effect of infant irritability on mother_infant interaction: A growth_curve analysis. Developmental Psychology, 30, 581_590.
Vaughn, B. E., Bradley, C. F., Joffe, L. S., Seifer, R., & Barglow, P. (1987). Maternal characteristics measured prenatally are predictive of ratings of temperamental "difficulty"on the Carey Infant Temperament Questionnaire. Developmental Psychology, 23,152_161.
Vaughn, B., Deinard, A., & Egeland, B. (1980). Measuring temperament in pediatric practice. The Journal of Pediatrics, 96, 510_514.
Vaughn, B. E., Taraldson, B. J., Crichton, L., & Egeland, B. (1981). The assessment of infant temperament: A critique of the Carey Infant Temperament Questionnaire. Infant
Behavior and Development, 4, 1_17.
Vondra, J. I., Shaw, D. S., & Kevenides, M. C. (1995). Predicting infant attachment classification from multiple, contemporaneous measures of maternal care. Infant
Behavior and Development, 18, 415_425 .
Weinraub, M., & Wolf, B. (1983). Effects of stress and social supports on mother_child interactions in single_ and two_parent families. Child Development, 54, 1297_1311.