J Fam Viol (2010) 25:73–80 DOI 10.1007/s10896-009-9271-2
ORIGINAL ARTICLE
Automatic Encoding of Ambiguous Child Behavior in High and Low Risk for Child Physical Abuse Parents Julie L. Crouch & Joel S. Milner & John J. Skowronski & Magdalena M. Farc & Lauren M. Irwin & Angela Neese
Published online: 11 August 2009 # Springer Science + Business Media, LLC 2009
Abstract Recent theory and research suggest that physically abusive parenting behavior might be understood as originating from: 1) greater accessibility of hostile/negative schema, and/or 2) lower accessibility of benign/positive schema. This study examined whether parents at high and low risk for child physical abuse (CPA) differed in the extent to which they spontaneously encoded ambiguous caregiving contexts in negative versus positive terms. Twenty-five high and forty-one low risk for CPA parents were asked to memorize a set of sentences that described ambiguous caregiving situations. After a brief delay, participants were asked to recall the sentences. During recall, cues were given (e.g., negative and positive words) to facilitate recall. According to the cued-recall paradigm, to the extent that recall was facilitated by negative/positive cues, it was inferred that negative/positive meaning was activated when the ambiguous sentences were encoded. Although all parents tended to recall more information in response to negative relative to positive cues, the influence of cue type on recall was greater for high CPA risk parents. That is, high, compared to low, CPA risk parents obtained significantly higher recall difference scores (M=4.6 versus M=2.3); with higher recall difference scores indicating
This research was supported, in part, by Grant #CA901490 from the Children’s Bureau, Administration of Children and Family Services, U.S. Department of Health and Human Services. Correspondence regarding this article should be addressed to Julie L. Crouch, Center for the Study of Family Violence and Sexual Assault, Northern Illinois University, DeKalb, IL 60115. J. L. Crouch (*) : J. S. Milner : J. J. Skowronski : M. M. Farc : L. M. Irwin : A. Neese Center for the Study of Family Violence and Sexual Assault, Northern Illinois University, DeKalb 60115 IL, USA e-mail:
[email protected]
greater recall in response to negative relative to positive cues. Present findings are consistent with the proposition that high and low CPA risk parents differ in how they spontaneously encode information in ambiguous caregiving contexts. Keywords Child physical abuse . Information processing . Encoding
Introduction Social information processing (SIP) models propose that individuals develop mental networks (i.e., cognitive schema) based on their experiences (e.g., histories of attachment, maltreatment, trauma) and that these representations exert an influence on how social information is perceived, processed, and interpreted. The SIP models of child physical abuse (CPA; e.g., Azar et al. 2008; Milner 1993, 2003) suggest that the cognitive schema of abusive parents create patterns of information processing that favor attention to aggressive or negative stimuli and increase the likelihood of interpreting behavior as aggressive or hostile. Such processing patterns are believed to increase the likelihood of subsequent parental aggression (Crouch and Milner 2005; Todorov and Bargh 2002). For example, the mental networks of parents at risk for CPA may be especially likely to contain readily accessible hostility constructs (Farc et al. 2008), which may influence interpretations, attributions, and behaviors automatically (i.e., without intention or awareness). Such implicit (i.e., unconscious) processes are obviously a challenge to study. Fortunately, social cognitive researchers have developed several methods for assessing implicit information processes and some of these innovative methodologies have been
74
employed in attempts to better understand the information processes associated with aggression. For example, Zelli and colleagues (Zelli et al. 1995, 1996) used the cued-recall task to examine implicit encoding biases in aggressive and nonaggressive individuals. In these studies, participants memorized sentences that each described an ambiguous social interaction. After a brief delay, participants were asked to recall the sentences and were given cues to aid their recall. The cues were either hostility-related or nonhostile words that were semantically related to the sentence content. The cued-recall paradigm assumes that recall cues whose meanings are activated at the time of sentence encoding will be especially likely to facilitate recall. Thus, if recall is facilitated by hostile cues, it is inferred that hostile meaning was activated as the ambiguous information was encoded. Zelli et al. (1995) found that aggressive, compared to nonaggressive, individuals differed in their spontaneous encoding patterns in two ways: 1) aggressive individuals showed a greater tendency to encode ambiguous information in hostile terms (mean recall to hostile cues: nonaggressive, M=1.63, SD=1.73; aggressive, M=2.31, SD=2.14); and 2) aggressive individuals showed a relative lack of encoding in semantically-related nonhostile terms (mean recall to semantic cues: nonaggressive, M=3.71, SD=3.0; aggressive, M=2.09, SD=2.20). Noteworthy is the fact that the largest difference between aggressive and nonaggressive individuals in the Zelli et al. study was found in the amount of information recalled to nonhostile semantic cues. More specifically, nonaggressive, compared to aggressive, participants recalled significantly more information to the nonhostile, semantically related cues. Results obtained by Zelli et al. (1995) suggest that when spontaneously processing ambiguous social stimuli, aggressive and nonaggressive individuals differ most in the amount of information they encode in nonhostile (i.e., semantic or positive) terms. Initial encoding of social information as benign or positive may increase the likelihood of producing benign or positive interpretations in ambiguous situations, and thereby reduce the likelihood of angry or aggressive responding. The notion that low CPA risk parents show positive interpretive tendencies when interpreting child behaviors has been suggested in other research. For example, Dopke et al. (2003) presented a series of vignettes depicting compliant and noncompliant child behaviors and asked high and low CPA risk parents to identify which vignettes represented the child being compliant. Dopke et al. found that low CPA risk parents were significantly more likely to interpret behaviors meant to represent noncompliance as compliance. Collectively, findings across studies suggest that low, compared to high, CPA risk parents are more likely to interpret child behaviors in positive terms (Dopke et al.) and that these positive
J Fam Viol (2010) 25:73–80
interpretive tendencies may originate at the time of encoding (Zelli et al.). Although the SIP model of CPA, as initially formulated, emphasized cognitive processes that increase risk of CPA, the studies reviewed above suggest that aggressive behavior also might be understood as originating from the relative absence of positive or benign information processes that may reduce the likelihood of aggression (Berkowitz 2008; Dodge 2006). In keeping with this notion, Dodge (2006) proposed a revision to his SIP model for understanding the development of aggressive behavior problems in which aggression is construed as a primitive response that becomes inhibited through the process of socialization. According to Dodge (2006), patterns of hostile attribution in response to negative, aversive, or frustrating experiences develop early in life; and it is only through positive socialization experiences that the capacity to formulate benign (i.e., nonhostile) attributional tendencies is developed. Examples of positive socialization experiences leading to benign attributional styles include secure attachment to a nurturing caregiver and modeling of benign attributions by significant others. Individuals who lack such positive socialization experiences will, in all likelihood, fail to develop the capacity for benign attributional styles, resulting in persistence of the more primitive hostile attributional style. Consistent across both the SIP model of CPA and Dodge’s model of aggression is the proposition that attributions of hostile/benign intent are believed to mediate how one responds to environmental challenges, with hostile/benign attributional styles leading to patterns of aggressive/nonaggressive responding, respectively. To recapitulate, the SIP model of CPA, as originally formulated, emphasized the risk potentiating role of negative/hostile schema in social information processing (Crouch and Milner 2005; Milner 1994, 2003). Accumulating evidence suggests that benign/positive schema may play an important function in reducing risk of aggression in low CPA risk parents (Dopke et al. 2003; Zelli et al. 1995). In all likelihood, both positive and negative schemata are involved in determining CPA risk and the influence of both may be evident as early as the encoding stage of information processing (Zelli et al. 1995). The present study was modeled after Zelli et al. (1995) and attempted to extend their findings to parental aggression by examining differences in encoding among parents with varying degrees of CPA risk. More specifically, the present study used a cued-recall task and asked parents to memorize sentences depicting ambiguous caregiving situations (e.g., “Miguel slapped his hand on the tray as his mother fed him.”). Cues used to prompt subsequent recall included both negative (e.g., hostile, difficult) and positive (e.g., happy, sweet) words. It was predicted that high, compared to low, CPA risk parents would display 1) greater
J Fam Viol (2010) 25:73–80
recall to negative cues, and 2) less recall to positive cues. Given that raw recall scores may be affected by individual differences in memory ability, recall difference scores (i.e., negative minus positive cued-recall) also were computed, such that higher recall difference scores reflected greater recall to negative relative to positive cues. It was predicted that high, compared to low, risk parents would obtain higher recall difference scores (i.e., recall more to negative relative to positive cues).
Methods Participants The initial participant pool included 100 general population parents (25 male, 75 female). Of the 100 participating parents, 4 parents were excluded due to missing data (i.e., left 10% or more items blank on any measure) and 30 were excluded for response distortion (3 faked bad, 6 randomly responded, and 21 faked good as indicated on the response distortion indices of the Child Abuse Potential [CAP] Inventory). Of the remaining 66 parents, 41.8% were African American, 52.2% Caucasian, 4.5% Latino, and 1.5% were other. Mean age was 29.2 years (SD=8.52). Most of the parents (68.2%) were not married and the mean number of children was 1.5 (SD=.66). Highest grade completed ranged from 10 to 20 years, with 40.3% of the parents in the sample reporting 12 or fewer years of education. Among parents with valid and complete protocols, 25 were classified as high risk and 41 as low risk for CPA, based on scores from the CAP inventory (see classification procedures described below). Mean CAP scores were 78.3 (SD=45.0) and 236.0 (SD=56.1) for low and high CPA risk parents, respectively. Low and high CPA risk groups did not differ with regard to the percentage of male participants (14.9%), nor did the groups differ on the mean age of participants or number of children. Low, compared to high, risk parents were more often married (41.5% vs. 16.0%) and reported more years of education (14.5 [SD=1.69] versus 12.7 [SD=2.46]). Low, compared to high, risk parents were less often African American (31.0% vs. 60.0%) and more often Caucasian (66.7% vs. 28.0%).
Materials Sentence Stimuli Twelve ambiguous sentence stimuli were selected from an initial set of 20 sentences. Each sentence contained the following elements: child’s name (e.g., Riley), child’s action (e.g., “kicked his legs”), caregiving context (e.g., “as his mother changed his diaper”). Children’s names
75
were selected from among the 100 most popular names in 2002. Children’s actions were selected from a list of subtle and potent disengagement cues commonly displayed by children younger than 3 years of age. Caregiving contexts included activities that occur routinely (e.g., feeding, bathing) when caring for young children. In order to identify sentences within the initial pool of 20 sentences that were ambiguous with regard to hostile intent, a preliminary study was conducted. Participants in the preliminary study were asked to read each of the 20 sentences, and each sentence was followed by a list of 24 descriptors. After reading each sentence, participants were asked to “rate the extent to which each word below describes the child’s behavior.” Twelve of the descriptors were negative attributes: uncooperative, aggressive, defiant, difficult, irritable, mean, negative, oppositional, hostile, unfriendly, cold, and violent. The remaining 12 descriptors were positive attributes: peaceful, animated, loving, attached, playful, content, happy, friendly, lively, sweet, accepting, and endearing. Participants marked their rating for each descriptor on a scale from 1 “not at all” to 5 “very much.” Participants in the preliminary study included 66 undergraduates and 18 parents. More specifically, a convenience sample of 66 female undergraduates was first obtained. Participation was limited to female (rather than male) undergraduates given that they were on average more likely to have had prior experience providing childcare. Their ages ranged from 18 to 21 years (M=19.6, SD=.85). The mean number of years of education was 12.6 (SD=.78). Racial/ethnic composition of the undergraduate sample was as follows: 53.0% African American, 34.8% Caucasian, 9.1% Latino, 1.5% Asian-American, and 1.5% other. The undergraduate women were heterogeneous with regard to CPA risk, with CAP scores ranging from 12 to 351 (M=127.78, SD=93.75). To examine whether the student ratings were similar to ratings provided by parents, a small parent sample was also obtained for the preliminary study. More specifically, 18 general population parents (3 fathers and 15 mothers) whose ages ranged from 20 to 34 years (M=25.9, SD=5.1) were also asked to participate in the preliminary study. The mean number of years of education among parents was 12.6 (SD=3.6). One-third of the parents (33.3%) were African American and the remaining (66.7%) were Caucasian. Most (55.6%) of parents were single, with 22.2% married and 16.7% separated or divorced. The parents were heterogeneous with regard to CPA risk, with CAP scores ranging from 27 to 305 (M=105.67, SD=82.46). Indices for the negative and positive ratings for each sentence were computed. The negative (positive) index represented the mean rating across the 12 negative (positive) traits. Mean negative and positive ratings did
76
J Fam Viol (2010) 25:73–80
not differ significantly for the undergraduate and parent samples, therefore the data from the two samples were pooled. A difference score was then calculated by subtracting the positive index from the negative index for each sentence. Difference scores closest to zero were considered ambiguous (i.e., difference scores near zero indicated that the sentence was rated as potentially negative and positive to equivalent degrees). A difference score greater than zero indicated that the child in the sentence was rated higher on the negative compared to the positive descriptors. A difference score less than zero indicated that the child in the sentence was rated higher on the positive compared to the negative descriptors. Twelve sentences with difference scores between −1 and +1 (i.e., were ambiguous in that they received nearly equal negative and positive ratings) were selected as the ambiguous sentence stimuli to be used in the encoding task. See Table 1 for mean negative ratings, positive ratings, and difference scores for the 12 sentences used in the encoding task. Recall Cues Cues for the recall task were selected from the 24 descriptors used to rate the sentences in the sentence selection process described above. More specifically, the correlations between the ratings for each of the 24 descriptors and the negative/positive indices were computed (computation of the negative/positive indices is described above). Each of the negative recall cues selected for use in the cued-recall task (uncooperative, difficult, irritable, negative, hostile, unfriendly) was significantly correlated (r’s ranged from .74 to .82; p’s