Abuse and the Accessibility of Childhood Memories

In the first investigation of trauma-related memory problems in our laboratory, Johnson, Greenhoot, Glisky, and McCloskey (2005) examined the extent to which both early and recent abuse experiences, as well as current depression, predicted adolescents' memory functioning during Year 6 of the longitudinal study. The sample consisted of a subset of 134 participants whose autobiographical memory assessments were transcribed and available for analysis and whose reports regarding family violence were corroborated by their mothers.1 The participants reported a broad range of family violence and abuse exposure, from no exposure to moderate exposure to highly frequent exposure. For both Year 1 and Year 6, we collapsed the indicators of mother- and child-directed violence into overall measures of the frequency of family violence reported at each interview, capping a few extreme outliers to prevent them from skewing the results. Because sexual abuse was only reported for a small number of these participants at Year 1 (n = 8) and Year 6 (n = 12), and because some of the disclosed abuse had taken place in the distant past, we created a dichoto-mous indicator of whether the participant had ever been sexually abused for these analyses.

During Year 6, when they were between 12 and 18 years of age, the participants' autobiographical memories for childhood were assessed with an adaptation of the AMT in which they were presented two positive (playing, present), two negative (arguing, punishment), and two neutral (car, shopping) cue words in random order. They were given 3 minutes to generate specific childhood memories (from before age 9) in response to each cue. Consistent with previous research using the AMT, each memory generated was coded as specific or overgeneral. But whereas most

1. This subset of participants did not differ from the remainder of the sample in family violence exposure or demographic characteristics.

previous research has focused almost exclusively on rates of overgeneral or specific memories, we examined two additional dimensions of the participants' AMT performance that might indicate an inability or unwillingness to recollect specific past experiences: the number of interviewer prompts required to elicit each memory and the mean length of each memory produced (indexed by the number of words in the narrative). Occasionally, portions of the audiotaped interviews were too muffled for the transcribers and coders to interpret, and memories for which more than two words were unintelligible were not included in the calculation of average memory length and were assigned a code of "unintelligible." We also examined the valence of the memories produced (i.e., negative or non-negative), as judged from the perspective of an objective observer; we thought that valence might provide additional information about whether memory problems reflect emotion regulation processes as opposed to broader cognitive or memory dysfunctions. To evaluate the possibility that abuse-related autobiographical memory problems reflect more general memory deficits, we also tested the adolescents' immediate and delayed recall of a list of 10 verbally presented paired associates. Finally, to measure depressive symptoms at Year 6, we administered the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977) and calculated a dichotomous indicator of depression using the clinical cutoff recommended by Radloff (1977). Although the Year 6 assessment included several other measures of child well-being and psy-chopathology (e.g., anxiety, attachment, attention problems, externalizing symptoms, and aggressive and delinquent behavior), preliminary analyses revealed no associations between autobiographical memory performance and these other measures of mental health; thus they were excluded from the remaining analyses.

A series of general linear models predicting the measures of childhood autobiographical memory indicated that the frequency of both recent and childhood exposure to family violence, as well as depression, were associated with performance on the AMT, even when other important variables such as age and gender were included in the models. Table 4.2 presents the standardized regression coefficients resulting from these analyses. Note that the model predicting memory length controlled for the number of unintelligible memories because these memories were excluded from the calculation of the mean length of each memory. Additionally, the model predicting the number of overgeneral memories generated per

table 4.2. Standardized Regression Coefficients from General Linear Models Predicting Measures of Memory Performance on Year 6 AMT


# Prompts


# Overgeneral

% Negative













Positive cues


Negative cues


Neutral cues


Year 1 family violence





Year 6 family violence



Positive cues



Negative cues



Neutral cues

.164 t


Sexual abuse










Unintelligible memories

-.177 t

Memory length




Specific memories


Parameter estimates are presented by cue type only when there was a significant interaction between a predictor and cue type. a Male = 0; female = 1. tp < .06; *p < .05; **p < .01; ***p < .001.

Parameter estimates are presented by cue type only when there was a significant interaction between a predictor and cue type. a Male = 0; female = 1. tp < .06; *p < .05; **p < .01; ***p < .001.

cue included memory length, number of prompts, and number of specific memories as covariates, because these variables might affect the number of responses participants were able to produce during each 3-minute period. As indicated in Table 4.2, teens who were exposed to higher levels of recent (Year 6) family violence provided shorter childhood memory narratives across all cue types, fewer negatively valenced memories in response to neutral cues, and more overgeneral memories in response to neutral cues, although the latter effect did not quite reach significance.2 Greater exposure to family violence during childhood (Year 1) was as

2. Note that although the univariate tests indicated that the effect of recent family violence on overgeneral memories for neutral cues did not quite reach significance, the test of the interaction between recent family violence and cue type in the repeated measures model predicting overgeneral memories was statistically significant, F(2, 119) = 4.15, p = 0.018.

sociated with more interviewer prompts on the AMT, suggesting that interviewers had to work harder to elicit specific childhood memories from teens with traumatic childhoods. Moreover, consistent with the extensive literature on depression and memory functioning, teens who met the clinical cutoff for depression produced more overgeneral memories across all cue types than teens who were not depressed. In contrast to previous research, reports of childhood sexual abuse did not relate to autobiographical memory patterns in this study, but it should also be noted that only 14% of the sample reported histories of sexual abuse, and this may not have been a large enough group to detect an effect on autobiographical memory.

Importantly, performance on the paired associates test, our measure of nonautobiographical episodic memory, was not associated with family violence, depression, or any of the measures of AMT performance. Thus, as in previous work by de Decker et al. (2003), the links between trauma history and autobiographical memory functioning observed in this study do not seem to be explained by basic memory impairments. The overall pattern of results is more consistent with the argument that trauma-related memory problems reflect cognitive strategies for affect regulation that involve avoiding thinking and talking about the details of past experiences so as to blunt potentially negative emotions. The fact that teens exposed to recent (Year 6) family violence produced shorter and more generic memories, as well as fewer negative memories in response to neutral cues, than other teens provides convergent evidence that they may have been avoiding the retrieval of potentially painful memories, or that they were unwilling to report memories that were retrieved. Similarly, the link between childhood family violence and the need for more interviewer prompting to elicit specific memories suggests that participants with childhood abuse histories had encoded and stored specific memories but were reluctant to recollect them. Whether depression is a cause or an outcome of this apparent avoidance cannot be determined from this study. Our results also suggest that avoidance may be more easily applied under some conditions than others. For instance, the link between Year 6 abuse and the production of fewer negative memories on the AMT held only for neutral cues, which offer more latitude in the types of memories recalled than positive or negative cues. Similarly, Year 6 abuse was associated with overgeneral memories for neutral cues alone. We are not aware of any other study that has used neutral cues on the AMT, but previous findings of cue type effects (i.e., positive versus negative) on trauma-related specificity problems have been quite inconsistent (e.g., de Decker et al., 2003; Henderson et al., 2002; Kuyken & Brewin, 1995). Although we cannot draw any definitive conclusions about these divergent cue type effects, one interpretation is that the semantic (e.g., strength and extensiveness of associative networks) and emotional qualities of different cues may create different contexts for recall, which may influence the degree to which avoidance is utilized.

An additional implication of our findings is that this pattern of avoidance does not necessarily have to emerge out of early adverse experiences, as both recent and childhood exposure to family violence predicted performance on the AMT. Thus, even though the tendency to avoid childhood recollections might reflect an enduring style that develops in response to early adversity, this pattern could also reflect a more transient, strategic response to current stressors. Indeed, given that research on the development of emotion regulation suggests that cognitive strategies for affect regulation develop late in childhood, it is possible that the tendency to avoid details and blunt affect does not develop until later in life, in response to negative thoughts about past experiences.

0 0

Post a comment