FINDINGS

The appendix for this chapter lists the sources used for each of the tables. There were 8 studies reporting associations between violence/physical aggression and parental loss or separation, 11 studies using attachment categories, 45 studies using continuous measures of attachment, 16 studies of other measures of parental bonding, and 10 studies of parental sensitivity or responsiveness.

Parental Loss and Separation and Violent Behavior

Parental loss and separation, for some, is the quintessential indicator of attachment failure. Studies of nonhuman primates suggest an evolutionary basis to separation behavior that we see in human children (e.g., Hoff, Nadler, Hoff, & Maple, 1994).

The preponderance of comparisons (PoC3), 17 out of 20 comparisons, in 6 out of 8 studies, show that parental loss or separation was positively, not always significantly, associated with violence (see Tables 6.1 and 6.2). The association was positive and significant in 3 of 8 studies, which is higher than we would expect by chance (which would be less than 5% if a = .05), but lower than what we would rely on to draw firm conclusions. The most convincing effects were reported by Corvo (2006) who found that separation and loss events, seen as indicators of attachment problems, were significantly associated with current levels of domestic violence, even controlling for measures of family-of-origin violence. Findings by Tupin, Mahar, and Smith (1973) were coded as “ambiguous” applying our conservative decision rule, though some of their findings were supportive. In that

Table 6.1 Summary of Study-Level Findings Related to Attachment, Violence and Nonviolent Offending

Summary of STUDY Results

Type of Study (Indicator of Attachment)

Number of Studies (k)

@

о

©

X

Parental Loss or Separation

Violent

8

2

0

0

3

3

Nonviolent

3

0

1

0

2

0

Attachment Categories

Violent

11

3

1

1

2

4

Nonviolent

1

1

0

0

0

0

Continuous Measures of Attachment

Violent

45

2

3

1

21

18

Nonviolent

25

1

2

0

9

13

Measures of Parental Bonding

Violent

16

0

0

2

4

10

Nonviolent

7

0

0

0

1

6

Caregiver Sensitivity

Violent

10

0

1

0

3

6

Nonviolent

1

0

0

0

0

1

Щ Findings are ambiguous

О Findings are in the opposite direction of the attachment hypothesis (not necessarily statistically significant)v

© Relationship is reported as “null” or coefficient = 0

Findings are in the expected direction of the attachment hypothesis but are not statistically significant

X Findings are in the expected direction of the attachment hypothesis and are statistically significant

Table 6.2 Summary of Comparison-Level Findings Related to Attachment, Violence and Nonviolent Offending

Summary of COMPARISONS

Type of Study (Indicator of Attachment)

Number of Comparisons

О

©

X

Parental Loss or Separation

Violent

20

2

1

5

12

Nonviolent

4

1

0

3

0

Attachment Categories

Violent

31

7

4

6

14

Nonviolent

4

2

0

1

1

Continuous Measures of Attachment

Violent

156

23

5

56

72

Nonviolent

96

12

3

26

55

Measures of Parental Bonding

Violent

64

0

11

12

41

Nonviolent

27

1

1

4

21

Measures of Parental Sensitivity or Responsiveness

Violent

42

3

5

7

27

Nonviolent

2

0

0

0

2

О Findings are in the opposite direction of the attachment hypothesis (not necessarily statistically significant)

© Relationship is reported as “null” or coefficient = 0

• Findings are in the expected direction of the attachment hypothesis but are not statistically significant

X Findings are in the expected direction of the attachment hypothesis and are statistically significant

study, being raised by persons other than parents was significantly associated with habitual violence but not “non-habitual” violent behavior among offenders.

There were no studies of young children. If loss and separation do wield a causal effect, we would expect to see it at its peak not too long after the changes in attachment figures took place, so the lack of studies of young children limits any conclusions from the body of literature as it exists to date. It is also important to note that most of these studies were not designed specifically to test the research question of interest here. Loeber et al. (2005) and Saner and Ellickson (1996) were generally looking for “risk factors”; Miura (2009) was interested in frontal lobe function; and Pogarsky et al. were interested in maternal age (Pogarsky, Lizotte, & Thornberry, 2003). The indicators of parental loss or separation are not ideal in this set of studies, either. Loeber et al. (2005) used “number of caretaker changes” which does not capture attachment loss in all cases. Model specification was also not the best for our purposes. For example, Cusick et al. (2012), who report nonsignificant findings in the right direction, include five indicators of parental loss and separation in the same model together, as well as a control for social support, abuse, and number of foster care placements. Given this “overspecification4” of the model (for our purposes), it is unlikely that a significant effect for any particular measure of parental separation would be statistically significant. Pogarsky et al. (2003) use separation from parent as a control variable in their model which also includes “number of family transitions” among other factors. Saner and Ellickson (1996) include controls for many overlapping constructs such as parent support, with recent separation/divorce and parent loss in the models at the same time; their findings are in the right direction and one of three is still statistically significant in spite of this practice.

Thus while we speculate that a reanalysis of these data sets, using the best indicators of parental loss and separation, and a set of control variables carefully designed to disentangle the effects of attachment loss on offending, would yield stronger results, we can only conclude at this time that the evidence is suggestive of an association between parental loss and separation and violence.

 
Source
< Prev   CONTENTS   Source   Next >