Preference for Violent Electronic Games and Aggressive Behavior among Children: The Beginning of the Downward Spiral?

Abstract

A one-year longitudinal study with 324 German third and fourth graders was conducted in order to find out whether a preference for violent electronic games socializes children to become more aggressive or whether aggressive individuals tend to select this type of game. Cross-lagged panel analyses suggest that children who were rated as openly aggressive at Time 1 intensified their preference for violent electronic games over time. We determined that it could be ruled out that this selection effect was due to a number of underlying variables ranging from ecological variables (neighborhood) to family variables (migration status, older brother) and child variables (gender, self-esteem, level of achievement). Discussion focuses on the emerging preference for violent electronic games among children.

Notes

Note. r (pairwise deletion), *p ≤ .05

**p ≤ .01. PNI = peer nomination index, TR = teacher rating.

Note. *p ≤ .05

**p ≤ .01.

1. The nature of the data matrix did not meet the criteria necessary to calculate Kendall's W. Instead, bivariate intercorrelations between all raters on the target variable “brutal and bloody plot” were calculated and transformed by means of Fisher's Z standardization. Then, the overall mean score was calculated and retransformed into a correlation coefficient indicating the inter-rater reliability.

2. The positive selection bias in the diary sample was another reason why we chose the preference measure over an exposure measure that could be derived from the diary data.

3. Because brutal and bloody electronic games model and reward physically aggressive behavior, stronger effects may be expected for this variant of aggression. However, when openly aggressive behavior was substituted by its component variable physically aggressive behavior, the same pattern of results emerged. Again the selection path emerged as significant (β = .25, p < .05) and the socialization path as negligible (β = .02, ns). The overall goodness fit of this model was also very good, χ²(4, N = 238) = 6.08, p = .19, RMSEA = .05, SRMR = .03.

4. The power analysis showed that this cross-lagged effect was statistically significant (α = .05) in 94% of all replications, while the other cross-lagged effect of β = −.01, ns, was statistically insignificant in 100% of all replications.