Corporal Punishment and Antisocial Behavior

Straus et al¹ purport that corporal punishment (CP) is correlated with long-term increases in antisocial behavior among children. This highly publicized article is riddled with methodological and statistical flaws that cast light more on the long-established biases of Dr Straus and colleagues than on CP's use in discipline.

While the study controls for demographic and socioeconomic variables, baseline level of antisocial behavior, maternal warmth, and cognitive stimulation, it fails to consider several equally germane factors, including school setting, presence of attention-deficit/hyperactivity disorder, and family dynamics. The utter lack of evaluation of paternal involvement raises serious concerns about the validity of this study. As fathers (when present) generally are the primary disciplinarians, the study's absence of assessment of fathers begs key questions. Are maternal spanking and child antisocial behavior both correlated with the presence of single-mother homes? As CP's use and effects may vary on boys and girls, does the effect of CP vary with whether it is delivered by the mother or father? Do poor (and possibly violent) mother-father relationships translate into both maternal spanking of children and increased child antisocial behavior?

The study also does not control for the possibility that mothers who spank may already be biased toward perceiving their children to have antisocial behavior. Furthermore, the indicators of antisocial behavior are vague and subjective, depending on maternal perceptions alone. Objective markers, such as truancy, fights at school, drug abuse, suspension or expulsion from school, premature sexual activity and pregnancy, declining academic performance, crime, and delinquency, are not at all assessed. These types of behaviors are the true measures of CP's efficacy, not a questionnaire gauging parental perception.

The weakness of the study's design is reflected in its underwhelming results: the strongest presented coefficient of correlation is 0.312, which yields a coefficient of determination (r²) of 0.096; ie, more than 90% of the variance in antisocial behavior is not attributable to spanking. Furthermore, the authors used 1-tailed tests for 30 correlations, and Bonferroni (or other) corrections were not performed to adjust for the testing of so many correlations. In the absence of raw data, assuming independence of the 30 correlations, the actual type 1 error, even using 1-tailed tests, is .26; using unmotivated 2-tailed tests, it can be as high as .45!

Corporal punishment is a time-tested tool employed in the disciplinary armamentarium of many cultures. The erosion of traditional family values in the United States has paralleled efforts to impugn this practice and rises in social decay, while Eastern societies (where CP is common) continue to best the United States in both academic and social indicators of youth welfare. The authors note that since the 1950s, the prevalence and approval of CP in the United States has steadily declined. The same period is witness to disturbing rises in youth crime and delinquency. The sweeping conclusion of Straus et al that eliminating CP will reduce the level of violence in US society is based on little more than statistical quicksand and methodological thin ice. The study uses statistics much the way drunkards use lampposts: for support rather than illumination. It is a sad commentary on the state of science-society relations that this study garnered such intense global publicity instead of the careful scrutiny it deserves.