Do Men With Excessive Alcohol Consumption and Social Stability Have an Addictive Personality?

¶ … Berglund et al. (2011) are addressing in this research are actually quite old questions regarding the relationship between addiction and personality. They discuss Cloninger's hypothesis of type I and type II alcoholics and differences in personality styles. The researchers are interested in determining if an empirical relationship between excessive alcohol consumption and certain personality traits in males exists.

The idea of an addictive personality is actually quite old dating back to psychodynamic concepts of addiction and early researchers tried to find profiles on personality inventories such as the Minnesota Multiphasic Personality Inventory (MMPI) that were distinctive for additive behavior (e.g., Lester, Burkman, Gandica, & Narkunski, 1976). However, despite a few scales that have some predictive power towards potential addiction the idea of an "addictive personality" has never gained much empirical support.

Technically the null hypothesis in this study would be that there is no difference between heavy drinkers and controls on scales of the Karolinska Scales of Personality (KSP) which is actually the hypothesis of the researchers as well. The alternative hypothesis would be that there are differences between the two groups on the KSP scales. The first issue here is that the researchers are looking to support the null hypothesis.

Quite frankly the null hypothesis can never be confirmed (Nunnally & Bernstein, 1994); however, it can be disconfirmed giving the specific parameters of a study. Thus, the researchers put themselves in awkward position of trying to confirm an unconformable hypothesis. In fact a bigger problem here is that there is no "addictive personality" profile specified by the KSP or any other personality inventory. That fact that there is not an addictive personality profile indicates the researchers are simply fishing for something that is not even there.

Thus, this makes the study even more awkward because simple differences on a few scales of a large personality inventory did not indicate a specific personality type. Perhaps using the scales on the MMPI or other inventory that have empirical support to suggest a higher potential for addiction would have been better. A second issue here (and the major issue) is that the normative group did not appear to be analyzed for alcohol abuse and dependence in the same way that the self -- reported excessive alcohol abusers were.

The researchers appear to have taken the normative group at their word regarding the use of alcohol. The researchers did not present data showing that the excessive drinking group really drank significantly more alcohol than the normative group. We do not know if the two groups were statistically equivalent in their alcohol use and the fact is that outliers in excessive drinking group did demonstrate some potential differences on scales of the KSP. Perhaps these "outliers" were actual alcoholics and the rest of the subjects in both groups were not.

There is no way to tell with the information in the study. The identity of the groups and their validity was not investigated here. There is also a major problem with the variables and methodology in the study. Instead of using more variables in the researchers should have attempted to get more participants. The reason for this is that in PCA it is generally accepted that at least a 10:1 subject to variable ratio ensures that the analysis will not be spurious (see MacCallum, Widaman, Preacher, & Hong, 2001).

The excessive drinking group only had 100 subjects in the normative group 131 subjects and the KSP has 15 scales (thus each sample should have contained at least 150 subjects). Moreover, despite the researchers' claims, PCA is probably not the best statistical method to achieve the goals the researchers wished to achieve.

Instead, some form of confirmatory factor analysis or model fitting would have been appropriate where a model on the KSP for the excessive drinkers would be developed and then the researchers could try to see if this model fit with the normative group. If the models fit with little error then the researchers' hypothesis would be supported, whereas if.