Dr. Herbst and Colleagues Reply

To the Editor: In our article, we tested the psychobiological model of temperament and character of Cloninger et al. (1) on both molecular genetic and phenotypic levels. Our molecular genetic analyses offered no support for the model. There were no significant associations between polymorphisms in D4DR and the personality trait of novelty seeking, nor were there associations between 5-HTTLPR and harm avoidance. At the phenotypic level, a factor analysis of the 25 subscales of the Temperament and Character Inventory did not reveal the hypothesized seven-factor structure. In particular, the subscales of novelty seeking defined two separate factors.

In their comment on our article, Dr. Lichtermann and colleagues report the results of a factor analysis conducted in their Finnish sample (Ekelund et al., 1999). Like us, they found a clear harm avoidance factor, but they also found a relatively clear novelty seeking factor. They suggest that this “may have allowed detection of a DRD4 association in our sample but not in theirs.”

This is unlikely to be the explanation. The structure that Dr. Lichtermann and colleagues describe resulted from a different analysis than the one we reported. They extracted four factors from the 12 temperament subscales of the Temperament and Character Inventory, whereas we extracted seven factors from all 25 subscales of the same inventory. When we restricted our analysis to four factors from the 12 temperament subscales, we, too, observed a single factor defined by the novelty seeking subscales. The phenotypic structure of the Temperament and Character Inventory appears to be similar in the U.S. and Finnish samples when analyzed in the same manner.

However, we did not observe the same genotypic associations. Although our smaller sample made it impossible to duplicate the analyses Dr. Lichtermann et al. reported, we did examine the distribution of alleles in groups with high and low levels of novelty seeking by using a median split and groups found at the extremes, which were defined both as those scoring in the upper and lower 27% and as those scoring more than one standard deviation from the mean. No significant associations were found.

However, we believe the most important point has been missed by Dr. Lichtermann and colleagues. The significant associations they reported in their article (Ekelund et al., 1999) were in the wrong direction and did not support the temperament and character model. Their original hypothesis was that long alleles of D4DR would be associated with novelty seeking, whereas in their study they found that short alleles were more frequent among individuals with high novelty seeking levels—a finding with no obvious biological rationale. Neither their study nor ours supports the temperament and character model.