In line with our hypothesis the TPQ profile of violent ASPD mostly suits Cloninger's definition of ASPD. We also hypothesized that ASPD distinguishes types of violent crimes. Results suggest that arson is more common in the non-ASPD group than among offenders who had ASPD but no other differences occurred.
The typical violent offender temperament profile comprised high NS, high HA, and low RD. This profile suits Cloninger's proposal of explosive personality. Further, an explosive personality linked to alcoholism and antisocial behavior – Cloninger claims – corresponds with the impulsive-aggressive antisocial personality disorder described in DSM-III, distinct from his definition of antisocial personality disorder ("primary psychopath") with high NS (including second order traits of impulsive-aggressive behavior), low HA, and low RD . Our results support this kind of distinction among alcoholic persons with ASPD because we observed, embedded in the general high harm avoidance trend, a notable offender minority with exceptionally low harm avoidance scores. Results suggest that low HA connects to ASPD on its own, whereas high HA corresponds with the comorbidity of ASPD and borderline personality disorder. This suits Cloninger's hypothesis as well. Though, in contrast to Cloninger, our "primary psychopaths" who featured low HA, connected to second-order traits of low impulsiveness (NS2) and high degree of social attachment (RD3) similar to the traits of our controls. We note that 45% of the low HA offenders lacked an ASPD diagnosis.
According to Cloninger, low HA manifests a temperament of uninhibited optimism, confidence, gregariousness, and vigor  that presumably results in character features of excellent consecutive behavioral ability. Taking this further, and combining it with our finding of low impulsiveness in low HA offenders and elevated impulsiveness in high HA offenders we argue that low HA offenders are prone to commit premeditated violent crimes (murders in our sample) whereas impulsive violence (assaults and batteries) connects to the high HA offenders.
The temperament features of the low harm avoidance subpopulation springs a hypothesis that low HA may correlate with high factor-1 scores in Hare's Psychopathy Checklist-Revised , as these seem to have many nuances in common: (1) gregariousness/vigor (TPQ) versus glibness/superficial charm (PCL-R), (2) uninhibited optimism versus a grandiose sense of self-worth/confidence, and (3) a carefree attitude versus shallow affect. Moreover, we observed a high resemblance in the total temperament profiles of controls and low HA offenders implying that low HA offenders may respond to stimulus in a relatively normal manner. Studies combining TPQ and PCL-R are necessary to test this hypothesis.
Our observation of high harm avoidance in these alcoholic offenders with severe personality is in line with research that suggests an association of a broad spectrum of specific personality disorders with high HA temperament [17, 18], and recently with comorbid alcoholism and ASPD, as well [10, 11]. Preliminary genetic polymorphism discoveries in alcohol research show that GABRA2 haplotype-  and galanin haplotype  frequencies differ in low and high HA alcoholics. In this study, however, the HA extremes showed no difference in clinical alcohol dependence but rather a difference in severe impulsive cluster B psychopathology. All offenders in this study suffered from an alcohol usage disorder, in contrast to alcohol studies in which alcoholics are compared to non-alcoholics. Severe impulsive cluster B personality disorders and alcoholism may have an additive association to deviation in trait HA.
In line with Cloninger's hypothesis, we found that trait novelty seeking was higher in ASPD than in non-ASPD offenders. Similar results have emerged for substance users , and psychiatric patients with cluster B disorders . Moreover, ASPD offenders showed particularly high impulsiveness (NS2) and disorderliness (NS4), whereas non-ASPD offenders corresponded to the control figures in these dimensions. These associations are comprehensible since the TPQ surveys features that are similar to the features surveyed in the ASPD and BPD sections of the structured interview for DSM-III-R . All 8 NS2 items survey impulsiveness with questions such as "I often act on hunches, momentary whims, or by intuition without making a detailed analysis of facts" . The 10 NS4 items survey issues such as breaking rules, lying, and losing one's temper . Impulsiveness (NS2) and disorderliness (NS4) are likely to be linked to biologic correlates of ASPD. Brain imaging studies strongly suggest that among habitually violent offenders with ASPD, impulsive aggressive behavior is associated with decreased glucose uptake in the frontal cortex (reviewed by Bufkin & Luttrell, 2005 ), and that a low brain serotonin turnover, indicated by low cerebrospinal fluid 5-hydroxyindoleacetic acid (CSF 5-HIAA) levels, is also connected with impulsive aggression [19, 20]. Our preliminary observation of NS1 suggests that high HA offenders are more indifferent to stimulus than the low HA offenders.
Examining the third dimension, ASPD offenders featured low reward dependence and considerable detachment (RD3) compared to controls in agreement with Cloninger's hypothesis, but no difference to offenders lacking ASPD emerged. Trait harm avoidance, however, distinguished offenders with higher reward dependence (low HA offenders) from their counterparts.
To the best of our knowledge, this is the first systematic study of temperament traits in alcoholic violent offenders in relation to categorical ASPD diagnosis and the type of violent offense. The strengths of this study comprise its large sample size, accurate diagnoses, and simple study design. The limitations, on the other hand, were that participants were a selected group of males who featured substantial psychopathology. Consequently, results may not be generalized to other patient groups such as the ASPD patients with infrequent or unregistered minor property crimes. Secondly, we chose the conservative Kruskal-Wallis and Dunn's tests for variance and pairwise comparisons due to the non-normally distributed data and unequal group sizes, which left potential confounding factors uncontrolled for (for instance age). However, our population is markedly homogeneous compared to community based settings were demographic data is more relevant. According to Cloninger's original proposal, the TPQ dimensions are hereditary, and stable basic stimulus-response characteristics, which, being true, also decreases the risk for bias due to demographic differences. Another source of bias is the temperament figures of our control group since curious and novelty-seeking individuals may be over-represented in control groups attracted by newspaper advertisements , but still, the TPQ scores were in line with the results of a large psychometric TPQ testing project in the Finnish population . Inter-rater agreement analysis was not performed, which is a limitation even though we consider the SCID administration in the unit as well standardized due to frequent administration by experienced psychiatrists and the diagnoses were double-checked by other psychiatrists. Moreover, ASPD offenders were more frequently alcohol dependent than non-ASPD offenders. On the other hand the non-ASPD offenders who lacked an alcohol dependence diagnosis suffered from severe alcohol abuse. This difference in alcohol-disorders could be a source of bias even though the overall contrast to the controls was robust, as any substance usage disorder was an exclusion criterion for controls. Finally, every fifth offender chose not to complete a TPQ (resulting in the sample of 198 offender analysed in this paper) and thus, returned questionnaires could be biased due to the desire for judiciary benefit. The non-returnees had no obvious differences compared with other offenders in this homogeneous sample.