RESULTS

Subjects were dichotomously classified as pedophilic or not pedophilic according to one of two similar rules, depending on which version of the phallometric test they had

[Page 516]    

undergone.

Very similar proportions of subjects were classified as pedophilic by the two tests: 33.5 and 36.5%, for the nine-and five-category tests, respectively.

Because these are z-transformed data, 0.00 on the Y-axis represents the grand mean of responses to all stimulus categories, not the absence of erotic response. The absence of erotic response is located by the mean response to the neutral stimulus category.

These figures show that the non-pedophilic subjects primarily responded to adult women on the phallometric test. These data are consistent with the breakdown of known sex offenses given earlier, which indicated that relatively few patients in this sample were referred because of sexual behavior involving adult men.

Table I shows the responses of pedophilic and non-pedophilic subjects to the EPES questionnaire item, “Before you were age 6, did you ever have an accident which left you unconscious for at least half an hour?” The sample size is 899 rather than 902 for the nine-category subjects be-cause 3 subjects with the nine-category test did not answer this item.

The table shows that among the subjects who had the nine-category test, 10.6% of the pedophilic subjects indicated that they had experienced an accident resulting in unconsciousness before age 6, compared with 4.5% of the non-pedophilic subjects.

The results of a chi-square test comparing these proportions, which are presented in the table notes, indicate that the difference was statistically significant.

The corresponding percentages for the subjects who had the five-category test were 9.0 and 4.1%.

Although these results are similar to those for the subjects with the other phallometric test, they did not quite reach statistical significance in the (two-tailed) chi-square test (p = .08), probably because the sample size was much smaller.

[Page 517]

In the combined group of all subjects, 10.2% of pedophiles and 4.4% of non-pedophiles reported an accident before age 6 that left them unconscious. The difference was statistically significant.

Subjects’ responses to the questionnaire item, “Between the ages of 6 and 12, did you ever have an accident which left you unconscious for at least half an hour?” are shown in Table II. 

In this case, the percentages of pedophiles and non-pedophiles responding in the affirmative differed significantly for the five-category group (9.9 and 4.1%, respectively), but not for the nine-category group. The difference was also significant for the combined sample.

Table III presents the results for a third, composite variable: Whether the subject responded “Yes” to either of the previously mentioned items — in other words, whether the subject ever had an accident resulting in unconscious-ness prior to age 13. 

Positive histories were significantly associated with pedophilia in subjects with both phallometric tests and in the combined group.

In Tables  IV  and  V  are presented the results for the questionnaire items, 

Table VI shows the results for the composite variable: Whether the subject responded “Yes” to either of the preceding items — that is, whether the subject ever experienced a period with frequent losses

[Page 518]

of consciousness prior to age 13. Hardly any subjects reported such a history prior to age 6, and the somewhat greater number of subjects who reported this between ages 6 and 12 were about equally likely to be pedophiles or non-pedophiles. None of the statistical comparisons for these three variables was statistically significant.

The analyses for the four items 

that we initially targeted did not strongly suggest that the findings for child-hood accidents occurred because the pedophiles were, for some unknown reason, prone to claim pathology of all types, or because the pedophiles answered this section of the questionnaire carelessly, checking the first response-option of each item regardless of its content. 

(For all four items, the first response option was “yes” and the second was “no or don’t know.”) 

Nevertheless, we decided to look further for evidence of such response biases.

The section of the EPES that contained the four items regarding childhood unconsciousness included four other  

[Page 519]

items that pertained, in a general way, to potential neuro-developmental correlates of adult sexual behavior.

Three of these preceded, and one followed, the four items on childhood unconsciousness. 

These items had parallel response options: 

The third item  

was, “Do you know anything about your birth?” This item offered five response options: 

The fourth item 

(which followed the unconsciousness items) asked, “When did you stop wetting your bed?” This item had six response options: 

We investigated these items to ascertain whether pedophiles were differentially prone to endorse first-listed response-options or pathology-claiming response-options.

None of the items concerning father’s psychiatric history, obstetrical complications, or bed-wetting produced any statistically significant results. A summary of the findings for these variables is presented in Table VII.

The one variable that produced statistically significant results was the item concerning the mother’s psychiatric history.

[Page 520]  

The data are presented in Table VIII.

The results, which suggest that the mothers of pedophiles were more likely to have undergone psychiatric treatment, were similar for the subjects with the nine- and five-category tests, but they were statistically significant only for the former, probably because of their larger sample size. The results were also statistically significant for the combined subjects.

The foregoing item was investigated more thoroughly with a logistic regression analysis, which used the combined group of all subjects. The criterion was the subject’s phallometric diagnosis (coded “1” for pedophilic and “0” for not pedophilic). 

The predictors were dummy variables representing a set of Helmert contrasts. 

(This type of contrast is a standard option for various statistical procedures in numerous software packages. The present authors used SPSS, Version 10.1.) 

Our Helmert contrasts were designed for two comparisons: 

The results of this analysis are presented in Table IX.

In Table IX, each coefficient B represents the change in the log odds of pedophilia associated with the corresponding contrast. The next column presents the standard error (SE) for each B The Wald statistic was the quantity used to determine the significance level (p) of each contrast. The quantity e^B is the multiplicative change in the odds of pedophilia for a given contrast, and thus 100 x (e^B - 1) represents the percentage change in the odds for that contrast.

The results show that there was no difference in the odds of pedophilia between the subjects who endorsed the response-option “No, or not sure” and those who endorsed “Know little or nothing about your mother.” 

However, those subjects who endorsed the “Yes” option, indicating that their mother had undergone psychiatric treatment, were over twice (i.e., 216%) as likely to be pedophilic as those subjects who endorsed one of the negative options (e^B = 2.16). This result was statistically significant.

In summary, 

the foregoing analyses did not provide evidence that the results for the childhood accident items were an artifact of response bias. They did, however, yield evidence that pedophilia might be related to maternal psychiatric disorders, in addition to whatever factors were reflected by the unconsciousness items.

The uni-variate analyses for the items regarding accidents before age 6 and accidents between 6 and 12 suggested that accidents in the earlier age range were more strongly associated with pedophilia. 

We investigated this further 

in another logistic regression analysis, in which the criterion variable was again the subject’s phallometric diagnosis. The predictor variables were the two accident items, with the subjects’ responses coded “1” for “Yes” and “0” for “No or don’t know.” 

The results of this multi-variate analysis are presented in Table X.

In Table X, each coefficient B represents the change in the log odds of pedophilia associated with a “Yes” response to the corresponding item, controlling for the subject’s response to the other item. The quantity e^B is the multiplicative change in the odds of pedophilia associated with a “Yes” response, and 100 x (e^B - 1) represents the percentage change in the odds associated with a “Yes” response.

The results indicated that controlling for accidents at the later age, accidents at the earlier age more than doubled

[Page 521]

the odds of pedophilia (e^B = 2.30, that is, the odds increased 230%). This association was statistically significant.

In contrast, accidents at the later age, controlling for accidents at the earlier age, did not significantly increase the odds of pedophilia.

The finding that pedophilia was associated with accidents in earlier rather than later childhood prompted us to investigate whether other signs of problematic neuro-development would also be associated, in this sample, with accidents in earlier rather than later childhood. To test this, we used two linear regression analyses. In one of these, the criterion variable was the previously described interviewer rating of the subject’s intelligence, and in the other, the criterion variable was the subject’s self-reported educational level. The predictor variables were the same in both analyses — accidents causing unconsciousness before age 6, and accidents causing unconsciousness between the ages of 6 and 12 — coded as described above. 

These analyses were run on the subjects phallometrically diagnosed as pedophiles (on either test), on the subjects diagnosed as non-pedophiles, and on the combined group of all subjects.

The results are given in Table XI.

The results were similar for the pedophiles and the non-pedophiles. They indicated that both lower intelligence and lesser educational achievement were more strongly associated with childhood accidents before age 6 than with childhood accidents between the ages of 6 and 12.

Finally, 

we examined the relations between the subject’s intelligence rating and educational level, on the one hand, and his mother’s history of psychiatric treatment, on the other. For this analysis, the subjects’ responses to the maternal psychiatric treatment item were coded “1” for “Yes” and “0” for either of the two types of negative responses. 

There were 405 pedophiles and 791 non-pedophiles with complete data for mother’s psychiatric history and subject’s rated intelligence, and there were 410 pedophiles and 792 non-pedophiles with complete data for mother’s psychiatric history and subject’s education.

None of these correlations was statistically significant.