Large Experiment to Test Intertype Relations

RESULTS (Part 1)

We begin presenting the results of the experiment testing intertype relations, which has already been ongoing for more than a year. All data were collected by V. L. Talanov using 11 of his questionnaires from series 065-072 and 078-080. The data were collected by means of paired socionic testing, in which both members of a pair, connected to each other by blood kinship, purely friendly relations, or romantic/intimate relations, completed a socionic diagnostic questionnaire of the same form.

Taking into account whether the members of a pair were of the same or different sex, the questionnaire survey yielded the following 6 groups of participants, combined into ordered pairs:

1. 55 same-sex blood-relative pairs
2. 37 opposite-sex blood-relative pairs
3. 165 same-sex friendship pairs
4. 101 opposite-sex friendship pairs
5. 48 same-sex romantic/intimate pairs
6. 257 opposite-sex romantic/intimate pairs

The ordering of the pairs consists in the fact that all questionnaires of pair members submitted as number 1 were assigned to the first subgroup of each group, and corresponding to them are questionnaires from the respective pairs submitted in the letters of anonymous respondents as number 2 and forming the second subgroup (male questionnaires in opposite-sex pairs were always submitted, under the typing conditions, as number 1)

15 socionic traits were measured in all participants in two different ways.

Under the first method (“type-based method”), correlations of the respondent’s answers with the diagnostic vectors of 16 psychotypes are first measured; these correlations (more precisely, their Fisher transformations) form the type profile, from which the magnitudes of all traits are calculated – the trait profile, and from the trait profile the functional one is calculated for the magnitudes of 12 functions in arbitrary score units.

Under the second method (“trait-based method”), correlations of the respondent’s answers with the diagnostic coefficients of 15 socionic traits are first measured (the coefficient vectors are strictly orthogonalized). The resulting correlations (their Fisher transformations) constitute each respondent’s trait profile, and the type and functional profiles are then calculated from the trait profile.

Correlations between trait values of pair members can be measured within each participant group as the correlation between the series of values of a particular trait in the first subgroup of participants in that group and the series of values of the same trait among the corresponding members of the second subgroup of participants (submitted in pairs as number 2). In this case, we speak of the magnitude of trait correlations in ordered pairs of participants.

But the two subgroups of participants can be combined and the correlation of trait values within their pairs can be measured in series of values from two sets of respondents, one of which is composed of subgroups 1+2, and the other – respectively, of subgroups 2+1. In this case the number of experimental points doubles (although the correspondence between pair members is not disrupted, but in only half of the experimental points the pair members exchange places according to their number: 1 changes to 2 and vice versa), and in this case we will speak of measuring correlations in unordered pairs of participants.

Today we will begin by presenting the results of correlations of trait magnitudes within pairs (which already gives an idea of the preferred intertype relations in pairs of various kinds).

Next time we will consider correlations between the corresponding mental functions of pair members, then - the actual varieties of statistically preferred intertype relations calculated in different ways, and so on.

Fig. 1 shows trait correlations between pair members in all 6 groups of respondent pairs. The correlations are calculated using respondents’ traits recalculated from their type profiles.

Fig. 2 – the same, but the traits from which the correlations between pair members are calculated are measured directly by the questionnaire, rather than by recalculation from type profiles.

Fig. 3 separately examines correlations in opposite-sex romantic pairs, with the sample of pairs divided into subgroups of socionically naive and socionically experienced participants, and also into those with a short relationship duration (up to 2 years) and a long relationship duration (over 2 years).

Figures 4, 5, 6, and 7 show intercorrelation matrices of traits in different participant datasets. Figs. 4 and 5 – in the dataset of participants of different sexes connected by romantic relations. Figs. 6 and 7 – in the dataset of participants connected by purely friendly relations (in this case respondents from both same-sex and opposite-sex pairs are included).

CONCLUSIONS:

1. Although perfectly orthogonalized vectors of diagnostic coefficients are used in the questionnaires to diagnose 15 socionic traits, ultimately nonorthogonality re-emerges in the array of resulting trait profiles (see Figs. 4, 5, 6, 7) (moreover, without prior orthogonalization of the diagnostic vectors it would have the same sign, but an even larger absolute magnitude). This phenomenon concerns not only the examined samples of paired questionnaires, but also other, much larger samples composed of questionnaire respondents. The phenomenon itself is important, but we will not analyze its possible causes here. We draw attention to it because correlations of weak traits with strong ones (even comparatively weak correlations) can substantially contaminate the variance of the corresponding weak trait with the variance of another, strong trait. Thus, if a weak trait has a sigma (spread of measured values) three times smaller than that of a strong trait, and exhibits a correlation of about 0.33 with this strong trait, then the parasitic contribution of the variance of the strong trait to the variance of the weak trait will be exactly equal in magnitude to the variance of the weak trait itself. If, with the same correlation magnitude, the sigma of the weak trait is 4-5 times smaller than the sigma of the strong trait, then the influence of the strong trait will already predominate in the obtained results, completely masking the role of the weak trait. This may, among other things, distort the results shown in Figs. 1, 2, 3.

2. The principal traits Extraversion, Intuition, and Logic show clearly negative correlations between pair members in the romantic-pair group, whereas in the opposite-sex friendship-pair group (and in the same-sex group – even more so) these correlations are much smaller and close to neutral zero. The value traits Judiciousness, Merry, and Democracy show a positive correlation between pair members in the romantic-pair group (as expected for Duality and other intraquadra relations). Among all friendship pairs, these same value traits also have positive correlations (only slightly smaller in magnitude), which indicates a tendency toward the same polarity of all value traits among members of friendship pairs, that is, a tendency toward intraquadra relations. However, among friendship pairs the absolute correlation is maximal specifically among the value traits, and this indicates a tendency toward Identity rather than Duality relations.

3. Statics behaves among romantic pairs as expected for a tendency toward Duality relations – that is, the Statics-Dynamics correlation among members of romantic pairs is negative. In friendship pairs it tends toward zero, although it retains a negative sign.

4. Less significant traits, such as Constructivism (negative correlation) and Process (weakly positive correlation), also correspond to a tendency toward Duality relations among romantic pairs. However, a number of traits behave in an “incorrect” way, refuting this tendency. These include Questimity, Positivism, Tactics, and Yielding. It should, however, be noted that the magnitudes of these correlations, first, do not reach the 5% significance level. Second, their paradoxical values may be explained by their parasitic correlation with strong value traits (this question requires additional mathematical investigation).

5. There is no difference in the sign of the resulting correlations between subsamples of socionically naive and socionically experienced pairs (see Fig. 3). The difference in the absolute magnitudes of correlations in these subsamples is explained not by differences in the level of cognitive distortions between these two subsamples, but merely by the different reliability of the trait profiles obtained in these subgroups (among the socionically naive, all reliability indicators are lower on average).

6. There is no noticeable fundamental difference in the pattern of trait correlations between subsamples composed of pairs who have known each other for a long time and those who have known each other recently. Moreover, romantic pairs with a short relationship duration tend even to correspond slightly better to the trait pattern of Duality relations (greater agreement in Irrationality and Judiciousness among pair members, while Constructivism and Tactics more often take opposite pole signs among pair members, which again corresponds better to Duality relations)

7. Examining blood-relative pairs for a hereditary factor turns out to be a rather meaningless exercise. There are two reasons for this. First, the size of the obtained sample is too small. Second, the sample is formed from volunteers – in this case one should expect blood relatives to submit their questionnaires for simultaneous consideration mainly in those cases when there is either deep mutual understanding between them, or dual attraction, or, conversely, a deep conflict seeking psychological assistance from outside. This second reason, unfortunately, makes the sample inherently invalid for analysis of the genetic component on the basis of the obtained statistics.

8. Comparison of the groups of romantic and friendship relations permits the assertion that a substantial predominance of Duality relations is observed only among romantic pairs. At the same time, existing socionic theoretical concepts make no distinction between romantic and friendship relations, prescribing predominant duality equally to friendship relations. Since this is not observed in the experiment, this makes it possible to conclude that the results are in no way explained by cognitive distortions (suggestive pressure from socionic theory). The same is evidenced, as we indicated above, by the absence of fundamental differences in results between subgroups of socionically experienced and socionically naive pairs, as well as by the fact that pairs with a longer duration of living together do not at all demonstrate deeper manifestations of duality (which would be expected if cognitive distortions had any appreciable influence on the results).

Fig. 1. Trait correlations between pair members in 6 groups of respondent pairs

Ordered pairs. Recalculation of traits from type profiles

Unordered pairs. Recalculation of traits from type profiles

Group legend:

1. same-sex blood-relative pairs
2. opposite-sex blood-relative pairs
3. same-sex friendship pairs
4. opposite-sex friendship pairs
5. same-sex romantic/intimate pairs
6. opposite-sex romantic/intimate pairs

Fig. 2. Trait correlations between pair members in 6 groups of respondent pairs

Ordered pairs. Direct measurement of traits

Unordered pairs. Direct measurement of traits

Group legend:

1. same-sex blood-relative pairs
2. opposite-sex blood-relative pairs
3. same-sex friendship pairs
4. opposite-sex friendship pairs
5. same-sex romantic/intimate pairs
6. opposite-sex romantic/intimate pairs

Fig. 3. Trait correlations between members of romantic pairs

Direct measurement of traits. Unordered pairs

Direct measurement of traits

Fig. 4. Intercorrelation matrix of traits. 514 respondents from opposite-sex romantic pairs

Recalculation of traits from the type profile obtained as direct correlations with types

Empirical number of degrees of freedom: 9.562

Fig. 5. Intercorrelation matrix of traits. 514 respondents from opposite-sex romantic pairs

Direct measurement of traits (trait profile as correlations of questionnaire responses with vectors of trait coefficients)

Empirical number of degrees of freedom: 9.630

Fig. 6. Intercorrelation matrix of traits. 532 respondents from friendship pairs (both same-sex and opposite-sex)

Recalculation of traits from the type profile obtained as direct correlations with types

Empirical number of degrees of freedom: 8.718

Fig. 7. Intercorrelation matrix of traits. 532 respondents from friendship pairs (both same-sex and opposite-sex)

Direct measurement of traits (trait profile as correlations of questionnaire responses with vectors of trait coefficients)

Empirical number of degrees of freedom: 8.643