Path Analysis: Too Young for Intimacy?

Shelly spends all her time with friends—if not with them at school, then on the phone. Shrieks of laughter and silent smiles punctuate their conversations. It’s clear who’s “in” and who’s not when they’re with others. Shelly is 16. Her life revolves around her friends.

Yet some experts would question how close Shelly really is to her friends. These developmentalists argue that adolescents can form intimate relationships only after they have established a stable identity—a task several years beyond this 16-year-old. Erik Erikson assumes that adolescents must resolve the psychosocial crisis of identity before they can become intimate with others—that is, that intimacy is contingent on identity. Other theorists, such as Carol Gilligan and Ruthellen Josselson, argue that Erikson’s developmental model fits the experience of males better than that of females. These theorists note that females’ interpersonal skills prepare them to define themselves through their relationships with others—that is, that intimacy contributes to identity. Is the connection between identity and intimacy different for adolescent males and females? How might we tell?

Couldn’t we simply measure identity achievement and intimacy in a group of adolescents and see whether those who have a better sense of themselves also have closer relationships with others? Let’s say we do and discover that our assumption was correct. Does this finding support Erikson? Gilligan and Josselson? Actually, we have no way of knowing. There is no way to tell from this single correlation which factor is responsible for the other.

Would it help to separate the adolescents by sex and look at the degree to which the two measures are correlated for each? Not really. Even if we found a stronger relationship for one group than the other, we still would not know for that group which quality contributed to, or caused, the other—that is, whether identity is necessary for intimacy to develop or whether intimacy contributes to the development of identity. All we would know is that adolescents who are high in one attribute are also high in the other, and vice versa.

Path analysis is a statistical procedure that allows developmentalists to infer the direction, or path, of an effect from correlational data. To use this procedure, one must obtain measures for the same variables on more than one occasion. Because causes precede their effects, we need this time difference to trace the direction of the relationship. But how is this procedure any better than a single correlation? We’re still measuring both factors at the same time, just doing it twice instead of once.

That’s true. However, because causes precede their effects, we can look for differences in the strength of relationships that differ only in which factor precedes the other. If Factor A causes Factor B, these factors should be most strongly correlated when measurements for A precede those for B, that is, between Factor A at Time 1 and Factor B at Time 2. The opposite correlation (Time 1 measures of Factor B with Time 2 measures of Factor A) should be relatively weak. The figure shows these correlations as diagonals stretching from Time 1 (T1) measures at the top to Time 2 (T2) measures at the bottom.

But let’s get back to Shelly. Are her friendships likely to be as intimate as they seem, or is their closeness illusory, awaiting further identity development?

Patricia Dyk and Gerald Adams (1990) conducted the study we have been describing (except their subjects were older than Shelly). These investigators found no simple gender difference in the relationship between identity and intimacy: Identity predicts intimacy for both sexes. However, when factors such as gender typing and empathy are taken into account, different patterns emerge for females and males. In males, and in females high in masculinity, identity predicts intimacy. In highly feminine females the association between identity and intimacy is more fused, suggesting that intimacy and identity develop together as Gilligan and Josselson also suggest. Some of these relationships appear in the figure in the source paper.



P. H. Dyk & G. R. Adams. (1990). Identity and intimacy: An initial investigation of three theoretical models using cross-lag panel correlations. Journal of Youth and Adolescence, 19: 91–110.