Many motivational theories are built on the analysis of correlations between variables, factors analyses, structural equation models, and related approaches. These ‘variable-oriented’ statistical approaches yield many insights, but they do not answer to all research questions. What analyses of inter-individual co-variation do not show, is whether the relations between variables describe the experiences of real individuals. For instance, the finding that harmonious passion is related to positive and adaptive experiences while obsessive passion is related to negative and maladaptive experiences does not imply that individuals experience alternatively a dominating harmonious, or a dominating obsessive passion, and it also does not imply that you are well-off if your harmonious passion is higher than the harmonious passion score of most other people.
Recently, an increasing number of publications in the field of motivational psychology have reminded of the fact that the analysis of profiles and intra-individual variation leads to often different conclusions than the analysis of correlations across a sample of individuals. These articles often emphasize that we need to examine the intra-individual variation of constructs in order to describe the experiences of individuals, of groups of individuals, and in order to be able to diagnose, predict, and treat individuals. A good overview about the past and future directions of such so-called person-oriented analyses was provided in the first issue of the recently founded open-access Journal for Person-Oriented Research.
The previous research on passion in psychology was to a big part built on ‘variable-oriented’ analyses of correlations between variables across individuals. In contrast, the conclusions drawn in this field of research often refer to differences between individuals (the mainly harmonious people versus the mainly obsessive people), or to the profiles of harmonious and obsessive passion within individuals (having a dominating harmonious passion, versus having a dominating obsessive passion). We often read or hear statements such as ‘if you have a harmonious passion, then you are lucky, probably happy and in control of your passionate activity’ or ‘people with a harmonious passion loose control over the activity and develop an uncontrollable urge to persist when they should let go’. Such statements require analyses of profiles, cluster analyses, but do not necessarily follow from the well-established finding that harmonious and obsessive passion are distinct variables with different outcomes.
To find out if these statements would hold true when examining profiles and clusters, my colleagues Melanie Keiner (University of Erfurt, Germany), Robert Grassinger (University of Augsburg, Germany) and I recently conducted several analyses of profiles of harmonious and obsessive passion, and other analyses of distinct subgroups of individuals. The results were quite surprising! First, we found that the results of cluster analyses suggested very different interpretations than the so-far known correlation-based analyses. There were virtually no individuals with higher obsessive than harmonious passion, but instead, harmonious passion was higher for almost everyone in each of our four samples. Furthermore, most individuals reported either aligned high scores of both harmonious and obsessive passion, or aligned low scores of both passion variables. In other words, most probably you won’t get the downsides of passion (obsession) without the upsides, which is a good message for everyone who was worried about “How can we help people with an obsessive passion?”.
Another interesting finding of our study more generally regards the usage of z-scores in the analyses of profiles and group differences. We found three reasons why z-scores might be misleading in the analyses of profile differences:
- Sometimes z-scores suggest that response A was higher than response B, even though it was the other way around when the person answered to the original response scale.
- z-scores can also suggest that an answer was “high”, even though this answer negated the item statement on the original response scale (=was a rating below the scale midpoint on a scale from 1 = don’t agree at all to 7 = totally agree).
- Plotting profile differences using z-scores often equals plotting a graph with a truncated Y-axis, because rarely the whole area of possible answers is shown in such z-score graphs.
- When we presented our findings at the SELF Conference in Kiel this year, Olaf Köller added the argument that z-scores are very sample-specific. The same z-score can represent widely different raw scores (and thus degrees of item affirmation) in different samples. If the sample is representative for the population, this might not be a problem, but in most studies samples are not representative in all relevant aspects, and then z-scores are not comparable across samples (although many people would think they are).
The paper has been published in the latest issue of the Journal for Person-Oriented Research and can be downloaded here (open access).