B&A argue that some people are consistent on some traits, but not on others. For example, I may be trait like on the personality dimension of aggressiveness, but situational with regard to friendliness. Another person may be the exact opposite. So B&A suggest that the most that we can do with personality variables is PREDICT SOME OF THE PEOPLE SOME OF THE TIME (i.e., only on those traits on which they are consistent).

Researchers who are firm believers in trait theory have a priori hypotheses that 1) a variety of behaviors cluster into a unitary class of responses which reflect a trait and 2) that these generalize across situations. For example; hitting, yelling, screaming, biting, kicking, and having a red face would all be considered to be a unitary class of responses reflecting the trait of AGGRESSION. These behaviors would be expected to be displayed by people high in trait-aggression in relevant situations: Frustration, pain, insult, etc. Researchers design their studies from this frame of reference. B&A question whether research subjects share this philosophy of the researchers. B&A say that researchers must relinquish this omniscience to the subjects and let the subjects define which traits and behaviors are relevant for them. In an attempt to find subjects who were consistent on some traits, B&A had Stanford undergraduates involved in a semester long study where the following data were recorded:

1. A 24 item friendliness scale

2. A 24 item conscientiousness scale

3. “In general, how friendly and outgoing are you?” 7 point scale: Not at all to extremely.

4. “In general, how conscientious are you?” 7 point scale: Not at all to extremely

5 & 6 “ How much do you vary from 1 situation to another in how (Friendly or conscientious) you are. NOTE THAT 5 & 6 ARE THE MAIN MODERATOR VARIABLES IN THE STUDY. ALSO NOTE THAT 1 & 2 INVOLVE EXPERIMENTER DEFINITIONS OF THE 2 TRAITS. 3,4,5 & 6 INVOLVE SUBJECT DEFINITIONS OF THE MEANING OF THE 2 TRAITS.

B&A also got reports from the subjects mom, dad, and a peer on the 2 24 item scales and the 2 global ratings. There were also the following behavioral measures: FRIENDLINESS-- Group discussion, and Spontaneous friendliness; CONSCIENTIOUSNESS--Returning course evaluations, Course readings, and Neatness.

Subjects in the study were divided into two groups (by median split) on each trait as being highly variable or consistent.

RESULTS:

FRIENDLINESS. The average intercorrelation among friendliness measures was .57 for the self described consistent folks and .27 for the self-described high variability folks (sig). 13 of the 15 comparisons were in the predicted direction, with 6 of them significantly different. Quite a few individual correlations exceeded the “personality coefficient” of .3 to .4. It is important to note that the correlation between experimenter defined friendliness (the 24 item scale) and subject defined friendliness was quite large r = .84, accounting for 70% of the variance. There was much less agreement between subject defined conscientiousness and the experimenters 24 item scale, r = .62, accounting for 38% of the variance. B&A believe that this is the reason that the Conscientiousness data did not show the same pattern as the Friendliness data.

B&A then re-analyzed the Conscientiousness data using a different moderator variable. B&A calculated high versus low variability on Conscientiousness on the basis of answers to the 24 item conscientiousness scale: After doing this, the results for Conscientiousness replicated those for Friendliness. 15 out of the 21 comparisons of high and low variability subjects were in the predicted direction, with 9 of them significant. This pattern held except for the variable of neatness: For low variable subjects neatness did not correlate with any other measure. For high-variability subjects, neatness correlated -.61 with course readings, suggesting that these subjects either had time to keep things neat and clean or do school work, but not both together. Obviously the subjects are saying that school related conscientiousness (returning evaluations) and personal neatness do not belong to the same “equivalent” class of responses, which the experimenters had assumed in designing the study.