مدارک و شواهد برای ساختار صفت پنج عامل بزرگ شخصیتی در سازمان حافظه

The extent to which the “Big Five” personality trait factors guide organization in free recall was examined. Unlike previous studies, which failed to find effects, this study used traits with factor purity as stimuli. Significant levels of clustering in free recall around the Big Five factors were found. In addition, this level of clustering was compared with that found for common object nouns and personality type nouns.

The “Big Five” approach currently dominates the study of personality trait structure (Allport & Odbert, 1936; see John & Srivastava, 1999, for an overview). An impressive amount of research suggests that personality traits can be organized into the five dimensions of extraversion–introversion, neuroticism, agreeableness, conscientiousness, and openness to experience. These dimensions were uncovered primarily by factor-analyzing trait ratings. This strategy has proven quite robust – the five dimensions have been obtained across different types of targets (e.g. ratings of self, ratings of other people; McCrae & Costa, 1987), in a variety of languages and cultures (McCrae & Allik, 2002), and across different factor analysis methods (Goldberg, 1990). Smith and Kihlstrom (1987) postulated that the ubiquity of the Big Five in trait ratings suggests that these factors might serve as conceptual structures (i.e., natural categories; Cantor & Mischel, 1979) that people use when perceiving others. If so, the individual trait concepts contained in memory should be associated with each other according to their degree of similarity to the five dimensions, and these structures should guide person perception (that is, that they function as schemas; Smith & Kihlstrom, 1987). Is there evidence for this assumption? The factor analysis of trait ratings provides one type of evidence, of course. However, there are other ways that one might estimate the cognitive organization of personality traits. For instance, research has examined the perceived similarity of traits (e.g. Rosenberg & Sedlak, 1972). These studies suggest that traits can be ordered in a conceptual space defined by the dimensions of evaluation (which can be split into social and intellectual aspects) and dynamism, which is akin to the activity or strength implied by the trait. The results of this research are not entirely consistent with the five dimensions uncovered by the trait rating approach, although the five factors have been obtained using this methodology (D’Andrade, 1965). Smith and Kihlstrom (1987) conducted several studies examining the possibility that the Big Five function as schemas by using methods designed to reveal schema-like effects. They found evidence for the five factors using a conceptual-similarity methodology in which people chose traits they thought were likely to be true of a person who scored highly on one of the Big Five dimensions. They also found evidence for the Big Five using an illusory correlation-based method. However, further analyses indicated that the results of these two studies were driven by a relatively small set of traits. More importantly, Smith and Kihlstrom attempted to find evidence for trait organization around the Big Five by assessing clustering in free recall but failed to find such clustering in two studies. Dabady, Bell, and Kihlstrom (1999) also examined the influence of the Big Five on clustering, this time using modified items from the NEO-PI-R (Costa & McCrae, 1992) as stimuli, but they also failed to find clustering around the Big Five categories. The failure to find evidence for the Big Five in clustering in free recall is disconcerting, as a large amount of research from cognitive psychology suggests that mental organization should be reflected in such measures (see, e.g. Puff, 1979). Indeed, these measures are a primary way in which cognitive organization is shown. Combined with the equivocal evidence from the other techniques used by Smith and Kihlstrom (1987), this casts doubt on the extent to which the Big Five function as cognitive structures. Perhaps this is not surprising. The idea that the Big Five constitute the fundamental dimensions of personality has its critics (e.g. Block, 1995). Furthermore, even if the Big Five do constitute such dimensions, it may be that people’s mental organization of traits does not reflect this. Perhaps people simply do not mentally organize personality traits in terms of the Big Five. It is possible that an aspect of the procedure used by Smith and Kihlstrom (1987) and Dabady et al. (1999) limited the power of these studies to detect the predicted effects. For a relatively powerful test of cognitive organization, the stimuli should be strongly representative of a target category and not linked to other tested categories. That is, each stimulus should have a mental association with only one of the categories under examination. So, for instance, a researcher that is interested in the extent to which traditional animal categories (e.g. birds, mammals, insects) can organize memory for animals would be better off choosing “lions” as a representative of the category of mammal than “bats”, which have birdlike qualities and therefore probably have mental associations with both mammals and birds. For stimuli derived from factor analysis, as is the case for the Big Five trait structure, this issue corresponds to the problem of “factor purity” – the extent to which each item loads strongly on one and only one factor. Traits that load on more than one factor suggest that the raters have a mental “link” between that item and more than one category. In a memory study examining organization according to these categories, items loading on more than one category make it unclear which category should be guiding clustering for that item (and therefore what constitutes an a priori within-category repetition). Just how factor (im)pure were the items used in the Smith and Kihlstrom (1987) and Dabady et al. (1999) studies? Saucier (1994) suggests that an appropriate criterion for factor purity is that an item load highly on the relevant main factor and have loadings on that factor that are more than twice its loading on other factors. Smith and Kihlstrom (1987) used as stimuli the traits identified in Norman’s (1963) seminal Big Five study. An examination of factor-loading tables for the Norman items for three different participant samples (Norman, 1963 and Passini and Norman, 1966) reveals that out of the 20 bipolar trait scales, 6, 7, or 8 (depending on the participant sample) do not meet this criterion. In some instances, the Norman items actually have higher loadings on alternative factors than they do on the predicted factor for some participant samples (e.g. “polished-crude” loads more highly on Conscientiousness than it does on the a priori category of Openness). The status of the sentences used by Dabady et al. (1999) is less clear with regards to factor purity. The items used were modified items from NEO-PI-R (Costa & McCrae, 1992). Costa and McCrae (1992) report associations between the five factors and trait facets, but not for individual items. However, since the facets are composed of small numbers of items, it is likely that the facet-level statistics are informative with regards to the factor purity of the NEO. Out of the 30 NEO facets, 12 of them do not meet Saucier’s factor purity criterion. Therefore, roughly a third of the Norman items used by Smith and Kihlstrom were not factor pure, and most likely a similar number of the NEO items used were not as well. It might be argued that the lack of factor purity in these items is an indication that the Big Five are a poor candidate for an overarching conceptual structure of traits. However, the fact that people’s categories are “fuzzy” – that some category members are seen as better exemplars than others and that some members show overlap with other categories – does not mean that the category is not important for the person. “Fuzziness” is probably the case for most if not all natural categories (Rosch & Mervis, 1975). The main goal of the current studies was to find evidence in the organization of free recall that the Big Five serves as an organizing structure for personality traits. Efforts were made to ensure that the traits used as stimuli loaded strongly on their primary factor and weakly on all other factors; that is, that they showed factor purity.