Personality impression formation: a correlational-

Personality impression formation: a correlational-

Personality impression formation: a correlational- experimental design* JOHN T. PARTINGTON Brock University

LOUISE CLARKE University of Western Ontario


A correlational-experimental design was used to study personality impression forma- tion. Subjects rated how they would accept a number of hypothetical stimulus people represented by combinations of self-referent statements denoting opposite poles of four personality dimensions. They also gave self-endorsement responses to a person- ality battery which sampled the same four dimensions. Perceivers were classified into personality types defining levels of a subject’s factor which together with four stimulus cue factors comprised a complete factorial design for analyzing l ie accep- tance ratings. Although the main experimental results suggested that stimulus cue integration may be more linear than configural, the nature of subject-cue interactions illustrated the need for more representative design in social perception research.

The present investigation focussed on how people form impressions of others based on limited stimulus information. Asch’s (1946) pioneering studies on this problem suggested that people appear to combine stimulus cues configurally and that certain traits are more central than others in determining impressions. However, subsequent work within the Asch paradigm yielded conflicting findings regarding trait centrality (e.g., Kelley, 1950; Wishner, 1960). In addition, considerable evidence favouring more parsimonious linear models of impression formation has also been obtained (Anderson, 1962; Triandis & Fishbein, 1963; Goldberg, 1968), even under widely varied stimulus conditions (Partington, 1967). Not- withstanding this, it is conceivable that die “fit” between such simple models and obtained impression formation data may be a design artifact. That is, the predictive power of these simple linear models may have been a function of “rigorous” experimental paradigms which varied stimulus cues in a controlled manner while holding constant other poten- tially important covariates such as perceiver characteristics, nature of the responses obtained, and situational effects (cf., Bieri, Atkins, Briar, Lea- man, Miller, & Tripodi, 1966).

The purpose of the present study was to determine how different types of perceivers would utilize and integrate information when forming im-

” This research was supported by Canada Council Grant 68-0667 and by the University of Western Ontario Research Fund. Thanks are also due D. N. Jackson for granting permission to use subscales from the Personality Research Form.

CANAD. J. BEHAV. SCI./REV. CANAD. SCI. OOMF., 3 ( 1 ) , 1971


pressions of different types of stimulus people. A combined correlational- experimental design was used (cf., Owens, 1968). Correlational proce- dures were required to classify perceivers into personality types (Nunally, 1967). These types were then included as levels of a “subjects” factor in a complete factorial ANOVA design developed for studying cue utilization in social and clinical judgments (Anderson, 1962, 1969; Hoffman, Slovic, & Rorer, 1968). Since the sets of cues representing the experimental stimuli were selected from the same personality dimensions used to clas- sify the perceivers, this design made it possible to evaluate the relative effects of perceiver-stimuli interactions, as well as the effects of stimulus cues separately and in combination.


Subjects Sixty-two University of Western Ontario undergraduates comprised the original sample. They were selected from the introductory psychology subject pool to avoid possible contamination due to familiarity with measures used in die study. Males and females were equally represented.

Materials A paper and pencil questionnaire was developed which included demographic items, a personality battery, and a social judgment section. The personality battery com- prised three 20 item scales, Dominance, Order, and Affiliation, from the Personality Research Form A (Jackson, 1967), plus a 10 item Masculinity (mf) scale from the MMPI. The social judgment section required Ss to indicate the degree to which they themselves, and others, would accept each of a number of hypothetical stimulus people.

Stimulus People Each hypothetical stimulus person was described in terms of four self-referent state- ments. These statements represented the same four personality dimensions sampled by the personality battery described above. These dimensions are not only representa- tive but also relatively independent (Lay & Jackson, 1969). Table 1 lists the stimulus items together with their operating characteristics. It may be seen that the items were selected according to high denotative relevance, moderate evaluative connota- tions (Mean absolute deviation from neutral: Denotation x = 2.1; Connotation X = 0.75; t = 6.83, p < 0.001), and moderate endorsement frequencies. The hypo- thetical stimulus people were represented by all possible combinations of the stimulus items in a completely crossed factorial design. Thus there was a set of 16 stimuli (2*) since each pole of the four personality dimensions was represented (e.g., Dominance-Submission). Two example stimuli are given below:

1. I go out of my way to meet people; I enjoy reading love stories; I think it is better to be quiet than assertive; I often decide ahead of time exactly what I will do on a certain day. 2. Most of the dungs I do have no system to them; I have relatively few friends; I like adventure stories better than romantic stories; I am quite good at keeping others in line.


TABLE 1 Operating characteristics of the stimulus items








I like adventure stories better than romantic stories

I enjoy reading love stories I often decide ahead of time exactly what I will do on a certain day

Most of the things I do have no system to them I am quite good at keeping others in line

I think it is better to be quiet than assertive I go out of my way to meet people

I have relatively few friends



















Endorsement frequency









Note. Denotative and connotative item weights are 7-point scale averages. Denotative values reported in Partington (1966). Connotative values and endorsement frequencies compiled by D. N. Jackson and A. Conger based on University of Western Ontario samples.

Careful counterbalancing was undertaken to avoid possible intrastimulus and inter- stimulus order effects. The former was accomplished by taking a stratified sample of 16 permutations from the 24 possible permutations of the four personality dimen- sions to be represented. These 16 were then reviewed to ensure that each pole of each personality dimension would occur first and last exactly twice in the person descriptions. Interstimulus order effects were counterbalanced by creating two dif- ferent orderings of the 16 stimuli and nesting one within the other. This resulted in a total of 32 stimuli-to-be-judged.

Dependent Measures Subjects were required to indicate on a 7-point scale how they themselves (Own-ac- ceptance) and how others in general (Others-acceptance) would accept each of die hypothetical stimulus people. The poles of the acceptance scale were defined as fol- lows: “This person would be readily acceptable as a friend;” “This person would be completely rejected.” This social distance continuum was used because of its poten- tial for involving Ss personally in the judgment task. Own-acceptance was die pri- mary dependent measure. The Others-acceptance set was included to provide Ss with the opportunity to express their conception of a “normative” and/or desirable response. This multi-set response format was designed to enhance the probability of obtaining idiosyncratic Own-acceptance judgments relatively free from response styles.


Procedure Subjects were tested in groups of 10-20 each. Several steps were incorporated to re- duce the possibility that Ss might be alienated by die standardized testing procedure (cf., Argyris, 1968) and to lessen possible boredom, fatigue, and defensiveness (cf., Loevinger, 1957). First, instructions were in the form of a client contract (LoveH, 1967) which emphasized reciprocal openness between the £ and S. A realistic set was attempted by encouraging Ss to consider the stimulus people as potentially real, and by requiring Ss to indicate how they themselves would accept each stimulus person “as a friend.” Finally, personal involvement was fostered by requiring Ss to respond to the personality battery before undertaking the social judgments. In addition, to insure optimum freedom for Ss to communicate to the E how they felt about their social judgments, a “Certainty” scale and a space for “Comments” was provided beside each stimulus-to-be-judged.


Subjects were classified into types based on similarities in their own per- sonality profiles, and their impression responses toward a number of hypothetical stimulus people represented in terms of die same dimensions of personality were analyzed.

Classification of subjects’ personality profiles was accomplished by “obverse” principal components of sums of squares and cross-products between subjects (Nunally, 1967). Examination of the distribution of re- sulting eigenvalues revealed a marked drop after the second value. This suggested that the first two “Subjects” factors adequately represented covariation in the original matrix. These factors were rotated to simple structure according to die Varimax criteria (Harman, 1960). Subjects with high loadings (above the median) on the first rotated factor and low loadings (below the median) on the second rotated factor were classified as Type i. Subjects with a reverse pattern of loadings were classified as Type n. Those with approximately equal loadings on each factor were eliminated from further analysis. This procedure resulted in 25 Type i and 24 Type n subjects. Results of a discriminant analysis (Anderson, 1958) confirmed that Types I and n were clearly separable (D2 = 21.9, p < 0.001), especially in terms of their Dominance and Orderliness (per cent Contribution to D* = 65 and 23 respectively). Separate analysis of demographic items showed that these subsamples also differed in sex representation (Males: Type i = 40 per cent, Type n = 83 per cent, x2

= 5.2, p< 0.05). To determine how subjects formed impressions of the stimulus people,

a five-way ANOVA was performed on the Own acceptance data. The nested replications were combined and processed on the University of Western Ontario BBM 7040 system according to BALANOVA using the approximate unweighted means method (Winer, 1962, pp. 224-227). The independent


variables under consideration were represented by die following two-level fixed factors: Affiliation (Friendly-Unfriendly); Orderliness (Orderly- Disorderly); Dominance (Dominant-Submissive); Masculinity (Male interest-Female interest); and Subjects (Type i-Type n) . Main effects of all stimulus cues for the total sample were significant and variance esti- mates (Hays, 1963) indicated the following rank order of salience: Affilia- tion (F = 82.30, df = 1,1536, p < 0.01, to2 = 0.31); Orderliness (F = 72.17, df = 1,1536, p < 0.01, to2 = 0.27); Dominance (F = 29.07, df = 1,1536, p < 0.01, to2 = 0.10; and Masculinity (F = 18.49, df = 1,1536, p < 0.01, ttf2 = 0.05). Most noteworthy, however, was the complete absence of significant interactions among the stimulus cues.

Main effects were significant for the Subjects factor with Type i sub- jects indicating relatively greater acceptance of the stimulus people (F = 24.00, df = 1,1536, p < 0.01). The major source of this difference was a subject type-by-Masculinity cue interaction (F = 12.77, df = 1,1536, p < 0.01), cell means of which suggest that Type i subjects expressed greater acceptance of those represented with a Feminine interest cue than did Type n subjects.

Validity of these results was inferred from a comparison of correlations between sum totals of the connotative values of die items representing each stimulus person and the average “Own” and “Others” acceptance judgments of each group. The following values clearly indicate that the Own acceptance judgments of both groups were significantly less asso- ciated with item connotation than were the Others acceptance judgments: Type i -Own (0.75), Others (0.90) (t = 7.38, p < 0.01); Type n -Own (0.43), Others (0.82) (t = 9.59,p < 0.01). This suggests that the findings in this study may be considered representative of subjects’ idiosyncratic Own acceptance judgments. It also confirms die utility of a multi-set response procedure for social perception studies.


This investigation of personality impression formation involved a corre- lational-experimental design in which subjects were classified in terms of the same personality dimensions as those used to represent the experi- mental stimuli. Results suggested that subjects’ impression responses were determined by individual stimulus cues rather than by combinations of cues. That is, they appeared to operate as though traits in other people function independently. Moreover, subjects seemed to be relatively more influenced by some cues, for example those denoting friendliness, than by other cues. Such findings are consistent with a growing body of evidence which points toward “weighted linear” as opposed to “configural” inter-


pretations of impression formation (Anderson, 1968; Goldberg, 1968). They also confirm previous evidence concerning the centrality of the warm-cold dimension for social perception (Asch, 1946; Kelley, 1950).

In addition to these experimental effects, a subject type-by-Masculinity cue interaction was evident. Cell means indicated that submissive, orderly Type i subjects appeared more willing to accept those represented as enjoying reading love stories than did the dominant, less orderly Type n subjects. This interaction, while understandable in terms of the measured personality differences between the two groups, may have been mediated by significant differences in actual sex representation in the groups since there were 40 per cent males in the Type i classification and 83 per cent in Type n. Certainly it makes just as much sense to think of male and female groups expressing different acceptance toward someone who likes to read love stories as it does to consider the same finding in terms of differences in dominance and/or orderliness between the groups. In any event, this interpretation is consistent with the observation that sex has frequently been found to be a potent moderator variable in person per- ception (Schrauger & Altrocchi, 1964).

The absence of other possible perceiver-perceived interactions invites comment even at the risk of Type n error: For example, why were not the dominant Type n subjects relatively more attracted to the submissive stimulus people? Or, why did not the orderly Type i subjects respond more favourably to the orderly stimulus people than did the less orderly Type n subjects? Both substantive and methodological explanations may be considered. On the one hand, it is possible that models of interpersonal attraction involving need-complementarity or similarity (e.g. Winch, 1958; Newcomb, 1961) simply may not be tenable. However, the scope of the present study precludes such a firm conclusion. An alternative explanation concerns the nature of the information used to represent stimulus people, and the kinds of measures used to differentiate perceiver types. Specific- ally, it is an open question whether such items as, I often decide ahead of time exactly what I will do on a certain day, and, I am quite good at keeping others in line, are sufficiently potent to transmit a satisfactory image of Orderly and Dominant people respectively. Equally questionable is the representativeness of information about subjects yielded by true- false psychometric devices which purport to reflect basic personality traits. This line of reasoning raises the possibility that the obtained interaction might have been different in degree had the actual sex of the stimulus people been represented. Other possibilities also come to mind. For ex- ample, would a “complementary-needs” interaction have occurred had the dominance of the perceivers and the submissiveness of those perceived


been assessed and described in other than psychometric terms, such as demographically or via audio-visual channels?

In closing, it is suggested that Brunswick’s (1947) conception of repre- sentative design be reincarnated for studying impression formation. It is conceivable that with richer stimulus information subsequent investiga- tions may obtain quite different results, just as more general conditions of social judgment have yielded less simple impression strategies (Partington & Jackson, 1968).


Etude corr61ationnelle-experimentale sur la maniere dont on se forme une impression sur la personnalite d’autnd. Les sujets doivent estimer comment ils accepteraient certaines personnes-stunuli hypothetiques representees par des constellations d’enonces auto-descriptifs denotant les poles opposes de quatre dimensions de la personnalite. Us doivent egalement dormer des reponses d’auto-acceptation aux item d’une batterie de personnalite cemant les memes quatre dimensions. Les sujets sont repartis en types de personnalite constituant les differents niveaux d’un facteur lie au sujet: les relations de ce facteur avec les quatre faeteurs lies aux indices stimuli donnent lieu a un plan factoriel complet servant a l’analyse des jugements d’acceptation. Meme si les principaux resultats expenmentaux suggerent que l’integration des indices stimuli soit plus lineaire que configurationnelle, la nature des interactions sujet-indices illustre la necessite d’un modele plus representatif dans les recherches en perception sociale.


ANDERSON, N.H. Application of an additive model to impression formation. Science, 1962, 138, 817-818. ANDERSON, N.H. A simple model for information integration. In R. P. Ableson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, and P. H. Tannen- baum (eds.), Theories of Cognitive Consistency: A Sourcebook. Chicago: Rand McNally, 1968. ANDEBSON, N.H. Comment on “An analysis of variance model for die assessment of configural cue utilization in clinical judgment.” Psychological Bulletin, 1969, 72, 63-65. ANDERSON, T.W. An introduction to muttivariate statistical analysis. New York: Wiley, 1958, pp. 126-138. ARGYRIS, c. Some unintended consequences of rigorous research. Psychological Bulletin, 1968, 70, 185-197. ASCH, s.E. Forming impressions of personality. Journal of Abnormal and Social Psychology, 1946, 41, 258-290. BIER], J., ATKINS, AX., BRIAR, S., LEAMAN, R.L., MILLER, H., & TRIPODI, T. and social judgment. New York: Wiley, 1966. BRUNSWICK, E. Systematic and representative design of psychological experiments. Berkeley, California: University of California Press, 1947. CATTELL, R.B. The scientific analysis of personality. Harmondsworth: Penguin Books, 1965.


GOLDBERG, L.R. Simple models or simple processes? Some research on clinical judg- ments. American Psychologist, 1968, 23, 483-496. HAYS, W.L. Statistics for Psychologists. New York: Holt, Rinehart, and Winston, 1963. HAHMAN, H.H. Modern factor analysis. Chicago: The University of Chicago Press, 1960. BOFFMAN, F.J., SLOVIC, P., * BORER, L.G. An analysis-of-variance model for the assessment of configured cue utilization in clinical judgment Psychological Bulletin, 1968, 69, 338-349. JACKSON, D.N. Personality Research Form Manual. Goshen, New York: Research Psychologists Press, 1967. KELLJEY, H.H. The warm-cold variable in first impressions of persons. Journal of Personality, 1950, 18, 431-439. LAY, CH., & JACKSON, D.N. An analysis of the generality of trait inferential networks. Journal of Personality and Social Psychology, 1969,12,12-21. LOEVINGER, j . Objective tests as instruments of psychological theory. Psychological Reports, 1957, 3, 638-694. LOVEIX, V.R. The human use of personality tests; a dissenting view. American Psychologist, 1967, 22, 383-393. NEWCOMB, T.M. The acquaintance process. New York: Holt, Rinehart, and Winston, Inc. 1961. NUNAIXT, j . Psychometric Theory. New York: McGraw-Hill, 1967. OWENS, W.A. Toward one discipline of scientific psychology. American Psychologist, 1968, 23, 782-785. PARTINGTON, J.T. A test of averaging, adding, and dynamic models in personality impression formation. Unpublished Doctoral Dissertation, University of Western Ontario, 1966. PARTINGTON, J.T. Generality of the simple averaging model in personality impression formation. Psychonomic Science, 1967,8,339-340. PARTINGTON, J.T. Dr. JekyD and Mr. High: Multidimensional scaling of Alcoholics’ self-evaluations. Journal of Abnormal Psychology, 1970, 75,131-138. PARTINGTON, J.T., * JACKSON, D.N. A multidimensional scaling evaluation of alter- native models in personality impression formation. Psychonomic Science, 1968,10, 331-332. SCHRAUGER, s., & ALTROCCHi, j . The personality of the perceiver as a factor in person perception. Psychological Bulletin, 1964,62,289-308. TRIANDIS, H.C., & FisHBEiN, M. Cognitive interaction in person perception. Journal of Abnormal and Social Psychology, 1963, 67,446-453. WINCH, HJ . Mate-selection: A study of complementary needs. New York: Harper and Row, Publishers, Incorporated, 1958. WINER, B. Statistical Principles in Experimental Design. New York: McGraw-Hill, 1962. WISMNKH, j . Reanalysis of Impressions of personality.” Psychological Review, I960, 67, 96-112.

First recetoed 22 January 1970

You may also like...