5
Astrological Data and Personality 713 astrological analyses were thorough and far more sophisticated than those found in the popular literature. However, the approach taken here does not necessarily imply that astrological data may not be useful as part of a comprehensive personal- ity assessment conducted on an individual basis by an experienced clinician com- petent in astrology. Perhaps the present study will serve as a challenge to such individuals to put their beliefs and hypotheses to an empirical test that is more acceptable to them. REFERENCES DOBYNS, Z. Personality assessment through astrology. The Apuarian Agent, 1970, 1, 2. HATHAWAY, S. R., & MCKINLEY, J. C. Minnesota Multiphasic Personality Inventorg. Manual for LAFORGE, R., & SUCZEK, R. The interpersonal dimension of personality (111). An interpersonal LEARY, T. Multilevel measurement of interpersonal behavior: A manual for the use of the Interpersonal LEARY, T. Interpersonal diagnosis of personality. New York: Ronald Press, 1957. administration and scoring (rev. ed.). New York: Psychological Corporation, 1951. check list. Journal of Personality, 1955, 2'4, 94-112. System of Personality. Berkeley, Calif. : Psychological Consultation Service, 1956. SELF-RATINGS AND THE EYSENCK PERSONALITY INVENTORY (EPI) M. J. STONES Memorial University of Newfoundland St. John's, Netofoundland An attempt was made to assess the relationships between self-ratings and psychometric (EPI) estimates on the dimensions of extraversion and neu- roticism. Ninety-five individuals served as 8s. The correlations between self-ratings and EPI estimates achieved high significance on both dimensions. Contrary to earlier suggestions (Bem & Allen, 1974), ratings of behavioral variability on the respective dimensions failed to produce a significant mod- erating effect. However, a tentative trend was observed for cross-procedural comparability (between self-ratings and the EPI) to be higher near the mid- points of the respective distributions. A factor contributory to this was a central regressive tendency (noted at both poles on both dimensions and for both sources of estimation), whereby an extreme estimate from one source was associated with a less extreme estimate from the other. Acker and McReynolds (1966) and Harrison and McLaughlin (1969) suggest that self-ratings can be regarded as a valuable procedure for personality assess- ment. In fact, self-ratings frequently have been employed as a validity criterion for psychometric estimates of position on a trait dimension. An example of such is the Eysenck Personality Inventory (EPI), on which self-ratings have been found to correlate significantly with both the extraversion (Bem & Allen, 1974; Harrison & McLaughlin, 1969; Vingoe, 1966) and neuroticism scales (Harrison & McLaughlin, 1969). The variance in common between self-ratings and psychometric estimates typically has been found to lie in the region of 25% (Nickels & Renzaglia, 1958). It therefore is pertinent to inquire whether that unaccounted-for variance reflects random effects or identifiable, systematic departures from linearity. A concrete suggestion, with regard to the latter, is provided by B e p and Allen (1974). These authors found that a student population could be differentiated into two groups, on the basis of self-ratings of behavioral variability (variability ratings), relevant to the extraversion dimension. For the low-variability group, the correlations between EPI extraversion and various criterion measures were higher than for the high-variability group.

Self-ratings and the Eysenck personality inventory (EPI)

Embed Size (px)

Citation preview

Page 1: Self-ratings and the Eysenck personality inventory (EPI)

Astrological Data and Personality 713

astrological analyses were thorough and far more sophisticated than those found in the popular literature. However, the approach taken here does not necessarily imply that astrological data may not be useful as part of a comprehensive personal- ity assessment conducted on an individual basis by an experienced clinician com- petent in astrology. Perhaps the present study will serve as a challenge to such individuals to put their beliefs and hypotheses to an empirical test that is more acceptable to them.

REFERENCES DOBYNS, Z. Personality assessment through astrology. The Apuarian Agent, 1970, 1, 2. HATHAWAY, S. R., & MCKINLEY, J. C. Minnesota Multiphasic Personality Inventorg. Manual for

LAFORGE, R., & SUCZEK, R. The interpersonal dimension of personality (111). An interpersonal

LEARY, T. Multilevel measurement of interpersonal behavior: A manual for the use of the Interpersonal

LEARY, T. Interpersonal diagnosis of personality. New York: Ronald Press, 1957.

administration and scoring (rev. ed.). New York: Psychological Corporation, 1951.

check list. Journal of Personality, 1955, 2'4, 94-112.

System of Personality. Berkeley, Calif. : Psychological Consultation Service, 1956.

SELF-RATINGS AND THE EYSENCK PERSONALITY INVENTORY (EPI)

M. J. STONES

Memorial University of Newfoundland St. John's, Netofoundland

An attempt was made to assess the relationships between self-ratings and psychometric (EPI) estimates on the dimensions of extraversion and neu- roticism. Ninety-five individuals served as 8s. The correlations between self-ratings and EPI estimates achieved high significance on both dimensions. Contrary to earlier suggestions (Bem & Allen, 1974), ratings of behavioral variability on the respective dimensions failed to produce a significant mod- erating effect. However, a tentative trend was observed for cross-procedural comparability (between self-ratings and the EPI) to be higher near the mid- points of the respective distributions. A factor contributory to this was a central regressive tendency (noted at both poles on both dimensions and for both sources of estimation), whereby an extreme estimate from one source was associated with a less extreme estimate from the other.

Acker and McReynolds (1966) and Harrison and McLaughlin (1969) suggest that self-ratings can be regarded as a valuable procedure for personality assess- ment. In fact, self-ratings frequently have been employed as a validity criterion for psychometric estimates of position on a trait dimension. An example of such is the Eysenck Personality Inventory (EPI), on which self-ratings have been found to correlate significantly with both the extraversion (Bem & Allen, 1974; Harrison & McLaughlin, 1969; Vingoe, 1966) and neuroticism scales (Harrison & McLaughlin, 1969).

The variance in common between self-ratings and psychometric estimates typically has been found to lie in the region of 25% (Nickels & Renzaglia, 1958). It therefore is pertinent to inquire whether that unaccounted-for variance reflects random effects or identifiable, systematic departures from linearity. A concrete suggestion, with regard to the latter, is provided by B e p and Allen (1974). These authors found that a student population could be differentiated into two groups, on the basis of self-ratings of behavioral variability (variability ratings), relevant to the extraversion dimension. For the low-variability group, the correlations between EPI extraversion and various criterion measures were higher than for the high-variability group.

Page 2: Self-ratings and the Eysenck personality inventory (EPI)

714 Journal of Clinical Psychology, July , 1977, Vol. 33, No . 3.

A further suggestion concerns the degree of polarization (extremity) of the estimates obtained. It is possible that extreme estimates derived from one source (e.g., the EPI) might have systematic but nonlinear implications for the accuracy of estimation that pertains to the other source, (e.g., self-ratings). ,4 modification of this hypothesis, which emphasizes extremity in conjunction with direction, is one of social desirability response bias. Persons who perceive themselves as possessing socially desirable traits (both stability and extraversion are desirable traits in the North American culture) might display less defensiveness and hence greater judgmental accuracy and comparability across the two procedures. Ques- tions that relate to the cross-procedural generality of estimate extremity are im- portant both for clinical inference and the study of individual differences.

This article attempted to replicate earlier findings of positive relationships between self-ratings and the EPI scales and to identify, if possible, nonlinear components that might account for portions of the residual variance. With regard to the latter, attention was focused specifically upon estimate extremity and the moderating effect of variability ratings.

METHOD Subjects

The division between males and females was approximately even.

Procedure Ss first were presented with a description of introversion-extraversion and

neuroticism-stability. The descriptions were taken verbatim from the EPI manual (Eysenck & Eysenck, 1968). They then were required to rate themselves in terms of extraversion and neuroticism, on 24-point scales, after they had been told that the “average” North American student scores 14 on EPI extraversion and 11 on EPI neuroticism (these values were derived from the manual). Next, Ss were instructed to rate the consistency-variability of their behavior, relevant to the two dimensions. The specific question was of the form employed by Bem and Allen (1974), i.e., “HOW much do you vary from one situation to another in how. . . you are?” Again, 24-point scales were used and the information that the “typical” North American student scores 12 was provided. Finally, Forms A and B of the EPI were completed.

Ninety-five student volunteers were obtained from three separate classes.

RESULTS The initial analyses involved an attempt to replicate earlier findings of signifi-

cant relationships between self-ratings and the corresponding EPI scales. The correlation between self-rated extraversion and EPI extraversion was r (93) = .57 ( p < .OOl) and between self-rated neuroticism and EPI neuroticism was r (93) = .47 ( p < .001). These results are in accord with those of Harrison and McLaughlin (1969).

Correlational analyses also were performed in an attempt to validate Bem and Allen’s (1974) suggestion that variability ratings might moderate the extent here of relationship between self-ratings and psychometric estimates. On each dimension, Ss were divided into high-variability and low-variability groups, according to whether the variability rating lay above or below the median. The groups were found not to differ significantly in terms of mean self-ratings or mean EPI estimates. For the extraversion dimension, the self-rating/EPI correlations were r (52) = .48 ( p < .OOl) for the high-variability group and r (39) = .62 ( p < .OOl) for the low-variability group. These values do not differ significantly ( z = .94). For the neuroticism dimension, the corresponding correlations were r (42) = .33 ( p < .05) for the high-variability group and r (49) = .54 ( p < .OOl) for the low-variability group. Again, the values do not differ significantly (z = 1.30). The results ob- tained are in the direction suggested by Bem and Allen (1974), but are insignificant.

Page 3: Self-ratings and the Eysenck personality inventory (EPI)

Self-ratzngs and the Eysenck Personality Inventory 715

Further analyses attempted to assess intergroup differences rather than intragroup correlations. The intention here was (a) to provide additional appraisal of any moderating effects attributable to variability ratings and (b) to evaluate the effects of extremity and direction of estimates from one source on the com- parability of estimates from the other source. The comparability of self-rating and EPI estimates was operationalized by deriving discrepancy scores over the two dimensions. The extraversion discrepancy scores were obtained by first trans- forming the self-rating and EPI estimates into standard score format and then subtracting the transformed self-rating score from the transformed EPI score, i.e., EPI minus self-rating. I n the form described, the extraversion discrepancy score takes into account both extent and direction of cross-procedural compar- ability. In order to assess extent of comparability in an absolute sense, a further set of extraversion discrepency scores was obtained where sign (plus or minus) was ignored. Similarly, two sets of discrepancy scores were obtained from the neuroticism dimension.

TABLE 1

LOWER 25%, MIDDLE 50y0 AND UPPER 25% ON THE BASIS OF SOURCE INDICATED IN LEFT-MOST COLUMN

MEAN EXTRAVERSION DISCREPANCY SCORES (WITH AND WITHOUT SIGN) FOR f l S GROUPED INTO

Discrepancy Groups score Sign

Source of taken Lower Middle Upper Overall Individual groups account of? 25% (L) 50% (M) 25% (U) significance comparmons

Extraversion Yes - .17 .06 .05 F(2,92) = -

No .77 .62 .78 F(2,92) = - Variability- .55, ns rating

.78, ns

EPI, Yes - .60 .08 .46 F(2,92) = All significant Extraversion 9.32 p < .05, or

p < .001 beyond No .84 .60 .79 F(2,92) = -

1.69, ns -~ ~ ~~~~ ~ ~

Self-rating, Yes .58 - 01 - .50 F(2,92) = All significant Extraversion 8.81 a t p < .05, or

p < .001 beyond

3.48 significant a t No .94 .62 .72 F(2,92) = L vs. M,

p < .05 p < .05

Ss were separated into groups on the basis of their variability ratings, self- ratings and EPI scores, respectively. I n all cases, the division was approximately between the lower 25%, the middle 50% and the upper 250j0. The discrepancy scores associated with the respective groupings fqr the. extraversion dimension are presented in Table 1 and for the neurotlclsm dimenslon in Table 2. I n both tables overall significance levels are provided, as well as the results of individual comparisons (by multiple t ) where appropnate.

Separation into groups on the basis of variability ratings failed to provide significant overall differences for any of the four comparisons. This result is in accord with the aforementioned correlational analysis, where no significant moderat- ing effects of variability ratings were obtained.

Separation on the bases of self-ratings and the EPI produced results ap- proximately similar across both dimensions. Where sign of the discrepancy score

Page 4: Self-ratings and the Eysenck personality inventory (EPI)

716 Journal of Clinical Psychology, Ju ly , 197’7, Vol. 33, N o . 3.

TABLE 2

MEAN NEUROTICISM DISCREPANCY SCORES (WITH AND WITHOUT SIGN) FOR 8s GROUPED INTO LOWER 25%, MIDDLE 50% AND UPPER 25% ON THE BASIS OF SOURCE INDICATED IN LEFT-MOST COLUMN

Discrepancy Groups score Sign

Source of taken Lower Middle Upper Overall Individual groups account of? 25Oj, (L) (M) 25% (U) significance comparlsons

Neuroticism Yes - .03 - .09 .23 F(2, 92) = Variability- .71 ratings

.48 No .91 .77 .78 F (2,92) =

EPI Yes - .66 - .04 .81 F(2,92) = Neuro ticism 17.03

p < .001

No .87 .62 1.11 F(2, 92) = 5.69 p < .01

All significant a t p < .01 or beyond

M vs. H, sig- nificant a t p < .01

Self-rating, Yes .54 .04 - .58 F(2, 92) = All significant Neuroticisin 9.81 at p < .05 or

p < .001 beyond

No .83 .82 .79 F(2, 92) = - .04, ns -

was disregarded, significant overall effects were obtained on two occasions (extra- version discrepancy scores when grouped by self-ratings and neuroticism discre- pancy scores when grouped by the EPI) and on three of the four occasions, the lowest mean score was associated with the middle 50% group: If any general trend is discernible from these data, i t is in the direction of higher cross-procedural comparability where estimates lie near the midpoints of the respective distributions.

With sign taken into account, overall significance was achieved on all four occasions and the individual comparisons were uniformly significant. A phenomenon of “central regression” was noted, whereby extreme estimates from one source were associated with less extreme estimates from the other. This phenomenon is apparent from a perusal of the directionality of the group means obtained and is present a t both extremes and over both dimensions.

DISCUSSION The first aim of the experiment was successful in that correlations between

self-ratings and EPI scales achieved satisfactory levels of significance. For extra- version, variance in common was 32% and for neuroticism, it was 22%. The second aim, to identify nonlinear components that might contribute to the residual variance, met with only partial success.

With regard to the suggestion (Bem & Allen, 1974) that variability ratings might exert a moderating effect, the relevant statistical analyses failed to provide confirmatory evidence even though the correlational data were in the appropriate direction. This might be considered as surprising because Bern and Allen (1974) found their low variability-rating group to obtain higher correlations with EPI extraversion on each of six criterion measures (including that of self-ratings). However, upon reanalysis by the author, Bem and Allen’s (1974) conclusion ap- pears less substantial. For none of their six comparisons, taken independently, did the correlation obtained by the low variability-rating group signijcantly exceed that of the high variability-rating group. This finding is compatible with that of

Page 5: Self-ratings and the Eysenck personality inventory (EPI)

Self-ratings and the Eysenck Personality Inventory 717

the present study, where slight moderating effects were present, but failed to achieve significance in individual comparisons.

Greater success was obtained from the consideration of estimate extremities. In no case was the absolute discrepancy between self-rating and EPI estimates greater at one pole of the relevant dimensions than at the other. Thus no evidence was provided for the hypothesis that a social desirability factor might affect degree of cross-procedural comparability. Similarly, no support was provided for Vingoe’s (1966) assertion that introverts display greater cross-procedural comparability than do extraverts. However, some evidence was obtained for a lower absolute discrepancy, where estimates lay near the midpoints of the respective distributions. Although not all comparisons achieved significance, the tentative implication is that greater confidence can be invested in the cross-procedural comparability of estimates, where one lies a t the middle range of the distribution rather than the extremity.

A probable reason for the low cross-procedural comparability, a t the extremes of the distributions, is the presence of a strong central regressive trend. Where an estimate from one source was highly polarized, that from the alternate source was less polarized. In general, the extent of regression was half a standard devia- tion or greater. The trend was noted over both dimensions, at both poles, and for S grouping based on both self-ratings and the EPI.

In conclusion, the present study was successful in demonstrating significant variance in common between self-ratings and formal psychometric estimates of position on a personality trait dimension. Departures from linearity tended to be greater a t the extremes of the distributions, where cross-procedural compar- ability was lower and a central regressive tendency was apparent. This has some- what unfortunate implications for clinical inference and research into individual differences because interest typically is focused more on the highly polarized estimates. Finally, little support was provided for the suggestion (Bem & Allen, 1974) that variance accounted for would be increased significantly for the subgroup associated with low variability ratings.

REFERENCES ACKER, M., & MCREYNOLDS, P. On the assessment of anxiety: 111. By self-ratings. Psychological

BEM, D. J., & ALLEN, A. On predicting some of the people some of the time: The search for cross- Psychological Review, 1974, 81, 506-520.

EYSENCK, H. J., & EYSENCK, S. B. G. Manual for the Eysenck Personality Inventory. San Diego,

HARRISON, N. W., & MCLAUGHLIN, R. J. Self-rating validation of the Eysenck Personality Inven-

NICKELS, J. B., & RENZAGLIA, S. A. Some additional data on the relationship between expressed

VINGOE, F. J. Validity of the Eysenck extraversion scale as determined by self-ratings in normals.

Review, 1966, 19, 251-254.

situational consistencies in behavior.

Calif. : Educational & Industrial Testing Service, 1968.

tory. British Journal of Social and Clinzcal Psychology, 1969, 8, 55-58.

and measured values. Journal of Applaed Psychology, 1958, 42, 99-104.

British Journal of Social and Clinical Psychology, 1966, 6, 89-91.