Page 1 PERSONALITY PROCESSES AND INDIVIDUAL ...

9 downloads 0 Views 2MB Size Report
The author would like 10 thank Jerry Wiggins. Harold. Sackeim, and two ... Byrne's (1961) Repression-Sensitization scale. Other instruments falling within this ...
PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Two-Component Models of Socially Desirable Responding Delroy L. Paulhus University of British Columbia, Vancouver, British Columbia, Canada A recent two-factor model of socially desirable responding based on denial and allribution components was reviewed and disputed. A second model distinguishing self-deception and impression management components was reviewed and shown to be related to early factor-analytic work on desirability scales. Two studies were conducted to test the model. A factor analysis ofcommonly used desirability scales revealed thai the two major factors were best interpreted as self.Deception and Impression Management. A second study employed confirmatory factor analysis to show that the attribution/denial model does not fil the data as well as the selfdeception/impression management model. A third sludy compared scores on desirability scales under anonymous and public conditions. Results showed that those scales that had loaded highest on the Impression Management factor showed the greatcst mean increase from ilnonymous 10 public conditions. It is recommended that impression management, but not self-deception, be controlled in self-reports of personality.

Over the last 25 years a variety of instruments have been designed to assess individual differences in socially desirable responding. Wiggins (1968) cited references to at least 12 such scales. Other scales introduced since that time include the Approval~Motivation scale (Larsen, Martin, Ettinger, & Nelson, 1976), the Self~ and Other-Deception Questionnaires (Sackeim & Gur, 1978), and the Social Desirability Inventory (Jacobson, Kellogg, Cauce, & Slavin, 1977). Evidence has been accumulating that the various measures of socially desirable responding can be incorporated within a two-factor model. This article reviews several such approaches and reports some new evidence concerning the underlying structure of social desirability. Attribution Versus Denial One approach has focused on the distinction between attribution and denial responses to The author would like 10 thank Jerry Wiggins. Harold Sackeim, and two anonymous reviewers for comments on an earlier draft. Requests for reprints should be sent to Delroy L. PaUlhus, Department of Psychology, Universily of British Columbia. Vancouver, British Columbia, Canada V6T IWS.

self-report items (see Millham & Jacobson, 1978). Attribution responses involve claiming socially desirable characteristics for the self; denial responses involve disclaiming that undesirable characteristics apply to the self. To assess these cornJXInents Millharn (1974) partitioned the items ofthe Marlowe-erowne scale into attribution and denial subscales. He found mixed evidence that the two comJXInents were differentially related to cheating behavior. Ja~ cobson et a1. (1977) used a larger item pool to develop a social desirabiJity inventory that explicitly separated attribution and denial subscales. They found that the subscale intercorrelations (average, .60) were substantially lower than the correlations of the subscales with total scores (average, .82) and concluded that tbe subscales tapped different constructs. This conclusion, however, was not justified by their data. The subscaJe intercorrelations must be stepped up to the total test length and then compared with the observed reliability of the total test. Indeed, after being stepped up with the Spearman.Brown formula, the subscale intercorrelations range from .85 to .89, figures very close to the observed reliability of the whole test (Kuder-Richardson 20 "" .90). Therefore, a more appropriate conclusion

J""mal of ''''ItOOalily &1Id Social l's)"ChotolY. 198&. Vol. #, No, J. S98-609 COPl"i&h' 1980t by ,"" AmO you tell the truth?" Another strong item is "When you take sick leave, are you always as sick as you say you are'!" Note that they do not have the personal threat Quality typical of the items loading on Factor I. Finally, the ODQ is the best marker variable, providing further evidence that the second factor represents impression management. ~

In the factor analysis of the scale totals the first three factors explained 43%, 24%, and 10% of the total variance. In the analysis of the odd items, the first three factors explained 15%, 13%, and 7% of the total variance. For the even items, the comparable figures were Study 2 16%, 13%, and 5%. Because the same "elbow" Although the factor pattern of the SDQ and appeared in all three analyses, only the first two factors were retained for rotation. The 000 on the Alpha and Gamma factors supfactor loadings of the six scale totals from the ports the self-deception/impression manage· ment model, the attribution/denial model is three analyses are presented in Table 2. The configuration of the six total scales was not entirely ruled out. An ambiguity arises very similar whether or not the individual from the method of keying the SDQ and ODQ items were included in the factor analysis. The scales. Specifically, the SDQ contains only correlations of their factor loadings across the negatively keyed items and the ODQ contains three analyses were above .90 for both Factor only positively keyed items. Thus the SDQ 1 and Factor 2. It is clear from Figure I that items are primarily denials of negative attri· the pattern of loadings is quite consistent with butes: The scale confounds self-deception with previous factor-analytic studies: Edwards's SO denial. Similarly, the ODQ scale confounds scale loads strongly on the first factor; Wiggins' impression management with attribution. Sd scale has its highest loading on the second Note that in Study I, an examination of the factor; and the Marlowe-Crowne scale loads items loading on Alpha and Gamma indicated highly on both factors. Results for the SIX2 that the factors were not fuUy confounded 'Nith and ODQ were even more cleareut. The SDQ positive and negative keying: Items of both is the highest loading scale on the first factor, the ODQ is the purest marker of the second • The complele list of items and their factor 1oadi~ is factor. available from the author. , Neither factor seems to be related to the oonstruCi of The highest loading items on each factor4 are listed in Table 3. An examination of the .self-monitoring (Snyder, 1974), which appears to tap skill highest loading items on Factor I aids in its I3.tberthan motivation in impression ~t (Collins, Paulhus, &: Grariano, 1983; Danheiser &: Paulhus, 19111). interpretation. Five of the top 10 items are The correlations of .self·monitoring witb the SDQ and drawn from the SDQ, and the other five are ODQ v.oer't -.10 and .23, respectively.

602

DELROY L PAULHUS SELF-DECEPTION

.'" co.

----;,.,.,--7''''---,.,.,---!----,.,.,---.,..---..-•• 7~

-.50

-.25

.25.50.75

IMPRESSION lUNAGEM[NT

,.

·.25

-. '" 000* Ott,... -Oec;ept Ion OUest lonna

t.-.

·,75 50·Edwards social de.I.-abl' ltv scal.

Figure 1. fu of foctor loadings of six social desirability scales.

valences showed up on each factor. Nonetheless, a clear separation of keying direction and content in the SDQ and ODQ would facilitate

dorsc 10 socially desirable attributes and deny 10 socially undesirable attributes. The overall

a comparison of the two models of socially desirable responding. In the present study, the items from the SDQ and 000 were rewritten so that (a) all items were worded as statements rather than Questions, (b) all statements were worded as trait affirmations (I am nice); negations (I am not nice) were eliminated; and (e) equal num· bers of attribution and denial items appear on each scale.' For instance. the ODQ item, "I am honest," was changed to '" sometimes tell lies if I have to," The SDQ item, "Is it important to you that others think highly of you," was changed to ..It's alright with me if others happen to dislike me." To get a perfect score on either scale, the respondent must now en·

• There is an ambiguity in operationaliring lhe ami· bution/denial model. Millham (1914, p. 382) partitioned the 33 Maf1owe.Crownc items into positMlyand ncgatiYdy keyed statemellts. Theaunbutton toore "''3Sdefinc:d as the number of agrccmenlS with positively keyed items; the denial score was the number of disagreemenlS wilh negatively ke~ items.. Thus the items '" am a saine and ") am not a crool: would both be considered attribution item$. The items "I am I crook and "I am 001 a saint" would be denial items because disagreemenl, is the socially desirable response, ~ Millham (1974, p, 380) and Millham and KdklI& (1980, P. )71) ddined the altributiOll compQQent u a .. tendency to attribute KJciaIIy dc:sirabIe Slltement5 to one's selr and denial as a -tendency to deny undesirable charactcristics." This definition of al_ tribution implies that the: attributioo items ~ "1 am a saint" and"l am nota saine because agreement with the M

M

SOCIALLY DESIRABLE RESPONDING

603

Table 3 Highest Loading Items on FaclOrs J and 1 From Separate Factor Analyses a/Odd and Even Items Loadings

II

Ilem Faclor I; Self-Deoeption I. H.Ye you e-u aljoycd your bowel trIO'"'C'menlS? (SDQI 2. HlIYC' you e-u been uncenain as 10 .....bether or not you are OOmose.o;uaf? (SDQ) 3. ~ often disappoinl me. (SO) 4. Life is • Slrain for me Il'105l of the time. (SO) 5. HlIYC' you e-u doubted your sexual adequacy? (SDQ) 6. I worry quite a bil OYC'I" possibJe misl"OI1UIlC'$. (SO) ,. HlIYC' you e-u thou&hl thai your parenlS haled you? (5DQ) 8. I have several times Pveo up doinc something because I thoupl too linle of my ability. (SO) 9. In I IJ'llUp of people I have trouble lhinkins of the rilhl thinss 10 talk about. (SO) 10. HlIYC' you eYer thoughl of eommiltilll suicide in orde!" to It' black al SOInC'OIlC'? (500)

Factor

. 61

63 63 62 62 61 61

"

36 20 -IS

-18

.," 12

16

28

60

I.

25 14

63 60

2; Impressioo Manatemenl

I. Do you tell the truth? (ODQ) 2. When you take sick-leave ff'OOl work or school•• re you as sick as you say you are? (ODQ) 3. I 1m always couneous. even 10 peopk .....ho are disagreeabk. (Me) 4. Ooce in • while I laugh .1 • dirty joke. (Sd) 5. I iOO1C'Iimes try 10 Jel even, rathc:f than forgive and fOJ'lC'l. (Me) 6. I always apolo&ize 10 others for my mistakes.. (OOQ) 7. Would you deelare everythifl3 at customs, even if you knew thaI you cook! never be found

out? (ODQ) 8. I never .nC'nd a sexy show if I can avotd iL (Sd) 9. Sometimes at ekctions J VOle for candidates I know linle about. (Sd) 10. J am 5OIJlC'times irriuted by peopk who asIc f2V0f'5 of me. (Mq

-18

" "

-