10
The convergent validity of acquiescence: an empirical study relating balanced scales and separate acquiescence scales Pere J. Ferrando * , Lorena Condon, Eliseo Chico Facultad de Psicolog ıa, Universidad ‘Rovira i Virgili’, Carretera Valls s/n, 43007 Tarragona, Spain Received 15 May 2003; received in revised form 10 December 2003; accepted 21 January 2004 Available online 28 February 2004 Abstract The convergent validity of the hypothetical construct of acquiescence was investigated in a sample of 207 psychology undergraduates by using three different methods: (1) separate acquiescence scales, (2) balanced scales, and (3) the procedure proposed by Winkler, Kanouse and Ware (1982). The study used recently developed factor-analytic procedures to obtain acquiescence scores, and was designed to overcome some of the drawbacks of previous empirical research. The results suggested that there was some degree of gen- eralizability between the different measures. However, the correlational structure was very weak. Fur- thermore, the analyses suggested that ÔmethodÕ effects accounted for some of the relations observed between the separate acquiescence scales. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: Acquiescence measurement; Response biases; Balanced scales; Structural validity; Factor analysis 1. Introduction Since the pioneering study by Cogan, Conklin, and Hollingworth (1915) the problem of acquiescent responding has been, and continues to be, a controversial topic for personality the- orists and researchers. One of the controversial points is the conceptualization of acquiescence as a variable. On the one hand, some authors consider acquiescence as a statistical nuisance that must be controlled or suppressed with appropriate procedures (e.g. Hofstee, Ten Berge, & Hendriks, 1998; Nunnally, 1978; Ray, 1979, 1983). On the other hand, other authors believe that acquiescence is a meaningful personality construct that is potentially measurable (e.g. Couch & * Corresponding author. Tel.: +34-9-7755-8075; fax: +34-9-7755-8088. E-mail address: [email protected] (P.J. Ferrando). 0191-8869/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.paid.2004.01.003 Personality and Individual Differences 37 (2004) 1331–1340 www.elsevier.com/locate/paid

The convergent validity of acquiescence: an empirical study relating balanced scales and separate acquiescence scales

Embed Size (px)

Citation preview

Personality and Individual Differences 37 (2004) 1331–1340www.elsevier.com/locate/paid

The convergent validity of acquiescence: an empiricalstudy relating balanced scales and separate acquiescence scales

Pere J. Ferrando *, Lorena Condon, Eliseo Chico

Facultad de Psicolog�ıa, Universidad ‘Rovira i Virgili’, Carretera Valls s/n, 43007 Tarragona, Spain

Received 15 May 2003; received in revised form 10 December 2003; accepted 21 January 2004

Available online 28 February 2004

Abstract

The convergent validity of the hypothetical construct of acquiescence was investigated in a sample of 207

psychology undergraduates by using three different methods: (1) separate acquiescence scales, (2) balanced

scales, and (3) the procedure proposed by Winkler, Kanouse and Ware (1982). The study used recently

developed factor-analytic procedures to obtain acquiescence scores, and was designed to overcome some of

the drawbacks of previous empirical research. The results suggested that there was some degree of gen-

eralizability between the different measures. However, the correlational structure was very weak. Fur-thermore, the analyses suggested that �method� effects accounted for some of the relations observed between

the separate acquiescence scales.

� 2004 Elsevier Ltd. All rights reserved.

Keywords: Acquiescence measurement; Response biases; Balanced scales; Structural validity; Factor analysis

1. Introduction

Since the pioneering study by Cogan, Conklin, and Hollingworth (1915) the problem ofacquiescent responding has been, and continues to be, a controversial topic for personality the-orists and researchers. One of the controversial points is the conceptualization of acquiescence asa variable. On the one hand, some authors consider acquiescence as a statistical nuisance thatmust be controlled or suppressed with appropriate procedures (e.g. Hofstee, Ten Berge, &Hendriks, 1998; Nunnally, 1978; Ray, 1979, 1983). On the other hand, other authors believe thatacquiescence is a meaningful personality construct that is potentially measurable (e.g. Couch &

* Corresponding author. Tel.: +34-9-7755-8075; fax: +34-9-7755-8088.

E-mail address: [email protected] (P.J. Ferrando).

0191-8869/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.paid.2004.01.003

1332 P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340

Keniston, 1960; Jackson & Messick, 1965; Morf & Jackson, 1972). Mellenbergh (2001) made aclear methodological distinction between these contrasting conceptualizations. For those in thefirst group, acquiescence arises in a person · item interaction situation in which the person re-sponds to item properties other than the location of the item in the construct to be measured.Acquiescence, then, is a specific variable that depends largely on the characteristics of the mea-surement instrument which is used. Understood in this way, acquiescence can be general (when itaffects all the respondents to the same extent) or differential. General acquiescence would act as anadditive constant and would therefore not affect the inter-item correlational structure. Differentialacquiescence, however, would alter this structure and generate more important problems whenevaluating the properties of the measurement instrument (Hofstee et al., 1998).

For those authors who belong to the second group, acquiescence is considered to be a stablepersonality trait not intended to be measured by the items of the measurement instrument. Itshould be pointed out, however, that this conceptualization does not prevent acquiescence frombeing considered as a nuisance variable since this hypothetical trait, which is not measured, canhinder the measurement of the main trait.

If acquiescence is a meaningful personality trait rather than a scale-specific variable, thenacquiescence measures obtained by different instruments and procedures should be expected to besomewhat consistent. However, the empirical evidence on this point is not very clear. Somestudies have found no significant relations between scores obtained from different acquiescencemeasures, whereas others have found moderate positive correlations. These studies have severallimitations, which are discussed below in greater detail.

1.1. Procedures for assessing acquiescence

The acquiescence measures reported in the literature have been mainly derived from two typesof instruments: separate acquiescence scales and balanced scales. The separate acquiescence scaleswere developed from the assumption that acquiescence is a generalizable trait, whereas balancedscales were developed from the assumption that acquiescence largely depends on the particularinstrument that is used (e.g. Martin, 1964; Messick, 1967).

Separate acquiescence scales are derived from item pools with heterogeneous content, and theyare scored in the direction of agreement regardless of the item content. The items that make upthese scales are chosen to be of medium difficulty (i.e. proportions of endorsement between 0.4and 0.6), and of neutral value in terms of social desirability. The rationale behind these instru-ments is that the tendency to agree across such a heterogeneous and neutral pool must essentiallyreflect a general trait of acquiescence. Scales of this type have been proposed by Bass (1956),Fricke (1957), Couch and Keniston (1960), and Hanley (1961) among others. Couch and Ken-iston�s scale is perhaps the best known of these instruments (Paulhus, 1991). We reviewed theempirical evidence and found that these scales tend to have low reliabilities: the average was about0.30. Clearly, the low reliabilities made these scales unsuitable for clinical measurement (Paulhus,1991). However, they can still be of interest in basic research, as is the case in the present study.

Balanced scales are usually made up of Likert-type items. In a balanced scale all the item stemsare positively worded; however, half of the items measure in one direction of the trait whereas theother half measure in the opposite direction. The main assumption of this type of measures is thatacquiescence to the items in one direction will be canceled out by acquiescence to the items in the

P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340 1333

opposite direction. So, the sum of the appropriately reversed item scores (content score) is ex-pected to be reasonably free of acquiescence (Hofstee et al., 1998; Nunnally, 1978; Ray, 1983). Onthe other hand, the sum of the item scores without reversing is considered to be a measure ofacquiescence (acquiescence score; Ray, 1979, 1983). A review of the literature suggests thatdeveloping a good balanced scale is not an easy task. It also shows that, as in the case of separatescales, the reliabilities of the acquiescence scores are generally low. In the studies reviewed by Ray(1979, 1983) the average reliability was about 0.35.

Winkler, Kanouse, and Ware (1982) proposed a procedure for assessing acquiescence that canbe considered to be a variant of the balanced scale. They developed a measure of attitudes towarddoctors made up of 12 pairs of logically opposite statements and proposed a scoring system thatinvolved counting up double agreements to opposed paired items. Because in such a short scale itis quite possible that respondents are likely to notice that they are repeatedly asked the samequestions in opposite ways, the scale was embedded within a 42-item questionnaire about generalattitudes toward health care. The review of the literature suggests that this scale has some use inapplied research (e.g. Gudjonsson & Clare, 1995).

Recently Ten Berge and coworkers (Hofstee et al., 1998; Ten Berge, 1999) proposed a PrincipalComponent-based approach for analyzing balanced scales, and the present writers adapted theprocedure for the common factor analysis model. For a balanced scale intended to measure a singletrait, the factor analytic procedure can be summarized as follows: (1) to obtain an arbitrary two-factor orthogonal solution, (2) to rotate the solution so that one of the factors be as similar aspossible to the first centroid of the correlation matrix and all the signs positive, and (3) to obtainfactor scores on these two factors. In a well designed scale, the solution obtained in step 2 willconsist of: (1) a bipolar factor (content factor), in which the items in the positive and negativedirections have loadings with opposite signs, and (2) a factor with positive loadings (acquiescencefactor). Furthermore, the variance explained by the content factor will be far larger than the vari-ance explained by the acquiescence factor. It should be noted that the conventional scoring methodfor balanced scales can be considered to be an approximation to the factor procedure. Thisapproximation is obtained by using unit weights in all cases and only taking into account the signsof the factor loadings (i.e. positive and negative unit weights for obtaining the content scores, andpositive-only unit weights for obtaining the acquiescence scores). In contrast, the factor scores areweighted sums, and are expected to be more accurate than the conventional scores, especially inshort scales and moderate-to-large samples.

1.2. A review of previous related research

Until 1965 there were a considerable number of studies in which the scores obtained fromdifferent acquiescence measures were related. Most of this empirical evidence was reviewed inMcGee (1962) and Rorer (1965), and was based on separate acquiescence scales. The correlationsbetween the scores produced by these measures were generally low. In some cases they were non-significant, and usually they were smaller than 0.30.

After the influential paper by Rorer (1965) this type of study practically disappeared. Interestrevived at the end of the 1970s and some new studies about convergent validity were carried out(Ray, 1979, 1983). However, in this second phase the studies were based on acquiescence measuresderived from balanced scales. We reviewed these studies, and found that the correlations between

1334 P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340

acquiescence scores derived from scales that measured different contents ranged from 0.13 to 0.40with an average of about 0.25.

An important shortcoming found in the empirical evidence reviewed is the �mono-method�character of most of the studies. While there are different approaches for measuring acquiescence,very few studies relate acquiescence scores obtained from different methods (mainly separatescales and balanced scales). Actually, as far as we know, the only study of this type is Martin�s(1964). This shortcoming has already been noted by Nunnally (1978) and Ray (1983). In par-ticular, Nunnally (1978) argued that the moderate correlations that are sometimes observed be-tween separate acquiescence scales may simply be due to a method effect.

A second important weakness of previous empirical research is that the studies mentionedabove did not previously assess the reliability of the acquiescence measures. Ray (1979) criticizedthis, and pointed out that in any standard study, evidence that a set of scores is reliable is arequisite if these scores are to be used. Now, a common point in all the procedures reviewed aboveis the (relatively) low reliability of the scores they produce. As is well known in psychometrics, thelack of reliability attenuates the estimates of validity between the measures of interest, and pro-duces a downward bias in the correlations. So, the low validities in previous studies are notsurprising given the low reliabilities of the measures that were intercorrelated. In spite of this, theauthors are not aware of any study in which the correlations between measures of acquiescencehave been corrected for attenuation.

1.3. Purposes and contributions of the present study

The main purpose of the present study is to assess the convergent validity of a series of acqui-escence scores derived from different measurement procedures. Furthermore, the study attempts toovercome some of the drawbacks noted in previous research. First, it relates measurements derivedfrom three procedures: (1) separate acquiescence scales, (2) fully balanced scales and (3) theWinkleret al. (1982) procedure. It should be pointed out that Martin�s (1964) study of a similar issue usedscales that were not completely balanced and his results have been criticized for this reason (Ray,1983). Second, the acquiescence scores derived from the balanced scales are obtained by using thefactor analytic procedure described above. Third, the reliabilities of the different scores are esti-mated and used to estimate the disattenuated correlations between the different measurements.

As Ray (1979) noted, the reliabilities of the different acquiescence scores can be considered as ameasure of the extent to which acquiescence generalizes across the items within a particularinstrument. Generalizability within a single measure is clearly a prerequisite for generalizabilitybetween different measures, which is the main topic of the present research. The correction forattenuation is used not as a correction, as the name suggests, but simply as an estimate of howmuch the correlation could be if the measures were made perfectly reliable.

2. Method

2.1. Instruments

We used the following separate acquiescence scales:

P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340 1335

(1) Couch and Keniston�s (1960) 15-item agreement response scale (ARS).(2) Hanley�s (1961) Bn scale. Hanley developed the Bn scale by selecting those items from Fricke�s

(1957) 63-item B acquiescence scale that were neutral in social desirability. As the selecteditems were not available in the literature, we had Fricke�s 63 items rated for social desirabilityby two judges (�socially desirable�, �socially undesirable� and �neutral�) and used the 32 itemsthat were judged as �neutral� by them both. The 32 items of the Bn scale come from theMMPI, and we used these items as they appear in the published MMPI Spanish adaptation(Seisdedos & Cordero, 1975).

As noted above, good, fully balanced scales are very scarce in the literature. We chose threemeasures that had shown good properties in previous research. They were:

(3) The perceived stress scale (PSS; Cohen, Kamarck, & Mermelstein, 1983), a 14-itemscale that attempts to measure the degree to which situations in one�s life are perceived asstressful.

(4) Rosenberg�s self esteem scale (SES, Rosenberg, 1989), a well-known 10-item personality ques-tionnaire for measuring self-esteem.

(5) Ray�s (1974) balanced version of Rokeach�s dogmatism scale (BDS; Ray, 1974). The BDS is a36-item scale in which the 18 items measuring the positive pole of dogmatism belong to Rok-each�s well-known dogmatism scale, and the 18 negative-pole items were newly written byRay.

Previous research about the structure and balance of these measures was carried out by thepresent authors (PSS), Ten Berge (1999, SES) and Ray (1974, 1983, BDS).

(6) Finally, the acquiescence scale by Winkler et al. (1982, WKW) was used as it was in the ori-ginal study: embedded within the same 42-item questionnaire about general attitudes towardhealth care that Winkler et al. used.

Except for the Bn scale, all the measures were adapted by two of the present authors separatelyand then the results were discussed until an agreement was reached. Finally, the adapted versionswere revised by a native English teacher of our University�s Language Service.

2.2. Participants and procedure

The respondents were 207 undergraduate psychology students (156 female and 51 male), fromthe Rovira i Virgili University (Spain). Data were collected in the academic year 2002–2003.

Because of limited time, the six measures had to be administered in two sessions. The ARS, Bnand WKW measures were administered in one session, and the PSS, SES and BDS in the other.The order of presentation was balanced between sessions and, within each session, the order inwhich the three measures were administered was also balanced in the 6 possible permutations. Themeasures were completed voluntarily and were administered in classroom groups of about 60pupils. The same individual (one of the present authors) administered the scales in all caseswithout time limits.

1336 P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340

3. Analysis and results

3.1. Preliminary analyses

The procedures involved in this stage will be described only briefly. Additional information anddetails are available on request. First, the factor analytic procedure described above was appliedto the balanced scales. In the three measures, the MAP procedure (Velicer, 1976) suggested thatthe two-factor solution was the most appropriate. Furthermore, the factor solutions were quiteclear, with a first bipolar �content� factor, and a second �acquiescence� factor with loadings thatwere positive and generally smaller than the content loadings. So, in all scales, the variance ex-plained by the content factor was larger than the variance explained by the acquiescence factor.Next, the content scores and the acquiescence scores were obtained as explained above. As for theseparate acquiescence scales, the raw scores were obtained by simply summing the items scored inthe �agree� direction. The means and standard deviations of these scores were 7.56 and 2.52 (ARS)and 16.62 and 2.22 (Bn).

The WKW scale was the only one to present problems. The scores obtained with the systemdescribed above were nearly all zeros, thus indicating that, in the present sample, there werevirtually no double agreements to opposed paired items. Because of this �floor� effect, the reli-ability of the WKW was practically zero, and the correlations between the remaining measureswere virtually null. For this reason we decided not to use this scale in the analyses below.

3.2. Reliability and correlational analyses

For the 5 remaining measures, Table 1 shows the estimated product-moment correlations(below diagonal), reliabilities (diagonal) and disattenuated correlations (above diagonal). For theseparate scale scores, the reliabilities were estimated by coefficient alpha. For the acquiescencefactor scores derived from the balanced scales, the reliabilities were estimated by the split-halfmethod.

The reliabilities in the diagonal were tested using Feldt�s F-test (Feldt, 1965), and they were allfound to be statistically significant. This suggests that the acquiescence scores have some within-scale degree of generalizability, as Ray (1983) hypothesized. It is noted that all the five reliabilityvalues are similar, and that they are quite low for a psychometric instrument. However, they arehigher than most of the previous results we reviewed in the literature.

Table 1

Correlations and reliabilities of the acquiescence measures (N ¼ 207)

Acq-PSS Acq-SES Acq-BDS Bn ARS

Acq-PSS 0.551 0.307 0.515 0.259 0.418

Acq-SES 0.168 0.545 0.198 0.141 0.322

Acq-BDS 0.274 0.105 0.514 0.345 0.569

Bn 0.141 0.076 0.181 0.536 0.886

ARS 0.229 0.175 0.301 0.478 0.544

P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340 1337

The raw product-moment correlations are positive and low, as expected, and the values aresimilar to those reported previously in the literature. The significance of these correlations wastested simultaneously by means of Bartlett�s test of sphericity. The null hypothesis was rejected,thus suggesting that there is some degree of correlation between the measures.

The disattenuated correlations are higher than the raw correlations (as predicted from the lowreliabilities) but they are not so high in the situation we are considering. To see this point, considerthat, if the different scores measured a single strong common trait, then all the disattenuatedcorrelations should be near unity, which clearly is not the case. The only corrected correlation thatis high is the one between the Bn and the ARS scores.

The results described so far suggest that there is some degree of correlation among the measures(although it is very low), but that they do not seem to measure a strong common construct. So, itseems appropriate to factor analyze the data. Note that factor analysis is appropriate here becauseit takes into account that the measures contain measurement error (a considerable degree in thepresent case).

The correlation matrix obtained in the previous step was factor analyzed using the maximumlikelihood procedure. The Kaiser–Meyer–Olkin test of sampling adequacy was 0.64. This result,together with the previous result of Bartlett�s test, suggests that there is some commonality in thematrix and that it can be factor analyzed.

The first two rows of Table 2 show the assessment of fit for the one-factor and the two-factormodels. It should be pointed out that the fit of the one-factor model is almost acceptable, whereasthe fit of two-factor model is excellent.

The obliquely rotated two-factor solution (direct Oblimin) is shown in Table 3. The estimatedinterfactor correlation was 0.42.

The oblique two-factor solution is quite clear and essentially exhibits positive manifold. Thefirst factor, which is very weak, is mainly determined by the acquiescence measures derived from

Table 2

Fit indices for the different models

Model v2 df RMSEA 90% CI NNFI

One-factor 11.30 5 0.077 (0.007; 0.139) 0.87

Two-factor 0.20 1 0.001 (0.00; 0.138) 1.00

One-factor with correlated residuals 1.76 4 0.001 (0.00; 0.019) 1.00

Note: RMSEA, root mean square error of approximation; 90% CI, RMSEA 90% confidence interval; NNFI, non-

normed fit index.

Table 3

Rotated factor pattern of the two-factor solution

Measure F1 F2

Acq-PSS 0.80 )0.09Acq-SES 0.19 0.10

Acq-BDS 0.30 0.19

Bn 0.03 0.50

ARS 0.01 0.93

Table 4

One factor solution with correlated residuals between Bn and ARS

Measure F1

Acq-PSS 0.48

Acq-SES 0.28

Acq-BDS 0.54

Bn 0.31

ARS 0.52

1338 P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340

the balanced scales. The second factor is essentially a doublet defined by the separate acquiescencescales. Conceptually the factor solution again suggests that there might be a weak degree ofconsistency among the measures, but that there is no strong general factor. Furthermore, the only(relatively) high correlation in Table 1 may merely reflect a method effect.

The results discussed so far prompted us to fit a third factorial model. This was a single-factormodel in which the Bn and ARS residuals were allowed to correlate. Conceptually, this modelhypothesizes that the five scores measure a common factor, but that the Bn and ARS scores sharesome common variance that cannot be explained by this factor.

The fit of this model is summarized in the third row of Table 2, and can be qualified asexcellent. The factor solution is shown in Table 4. Note that in this solution all of the factorloadings are moderate or low, so none of the variables can be considered to be a good marker forthe factor. Overall, these results suggest that, once the specific relation between Bn and ARS ispartialled out, the remaining covariance among the measures can be accounted for by a singlefactor. However, these remaining covariances are very small, and so the factor is weak and poorlydefined.

4. Discussion

Ray (1979, 1983) considered that acquiescence has a certain degree of generalization within asingle scale but that the generalization between scale types is far more dubious. The present resultssuggest that there may be some degree of convergent validity among the measures, and that theremay be a common factor that underlies them all. However the degree of convergent validity isvery low, and the factor is weak and poorly defined. Furthermore, part of the covariance betweenthe separate acquiescence scores might be due to method effects. This result and its interpretationcan be related to previous research: Couch and Keniston (1960) correlated the overall agreementscale with the MMPI acquiescence scale (essentially longer versions of the ARS and Bn,respectively) and obtained a correlation of 0.64. In response to this sort of result, Nunnally (1978)conjectured that, in spite of the efforts to make them heterogeneous, the separate scales tend toshare common variance because to some degree they measure common, non-stylistic personalitytraits. For this reason separate scales tend to correlate moderately among themselves, but theyshow far lower correlations with external, non-content-related acquiescence criteria. Nunnally�sconjecture is relevant in the present case because the remaining acquiescence scores used in thepresent study are derived from the balanced scales, and so they are uncorrelated with theirrespective contents.

P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340 1339

Itemmetric analyses of the measures used in this study suggest that other factors should beconsidered. The type of items in the balanced scales are quite different: The BDS items are generalaphorisms, the SES items refer to internal feelings and the PSS items are numerical ratingscontaining expressions of frequency and amount. On the other hand, even though the Bn and theARS contain some item mixture, most of the items in both scales are typical personality itemsreferring to reaction descriptions and trait attributions (see Angleitner, John, & L€ohr, 1986). So,the Bn and ARS scales might share some common variance not only because of some degree ofcontent overlap, but also because the type of items they use are quite similar. Finally it cannot bediscounted that the Bn and ARS scores are affected by socially desirable responding and that it isthe common social desirability variance that is responsible, to a greater or lesser extent, for theobserved correlation. Hofstee et al. (1998) suggest that in non-balanced scales (such as the Bn andARS separate scales) acquiescence and social desirability are inextricably confounded. Eventhough the items of the Bn and ARS scales were chosen so as to be neutral in social desirability,this possibility cannot be totally discounted.

In our opinion, the present results do not encourage further studies into acquiescence as ameaningful and potentially valuable personality trait: Both the reliabilities and the convergentvalidity are too low even for purely research purposes. Faced with these results, nobody can reallyexpect to find substantive relations with external variables: continuing the validation process,therefore, does not seem to make too much sense. To sum up, there seems to be a certain degree ofconvergent validity that makes it possible to consider acquiescence as a trait. If this were to be so,however, the trait would be weak, very unreliable and probably externally invalid. In theseconditions, possibly the most appropriate thing to do would be to focus on controlling andsuppressing acquiescence. If �cleaner� (acquiescence-free) content scores can be obtained, perhapsthe low predictive validity that personality scores usually have could be improved.

Bock, Dicken, and van Pelt (1969) conjectured that university students tend to respond moreconscientiously to scale content than other populations, and some of the present results tend togive support to this conjecture. In particular, in our sample, there were virtually no doubleagreements to opposed paired items in the WKW scale, which made this scale unsuitable for thepresent study. Perhaps in a more heterogeneous sample, which contained individuals expected �apriori� to be less conscientious, a clearer and stronger structure might have been obtained.However, it should be pointed out that both the reliabilities and the correlations obtained in thepresent study are similar or even higher than those reported previously in the literature.

References

Angleitner, A., John, O. P., & L€ohr, F. J. (1986). It�s what you ask and how you ask it: an itemmetric analysis of

personality questionnaires. In A. Angleitner & J. S. Wiggins (Eds.), Personality assessment via questionnaires (pp.

61–107). Berlin: Springer-Verlag.

Bass, B. M. (1956). Development and evaluation of a scale for measuring social acquiescence. Journal of Abnormal and

social Psychological Measurement, 26, 291–309.

Bock, R. D., Dicken, C., & van Pelt, J. (1969). Methodological implications of content-acquiescence correlation in the

MMPI. Psychological Bulletin, 71, 127–139.

Cogan, L. C., Conklin, A. M., & Hollingworth, H. L. (1915). An experimental study of self-analysis, estimates of

associates, and the results of tests. School & Society, 2, 171–179.

1340 P.J. Ferrando et al. / Personality and Individual Differences 37 (2004) 1331–1340

Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress. Journal of Health and Social

Behavior, 24, 385–396.

Couch, A., & Keniston, K. (1960). Yeasayers and naysayers: agreeing response set as a personality variable. Journal of

Abnormal and Social Psychology, 60, 151–174.

Feldt, L. S. (1965). The approximate distribution of Kuder–Richardson reliability coefficient twenty. Psychometrika,

30, 357–370.

Fricke, B. G. (1957). A response bias (B) scale for the MMPI. Journal counselling Psychology, 4, 149–153.

Gudjonsson, G. H., & Clare, I. C. H. (1995). The relationship between confabulation and intellectual ability, memory,

interrogative suggestibility and acquiescence. Personality and Individual Differences, 19, 333–338.

Hanley, C. (1961). Social desirability and response bias in the MMPI. Journal of Consulting Psychology, 25, 13–20.

Hofstee, W. K. B., Ten Berge, J. M. F., & Hendriks, A. A. J. (1998). How to score questionnaires. Personality and

Individual Differences, 25, 897–909.

Jackson, D. N., & Messick, S. J. (1965). Acquiescence: the nonvanishing variance component. American Psychologist,

20, 498.

Martin, J. (1964). Acquiescence––measurement and theory. British Journal of Social and Clinical Psychology, 3, 216–

225.

McGee, R. K. (1962). Response style as a personality variable: by what criterion? Psychological Bulletin, 59, 284–295.

Mellenbergh, G. J. (2001). Outline of a faceted theory of item response data. In A. Boomsma, M. A. J. van Duijn, &

T. A. B. Snijders (Eds.), Essays on item response theory (pp. 415–432). New York: Springer.

Messick, S. J. (1967). The psychology of acquiescence: an interpretation of research evidence. In I. A. Berg (Ed.),

Response set in personality assessment (pp. 115–145). Chicago: Aldine.

Morf, M. E., & Jackson, D. N. (1972). An analysis of two response styles: true responding and item endorsement.

Educational and Psychological Measurement, 32, 329–353.

Nunnally, J. C. (1978). Psychometric theory. New York: McGraw Hill.

Paulhus, D. L. (1991). Measurement and control of response bias. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman

(Eds.), Measures of personality and social psychological attitudes (vol. 1, pp. 17–60). New York: Academic Press.

Ray, J. J. (1974). Balanced dogmatism scales. Australian Journal of Psychology, 26, 9–14.

Ray, J. J. (1979). Is the acquiescent response style not so mythical after all? Some results from a successful balanced F

scale. Journal of Personality Assessment, 43, 638–643.

Ray, J. J. (1983). Reviving the problem of acquiescent response bias. Journal of Social Psychology, 121, 81–96.

Rorer, L. G. (1965). The great response-style myth. Psychological Bulletin, 63, 129–156.

Rosenberg, M. (1989). Society and the adolescent self image (2nd ed.). Middletown: Wesleyan University Press.

Seisdedos, N., & Cordero, A. (1975). Cuestionario de personalidad MMPI. Madrid: TEA Ediciones.

Ten Berge, J. M. F. (1999). A legitimate case of component analysis of ipsative measures, and partialling the mean as an

alternative to ipsatization. Multivariate Behavioral Research, 34, 89–102.

Velicer, W. F. (1976). Determining the number of components from the matrix of partial correlations. Psychometrika,

41, 321–327.

Winkler, J. D., Kanouse, D. E., & Ware, J. E. (1982). Controlling for acquiescence response set in Scale Development.

Journal of Applied Psychology, 67, 555–561.