British Journal of Social & Clinical Psychology, 1972, 11, pp. 346-352

(With two post-publication addenda following the original article)



University of New South Wales

It is proposed that the low negative correlation observed between positive and negative halves of balanced F scales is not atypical of scales in this general content area. An extreme form of this view is Kerlinger's theory that liberalism and conservatism are innately orthogonal. This theory was tested using two conservatism scales: one a revision of Wilson & Patterson's C Scale and one a scale using items in the conventional format. Low negative correlations between positive and negative halves were found in both cases. It was reasoned that orthogonality between one-way worded scales indicates a negative relationship in the real attitudes being measured. Kerlinger's theory is therefore rejected as inadequate and as contra-indicated by the results. It is concluded, however, that some balanced versions of the California F Scale may have been too hastily rejected. Conversely, it is held that the Wilson & Patterson C Scale might be rejected in favour of one of the new conservatism scales developed in this study. This is because the original C Scale showed a positive correlation between its supposedly negative and positive halves.

The title of this article is a deliberate allusion to the important article by Christie et al. (1958), wherein an extensive report was made of attempts to construct versions of the California F Scale such that half the items were negatively worded. The authors were sufficiently cast down by the generally low levels of the negative correlation observed between original and reversed versions of the F Scale to conclude that the F Scale might be inherently 'irreversible'.

The question that is advanced here, however, is: What levels of negative correlation ought we to expect between the two halves of any balanced scale? Are the correlations found with the F Scale in fact atypical? Grounds for believing that a low negative correlation between positively and negatively scored halves might be typical not only of F Scale versions but also of conservatism scales in general are to be found in an article by Kerlinger (1967). Kerlinger in fact asserts that complete orthogonality is the normal and proper relationship between liberalism and conservatism items. If this is so, the absence of high negative correlations between the two halves of a balanced F Scale might be seen as less surprising than has heretofore been the case.

If Kerlinger is right, what are the implications for the practice of using balanced scales to control out acquiescent response set (Cloud & Vaughan, 1970)? This practice seems to be predicated on the assumption that positive and negative items are equivalent in content and differ only in the effect on them of acquiescent set. If they are not equivalent in content (i.e. if they are uncorrelated) can they continue to be used both in the one scale?

Before this issue can be cleared up, a thorough empirical test of Kerlinger's theory on several sorts of attitude scale seems called for. Let it initially be noted, however, that it is not in fact impossible for scales to be constructed wherein the correlation between the two halves is high and negative (see, for example, Ray, 1970)

One of the most interesting developments in the attitude measurement field in recent years is the proposal by Wilson & Patterson (1968), not for a new type of measurement, but for a new type of attitude scale item. Instead of using a full, multicomponent sentence to which a subject must respond, Wilson & Patterson propose that a single word or phrase may be used. This, they believe, makes for a more valid and more reliable scale. Such items also fit in with Kerlinger's (1967) thinking on the meaning of attitude measurement. The 'short-form' item seems indeed to get very close to what Kerlinger calls the 'criterial referent' in attitude measurement. It is rather unsatisfactory to test Kerlinger's theory that the criterial referents of liberal and conservative items are independent (orthogonal) using the conventional type of attitude item. One is inclined to distrust the results as being too far removed from the central concern of Kerlinger's theory. With the short-form items, however, a more convincing test can be carried out.

Insofar as it is an assertion about general social attitudes, the only empirical evidence Kerlinger marshals in support of his theory is the correlation among a group of university students between two sets of items especially written by himself. Obviously, his results need not be typical of the relationship between the two halves of fully developed attitude scales. It could well be the case that the normal item-analysis methods used in scale construction eliminate liberalism and conservatism items which are not negatively related with one another.


It was proposed to apply both the Wilson & Patterson C Scale and a conservatism scale in the more conventional format to a general population sample. It was desired to supplement the C Scale with a more conventional scale in order to increase the generalizability of the results. This is particularly important because the relationship between scales in the new and old formats does not yet appear to have been documented. It could well be the case that conservatism as measured by the new-format scale differs substantially from conservatism as measured by an old-format scale.

The general conservatism scale in conventional format which was finally chosen was a balanced instrument devised by the present author to measure the concept of non-economic conservatism developed by Lipset (1960) in his paper on 'Working Class Authoritarianism'. It did therefore include a considerable number of California F Scale items but was nonetheless completely balanced (12 negative and 12 positive items) and had satisfactory reliability. Note that there is a necessary relationship between internal consistency and test-retest reliability (Cronbach, 1951). On both the samples described below scale reliabilities will therefore be presented using the "alpha" statistic. Although not normally so calculated, this statistic may be viewed as a ratio between internal consistency and test length (see Lord & Novick, 1968). As a case of the Spearman-Brown prophecy formula therefore it is an estimate of test-retest reliability.

To avoid unnecessary duplication, the general population results for the old-format scale will in fact be derived from the norming sample data for that scale. This sample comprised 404 Australian National Servicemen (Army conscripts) at the Singleton training battalion in 1968 (the entire intake after incomplete forms had been discarded).* As National Servicemen are selected at random by a birth-date ballot procedure from the entire 20-year-old male population, they provide a good sample for social research where wide generalizability of the results is desired. They are superior to a door-to-door sample in that there is little or no volunteer artifact (see Patterson & Wilson, 1969; Sheridan & Shack, 1970). The final form of the scale was produced from an initial pool of 65 items. Item analysis was carried out by the usual procedure of item-to-total correlations through successive reductions in length of the scale. This method is characterized by greater precision than the upper-lower thirds procedure used by Wilson & Patterson.

The scale so produced was then readministered in its original form (i.e. of 65 items) to the entire class (n = 262) of first-year psychology students at Macquarie University. (Only the items finally selected on the Army sample were, of course, used.)

The properties of the Wilson & Patterson scale were examined in a separate study with 110 National Servicemen at the Kapooka training centre in 1969. Because Ray (1971a) has shown that the Wilson & Patterson scale is characterized by very poor reliability on such a sample, 80 new items were included in the battery to allow construction of a revised and more reliable scale. These 130 items were also included in the questionnaire given to the 262 psychology students mentioned above.


The reliability of the revised C Scale on the Army sample was 0.84 and on the student sample 0.87. Both these coefficients represent levels of reliability that are more than acceptable (see Shaw & Wright, 1967). The correlation between the 25 positive and 25 negative items was -0.373 for the Army sample and -0.181 for the student sample. The equivalent correlation for the original C Scale on the student sample was 0.288. Because a positive correlation is so anomalous, the result was rechecked using the most basic FORTRAN and found to be confirmed. See Table 1 for the final form of the revised scale. Twenty-nine out of 50 items come from the Wilson & Patterson original.

Table 1. The items of the revised C Scale: a measure of general conservatism using short-item format and normed on a sample of 110 Australian National Servicemen.

(Items are scored 3 for 'yes', 2 for '?' and 1 for 'no'. Preamble is as in Wilson & Patterson (1968). Items marked R are reverse-scored. Reliability (alpha) is 0.84 for National Servicemen and 0.87 for first-year university students.)

1 . R Evolution theory..............26. R Jazz
2. .. School uniforms...............27. R Casual living
3. R Striptease shows.............28. R Divorce
4. R Beatniks..........................29. R Pyjama parties
5. R Modern art.......................30. R Mercy killing
6. R Working mothers.............31. R Protest marches
7. R Birth control.....................32. .. Debutantes
8. .. Military drill.......................33. R Law reform
9. .. Moral training...................34. .. Respect for parents
10. R Suicide...........................35. .. Common decency
11. .. Chaperones....................36. R Legalized homosexuality
12. R Legalized abortion..........37. R Pot-smoking
13. .. Empire-building...............38. R Love-ins
14. R Student pranks...............39. .. Imperialism
15. .. Licensing laws.................40. .. Conformity
16. .. Chastity............................41. .. The police
17. .. Royalty............................42. R Soviet Russia
18. .. Conventional clothes.......43. .. Teetotalling
19. .. Apartheid.........................44. .. Defence spending
20. R Nudist camps..................45. .. Religion
21. R Disarmament..................46. .. God
22. .. Censorship......................47. .. Authority
23. R White lies........................48. R Political demonstrations
24. R Mixed marriage...............49. .. Nationalism
25. .. Strict rules.......................50. .. My country

The means and S.D.s for the student sample were 94.55 and 15.52. For the Army sample they were 103.03 and 13.60. Z for the difference (5.30) is significant (P< .0001) and this represents validation for the new scale of the criterion groups type. That students are more radical on social issues is of course known a priori (Lipset, 1960, 1965).

As a matter of interest it is noted that the new scale has the characteristic of presenting respondents as more conservative in the absolute sense than does the original. When both scales are scored so that a high score represents high conservatism, the mean on the original scale for the student sample was 87.17 (cf. 94.55 above). As will be seen later this is probably the outcome of the greater degree of meaning opposition between negative and positive items observable in the revised scale.

The validity check for the new scale was of the same sort as that used by Wilson & Patterson (criterion groups).

Table 2. The items of the General Conservatism Scale

(alpha = 0.70 on 404 conscripts and 0.78 on 262 university students. Response options are: strongly agree (scored 5), agree (4), not sure (3), disagree (2), strongly disagree (1). Items marked R are reverse-scored, i.e. 1 to 5.)

1. Allowing educated Asians to immigrate benefits Australian society. (R)
2. Australia should aim at closer contact with communist China. (R)
3. The white-Australia policy is a good policy because it helps to keep Australia white.
4. We must be careful not to let too many Asians into the country or they'll take over the place.
5. People who hold communist beliefs should not be allowed to hold high positions in the public service.
6. Asians should be allowed to emigrate to Australia. (R)
7. The Japanese are very productive people and should be allowed to settle in Australia. (R)
8. Although many details still remain to be worked out, we now have definite answers to most practical problems in life.
9. Human beings are more important than efficiency. (R)
10. Young people sometimes get rebellious ideas, but as they grow up they ought to get over them and settle down.
11. Human life is sacred. (R)
12. There is seldom any reason to hurt people's feelings. (R)
13. Nobody ever learned anything really important except through suffering.
14. An insult to our honour should always be punished.
15. Sex crimes, such as rape and attacks on children, deserve more than mere imprisonment; such criminals ought to be publicly whipped or worse.
16. All men are equal. (R)
17. Homosexuals are hardly better than criminals and ought to be severely punished.
18. Most of our social problems would be solved if we could somehow get rid of the immoral, crooked and feebleminded people.
19. People can be divided into two distinct classes: the weak and the strong.
20. Individual freedom is a basic human right. (R)
21. Crimes of violence should be punished by flogging.
22. The death penalty is barbaric and should be abolished. (R)
23. It is wrong to punish a man if he helps another country because he prefers it to his own. (R)
24. Our treatment of criminals is too harsh; we should try to cure them, not punish them. (R)

The alpha for the conventional format scale was 0.70 on the army sample and 0.78 on the student sample. Means and S.D.s were 53.18 (11.31) and 71.44 (10.38), respectively. Again this difference is significant at the P < 0.0001 level. The correlation between positive and negative halves was -0.206. The correlation between the two new scales on the student sample was 0.473. This is highly significant but, considering that both scales are said to measure the same thing, surprisingly low. It is nonetheless high enough to say that the two scales provide concurrent validation for one another.

Both scales were similar in predictive validity. The revised C Scale correlated 0.354 with conservatism of actual political party preference, while the standard format scale correlated 0.312. (Again these results were available for the student sample only.)

Because of the great difference in reliabilities (0.63 v. 0.84) most emphasis has been given in the above results to data derived from the revised rather than the original C Scale.

The items of the two new scales are given in Tables 1 and 2.


It has been shown that both the revised C Scale and the old-format general conservatism scale have satisfactory reliability on both general population and student samples (cf. Shaw & Wright, 1967). Both scales have been provided with validity evidence of the criterion-group, concurrent and predictive sorts. The relation between their positive and negative halves should therefore provide a good test of Kerlinger's theory and give us a base with which to compare the results derived from balanced versions of the F Scale.

Kerlinger's theory maybe decisively rejected on the basis of the general population samples (r= -0.206 and -0.373; P< .0001), but the result on the student sample with the revised C Scale is quite close to the orthogonality he claims. This would seem to imply that the measured attitudes of students are more multidimensional than those of people at large. The interesting thing, however, is that by normal criteria both the original and revised C scales are fully satisfactory on this sample -- with alpha coefficients of 0.83 and 0.87 respectively. The revised scale was in fact constructed by almost completely mechanical means -- using the built-in decision procedures of ITRO, a conventional item-analysis program (Ray, 1972). This program takes no account of which items are positive and which are negative -- and yet the maximally reliable scale produced by this program proved to be almost completely balanced (22 out of 48 items were negative) and only minor discretionary adjustments were needed to produce the scale given in Table 1. It would seem therefore that it must be regarded as an open possibility that any balanced conservatism scale in normal use might prove to be composed of roughly orthogonal subsets of negative and positive items. If the balanced F scales are to be regarded as unsatisfactory because of the low levels of correlation between positive and negative halves, they are certainly not alone in this.

We have above all three possibilities demonstrated. In the revised C Scale the positive and negative items are negatively correlated on the norming sample and almost orthogonal on the student sample, while in the original C Scale they are positively correlated. This is certainly a more complicated picture than the one presented by Kerlinger, or, indeed, by any previous writer. How may we give a systematic account of these phenomena?

The first distinction that needs to be made is between measured attitudes and actual attitudes. The two will differ because of acquiescent and social desirability response sets. It is, of course, measured attitudes that we have been considering here and it is believed that the above results may be explained by variations in acquiescent response set. This set is one which causes items to be responded to as if they were alike. If the attitudes they reflect are in fact neither alike nor opposed, acquiescent response set should be enough to ensure an observed positive correlation. If the real attitudes are in fact alike, this set should cause the correlation to be even higher than it otherwise might. Therefore, if orthogonality is observed between two sets of items and we know acquiescent response set to be operative, it is thereby implied that the real attitudes must be opposed. In other words, the orthogonality is the outcome of the countervailing forces of real meaning and acquiescent response set. Both degree of meaning opposition and acquiescent response set are, however, variable quantities. If the former is weak and the latter strong we will have results much as are observed with Wilson & Patterson's scale. If both are strong we will have results such as were obtained on the student sample with the revised scale. If the meaning opposition is exceptionally strong (as with the more simplistic, black and white attitudes of National Servicemen) we have results such as the -0.373 correlation in the revised C Scale.

At this point one may ask what policy decisions we should now take regarding the use of balanced F and conservatism scales. Can we, as Cloud & Vaughan (1970) contend, continue to use them for controlling out acquiescent set in our final results? Let it be observed that no matter whether the two halves are orthogonal or negatively correlated, the total score will still represent the intersection of what is measured by both. This can be seen in the fact that the correlation with the total score of the revised C Scale on the student sample was .759 for the negative half and 0.777 for the positive half. The controversy over the F Scale might then have been better couched in terms of producing a balanced scale of high reliability. If the two halves within themselves are internally consistent, the reliability can still be high. Indeed, a high internal consistency within the two halves can more than offset lack of correlation between the two halves. As was shown in the student sample here, the reliability for the revised C Scale rose even though the correlation between the two halves dropped. The importance of the correlation between the two halves then lies in the area of evidence about construct validity. How successful is our measure of a construct if our expectations about the meaning of our items are so violated that supposedly negative items in fact correlate positively? Our knowledge of the distorting influence of acquiescent set may make us prepared to accept low negative correlations between both halves, but we would have to have especially strong evidence of validity from other sources before we were prepared to accept scales wherein the correlation between the halves was non-significant or significantly positive. This being so, it seems to the present author that the original Wilson & Patterson scale might be deemed unsatisfactory for further use (cf. Oskamp & Thompson, 1970) but several of the balanced F scales produced may in fact be satisfactory (providing their reliability can be shown to be high). See, however, Ray (1971b) on the question of F Scale validity.


The role of balanced scales in controlling out response set is therefore still seen as valid. The only proviso is that attention should be given to their reliability as much as to the correlation between their positive and negative halves.

Kerlinger's theory is seen as taking insufficient account of acquiescent response set. Although his finding of orthogonality in liberal and conservative attitudes measured by one-way worded scales has received some support, the implication of this is seen to be that the real attitudes which underlie responses to such scales are opposed (negatively correlated).

The revised form of the C Scale which has been presented is superior to the original in general population reliability and in the more satisfactory relationship between its positive and negative halves.


* I would like to thank the Department of the Army for making this work possible.


CHRISTIE, R., HAVEL, J. & SEIDENBERG, B. (1958). Is the F scale irreversible? J. abnorm. soc. Psychol. 56, 143-158.

CLOUD, J. & VAUGHAN, G. M. (1970). Using balanced scales to control acquiescence. Sociometry 33, 193-202.

CRONBACH, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika 16,297-334.

KERLINGER, F. N. (1967). Social attitudes and their criterial referents: a structural theory. Psychol. Rev. 74, 110-122.

LIPSET, S. M. (1960). Political Man. New York: Doubleday.

LIPSET, S. M. (1965). Students and politics. In S. M. Lipset & S. S. Wolin (eds.), The Berkeley Student Revolt. New York: Doubleday.

LORD, F. M. & NOVICK, M. R. (1968). Statistical Theories of Mental Test Scores. Reading, Mass.: Addison-Wesley.

OSKAMP, S. & THOMPSON, G. (1970). Internal consistency in the stereopathy-acquiescence scales J. soc. Psychol. 81, 73-77.

PATTERSON, J. R. & WILSON, G. D. (1969). Anonymity, occupation, and conservatism. J. soc. Psychol. 78, 263-266.

RAY, J.J. (1970) The development and validation of a balanced Dogmatism scale. Australian Journal of Psychology, 22, 253-260.

RAY, J.J. (1971a) "A new measure of conservatism" -- Its limitations. British Journal of Social & Clinical Psychology, 10, 79-80.

RAY, J.J. (1971b) An "Attitude to Authority" scale. Australian Psychologist, 6, 31-50.

RAY, J.J. (1972) A new reliability maximization procedure for Likert scales. Australian Psychologist 7, 40-46.

SHAW, M. E. & WRIGHT, J. M. (1967). Scales for the Measurement of Attitudes. New York: McGraw-Hill.

SHERIDAN, K. & SHACK, J. R. (1970). Personality correlates of the undergraduate volunteer subject. J. Psychol. 76, 23-26.

WILSON, G. D. & PATTERSON, J. R. (1968). A new measure of conservatism. Br. J. soc. clin. Psychol. 7, 264-269.

Manuscript received 14 April 1971

1. The correlation between the positive and negative halves of the original C-scale is not given for the sample of 110 army conscripts mentioned above. That the correlation was -0.199 is however given in Ray & Pratt (1979). This is at least in the right direction but still indicates that the two halves of the scale give essentially unrelated scores on conservatism.
2. A subsequent analysis of results from the original C-scale obtained by others shows that satisfactory (highly negative) correlations between the positively-scored and negatively-scored halves of the original C-scale were obtained on all seven occasions concerned. Such a finding does of course constitute some warrant for continued confidence in the usefulness of the scale. That highly unsatisfactory correlations (such as the significant positive correlation between supposedly opposed items mentioned above) can also be obtained, even with students, does mean, however, that the many studies which fail to report the correlation between the "liberal" and "conservative" items continue to be of dubious meaning. Accepting their results at face value is therefore a matter of faith rather than of science.

Ray, J.J. & Pratt, G.J. (1979) Is the influence of acquiescence on "catchphrase" type attitude scale items not so mythical after all? Australian Journal of Psychology 31, 73-78.

Go to Index page for this site

Go to John Ray's "Tongue Tied" blog (Backup here)
Go to John Ray's "Dissecting Leftism" blog (Backup here)
Go to John Ray's "Australian Politics" blog (Backup here)
Go to John Ray's "Gun Watch" blog (Backup here)
Go to John Ray's "Education Watch" blog (Backup here)
Go to John Ray's "Socialized Medicine" blog (Backup here)
Go to John Ray's "Political Correctness Watch" blog (Backup here)
Go to John Ray's "Greenie Watch" blog (Backup here)
Go to John Ray's "Leftists as Elitists" blog (Not now regularly updated)
Go to John Ray's "Marx & Engels in their own words" blog (Not now regularly updated)
Go to John Ray's "A scripture blog" (Not now regularly updated)
Go to John Ray's recipe blog (Not now regularly updated -- Backup here)

Go to John Ray's Main academic menu
Go to Menu of recent writings
Go to John Ray's basic home page
Go to John Ray's pictorial Home Page (Backup here)
Go to Selected pictures from John Ray's blogs (Backup here)
Go to Another picture page (Best with broadband)