THE MEASUREMENT OF ATTITUDES TOWARD AFRICANS: A UNIDIMENSIONAL SCALE WITH HIGH DISCRIMINATING POWER. A. M. COLMAN 1 University of Cape Town

Transcription

1 Reprinted from "Psycilologia Ajricallu." Volume 14. No. /. / PSYCHOLOGIA AFRICANA. 1971, 14, THE MEASUREMENT OF ATTITUDES TOWARD AFRICANS: A UNIDIMENSIONAL SCALE WITH HIGH DISCRIMINATING POWER A. M. COLMAN 1 University of Cape Town The requirements of a sophisticated scale for measuring attitudes toward Africans are outlined, and the need for such a scale is emphasized. An account is given of the construction of a scale by means of a procedure based on a modification of the scale-discrimination technique. The procedure included Thurstone scaling, item analysis and scalogram analysis, and resulted in a largely unidimensional scale with a wide diversity of item content and range of scores, corrected as far as possible for the operation of response styles, and possessing high reliability. Evidence for the concurrent and construct validity of the scale is discussed. A scale to measure attitudes toward Africans, if it is to provide scores which can be interpreted confidently and meaningfully, should possess the following attributes: (a) it should be unidimensional, measuring as far as possible only attitudes toward Africans; (b) it should possess wide diversity of item content (its unidimensionality should not be due to the fact that it consists merely of rephrasings of the same statement); (c) it should characteristically generate a wide range of scores; (d) it should be corrected as far as possible for the operation of response styles; (e) it should be reliable; and (f) it should possess high discriminating power and validity. No scale which possesses all these attributes has yet been published, and in particular no attempt appears to have been made to construct a unidimensional scale for measuring attitudes toward Africans. The most widely-known scale for this purpose, the simple Thurstone scale developed by MacCrone (9) has, in any event, been largely abandoned, partly because of the use of the word "Native" in each item. Particularly in Africa, this word has acquired a negative connotation, and has quite a different meaning today from the meaning it possessed before the decolonization process began. As a result, more recent research involving the measurement of attitudes toward Africans has usually been carried out with the use of crude measuring instruments (e.g. Pettigrew, 11; Bloom, De Crespigny, and Spence, 2; Van den Berghe, 15; and Mann, 10). This paper describes the construction of a more sophisticated scale than the ones currently available. Method The method of scale construction was based on a modification of the scalediscrimination technique developed by Edwards and Kilpatrick (4). This consists of a synthesis of several well-known procedures, and yields a method of attitude scaling which retains all the advantages, and eliminates most of the disadvantages of Thurstone (13, 14), Likert (8) and Guttman (7) scaling procedures. In particular, unidimensionality, a wide diversity of item content, and a wide range of scores are guaranteed, the operation of response styles is kept at a minimum, and high discriminating power is assured, thereby giving excellent chances of achieving high validity. INow at the University of Leicester, U.K.

2 THE MEASUREMENT OF ATTITUDES TOWARD AFRICANS: A UNIDIMENSIONAL SCALE 33 WITH HIGH DISCRIMINATING POWER Scale and Q values In general, psychologists tend to be vague about the sources from which they obtain statements as raw material for developing attitude inventories. Some are written by the investigators, while others are obtained from newspaper articles, books, and magazines relating to the psychological object. Such items can be criticized on the grounds that they do not always represent common and currentlyheld opinions. It follows that the investigator's intuition sets an upper limit to the usefulness of the resulting scale. The common source from which all the inventory statements used in the present scale were drawn, was a series of protocols of small-group discussions among first-year psychology students at various South African universities, stimulated by a controversial tape-recorded political speech. The statements extracted from these group discussions had been carefully edited for purposes of an undergraduate research project at the University of Cape Town (Callias, unpublished). One hundred evaluative statements referring to Africans were selected from these protocols, according to the informal criteria which have been set out by Wang (16), Likert (8), and Edwards (3). Some of the statements were slightly rephrased: wherever the words "non-white", "native", "non-european", "Bantu", etc., appeared, the word "African" was substituted, since these words are themselves attitude-expressive. In this way the value-loading of the word was held constant, and it was ensured that the psychological object was unambiguously implied by the items in the scale. Each statement was printed on a card, and the cards were arranged in bundles each containing a hundred statements in random order together with eleven cards numbered 1 to 11. The bundles were distributed among a judging group of secondyear psychology students at the University of Cape Town, who were requested to spread the numbered cards out in front of them, and on card 6 they were asked to write "neutral". The judges were then requested to sort the hundred statements into eleven piles, from the least favourable toward Africans on the left, to the most favourable toward Africans on the right, with pile 6 containing statements judged to be neutral toward Africans. Inevitably, several judges responded with their own agreement or disagreement instead of the degree of favourableness or unfavourableness implied by the statements themselves. In an effort to detect this, two statements which seemed prima facie to be most unfavourable toward Africans, and two which seemed to be most favourable were selected, and any arrangement by a member of the judging group which contained one or more of these statements in one of the middle five categories was discarded. Furthermore, following the traditional procedure in constructing Thurstone scales, any arrangement which contained 30 or more statements in one category was discarded on the assumption that the judge had been careless or lazy. The responses of 50 judges remained, and from these responses scale and Q values were calculated with the method described by Edwards (3). Summated Ratings and Item Analyses The 45 statements displaying the highest Q values were discarded and the remaining 55 were retained for Likert scaling. These statements were typed in the form of an inventory with four response categories: "strongly disagree", "disagree",

3 34 CoLMAN, A. M. "agree", "strongly agree" after each item. For purposes of preliminary analysis these response categories were weighted 0 through 3, such that 3 always indicated the response most favourable toward Africans. The appropriate direction of the weighting (which is not at all obvious in statements approaching neutrality on the Thurstone continuum) was not decided upon arbitrarily as is customary in the construction of Likert scales, but was determined by the position of the statements on the Thurstone continuum. The resulting summated rating scale was distributed among a group of students enrolled for the introductory course in psychology at the University of Cape Town, and a score was obtained for each respondent by summing his scores for each item. Once again the returns were scrutinised for any obviously careless or facetious responses. Such returns, together with those in which items had been omitted, were discarded, leaving 165 returns for item analysis. The top 25 % and bottom 25 % of respondents (in terms of their total Likert scores) were selected, and for each statement a distribution was obtained showing the frequency of responses in each response category for these two groups. The response categories were then dichotomised so as to minimise the total number of responses of members of the second group in the pro-african direction and maximise the number of responses of the first group in the pro-african direction, after the manner described by Edwards (3). New weights of 0 and 1 were assigned to each statement according to the dichotomy. The four forcing response categories were retained, but they were scored dichotomously in subsequent analyses. On the basis of the item analysis, each statement was assigned a phi coefficient, determined from Guilford's (6) nomographs. Scalogram Analysis The 55 statements were then ranked in ascending order of their Thurstone scale values. From this ensemble, 28 statements with relatively high phi coefficients were selected, in such a way that a wide range of Thurstone scale values was represented in the items selected. Two forms of the scale (A and B) were then prepared by assigning items to either form alternately in order of their scale values. For purposes of scalogram analysis, the responses of 100 of the 165 respondents used in the summated rating procedure were randomly chosen, and their responses were re-scored using the new dichotomised response weights. Application of scalogram analysis to each form separately resulted initially in coefficients of reproducibility of in each case. After removing the two items from each form which accounted for the greatest amount of error, Reps were increased to O 825 for form A, and for form B. The two equivalent forms of the resulting scale are given in Table 1. A line is drawn under the response category or set of response categories for each statement which is scored 1. All other responses are scored O. The underlining should be omitted when testing SUbjects. Results and Discussion The scale consists of two equivalent forms containing 12 items each. The response to each item is scored 1 or O. giving the scale as a whole a range of scores from 0 to 24. The exclusion of statements with high Q values ensures that the items in the scale are all relatively unambiguous.

4 THE MEASUREMENT OF ATTITUDES TOWARD AFRICANS: A UNIDIMENSIONAL SCALE 35 WITH HIGH DISCRIMINATING POWER Table 1 The Attitudes Toward Africans Scale Below you will find a list of statements about Africans. Please respond to each statement by circling the code opposite it. Thus, if you strongly disagree, circle "SD"; if you disagree on the whole, circle "D"; if you agree on the whole, circle "A"; and if you strongly agree, circle "SA". FORM A I. The Africans should be liberated now. 2. The Africans are not capable of being educated or civilized. 3. Africans are still essentially barbarians. 4. Africans are not fit for the opportunities given to Europeans. 5. Trying to educate the African only results in uprisings and clashes. 6. Africans are uncivilized. 7. If the Africans enjoyed better environmental conditions, in time they would have as many prominent people as Whites have. 8. Africans enjoy their low standard of living. 9. Africans are slovenly, dirty, and not house-proud. 10. The African is basically peace-loving, but the government's racial policy incites him to violence. 11. Africans would not be able to appreciate their rights if we had complete democracy. 12. It will take years before we can give the African equal rights without disastrous results. SV D A SA FORM B 1. If the Africans absorb the Europeans, civilization will flourish. 2. The African is ignorant and savage. 3. Africans are not ready for freedom yet. 4. Africans are quite happy to live in absolutely disgraceful conditions. 5. The African is essentially lazy. 6. Intermarriage with Africans would degrade the white race. 7. Africans should gradually be given more and more say in the government. 8. Africans are on a par with whites. 9. The Africans are several decades behind the Europeans in moral development. 10. Africans are too lazy and ignorant to support themselves. 11. Educated Africans are as good as whites. 12. Africans bring their bad habits to town with them. SDDASA SD 0 A SA SO D A SA The items are drawn from widely divergent points on the Thurstone continuum; they therefore represent both favourable and unfavourable opinions concerning Africans. This ensures that acquiescence response style is kept at a minimum, and that the scale does not consist of simple rephrasings of the same opinion. Extreme response style, "evasiveness" and "criticalness" (Frederiksen and Messick, 5) are probably minimised by the use of four forcing response categories which do not include a "neutral" category. The coefficients of reproducibility of the two forms (0' 864 and O' 825) give the scale as a whole a Rep. of 0'85, which implies that the scale is satisfactorily unidimensional. The value of Rep. should be considered in conjunction with the

5 36 COLMAN, A. M. mean of the modal response categories in making inferences about dimensionality. The mean of the modal response categories in this scale is O' 67 (the range is O' 51 to O 79 for form A, and O' 51 to O' 84 for form B). Since this figure is fairly low, the inference concerning the scale's unidimensionality is justified. The phi coefficients in form A range from O' 51 to o 73, with a mean of O' 65; in form B the range is from 0 54 to O' 83 with a mean of O' 69. Thus the scale as a whole has a mean phi coefficient of O' 67 indicating that the discriminating power is extremely high. This is a most encouraging figure, especially when one notes that Edwards and Kilpatrick (4) did not originally obtain such a high discriminating power for their scale. The equivalent-form reliability of the scale has been computed from the scores of a group of 60 first-year psychology students at the University of Cape Town. Applying the Brown-Spearman formula to the correlation between the two forms, a coefficient of equivalence of O' 88 was obtained. The coefficients of reproducibility, of course, set a lower bound to the split-half reliability of each form. The split-half reliability of form A, is therefore above O' 83, and of form B, above With regard to the validity of the scale, in the absence of a sound criterion, indirect evidence has to be sought, unless one accepts the interpretation of the mean phi coefficient which equates it with concurrent validity (Guilford, 6). Three independent unpublished studies at the University of Cape Town. by Lambley, Morris, and Colman (unpublished), have provided such indirect evidence. In the first, highly significant correlations were obtained between each of the forms and a simple social distance questionnaire, with P<' 01 in each. Secondly, high scorers on the scale as a whole have been shown to display significantly greater accuracy (P<' 05) in a race identification task involving binocular resolution than low scorers. Finally, it has been demonstrated in a factorial study that authoritarianism as measured by a forced-choice F scale (Strickland and J anicki. 12), and independence of judgement as measured by the Barron-Asch Scale (Barron, 1, Ch. 14) contribute significantly (P<' 01 in each case) to the variance in attitudes toward Africans given by the scale as a whole. Taken together, these findings bestow on the scale a measure of construct validity. A word needs to be said, in conclusion, about the applicability of the scale outside the population on which it was standardised. Seven of the items (form A, items 1, 10, 11, 12; form B, items 1, 7. 12) assume conditions which obtain chiefly in the subcontinent of Southern Africa. in which a white hegemony is dominant. If these items are omitted from the scale, however, there is no apparent reason why it should not give satisfactory results elsewhere, and in view of the high split-half reliability of the scale, the omission of these items should not seriously affect its important properties. REFERENCES 1. BARRON, F. Creativity and Personal Freedom. Princeton, N. J., Van Nostrand, BLOOM, L., DE CRESPIGNY, A. R. C. and J. E. SPENCE. An interdisciplinary study of social, moral, and political attitudes of white and non-white South African university students. J. soc. Psychol., 1961, 54, EDWARDS, A. L. Techniques of Attitude Scale Construction. New York, Appleton-Century Crofts, EDWARDS, A. L. and F. P. KILPATRICK. A technique for the construction of attitude scales. J. appl. Psychol., 1948, 32,

6 THE MEASUREMENT OF ArrrrunES TOWARD AFRICANS: A UNlDIMENSlONAL SCALE 37 WITH HIGH DISCRIMINATING POWER 5. FREDERIKSEN, N. and S. MESSICK. Response set as a measure of personality. Educ. psychol. Measur., 1959, 19, GUILFORD, J. P. The phi coefficient and chi square as indices of item validity. Psychometrika, 1941,6, 11-19, 7. GUTTMAN, L. A basis for scaling qualitative data. Amer. sociol. Rev., 1944, 9, LlKERT, R. A. A technique for the measurement of attitudes. Archs Psychol., 1932, MACCRONE, I. D. Race Attitudes in South Africa. London, Oxford University Press MANN, J. Inconsistent thinking about group and individual. J. soc. Psychol., 1967,71, PETTIGREW, T. F. Social distance attitudes of South African students. Soc. Forces. 1960, 38, STRICKLAND, J. H. and W. P. JANICKI. An alternative form of a forced-choice F scale. Psychol. Rep., 1965, 16, THURSTONE, L. L. Theory of attitude measurement. Psychol. Rev., 1929, 36, THURSTONE, L. L. The measurement of social attitudes. J. abnorm. soc. Psychol., 1931, 26, VAN DEN BERGHE, P. L. Race attitudes in Durban, South Africa. J. soc. Psychol., 1962, 57, W ANG, C. K. A. Suggested criteria for writing attitude statements. J. soc. Psychol., 1932, 3, Mr. A. M. Colman, Department of Psychology, University of Leicester, Leicester LE I 7RH, United Kingdom. Manuscript received: 26th February Revised manuscript received: 29th June C,T.P.LTO.JHB.