Reference scale and placement of items with the own categories procedure
Lawrence LaFave and Muzafer Sherif
Department of Psychology, University of Windsor, Canada;
Pennsylvania State University
Various experiments have reported systematic variations in judgments of verbal statements on controversial social issues made with the method of equal-appearing intervals by subjects (Ss) with differing attitudes toward the item content (2, 3, 4, 7, 8, 11). Hovland and Sherif (4, 6) demonstrated that Ss with widely divergent stands concerning the social position of Negroes judged certain items on that issue differently, in particular the so-called "neutral" items (cf., 1). In a survey of the literature, Webb (11) concluded that under certain conditions the judgment of items under equal-appearing intervals procedure is affected by strong attitudes of the judges. More recently, Torgerson (9, p. 424) commented that variability of judgment is greatest in precisely those areas where judgment scales most frequently and routinely are used for measurement purposes: viz., the investigation of personality traits, attitudes, and preferences. Yet, he continued, comparatively little research in these areas specifies the functional relationships producing such variability.
The present analysis concerns certain formal characteristics of the categorizations by Ss who were publicly identified as representing differing stands on the issue: viz., the number of categories used; and the frequency of accepting, rejecting or remaining noncommittal on the definitions presented for judgment. Prior research results prompted Es to predict that under the stated conditions of judgment:
1. Ss upholding extreme positions on the issue will tend to use fewer categories than will the less ego-involved, unselected Ss (i.e., ego-involved Ss will tend "to see things in terms of black and white").
2. Ego-involved Ss upholding an extreme position will tend to reject more items than they accept, while Ss less involved in an issue will accept as many or more items than they reject. In other words, the relative sizes of the latitudes of acceptance and rejection will differ in the predicted directions for ego-involved and less involved Ss (i.e., ego-involved Ss will tend to "see" more "black" than "white").
3. Given the opportunity to remain noncommittal as to acceptance or rejection of items, unselected Ss will more frequently refrain from either acceptance or rejection of items than will ego-involved Ss with extreme stands.
4. The more ambiguous definitions will be displaced further as a function of the attitude of the S than will statements clearly defining a position on the issue.
A total of 317 Ss participated in the experiment-216 males, 89 females, and 12 who did not report their sex. These Ss were partitioned into three subsets: (a) 95 Negroes from a state Negro university in the Southwest; (b) 144 white Ss from various lower-division classes either in a state or private university in the Southwest; and (c) 78 male Ss from chapters of a southern fraternity at three universities (two state and one private) in the Southwest. Unselected Ss consisted of 105 males, 33 females, and six unreported. For Negro classes the figures were 33, 56, and six, respectively. The median age for all Ss was 20.0 years.
The Negro Ss attending an all Negro university in a southern city (which maintained a completely segregated public school system) represent the highly ego-involved subject classification in the study.
The white Ss from classes are heterogeneous and represent various shades of opinion on the issue. They are considered least ego-involved of the three subject classifications.
The fraternity chosen was one whose chapters are mainly located in the South and is known for its adherence to southern traditions. Therefore, members who took part in this experiment are considered more ego-involved than unselected white Ss, but less so than Negroes.
The 25 definition slips used in this experiment were selected from 35 employed in a pretest (5). Q values and medians were computed for these original 35 slips. This set of 25 slips was partitioned into five subsets (ordered along the
( 77) integration-segregation dimension) of five slips each. Which subset each of the 25 slips was assigned to was determined by the median value of that slip. The five subsets were Very Integrationist, Integrationist, Moderate, Segregationist, and Very Segregationist.
Sample slips from each of the five subsets read as follows:
SI, Very Integrationist: "The man who insists that discrimination against the Negro is contrary to every ideal which made this country great."
SII, Integrationist: "The person who wants integration of whites and Negroes to proceed at a reasonable pace."
SIII, Moderate: "A man who does not mind eating in the same restaurant as Negroes do as long as they do not sit at his table."
SIV, Segregationist: "The person who feels the present pace of integration of white and Negro students is too rapid for the good of our schools."
SV, Very Segregationist: "The individual who would prevent Negroes from voting."
First, Ss were asked to fill out an information sheet which called for such standard information as age, sex, school, residence, educational level, and so on. To insure anonymity, Ss were not required to write their names.
Next, Ss were asked to sort the definitions into piles. "These definitions," Ss were told, "all deal with the issue of integration-segregation in Negro-white relations." Before sorting, however, Ss were asked to become familiar with all the definitions by quickly reading them. Ss were then instructed to place all those definitions that "seem very integrationist on the issue of Negro-white race relations" into Pile 1. "Next to this pile," Ss were further informed, "place the definitions which seem next most integrationist, and so on, so that your last pile will consist of those definitions that represent very segregationist definitions in Negro-white relations." In brief, the "own category" procedure was used.
Having sorted all the slips into piles, Ss were then asked to label the piles of definitions they found "most acceptable" and "most objectionable." Ss were told that if any remaining piles seemed either acceptable or objectionable to label them thus. This procedure allowed Ss who used more than two piles to leave some piles unlabeled, if they chose. These unlabeled piles were designated "noncommittal."
Last, Ss were asked to indicate their own position on a graphic scale ranging from a "Very Integrationist" to a "Very Segregationist" position, and to write "a brief statement indicating what seems to you the correct position on the integration-segregation issue."
Since the present experiment concerns typical categorizations by Ss with differing ego-involvements in an issue, the analysis compares the frequencies with which predicted characteristics actually occurred for each of the three subsets of Ss.
If the first hypothesis be supported, unselected Ss should more frequently use a large number of categories in sorting the definitions than should either southern fraternity members or Negro Ss. To test it, Ss were dichotomized according to whether or not they used five or more categories. A trend is revealed in this direction. Whereas 36 per cent of the unselected Ss used five or more categories, only 23 per cent of southern fraternity Ss and 7 per cent of Negro Ss employed as many as five categories. In view of the large and significant X2 value for main effects, separate comparisons were made between unselected Ss and Negro Ss, unselected Ss and fraternity Ss, and fraternity Ss and Negro Ss. All X2 values arc significant: p < .025 for unselected and fraternity Ss and p < .005 or less for all other comparisons (one-tailed tests). Any evaluation of the strength of the trend must consider that the unselected Ss and fraternity Ss were by no means homogeneous in expressed stand on the issue. The Negro Ss were most homogeneous in stand and ego-involvement, and here the trend is most clear-cut. In fact, fully 68.6 per cent of Negro Ss used three or fewer categories, while only one Negro S used as many as six.
The second hypothesis concerns the relative sizes of the latitudes of acceptance and rejection. To test it, Ss were dichotomized according to whether the number of items they placed in the union of categories labeled "most acceptable" and "acceptable" equaled or exceeded the number of items placed in "objectionable" categories; or, if the number of acceptable items was smaller than the number of objectionable. The results reveal that the latitudes of acceptance of unselected Ss are more frequently equal to or larger than their latitudes of rejection. On the other hand, fraternity Ss, and especially Negro Ss, show the reverse tendency. Over 87 per cent of Negro Ss and 59 per cent of fraternity Ss placed fewer items in categories they indicated acceptable than in those they labeled objectionable. Comparisons between all pairs of subject groupings also were significant (p < .00I).
Hypothesis 3 compares Ss' placement of items into categories which they neither accept nor reject: i.e., noncommittal placements. The prediction is that less ego-involved, unselected Ss would more often have noncommittal placements than either Negro or southern fraternity Ss. To test it, Ss were dichotomized according to whether or not they employed any noncommittal place-
(79) -ments. While 41.7 per cent of the unselected Ss used one or more unlabeled categories, only 30.7 per cent of fraternity Ss and 29.5 per cent of Negro Ss employed such categories. Interestingly, relative frequencies of noncommittal placements for Negro Ss and fraternity Ss are similar, both differing from the frequency for unselected Ss. The trend is both consistent with expectations and significant (p < .05).
A test of the fourth hypothesis required determination of the median and Q values for each of the 25 items. Because the "own categories" procedure was employed, a conversion formula was needed to generate a common scale to compare the sortings of Ss using different numbers of categories. If p equals the pile in which the ith item is placed by Si, and k the number of piles or categories used by the 5th S, the scale value (v) of item i for the 5th S is given by vij = 100p/(k + 1), where the coefficient (I00) is used simply to eliminate decimals.
In this formula, p = 1 if the 5th S assigns item i to his "Very Integrationist" category; p = 2 if S places this item under his second most integrationist category, and so on. If S categorizes item i under his most segregationist pile, then p = k. For example, if Sj places item i in his second most integrationist of five piles, then vii = 200/6 = 33. The median for item i is simply its middle v, with integrationist items having low medians and segregationist high. To test Hypothesis 4, items were dichotomized in terms of their medians. With one exception, items with moderate medians (within the closed interval 40-60) had higher Q values than items with extreme medians. Thus, displacements of the moderate items (which were more ambiguous though not necessarily "neutral") were consistent with Hypothesis 4.
The trends in the foregoing analysis support the hypotheses statistically: the third at the .05 level, and the other three at .001 (on one-tailed tests). However, to evaluate the theoretic significance of the trends, more must be known about the three samples. The Negro Ss possess the most homogeneous background characteristics (including the stand on the issue). That these Negro Ss were highly ego-involved in a strong integrationist position is indicated by Es' observations of Ss' university, Ss' self ratings, written comments, and behavior in the judging situation. Strong support for the main hypotheses is given by Ss' categorizations which display, with few exceptions, the predicted characteristics. The trends for extremist white Ss are also in the predicted direction.
The manner in which the Negro Ss categorized reflects their intense ego-involvement in an extreme position. The significance of such categorizing is fur-
( 80) -ther indicated by a specific case in the pilot study: One week before a particular Negro S was to appear, Es were informed that his Negro peers called him "Uncle Tom." Es predicted that, unlike other Negro Ss, he would use both a large number of categories and be noncommittal on many items. This Negro S used eight categories (no other Negro S more than six).
The present experiment, in at least two important ways, apparently helps answer criticisms of the one by Sherif and Hovland. First, Sherif and Hovland did not employ an anti-Negro group of Ss selected a priori. Consequently, some critics might argue that certain Sherif and Hovland hypotheses were substantiated for the wrong reasons. For instance, why did the pro-Negro Negro Ss typically use fewer piles than did moderate white Ss? Perhaps because the former were more culturally deprived (less educated), instead of more ego-involved, than these white Ss. This "educational" hypothesis seems more difficult to maintain as the result of the La Fave and Sherif findings, since anti-Negro whites also used fewer categories than did white moderate Ss. (Of course both experiments attempted to control the educational variable by using undergraduate Ss in each group. Since one might contend that subtle educational differences between the groups exist, the La Fave and Sherif results seem reassuring to the ego-involvement hypothesis.)
A second criticism of the Sherif and Hovland experiment also seems counteracted by the present study. Upshaw (10, pp. 93 and 96) believes Sherif and Hovland obtained their predicted displacement effects because they employed out-of-range Ss relative to the range of scale values of the Hinckley items. Yet the present experiment also obtained displacements predicted and found by Sherif and Hovland. The Ss used in the present experiment could hardly have been "out-of-range," since items at both extremes were included, such as "A white person who would marry a Negro," and "The person who insists that the only good Negro is a dead Negro."
Neither unselected nor fraternity Ss were homogeneous with respect to either stands taken on or concerned with the issue. These were two of several difficulties in testing the hypotheses: i.e., neither all unselected Ss were moderates, nor were all southern fraternity Ss segregationists. The small number of items used-only 25-may represent a third condition that also weakened trends (both by reducing variability in number of categories employed and in distribution of items within these piles).
A fourth condition that probably attenuated the trend in the present experiment was that Ss were asked to sort the items along the integration-segregation dimension, whereas Sherif and Hovland requested Ss to sort on a pro-Negroanti-Negro scale. The Es in the present experiment assumed the integrationist
( 81) end to be pro-Negro and the segregationist end anti-Negro. But the correlation is obviously less than one-to-one. (For instance, Black Muslims are pro-Negro segregationists; some anti-Negro businessmen, in order to sell to Negroes, are willing to "integrate" with them; and so on.)
The findings of this experiment call to question certain established procedures in the investigation of attitudes and judgment of social issues. Because of these questions, the Es would suggest that adequate measures of attitude may be derived from characteristics of S's categorization of pertinent items on a dimension apparently unrelated to his attitude, made spontaneously by S, and without instructions for imposed categories.
The present experiment relates Ss' attitudes towards a controversial social issue to certain characteristic ways in which they place statements representing various positions on the issue. The underlying assumption is that S's attitude will reflect itself in his placement of items when the procedures permit him his own categories for judgment.
Consistent with earlier results, predictions were made as follows:
I. Highly ego-involved Ss will use fewer categories than Ss only moderately concerned with the issue.
2. The latitudes of rejection of highly ego-involved Ss will be greater than their latitudes of acceptance, whereas the latitudes of acceptance of less involved Ss will more frequently equal or exceed their latitudes of rejection.
3. Given the opportunity, more Ss only moderately concerned with the issue will remain noncommittal on some items than will more highly ego-involved Ss.
4. The more ambiguous definitions (i.e., items with intermediate scale values) will be subject to greater displacement (i.e., will have higher Q values) than will the more clearly defined end items.
Results support all four hypotheses, the third suggestively and the others strongly, particularly if the heterogeneity of stands represented in the white samples be considered. Problems related to attitude measurement procedures and social judgment studies, and the rationale for developing assessment techniques based on indices of S's categorizations were discussed.
1. EDWARDS, A. L. A critique of "neutral items" in attitude scales constructed by the method of equal-appearing intervals. Psychol. Rev., 1946, 53, 156-169.
2. FISHMAN, J. A., & LORGE, I. The role of the culture-group affiliation of the "judge" in Thurstone attitude-scale construction. Amer, Psychol., 1954, 9, 368-369 (Abstract).
3. GRANNEBERG, R. T. The influence of individual attitude and attitude-intelligence upon scale values of attitude items. Amer. Psychol., 1955, 10, 330-331 (Abstract).
4. HOVLAND, C. I., & SHERIF, M. Judgmental phenomena and scales of attitude measurement: Item displacement in Thurstone scales. J. Abn. & Soc. Psychol., 1952, 47, 822-833.
5. LA FAVE, L., & SHERIF, M. Categorization in judgment on a controversial social issue. Paper presented at the Southwestern Social Science Convention, Dallas, Texas, April 3, 1958.
6. SHERIF, M., & HOVLAND, C. I. Judgmental phenomena and scales of attitude measurement: Placement of items with individual choice of number of categories. J. Abn. & Soc. Psychol., 1953, 48, 135-141.
7. _____. Social Judgment: Assimilation and Contrast Effects in Communication and Attitude Change. New Haven, Conn.: Yale Univ. Press, 1961.
8. SHERIF, C. W., SHERIF, M., & NEBERGALL, R. E. Attitude and Attitude Change: The Social Judgment-Involvement Approach. Philadelphia: Saunders, 1965.
9. TORGERSON, W. S. Theory and Methods of Scaling. New York: Wiley, 1958.
10. UPSHAW, H. S. Own attitude as an anchor in equal-appearing intervals. J. Abn. & Soc. Psychol., 1962, 64(2), 85-96.
11. WEBB, S. C. Scaling of attitudes by the method of equal-appearing intervals: A review. J. Soc. Psychol., 1955, 42, 215-239.
Department of Psychology
University of Windsor
Windsor, Ontario, Canada
- Received in the Editorial Office, Provincetown, Massachusetts, on July 14, 1967. Copyright, 1968, by The Journal Press.
- Persons in various institutions helped in carrying out this study. The authors are especially indebted to Dr. Carolyn W. Sherif, Professor H. A. Bullock, Dr. Albert T. Milam, Dan Eddy, Jerry Jordan, William Maesen, La Vergne Gill, and Barbara Reid. The research was conducted as a unit of the research program of the institute of Group Relations, The University of Oklahoma, while Muzafer Sherif, currently Director, Psychosocial Studies Program, Penn. State U., was Director and Lawrence La Fave was on the research staff.