A presumption underlying the semantic differential as a measure of attitude is that summated scores represent attitude intensity as well as direction. Several studies of semantic differential scales have pointed to the acceptability of this assumption. In two studies when independent intensity measures were plotted against bipolar scale scores the resulting curves indicated that the bipolar scales measured intensity. In a study reported by Osgood, Suci, and Tannenbaum the curve was U-shaped.¹ In a study reported by Mehling a V-Shaped curve was obtained.² In both these studies the point of lowest intensity was the mid-point of the seven-point scale. This finding led Mehling to conclude that the mid-point on bipolar scales "represents the neutral point in attitude."³

Sherif and Hovland, however, stress that some people may hold "extreme neutral" attitudes.⁴ Diab also suggests that neutral responses to semantic differential scales may have different meanings for different subjects.⁵

In a study using the semantic differential to measure attitudes toward a wide variety of concepts, the present writers employed independent intensity measures to test the intensity function of the semantic differential. The study involved two sets of ten concepts. Each set of concepts included five that were closely related to campus life and five that were of more general interest.⁶ The eighty-three subjects involved in the study were college students enrolled in a required speech fundamentals course at the Pennsylvania State University.

In the first part of the study the procedure developed by Mehling was employed.⁷ The subjects completed six bipolar scales⁸ for each of the ten concepts. After they had completed the semantic differentials for all ten concepts, they responded to an independent intensity measure for each scale on each concept. This measure was a ten-step continuum adjacent to each bipolar scale. For example:

The subjects were asked to indicate on this measure "how strongly" they felt about their responses on the bipolar scale.

In the second part of the study the subjects responded to a single-item independent intensity measure after completion of the semantic differential for each concept. This measure was a ten-step bipolar scale with one pole defined as "I am very certain of my attitude toward [concept]" and the other pole defined as "I am very uncertain what my attitude is toward [concept]."

When the semantic differential scale scores were plotted against the independent intensity measure across the first ten concepts, a U-shaped curve was obtained. The mid-point of the bipolar scales received the lowest average intensity rating. This would appear to confirm the conclusion of Mehling that the mid-point of bipolar scales represents the neutral point of attitude. Further analysis, however, suggests that this conclusion cannot be accepted.

When scale scores were plotted against average intensity scores for each of the six bipolar scales across the ten concepts, skewed U-shaped curves were obtained in every case. In no case was the mid-point of the bipolar scale the point of lowest intensity. The lowest intensity point was 3 or 5 (on a 7-point continuum) for each of these scales.

The plots based on average intensity ratings, however, obscure an important facet of the data. A large majority of the mid-point responses on each of the scales were accompanied by low-intensity ratings. Other responses to the mid-point were accompanied by extremely high-intensity responses. When these "extreme neutrals" were excluded, an approximately V-shaped curve was obtained in every case. This finding lends support to Sherif and Hovland's and Diab's suggestions that neutral responses may mean very different things to different respondents.

In short, the assumption that the mid-point on individual bipolar scales represents the neutral point of attitude was demonstrated to be unacceptable. These results clearly deny the Mehling conclusion, but does it follow that summated semantic differential scores do not adequately measure intensity of attitude? The second part of the study provides data relevant to this question.

When the summated semantic differential scores were plotted against the single-item independent intensity measure across the second ten concepts, a V-shaped curve was obtained. The point of the V was 24, the hypothetical neutral point of the semantic differentials. This would tend to indicate that the hypothetical neutral point of summated semantic differential scores represents the neutral point in attitude intensity. Subsequent analysis provides further confirmation for this conclusion. Plots of the summated semantic differential scores and the single-item independent intensity measure for each of the ten concepts individually produced approximately V-shaped curves in every case. On some of the concepts so few subjects responded on one side of the issue (only two subjects indicated a moderately unfavorable attitude toward "freedom of speech") that one side of the V was missing. However, the lowest intensity point was always the hypothetical neutral point of 24. Of the 830 responses, 65 were at the hypothetical neutral point of 24. Of these 65 responses, only 4 were accompanied by intensity ratings in the three steps of the intensity scale representing the highest intensity. Thus, it appears that although the mid-point of a single bipolar scale may not represent the neutral point in attitude intensity, the mid-point of the range of possible summated scores across six bipolar scales does represent the neutral point in attitude intensity.

A supplementary finding is noteworthy. A plot of the summated scores and average independent intensity scores for the second ten concepts, disregarding direction of attitude, indicated a linear intensity function. However, a plot of the single scale scores and independent intensity scores for the first ten concepts, disregarding direction of attitude, indicated a curvilinear intensity function. The increase in intensity between the "moderate" and "strong" responses was more than double the increase in intensity between the "somewhat" and "moderate" responses. This raises a question as to whether the usual 1-7 scoring procedure is the best method of scoring semantic differential responses. For purposes of comparison, a new scoring procedure was employed. Scores were assigned to the seven steps of the scale as follows: 1, 3, 4, 5, 6, 7, 9. When replotted using this scoring procedure, the single scale scores and average independent intensity scores for the first ten concepts yielded an approximately linear intensity function.

This supplementary finding suggested that the revised scoring procedure should be used in attitude-change studies employing the semantic differential. An attitude shift of one scale unit for extreme subjects represents more than twice as much attitude change as a one-unit shift for moderate subjects. Data were available from three experiments that employed the semantic differential as an attitude measure. The attitude scores were computed by both the 1 to 7 and the 1 to 9 procedures. Resulting F-ratios produced by analysis of variance were increased in every case by the 1 to 9 scoring procedure. Reliability of the measures, however, was not affected. This suggests that greater precision in measurement may be provided by the revised scoring procedure. However, further studies comparing the results of attitude change employing both procedures are needed before the new procedure can be recommended for general adoption.

The results of this study indicate that semantic differential summated scores across six scales reflect attitude intensity and that the hypothetical neutral point of a six-scale semantic differential is the neutral point in attitude intensity. However, the mid-point of individual bipolar scales does not appear to represent the neutral point in attitude intensity. This raises a question as to how many scales are needed to establish the hypothetical neutral point as the actual neutral point. It appears that six scales are adequate but that one is inadequate. Further research is needed to determine whether less than six scales are adequate. The results of this study also indicate that scoring semantic differentials with double weighing for the extreme positions may give a more accurate picture of subjects' attitudes. Further studies comparing this procedure with the usual equal weighing procedure are needed.

1. Charles E. Osgood, George J. Suci, and Percy H. Tannenbaum, The Measurement of Meaning, Urbana, Ill., University of Illinois Press, 1957, pp. 155-159.

2. Reuben Mehling, "A Simple Test for Measuring Intensity of Attitudes," Public Opinion Quarterly, Vol. 23, 1959, pp. 576-578.

3. Ibid., p. 578.

4. Muzafer Sherif and Carl I. Hovland, Social Judgment, New Haven, Conn., Yale University Press, 1961, p. 190.

5. Lutfy N. Diab, "Some Limitations of Existing Scales in the Measurement of Social Attitudes," Psychological Reports, Vol. 17, 1965, pp. 427-430.

6. The first ten concepts were dormitory life, apartment visitation rule, Penn State, cheating at college, jammies (informal student-sponsored dances), Ku Klux Klan, modern art, the Federal income tax, nuclear disarmament, and patriotism. The second ten concepts were Speech 200, draft deferments for college students, Daily Collegian, Speech 200 three-hour research requirement, teenage marriages, capital punishment, freedom of speech, Federal control of education, Viet Cong, and U.S. Supreme Court.

7. Mehling, op. cit., p. 577.

8. Scales used were beneficial-harmful, good-bad, right-wrong, fair-unfair, positive-negative, and wise-foolish.

Click Here To Go Back To PERIODICALS