Attitude Tests

Multiple-Item Direct Attitude Measures

Formal attitude measurement began with the work of Louis L. Thurstone. Although he proposed several procedures, his most popular approach was the equally appearing interval scale. This procedure requires development of a series of opinion statements that represent varying levels of positivity or negativity toward the attitude object. Judges then sort these statements into equal-interval categories of favorability toward the attitude object (e.g., 11 intervals, where 1 = extremely unfavorable and 11 = extremely favorable). Next, a scale value is computed for each statement that corresponds to the median (or mean) score of judges' ratings, eliminating items with highly variable ratings. From the remaining items, the researcher selects a final set of statements representing equal intervals on the evaluative dimension. The final scale consists of these items in random order, with instructions for respondents to check the items with which they personally agree. The mean scale values of marked items are computed to obtain individual attitude scores.

Although Thurstone scales generally work well, the procedure is time consuming because it requires items to initially be rated by judges. In response to this and other concerns, in 1932 Rensis Likert developed the method of summated ratings. This procedure requires a set of opinion items that are clearly positive or negative in relation to the attitude object. The items are then administered to the sample of interest, whose members are instructed to indicate their level of agreement on a 5-point continuum (where strongly agree is assigned a value of 5 and strongly disagree is represented by 1). When the sample has completed the items, negative items are reverse coded, and each respondent's item scores are summed to create an overall attitude score. Item-total correlations are computed to identify poorly performing items (i.e., items with low item-total correlations). These items are discarded, and the final attitude scale scores are computed. Research has suggested this procedure tends to produce highly reliable scales.

In response to concerns that prior procedures did not guarantee unidimensional attitude tests, Louis Guttman proposed scalogram analysis. This approach involves constructing a set of opinion statements that are ranked in order, from least extreme to most extreme (i.e., a set of statements ranging from mildly positive to extremely positive or a set of items ranging from mildly negative to extremely negative). Scalogram analysis assumes that agreeing with a more extreme position implies agreement with less extreme positions. The pattern of responses to the set of items is examined to assess the extent to which items satisfy this assumption. Items are discarded if they frequently fail to meet this criterion. Although Guttman scales have the advantage of producing scales that are likely to be unidimensional, the difficulty of constructing scales that satisfy its stringent requirements has [Page 54]prevented it from becoming a widely used method of attitude measurement.

...