Abstract:
Generalization gradients indicate the relative strength of a subject’s response to stimuli ordered on a continuum. A theory is proposed which envisages attitudes being measured in this fashion. Attitude stimuli ranging from those extremely favourable to an attitude object, through neutral stimuli, to those extremely unfavourable to an object, are ordered on a continuum. The subject is required to graduate his responses to these stimuli in terms of his strength of agreement with them. A generalization gradient may then be plotted with attitude stimuli on the abscissa, and strength of agreement on the ordinate.
Two major components of attitudes may be separately scored with this method. Extremity is scored as the attitude stimulus the subject agrees with the most. Intensity is scored as being proportional to the steepness of the generalization gradient.
Various response languages which enable the subject to graduate his strength of agreement are reviewed, and Magnitude Estimation finally selected. The reliability and validity of the resulting "Generalization Gradient" method is tested with the following conclusions: Reliability is adequate, the method is comparable in validity to other attitude measures in relating to behaviour, and the method is superior to other measures for recording change in both extremity and intensity components.
Results of secondary importance indicate lack of validity in some other measures of attitude intensity, the existence of a W curve relating extremity to intensity, and the effect of attitude intensity on the reliability of attitude scales.
It was concluded that intensity is a much neglected component in attitude research. This neglect may have arisen from inadequate conceptualization of the structure of an attitude. The Generalization Gradient method provides a research instrument which enables investigation of attitude intensity within the framework of a theory providing strong emphasis on this component.