Information Processing Capacity and Attitude Measurement

Green and Rao, using the criterion of information recovery and simulated data, concluded that sevenpoint scales were best [3]. Benson, pointing out that their criterion was not always applicable, suggested that wherever only averages are required, dichotomous or trichotomous scales suffice [1]. Jacoby and Matell reached similar conclusions using summary rating indices [9]. Lehmann and Hulbert's simulation concluded that a minimum of five categories was desirable unless there was averaging, either across individuals or scales [10]. Most research on the number of scale categories has focused on the desirable lower limit. At the upper limit, there are possible costs of respondent confusion and increased questionnaire length associated with increasing the number of categories [7]. Apart from this type of measurement error, however, Miller has found that over a wide range of unidimensional sensory variables, humans are incapable of discriminating among more than about 10 discrete categories. He concluded that this resulted from limited information processing capacity which appeared reasonably invariant over a number of different sensory attributes [11]. If this phenomenon were to apply to such common marketing measurement problems as rating stimuli on dimensions, then it would raise legitimate