Selecting the Number of Response Categories for a Likert-type Scale

On theoretical grounds, one would expect there to be a positive correlation between reliability and the number of response categories used in a Likert-type scale. Many psychometrists recommend that at least 20 response categories be used. However, when respondents are presented with either too many or too few response categories, it is possible that respondent fatigue might occur with a corresponding drop-off in response rate and reliability. This hypothesis was tested by comparing the results of three simultaneous surveys done using comparable versions of the Multidimensional Health Locus of Control Questionnaire employing two, six, and fourteen response categories. These were mailed to independent samples of 300 randomly selected adults. There was no significant correlation between the number of response categories and the response rate. Significant differences in reliability were found. Reliability tended to be highest with the questionnaire having six response categories. These data support the practice of employing about seven response categories.