Psychological aspects of survey methodology : experiments on the response process
This thesis examines the psychological processes involved in responding to survey questions. Minor variations in questions have been shown to lead to variation in responses. These findings are inconsistent with the assumption that survey questions are tapping stable responses. Recently, psychological theories have been used to provide an explanation for these response effects. Research applying psychological theory to survey response is reviewed, covering research on both behavioural and attitudinal questions. These reviews illustrate a reconceptualisation of the basis of the survey response. The need for more detailed data on the response process is identified. Verbal reports are identified as a potential method for producing process data, yet, uncertainty over their validity is noted. The use of verbal reports as data is then reviewed, covering both their historical and more recent use. In the present research verbal report techniques are first experimentally examined to find an appropriate technique for obtaining process data in surveys. Think-aloud techniques are then used to examine the processes involved in responding to questions. A split-ballot questionnaire was administered, varying a number of questionnaire features where response effects have been hypothesised or shown to occur. Generally, the verbal protocols showed processing differences between the different question forms, and provided information about general types of cognitive processing during response. The next study moved on to look at context effects for attitudinal questions. An experiment was carried out in which a number of factors hypothesised to be influential in producing context effects were examined. A questionnaire was administered via computer and response latencies were measured. The results provide further information about the nature of context effects at attitude questions. The findings from this study are then discussed in terms of the methodologies used, the specific response effects addressed, and the survey response process generally.