Questionnaire design, pretesting, and evaluation can be enhanced by the use of planned experiments that manipulate variables like questionphrasing and ordering, and that assess test-retest reliability. However, experimentation involving survey questions has been limited because of cost, burden, and time constraints in obtaining appropriate samples. This presentation will describe the use of Amazon’s Mechanical Turk (MTurk), an online labor market, to examine question order effects, test-retest reliability, and the potential utility of an online crowdsourcing platform for the conduct of survey experiments. We selected survey self-report items on perceptions of the environment related to walking, and on the frequency and duration of walking for transportation and leisure, from the 2015 Cancer Control Supplement of the National Health Interview Survey (NHIS). A total of 1,446 respondents were recruited through MTurk and completed the web survey using Qualtrics. At Time 1, respondents were randomly assigned to one of two conditions, in which the order of questions on walking behavior, versus perceptions of the walking environment, was varied. Four weeks later, respondents were invited to complete the survey in the same order as in Time1. We obtained a wide range of demographic characteristics (56.2% male, 84% Non-White Hispanic, 50% with at least a Bachelor’s degree, and mean age of 31.8 (SD = 10.6). Results demonstrated reasonable item test-retest reliability for the items administered (Kappa = .48-.84). Further, as in past studies of walking-related survey items, we observed a significant question order effect. The total cost for respondent incentives and MTurk administrative fees was $750. Based on success in carrying out the operational aspects of the experiment in a quick, low-cost manner, and in obtaining information useful for driving questionnaire-design decisions, we conclude that an online platform like MTurk is useful for embedding survey experiments.