Likert, slider, or text? reassurances about response format effects

Abstract

Few experiments in the domain of addictive behaviors address how response format might impact self-reports of drug use, drug problems, or their correlates. In a preregistered study, participants (N = 1,546) reported on their drinking habits, problematic drinking behaviors, cigarette smoking, Big Five traits, and some additional health behaviors using either a Likert scale, slider, or free-response text entry format. Response format exerted little observable effect on mean values despite the large sample size. Exploratory analyses revealed that the magnitude of the association between frequency of drinking and problematic alcohol use was substantially and reliably greater when participants responded on a slider scale (vs. Likert or text entry), and that responses to items inquiring about longer-term estimates (e.g. number of days per year one exercises) were smaller when participants reported in the text-entry format. Likert-style measures appeared easiest to use and generated few problematic associations with other variables. Although variation in response format therefore does not appear to pose a large threat to the absolute value of many variables (with the possible exception of text entry deflating some types of responses), associations among variables may inflate when participants use sliders instead of Likert-type scales or text entry.