**A Study on the Use of Plant-Based Distractors in FFQs** By Kathryn Asher **Introduction** When used to assess diet change in response to even a modest dietary intervention, food frequency questionnaires (FFQs) may be vulnerable to an overestimation of adherence to intervention goals, termed response set bias. Two notable response set biases are social desirability and social approval. Social desirability is the tendency to respond in a normative way to avoid criticism, while social approval is the inclination to seek praise in testing situations. Though they have often been the preferred instrument for large-scale dietary assessment, the potential for bias using FFQs is thought to be greater than for instruments that cover shorter periods such as a 24-hour recall. While animal products (meat, dairy, eggs, etc.) may not have the social stigma that certain “junk foods” do, there is an increasing unease surrounding them, particularly concerning meat. Given this, an assessment of animal product consumption, even without the associated dietary intervention, may be subject to this type of bias. The social desirability bias that stems from diet self-report is magnified when connected to an intervention study. The degree to which respondents can ascertain which responses are considered desirable and undesirable affects the likelihood of bias, which is quite evident in a dietary intervention. **Study Purpose** One approach to limiting bias is the use of deception or partial disclosure, which is commonplace in social science research including studies on attitudes and behaviors toward food. One aim of this study was to determine whether adding non-animal foods (i.e., plant-based “distractors”) to an FFQ’s food list could effectively disguise the purpose of the instrument and in turn limit under-reporting of animal products in both a dietary intervention and non-intervention setting. **FFQ Instruments** A non-quantitative FFQ with a 13-item animal products food/beverage list was administered to some participants. It included the following foods/beverages with memory cues for each: cheese; other dairy products; dairy from non-cow sources; eggs (main dish); eggs (as ingredient); chicken (main dish); chicken (mixed dish); turkey; fish, not including seafood; seafood; pork; beef; and other meats. Other respondents received a longer 26-item FFQ with the same list of animal products mixed with an equal amount of non-animal foods/beverages. Respondents were asked to indicate how often they usually consumed the listed foods and beverages using a 9-point scale: never, less than 1 time per month, 1–3 times per month, 1-3 times per week, 4–6 times per week, 1 time per day, 2 times per say, 3 times per day, and 4 or more times per day. **Conditions** There were four treatment groups in this study. Half of the respondents were presented with a dietary intervention in the form of a short video about factory farming. The 13-item animal products FFQ was self-administered to half the respondents in the video group. The other half of respondents in this group received a longer 26-item FFQ with the same list of animal products mixed with an equal amount of non-animal foods/beverages. The non-video group was split in the same way, resulting in four unique groups: 1) the animal-based FFQ and a video dietary intervention (FFQANIMALVIDEO); 2) the plant and animal-based FFQ and a video dietary intervention (FFQMIXEDVIDEO); 3) the animal-based FFQ and no video dietary intervention (FFQANIMAL); and 4) the plant and animal-based FFQ and no video dietary intervention (FFQMIXED). **Dietary Intervention** A shortened version of the animated video “Here’s How We're Going to End Factory Farming” served as the dietary intervention. Exposure to the dietary intervention was verified using two methods. The video was hosted by Wistia—an online video hosting and analytics service—which provided data on the proportion of the video that respondents viewed. Individuals who watched less than 90% were said not to have received the full intervention and were removed from the study population. Viewers were also asked to explain what the video was about as well as what they thought about its message. Respondents who answered these in a way that suggested they did not give their attention to the video were also removed from the study. **Sample and Data Collection** The study was conducted by Kathryn Asher, PhD Candidate at the University of New Brunswick. The study was reviewed by the Research Ethics Board of the University of New Brunswick and respondents were provided with an informed consent agreement. Data was collected using an online survey between November 14 and 26, 2014. Survey Sampling International distributed invitations on behalf of the author to a selection of their panelists, aiming for four U.S. nationally representative samples of 200 individuals each aged 21 years and over. Representativeness was based on non-interlocking quotas for age, gender, education, and region. **Data Cleaning** Each of the samples were deliberately overpopulated to allow room for the removal of unreliable respondents and to address the discrepancy between the demographics collected for this study and those on file with Survey Sampling International. The initial sample sizes were: FFQANIMALVIDEO (n=428), FFQMIXEDVIDEO (n=368), FFQANIMAL (n=250), and FFQMIXED (n=248). Data cleaning was conducted where respondents were removed if they: 1) did not complete the full survey (all questions were mandatory); 2) did not meet the age requirement; 3) were without, or had a duplicate, respondent ID which served as their unique identifier in the panel; 4) did not watch a sufficient proportion of the video (for the FFQANIMALVIDEO and FFQMIXEDVIDEO groups); and 5) entered text that suggested their answers were unreliable to the open-ended questions that asked what the video was about and for their impressions of its message (for the FFQANIMALVIDEO and FFQMIXEDVIDEO groups) and what the survey was measuring (for all groups). Following this, the four samples were overpopulated by between 11 and 18 respondents each. A systematic procedure that aimed to closely match the target demographic quotas was used to reduce each sample to 200. **Data Analyses** The 9-point frequency scale was translated into a quantitative measure of daily frequency of consumption, which was assumed to represent daily servings consumed (i.e., if an individual consumes pork once per day they are assumed to have consumed one serving). This is a common procedure in the nutritional epidemiology literature. **Results** Preparation for a linear regression analysis with experimental group as the predictor showed that the model was not significant with total servings of animal products as the dependent variable (p = 0.583) nor with total servings of meat as the dependent variable (p = 0.555), indicating the type of experimental group is not associated with a change in reported consumption, whether of meat or all animal products. In a multiple linear regression analysis with total servings of animal products as the dependent variable, 7.9% of the variance (R Square) was accounted for by the 10 explanatory variables: the type of experimental group and the eight demographic variables (age, gender, U.S. region, education, race/ethnicity, religion, community size, household income, and political ideology). In this model, the type of experimental group was not associated with a change in total animal product consumption when controlling for the demographic variables. A second model was run with total servings of meat as the dependent variable. In this model, 8.6% of the variance (R Square) was accounted for by the explanatory variables. The type of experimental group was again not associated with a change in meat consumption when controlling for the demographic variables. Separate multiple linear regression analyses were run with each of the 13 types of individual animal product consumption as the dependent variable. The results show that in most cases the type of experimental group was not associated with a change in reported consumption when controlling for age and gender. The exceptions include (with FFQANIMAL as the reference category): reduced pork consumption for the FFQMIXEDVIDEO (β = -0.130; p < 0.05), and for the FFQMIXED group reduced consumption of fish (β = -0.114; p < 0.05), pork (β = -0.134; p < 0.05), and beef (β = -0.135; p < 0.05). These run counter to what one might assume in that the FFQs with plant-based distractors resulted in some instances of lower reported consumption than the purely animal-based food list. **Discussion** The study’s results suggest that adding plant-based distractors to an FFQ’s food list does not limit under-reporting of animal products nor total meat consumption in dietary intervention and non-intervention settings. **Limitations** There are a variety of limitations to the study. Of particular importance for these purposes is that a power analysis was not conducted, so it is possible a meaningful difference in consumption exists, however the study was underpowered to detect this. Another limitation is that there may have been notable differences in consumptipon but the FFQ was not responsive enough to detect this. *Note: the text provided in this write-up is borrowed from a draft of an article that will be submitted to a peer-reviewed journal for review. This study was funded by an Animal Welfare Trust Student Internship grant.*