The use of a covariate for randomized response (RRT) research has been shown to reduce standard errors of sensitive trait proportion estimates. At the same time, the model has been shown to be subject to serious misspecification when the relationship between the covariate and the sensitive trait is non-monotonic. The RRT covariate model is adapted to the non-monotonic case, and the efficacy of this approach is demonstrated through the analysis of a set of data from a business school study of bluffing behavior by business executives. Approximately 500 executives from a broad range of industries participated in a study of their tendencies to bluff when announcing a price increase. While the study does not develop a general non-monotonic covariate RRT model, it shows the potential utility of alternatives to the standard covariate approach and the reasonableness of attempting a non-monotonic model through the demonstrated capacity of the merged logistic function to estimate the sensitive trait proportions more precisely. Three tables present analysis results. (Contains 5 references.) (SLD)