Preference tests help to determine how highly individuals value different options to choose from. During preference testing, two or more options are presented simultaneously, and options are ranked based on the choices made. Presented options, however, influence each other, where the amount of influence increases with the number of options. Multiple binary choice tests can reduce this degree of influence, but conventional analysis methods do not reveal the relative strengths of preference, i.e., the preference difference between options. Here, we demonstrate that multiple binary comparisons can be used not only to rank but also to scale preferences among many options (i.e., their worth value). We analyzed human image preference data with known valence scores to develop and validate our approach to determine how known valence ranges (high vs. low) converge on a scaled representation of preference data. Our approach allowed us to assess the valence of ranked options in mice and rhesus macaques. By conducting simulations, we developed an approach to incorporate additional option choices into existing rank orders without the need to conduct binary choice tests with all original options, thus reducing the number of animal experiments needed. Two quality measures, consensus error and intransitivity ratio, allow for assessing the achieved confidence of the scaled ranking and better tailoring of measurements required to improve it further. The software is available as an R package (“simsalRbim”). Our approach optimizes preference testing, e.g., in welfare assessment, and allows us to efficiently and quantitatively assess the relative value of options presented to animals.
