A proximate-level hypothesis was tested about decision making in honey bees, Apis mellifera. Fifteen bees were given binary choices between artificial flowers that varied in depth and in volume of sucrose-water delivered. Three bees violated weak stochastic transitivity, under an assumption about the order in which the flowers ranked along a utility scale; they preferred flower A to flower B, B to C, C to D, but D to A. These bees also violated strong stochastic transitivity, which focuses on the strength rather than the direction of preference. An intransitive pattern of choice supports the hypothesis of 'comparative' evaluation of flowers, where options are compared along their different dimensions, each dimension separately. Intransitivities cannot be explained by an 'absolute' model, where each option is evaluated independently and assigned a fixed value (utility). Twelve of the bees made choices that were consistent with both absolute and comparative models of decision making. In light of the above results it is suggested that the assumption that animals employ an absolute method of decision making be reconsidered, and perhaps supplemented with a comparative account. Possible evolutionary advantages of comparative evaluations are discussed.