We present a computational model of the mammalian olfactory bulb (OB) designed to investigate how cholinergic inputs modulate olfactory sensory representations. The model integrates experimental data derived from diverse physiological studies of cholinergic modulation of OB circuitry into a simulation of bulbar responses to realistic odorants. Experimentally-observed responses to a homologous series of odorants (unbranched aliphatic aldehydes) were simulated; realistic cholinergic inputs to the OB model served to increase the discriminability of the bulbar responses generated to very similar odorants. This simulation predicted, correctly, that missing cholinergic inputs to the OB would result in greater generalization between similar aliphatic aldehydes. Based on the assumption that the overlap between the neural representations of two sensory stimuli is predictive of their perceptual similarity, we tested this prediction in a behavioral experiments with rats. We show that, indeed, rats with selective lesions of cholinergic neurons that project to the OB and cortex discriminate less well between aliphatic aldehydes with similar carbon chain lengths than do rats that received sham lesions.