Human dietary preferences vary highly between individuals yet also conform to common principles. However, the neural encoding of specific nutrient and oral-sensory food properties and related decision-making are poorly understood. This investigation of human food choice sought to elucidate these mechanisms by utilising behavioural and neuroimaging techniques while participants tasted, evaluated, and selected specifically designed liquid foods, where food information was derived exclusively from oral sampling. Healthy, human participants (N = 42) provided valuations (as willingness-to-pay (WTP) bids and ratings of food qualities) while sampling eight liquid food stimuli varying in nutritive (fat, sugar) and sensory (oral texture, taste) properties. During a subsequent binary choice task, participants chose their preferred option from two of these foods cued by conditioned visual stimuli. A subset of participants (N = 30) completed this choice task during functional magnetic resonance imaging (fMRI). Valuations and choices depended on nutrient content (fat, sugar) and were best explained by participants’ subjective perceptions of sensory food properties. Key oral-texture properties (viscosity, sliding friction) partially explained the influence of fat content on choices. Orbitofrontal cortex, amygdala, and ventromedial prefrontal cortex encoded subjective economic valuations of cued foods. Value of chosen foods was signalled in ventromedial prefrontal cortex activity at the time of choice. These findings demonstrate that human food preferences depend on subjective perceptions of specific nutritive and oral-sensory attributes that are integrated into subjective economic values in neural reward systems to guide choices.