Sensorimotor information plays a fundamental role in cognition. However, datasets of ratings of sensorimotor experience have generally been restricted to several hundred words, leading to limited linguistic coverage and reduced statistical power for more complex analyses. Here, we present modality-specific and effector-specific norms for 39,954 concepts across six sensory modalities (touch, hearing, smell, taste, vision, and interoception) and five action effectors (mouth/throat, hand/arm, foot/leg, head excluding mouth, and torso), which were gathered from 4,557 participants who completed a total of 32,456 surveys using Amazon's Mechanical Turk platform. The dataset therefore represents one of the largest set of semantic norms currently available. We describe the data collection procedures, provide summary descriptives of the data set, demonstrate the utility of the norms in predicting lexical decision times and accuracy, as well as offering new insights and outlining avenues for future research. Our findings will be of interest to researchers in embodied cognition, cognitive semantics, sensorimotor processing, and the psychology of language generally. The scale of this dataset will also facilitate computational modelling and big data approaches to the analysis of language and conceptual representations.