Previous studies have demonstrated that comprehension of conceptual metaphors elicits embodied representations. This finding is non-trivial, but begets the question whether alternative explanations are to be dismissed. The current paper shows how a statistical linguistic approach of word co-occurrences can also reliably predict metaphor comprehension. In two experiments participants saw word pairs with positive (e.g. happy) and negative (e.g. sad) connotations. The pairs were presented in either a vertical configuration (Experiment 1) or a horizontal configuration (Experiment 2). Results showed that response times could be explained by both the statistical linguistic approach and the embodied approach. However, embodied information was most salient in the vertical configuration and statistical linguistic information was most salient in the horizontal configuration. Individual differences modulated these findings, with female participants being most sensitive to the statistical linguistic approach, and male participants being most sensitive to the embodiment approach. These findings suggest that comprehension of conceptual metaphors can be explained by both linguistic and embodiment factors, but that their relative salience is modulated by cognitive task and individual differences.