Investigating respiratory acts using motor imagery has the advantage that motion artifacts are much less likely to occur. To test whether motor imagery of voluntary coughing shows similar spatiotemporal activity patterns as compared to overt coughing, 12 participants underwent functional MRI scanning performing both tasks. We analyzed the data using a pattern classifier, that is, a support vector machine. Results showed that during imagined coughing, a number of brain areas reported previously to be involved in respiration showed more similarity in their spatiotemporal activity patterns with overt coughing than with a resting baseline. We conclude that motor imagery can be a suitable paradigm to investigate respiration, and that support vector machine analysis is potentially more sensitive and specific than a standard univariate analysis.