This article explores the agreement between the predictions of modified Newtonian dynamics (MOND) and the rotation curves and stellar velocity dispersion profiles measured by the DiskMass Survey (DMS). A bulge–disk decomposition was made for each of the thirty published galaxies, and a MOND Poisson solver was used to simultaneously compute, from the baryonic mass distributions, model rotation curves and vertical velocity dispersion profiles, which were compared to the measured values. The two main free parameters, the stellar disk's mass-to-light ratio (M/L) and its exponential scaleheight (hz), were estimated by Markov Chain Monte Carlo modelling. The average best-fitting K-band stellar mass-to-light ratio was M/L ≃ 0.55 ± 0.15. However, to match the DMS data, the vertical scaleheights would have to be in the range hz = 200–400 pc which is a factor of 2 lower than those derived from observations of edge-on galaxies with a similar scalelength. The reason is that modified gravity versions of MOND characteristically require a larger M/L to fit the rotation curve in the absence of dark matter and therefore predict a stronger vertical gravitational field than Newtonian models. It was found that changing the MOND acceleration parameter, the shape of the velocity dispersion ellipsoid, the adopted vertical distribution of stars, as well as the galaxy inclination, within any realistic range, all had little impact on these results.