Small‐magnitude earthquakes tend to be more numerous and broadly distributed than large‐magnitude events, so determining the directions of maximum and minimum stress indicated by small‐magnitude events can help describe regional stress patterns. However, because these events are often recorded by a small set of stations and have significant azimuthal and epicentral gaps, a comprehensive search usually produces a large set of acceptable fault‐plane solutions. Uncertainties in estimates of stress directions must therefore be computed and represented before meaningful comparisons can be performed. We present two techniques for representing earthquakes’ maximum and minimum compressive stress directions (i.e., P and T axes) as 2D probability density functions on a sphere. The first method is modified from an approach to contouring fracture directions by Kamb (1959). The second, which we call PaTaPs (“P‐ and T‐axis probabilities”), performs a chi‐square comparison between P and T axes and Monte Carlo samples of a Gaussian distribution and displays the resulting probability distributions via color shading on a sphere. PaTaPs and the Kamb method are applied to intermediate‐depth earthquakes recorded by a temporary broadband seismic network in the Dominican Republic during the period 2013–2017. We find that stresses at depths of 50–200 km are represented by an overall north–south maximum compressive stress that is more consistent with a model in which regional tectonics are dominated by a collision between Hispaniola and the Bahamas Bank. Stress orientations at these depths also differ from the generally east–northeast orientations of surface stresses as determined by Global Positioning System (GPS) measurements.