Mount St. Helens has been closely monitored by a permanent network including 16 seismometers located within a 15 km radius, yet some of its diverse seismic signals remain challenging to detect and relate to deformation processes. Data from a dense array of 904 geophones within 15 km of Mount St. Helens provide novel constraints on weak seismic signals observed during a two-week deployment in 2014. Here we report on the identification of weak low-frequency (LF) events that were either not detected by the permanent network or were misclassified as likely rockfalls. We refer to the signals as LF earthquakes because they have less power above ~10 Hz and longer durations compared to similar magnitude volcano-tectonic (VT) events. Reverse-time imaging with the entire geophone array indicates that the LF hypocenters overlap with the frequent VT seismicity in a sub-vertical column beneath the summit crater at ≤6 km below sea level. Beamforming of P-wave arrivals with an exceptionally dense sub-array yields slowness estimates that confirm a subsurface origin for 16 LF events. Occurrence of VT and LF events in approximately the same volume suggests distinct source processes rather than path effects as the primary explanation for their different signals. Some of the LF and VT events have multiple spectral peaks that may be consistent with source mechanisms involving resonance in fluid-filled cavities. However, the majority of LF and VT event spectra exhibit smooth amplitude decay with increasing frequency, consistent with shear failure sources. The depletion of high frequencies for LF events could be caused by unusually low rupture velocities or stress drops. Our results indicate a variety of earthquake source processes occur in a sub-vertical column at ≤6 km depth, which is thought to overly the primary upper crustal magma reservoir. LF events are typically associated with eruptive or near-eruptive activity and are generally not reported at depths greater than ~1 km at Mount St. Helens, but based on our short-duration high-density experiment we suggest that upper crustal LF events may be more common than previously recognized and occur across a broader depth range.