Theoretical studies using seismic sources that radiate pulses of around 1-s duration or less show that for a double-couple source, in what are presumed to be realistic crustal structures, the most prominent arrivals on short-period (SP) P seismograms should be P, pP and sP. Some seismograms that show P and the surface reflections are observed and can be adequately modelled. However the number of such seismograms observed from shallow earthquakes is small—usually observed seismograms are either simple, that is they show a prominent first arrival and little else, or the seismograms are complex showing 30s or more of arrivals with amplitudes greater than or equal to the amplitude at the onset. The absence of clear surface reflections on both simple and complex seismograms is difficult to account for. It may be of course that the absence of such reflections is the result of scattering and defocussing of the P- and S-waves by lateral variations in the crust above the source. Such a hypothesis is difficult to test without very detailed crustal structures for earthquake zones. An alternative explanation of simple and complex SP seismograms, proposed here, is that many earthquakes may be due to unilateral fractures that propagate to increasing depth with time on steeply dipping fault planes. For such sources the corner frequencies of pP and sP will be below the corner frequencies of P (and S) radiated downwards. The consequence of this could be that the amplitude of pP and sP in the SP recording band is reduced relative to P. If this happens then on paths where P leaves the source near an antinode in its radiation pattern, simple seismograms would be observed. On paths where P leaves the source near a node in the radiation pattern, S-to-P conversions due to the incidence of downward-travelling S on minor discontinuities in wavespeed in the mantle could be of significant amplitude relative to P (and pP, sP) and produce complex seismograms. Examples are shown of observed complex and simple seismograms that can be explained on the unilateral fracture model.