Context. Solar flare hard X-rays (HXRs) are thought to be produced by nonthermal coronal electrons stopping in the chromosphere or remaining trapped in the corona. The collisional thick target model (CTTM) predicts that more energetic electrons penetrate to greater column depths along the flare loop. This requires that sources produced by harder power-law injection spectra should appear further down the legs or footpoints of a flareloop. Therefore, the frequently observed hardening of the injected power-law electron spectrum during flare onset should be concurrent with a descending hard X-ray source. Aims. We test this implication of the CTTM by comparing its predicted HXR source locations with those derived from observations of a solar flare which exhibits a nonthermally-dominated spectrum before the peak in HXRs, known as an early impulsive event.Methods. The HXR images and spectra of an early impulsive C-class flare were obtained using the Ramaty High-Energy Solar Spectroscopic Imager (RHESSI). Images were reconstructed to produce HXR source height evolutions for three energy bands. Spatially integrated spectral analysis was performed to isolate nonthermal emission and to determine the power-law index of the electron injection spectrum. The observed height-time evolutions were then fitted with CTTM-based simulated heights for each energy, using the electron spectral indices derived from the RHESSI spectra. Results. The flare emission was found to be dominantly nonthermal above ∼7 keV, with emission of thermal and nonthermal X-rays likely to be simultaneously observable below that energy. The density structure required for a good match between model and observed source heights agreed with previous studies of flare loop densities. Conclusions. The CTTM has been used to produce a descent of model HXR source heights that compares well with observations of this event. Based on this interpretation, downward motion of nonthermal sources should occur in any flare where there is spectral hardening in the electron distribution during a flare. However, this is often masked by thermal emission associated with flare plasma preheating. To date, flare models that predict transfer of energy from the corona to the chromosphere by means other than a flux of nonthermal electrons do not predict this observed source descent. Therefore, flares such as this will be key in explaining this elusive energy transfer process.