Amphitheatre-headed canyons are common on Earth and Mars and researchers have long sought to draw inferences about canyon-forming processes from the morphology of canyon heads and associated knickpoints, often suggesting that amphitheatre heads indicate erosion by groundwater seepage erosion. However, the conditions and processes that lead to amphitheatre-headed canyon formation have been debated for many years. We consider two hypotheses that attribute the amphitheatre-headed canyon formation to fluvial erosion of strong-over-weak stratigraphy or, alternatively, groundwater spring discharge and seepage erosion. A spatial analysis of canyon-form distribution with respect to local stratigraphy along the Escalante River and on Tarantula Mesa, Utah indicates that canyon form is most closely related to variations in local sedimentary rock strata, rather than inferred groundwater spring intensity. Lateral facies variations that affect the continuity of strong layers can induce or disrupt the formation of amphitheatres. Furthermore, we find that amphitheatre retreat rate is dictated by the interaction of fluvial processes downstream of the amphitheatre headwalls and stratigraphy, rather than waterfall and groundwater processes that likely importantly influence headwall form. We conclude that fluvial erosion of strong-over-weak stratigraphic layering alone is sufficient to form amphitheatres at knickpoints and canyon heads. Thus, we reaffirm that formation process should not be inferred from canyon-head morphology, particularly where a strong-over-weak layering is known or plausible.