The steady-state visual evoked potential (SSVEP) is prevalent in cognitive neuroscience research yet much remains to be learned about how it is generated, including the relative contributions from different visual areas and the temporal delays among them. To address these questions, we exploited the idiosyncratic cortical folding patterns of retinotopically mapped visual areas to generate predictable variations in SSVEP topography across the visual field. Specifically, we regressed empirical topography variations against forward-model predictions generated by the Benson retinotopy atlas for different visual areas (V1, V2 and V3). In experiment 1 (N=10), we evoked SSVEPs at two frequencies, 7.5 Hz and 18.75 Hz, using contrast flickering checkerboards at 16 locations across an annular region of the visual field. The pattern of topography variations matched V1 predictions, and models suggested that the V1 contribution was twice the strength of either V2 or V3. Models without V1 performed substantially worse. Surprisingly, a large shift in SSVEP phase emerged in the lower visual field for the low-frequency SSVEP, and follow-up modelling supported an interpretation centred on transmission delays across visual areas, with V2/V3 trailing V1 by approximately 20ms. In a second experiment (N=12), we have extended the test to include 16 frequencies ranging from 4-40 Hz to provide novel constraints to estimate transmission delays between visual areas, a topic of longstanding contention. Initial results indicate that the phase shift was replicated, occurring at low frequencies and disappearing abruptly between 10-12 Hz. Modelling work to estimate the transmission delays is ongoing.