Purchase this article with an account.
Magdalena Bieniek, Guillaume Rousselet; The contribution of Fourier phase and amplitude spectra to image categorisation: An EEG study. Journal of Vision 2011;11(11):885. doi: 10.1167/11.11.885.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We quantified the relative contribution of Fourier phase and amplitude spectra to the time course of categorical object processing in humans. Although phase is the main contributor to the visibility of local edges, amplitude might carry global image categorical information; additionally, a special type of phase-amplitude interaction seems to exist, which when disturbed, hampers image categorisation [Gaspar & Rousselet, 2009, Vision Research, 49(24), 3001–3012]. We addressed this debate by recording ERPs from 8 normal subjects twice. Participants viewed 9 categories of images containing face, house or random phase information with the amplitude spectrum of a face, a house, or the average amplitude of faces and houses (a 3 × 3 design). Single-trial ANOVAs were performed at all time points and electrodes, with bootstrap spatial-temporal clustering multiple comparison correction. The results indicate major phase sensitivity around 170 ms and 250 ms post-stimulus in all eight subjects, reliably across sessions. Weak amplitude sensitivity occurred before 200 ms in three subjects, reliably across sessions. Additionally, in three subjects, phase and amplitude interacted in modulating ERPs around 240–250 ms, and around 330–350 ms. Both amplitude effects and phase × amplitude interactions had different spatial-temporal scalp distributions compared to phase effects. Our evidence supports the idea that phase information is the main contributor to early categorical ERP differences (<200 ms). Although amplitude spectrum alone seems to be insufficient to support image categorisation, it interacts with phase spectrum in later phases of visual processing. To investigate this interaction, we are currently undertaking a follow-up study using images of faces and houses with parametrically manipulated amount of phase and amplitude information (from 0 to 100% in steps of 10%). Subjects will be tested twice to assess reliability. This new design will allow us to describe precisely the time-course of phase and amplitude ERP sensitivity.
This PDF is available to Subscribers Only