Purchase this article with an account.
Anastasia Flevaris, Lynn Robertson, Shlomo Bentin; Methodological issues in using spatial filters in ERP studies of face processing. Journal of Vision 2007;7(9):616. doi: 10.1167/7.9.616.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Research exploring the influence of spatial filtering on the face-sensitive N170 ERP has yielded mixed results. Goffaux and colleagues (2003) found that the N170-effect (increased amplitude elicited by faces compared to non-face objects) was absent when low spatial frequencies were filtered out of the stimuli. Using MEG, however, Hsiao et al. (2005) found the opposite result: a decrease in the M170 when mid and high spatial frequencies were removed from the stimuli. One difference between these two studies is that the stimuli used by Goffaux et al. were hybrids of low- or high-pass filtered foreground images with conversely filtered background textures, whereas Hsiao et al. did not use such hybrids. Hybrid stimuli have the advantage of equating spectral content across all conditions. However, the addition of a reciprocally filtered texture may add noise that is not equal across filtering conditions. We compared the N170 effects elicited by hybrid vs. non-hybrid stimuli, and found asymmetrical influences of the background texture. Participants viewed face stimuli and non-face objects that were either high-pass ([[gt]]5 cycles/degree) or low-pass ([[lt]]0.9 cycles/degree) filtered, and performed an unrelated detection task. Half of the blocks contained hybrid stimuli and half of the blocks contained non-hybrid stimuli. Adding a high spatial frequency background to a low spatial frequency stimulus did not have much influence. In contrast, adding a low spatial frequency background to a high spatial frequency stimulus diminished the N170 component. For non-hybrid faces, the N170 was greater for high than low spatial frequency stimuli. This effect was diminished in hybrid stimuli, presumably due to the asymmetrical effect of the background textures. Thus, when designing such studies, it is important to consider not only the benefit of equating spectral content across conditions but the cost of introducing a new source of noise in the background texture.
This PDF is available to Subscribers Only