ИСТИНА

Previous work has demonstrated that facial expression recognition relies on a holistic representation. It was proved also that holistic processing may be disrupted when faces are inverted. We investigated expression discrimination at upright and upside-down orientations to obtain a measure of the influence of orientation on holistic face processing. Stimuli were images of six basic emotional expressions of a male poser, and 24 composite expressions which combine Happiness, Anger and Fear in upper or lower face with other 5 expressions. The experiment consisted of two parts in which images of faces were upright (in the first part) or upside-down (in the second). Twenty one participants were tested. Faces were presented in pairs, each for 2000 ms. The participants were asked to rate similarity between face expressions on a 9-point scale. Full matrices were processed with PROXSCAL multidimensional scaling program. We retained a 6D solution for both upright face (Stress = 9.8%) and upside-down face data (Stress = 10.5%). For both data the arrangement of points (i.e. basic and composite expressions) in the “expression space” was very similar: D1 separated X-Happiness (smiling mouth) from other stimuli, with X-Anger (compressed mouth) at the other extreme; D3 opposed X-Disgust (closed mouth) and X-Fear, X-Surprise (open mouth); D2 separated Fear-X, Surprise-X (wide-opened eyes) from other expressions; D4 opposed Surprise-X from others; D5 separated Fear-X from others. The proposed method proved of low efficiency for assessing holistic processes.