Face detection is an initial step of many social interactions involving a comparison between a visual input and a mental representation of faces, built from previous experience. Furthermore, whilst emotional state has been found to affect the way humans attend to faces, little research has explored the effects of emotions on the
mental representation of faces. In four studies and a computational model, we investigated how emotions affect mental representations of faces and how facial representations could be used to transmit and communicate people’s emotional states. To this end, we used an adapted reverse correlation techniquesuggested by Gill et al. (2019) which was based on an earlier idea of the ‘Superstitious Approach’ (Gosselin & Schyns, 2003). In Experiment 1 we measured how naturally occurring anxiety and depression, caused by external factors, affected people’s mental representations of faces. In two sessions, on separate days, participants (coders) were presented with ‘colourful’ visual noise stimuli and asked to detect faces, which they were told were present. Based on the noise fragments that were identified by the coders as a face, we reconstructed the pictorial mental representation utilised by each participant in the identification process. Across coders, we found significant correlations between changes in the size of the mental representation of faces and changes in their level of depression. Our findings provide a preliminary insight about the way emotions affect appearance expectation of faces. To further understand whether the facial expressions of participants’ mental representations can reflect their emotional state, we conducted a validation study (Experiment 2) with a group of naïve participants (verifiers) who were asked to classify the reconstructed mental representations of
faces by emotion. Thus, we assessed whether the mental representations communicate coders’ emotional states to others. The analysis showed no significant correlation between coders’ emotional states, depicted in their mental representation of faces and verifiers’ evaluation scores. In Experiment 3, we investigated how different induced moods, negative and positive, affected mental representation of faces. Coders underwent two different mood induction conditions during two separate sessions. They were presented with the same ‘colourful’ noise stimuli used in Experiment 1 and asked to detect faces. We were able to reconstruct pictorial mental representations of faces based on the identified fragments. The analysis showed a significant negative correlation between changes in coders’ mood along the dimension of arousal and changes in size of their mental representation of faces. Similar to Experiment 2, we conducted a validation study (Experiment 4) to investigate if coders’ mood could have been communicated to others through their mental representations of faces. Similarly, to Experiment 2, we found no correlation between coders’ mood, depicted in their mental representations of faces and verifiers’ evaluation of the intensity of transmitted emotional expression. Lastly, we tested a preliminary computational model (Experiment 5) to classify and predict coders’ emotional states based on their reconstructed mental representations of faces. In spite of the small number of training examples and the high dimensionality of the input, the model was successful just above chance level. Future studies should look at the possibility of improving the computational model by using a larger training set and testing
other classifiers. Overall, the present work confirmed the presence of facial templates used during face detection. It provides an adapted version of a reverse correlation technique that can be used to access mental representation of faces, with a significant reduction in number of trials. Lastly, it provides evidence on how
emotions can influence and affect the size of mental representations of faces.
What you see is what you feel: Top-down emotional effects in face detection
Lombardi, F. (Author). 3 Jul 2024
Student thesis: Doctoral Thesis