Final Report Summary - SOCIAL_PREDICT (Predictive coding in social perception: a social neuroscientific approach to study the dynamic social brain.)
Our perception, interpretation and evaluation of our environment is, to some extent, always influenced by our social knowledge. A number of scientific studies indeed show that expectations about race, class and gender can, for example, alter how children interpret ambiguously aggressive acts, how doctors evaluate their patients, or whether a policeman decides to shoot a person carrying an object that could be a gun or a tool. Findings such as these suggest that social expectations directly change how we see people and the actions and objects related to them. However, responses to social stimuli are based on an interplay of different processes: initial sensory perception (seeing, hearing, feeling) of the stimulus, interpretation and evaluation of that sensory input, and response preparation and execution. For example, when anger on a black face looks more angry than the same expression on a white face, this could have different reasons. First this could mean that the initial basic (conscious) perception of the expression is directly changed, such that people ‘hallucinate’ more anger in the context of a black face. However, it could just as well mean that the percept is unaltered (the initial neural representation of the expression is the same, independent of whether it is presented on a black or white face), but that the interpretation or evaluation of that expression and/or response to that expression differs depending on the race of the observed person. Therefore, studies that show differential responses (in the form of reaction-times, evaluations etc.) depending on social expectations could be taken to suggest that basic sensory perception is shaped by social predictions. These results could also depend on post-perceptual attention, interpretation, evaluation or response bias. Here we present a electrophysiological study that shows that social contextual expectations directly change perception.
The aim of this project was to test whether expectations about people’s emotional state can make someone see that emotion in their face. We established a set of sentences that generate specific expectations about a person’s emotional state. The expected emotion is either positive (‘Someone just gave her a compliment.’) or negative (‘Her boss just accused her of stealing.’). Subsequently, we established that the expectation of a positive emotion actually induced participants to rate a subsequent hard-to-see face with a neutral expression as having a positive (happy) expression and the expectation of a negative emotion induced participants to rate the neutral face as having a negative (angry) expression (see figure 1 for examples of the stimuli). However, as explained above, reported differences in experiences do not necessarily indicate actual differences at the perceptual stage. To test this, we ran an EEG experiment, measuring the electrical response of the brain in the earliest stages of face perception. In this experiment we combined the emotional sentences with neutral faces to induce the perception of expectation-based emotional expressions (angry and happy). In addition, we combined neutral sentences with actually happy and angry faces to induce the perception of perception-based emotional expressions (angry and happy).
The results (Figure 1) show that expectations about emotions lead to a similar pattern of activation in the EEG as the visual actual perception of that emotion. In the case of induced emotion – where the actual visual stimulation is identical in the positive and negative condition - negative stimuli evoked an increased N1, followed by a decreased N170. When participants saw faces that actually displayed positive and negative emotions, these early cognitive components show the same pattern. Statistical tests of activation in the N1 and N170 timewindows show that processing of actual emotion is identical to processing of induced emotion.
Combined with the finding that participants report seeing positive and negative emotions in the neutral faces that follow the emotional sentences, these findings strongly suggest that expectations about the emotional state that another person is in can alter which emotions we see in the face of that person. This finding is only the first step in the total analysis procedure, which will also include a Bayesian approach of the data (very useful to test for the absence of statistical differences) and Multivariate Pattern Analysis (highly useful to test for similarities between patterns of activation, such as the similarities between real and induced emotion). The broader implication of this research is that social perception can be a mixture of actual sensory input (such as someone else’s facial expression, or their actions and utterances) with the perceiver’s internal expectations (based, for example, on the current context, prejudice or personal memories) about that person.
The aim of this project was to test whether expectations about people’s emotional state can make someone see that emotion in their face. We established a set of sentences that generate specific expectations about a person’s emotional state. The expected emotion is either positive (‘Someone just gave her a compliment.’) or negative (‘Her boss just accused her of stealing.’). Subsequently, we established that the expectation of a positive emotion actually induced participants to rate a subsequent hard-to-see face with a neutral expression as having a positive (happy) expression and the expectation of a negative emotion induced participants to rate the neutral face as having a negative (angry) expression (see figure 1 for examples of the stimuli). However, as explained above, reported differences in experiences do not necessarily indicate actual differences at the perceptual stage. To test this, we ran an EEG experiment, measuring the electrical response of the brain in the earliest stages of face perception. In this experiment we combined the emotional sentences with neutral faces to induce the perception of expectation-based emotional expressions (angry and happy). In addition, we combined neutral sentences with actually happy and angry faces to induce the perception of perception-based emotional expressions (angry and happy).
The results (Figure 1) show that expectations about emotions lead to a similar pattern of activation in the EEG as the visual actual perception of that emotion. In the case of induced emotion – where the actual visual stimulation is identical in the positive and negative condition - negative stimuli evoked an increased N1, followed by a decreased N170. When participants saw faces that actually displayed positive and negative emotions, these early cognitive components show the same pattern. Statistical tests of activation in the N1 and N170 timewindows show that processing of actual emotion is identical to processing of induced emotion.
Combined with the finding that participants report seeing positive and negative emotions in the neutral faces that follow the emotional sentences, these findings strongly suggest that expectations about the emotional state that another person is in can alter which emotions we see in the face of that person. This finding is only the first step in the total analysis procedure, which will also include a Bayesian approach of the data (very useful to test for the absence of statistical differences) and Multivariate Pattern Analysis (highly useful to test for similarities between patterns of activation, such as the similarities between real and induced emotion). The broader implication of this research is that social perception can be a mixture of actual sensory input (such as someone else’s facial expression, or their actions and utterances) with the perceiver’s internal expectations (based, for example, on the current context, prejudice or personal memories) about that person.