Clearly, this shows a substantial degree of invariance against changes of the twodimensional image features.Facial expression dynamics seems thus represented independently of the detailed shape features of the primate head and of the stimulus view.Yet, we found a clear and highly robust ownspecies advantage in terms of the accuracy of the tuning for expression dynamics: the tuning along the species axis of our motion style space was narrower for human than for monkey expressions.This remained true even for stimuli that eliminated speciesspecific features, such as ear motion, or which were carefully equilibrated in terms of the amount of lowlevel information.Both key results support our initial hypotheses: perception can exploit the similarity of the structure of dynamic expressions across different primate species for fast learning.At the same time, and consistent with a coevolution of the visual processing of dynamic facial expressions with their motor control, we found a largely independent encoding of facial expression dynamics from a basic facial shape in primate expressions.This observed independence has to be further confirmed in more <a href="https://www.targetmol.com/compound/Alkaline%20Phosphatase">searchAlkaline Phosphatase</a> extended experiments, including a bigger spectrum of facial <a href="https://pubmed.ncbi.nlm.nih.gov/24966474/">buyAlkaline Phosphatase</a> shapes and, probably, even faces from nonprimate species.In fact, the observation that humans observe facial expressions readily from comic characters, which do not even correspond to existing species, suggests that the observed invariance goes far beyond primate faces.However, further experiments including a much wider spectrum of facial shapes will be required to confirm this more general hypothesis.The observed independence of basic facial shape and expression encoding seems inline with results from functional imaging studies that suggest a modular representation of different aspects of faces, such as changeable and nonchangeable ones. At the same time, it is difficult to reconcile our experiments with several popular neural network models that represent facial expressions in terms of sequences of learned key shapes. Since the shape differences between human and monkey faces are much larger than the ones between the keyframes from the same expression, the observed spontaneous generalization of dynamic expressions to faces from a different primate species seems difficult to account for by such models.Concrete circuits for a shapeindependent encoding of expression dynamics still have to be discovered.One possibility is that they might exploit opticflow analysis, since the optic flow of the expressions on different head models might be similar.We have demonstrated elsewhere that a very robust recognition of dynamic facial expressions can be accomplished by a neural recognition model that is based on this encoding principle, where normreferenced encoding had been shown to account for the identity tuning of faceselective neurons in area IT. The presented novel technology for the generation of highly realistic dynamic face avatars of humans and monkeys enables electrophysiological studies that clarify the exact underlying neural mechanisms.A similar methodological approach was quite successful for discovering of the neural mechanisms of the identity of static faces. All participants had no prior experience with macaque monkeys and normal or tonormal corrected vision.Participants gave written informed consent and were reimbursed with EUR per hour for the experiment.The monkey head model was derived from a structural magnetic resonance scan of a rhesus monkey that imitates the deformations induced by the major face muscles of macaque monkeys.
