Face adaptation following a 3-second interval: persistence and transferability of blurring and sharpening filter information

To maintain the ability to recognize familiar faces despite ongoing changes over time (e.g., through aging), the human face memory continually updates in response to new visual input. This dynamic adjustment can be investigated experimentally using the face adaptation paradigm. In such experiments, participants are first exposed to highly manipulated faces during an adaptation phase. In a subsequent test phase, they are asked to identify a veridical face from a set of non-manipulated and slightly modified versions. Participants typically show a bias toward choosing faces that resemble the previously viewed manipulated stimuli (the adaptor), reflecting a shift in their perceptual norm, a phenomenon referred to as the Face Adaptation Effect (FAE; e.g., Carbon et al., 2007). In recent years, social media use has increased, reflecting a substantial shift toward digital communication and visual self-presentation. This widespread use of online platforms has also popularized the application of artificial facial filters, which modify visual appearance in systematic ways. Blurring filters smooth the skin and create a more uniform facial surface, whereas sharpening filters enhance contrast and emphasize fine details. FAEs have been reported for a wide range of facial features, and our previous experiments examined whether such commonly used artificial filters can similarly influence face perception. Using the face adaptation paradigm with a brief 300ms adaptation-test interval, we found evidence for adaptation to blurring: after exposure to strongly blurred faces, participants tended to classify those faces as “unchanged” that most closely resembled the blurred adaptor, while the veridical, non-manipulated versions were more likely to be perceived as altered. While previous studies have demonstrated both short-term and long-term FAEs for different types of face information (e.g., Carbon et al., 2007; Mueller et al., 2021a, 2021b; see Strobach & Carbon, 2013), it has remained an open question whether such temporal stability also extends to artificial filter information. In our earlier work, we examined whether FAEs induced by blurring and sharpening filters would persist over a five-minute interval - a duration at which higher-level memory components, rather than early sensory processes, are typically assumed to support adaptation. This experiment revealed an overall face adaptation effect; however, the structure of filter-specific and transferable adaptation at shorter delays remained unclear. The present study therefore examines whether such filter-related FAEs can still be observed with a substantially shorter adaptation-test interval of three seconds, allowing a more fine-grained assessment of their temporal dynamics. As in our previous 300ms and five minutes experiments, adaptation effects were measured across three transfer conditions: (1) pictorial (same image, same identity in adaptation and test phase), (2) structural (different image, same identity in adaptation and test phase), (3) cross-identity (different image, different identity in adaptation and test phase). The investigation of the transferability of face adaptation effects aims to clarify whether adaptation effects to blurring and sharpening filters remain confined to the pictorial level of image-specific representations or whether they extend to more abstract, face-specific representations, thus indicating control by high-level mechanisms of face perception rather than by general visual processes.