CHANGING FACES: A Detection Advantage in the Flicker Paradigm Tony Ro, 1 Charlotte Russell, 2 and Nilli Lavie 2 1 Rice University and 2 University College London, London, United Kingdom Abstract—Observers seem surprisingly poor at detecting changes in images following a large transient or flicker. In this study, we com- pared this change blindness phenomenon between human faces and other common objects (e.g., clothes). We found that changes were detected far more rapidly and accurately in faces than in other ob- jects. This advantage for faces, however, was found only for upright faces in multiple-object arrays, and was completely eliminated when displays showed one photograph only or when the pictures were in- verted. These results suggest a special role for faces in competition for visual attention, and provide support for previous claims that human faces are processed differently than stimuli that may be of less bio- logical significance. Faces are meaningful stimuli, of great social and biological im- portance. Indeed, single-cell recordings (e.g., Perrett, Hietanen, Oram, & Benson, 1992), neuroimaging studies (e.g., Kanwisher, McDermott, & Chun, 1997), and neuropsychology (e.g., De Renzi, 1986) have pointed to brain areas selectively activated by faces. Such evidence has led to proposals that face processing may be subserved by a specialized module that results in automatic and mandatory process- ing of faces (Farah, 1995). However, despite this neural evidence that face processing may be special, there is little behavioral evidence for this claim. Although some studies have demonstrated unique sensitivity of face perception to configural information (Tanaka & Farah, 1993), other studies have shown that objects other than faces can acquire similar configural sensitivity after extensive training (Gauthier & Tarr, 1997). In addition, although facial expressions and gaze direc- tion seem to have a special capacity to attract attention (Driver et al., 1999; Friesen & Kingstone, 1998; Langton & Bruce, 1999; Mack & Rock, 1998), neutral faces typically require serial search, which im- plies that their perception is not automatic, but rather requires atten- tion (Brown, Huey, & Findlay, 1997; Kuehn & Jolicoeur, 1994; Nothdurft, 1993; Suzuki & Cavanagh, 1995). These visual search studies, however, have used fairly artificial experimental situations. For instance, most of these studies have used inverted or scrambled faces for nontarget stimuli, and schematic face drawings that do not convey all the information real faces convey have been used as targets. It remains possible, then, that faces would draw more attention than other objects if all stimuli in the array were photographic images of real objects. Moreover, even if face percep- tion is not automatic but rather depends on attention, faces may still have a special capacity to attract attention for further processing. This attention-capturing property of faces should become apparent, how- ever, only in situations in which real face stimuli are competing with other real objects for attention. This was the hypothesis of the present study. To investigate it, we employed a recent paradigm that has been used to investigate visual attention in more lifelike settings. Rensink, O’Regan, and Clark (1997) demonstrated that subjects are sometimes remarkably poor at detecting changes between two images of real-life scenes when the images are separated by a large transient, so that they appear to flicker. This phenomenon is termed change blindness (for reviews, see Simons, 2000; Simons & Levin, 1997). For example, subjects failed to detect that an airplane engine was removed between scenes. Attention seems to play an important role in change detection, as subjects are more likely to detect a change when attention is precued to the object of change. From our hypoth- esis about the special attention-capturing quality of faces, we reasoned that attention would be spontaneously cued to a face even in the absence of any instruction (cf. Levin & Simons, 1997). Thus, we predicted that any changes concerning faces would be noticed more readily than changes concerning other competing objects. We tested this prediction in two experiments. EXPERIMENT 1 Figure 1 depicts the sequence of events in each trial in Experiment 1. Photographs of six objects from six different categories (faces, food, clothes, musical instruments, appliances, and plants) were pre- sented on each display, with the center of each object placed 5° from fixation. Six instances (e.g., six different faces, six different appli- ances) from each category were used throughout the experiment. Apart from the faces, the objects selected for each category were chosen to be maximally different from one another in terms of their overall shape, (e.g., a rectangular toaster and a round fan were in- cluded in the appliance category). All the faces used, however, had a similar overall shape and were of the same sex (female). We selected the stimuli in this way to minimize low-level visual differences be- tween the changing faces and to maximize such differences for the changing objects in the other categories, so that any advantage for faces in change detection could not be attributed to the faces being less visually similar than the objects in the other categories. (We also addressed this issue directly in Experiment 2.) Each trial cycled through two displays (533 ms each) separated by a blank interval (83 ms) until a response was made or 20 s elapsed. On half of the trials, the displays were identical to one another (no-change condition). In the other half (change condition), between displays one of the objects changed to another object of the same category (e.g., a face changing to another face). Twelve subjects were asked to make a speeded forced-choice response to indicate the presence or absence of a change between displays. After this speeded response, subjects were asked to report which, if any, of the six categories had changed. A total of 144 trials was run after a short block of practice. Address correspondence to Tony Ro, Rice University, Department of Psy- chology - MS 25, 6100 Main St., Houston, TX 77005-1892; e-mail: tro@rice.edu. PSYCHOLOGICAL SCIENCE Research Report VOL. 12, NO. 1, JANUARY 2001 94 Copyright © 2001 American Psychological Society