There is some disagreement on the underlying cause of the inversion effect. Some authors suggest that the inversion effect is due to impairments in the ability to extract 2^nd order configural information from an inverted face. Second-order configural cues refer to the relative size and spacing of features (Diamond and Carey 1986; Freire et al. 2000; Maurer et al. 2002). These claims are supported by studies that show quantitatively greater inversion effects when subjects are required to utilize the spacing of features compared to the identity of features when discriminating faces (Mondloch et al. 2003). An alternative account, however, suggests that the inversion effect results as a consequence of holistic processing for upright faces and thus can be used as a marker of holistic face processing (Rossion 2008). This holistic approach refers to a perceptual process whereby the features of a face, their relative size, placement, and spacing, are simultaneously integrated into a perceptual whole (Tanaka and Farah 1993). The inability to extract holistic information from an inverted face qualitatively shifts the strategy towards detecting spacing differences. Moreover, in several classic tasks of holistic processing, the effects are minimized or eliminated when the stimuli are inverted. The composite face effect, for example, is a task that combines the upper and lower face parts of two different individuals. The effect is that the individual in the upper face part is easier to identify when it is offset from the lower part, disengaging holistic processing, compared to when the face parts are aligned in a face-like configuration (Young et al., 1987). The misaligned advantage disappears when the faces are inverted because inverted faces do not engage holistic processing. Similarly, the Thatcher illusion, the grotesque appearance of faces after selective inversion of the eyes and the mouth in an upright face, disappears when the faces are inverted (Thompson, 1980). Finally, the other-race effect, poorer discrimination of faces from an unfamiliar race, is believed to result from the lack of holistic processing for other-race faces. Deficits in identifying own versus other-race faces disappears when the faces are inverted (Rhodes et al., 1989; Sangrigoli & de Schonen, 2004). While it is not the purpose of this paper to disentangling whether the inversion effect is a cause or consequence of holistic processing, these studies confirm the important relationship between these two face processing phenomena.

There is also considerable debate about whether the numerous face processing specializations observed in humans are shared by other species or whether they reflect true cognitive specializations. As is the case for the human literature, the inversion effect is by far the most widely tested phenomenon in comparative face recognition studies, but the results of these studies have been inconsistent and difficult to interpret for several reasons. The main source of this confusion is whether conspecifics’ faces share a special status in holistic processing, eliciting stronger inversion effects than heterospecific faces or nonface stimuli, in nonhuman primates that is comparable to human reports (Parr, 2011 for a detailed review of this literature). In brief, studies of inversion effects in chimpanzees report consistent findings. Using a matching-to-sample (MTS) task, Parr and colleagues (1998) found clear expertise effects in five chimpanzees. Subjects were significantly worse discriminating inverted compared to upright faces of expert categories only (chimpanzee and human faces), but no significant inversion effects were found for unfamiliar, heterospecific faces (capuchin monkeys) or nonface categories (automobiles or clip art). Using a similar MTS task format, Tomonaga (1999) found the inversion effect for unfamiliar human faces compared to houses in one chimpanzee. The same subject was significantly faster identifying upright human faces compared to upright chairs or hands in a visual search task (Tomonaga 2007). A follow-up study showed that chimpanzees were faster to identify the inner features of an upright face compared to inverted distracters (Tomonaga 2007).

In contrast to chimpanzees, data on inversion effects in monkeys are inconsistent and often fail to support a special status for conspecifics’ faces, or even face-selective effects more generally. Some studies have reported evidence of the face inversion effect in monkeys (Neiworth et al. 2007-cotton-top tamarins; Overman and Doty 1982-pigtail macaques; Tomonaga 1994-Japanese macaques; Vermeire and Hamilton 1998-rhesus monkeys), while others have failed to find evidence of inversion effects specific for faces (Bruce 1982; Dittrich 1990-longtail macaques; Gothard et al. 2004; Parr et al. 1999; Rosenfeld and van Hoesen 1979-rhesus monkeys; Weiss et al. 2001-cotton-top tamarins). When rhesus monkeys were tested using procedures comparable to chimpanzees, Parr and colleagues (1999) reported inversion effects for conspecifics’ faces, unfamiliar capuchin monkey faces, and automobiles, but not human faces (which could have been considered an expert category as the monkeys were born and raised in captivity with human caregivers), or clip art (Parr et al. 1999). Using an identical testing method in a separate group of rhesus monkeys, inversion effects were found for conspecifics’ faces, human faces, and unfamiliar chimpanzee faces, but not houses or clip art (Parr et al. 2008). Using eye-tracking as a dependent measure in rhesus monkeys, Gothard and colleagues (2009) demonstrated a reduction eye-fixations when subjects viewed inverted compared to upright conspecifics, but not human faces. These authors speculated that monkeys have different perceptual strategies for processing conspecifics’ faces compared to human faces. Using a similar methodology in rhesus monkeys, Dahl and colleagues (2009) also reported reduced fixations to the eye region of inverted conspecifics compared to human faces. In both of these studies, no control stimuli were used to assess the face-selectivity of the scan patterns, and the monkey subjects were all born and raised in captivity by humans, so the quality of the expertise manipulation was also unclear.

As a separate analysis of the inversion effect, we adjusted the difference scores between performance on upright and inverted trials based on the overall cost of inversion shown for the clip art trials. Changing the orientation of stimuli between the training and test trials is expected to produce a deficit in performance regardless of the identity of the stimulus categories. Our choice to adjust performance using the results from the clip art trials was based on previous data from numerous studies in our lab showing that clip art fail to reveal significant inversion costs. Therefore, general impairments on inverted clip art were used to compensate for the general costs of changing the quality (orientation) of stimuli between the training and test trials, and not differences in cognitive and/or perceptual processing strategies. However, it could also be argued that clip art does not reflect a stimulus category for which that are many within-category exemplars, like faces, and therefore may not be the most appropriate category. If, however, one were to utilize a different category, such as houses, the magnitude of inversion effects shown by the monkeys would be reduced, as houses revealed significant inversion effects. For example, for conspecific faces, the DI score adjusted for clip art was 10.78%, but when this was adjusted for houses, it dropped to 0.22%. The chimpanzees, however, showed slightly better performance discriminating inverted compared to upright houses, so the magnitude of their inversion effects would be increased if houses were chosen. For conspecific faces, the DI score adjusted for clip art was 16.39%, but when adjusted for houses this increased to 22.13%. These adjustments for the inversion effect in rhesus monkeys highlights the important finding; that the conspecific face lacks special status for the rhesus monkeys compared to the chimpanzee. Regardless of the approach taken, the results of these studies reveal significant species differences in the pattern of inversion effects and confirm the importance of holistic processing for conspecifics’ face in the chimpanzee, similar to humans.

Previous studies in monkeys have utilized the inversion effect as a marker for holistic processing and concluded that rhesus monkeys show face-selective holistic processing, just like humans (Dahl et al. 2009; Adachi et al. 2009). This requires that the inversion costs for conspecifics’ faces be compared to the results from both nonexpert face and nonface categories, but neither control was used in these studies. Without such comparisons, no conclusions can be made about the importance of holistic processing for conspecifics’ faces or its similarity to humans. Adachi and colleagues (2009) reported similar face processing strategies between humans and monkeys using the Thatcher Illusion (Bartlett and Searcy 1993; Thompson, 1980). This is a test of holistic processing as it shows that information about the specific features of a face, e.g., orientation of the eyes and mouth, is hard to extract when the features are embedded in the context of a whole face (Tanaka & Farah, 1993). Moreover, this effect is minimized when the faces are inverted, implicating its dependence on holistic processing. Using a viewing preference paradigm, monkeys looked longer at upright compared to inverted thatcherized faces, suggesting that they share similar face processing mechanisms as humans (Adachi et al. 2009). This conclusion is premature, however, because the study failed to utilize any control stimuli, e.g., thatcherized faces of another species, or other manipulated nonface categories, to confirm that the effect was indeed selective for conspecifics’ faces. Problematically, what is often concluded from these and similar studies is that if monkeys show the inversion effect then, like humans, they must be sensitive to 2^nd order configural cues. This is an important comparative issue as the ability to utilize 2^nd order configural cues in humans requires a protracted period of development (Mondloch et al., 2003) and is used to define adult-levels of face expertise (Crookes & McKone, 2009). Moreover, whether monkeys are sensitive to the spacing of facial features has only been directly tested a few times with different results (Dahl et al. 2007; Parr et al. 2008; Sugita, 2008). Therefore, considerably more testing is needed before any conclusions can be drawn about species similarities or differences in sensitivity to both holistic processing and 2^nd order cues in faces.

There are some discrepancies in our published data from chimpanzees that should be discussed. Parr and colleagues (1998) reported evidence of the inversion effect in chimpanzees for human faces, presumed to be an expert category, as subjects were raised in peer-groups by humans in the primate center nursery. However, a decade later, Parr and colleagues (2008) failed to demonstrate holistic processing for human faces in the same chimpanzee subjects using the composite face task, another task of holistic processing (Young et al., 1987). Although the exact cause of this discrepancy may never be known, it is very likely that a change in the chimpanzees’ expertise with human faces is responsible. Chimpanzees in both studies (4 of the same subjects were included in both studies) were all reared by humans in the Yerkes Primate Center nursery during the late 1980’s. After moving into the main colony at 4 years of age, the chimpanzees’ experience with humans was almost exclusively limited to caretakers and researchers wearing personal protective equipment (PPE), e.g., masks, face shields, hair covers, a procedure that was uncommon in the nursery. In the event that humans were seen without PPE, it was by regulation at a considerable distance from the subject and viewing distance is a critical factor in the ability to integrate facial features via holistic processing (McKone, 2009). Therefore, the earlier expertise shown by chimpanzees for human faces appears to have diminished as their visual experience with human faces (not wearing PPE) became more restricted. It is because of this uncertain degree of expertise with human faces that they were not used in the present study.