Perception

A study of portrait paintings by various artists revealed that the sitter's face is more likely to be illuminated by a light source to the artist's left than to his right. Although the sitter's face may be turned towards the artist's left or right side with equal frequency, the interaction of illumination and profile orientation does not appear to be arbitrary. This finding is interpreted in the context of other findings on left and right in picture perception, and a new explanation of the relationship between profile orientation and the personality of the sitter is offered.

Recordings were made of the eye fixations of three subjects in two tasks involving black-and-white photographs of faces. In the first task, subjects matched a test face with a previously viewed target face; in the second task, subjects compared two simultaneously presented faces. The eye movements were recorded with a corneal reflection technique.

Each subject showed an individual fixation strategy for the tasks; in particular each subject had one or more preferred facial features which were viewed foveally in both tasks. The subjects also showed some tendency to use a regular sequential pattern of eye movements. However, the sequences used differed from one task to the other. Although some aspects of the results support the scanpath hypothesis of Noton, it is suggested that an alternative interpretation is possible.

Eye movements were recorded while subjects viewed ordinary portraits and photographic negatives of those portraits. Under both conditions they first studied sixteen portraits and then tried to decide which of forty-eight portraits they had just seen. They made more errors of recognition while viewing negatives, and their fixation patterns were significantly altered: there was a decrease in the percentage of fixations directed to the eyes, nose, and mouth, and an increase for such details as the ears, cheeks, chin, cap, and necktie. There was also a decrease in the ratio of fixations to the most fixated detail compared to the least fixated detail.

The experiment was designed to test the hypothesis that caricatures, relative to photographs, are ‘superfaithful’ carriers of information for facial recognition. Subjects were shown fifteen pictures of people's faces and were then asked to pick those same people out of a set of fifty-four pictures. There were three sets of pictures: caricatures, profile-view photographs, and three-quarter-view photographs. There were nine groups of subjects: for three groups the exposure and test stimuli were in the same medium, for six groups the test stimuli were in one of the media not previously seen. Points were scored for the number of people correctly identified and the number of false positives. Facial recognition within medium was very good, but was seriously disrupted by any medium shift, especially those involving caricatures. It is argued that the superfidelity of caricature may be manifest only when the task involves recognition of actual persons rather than their pictures.

The relevance of low and high spatial-frequency information for the recognition of photographs of faces has been investigated by testing recognition of faces that have been either low-pass (LP) or high-pass (HP) filtered in the spatial-frequency domain. The highest resolvable spatial frequency was set at 15 cycles per face width (cycles fw⁻¹). Recognition was much less accurate for images that contained only the low spatial frequencies (up to 5 cycles fw⁻¹) than for images that contained only spatial frequencies higher than 5 cycles fw⁻¹. For faces HP filtered above 8 cycles fw⁻¹, recognition was almost as accurate as for faces LP filtered below 8 cycles fw⁻¹, although the energy content of the latter greatly exceeded that of the former. These findings show that information conveyed by the higher spatial frequencies is not redundant. Rather, it is sufficient by itself to ensure recognition.

Human beings possess a remarkable ability to recognise familiar faces quickly and without apparent effort. In spite of this facility, the mechanisms of visual recognition remain tantalisingly obscure. An experiment is reported in which image processing equipment was used to displace slightly the features of a set of original facial images to form groups of modified images. Observers were then required to indicate whether they were being shown the “original” or a “modified” face, when shown one face at a time on a TV monitor screen. Memory reinforcement was provided by displaying the original face at another screen position, between presentations. The data show, inter alia, the very high significance of the vertical positioning of the mouth, followed by eyes, and then the nose, as well as high sensitivity to close-set eyes, coupled with marked insensitivity to wide-set eyes. Implications of the results for the use of recognition aids such as Identikit and Photofit are briefly discussed.

A new facial composites technique is demonstrated, in which photographs of the top and bottom halves of different familiar faces fuse to form unfamiliar faces when aligned with each other. The perception of a novel configuration in such composite stimuli is sufficiently convincing to interfere with identification of the constituent parts (experiment 1), but this effect disappears when stimuli are inverted (experiment 2). Difficulty in identifying the parts of upright composites is found even for stimuli made from parts of unfamiliar faces that have only ever been encountered as face fragments (experiment 3). An equivalent effect is found for composites made from internal and external facial features of well-known people (experiment 4). These findings demonstrate the importance of configurational information in face perception, and that configurations are only properly perceived in upright faces.

The encoding and relative importance of first-order (discrete) and second-order (configural) features in mental representations of unfamiliar faces have been investigated. Nonmetric multidimensional scaling (KYST) was carried out on similarity judgments of forty-one photographs of faces (homogeneous with respect to sex, race, facial expression, and, to a lesser extent, age). A large set of ratings, measurements, and ratios of measurements of the faces was regressed against the three-dimensional KYST solution in order to determine the first-order and second-order features used to judge similarity. Parameters characterizing both first-order and second-order features emerged as important determinants of facial similarity. First-order feature parameters characterizing the appearance of the eyes, eyebrows, and mouth, and second-order feature parameters characterizing the position of the eyes, spatial relations between the internal features, and chin shape correlated with the dimensions of the KYST solution. There was little difference in the extent to which first-order and second-order features were encoded. Two higher-level parameters, age and weight, were also used to judge similarity. The implications of these results for mental representations of faces are discussed.

The extent to which faces depicted as surfaces devoid of pigmentation and with minimal texture cues (‘head models’) could be matched with photographs (when unfamiliar) and identified (when familiar) was examined in three experiments. The head models were obtained by scanning the three-dimensional surface of the face with a laser, and by displaying the surface measured in this way by using standard computer-aided design techniques. Performance in all tasks was above chance but far from ceiling. Experiment 1 showed that matching of unfamiliar head models with photographs was affected by the resolution with which the surface was displayed, suggesting that subjects based their decisions, at least in part, on three-dimensional surface structure. Matching accuracy was also affected by other factors to do with the viewpoints shown in the head models and test photographs, and the type of lighting used to portray the head model. In experiment 2 further evidence for the importance of the nature of the illumination used was obtained, and it was found that the addition of a hairstyle (not that of the target face) did not facilitate matching. In experiment 3 identification of the head models by colleagues of the people shown was compared with identification of photographs where the hair was concealed and eyes were closed. Head models were identified less well than these photographs, suggesting that the difficulties in their recognition are not solely due to the lack of hair. Women's heads were disproportionately difficult to recognise from the head models. The results are discussed in terms of their implications for the use of such three-dimensional head models in forensic and surgical applications.

The importance of different perspective views for the recognition of model heads was studied. In experiment 1 subjects were instructed to learn the appearance of six heads placed individually on a turntable free to rotate through 360°. Subjects did not distribute their time evenly but focussed their inspection on particular views (the full face view and a view close to the profile). Despite differential inspection of these two views during the learning phase, the face, half profile, and profile views were recognized with equal efficiency in a subsequent recognition task with static views. Experiment 2 used the inspection paradigm to investigate view preference during the recognition of heads from memory. In this experiment subjects were asked to learn the appearance of three heads each seen rotating at an even speed. In a subsequent retrieval task the subjects actively inspected six model heads on the turntable and were asked to differentiate the three heads previously seen rotating from three novel heads. The pattern of inspection in this retrieval task was equivalent to that in experiment 1. Results suggest that during the encoding into memory subjects construct descriptions of specific prototypical views of the head and that descriptions of these same views are preferentially utilised during recognition.

When information about three-dimensional shape obtained from shading and shadows is ambiguous, the visual system favours an interpretation of surface geometry which is consistent with illumination from above. If pictures of top-lit faces are rotated the resulting stimulus is both figurally inverted and illuminated from below. In this study the question of whether the effects of figural inversion and lighting orientation on face recognition are independent or interactive is addressed. Although there was a clear inversion effect for faces illuminated from the front and above, the inversion effect was found to be reduced or eliminated for faces illuminated from below. A strong inversion effect for photographic negatives was also found but in this case the effect was not dependent on the direction of illumination. These findings are interpreted as evidence to suggest that lighting faces from below disrupts the formation of surface-based representations of facial shape.

A computer graphic method for extracting a natural image of an individual's facial prototype, or average appearance, from a number of different images of that individual is presented. The process improves upon previous photographic and computational techniques. Synthesis of a person's average expression and pose from a sample of images is derived in an automatic and quantitative way. Possible uses of composite faces produced in this manner in psychological investigations of facial qualities (eg attractiveness) and in applied areas such as telecommunication are pointed out.

The visual preferences of human infants for faces that varied in their attractiveness and in their symmetry about the midline were explored. The aim was to establish whether infants' visual preference for attractive faces may be mediated by the vertical symmetry of the face. Chimeric faces, made from photographs of attractive and unattractive female faces, were produced by computer graphics. Babies looked longer at normal and at chimeric attractive faces than at normal and at chimeric unattractive faces. There were no developmental differences between the younger and older infants: all preferred to look at the attractive faces. Infants as young as 4 months showed similarity with adults in the ‘aesthetic perception’ of attractiveness and this preference was not based on the vertical symmetry of the face.

Japanese male and female undergraduate students judged the gender of a variety of facial images. These images were combinations of the following facial parts: eyebrows, eyes, nose, mouth, and the face outline (cheek and chin). These parts were extracted from averaged facial images of Japanese males and females aged 18 and 19 years by means of the Facial Image Processing System. The results suggested that, in identifying gender, subjects performed identification on the basis of the eyebrows and the face outline, and both males and females were more likely to identify the faces as those of their own gender. The results are discussed in relation to previous studies, with particular attention paid to the matter of race differences.

Two experiments were conducted to investigate the role played by dynamic information in identifying facial expressions of emotion. Dynamic expression sequences were created by generating and displaying morph sequences which changed the face from neutral to a peak expression in different numbers of intervening intermediate stages, to create fast (6 frames), medium (26 frames), and slow (101 frames) sequences. In experiment 1, participants were asked to describe what the person shown in each sequence was feeling. Sadness was more accurately identified when slow sequences were shown. Happiness, and to some extent surprise, was better from faster sequences, while anger was most accurately detected from the sequences of medium pace. In experiment 2 we used an intensity-rating task and static images as well as dynamic ones to examine whether effects were due to total time of the displays or to the speed of sequence. Accuracies of expression judgments were derived from the rated intensities and the results were similar to those of experiment 1 for angry and sad expressions (surprised and happy were close to ceiling). Moreover, the effect of display time was found only for dynamic expressions and not for static ones, suggesting that it was speed, not time, which was responsible for these effects. These results suggest that representations of basic expressions of emotion encode information about dynamic as well as static properties.