The interaction between face context and congruency was also significant (F(l,7)=10.19, p<.05), with the mean comparisons analyses showing
significantly faster RTs (F(l,7)=31.39, p<.001) for congruent (450 ms) than incongruent (496 ms) trials for the whole-face displays only (i.e. when the whole head was visible), suggesting that in the present study the congruency effect was due primarily to external features of the seen head (see Fig. 5.4). When the head was not visible, the difference between congruent and incongruent trials was not significant (468 ms vs. 477 ms, respectively, F(l,7)=1.18, p>.l). The other terms in the ANOVA were nonsignificant, with no influence of the nose manipulation on RTs.
N ose No N ose C ongruent W hole face J u s t ey e s Incongruent W hole fac e J u s t e y e s 3.71 3.32 8.60 6.25 2.93 3.12 7.62 3.91
T able. 5.3. Summary table o f means percentages o f errors for all conditions in Experiment 12.
The percentages of errors made by the subjects were also analysed. The overall percentage was only around 5% (see Table 5.3). The percentages of errors for each condition were entered into a three-way ANOVA with the same factors as before. This showed only a significant effect of the nose (F(l,7)=6.47, p<.05) due to a significant increase in errors when the nose was visible (5.5%) compared to when the nose was masked (4.4%).
Discussion
These findings were very intriguing, as the (positive) congruency effect for full-face stimuli in RTs goes in the opposite direction to previous data from Anstis et al (1969), and to my analysis o f response times (see earlier on in this chapter) of the experiments presented in Chapter 2 and 3. However, they provide a conceptual replication of the findings reported by Langton and Bruce (2000), in showing faster RTs forjudging gaze direction while ignoring head orientation when head and gaze were congruent. I suggest that a possible resolution to the apparent discrepancy between the positive congruency effect in this study and Langton and Bruce (2000), versus the reverse congruency effect in Anstis (1969) plus the previous studies in this thesis, may lie in the fact that no great time pressure was put on responses in Anstis et al. (1969) or the previous chapters in this thesis, while in both the present experiment and in Langton and Bruce’s study, highly speeded responses were required.
Therefore, it is possible that rather than being contradictory results, the different outcomes of these studies (i.e. positive versus reverse congruency effects) may be due to the change in methodology from speeded to unspeeded responses respectively. In the present reaction-time experiment, considerable emphasis was given to the speed of performance, while in the previous studies of Anstis et al. (1969), and the earlier chapter of this thesis, no great time pressure was exerted.
In addition, the fact that the possible congruency effect vanished when the eyes alone were present suggests that the information within the eye region
itself as emphasised by Anstis, was not responsible for the present positive congruency effect.
Experiment 13
In the follow-up experiment, only the speed instructions were changed, leaving everything else unchanged from Experiment 12 (e.g. same stimuli and task). The participants were now asked not to rush and to perform the task as naturally as possible. If the previous time pressure was crucial to explain the positive congruency effect in the preceding experiment, we should obtain a different pattern of results in the present experiment, possibly even a reverse congruency effect.
Method
Subjects: Twelve new volunteer subjects (8M and 4F), drawn from a
similar age range as in Experiment 12, participated in the experiment. As before they were unaware of the aim of the experiment.
Apparatus, Materials, Design and Procedure: These were exactly the
same as for the previous experiment, except that now no emphasis was given to the time taken to make a response, and so a longer interval (i.e. a maximum of 5578 ms from the onset) was allowed for response execution.
Results and discussion: The data were analysed in the same way as
before. Crucially, the observed congruency effect now tended to be in the reverse direction (i.e. slower response time for congruent than incongruent:
566 ms vs. 539 ms). This effect of congruency was not significant overall (F (l,l 1)=1.70, p>.l). However, there was now a significant interaction between nose and congruency (F (l,l 1)=8.44, p<.05). This was due to the reverse congruency effect being stronger when the nose was absent. There was also a significant three-way interaction between context, nose and congruency (F( 1,11 )= 11.48, p<.01 ); (see Table 5.4). N ose No N ose C ongruent W hole face J u s t e y e s Incongruent W hole fa c e J u s t e y e s 563.92 547.88 542.54 540.00 575.29 577.50 544.38 530.33
T ab le 5.4. Summary table o f means o f median RTs (m s) for all conditions in Experiment 13.
A subsequent ANOVA on the response time difference values between
congruent minus incongruent conditions showed that the factor of nose played a significant role (F (l,l 1)=8.89, p<.05) in the reverse congruency effect, and further that nose direction interacted significantly with the context in which it appears (F(l,ll)=10.32, p< .01).
100 -I (/> E 80 - Sc 2 60 - i T3 C g 40 - O Whole face ■ Just eyes O ) § 20 - «
I
0 - c g -20 L . nose no noseF ig. 5.5. Congruent minus incongruent difference values for inter-subject means o f median
RTs from Experiment 13. RT reverse congruency effects are plotted as a function o f nose and
context.
Specifically, the presence of a nose reduced the difference between congruent and incongruent conditions significantly more when just the geometry of the eyes was seen (F (l,l 1)=36.00, p<.001) compared to when the whole head was shown (F( 1,11 )=2.12, p>.l n.s.) (see Fig. 5.5).
N ose No N ose C ongruent W hole fac e J u s t e y e s Incongruent W hole fac e J u s t e y e s 2.73 2.22 3.65 1.17 1.82 2.74 3.52 2.08
The overall percentage of errors was about 2.5%, less than that reported in the previous experiment (presumably due to the reduced time pressure). The three-way ANOVA with the same factors as before did not showed any significant effect (all ps> .1); see Table 5.5.
In sum, the above results suggest that changing the speed instructions
drastically affect subjects’ performance as a function of head/eye congruency, thus implying that time pressure might influence the facial cues taken into account by the visual system to judge gaze direction.
G eneral Discussion
An effect of the congruency of head orientation with gaze direction was found on gaze judgements, which varied depending on time constraints given in the instructions to the subjects. Interestingly, when a speeded judgement was required (Experiment 12) for gaze direction, faster RTs were
found when head and gaze were pointing in the same direction (positive congruency effect, as in Langton and Bruce (2000)), but only with the full face visible. However, when no emphasis was given on speed (Experiment 13), a more complex pattern of results was found, including now the pattern of a reverse congruency effect (i.e. faster response for incongruent conditions) which was maximum when only the eyes were visible and progressively reduced when other facial cues became visible, for example, adding the nose.
There are two main conclusions which can be drawn from these experiments. Firstly, the reverse congruency effect (as in Anstis et al, 1969; plus my RT analysis of the experiments in the previous chapters) is apparently found only under conditions which do not require a highly speeded responses. Under considerable time pressure, the head angle of the whole face apparently becomes weighted more highly, so that gaze in the same direction as the head then becomes easiest to judge. Secondly, contrary to the classic claim that the geometry of the eyes alone is sufficient to explain gaze perception (Anstis et al., 1969), my data indicate a more complex pattern o f interaction between head-orientation cues and eye geometry. The nose and where it is pointing seem undoubtedly to help the human visual system to extract gaze direction from a tilted head. However, this influence is strongest when other head cues are not available, as shown by Fig 5.4 (c,d) upper panels. This is in accordance with Wilson et al.’s (2000) study, showing that nose orientation is the cue for head direction based on internal features when head orientation is otherwise hard to discriminate. Instead, when the whole head is visible, head cues other than the nose seem to take over (perhaps information from the profile or outer contours of the head). Moreover, people’s sensitivity to these different sources of information varies with task demands, in particular as regards the required speed of the judgement. Given sufficient time, the human visual system can evidently work out a more precise geometry by integrating gaze information with other sources of information about the angle of the head.
Interestingly, what emerges from the present study can be related to what has been uncovered by Perrett in his single cell recording work (1985, 1992,
1994). My findings provide some behavioural evidence that the mechanisms responsible for the processing of head and gaze may be organised
hierarchically, and that they may not be extracted completely independently, as contrary to Langton (Langton et al., 2000; Langton and Bruce, 2000). In fact, two different outcomes can be possible depending at least on two things: first, how fast the judgement of gaze direction has to be made; and second, which sources of information are currently available to the visual system. On the basis of my results one could argue that the visual system might give different weight to different facial cues according to the environmental demands (i.e. their visibility, or the required speed of the desired response). However contrary to what was claimed by Perrett, my results also suggest that the hierarchy emerging from the present study seems to weight the head more strongly than just the eye region when the visual system must rush to extract directional cues. This could be due to a difference in salience or speed of extraction due to size, with head being extracted faster than the eyes. The same neurones responding to the conjunction of head and eye (Perrett et al.
1985) might be “read out” earlier under time pressure.
Anstis’ claim that perceiving the direction of gaze does not depend directly on head cues has only been partially confirmed by my study, as the reverse congruency effect this predicts was found only in unspeeded conditions. Moreover, other cues than just the amount of visible sclera are evidently taken into account by the visual system even when enough time is allowed for the decision; for instance, nose orientation (Experiment 12). It is worth noting that in both Langton and Bruce’s (2000) and Anstis’s (1969) studies, facial cues
were visible within the stimulus. The present study showed that the effects previously described in literature (possible congruency, and reverse congruency effect) may depend exactly on which cues are available to the visual system, and the speed of the required response.
If we consider the time-course of processing information coming both from the head and from the eyes, one might expect that head orientation is extracted quicker than the eyes, thus leading to a stronger influence of head orientation over the eye direction at early stages of processing. With respect to that, evidence has been provided by a recent electrophysiological study carried out by McCarthy et al. (1999), aiming to assess the responsiveness of a face- related ERP component (N200) to particular perceptual features of faces. The authors found that the N200 amplitude was largest and shortest in latency when full faces were presented, and decreased and delayed when eyes, face contour, lips and noses were presented in isolation. Interestingly, the latencies to the latter facial parts were the longest, suggesting that such face parts may require additional processing time compared to full faces. Furthermore, the joint effects of eye and head position also affected the amplitude and latency
of N200 in the right hemisphere. In fact, when head and eye were both diverted in the same direction (i.e. congruent) they evoked the smallest and shorter N200, whereas when the head was directed at the viewer but the eyes were diverted, N200 was largest and longer. In other words, the authors found that N200 amplitude and latency was affected by different conjunctions of eye and head. Conflicting directional information from the head and the eyes may initially require more time to be processed and integrated. However according
to my findings (Experiment 13), at later stages of processing, eye direction may now become a more salient cue to infer the attentional status of an observer, compared to head orientation at an earlier stage. For example, the larger amount of sclera visible when eyes are turned away from the head could at a later stage totally or partially override head orientation, making the
judgement of gaze direction for incongruent conditions easier, as Anstis reported in his unspeeded study.
As well as being consistent with recent electrophysiolocial studies (e.g. Sugase et al., 1999; McCarthy et al., 1999), the present findings are the first
behavioural results which may resolve the previous apparent discrepancies concerning the combined effects of head and eye direction, with positive congruency effects arising in speeded situations (e.g. Langton & Bruce, 1999, 2000; plus the present Experiment 12); but the opposite reverse congruencv effect arising in unspeeded situations (e.g. Anstis et ah, 1969; plus Experiment 13), due to the relative processing speeds of head versus eye cues.