Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Exp Brain Res (2006) DOI 10.1007/s00221-006-0455-1 RE SE AR CH AR TI C LE Iain D. Gilchrist · Henning Proske Anti-saccades away from faces: evidence for an influence of high-level visual processes on saccade programming Received: 10 August 2005 / Accepted: 8 March 2006 Springer-Verlag 2006 Abstract The anti-saccade task is an important tool for investigating both the generation of voluntary saccades and the suppression of involuntary, stimulus driven, saccades. In the anti-saccade task participants have to suppress an involuntary saccade to the stimulus in order to generate a voluntary saccade away from the stimulus. The extent to which errors occur in this task indicates the ability of the stimulus to trigger an orienting response that is beyond the control of the participant. Here we show an increase in anti-saccade error rates away from an upright face compared to a non-upright face. This suggests that complex high-level visual properties of the stimulus can inXuence whether involuntary orienting occurs. Introduction Saccadic eye movements move the high-resolution fovea to points of interest in the environment (e.g. Walls 1962). This happens for two, apparently distinct, reasons. First, saccadic eye movements can be generated in a voluntary manner—these are the saccades that are generated for example in exploration, search and reading. Second, saccades can be generated in an involuntary manner. In this case the eyes are drawn by an event or speciWc feature in the environment, for example a sudden movement or particularly salient visual object. To investigate when involuntary saccades are generated requires a speciWc set of experimental conditions. One task that can be used to investigate under what circumstances involuntary saccades are generated is the anti-saccade task (Hallett 1978). In the anti-saccade task, I. D. Gilchrist (&) · H. Proske Department of Experimental Psychology, University of Bristol, 12a Priory Road, Bristol, BS8 1TN, UK E-mail: [email protected] Tel.: +44-117-9299004 Fax: +44-117-9298450 participants are instructed to generate an eye movement in the opposite direction away from a visual onset when it appears. The anti-saccade paradigm pits the two types of saccades directly against each other. In a single trial, the involuntary programming of a saccade towards a stimulus is initiated by the onset and at the same time the participant is programming a voluntary movement in the opposite direction. Each trial represents a competition between the voluntary and involuntary orienting response. Typically, participants are able to control their saccadic responses and to execute a voluntary saccade in the correct direction on most trials (e.g. Hallett 1978). However, on a proportion of trials an error occurs and a saccade is directed towards the stimulus. In this case the onset has resulted in the generation of an orienting response that is performed against the participant’s intentions and therefore is best thought of as involuntary (see Everling and Fischer 1998; Munoz and Everling 2004). A number of studies have investigated what properties modulate the error rate in the anti-saccade task (see Munoz and Everling 2004). For example, error rates increase when the central Wxation point disappears a short interval before the onset of the target (Fischer and Weber 1997) and, when the target is more likely to appear on one side than another, errors are more likely to occur to that side (Koval et al. 2004). As far as we know there have been no studies of how the visual properties of the stimulus aVect the error rate in the anti-saccade task. Recent work from our group (Ludwig et al. 2005) using a diVerent saccade paradigm suggests that visual information within the Wrst 100 ms of Wxation determines where the eyes are directed and that the response is related to the contrast of the stimuli in a manner that closely mirrors the known response properties of cells in primary visual cortex. In the current experiment we were interested in investigating whether we could Wnd any evidence of an eVect of high-level object processes on saccade programming. The challenge for such a study is to exactly match the low-level visual properties of the stimuli in the experiment while at the same time changing the high-level visual properties. In pilot work we investigated anti-saccade performance using objects and faces as stimuli and compared them to scrambled objects and faces (cf., Kourtzi and Kanwisher 2000). We found a reliable diVerence in error rate between faces and scrambled faces, but no such diVerence between objects and scrambled objects. Having carried out these experiments, we were not convinced that these scrambling methods were useful in this context. SpeciWcally it is certainly not the case that the low-level characteristics of the stimuli remain unchanged by the scrambling process. Because of the properties of natural images, scrambling of this kind reduces the energy in the low spatial frequencies—the very kind of signal that may be most eVective at driving saccades (Ludwig et al. 2004). However, keeping the low-level properties of the stimulus constant while manipulating the higher-level properties can be achieved by using faces and inverted face stimuli as outlined below. There is extensive behavioural evidence that upright faces are processed in a qualitatively diVerent manner to inverted faces (see Valentine 1988). In addition neurons in the inferior temporal lobe (IT) not only respond selectively to faces but appear to be tuned to respond preferentially to upright faces (Perrett et al. 1988). This neurophysiological evidence is supported by fMRI evidence for reduced activation in IT to inverted faces compared to upright faces (Haxby et al. 1999). These data suggest that faces presented in an upright orientation are treated diVerently in high-level vision from non-upright faces. Importantly, because the same face stimuli can be used at diVerent orientations, lowlevel visual factors, such as brightness, energy at diVerent spatial frequencies and object size are identical across conditions. For this reason face stimuli uniquely allow us to look for an inXuence of high-level vision while matching the stimuli on low-level factors. In the current experiment we used an anti-saccade task to look for high-level visual eVects on the programming of involuntary saccades. SpeciWcally, we compared the rates of anti-saccades away from upright faces with the rates of anti-saccades away from non-upright faces (inverted, rotated 90° left, rotated 90° right). If saccade programming is only inXuenced by low-level factors, such as luminance and contrast, we would expect to Wnd no diVerence between these four conditions. However, if high-level factors can inXuence saccade programming then the upright face should lead to more errors. Method Design and procedure The trial sequence is illustrated in Fig. 1. Each trial commenced with the presentation of a Wxation point. The experimenter monitored the participant’s eye movements and initiated the peripheral presentation of the stimulus as soon as the eyes stabilised on the Wxation point. The task for the participant was to generate a saccade in the opposite direction away from the face and the instruction was to make the correct eye movement as quickly as possible. Participants completed three blocks of 96 trials. The stimuli were nine diVerent neutral expression faces taken from Matsumoto and Ekman (1988). All faces were presented in one of four orientations (0° [upright], 90°, 180° [inverted], and 270°). Face stimuli could appear at one of eight spatial locations equidistant around the centre of the screen and were presented at a centre-to-centre eccentricity of 8° of visual angle. Eye movements were monitored with an EyeLink II eye tracker (SR Research Ltd. Mississauga, ON, Canada). This infrared tracking system samples eye position at 250 Hz with a spatial accuracy of »0.3°. Saccades were detected using velocity and acceleration criteria of 30°/s and 8,000°/s2. The eye movement data were analysed oVline. Only the Wrst saccade after display onset was analysed. Trials were discarded if (1) the saccade latency was lower than 80 ms as these are assumed to reXect anticipatory movements (Wenban-Smith and Findlay 1991); (2) the starting point of the saccade was further than 2° away from the centre of the Wxation point and (3) the Wrst saccade-amplitude was less than 2°. These criteria lead to an exclusion of an average of 8.2% of trials per participant. Time Fixation display (1000ms) Response display, with example eye position data Participants There were 32 naïve participants (13 females, mean age 25.5 years). All were recruited from the University of Bristol and were paid for participation. The project was approved by the Ethical Committee of the Department of Experimental Psychology, University of Bristol. Fig. 1 The trial sequence over time in the experiment is shown. The second display shows the face stimuli overlaid with an example eye movement. Eye position was sampled at 250 Hz (white circles) As the instruction to the participants was to saccade away from the face we deWned an error in the task as a Wrst saccade towards the face (i.e. within §90° of the centre of the face). 30 800 29 700 28 600 27 500 26 400 25 300 24 200 23 100 Error rates The results are presented in Fig. 3. It is not clear why error rates in this experiment were somewhat higher than those previously reported. However, the displays used here where diVerent in a number of ways when compared to the stimuli used in a typical anti-saccade task (e.g. Hallett 1978; Koval et al. 2004). First, the face stimuli were large. Second, there were eight possible target locations rather than the more usual two. And third, the target and the Wxation point disappeared at the onset of the target. A one-way analysis of variance (ANOVA), revealed a reliable eVect of the orientation of the face on the percentage of errors—F(3, 93)=4.12, p<0.01. Participants made reliably more errors to upright faces (29.0%) than 22 0 90 180 270 Saccade Latency (ms) Figure 2 illustrates the spatial distribution of saccade landing positions across participants. Saccades successfully generated away from the target (downward movements in this Wgure) tend to have amplitudes consistent with the position of the target. However, saccades directed towards the target (upward movements in this Wgure) tend to be less spread and the majority fall within the area occupied by the face stimuli. It would appear that the error saccades in this task are genuinely target directed. Error rate (%) Results 0 Face Orientation Fig. 3 A graph of the results from the experiment. The grey bars plot the mean error rates (left hand y axis) with within-participant SEM error bars. The mean of the median saccadic reaction time for each participant is plotted for correctly directed anti-saccades responses (white circles) and incorrectly directed pro-saccades responses (black circles) to the faces in the other orientations (25.2%)—F(1, 31)=7.68, p<0.01. There was no reliable diVerence between error rates for the three non-upright orientations—F(2, 62)<1. We can also apply a stricter criterion that classiWes an error as a saccade that is directed directly at the face and a correct saccade as one directed in the opposite segment. The boundaries of these criteria are shown by the solid white lines in Fig. 2 and correspond to §38.7° towards and away from the face. With this deWnition a number of saccades that were previously classiWed as both errors and correct saccades are now excluded. However, there is still a reliable diVerence between upright faces (30.1%) and the faces in the other orientations (26.8%)—F(1, 31)=4.24, p<0.05. Fig. 2 A two dimensional frequency distribution plot of saccade landing positions in the experiment. The units on both axes are degrees of visual angle and each square in the plot represents the number of saccades that landed within that 0.5£0.5° area. The more saccades that landed in any given sector the brighter the square. The scale to the right of the plot indicates how the number of saccades maps onto the brightness of the square. The saccade landing position has been normalised so that upward saccades are towards the face and downward saccades are away from the face. The circle indicates the spatial extent and position of the face and the lines indicate one saccade classiWcation criteria discussed in the text Saccadic reaction times We also analysed saccadic reaction times for both the correct and incorrect saccades. The median saccadic reaction times were analysed using a 2£4 (correct or incorrect £ orientation: 0, 90, 180, 270) repeated measures ANOVA. There was a signiWcant main eVect of error—F(1, 29)=165.5, p<0.001. Error saccades were on average 82 ms faster than correct saccades. There was no main eVect of orientation (F(3, 87)<1) but there was an interaction between error and orientation—F(3, 87)=2.95, p<0.05. The diVerences between correct and incorrect saccades for each orientation condition were as follows 0°: 75 ms, 90°: 81 ms, 180°: 90 ms and 270°: 75 ms. The interaction in the ANOVA represents a reliable increase in the magnitude of these diVerences between the 180° and the 270° condition (t(29)=2.19; p<0.05). However, in a comparison between upright and non-upright faces this interaction was not signiWcant: F(1, 29)=2.07, ns (p=0.16). This provides strong evidence that the error rate diVerence between upright and non-upright faces was not a result of a speed-accuracy trade oV. Discussion Participants made more errors in the anti-saccade task when the stimulus was a face than when it was an inverted face. Upright faces generated a stronger involuntary saccadic orienting response. Controlling for other factors this in turn suggests that high-level visual information (in this case a face) is available to inXuence these involuntary saccadic responses. We chose to use face stimuli here for the methodological reasons detailed above. Because face processing is uniquely disrupted by changes in orientation (Valentine 1988), this allowed inverted faces to be used as a comparison stimulus. This comparison is useful because the overall low-level visual properties of the two stimuli are identical, but the high-level processing of these stimuli change. As a result we can conclude that the eVects reported are a result of these high-level visual properties. The current results do not rule out the possibility that other high-level meaningful object stimuli can also inXuence saccade programming in this way. However, it is not possible to test this using the current method as object processing, unlike face processing is not orientation sensitive in the same way (see Valentine 1988). In fact no other method except stimulus inversion will leave both the energy over spatial frequencies and the spatial structure intact. Saccade generation involves parietal and frontal cortices (see Schall and Hanes 1998) as well as sub-cortical structures. Frontal structures appear to be particularly important for successful performance in the anti-saccade task (see Munoz and Everling 2004). For example, in humans, error rates increase dramatically following damage to the frontal lobes (Guitton et al. 1985). In contrast, converging evidence suggests that neurons in the temporal lobe appear to be important for complex high-level visual processing, including the processing of faces (e.g. Perrett et al. 1988; Haxby et al. 1999). At a neural level, the current behavioural result suggests an interaction between the parietal–frontal system involved in spatial orienting and the temporal lobe, which is involved in complex high-level visual processes. This is a surprising result for theories in which these two systems are assumed to act in an independent manner (e.g. Milner and Goodale 1995). The current result suggests that the high-level visual properties of the target can inXuence the error rate in the anti-saccade task. One explanation for errors in the antisaccade task is the motor preparation hypothesis (Koval et al. 2004). Within this framework, errors are a result of an increased level of neural activity before the target appears, which results in the stimulus-related activity being more likely to reach threshold. The current result suggests that errors may not be solely motoric in origin but could also be visual. One possibility is that errors can result from an increase in the stimulus-related activity induced by the recognition of an upright face, which can also results in it being more likely that the threshold is reached. The increase in error rate found in the current study is relatively small. However, the size of the interference may not reXect, in a simple way, the strength of the interaction between high-level visual processes and saccadic orienting. To observe any interference in this task the signal that indicates the presence of an upright face has to arrive at ocular motor structures concurrently with the generation of the voluntary saccade. The latencies of visual responses in frontal areas are on average 40 ms faster than the visual responses in the inferior temporal lobe (Bullier 2001); as a result the correct saccade will often be generated before the inferior temporal area can signal the presence of a face. However, there is substantial variation in the speed of both of these signals from trial-to-trial. On some trials these two signals will therefore overlap in time and so interference can occur. Saccade latencies also vary considerably across diVerent tasks and so we might expect to observe more, or even less, interference under diVerent conditions. Whatever the exact neural mechanism, the current result provides evidence for an interaction between highlevel visual processing and involuntary saccadic orienting. Acknowledgements IDG is supported by the UK EPSRC and the Wellcome Trust. HP was supported by a Wellcome Trust vacation scholarship. We thank Casimir Ludwig for comments on a draft of this manuscript. References Bullier J (2001) Integrated model of visual processing. Brain Res Rev 36:96–107 Everling S, Fischer B (1998) The antisaccade: a review of basic research and clinical studies. Neuropsychologia 36:885–899 Fischer B, Weber H (1997) EVects of stimulus conditions on the performance of antisaccades in man. Exp Brain Res 116:191–200 Guitton D, Buchtel HA, Douglas RM (1985) Frontal-lobe lesions in man cause diYculties in suppressing reXexive glances and in generating goal-directed saccades. Exp Brain Res 58:455–472 Hallett PE (1978) Primary and secondary saccades to targets deWned by instructions. Vision Res 18:1279–1296 Haxby JV, Ungerleider LG, Clark VP, Schouten JL, HoVman EA, Martin A (1999) The eVect of face inversion on activity in human neural systems for face and object perception. Neuron 22:189–199 Kourtzi Z, Kanwisher N (2000) Cortical regions involved in perceiving object shape. J Neurosci 20:3310–3318 Koval MJ, Ford KA, Everling S (2004) EVect of stimulus probability on anti-saccade error rates. Exp Brain Res 159:268–272 Ludwig CJH, Gilchrist ID, McSorley E (2004) The inXuence of spatial frequency and contrast on saccade latencies. Vision Res 44:2597–2604 Ludwig CJH, Gilchrist ID, McSorley E, Baddeley RJ (2005) The temporal impulse response underlying saccadic decisions. J Neurosci 25:9907–9912 Matsumoto D, Ekman P (1988) Japanese and Caucasian facial expressions of emotion (JACFEE). Intercultural and Emotion Research Laboratory, Department of Psychology, San Francisco State University, San Francisco Milner AD, Goodale MA (1995) The visual brain in action. Oxford University Press, Oxford Munoz DP, Everling S (2004) Look away: the anti-saccade task and the voluntary control of eye movement. Nat Rev Neurosci 5:218–228 Perrett DI, Mistlin AJ, Chitty AJ, Smith PAJ, Potter DD, Broennimann R, Harries M (1988) Specialized face processing and hemispheric—asymmetry in man and monkey—evidence from single unit and reaction-time studies. Behav Brain Res 29:245–258 Schall JD, Hanes DP (1998) Neural mechanisms of selection and control of visually guided eye movements. Neural Netw 11:1241–1251 Valentine T (1988) Upside-down faces—a review of the eVect of inversion upon face recognition. Br J Psychol 79:471–491 Walls GL (1962) The evolutionary history of eye movements. Vision Res 2:69–80 Wenban-Smith MG, Findlay JM (1991) Express saccades—is there a separate population in humans. Exp Brain Res 87:218–222