Introduction to Chapter 3 The experiments in Chapter 2 were designed to investigate whether the presence of an upright face might capture and hence bias the early allocation of visual attention to its spatial proximity. In order to determine this, we tested participants on a task that relied heavily on visual spatial attention, in the peripheral presence/absence of an upright face. A significant performance decrement on the primary task, in the presence of an upright face would have supported the claim that faces captured visual-spatial attention (cf. de Gelder & Rouw, 2001). However, the consistent finding across all the four experiments showed that the orientation of a face-distractor had no impact on how well participants performed on the visual search task. In this chapter, a different visual attention paradigm is used to confirm whether an upright face’s inability to capture attention is isolated only to that used in visual search processes, or if it is a general rule. Furthermore, the experiment reported here addresses certain limitations of the general experimental design that the experiments in Chapter 2 were based upon. In the concluding section of Chapter 2, it was suggested that upright faces might be able to exert exogenous control in the absence of any endogenous control of attention. The attentional literature contains a dichotomy of exogenous and endogenous attention (e.g. {Corbetta, 2002 #85}. The former is said to be stimulus-driven and the latter, goaldirected. Having an goal-directed course of action such as a well-defined search-region or target object-category, could have had an over-riding effect over any exogenous control that upright faces might have exerted on attention (e.g. {Yantis, 1990 #115}.This could explain why upright faces failed to capture spatial attention and hence, disrupt the visual search process in the experiments reported in Chapter 2. If upright faces capture visual-spatial attention, it might only do so if doing so is not explicitly antagonistic to an internal goal-directed form of attention (such as with experiments 1, 2, 3, 4). Supposing this to be true, a task-paradigm is required that will not encourage the formation of any endogenous control of attention. Firstly, upright faces will have to appear within the search-space of the primary task, instead of outside it. In an experiment that also used a letter-detection task, face-learning were found for upright faces that served as a task-irrelevant background {Jenkins, 2002 #219}. Secondly, participants should not be primed towards the identification of any particular object category class, at the detriment of all others. As mentioned in Chapter 2, the active search strategy for alphabetic letters could have inadvertently suppressed the exogenous influence of upright faces on attention. The current hypothesis is that upright faces can act as effective exogenous cues, that orient spatial attention, but only in the absence of endogenous cues e.g. task-strategy. To test this, an experiment was carried out based on the dot-probe paradigm, a design that is commonly employed to demonstrate how early allocation of visual attention can be controlled by exogenous cues. Dot-probe paradigm The dot-probe method is usually used to assess preattentive and attentional biases to information embodied in visual stimuli, particularly in relation to threat and anxiety (e.g. {Broadbent, 1988 #229;Bradley, 1998 #227;Mogg, 1999 #225;Pishyar, 2004 #223}. A dot-probe trials simply requires participants to detect a probe that is initially hidden from view by two juxtaposed images. This format was adapted from paradigms used in experimental cognitive psychology that showed how spatial attention can be assessed from the speed of manual responses to visual probes (e.g. {Posner, 1980 #228}. Although the position of the dot-probe is not usually predictable by the nature of the image that hides it, certain images can result in faster probe-detection when they do share the same spatial location as the probe. For example, probes hidden behind threatening images were quicker to be detected by highly anxious individuals than those preceded by a neutral image under certain conditions {Mansell, 1999 #224;Bradley, 1999 #226;Bradley, 1998 #227}. Some researchers argue from such evidence that certain images can induce an attentional bias in the observer, that is based on its informational content {Pishyar, 2004 #223;Mogg, 1999 #225}. Existing dot-probe experiments with face-stimuli have mostly been interested in the emotional valence of face-stimuli, and how levels of anxiety in observers correlate with their exhibited levels of attentional bias exhibited towards threatening faces {Pishyar, 2004 #223;Mansell, 1999 #224;Mogg, 1999 #225}. Interestingly, this is an effect that is not consistently replicable with threatening non-face stimuli e.g words {Mansell, 2002 #230;Pishyar, 2004 #223}. Pishyar and colleagues (2004) recently concluded that the use of emotional face-stimuli in dot-probe tasks is a more sensitive index of attentional bias than emotional words. With dot-probe tasks, attentional-bias towards or away from a particular imagetype is always measured in relation to its accompanying control image. For example, there exists an attentional bias to threatening faces, but only in relation to a neutrally expressive face (e.g. {Pishyar, 2004 #223}. The question that will be addressed here is whether there exists an early attentional bias even to upright face-stimuli, when compared to inverted faces that differ only in terms of orientation. In the study reported here the element of emotional valence was removed from the face-stimuli and only faces with neutral expressions were used. Thus, according to de Gelder and Rouw (2001), upright faces which are said to capture spatial attention should Facilitatory and Inhibitory components of orienting The dot-probe task is essentially based on an experimental approach that has early precedence in the understanding of attentional capture; namely, the cuing-paradigm. Such studies explored the facilitatory and inhibitory influences that a brief non-informative cue could exert in the detection of a target that follows it (e.g. {Eriksen, 1973 #231;Posner, 1980 #228}. Thus, a brief non-informative visual cue presented in the periphery can speed up the detection response to a target that appears in the same location within 100msecs of the cue, compared to a target that might appear in a different location {Posner, 1980 #228}. The elegance of the cuing paradigm lied in its ability to prove that spatial shifts of attention can occur, even in the absence of observable motor behaviour e.g. eye-movements {Posner, 1980 #228;Posner, 1990 #110}. Effective orienting could be induced by either by peripheral and central cues. A peripheral cue oriented attention by drawing it towards itself, whereas a central cue appeared at fixation and directed attention symbolically e.g. arrow cues {Jonides, 1981 #232;Posner, 1980 #228}. Peripheral cues were said to exert exogenous shifts in attention whereas central cues appealed to the observers’ endogenous knowledge of its connotated meanings. The dot-probe paradigm which forms the basis of the study reported here, investigates exogenous rather than endogenous shifts of attention, since there is a simultaneous presentation of two images on both sides of fixation that act as competing peripheral cues. In such an experimental design, the greater efficacy of either image to act as a cue would indicate there there is something concerning the differences in visual characteristics across the two images, that allows one image to orient exogenous attention more effectively than the other. Such a paradigm is thus ideal for the purposes of the current study as it is the exogenous shifts in spatial attention that upright faces are claimed to induce {de Gelder, 2001 #33}. Furthermore, it allows for a direct comparison to be made between the upright face-image and a control image i.e. inverted-face, for the ability to capture attention. Unlike central cues, peripheral cues are said to require no interpretation or cognitive effort in order to capture attention. As a result, attentional shifts induced by peripheral cues are said to be reflexive and impossible to suppress compared to central cues, even if they are manipulated to be consistently incongruous to the probe-position {Jonides, 1981 #232}. Again, this mirrors the claim made by de Gelder and Rouw (2001), in that face-detection is a process that is fast and automatic. Usually, facilitation for the detection of a target is associated with the occurence of a peripheral cue that precedes it in the same location {Jonides, 1981 #232;Posner, 1980 #228}. As mentioned already, this is explained by imputing a spatial shift of attention to the appearance of a peripheral cue that goes on to make a positive contribution to the detection of items in the corresponding spatial region. However, there is a caveat to this rule. It has been found that facilitation for target detection by a valid cue is highly dependent on short cue-target onset asynchronies; that is, the time-interval between presentations of the cue and target. Specifically, facilitation only occurs with short cue-target onset asynchronies (~100msec). With longer cue-target onset asynchronies (~200msecs), a valid cue results in the inhibition of response for a targetdetection instead {Posner, 1984 #233;Klein, 2000 #234}. This late component of orienting has been termed ‘inhibition of return’, which suggests that early orienting of attention to a spatial location that does not yield a target is subsequently replaced with inhibition that serves to prevent attention from re-orienting back to a searched location {Posner, 1985 #235}. Understanding these two components of orienting – namely, the early component of faciliation and the late component of inhibition – allows for a further prediction to be made from using upright faces as cues in the dot-probe paradigm. If upright faces truly exhibit exogenous control over attention, performance in a dot-probe task ought to be facilitated or inhibited depending on the time-interval between the upright face cue and probe. Hence, if the spatial position of the upright face cue correspond to the probe’s, performance should not only be speeded up at short intervals, but it should also be slowed down for longer onset asynchronies between cue and probe. Experiment 5: Dot probe design In the final experiment, we investigate if upright faces are able to act as valid exogenous cues in visual-spatial attention. This experiment was based on the dot-probe task paradigm described in previous sections. On each trial, a probe appeared either on the left or right of fixation and participants were faced with a two alternative forcedchoice task that required them to make an appropriate keypress that indicated the probe’s position. Presentation of the probe was always preceded by 2 images that were presented on the both sides of the fixation cross that the probe could appear in. The images used were either upright or inverted face stimuli. In addition, each pair of images was either a mixed pairing of an upright and inverted face each or consisted of both upright or inverted face stimuli. Given that two cues were always provided on both sides of the fixation cross in each trial, there was no incentive for the participant to show an orienting bias to either cue. This was especially so, when the trials were balanced to ensure that the probe was just as likely to follow an upright face as it was, an inverted one. Despite this, the claim that upright faces captures attention predicted a spatial shift of attention towards the side of fixation that corresponded with the upright face cue, when presented in conjunction with a neutral cue i.e. inverted face. This was expected to result in faster probe-detection for when the upright face was a valid cue compared to when the probe followed the inverted face cue instead. However, this prediction is only valid for short cue-probe intervals less than 100 msec. Lengthening the cue-probe intervals to greater than 200 msec was expected to induce inhibition in the spatial regions that benefit from early facilitation. Thus, detection for probes cued by upright faces was expected to be slower than those cued by inverted faces instead, for long cue-probe intervals of about 500 msec. However, the above predictions could be only made for when the upright face cue is presented alongside an inverted face counterpart. As a control condition, same-cue pairings were introduced and probe detection that followed from these pairings were not expected to exhibited any facilitatory effect at all. Even if upright faces can serve as valid peripheral cues, Posner and colleagues (1980; 1984) have previously argued that selective attention cannot be simultaneously divided across 2 objects. They arrived at this conclusion when simultaneous cuing for both the peripheral positions that a probe could have appeared in failed to speed its detection over non-cued probes, at short cue-probe asynchronies {Posner, 1980 #228;Posner, 1984 #233}. Nonetheless, inhibition for this double-cuing procedure was maintained at both cued positions for long onset asynchronies of cue-probe despite this lack of an initial facilitatory effect. It will be interesting to note whether upright faces if they do capture attention, can demonstrate the same pattern of performance. In comparison to the results presented by Posner and colleagues (1980; 1984), a double-cuing procedure with upright faces ought to show no facilitatory effect for probe detection at short cue-probe onset asynchronies. However, it could still bring about an inhibitory aftereffect at long cue-probe onset asynchronies. This could be compared to trials of the same double-cuing design but with the use of inverted faces that are not thought to capture attention, and thus expected to demonstrate neither facilitatory nor inhibitory effect on the probes cued. Method Design This experiment followed a 2 (valid-cue type: upright vs. inverted) x 2 (pairing type: same vs. different) x 3 (cue-probe timings: see following for details) repeatedmeasures design. Altogether, there were 3 experimental blocks of 96 trials each. At the start of each trial, a pair of images was presented, each on either side of the fixation cross. One of the two images would share the same position as the probe and termed the valid- cue. This was either be an upright or inverted face. The image accompanying the valid cue was either an upright or inverted face; that is, same or different to the cue and thus determined the independent variable of pairing-type. Finally, the experimental blocks were each defined by the 3 varying time intervals of cue-probe and presentation length of cue. Specifically, these could be described as: i) cue (presentation time: 60 msec) – 0 msec cue-probe interval; ii) cue (presentation time: 60 msec) – 440 msec cue-probe interval; iii) cue (presentation time: 500 msec) – 0 msec cue-probe interval. Condition i) represented short cue-probe asynchronies whereas conditions ii) and iii) were variations of long cue-probe asynchronies. The trials within each block were equally distributed across the 4 trial conditions and counterbalanced for cue position; that is, left/right of fixation cross. In addition, the blocks are counter-balanced for serial order of presentation across participants. Participants 18 (age range: 18-19 years) undergraduate students of the Department of Psychology, University of Manchester, participated in this experiment as part of their course requirements. All participants reported normal or corrected-to-normal vision. Materials and apparatus The experimental set-up was similar to the experiments reported in Chapter 2. The experiment was controlled by an experimental generator software (Psyscope v.1.1.2), on a Macintosh G4. A viewing distance of 96cm was maintained throughout the experiment by use of a fixed chinrest. The same set of 24 photorealistic images of different faces, used in the experiments reported in Chapter 2, rendered a total of 48 upright and inverted faceimages that were used as valid cues. For the condition of pairing-type, each valid cue was randomly paired with an image of a different face, that was either of the same or different orientation to itself. Hence, there were a total of 96 unique pairings, each of which was used only once in each experimental block. Fig. 3.1. shows an example of the four different type of pairings. Figure 3.1: Examples of possible pairings between a valid cue and accompanying non-valid cue, represented by the left and right image respectively. Valid cue-type: upright face (a, b), inverted face (c,d); Pair-type: same (a,d), different (b,c). Procedure Each participant had to perform a 2 alternative forced-choice task in this experiment. On each trial a fixation cross was first presented for 1000 msec, followed by a pair of images (valid cue and accompanying image), each on either side of the fixation point. Participants were informed, prior to experimentation, that the these images were irrelevant to the task and held no information concerning probe location. Following this, the pair of images was replaced by a single probe that appeared on the side of the fixation cross, formerly occupied by the valid cue. Figure 3.1 illustrates the design of this experiment as well as the respective timings depending on the experimental block it occurred in. Participants were required to make fast and accurate keypress responses that indicated where they believed the probe to have appeared, relative to fixation. Nonetheless, the probe remained on the screen for as long as was necessary for participants to respond. The entire experiment was made up of 3 blocks of 96 experimental trials each and participants were allowed to take a break in between the experimental blocks. Each block of trials was defined by its unique combination of length of presentation for the cuepairing and cue-probe onset asynchrony (see Fig 3.2.). keypress response (b) i) 0 msec ii) 440 msec iii) 0 msec (a) i) 60 msec ii) 60 msec iii) 500 msec Fig. 3.2: Diagram illustrating the design of experiment 5. (i), (ii) and (iii) represent the collective timing conditions for the presentation time cue (a); and timing interval between cue-offset and probe-onset (b), that defined each of the experimental blocks. Inter-trial interval of 500 msec. At a fixed distance, the valid cue and accompanying non-cue subtend a maximum visual area of 1.5º x 2.1º each. Each cue is centred 3.1º on either side of the fixation cross which is itself, 0.42º x 0.42º. Results The dependent variable taken was mean detection latency, measured from probeonset. Collected timings were subjected to a 3 (Timing conditions) x 2 (valid cue type) x 2 (non-valid cue type) repeated measures ANOVA. There was only 1 main effect of that was statistically significant, timing conditions (F(2,34)=11.7, p<0.001). Neither the variable of valid cue-type (F(1,17)=0.02, p=0.89) nor pairing-type (F(1,17)=0.02, p=0.90) brought about significant diferences in detection latency. Table 1 presents a summary of the mean reaction times across the 3 independent variables. Upright-face Inverted-face valid cue Different 431 (18.9) Same 436.0 (20.3) Different 442 (19.7) Same 431.0 (16.0) 60(cue present.)+440 (blank interval) 417 (15.3) 431.6 (22.9) 418 (17.2) 420.5 (15.9) 500(cue present.) 365 (17.2) 370.1 (19.1) 381 (23.1) 361.8 (16.1) Timing conditions 60 (cue present.) Table 1: Summary of mean RTs in milliseconds (standard error is bracketed) across timing conditions for when i) the valid cue is an upright-face vs. inverted-face; and when the pairings were same or different. Pairwise comparisons indicate that long presentation times, i.e. 500ms, of cues resulted in significantly shorter mean RTs. When the presentation times of cues are short, i.e. 60ms, the addition of a 440ms blank interval did not bring about any significant difference in response times to probe. Finally, There were no significant interactions to report from this study’s results. Details are reported in Appendix A. Discussion In chapter 2, upright faces were shown to be unable to capture attention, at the expense of goal-directed behaviour. This study was carried out with the intention of determining whether upright faces could, nonetheless, orient attention when it was not contrary to goal-directed behaviour. Particularly, face-stimuli fell within the task designated area in Experiment 5. Furthermore, the target was a single item and detection was not reliant on object-identification at all. This is in contrast to the experiments previously reported, whereby target detection performance was reliant upon the participants’ ability to discriminate the target letter from an array of non-target items on the basis of the target’s identity. Despite these modifications, upright faces continued to have no effect in orienting spatial shifts of attention towards their locality. Having an upright face as a valid cue type did not influence probe responses in the same way as valid peripheral cues are generally reported to. Earlier, it was predicted that at shorter cue-probe onset asynchronies (less that 100msec), a valid upright face cue should facilitate probe-detection if it resembled a valid peripheral cue (cf{Posner, 1984 #233;Posner, 1980 #228}. Thus, faster detection ought to have been associated to probes cued by an upright faces, compared to inverted faces. This was not the case in Experiment 5. Furthermore, longer cue-probe onset asynchronies (~500 msec) ought to have resulted in an inhibition of attending to the spatial region formerly occupied by valid upright face cues, such as to induce longer detection latencies. Again, this prediction was not borne out. Inhibition was also thought to occur for long cue-probe onset asynchronies, when both cues were upright faces, without the facilitatory effect because early orienting of attention is not expected to be dedicated to more than one peripheral cue (cf. {Posner, 1984 #233;Posner, 1980 #228}. Specifically, this would have been expressed by a significant interaction across all the independent variables of timing conditions, valid cue-type and pairing-type. The facilitative effect to probe-detection for trials with valid upright-face cues would only be exhibited with when it was paired with an inverted face, and at short cue-probe onset asynchronies. Furthermore, long cue-probe onset asynchronies would have had an inhibitory effect on the detection of probes cued by upright faces, regardless of the face-image it was paired with. However, this critical interaction was not proven to be significant. Thus, the significant effects predicted by the original hypothesis that upright faces exert an exogenous control over spatial attention was not demonstrated in Experiment 5, even when attempts were made to remove any form of endogenous control. Instead, the significant main effect highlighted in this experiment was that of timing conditions. Regardless of the type of face-image that cued the probe, a longer interval between the cue and probe presentation resulted in quicker detection of the probe. A possible explanation for this could that participants were simply better prepared to respond when more time was given for the particpants to make a decision after the presentation of the paired cues . The findings reported here indicate that the image of an upright-face does not function as a valid peripheral cue any more so than an inverted-face image. This runs counter to the notion of an automatic face-detection system that automatically orients spatial shifts of attention to the visual presence of faces {de Gelder, 2001 #33}. If this claim was true, the results of Experiment 5 should have indicated early facilitation, and late inhibition, effects on detection towards probes cued by upright faces. Instead, the results did not reflect such a trend. These results mirror those reported in Chapter 2 and suggests that the upright face configuration might not orient attention any more than an inverted face configuration, whether or not endogenous controls of attention are in place.