جستجوی بصری برای حالات احساسات صورت در شبیه سازى کمتر تحت تاثیر قرارگرفته

Abstract Evidence in healthy human subjects has suggested that angry faces may be enhanced during spatial processing, perhaps even “popping-out” of a crowd. These contentions have remained controversial, but two recent reports in patients suffering from unilateral spatial neglect have lent some support to these views, suggesting that emotional faces capture attention more efficiently than neutral stimuli in the neglected field. Here, we investigate this phenomenon in a patient suffering from severe Balint's syndrome and consequent simultanagnosia. Using a visual search paradigm, we studied differences in the detection of angry, happy and neutral faces, as well as non-emotional stimuli. Results revealed that emotionally expressive faces, in particular anger, were detected more efficiently than other stimuli. These findings corroborate claims that facial expressions of emotion constitute a specific category of stimuli that attract attention more effectively, and are processed prior to attentional engagement.

Introduction In 1988, Hansen and Hansen (1988) reported that healthy control subjects could detect an angry face among a series of distracters consisting of happy or neutral faces more rapidly than the converse condition (i.e., detecting happy or neutral faces in an angry crowd). This so-called “face in the crowd effect” led to the contention that faces were processed pre-attentively for characteristics of threat, and consequently produced a pop-out phenomenon for angry faces. Subsequent research (Hampton et al., 1989 and Nothdurft, 1993) refuted the pop-out effect due to the observation that subjects' reaction times in these visual search tasks varied with the position of the target in the crowd, or with its size. Indeed, an increase in visual search time with a greater number of distracters is taken as an indication of attention being allocated to each element of the population in serial succession. Pre-attentive processing, or “pop-out”, requires the absence of any effect of population size on reaction time, denoting simultaneous processing of all elements in parallel (Treisman and Gelade, 1980). Since the findings with emotional faces demonstrated a serial search, a pop-out phenomenon for angry faces was ruled out. Further evidence (Purcell et al., 1996) demonstrated that the pop-out phenomenon obtained in Hansen and Hansen's original experiment was due to a low-level visual artefact that had appeared in the black on white versions of the (originally grey-scale) photographs used in the experiment. Indeed, when replicating the paradigm, this time with the original grey-scale pictures, happy and angry faces were detected after a self-terminating visual search. Fox et al. (2000) suggested a possible alternative while maintaining a special status for angry faces. Based on the reaction time data of a control population, these investigators noted that angry faces were searched at an estimated 16 msec per item and happy faces at 29 msec per item, compared to an expected <10 msec per item in automatic search. This, they pointed out, implied that angry face detection was not carried out in a fully automatic manner, although these stimuli were searched for and detected more efficiently. Clearly though, the effect of angry faces on attention remains controversial (see also Fox et al., 2000, Suzuki and Cavanagh, 1992 and White, 1995). However, two recent studies (Fox, 2002 and Vuilleumier and Schwartz, 2001), in patients suffering from unilateral spatial neglect following right hemisphere damage, suggested that brain damage did not affect attention for emotional faces and neutral stimuli to the same degree, again suggesting that emotional faces might constitute a particular category of stimuli. In this investigation, we examined whether a similar conclusion could be arrived at in a patient suffering from another spatial disorder, simultanagnosia, which occurred in the context of acute Balint's syndrome following bilateral parieto-occipital injury. This syndrome is relatively rare and was first described by Balint himself at the turn of the last century (Balint, 1909). It includes three major symptoms: optic ataxia, oculomotor apraxia and, of particular interest here, simultanagnosia (see Coslett and Chatterjee, 2003 or Rafal, 2001 for reviews). In simultanagnosia, patients are unable to perceive more than one object at a time, even though single stimuli are generally perceived correctly. This deficit occurs independently of the size of the objects, and is present even when they are in close proximity, or indeed superimposed (e.g., Luria, 1959 and Luria et al., 1963). Contrary to unilateral spatial neglect, simultanagnosia appears as an object-based constriction of the attentional field and is independent of the spatial location of the stimulus (Rafal, 2001). Obviously, this disorder produces severe difficulties in visual search and target detection tasks, and occurs independently of the visual field (Coslett and Saffran, 1991). The present study required that the simultanagnosic patients perform a series of visual search tasks in which emotional or non-emotional targets had to be detected among a series of distracters. The aim was to establish if target detection was impaired differentially for non-emotional stimuli or emotional faces despite severe simultanagnosia.

Results On initial pilot trials involving a serial search, it was obvious that the patient was rarely confident about the presence of a target, unless she accidentally came upon it at the onset. On occasions, she would continue exploring the scene seemingly endlessly, without reaching any definite conclusion. This, of course, reflected the erratic nature of the patient's visual search. Her visual exploration difficulties precluded any systematic search, leading to a continuous “open-ended” exploration of the visual scene. On the following trials, we therefore reassured the patient by emphasising the fact that a number of trials lacked targets and encouraging her to respond as quickly as possible, based on her general impression rather than certainty. With these instructions, the task became less arduous (roughly between 5 and 10 sec per item). However, since the search generally appeared incomplete, we surmised that the size of the set would not affect the correct response rate. The number of hits in 6 and 11 distracter conditions in each experiment was therefore compared to a “no difference” situation using chi square statistics with a significance level of .05. For 6 distracters, Experiments 1–7 yielded, respectively, 26/52; 48/52; 26/48; 26/48; 28/48; 34/48; 40/48 correct responses. For 11 distracters, these values were 31/52; 49/52; 38/48; 23/48; 30/48; 34/48; 42/48. Statistical comparisons between 6 and 11 distracter conditions using chi square tests were not significant and the data were subsequently collapsed across set size for analysis. The patient's and the control subjects' individual scores (correct responses and false alarms) were then used to compute d′ and c in each experiment. In the control group, several participants obtained response rates of 0 or 1; on these occasions, the rates were replaced by .01 and .99 in order to permit computation. 4.1. Stimulus detection: control subjects The results of the control subjects are presented in Table 2. These show the means and standard deviations of hits and false alarms, as well as the means, standard deviations and ranges of d′ and c. As can be noted, control participants showed excellent discrimination capacities with a mean d′ above 4 on all seven tasks. Response biases were all situated around 0. Table 2. Performance of patient MC and of control group on the seven search tasks Serial search Parallel search Face detection Joy control Anger control Joy Anger Patient MC Hit rate .55 .93 .67 .51 .60 .71 .85 FA rate .09 .03 .08 .17 .17 .08 .08 d′ 1.44 3.36 1.81 .99 1.23 1.93 2.44 c .6 .18 .48 .47 .35 .42 .16 Controls Hit rate (SD) .98 (.03) .99 (.01) .99 (.02) .98 (.01) .97 (.03) .98 (.02) .98 (.02) FA rate (SD) .03 (.03) .01 (.02) .0 (.01) .01 (.02) .01 (.03) .0 (.0) .01 (.02) d′ (SD) 4.10 (.41) 4.47 (.27) 4.42 (.27) 4.32 (.48) 4.23 (.43) 4.42 (.30) 4.35 (.31) Range 3.63–4.65 3.93–4.67 3.99–4.65 3.40–4.65 3.57–4.59 3.89–4.65 3.79–4.65 c (SD) −.06 (.25) −.05 (.11) .06 (.17) −.01 (.1) .06 (.25) .11 (.15) .03 (.21) Range −.51 – .51 −.23 – .13 −.3 – .33 −.3 – .03 −.44 – .54 .0–.38 −.3 – .43 Top row: Hits and false alarm rate for patient MC as well as the corresponding d′ and c values obtained in each experiment. Bottom row: Mean hit rate, false alarms, d′ and c for the control group. Standard deviations are given in parenthesis. Minimum–maximum d′ and c values produced by the control group are also indicated. Table options In order to test for possible differences between experimental conditions in controls, an analysis of variance for repeated measures was performed on d′ and c values. Compared to baseline serial search (Experiment 1), the pop-out task (Experiment 2) showed a marginally significant advantage (F(1,9) = 4.2; p = .07) for sensitivity. On the other hand, the response criterion did not differ between the two procedures (F(1,9) = .01; p = .91). The comparison of the sensitivities and response biases between serial search (Experiment 1), upright faces (Experiment 3), happy faces (Experiment 6) and angry faces (Experiment 7) showed no significant difference (F(3,27) = 1.93; p = .15 and F(3,27) = .81; p = .5, respectively). The two emotionally expressive face detection tasks (Experiments 6 and 7) were further compared each with their own specific control paradigms (Experiments 4 and 5, respectively). Neither the inverted versus upright happy face comparison, nor the inverted versus upright angry face comparison showed any difference in sensitivity (F(1,9) = .55; p = .48 and F(1,9) = .45; p = .52, respectively) or response criterion (F(1,9) = 1.89; p = .2 and F(1,9) = .07; p = .8, respectively). 4.2. Stimulus detection: comparison between MC and controls MC's hit rate, false alarm rate, d′ and c values are summarised in Table 2 (top). The results demonstrate that the patient's discriminative capacities were severely below normal limits in all visual search tasks, with d′ values situated over six standard deviations below the mean and well below the minimum value obtained in controls. One notable exception, however, was the pop-out task where she showed a less marked deficit (d′ = 3.36), although her score remained nonetheless ∼4 standard deviations below the mean. MC's response criteria, c, were generally more positive than in controls, situated at respectively (Experiments 1–7) 2.6, 2.1, 2.5, 4.8, 1.2, 2.1 and .6 standard deviations above the mean. Of note is the fact that only anger detection task showed a criterion within normal limits. The more positive c values on the tasks indicated a tendency towards a more conservative response bias, i.e., a global tendency to respond negatively more often than controls. 4.3. Stimulus detection: effect of stimulus-type in MC Although MC's scores were globally impaired in all tasks with respect to a group of matched controls, her performance did not appear to be homogenous throughout the seven experiments. In order to evaluate differences in MC's performance between experimental procedures, her d′ and c values were compared two by two in the different conditions ( Macmillan and Creelman, 1991). 4.3.1. Comparisons with baseline serial search MC's discriminative capacity on the pop-out task was significantly better than that in the serial search task (cumulative z-score = 3.31, p < .001) and the difference in response criterion was marginally significant (cumulative z-score = 1.432, p = .08). Compared to serial search, detection for neutral faces showed no statistically significant difference, either for sensitivity (cumulative z-score = .594, p = .28) or response criterion (cumulative z-score = .39, p = .35). Regarding the comparisons between serial search and the detection of emotionally expressive faces; no significant differences were observed for sensitivity (cumulative z-score = .78, p = .22) or response bias (cumulative z-score = −.574, p = .72) for joy compared to serial search. On the other hand, a marginally significant advantage in sensitivity in angry face detection was found when compared with the baseline serial search (cumulative z-score = 1.568, p = .058). These two conditions also showed a marginally significant difference when comparing the c values (cumulative z-score = 1.364, p = .09). Emotional faces compared to control conditions. The difference in sensitivity in the detection of upright happy faces (Experiment 6) compared to its corresponding inverted control condition (Experiment 4) was marginally significant (cumulative z-score = 1.339, p = .09), showing that the patient tended to detect happy faces more easily. In the case of angry faces, the patient was clearly better at detecting the upright angry faces, compared to the inverted control task (Experiment 5) (cumulative z-score = 1.709, p = .04). The response criterion did not differ in either of these comparisons (joy: cumulative z-score = −.152, p = .44; anger: cumulative z-score = −.531, p = .3). Finally, a comparison of d′ obtained in Experiments 6 and 7 (joy and anger upright) showed no significant difference (cumulative z-score = .662, p = .75).