Literature thesis
-Using the dot-probe task as a measurement of
emotion perception in comparative studies
Name: Rianne van Rooijen
University of Amsterdam
Student number: 6059333
Supervisor: Mariska Kret
Co-assessor: Annemie Ploeger
Abstract
For social animals the recognition of emotional facial expressions of other individuals is of importance for group survival. Whether these emotional expressions evolved over the course of evolution can be demonstrated by cross-species comparisons. Therefore, an experimental paradigm should be developed which is suitable for testing emotional
processing in several species. The dot-probe task is a good candidate as this test is implicit, does not need instruction and subjects need minimal training to perform the test
successfully. In this essay some important considerations when conducting the dot-probe task are discussed, regarding subject characteristics, type of stimuli and experimental procedure. When conducting comparative dot-probe studies with humans and non-human primates both groups should be matched on gender and relative age. In addition, emotional facial expressions which evoke similar levels of arousal in both groups should be used as stimuli. And last, probe detection instead of probe discrimination should be used to prevent problematic variability in reaction times. Conducting such cross-species dot-probe studies might eventually provide essential information about the evolution of emotional expressions.
Contents
1. Emotional signals as adaptations 5
2. Processing of emotional stimuli 9
3. Emotional signals and attention 11
4. The dot-probe task 13
5. Differences between dot-probe studies 17
Subjects 17
Age 17
Gender 19
Testosterone level 20
Serotonin transporter gene 21
Anxiety 21
Non-human primates 23
Type of stimuli 23
Level of threat 24
Differences in negative stimuli 24
Pictures vs. schematic faces 26
Face vs. whole body 28
Procedure 29
Stimulus presentation 29
Probe detection vs. probe discrimination 31
Emotional-neutral pair vs. emotional-emotional pair 33
Context 34
Reliability 35
5. Conclusion 35
References 39
Using the dot-probe task as a measurement of emotion perception in comparative studies For social animals the recognition of emotional signals of other individuals is of crucial importance for group survival and the maintenance of group bonds. For example, emotional signals play a significant role in warnings for predators and in partner choice. This is
supported by the observation that species which live in large groups are best able to
interpret facial expressions (Ellingson & Green, 2012). How these emotional signals originally developed is a topic of extensive research. Are the emotional signals of humans and other animals comparable? Do non-human animals perceive emotions of others in a similar way as humans? Humans can report on their emotion perception, but there is no direct
measurement of emotion perception in non-human animals. In order to investigate these questions a paradigm should be developed which is suitable for testing emotional processing in several species. A possible candidate for such a paradigm is the dot-probe task, as this test is implicit, does not need instruction and subjects need no or minimal training to perform the test successfully. In this essay will be discussed whether the dot-probe task is a relevant paradigm to investigate the evolution of emotion perception. Moreover, studies which conducted the dot-probe task with human subjects are reviewed. This provides an overview of some important considerations when conducting the dot-probe task.
1. Emotional signals as adaptations
It is hypothesized by evolutionary psychologists that emotional signals are
adaptations that evolved over the course of evolution. The expression of emotions and the perception of emotions of others is adaptive in social animals as it results in higher changes of group survival. Our ancestors evolved in an environment where the encounters with resources and danger were unpredictable. Chances of survival were therefore dependent on the ability of an individual to efficiently locate these events. Food and mating partners had to
be located whereas dangers had to be avoided (Öhman, Flykt & Esteves, 2001a). Emotional expressions convey important information about the presence of these positive and negative affairs in the environment. A fearful or threatening facial expression of another individual might signal danger in the environment. A sad facial expression is a request for help and a disgusted facial expression is an important signal for noxious stimuli (Seidel, Habel, Kirschner, Gur & Derntl, 2010).
Whether emotional expressions evolved over the course of evolution can be demonstrated in several ways. First, emotional expressions should be universal over many different cultures. Ekman and colleagues (1969) found that a tribe in New Guinea expressed and interpreted facial expressions in a similar way as people in the West even though this tribe never had had contact with the Western culture before. This study was extended to 21 countries around the world, which all showed similar results (Ekman, 1973). Thus, humans seem to have an innate module for expressing and recognizing emotional facial expressions. Although facial expressions appear to be universal, there are some slight differences in expressions between cultures. These differences can be compared to linguistic dialects, as cultural groups who live in proximity to each other are faster in recognizing each other’s facial expressions than more distinct groups (for a review see Elfenbein, 2013).
Second, emotional expressions should be universal over several species. It was found that non-human primates show similar emotional expressions as humans (Darwin, 1965). These emotional expressions are similar in both their morphological structure as in their social function (Preuschoft & van Hooff, 1995). For example, chimpanzees express the ‘relaxed open mouth display’ (ROM) during play, which is comparable with human laugh (see Figure 1; de Waal, 2003). When the ROM was bidirectionally expressed during play, the time
of playing increased (Waller & Dunbar, 2005). This suggests that the ROM signals social bonding and affection, similar to the function of human laugh.
In addition, non-human primates are not merely able to express emotions, but they also successfully distinguish emotional expressions. Research indicated that chimpanzees are capable of discriminating an emotional facial expression made by two different individuals from that of a neutral expression of a third individual. Successful discrimination of different emotional expressions was dependent on the amount of shared features between the two expressions (Parr, Hopkins & de Waal, 1998). Rhesus monkeys are also capable of
discriminating emotional expressions from neutral expressions, yet they had more difficulties with discriminating two different emotional expressions (Parr & Heintz, 2009).
Non-human primates also seem to process emotional signals in a similar way as humans. Chimpanzees are capable of recognizing the valence of an emotional expression as they successfully matched emotional videos (e.g., showing favorite food and veterinarian procedures) to videos of emotional expressions with the same emotional meaning (Parr, 2001). In addition, chimpanzees show cortical asymmetries in physiological responses when watching videos of conspecifics expressing a certain emotion similar to the responses observed in humans (Parr & Hopkins, 2000; for a review see Lindell, 2013).
Similar emotional processing in humans and other animals is further supported by the discovery of similar emotional circuits in the limbic system. For example, the brain structures responsible for the bodily responses associated with emotions in humans are homologue to brain structures found in the reptile brain (MacLean, 1993). Therefore,
investigation of the limbic system in non-human animals might provide relevant information about the functioning of human emotional processes (Panksepp, 1982). This highlights the importance of cross-species research.
Figure 1: Facial expressions of chimpanzees. (Image adapted from Cartwright, 2008; p.175)
For emotional expressions to be adaptive, the perception of these expressions should have an effect on behaviour. Expressions of fear and anger should elicit escape behaviour to efficiently avoid potentially dangerous situations. In contrast, positive expressions should elicit approach behaviour as these expressions signal desirable situations such as the presence of food (Lang, Bradley & Cuthbert, 1998). Indeed, it was found that different emotions are accompanied by different action tendencies. A total of eighteen action readiness modes are proposed, including approach, avoidance, attending and submission behaviour (Frijda, 1987).
To effectively initiate an action tendency in response to the perception of an emotional stimulus the perceptual system should be linked to the motor system. This link results in the unconscious predisposition to either approach or avoid a situation (Frijda, Kuipers & ter Schure, 1989; Öhman & Mineka, 2001). The link between the perceptual and motor system was investigated by presenting subjects with positive and negative words (Chen and Bargh, 1999). Half of the subjects had to push a lever away in response to negative
words and pull the lever in response to positive words. The other half of the subjects responded to the words in the opposite direction. Results showed that subjects were much faster in reacting to the stimuli in the congruent situation (positive/pull and negative/push) than in the incongruent situation (positive/push and negative/pull). This result indicates that positive stimuli indeed evoke an unconscious approach tendency, whereas negative stimuli evoke an avoidance tendency.
Similar results were found when emotional facial expressions were used as stimuli (Marsh, Ambady & Kleck, 2005; Seidel et al., 2010; but see Wilkowski & Meier, 2010).
Subjects showed an avoidance tendency for angry faces and an approach tendency for happy and sad faces. Faces with a disgusted expression did not evoke an action tendency, which might have resulted from the combination of disgusted expressions used in the task. In addition, these action tendencies not only apply to hand movements, but generalize to whole body movements (Stins, Roelofs, Villan, Kooijman, Hagenaars & Beek, 2011). Subjects stepped faster towards a smiling face compared to an angry face, yet no significant difference was found for stepping away from these faces. Overall, these studies indicate that
unconscious action tendencies play an important role in the processing of an emotional stimulus.
2. Processing of emotional stimuli
Currently, there are multiple views on the processing of emotional facial expressions. For instance, the discrete-category view (Ekman, 1992) states that facial expressions convey discrete emotions. In contrast, the dimensional view (Russell, 1980) states that facial
expressions do not explicitly convey a specific emotion, but rather provide values for valence and arousal. Subsequently, these values are used to attribute an emotion to the facial
expressions are unaffected by context (Carroll & Russell, 1996; Ekman & O’Sullivan, 1988). Which of these views describes the processing of emotional expressions best is still a topic of debate.
On a neuronal level emotional stimuli are consciously processed via the neocortex, supported by visual areas constituting the what- and where-pathways (for a review, see Mishkin, Ungerleider & Macko, 1983). The prefrontal cortex (PFC) also seems to play a major role in stimulus processing. The ventromedial PFC is involved in the detection of emotional expressions and the ventrolateral PFC is involved in discriminating between emotional expressions (Tsuchida & Fellows, 2012). Additionally, there also seems to exist a subcortical route which travels from the relay nuclei of the thalamus directly to the amygdala, superior colliculus and pulvinar (de Gelder, Frissen, Barton & Hadjikhani, 2003; Garvert, Friston, Dolan & Garrido, 2014). This route supports fast and unconscious processing of emotional stimuli which in turn modulates processes in cortical areas (for a review, see Johnson, 2005). The subcortical route is most commonly activated by fearful faces, but also by smiling faces (for a review, see Zald, 2003). The existence of a subcortical route is further supported by
neuroimaging studies which showed rapid and direct activation of the amygdala in response to the presentation of emotional information (e.g., Carlson, Reinke & Habib, 2009; Luo, Holroyd, Majestic, Cheng, Schechter & Blair, 2010).
The perception of a facial expression induces three consecutive processes (Phillips, Drevets, Rauch & Lane, 2003). First, the significance of the stimulus is identified, which is followed by the induction of an affective state by autonomic, neuroendocrinic and
behavioural responses. Last, the first two processes are modulated such that the produced behaviour is contextually appropriate. These processes seem to be supported by two neural systems (Phillips et al., 2003). A ventral system subserves the more automatic processes such
as the identification of the stimulus and autonomic responses. A dorsal system regulates the integration of the emotional input and cognitive processes such as selective attention.
The next section will focus on the relation between the perception of an emotional stimulus and attentional processes. The link between the perception of emotional stimuli and the automatic initiation of approach or avoidance behaviour by the motor system can only be beneficial when emotional stimuli are rapidly detected. Therefore, attentional processes should be specifically tuned for focusing on emotional stimuli.
3. Emotional signals and attention
For emotional signals to be beneficial in threatening situations it is essential that these signals are rapidly detected in the environment. Over evolution the stimuli relevant for survival became automatic triggers of attention, a process mediated by a
biologically-prepared mechanism (Öhman et al., 2001a). For example, attention was directed to the location of stimuli which acquired a negative emotional value through fear conditioning (Armony & Dolan, 2002). This process engaged the fronto-parietal network, associated with spatial attention. Moreover, in visual search tasks pictures of snakes elicited faster response times than pictures of flowers in both humans (Öhman et al., 2001a) and macaque monkeys (Shibasaki & Kawai, 2009). Emotional facial expressions also seem to have an effect on attentional processes. In general, faces with an angry expression are detected faster in a visual search task than happy or neutral faces (e.g., Öhman, Lundqvist & Esteves, 2001b; for a review, see Frischen, Eastwood & Smilek, 2008; but see Becker, Anderson, Mortensen, Neufeld, & Neel, 2011) and search times for emotional faces are not affected by array size (e.g., Eastwood, Smilek & Merikle, 2001; Fox, Lester, Russo, Bowles, Pichler & Dutton, 2000; Juth, Lundqvist, Karlsson & Öhman, 2005). The faster detection of emotional faces is
probably most influenced by the heightened arousal induced by these faces (Lundqvist, Bruce, & Öhman, 2014).
Instead of driving attention towards its location, a negative valued stimulus might also drive attention away from its location (Mogg, Bradley, Dixon, Fisher, Twelftree & McWilliams, 2000). In some situations it might be beneficial to maintain attention on current goals
without being distracted by task-irrelevant negative cues. In addition, driving attention away from negative information might serve to reduce anxious mood and maintain positive mood states. The process of driving attention away from negative stimuli seems mainly applicable to mildly aversive information.
Several concepts are related to the interaction between emotion and attention (Yiend, 2010). One of these concepts is “selection”, a process which selects that part of the total visual input which is relevant for further processing. This selection process depends on different aspects, which are the properties of the situation at hand and the goals and
expectations of the individual (Yantis, 1996). An important step in the selection process is the allocation of attention to a certain location. Stimuli which are simultaneously present
compete for processing space (Desimone & Duncan, 1995). The stimuli present in the
attended location become amplified and are therefore more prone to be selected for further processing. Attention is allocated more rapidly to a location containing emotional stimuli than to a location containing neutral stimuli. Moreover, the disengagement from a location containing emotional stimuli is more difficult. As a result, emotional stimuli are selected for further processing more often than neutral stimuli (e.g. Bradley, Mogg, Millar, Bonham-Carter, Fergusson & Jenkins, 1997; Yiend, 2010). The prioritized selection of emotional stimuli is even enhanced when arousal increases (Lee, Sakaki, Cheng, Velasco & Mather, 2014;
Mather & Sutherland, 2011). The emotional dot-probe task is based on this principle, which will be discussed in the next section.
4. The dot-probe task
The previous section described that there is a strong relation between emotional stimuli and attentional processes. From an evolutionary viewpoint, emotional signals should be detected rapidly to initiate avoidance behaviour in potentially threatening situations. This evolutionary viewpoint will be strengthened if research can indicate that humans and non-human animals possess similar attentional processes with regard to emotional stimuli.
An experimental paradigm which is potentially suitable for comparing the automatic attentional processes involved in emotion perception in humans and non-human animals is the dot-probe task, first described by MacLeod, Mathews and Tata (1986). This task is implicit, does not need instruction and participants need no or minimal training to perform the test successfully. Hence, the dot-probe task is suitable for testing children, clinical
patients, non-human primates and several other animals. Comparing the performances of all these subgroups gives insight in the developmental course of emotion perception and
enables cross-species comparisons. Eventually, essential information about the evolution of emotional signals as possible adaptations might be obtained.
In the dot-probe task two stimuli are simultaneously displayed, each on one side of the screen. One or both of these stimuli have an emotional value. The presentation of these two stimuli is followed by an emotionally neutral task, which involves the detection or discrimination of a probe in the location of one of the stimuli. An advantage of this paradigm is that the emotional stimuli are not present at the moment of response, which avoids response-biases.
A particular dot-probe task might present two pictures, side-by-side. One picture of a face with an angry expression and the other of a face with a neutral expression. After the offset of the pictures a dot emerges either in the location of the angry face (the congruent condition) or in the location of the neutral face (the incongruent condition). The participants have to indicate by a button press whether the dot appeared on the right or the left side of the screen. See Figure 2 for a schematic representation of the procedure. If the participants’ attention was automatically directed to one side of the screen by one of the pictures, then reaction times for detecting the dot at this location will be faster than reaction times for detecting the dot in the unattended location. Thus, reaction time is a function of the interfering emotional expressions. To determine which stimulus attracted attention the reaction time of congruent trails is subtracted from the reaction time of incongruent trials, which gives the attentional bias score (MacLeod et al., 1986). A positive value indicates an attentional bias towards the emotional face. In contrast, a negative value indicates an attentional bias away from the emotional face as the reaction times were faster when the dot appeared at the location of the neutral face. This suggests avoidance of the emotional stimulus. A value of zero suggests that both stimuli received a similar amount of attention.
An attentional bias towards emotional stimuli might result from either faster orientation towards the emotional stimulus or difficulties with disengaging from the emotional stimulus. These two processes are both components of the attentional system (Posner & Petersen, 1990). Faster orientation towards emotional stimuli is called vigilance (Koster, Crombez, Verschuere & de Houwer, 2004), which is often observed in individuals with high anxiety levels or individuals who suffer from hypersensitivity (Williams, Watts, MacLeod & Mathews, 1997). Vigilance results in heightened sensitivity for negative information in clinical patients, which leads to less efficient processing of information
important for the ongoing behaviour (Eysenck, 1992). More recent research suggests that an attentional bias towards emotional stimuli is not due to vigilance, but rather a result of difficulties with disengagement from the emotional stimuli (Fox, Russo & Dutton, 2002).
This issue was investigated by conducting a specific version of the dot-probe task (Koster et al., 2004). Participants were not only presented with trials consisting of a neutral and emotional stimulus, but also with trials consisting of two neutral stimuli. The authors argued that it would be possible to disentangle the process of vigilance and disengagement by comparing the reaction times on threat-neutral trials with neutral-neutral trials. Vigilance for threat should result in faster reaction times for congruent threat-neutral trials compared to neutral-neutral trials. In contrast, difficulties with disengagement should result in slower reaction times for incongruent threat-neutral trials compared with neutral-neutral trials. Results showed that participants reacted slower on congruent threat-neutral and
incongruent threat-neutral trials compared to neutral-neutral trials. This indicates that an attentional bias towards threat is not a result of faster orientation towards emotional stimuli, but rather of difficulties with disengaging from these stimuli.
Currently, the dot-probe task is widely used for assessing attentional biases in clinical groups. For example, the dot-probe task is often conducted to compare the attentional biases for emotional stimuli in people with an anxiety disorder and healthy controls. Meta-analyses indicated that anxious individuals show an attentional bias towards threatening stimuli, whereas non-anxious individuals show no attentional bias or even avoidance of threatening stimuli (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg & van IJzendoorn , 2007; Frewen, Dozois, Joanisse, & Neufeld, 2008). Additionally, anxious individuals show avoidance of positive stimuli, whereas non-anxious individuals show a bias towards positive stimuli (Frewen et al., 2008). A similar difference in attentional biases was found for
depressed and non-depressed individuals (Peckham, McHugh & Otto, 2010)
It is assumed that the dot-probe task is a valid measure of attentional biases. However, studies which investigated attentional biases by conducting the dot-probe task show inconsistent results. There is no general consensus in the field about how emotional stimuli affect response times in this task. Yet, it is difficult to compare the results of all studies because of the variability in the parameters and populations used for testing. These methodological issues might contribute to the discrepancies in the literature. Meta-analyses showed that threat intensity determines the magnitude of the attentional bias (Frewen et al., 2008). Moreover, the attentional bias is dependent on stimulus presentation time (Bar-Haim et al., 2007; Frewen et al., 2008). These methodological issues will be discussed in the next section.
The main goal of this essay is to examine whether the dot-probe task is suitable to be performed by children and non-human primates, which would provide information about the evolution of the interaction between attentional processes and emotional signals. Most studies discussed in the next section will therefore be studies which conducted the visual
dot-probe task, using images as stimuli and not words. Words lack ecological validity and may therefore provide limited insight in attentional biases for emotional information. Appendix A provides a table of the dot-probe studies reviewed. The table gives a reference of used parameters and results obtained with the visual dot-probe task. To enable good comparisons, studies were only included when attentional biases scores were calculated and analyzed. Moreover, studies which only reported group comparisons were excluded as this gives no information about the presence of attentional biases within the groups.
5. Differences between dot-probe studies Subjects
It is suggested that attentional biases are highly influenced by personal characteristics. Specific emotional and personal variables may determine whether an
individual shows an attentional bias towards or away from emotional stimuli. It is still unclear how these personal characteristics influence performances on the dot-probe task (Mogg et al., 2000a). Here will be discussed how age, gender, genotype and level of anxiety might influence performances on the dot-probe task.In addition, a comparison will be made between performances of humans and rhesus monkeys on the dot-probe task.
Age. It is hypothesized that over age the attention for emotional stimuli changes as older people have more emotionally meaningful goals than younger people (Mather & Carstensen, 2003). There is a significant correlation between age and attentional bias for happy faces. Older people show higher vigilance for happy faces than younger people (Lindstrom, Guyer, Mogg, Bradley, Fox, Ernst, et al., 2009). Moreover, older people apply an active emotion-regulation strategy where positive information is enhanced and aversive information avoided. This shift in attention is accompanied by reduced amygdala activation
and enhanced prefrontal cortex activation in response to aversive stimuli (Nashiro, Sakaki & Mather, 2011).
A meta-analysis of emotion processing in younger and older adults found a limited effect of age on emotional processing (Murphy & Isaacowitz, 2008). However, this meta-analysis comprised several experimental procedures. Differences between age groups are observed when merely focusing on dot-probe studies. Mather and Carstensen (2003) tested a younger age group (mean age of 25.4 years) and an older age group (mean age of 71.5 years) with the dot-probe task. Happy, sad and angry faces were used as the emotional stimuli, each paired with a neutral face. The younger group did not show differences in reaction times for neutral and emotional faces. In contrast, the older group showed an attentional bias towards happy faces and an attentional bias away from negative faces. This indicates that in older adults attention is drawn by positive information, whereas negative information is avoided. The authors replicated their own experiment and similar results were found, yet older subjects did not show an attentional bias towards happy faces in this
experiment. Two other studies also indicated that older subjects show an attention bias away from angry faces (Orgeta, 2011; Tomaszczyk & Fernandes, 2014).
In contrast, Isaacowitz and colleagues (2006) did not find an attentional bias away from negative faces in an older age group. There are several possible explanations for this discrepancy in results. First, all three studies which did find an attentional biases used photographs of faces as stimuli, whereas Isaacowitz et al. (2006) used schematic faces. Second, the stimulus presentation time differed between the studies. In Isaacowitz’ study the face pair was presented for 2000 ms, instead of 1000 ms in the other studies. Last, Isaacowitz et al. (2006) used sad faces as negative stimuli whereas the other studies used a combination of sad and angry faces (Mather & Carstensen, 2003) or just angry faces (Orgeta, 2011;
Tomaszczyk & Fernandes, 2014). The possible influences of these different parameters on results of the dot-probe task will be discussed in later sections.
Performance on the dot-probe task is also investigated in children and adolescents. The majority of studies did not find attentional biases for emotional stimuli in the normal population (Heim-dreger, Kohlmann, Eschenbeck & Burkhardt, 2006; Pérez-Edgar, Reeb-Sutherland, McDermott, White, Henderson et al., 2011; Roy, Vasa, Bruck, Mogg, Bradley et al., 2008; Susa, Pitică, Benga & Miclea, 2012; Waters, Kokkoris, Mogg, Bradley & Pine, 2010; Waters, Mogg, Bradley, & Pine, 2008) although one study found avoidance for negative faces and vigilance for positive faces in 8-year-old children (Brown, McAdams, Lester, Goodman, Clark & Eley, 2013). An explanation for the different result in the study of Brown et al. (2013) might be that they used a combination of facial expressions as negative stimuli whereas the other studies only used angry expressions. Moreover, in the studies which showed no bias the children had to press a button the moment they perceived the probe, whereas in Brown’s (2013) study the children had to determine whether the probe was a triangle or a square. The second method requires more cognitive load and might therefore measure different processes.
Gender. Most studies are conducted with psychology students, which results in a subject pool of primarily women. However, it is possible that men and women perform differently on the dot-probe task as a result of differences in emotional reactions and
attention. In ancient times men went out for hunting whereas women stayed home with the children. Consequently, men were exposed to danger more often than women and therefore might also show different attentional processes in reaction to threat.
A dot-probe study indicated that women showed attentional biases for angry faces, yet men did not express such a bias. Moreover, women showed vigilance for happy faces,
whereas men showed avoidance of happy faces. (Tran, Lamplmayr, Pintzinger & Pfabigan, 2013). Additionally, a study with children indicated that boys did not differ in their biases for fear-related and pleasant pictures whereas girls were more vigilant for fear-related pictures (Waters, Lipp & Spence, 2004). This effect was not present in the adult data. Furthermore, daughters of depressed mothers showed vigilance for sad faces, yet boys did not express such a bias (Kujawa, Torpey, Kim, Hajcak, Rose et al., 2011). These results indicate that caution is needed when interpreting data of subject groups which contain both genders.
Testosterone level. Research indicated that testosterone level in both men and women has influence on selective attention for angry faces (van Honk, Tuiten, Verbaten, van den Hout, Koppeschaar et al., 1999; Wirth & Schultheiss, 2007). King and colleagues (2012) conducted a dot-probe task to investigate the effects of testosterone administration on attentional biases in rhesus monkeys. The rhesus monkeys were first trained on the dot-probe paradigm by making use of a touch screen procedure. The monkey had to touch the dot on the screen in order to receive a food reward. During the actual task pictures of objects with an emotional value and facial expressions of rhesus monkeys were shown. At baseline monkeys were significantly faster in reacting to a dot replacing a negative face compared to a dot replacing a neutral face. Differences in reaction times for positive faces and objects were not significant. When treated with testosterone the monkeys showed an overall decrease in reaction times. Furthermore, they showed an attentional bias away from negative objects and an attentional bias towards positive faces. Thus, the administration of testosterone resulted in opposite attentional biases for threat. However, monkeys treated with a placebo showed the same changes in attentional biases. Therefore, it seems likely that repeatedly conducting the test, 4 days a week for 4 months, caused these changes in attentional biases and not the administration of testosterone.
Serotonin transporter gene. Individuals which have a short allele in the promoter region of the serotonin transporter gene (5-HTTLPR) show higher amygdala activity when observing threatening faces compared to individuals with two long alleles (for a review see Hariri & Holmes, 2006). This suggests that these individuals might also show differences in performance on the dot-probe task. Indeed, dot-probe studies showed vigilance for negative stimuli in short allele carriers (Carlson, Mujica-Parodi, Harmon-Jones & Hajcak, 2012), yet homozygous long allele carriers showed avoidance of negative stimuli (Carlson et al., 2012; Fox, Ridgewell, & Ashwin, 2009). Moreover, homozygous long allele carriers allocate attention towards positive stimuli (Fox et al., 2009). Overall, there seems to be a linear relationship between 5-HTTLPR genotype and attentional bias for emotional faces. Vigilance for angry faces decreases as a function of the number of long alleles, whereas vigilance for happy faces increases (Pérez-Edgar, Bar-Haim, McDermott, Gorodetsky, Hodgkinson et al., 2010).
Stimulus presentation time seems an important moderator of attentional biases. One study showed that short allele carriers only showed vigilance for spider pictures at long stimulus presentation (Osinsky, Reuter, Küpper, Schmitz, Kozyra, Alexander & Hennig, 2008). Another study showed that short allele carriers attended towards fearful faces only at short stimulus presentation whereas homozygous long allele carriers expressed a bias towards angry faces only at long stimulus presentation (Thomason, Henry, Paul Hamilton, Joormann, Pine et al., 2010).
Anxiety. Anxious and non-anxious individuals show differences in brain activation in response to emotional stimuli. EEG research indicated that anxious individuals show
increased brain responses for emotional faces, irrespective of emotional expression (Rossignol, Campanella, Bissot & Philippot, 2013). Moreover, anxious individuals show
abnormal brain activity in hippocampal areas during the disengagement from threat (Price, Siegle, Silk, Ladouceur, McFarland et al., 2014).
It is hypothesized that healthy people have an attentional threshold for emotional stimuli (MacLeod et al., 1986). Only highly threatening stimuli attract attention, whereas mildly threatening stimuli are ignored. People with an anxiety disorder might lack this threshold, such that mildly threatening stimuli are also attended. A dot-probe study indeed indicated that both anxious and non-anxious individuals avoid low threatening stimuli and show vigilance for highly threatening stimuli. Yet, anxious individuals shift earlier from avoidance to vigilance than non-anxious individuals as threat increases (Wilson & MacLeod, 2003). Thus, it is important to take the threat level of the stimuli into account when
conducting a dot-probe task with a mix of anxious and non-anxious individuals.
Furthermore, it is important to make a distinction between different measures of anxiety. Dividing participants into a high and low anxious group can be based on several anxiety measures. Which anxiety measure is used can influence the findings of the study. For example, a difference in attentional biases in high and low health anxious people in response to health threat pictures was found when the group was divided on the anxiety sensitivity index. However, such a difference was not found when the group was divided on the illness attitudes scale (Lees, Mogg & Bradley, 2005). It is therefore difficult to compare studies which used different anxiety measures. To enable better comparisons between studies which examine the relation between attentional biases and anxiety, a general set of questionnaires to measure anxiety should be proposed. This set can then be used in all future studies.
Non-human primates. To examine the evolution of emotional signals it would be interesting to directly compare non-human primates and humans on the same test. To my
knowledge, only two studies have conducted the dot-probe task in non-human primates. These studies showed that rhesus monkeys were significantly faster in reacting to a dot replacing a negative face compared to a dot replacing a neutral face (King, Kurdziel, Meyer & Lacreuse, 2012). In addition, the attentional bias for negative facial expressions in rhesus monkeys are affected by oxytocin administration (Parr, Modi, Siebert & Young, 2013) in similar ways as in humans (Kim, Oh, Corfield, Jeong, Jang, & Treasure, 2014).
The performance of humans and rhesus monkeys on the dot-probe task is directly compared by Lacreuse and colleagues (2013). Humans performed the dot-probe task by reacting to the probe with two keys on a keyboard, whereas the monkeys reacted directly on the probe by touching it on a touch screen. Humans showed an attentional bias towards negative faces, yet they avoided negative valenced objects. Monkeys also showed an
attentional bias towards negative monkey faces, but did not show a bias for objects. Thus, for social stimuli humans and monkeys show the same pattern in attentional biases. These findings support the hypothesis that over evolution facial expressions became important emotional signals.
Type of stimuli
The dot-probe task can be conducted with many different sorts of stimuli. Often human faces are used as stimuli, either pictures or schematic. Other studies use pictures of objects or scenes instead of faces. Furthermore, the used emotional facial expressions vary between studies. Sometimes fearful expressions were used, whereas other studies focused on anger or sadness. Do these differences in stimuli set have an effect on the performance on the dot-probe task?
Level of threat. In the section about anxiety was already mentioned that the threat level of the stimuli is important. It was shown in a dot-probe study that reaction times
slowed down linearly as the amount of threat increased (Koster et al., 2004). Thus, there seems to be a task-interference effect of threatening information. This finding is in line with theories suggesting that attending to threat disrupts ongoing behaviour (Mogg & Bradley, 1998). Most attentional resources are allocated to the threatening information which interferes with performance on the task. In accordance, the attentional bias towards threat increases with increasing threat level for both threatening faces (Wilson & MacLeod, 2003) and threatening scenes (Koster, Verschuere, Crombez & van Damme, 2005; Mogg,
McNamara, Powys, Rawlinson, Seiffer & Bradley, 2000)
It is therefore important to take the level of threat into account when making group comparisons. In comparative studies, the used stimuli should have similar threat levels and evoke similar levels of arousal. In studies with humans the participants can rate the stimuli on threat level, such that studies with similar rated stimuli can be compared. The level of arousal in non-human primates could for instance be measured with skin temperature. Decreases in skin temperature indicate negative arousal (Parr, 2001). Measuring arousal in humans in a similar way might enable the comparison of the threat level of stimuli used in dot-probe tasks conducted with humans and non-human primates.
Differences in negative stimuli. Several dot-probe studies used a combination of facial expressions as negative stimuli, such as sadness and anger (Mather & Carstensen, 2003) or sadness, anger, fear and disgust (Brown et al., 2013; Mansell, Clark, Ehlers & Chen, 1999). No distinction was made between these different facial expressions in the analysis. All of these expressions are interpreted as aversive as they evoke similar neurological and physiological responses (Marsh et al., 2005). However, responses to these expressions might differ if these expressions evoke different action tendencies. Angry faces evoke an avoidance tendency (Marsh et al., 2005; Seidel et al., 2010), whereas fearful faces evoke an approach
tendency (Marsh et al., 2005). These results indicate that anger automatically initiates avoidance behaviour, whereas fear facilitates approach behaviour.
Thus, fearful and angry expressions seem to evoke opposite action tendencies, which might also affect behaviour on the dot-probe task. A study by Cooper et al. (2011) indicated that adding angry faces to a stimulus set of neutral, happy and fearful faces resulted in opposite attentional biases. Unfortunately, no direct comparison between the reaction to fearful and angry faces has been measured with the dot-probe task. However, based on the discussed studies it becomes clear that caution is needed when using both fearful and angry faces as negative stimuli. If behaviour in reaction to these faces is pooled, this might result in averaging out effects.
In addition, gaze direction might be a complicating factor in the comparison of fearful and angry facial expressions. The findings of action tendencies are based on faces with direct gaze. Direct gaze is an important signal in angry faces, because the anger is then directed to the observer. Yet, averted gaze might be more important in fearful faces because this indicates threat in the environment (Adams, Gordon, Baird, Ambady & Kleck, 2003). This is supported by the finding that participants were faster to categorize angry faces with a direct gaze compared to angry faces with an averted gaze. In contrast, fearful faces were
categorized more rapidly with an averted gaze (Adams & Kleck, 2003). Thus, when angry and fearful faces are both used as negative stimuli it might be useful to include angry faces with a direct gaze and fearful faces with an averted gaze.
Pictures vs. schematic faces. The majority of dot-probe studies used pictures of human faces as stimuli. However, several studies used schematic faces instead of pictures (e.g., Heim-dreger et al., 2006; Isaacowitz et al., 2006; Orgeta, 2011). In pictures, visual
features other than facial expression might attract attention and evoke an attentional bias which is not attributable to the valence of the stimulus (Purcell, Stewart & Skov, 1996). In schematic faces these distracting visual features are eliminated. In the section about age was described that studies using pictures of human faces and studies using schematic faces have led to contrasting results. However, there were more methodological differences between these studies. The question is whether the difference in stimuli underlied the contrasting results.
One dot-probe study directly compared subjects’ responses to pictures of emotional faces and schematic emotional faces (Orgeta, 2011). Results showed that the performance of the subjects in both tests was comparable. Thus, the type of stimulus does not seem to have an effect on attentional biases. This was further supported by a control experiment
performed by Fox (2002). In this experiment the dot-probe task was performed with inverted faces, which disrupts holistic processing. If visual features other than the emotional
expression were responsible for the attentional bias, an attentional bias should also be found for inverted faces. No attentional biases were found in the control experiment, which
suggests that the emotional expressions indeed evoked the attentional bias rather than other visual features.
Facial vs. non-facial stimuli. Threat is not only signaled by facial expressions, but also by non-facial stimuli like predators or weapons (Carlson, Fee & Reinke, 2009). For instance, relevant pictures of snakes and spiders were detected more rapidly among fear-irrelevant pictures of flowers and mushrooms than the other way around (Öhman et al., 2001a). A similar result was found for angry and neutral faces (Öhman et al., 2001b), which suggests that threatening facial stimuli and threatening non-facial stimuli elicit similar attentional biases.
The results of dot-probe studies conducted with anxious individuals are mostly consistent. Studies which used facial stimuli found that anxious people show an attentional bias towards emotional faces (e.g., Bradley, Mogg, Falla & Hamilton, 1998; Ioannou, Mogg & Bradley, 2004; Mogg, Philippot & Bradley, 2004). Similar results were found in studies which used spider stimuli (e.g., Mogg & Bradley, 2006). A possible explanation for this vigilance might be that looking away from threatening stimuli does not remove the threat. Tracking the movements of threatening people and animals is more beneficial than ignoring them.
The results of dot-probe studies conducted with non-anxious individuals are less consistent. Studies with non-facial stimuli often found no bias for threat stimuli (e.g., Mogg & Bradley, 2006; Mogg, Bradley, Miles & Dixon, 2004; Lees et al., 2005). In contrast, some studies with facial stimuli found a bias towards threat (e.g., Tomaszczyk & Fernandes, 2014), some studies found a bias away from threat (e.g., Bradley et al., 1997) and yet other studies found no bias (e.g., Bradley et al., 1998). This inconsistency in results could be due to one or more parameters discussed in this essay.
A complicating factor in comparing studies with facial and non-facial stimuli is the high individual variability in attentional biases for non-facial stimuli. A dot-probe study with a stimulus set of pictures of angry faces, attacking dogs, attacking snakes, pointed weapons and violent scenes showed that 34% of the participants showed a general bias towards threat and 20.8% showed a general bias away from threat. However, 34% of the participants showed a bias towards some categories and away from other categories (Zvielli, Bernstein & Koster, 2014). Whether the same pattern in individual differences can be found for biases for negative faces needs to be investigated.
In sum, it is difficult to conclude whether threatening facial and threatening non-facial stimuli evoke similar attentional processes. More information should be provided by
conducting studies with both facial and non-facial stimuli. To my knowledge, only one study included both types of stimuli. This study showed that subjects displayed an attentional bias towards threatening faces and an attentional bias away from negative non-facial stimuli (Lacreuse, Schatz, Strazzullo, King & Ready, 2013). This result suggests that facial and non-facial stimuli have opposite effects on attentional biases. Yet, more studies should be conducted to draw firmer conclusions.
Face vs. whole body. Most dot-probe studies used human faces as emotional stimuli. Although the face is a very important part of the body for emotion perception, it is not the only body part which conveys emotional information. The human body as a whole is also an important messenger of the emotional state of a person. Emotion recognition is supported by both facial and bodily expressions as reaction times for categorizing emotional
expressions increased when face and body expressed different emotions instead of the same emotion (Kret, Stekelenburg, Roelofs & de Gelder, 2013; Shields, Engelhardt & Ietswaart, 2012).
Research indicated that facial and bodily emotional expressions are processed in a similar fashion. People are as good in categorizing facial as bodily expressions of fear and happiness (van de Riet, Grèzes & de Gelder, 2009). In addition, the observation of facial and bodily expressions elicits similar N170 waveforms, which were not found for the observation of objects (Stekelenburg & de Gelder, 2004). Moreover, the categorization of fearful
expressions activated brain areas associated with emotional processing, irrespective of the type of stimulus (de Gelder, Snyder, Greve, Gerard & Hadjikhani, 2004; van de Riet et al., 2009).
Facial and bodily emotional expressions also seem to have similar effects on
both facial and bodily expressions (Kret et al., 2013). Therefore, it would be interesting to examine whether facial and bodily expressions elicit similar attentional biases on the dot-probe task. To my knowledge, such a study has not been performed yet. Including bodily expressions as stimuli might add ecological validity to the test as faces are always
encountered in the presence of a body. Whole-body expressions might therefore be better stimuli for investigating the evolution of emotional signals.
Procedure
Not only stimulus type differs across dot-probe studies, there are also many differences in the experimental procedures of these studies. The possible influences of several of these differences will be discussed in the following section.
Stimulus presentation. The majority of dot-probe studies used a stimulus
presentation of 500 ms. A problem with a stimulus presentation of 500 ms might be that attention has already shifted between the stimuli during this time frame. Thus, responses given after a stimulus presentation of 500 ms do not necessarily provide information about initial orientation (Bradley, Mogg & Millar, 2000). Therefore, some dot-probe studies used a subliminal stimulus presentation. A meta-analysis showed that the effect size of subliminal presentation was twice as large as that of supraliminal presentation in data of anxious individuals (Bar-Haim et al., 2007). This suggests that attentional biases for threat are mostly supported by unconscious processes.
Direct comparison between subliminal and supraliminal stimulus presentation revealed that for fearful facial expressions the presentation duration does not have an effect on attentional biases. Fox (2002) compared a stimulus presentation of 500 ms with a
stimulus presentation of 17 ms after which the stimuli were masked with a scrambled face. Participants showed similar attentional biases for fear, irrespective of stimulus presentation
time. There is also no main effect of presentation time on responses to happy and sad facial expressions (Ortega, 2011). Presentation time did have an effect on attentional biases for angry faces in an older age group of participants (Ortega, 2011). Older subjects showed an attentional bias away from angry faces with a stimulus presentation of 1000 ms, whereas stimulus presentations of 17 ms and 500 ms did not reveal an attentional bias. However, this effect was not found in the younger age group. Thus, subliminal presentation or a
presentation of 500 ms does not seem to result in different attentional biases for facial stimuli. In contrast, stimulus presentation time seems to affect attentional biases for non-facial stimuli. Subjects showed attentional biases at a presentation of 500 ms, yet no biases were found at a presentation of 100 ms (Mingtian, Xiongzhao, Jinyao, Shuqiao & Atchley, 2011).
Longer presentation times than 500 ms were also used in dot-probe studies. It is hypothesized that people, and anxious individuals in particular, initially attend to a
threatening stimulus but then reallocate their attention to avoid the threat (e.g., Mathews, 1990). This hypothesis was confirmed by several studies which directly compared a stimulus presentation of 500 ms with longer stimulus presentation times (i.e., 1250 ms, 1500 ms or 2000 ms). Attention allocation in non-anxious individuals is not affected by stimulus presentation time for either threatening facial stimuli(e.g., Bradley et al., 1998; Ioannou et al., 2004; Mogg et al., 2004a; Pérez-edgar et al., 2010b) or threatening non-facial stimuli (Lees et al., 2005; Mogg et al., 2004b; Mogg & Bradley, 2006). Anxious individuals initially show an attentional bias towards threat, yet when stimulus presentation time increased this bias disappeared or turned into avoidance of threat (Ioannou et al., 2004; Koster et al., 2005; Lees et al., 2005; Mogg et al., 2004a; Mogg et al., 2004b).
However, an alternative explanation might be that with longer stimulus presentation anxious individuals make more eye-movements (Stevens, Rist & Gerlach, 2011). Research indicated that eye-movements might influence the findings on a dot-probe task (Petrova, Wentura & Bermeitinger, 2013). Attentional biases were calculated separately for trials in which the subjects did and did not make an eye-movement. An attentional bias towards angry faces was found for trials without eye-movements yet this bias was not found for trials with eye-movements and across all trials. An attentional bias towards angry faces across all trials was found when the participants were instructed to not make eye-movements. Thus, a specific instruction to not make eye-movements should be included in all future dot-probe studies. Whether the observation that anxious individuals show vigilance for threat with short stimulus presentations and show avoidance of threat with longer stimulus
presentations is due to more eye-movements should be further investigated.
Probe detection vs. probe discrimination. Most dot-probe studies measure probe detection times, which means that participants have to react on the appearance of the probe as quickly as possible. In contrast, other studies measure probe discrimination times, which means that participants react on the identity of the probe. The identity of the probe can for instance be based on the shape or the color of the probe. An advantage of probe
discrimination is that participants are encouraged to monitor both sides of the screen. Yet, probe discrimination requires more cognitive load than probe detection because of the arbitrary relationship between stimulus and response. Therefore, the findings of both versions of the dot-probe task might differ.
Both versions are directly compared by two studies which only differed in the required response to the probe. These studies measured attentional biases for threatening and happy facial expressions. In one study the subjects had to indicate whether the probe
appeared on the left or the right side of the screen (Mogg & Bradley, 1999). In the other study the subjects had to indicate whether the probe consisted of two vertically orientated dots (:) or two horizontally orientated dots (..) (Bradley et al., 1998). Both studies found an attentional bias towards threatening faces in anxious individuals and no bias for these faces in non-anxious individuals. The results of the attentional biases for happy faces were less consistent. Nonetheless, these two versions seem to have similar sensitivity in measuring attentional biases for threatening facial expressions. However, the response times in the discrimination task were longer and more variable and resulted in higher error rates. This might especially be a disadvantage when testing clinical groups, children and non-human animals.
Both versions of the dot-probe task were also conducted in a study which examined the attentional bias for threat words (Salemink, van den Hout & Kindt, 2007). Participants performed both the detection and the discrimination task. Results showed a correlation, yet small, between anxiety and attentional bias in the detection task. Participants with higher anxiety scores showed higher attentional biases towards threat compared to individuals with lower anxiety scores. Such a correlation was not found in the discrimination task. Moreover, there was no significant correlation between the results of both versions of the task. This might indicate that both versions of the dot-probe task measure different constructs, at least when word stimuli are used. However, the differences in results might also be due to the low reliability of the dot-probe task when comparing individual performances (Schmukle, 2005; Staugaard, 2009; Zvielli et al., 2014).
These findings indicate that both versions of the dot-probe task are only partially comparable. Therefore, both versions might measure different attentional processes. The detection task might measure low level attentional processes, whereas the discrimination
task measures a construct where higher processes play a role. For research aiming at
automatic emotion perception the probe detection task might be the best version in order to prevent higher cognitive processes to intervene.
Emotional-neutral pair vs. emotional-emotional pair. It is common in dot-probe studies that an emotional valenced stimulus is presented next to a neutral stimulus. In this way is examined whether that specific emotional stimulus is attended or avoided. However, very few studies paired two emotional stimuli to investigate whether one type of emotional expression is attended or avoided more than another type.
To my knowledge, three studies included angry-happy face pairs in the task and analyzed the response times on these trials. Two studies did not find an attentional bias for either angry-neutral, happy-neutral or angry-happy face pairs (Heim-dreger et al., 2006; Pineles & Mineka, 2005). The participants reacted to the probe equally fast, irrespective of the presented face-pairs. In contrast, the third study found an attentional bias towards angry faces, both when they were paired with a neutral and a happy face (Tomaszczyk &
Fernandes, 2014). This result indicates vigilance for angry faces, irrespective of the other face in the face-pair.
Unfortunately, to date no study investigated attentional biases when face-pairs consisted of two negatively valenced faces. For instance, attentional biases can be measured when an angry facial expression is paired with a fearful facial expression. An attentional bias towards or away from one of these expressions might provide information about which emotional expression was more important for survival.
Context. Several studies suggested that context is an important parameter in assessing attentional biases. Particularly subjects with an anxiety disorder seem to be sensitive to the context in which the experiment takes place. When the experiment is
conducted under stressful circumstances anxious participants might alter their behaviour. It is hypothesized that the additive stress causes a shift in processing priorities such that the attention for the task is reduced. For example, the Stroop interference effect of threatening words disappeared when snake-fearful participants were exposed to a snake during the test (Mathews & Sebastian, 1993).
Dot-probes studies also showed that context can have an effect on attentional biases (e.g., Everaert, Spruyt & de Houwer, 2013; Judah, Grant, Lechner, & Mills, 2013). More importantly, context can have opposite effects in anxious and non-anxious participants. In one study the dot-probe task was adjusted such that a prime word was shown before an angry-neutral face-pair was presented (Helfinstein, White, Bar-Haim & Fox, 2008). This prime word could either be a neutral or a social threat word, determining the context of the following trial. Results showed that participants with high social anxiety scores showed an attentional bias towards angry faces when the trial was preceded by a neutral prime word. In contrast, this effect was not found when the trial was preceded by a threatening prime word. Participants with low social anxiety scores showed the opposite effect. They showed an attentional bias towards angry faces when preceded by a threatening prime word, yet there was no bias after a neutral prime word.
Such opposite effects were also found when social stress was induced before the dot-probe task took place (Mansell et al., 1999). High anxious participants avoided both positive and negative faces in the social threat condition, whereas low anxious participants did not show an attentional bias. No attentional biases were found in the non-threat condition, irrespective of anxiety level. These findings suggest that context can affect performances on the dot-probe task on an individual level. Therefore, caution is needed when testing different clinical groups. Anxious participants can feel stressed by the procedure of the task even
when this was not the original intention of the paradigm. This might result in biased performances of some groups of individuals.
Reliability
The discussion of all these studies made clear that the dot-probe task has led to inconsistent results. A various set of parameters could be responsible for this inconsistency. However, another explanation might be that the dot-probe task has insufficient reliability. Research indicated that the word-version of the dot-probe task is neither internally consistent nor stable over one week (Schmukle, 2005). Similar results were found for the visual dot-probe task which included both happy and angry facial stimuli (Staugaard, 2009) or stimuli of threatening scenes (Schmukle, 2005).
Thus, the dot-probe task seems not suitable for investigating individual differences. During a second test session participants know what sort of stimuli will be presented and might therefore apply a coping strategy. However, the dot-probe task might still be useful in making between-group comparisons (Staugaard, 2009). The calculated bias scores towards emotional faces were consistent both within and between sessions. Thus, on a group level the dot-probe task is a reliable measure of attentional biases for emotional information.
5. Conclusion
The dot-probe task is widely used to assess attentional biases as this task is implicit, does not need instruction and subjects need no or minimal training to perform the test successfully. In this essay was discussed whether the dot-probe task is a valid measure of attentional biases. Research showed inconsistent results, but it was difficult to compare the studies because of the variability in parameters and populations used for testing.
The personal characteristics of the participants influence the results of the dot-probe task. Participants’ anxiety level in particular is a confounding factor. Anxious individuals
showed vigilance for mild and high threatening stimuli, whereas non-anxious individuals only showed vigilance for high threatening stimuli (e.g., Wilson & MacLeod, 2003). Moreover, several parameters such as context and variations in stimulus presentation influence the performances of anxious and non-anxious individuals in different ways (e.g., Helfinstein et al., 2008; Mogg et al., 2004a; 2004b). In addition, gender and age of the participants might influence attentional biases. For instance, older individuals seem to avoid negative
information (e.g., Mather & Carstensen, 2003). Thus, when comparing results of dot-probe studies it is important that the subjects of the compared studies have similar personal characteristics.
The type of stimuli presented in the test also has an impact on the results. The level of threat is an important factor as mild and high threat evoke different attentional biases, at least in non-anxious individuals (e.g., Koster et al., 2005). Furthermore, responses to
different negative facial expressions should not be merged for analysis. Angry and fearful facial expressions initiate opposite action tendencies (Marsh et al., 2005) and might therefore also evoke different attentional biases. Additionally, facial and non-facial stimuli might have opposite effects on attentional biases. Lacreuse et al. (2013) showed that subjects had an attentional bias towards threatening facial stimuli, whereas negative non-facial stimuli were avoided. However, no other studies have compared non-facial and non-non-facial stimuli directly, which makes it difficult to compare both types of stimuli.
Last, changes in the experimental procedure of the dot-probe task might lead to differences in results. Most studies used a stimulus presentation of 500 ms. Shorter, subliminal, presentation does not seem to influence the results (Fox, 2002; Orgeta, 2011). Longer stimulus presentation seems to affect performances of anxious individuals who shifted from vigilance to avoidance of threat with longer stimulus presentation. Yet,
non-anxious individuals did not show different attentional biases (e.g., Ioannou et al., 2004; Lees et al., 2005). Thus, performances of the normal population are not affected by changes in stimulus presentation. Whether the dot-probe task applied probe detection or probe discrimination does not seem to significantly affect results (Bradley et al., 1998; Mogg & Bradley, 1999). However, probe discrimination leads to more variable response times and might therefore be less sufficient to test children and non-human primates. A last point of importance is the context in which the dot-probe task is conducted. For example, the induction of an affective state alters performances (Everaert et al., 2013; Helfinstein et al., 2008). Comparing the results of dot-probe studies is only possible when the attentional biases were measured in the same context.
Despite the inconsistency in results of dot-probe studies there is an indication that emotional expressions are processed effectively in the normal population. Especially high threatening stimuli such as angry faces seem to capture attention. This suggests that
emotional facials expressions might have been adaptive signals. The rapid allocation towards threatening stimuli initiated action tendencies to effectively cope with the threat. Yet,
stronger support for the hypothesis that emotional facial expressions evolved as a signal of threat must be provided by comparative studies. A direction for further research might therefore be to conduct dot-probe studies with non-human primates and other animals. The comparison of attentional biases in several related species might provide essential
information about the evolution of emotional facial expressions. If individuals of several species attend to emotional stimuli in similar ways the underlying emotional circuits are probably also similar, which then indicates an evolutionary basis for the perception of emotional signals.
The dot-probe task is specifically suitable for comparative studies as the task can be conducted with non-human primates (King et al., 2012; Parr et al., 2013) and potentially also with other animals. Moreover, rhesus monkeys perform similar to humans when facial stimuli are used (Lacreuse et al., 2013). A good comparison between the performances of humans and non-human primates on the dot-probe task is possible when certain points are taken into account. Both groups should be matched on gender and relative age. In addition, emotional facial expressions which evoke similar levels of arousal in human and non-human primates should be used as stimuli. And last, probe detection instead of probe discrimination should be used to prevent problematic variability in reaction times. Conducting studies which enable reliable comparisons of emotional processing in humans and non-human primates might eventually provide essential information about the evolution of emotional signals.
References
Adams, R. B., Gordon, H. L., Baird, A. A., Ambady, N., & Kleck, R. E. (2003). Effects of gaze on amygdala sensitivity to anger and fear faces. Science, 300(5625), 1536-1536.
Adams, R. B., & Kleck, R. E. (2003). Perceived gaze direction and the processing of facial displays of emotion. Psychological Science, 14(6), 644-647.
Armony, J. L., & Dolan, R. J. (2002). Modulation of spatial attention by fear-conditioned stimuli: An event-related fMRI study. Neuropsychologica, 40, 817-826.
Arndt, J. E., & Fujiwara, E. (2012). Attentional bias towards angry faces in trait-reappraisal. Personality and Individual Differences, 52(1), 61-66.
Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., & Van IJzendoorn, M. H. (2007). Threat-related attentional bias in anxious and nonanxious individuals: a meta-analytic study. Psychological bulletin, 133(1), 1-24.
Becker, D. V., Anderson, U. S., Mortensen, C. R., Neufeld, S. L., & Neel, R. (2011). The face in the crowd effect unconfounded: happy faces, not angry faces, are more efficiently detected in single-and multiple-target visual search tasks. Journal of Experimental Psychology: General, 140(4), 637-659.
Bradley, B. P., Mogg, K., Falla, S. J., & Hamilton, L. R. (1998). Attentional bias for threatening facial expressions in anxiety: Manipulation of stimulus duration. Cognition & Emotion, 12(6), 737-753.
Bradley, B. P., Mogg, K., & Millar, N. H. (2000). Covert and overt orienting of attention to emotional faces in anxiety. Cognition & Emotion, 14(6), 789-808.
Bradley, B. P., Mogg, K., Millar, N., Bonham-Carter, C., Fergusson, E. & Jenkins, J. (1997). Attentional biases for emotional faces. Cognition & Emotion, 11, 25-42.
Brown, H. M., McAdams, T. A., Lester, K. J., Goodman, R., Clark, D. M., & Eley, T. C. (2013). Attentional threat avoidance and familial risk are independently associated with childhood anxiety disorders. Journal of Child Psychology and Psychiatry, 54(6), 678-685.
Carlson, J. M., Fee, A. L., & Reinke, K. S. (2009). Backward masked snakes and guns modulate spatial attention. Evolutionary Psychology, 7(4), 534-544.
Carlson, J. M., Mujica-Parodi, L. R., Harmon-Jones, E., & Hajcak, G. (2012). The orienting of spatial attention to backward masked fearful faces is associated with variation in the serotonin transporter gene. Emotion, 12(2), 203-207.
Carlson, J. M., Reinke, K. S., & Habib, R. (2009). A left amygdala mediated network for rapid orienting to masked fearful faces. Neuropsychologia, 47(5), 1386-1389.
Carroll, J. M. & Russell, J. A. (1996). Do facial expressions signal specific emotions? Judging emotion form the face in context. Journal of Personality and Social Psychology, 70, 205-218.
Cartwright, J. H. (2008). Evolution and human behaviour: Darwinian perspectives on human nature (2nd ed.). The MIT Press.
Chan, M. W., Ho, S. M., Law, L. S., & Pau, B. K. (2013). A visual dot-probe task as a measurement of attentional bias and its relationship with the symptoms of
posttraumatic stress disorder among women with breast cancer. Advances in Cancer: Research & Treatment, Vol. 2013, Article ID 813339.
Chen, M., & Bargh, J.A. (1999). Consequences of automatic evaluation: Immediate
behavioral predispositions to approach or avoid the stimulus. Personality and Social Psychology Bulletin, 25, 215-224.
