Humans are exposed to a plethora of concurrent visual stimuli and attention is preferentially directed to the most menacing of them. This is thought to be adaptive because quick detection of danger promotes survival by initiating necessary behavioral responses (Öhman et al., 2001). The selective allocation of attention toward a particular type of stimulus is referred to as attentional bias. Among more fearful individuals, this bias tends to be particularly strong (Öhman and Mineka, 2001; Mineka and Öhman, 2002). Furthermore, attentional biases are generally thought to be specific for particular fears: For example, specific phobias are related to pictures of the feared animals, social anxiety is related to negative (or neutral) emotional facial expressions (Cisler and Koster, 2010).

Among all the visual stimuli that are encountered, faces are especially relevant social cues for us (Alpers and Gerdes, 2007; Bublatzky and Alpers, 2017). Accordingly, humans are equipped with distributed neural systems that specialize in face processing (Haxby et al., 2000). Substantial evidence shows that faces are processed more efficiently than other neutral stimuli (Farah et al., 1998; Öhman et al., 2001) by facilitating perceptual and attentional processing (Alpers and Gerdes, 2007; Alpers et al., 2011; Bublatzky et al., 2014). Similarly specialized neural systems are likely involved in the preferential processing of other kinds of evolutionary significant stimuli. It has thus been suggested that humans are equipped with a preparedness for specific animals (i.e., spiders, snakes) that may be particularly relevant for our survival (Seligman, 1971). Indeed, compelling evidence indicates that fear-related animals are preferentially processed (Macleod and Mathews, 1988), which manifests itself in eye-tracking research as longer fixation durations or as greater skin conductance responses compared to neutral stimuli (Gerdes et al., 2008, 2009a; Wiemer et al., 2013). Still, the stimuli used in such studies are typically very diverse, e.g., potentially threatening animals like snakes, spiders, insects (Lipp, 2006; Rinck and Becker, 2006; Gerdes et al., 2008, 2009b; Weymar et al., 2013; Berdica et al., 2014, 2017).

Although different kinds of threatening stimuli (fear-related animals and social stimuli) evoke comparable reactions compared to neutral stimuli, this does not mean that all threatening stimuli are the same. For example, in an earlier study we demonstrated that pictures of emotional scenes and pictures of faces can result in very different psychophysiological response patterns (Alpers et al., 2011). Moreover, data from event-related potentials in another study (Kolassa et al., 2008) show enhanced attention to emotional face stimuli in socially anxious individuals regardless of facial expression; this was not specific to social phobia but was also observed in those participants who were afraid of spiders. No abnormalities were found in the processing of angry faces in particular (Kolassa et al., 2008). The authors also found that spider phobic individuals were generally more hypervigilant toward any kind of stimuli, suggesting a generalization of anxious behavior in spider phobic individuals. This finding is supported by previous research (Weymar et al., 2013) which showed an enhanced hypervigilance for all of the stimuli that they presented (both spiders and butterflies) in spider fearful participants.

Similar response patterns to different evolutionary fear-relevant stimuli are noticeable in infants younger than 9 months (Erlich et al., 2013). In this study, human voices and animal sounds (hissing snakes) were considered as part of the evolutionary fear-relevant category and they elicited similar reactions. At the same time, fear responses to threatening animal stimuli or to emotional facial expressions can differ in many ways, in terms of physiological or behavioral responses, fear learning, or fear conditioning. For example, in a recent review it was shown that fear conditioning to social stimuli (angry faces) was shown to be less robust as fear conditioning to snake and spider stimuli. The authors proposed that distinct mechanisms may underlie fear learning for social and animal stimuli (Mallan et al., 2013). In terms of brain activation, in a recent EEG study Langeslag and van Strien (2018) found that snakes capture attention faster than angry faces, manifested in more pronounced EPN for snakes compared to angry faces. In this study, pictures were presented separately in rapid serial visual presentation therefore no conclusions about initial orienting of attention when both pictures are presented side by side can be reached.

Hence, there is considerable evidence that both, pictures of threatening animals and of emotional facial expressions trigger fear responses, avoidant behavior, they differ in fear learning and in fear conditioning. Öhman (2009) postulate a predatory defense system related to evolutionary prepared animals like snakes, and a social submissiveness system corresponding to social fears. They support the hypothesis that these sorts of stimuli would elicit fear after minimal processing times (Öhman, 2009). There is, however, a lack of studies in which attentional processing of threatening animals and faces are directly compared. Faces have, nonetheless been compared with pictures of other emotional scenes before (Alpers et al., 2011; Eisenbarth et al., 2011), but not specifically with threatening animals like snakes or spiders.

Furthermore, no studies so far have addressed the specificity of attentional biases – is there a uniform bias to emotionally relevant pictures or is there a category specific bias? Are specific categories more dominant than others? Until now, the question how these types of stimuli compete for visual attention and whether specific fear (e.g., spider vs. face) can differentially modulate the underlying mechanisms of attentional biases remains still unanswered. In order to answer these questions, the main purpose of the present study was to examine whether angry faces and spiders elicit similar attentional biases (initial vigilance; difficulties to disengage).

Another important aspect we deal with in the present research is the time course of attention allocation to fear-related stimuli. This has also been tested in one or the other stimulus type (Animals: Lipp, 2006; Rinck and Becker, 2006; Gerdes et al., 2008, Faces: Mogg et al., 2004) but never for both types placed next to each other in one single experiment.

There is considerable evidence to support the notion that attentional biases are comprised of an initial hypervigilance for threat (Mogg and Bradley, 1998; Öhman, 2009). How aversive stimuli are subsequently processed — whether these stimuli are avoided or attended more — is less clear. Some researchers in the field suggest that an initial orienting toward threat is followed by a difficulty in disengaging from threat (Fox et al., 2001, 2002; Gerdes et al., 2008, 2009a). In contrast, others suggest that initial hypervigilance is followed by avoidant behavior (Mathews and Mackintosh, 1998; Mogg and Bradley, 1998; Wieser et al., 2009) making it challenging to reach conclusions on sustained attention on threatening stimuli.

Face stimuli are frequently used to investigate the aforementioned attentional processes. For example, in an eye-tracking study, Holas et al. (2014) found an initial vigilance for angry and happy faces, but at the same time inconsistent results for the sustained attention. Likewise, in another study, initial hypervigilance for angry faces was found but no subsequent bias in later processing (Gamble and Rapee, 2010). Furthermore, a recent study found no initial orienting bias in socially anxious individuals toward threat but a difficulty to disengage attention from threatening stimuli once their initial attention is oriented toward those stimuli (Liang et al., 2017).

Given the contradictory models for the sustained attention mentioned above, as well as the limited range of examined stimuli used in previous experiments, in the present study we directly compared faces and animals in a typical free viewing paradigm to identify attentional biases (in anxiety). Pairs of emotional (spiders, angry faces) and neutral (butterflies, neutral faces) visual stimuli were presented in one of several combinations: pictures of spiders and butterflies, angry and neutral faces, spiders and angry faces, butterflies, and neutral faces.

Taken together, the main purpose of our study was to examine whether spiders and angry faces elicit similar attentional biases (initial vigilance; difficulties to disengage) when directly compete for attentional resources. In order to achieve this, we use eye-tracking as an advantageous measure to investigate these attentional biases (Armstrong and Olatunji, 2012). We further aimed at replicating the preference toward threat hypothesis: spiders and angry faces should generally be detected faster than butterflies and neutral faces. Based on the findings reported above, spiders, and angry faces were not expected to differ in initial orienting of attention when pitted against each other but faces were expected to sustain attention longer compared to animal stimuli for all participants independent from specific fear. This pattern was expected to be enhanced for socially anxious individuals.

Attentional biases have been documented for fear-related animals (such as spiders) and emotional facial expressions (such as angry faces) but they have rarely been pitted against each other in one experiment. This is the first eye-tracking study to directly compare how threatening animal stimuli and social stimuli compete for visual attention. We recorded eye-movements during a free viewing task with pictures from different categories (animals and human faces), displaying emotional (spiders, angry faces) or neutral (butterflies, non-emotional faces) information. Our findings support an attentional preference toward threat, thus, this replicates previous findings in a more complex context. Overall, fear-related stimuli captured initial attention more than neutral pictures. As for the comparison between categories (animal vs. faces), there were no obvious differences in the initial allocation of attention or in sustained attention.

Eye-tracking methodology provides a useful tool to investigate attentional biases (see Armstrong and Olatunji, 2012); it does, in addition measure different attentional processes within different time frames: initial hypervigilance in early stages of processing and sustained attention in later stages. For this reason, we focused our analysis on two parameters: Entry time as an indicator of initial hypervigilance and fixation count as an indicator of sustained attention. The first parameter taps into the early component of the hypervigilance-avoidance hypothesis (Mogg and Bradley, 2006). Our results support the notion that unpleasant stimuli capture attention faster than neutral stimuli—threatening pictures were generally looked at faster than neutral pictures. In addition to this, our design also allows for a direct comparison of entry times in a single experimental design. This is in line with brain imaging data that show that negative emotional stimuli are preferentially processed in the brain (Alpers et al., 2009; Gerdes et al., 2010).

In addition, typical left bias was also observed in our study, which may represent the common reading direction. Pictures on the left were detected faster than the ones on the right, independent of stimulus type. A left bias for processing emotional information has been found before (Alpers, 2008) and it would have been expected but it was not supported from the current data.

Regarding sustained attention, fixation count data do not differ for spiders or faces. More fixations were observed for face stimuli compared to spider pictures in Block 3 but this did not reach significance. Similar to other studies (Calvo and Lang, 2004), we also did not find an emotion effect regarding fixation count — fear-related pictures did not hold attention longer than neutral stimuli. Furthermore, a left bias was also absent for the sustained attention parameter. It should be noted that a difficulty to disengage from threat (or sustained attention) may be visible only during visual search tasks but not during free viewing tasks (Armstrong and Olatunji, 2012). Therefore, task relevance should be investigated further in future studies.

Regarding the neutral category, butterflies were rated as more positive than neutral faces which can also be seen in the normative ratings of the few pictures depicting butterflies in the IAPS collection (Lang et al., 1999). One reason for this result might be the fact that butterflies are more unambiguous pictures while neutral faces sometimes appear ambiguous and tend to sometimes be regarded as negative (Melfsen and Florin, 2002; Blasi et al., 2009). Future research should investigate the extent to which attentional processes are similar for different types of neutral stimuli.

One reason for the differences in eye movements in participants with fear of spiders and social anxiety could be that social anxiety is generally considered to be more complex than specific phobias. Attentional biases in general in fear of spiders can be explained in terms of preparedness while social anxiety can be seen as a result of perceived dominance (Öhman, 2009). Interestingly, animal fears are relevant between species, while social phobia within-species. Moreover, while specific phobias are centered on a specified object, social anxiety comprises, among everything else, fear of criticism or disapproval (Hofmann et al., 2009). A large body of research documents attentional biases in individuals with specific animal fear (Gerdes et al., 2008, 2009a) and social anxiety (Amir et al., 2003; Mogg et al., 2004; Wieser et al., 2009). These studies generally differ in terms of initial hypervigilance, difficulties in disengagement, or avoidance that they find in the different kinds of fears. This, therefore, raises the question whether one single theoretical account can accommodate research findings on attentional biases in specific phobias and social anxiety.

Our results support the above-mentioned notion in that fear of spiders correlates with the observed initial hypervigilance but this is not the case for social anxiety. This suggests that distinct mechanisms may be underlying the two fears. It will be fruitful for the understanding of both fear domains to disentangle the specific mechanisms in future research. One plausible reason for the differences may in part be due to the intensity of reported fear in our sample. When each one of our groups is compared to normative controls, spider fearful participants report higher levels of spider fear compared to socially anxious participants who score lower on social anxiety (For spider fear: Rinck et al., 2002; for social anxiety: Sosic et al., 2008). One plausible explanation for this is that more socially anxious participants may have avoided participation in the experiment due to their fear. Even though social anxiety is highly prevalent in the population, individuals experiencing this fear are less expected to join experiments compared to spider fearful individuals.

Taken together, the present research is one of the few examples of a direct comparison of different kinds of emotional stimuli. Furthermore, it sheds light on how different categories of threat cues compete for visual attention and how attentional biases are manifested in the context of anxiety. The results suggest that both spiders and angry human faces preferentially guide attentional processes and they seem to be ingrained in our defense system as evolutionary fear-relevant stimuli.

EB contributed to the conceptualization and design of the work, data collection, data analysis, writing of the manuscript, and final approval. AG contributed to the conceptualization, programming, design, data analysis, revisions, and final approval. FB contributed to the conceptualization, data analysis, revisions, and final approval. AW contributed to the conceptualization, data analysis, and final approval. GA contributed to the conceptualization, funding, data analysis, continuous revisions, and final approval.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer ADC and handling Editor declared their shared affiliation.