Snakes seem to evoke a number of specific responses in humans, including a high fear reaction, preferential attention, and fast detection (Öhman and Mineka, 2003; Okon-Singer et al., 2011; Soares et al., 2014; Kawai and Qiu, 2020; Landová et al., 2020; Jensen and Caine, 2021). Each of these phenomena has been previously studied in detail and is described in its own terms and hypotheses. Taken altogether, however, the human mind seems to be specifically equipped to react to snakes in a certain manner (Isbell, 2006, 2009). Further, there are several compelling pieces of evidence that this reaction is at least partially innate (Tierney and Connolly, 2013; Kawai, 2019). First, it is shared amongst distinct cultures across the globe (Alves et al., 2014; Pandey et al., 2016; Landová et al., 2018; Onyishi et al., 2021), second, it can manifest itself very early in human ontogeny (Lobue and DeLoache, 2008; DeLoache and LoBue, 2009; Hayakawa et al., 2011; Borgi and Cirulli, 2015; Bertels et al., 2020), and third, we can observe a similar reaction in apes and other primates (Murray and King, 1973; Shibasaki and Kawai, 2009; Weiss et al., 2015; Kawai and Koda, 2016; Wombolt and Caine, 2016; Zhang et al., 2020). This body of evidence serves as the basis of what is now known as the Snake detection theory (Isbell, 2006 and references therein).

In the last two decades, the research of visual attention mainly focused on whether humans are able to detect a snake faster or more accurately than different types of stimuli in a challenging setup. In a visual detection task, snakes were detected faster than other animals (Lobue and DeLoache, 2008; Shibasaki and Kawai, 2011; Penkunas and Coss, 2013; Soares and Esteves, 2013), even under high perceptual load (Soares et al., 2014; Kawai and He, 2016; Kawai and Qiu, 2020) and regardless of their colouration (Prokop et al., 2018; Fančovičová et al., 2020). Similar results were obtained from a simulated virtual hike (Jensen and Caine, 2021), whilst LoBue (2014) and Kawai (2019) investigated which features facilitated the detection. Additional support for the Snake detection theory also comes from event-related potentials (ERP) studies (He et al., 2014; Grassini et al., 2016; Van Strien et al., 2016; Van Strien and Isbell, 2017) or neurobiological research (Van Le et al., 2013, 2014). Nonetheless, the issue seems more complex as, for example, guns (i.e., evolutionarily irrelevant inanimate objects) are detected as fast or even faster than snakes (Fox et al., 2007; Zsido et al., 2019a,b). Taken all together, there is a strong experimental support for primate brains being fine-tuned for snake detection, however, it seems that not all snakes are prioritised (see, e.g., Rádlová et al., 2019) and not under all circumstances (Subra et al., 2018; reviewed in Coelho et al., 2019).

Importantly, all these experiments assume – although sometimes inexplicitly – that humans and other primates pick on certain visual cues for snake detection provided by the snake’s appearance. It is worth pointing out that the snake does not provide these cues with the purpose of being detected; the primates rather take advantage of the cues the snake cannot conceal. In these scenarios, snakes are thought of as predators (Seligman, 1971; Öhman and Mineka, 2001; Isbell, 2006) and as such, they would not profit from being discovered. However, adult primates are rarely snake prey probably thanks to high vigilance, warning calls, and aggressive group defence (Seyfarth et al., 1980; Perry et al., 2003; Eberle and Kappeler, 2008; Etting et al., 2014; Teixeira et al., 2016). In fact, the roles might even reverse, and a snake might end up the prey itself (Headland and Greene, 2011; Falótico et al., 2018). What may have started as a clear predator–prey dynamics in evolutionary history, could now be seen rather as an equal-opponents situation.

Under these circumstances, it might be advantageous to signal one’s readiness to fight towards the opponent. This type of signalling is called a threat display. It is almost omnipresent in animals although it may take different forms in different species (e.g., chimpanzees – Nishida et al., 1999; frillneck lizards – Shine, 1990; cuttlefish – Langridge et al., 2007; pelicans – Gokula, 2011; tarantulas – Bennie et al., 2011). In snakes, the most famous example is the threatening posture of cobras – the animal puffs, its body front rises, and its neck-flack spreads (Greene, 1988). Another example comes from vipers – the animal puffs and coils its body in very tight loops with an elevated head held slightly back as if ready to strike (Greene, 1988). Both postures are quite conspicuous, and the animal often accompanies its display with hissing, which further facilitates its detection. The display is clearly intended to be seen by the opponent.

In this paper, we follow the line of thought previously introduced by Isbell’s Snake detection theory. We aim to explore whether the human mind is also fine-tuned for a snake’s intentional threat signalling rather than just unintentional cues of its presence. To this end, we employ an eye-tracking method utilising a simple design of spontaneous gaze preference when presented with two stimuli at once. We hypothesise that snakes in threatening postures will attract more attention than snakes in relaxed, non-threatening postures. In the past, it was demonstrated that emotions can guide visual attention (Vuilleumier, 2005), and, in particular, that fear-relevant animals are fixated faster, more often, or for longer time periods than fear-irrelevant animal targets (Öhman et al., 2001; Gerdes et al., 2009; LoBue, 2014). Moreover, the importance of snake posture for assessment of danger was previously shown in macaques (Etting and Isbell, 2014; Van Le et al., 2014), and humans (Masataka et al., 2010; Lobue and DeLoache, 2011).

To highlight the ecological aspect of our hypothesis, we focused on Somalis (specifically the population living in Somaliland). Whilst the culture is traditionally pastoral and therefore mobile, according to genetic and linguistic evidence they belong to the core populations of North and Northeast Africa, which have never left the African continent or the savanna environment. Somalis are thus characterised by the near-continuous presence in both the geographic region and the environment of human origin (Stringer, 2016; Gibbons, 2017). Moreover, evidence suggests that the snake species composition of the Horn of Africa has remained largely unchanged during the principal part of human evolution (Kelly et al., 2009; Barlow et al., 2019; Šmíd and Tolley, 2019; Zaher et al., 2019), and we previously found that Somalis consider snakes the most fear-eliciting animals amongst a wide variety of species (Frynta et al., 2023). This makes Somalis uniquely suited for research focusing on the possible co-evolution of snake signalling and human signal detection. In addition, we included participants from Czechia whose ancestors left Africa and, similarly to other Europeans, reached Europe about 30,000 years ago (Prüfer et al., 2021). As there have been virtually no dangerous snakes in Central Europe over the last 40,000 years (only mildly dangerous adder Vipera berus), Czechs seem a suitable match to Somalis for cross-cultural comparison. Similar responses across the participants, despite thousands of years of differential exposure to snakes, would suggest that the reaction is at least partially innate and a result of long coevolution between humans and snakes. Contrary, if cultural or more recent selection pressures are involved, clear differences between Somali and Czech participants should emerge. We know of no psychological study focusing on snakes and simultaneously utilising an eye-tracking experimental design in Sub-Saharan Africa.

To summarise, the aims of this study are as follows: (1) To test whether a snake in a threatening posture attracts more attention than one in a relaxed, non-threatening posture, (2) To investigate whether such phenomenon is universal or whether it is specific for certain snake morphotypes as those differ in their threatening postures, and (3) To compare the attention paid to snakes by Somalis and Czechs.

In the linear mixed effect model for the difference in the number of fixations on threatening versus relaxed posture, only one factor proved significant: snake morphotype (F_(2,2,593) = 24.42, p < 0.001). Other factors were successively taken out of the model since their effect did not prove significant (nationality: F_(1,138) = 0.05, p = 0.830; gender: F_(1,138) = 1.54; p = 0.217; age: F_(1,138) = 0.20, p = 0.652; snake morphotype – nationality interaction: F_(2,2,591) = 2.26, p = 0.078; nationality – gender interaction: F_(1,138) = 0.05, p = 0.831). The odds ratio (assessed with likelihood-ratio test) between the full and reduced model was 7.00, p = 0.321. For the estimated differences, the following logic can be applied. When the difference is positive, participants looked more at the threatening posture, when negative, participants looked more at the relaxed posture, zero difference signifies equal attention on both. To this end, we tested each difference against zero. Cobras in threatening posture captured 1.85 more fixations than cobras in relaxed posture (95% confidence interval 1.45–2.25), this difference was significantly different from zero (t₍₂₅₉₃₎ = 9.12, p < 0.001). Similarly, vipers in threatening posture captured 0.46 more fixations than vipers in relaxed posture (95% confidence interval 0.06–0.85). Whilst the difference was smaller than the one for cobras, it was also significantly different from zero (t₍₂₅₉₃₎ = 2.28, p = 0.023). Lastly, other snake morphotypes captured 0.28 more fixations in threatening posture than in relaxed posture (95% confidence interval − 0.23 – 0.79), however, this difference was not significantly different from zero (t₍₂₅₉₃₎ = 1.09, p = 0.276). For context, the average number of fixations per trial across all stimuli and participants was 18.07 for the left and right IAs combined. The results are shown in Figure 2.

There was no difference in the goodness of fit of the full and reduced model for the difference in dwell time (p = 0.300), hence we again chose the reduced model as the final model based on the AIC. The odds ratio between the models was 3.66. The final model for the difference of dwell time on threatening versus relaxed posture contained snake morphotype (F_(2,2,591) = 12.57, p < 0.001), nationality (F_(1,141) = 0.94, p = 0.333), and their interaction (F_(2,2,591) = 4.33, p = 0.013). Other factors were successively taken out of the model (gender: F_(1,138) = 3.15; p = 0.078; age: F_(1,138) = 0.50, p = 0.504; nationality – gender interaction: F_(1,138) < 0.01, p = 0.975). The results showed that both Somalis and Czechs gazed at cobras in a threatening posture longer than at cobras in a relaxed posture. In viper and other snake morphotype stimuli, only Czechs dwelled on the snakes in threatening posture longer; Somalis divided their attention equally. For more details, see Table 1 and Figure 3. The average dwell time per trial across all stimuli and participants was 4,368 ms for the left and right IAs combined. For average dwell times per trial of each stimulus species, see Supplementary Table S2.

To summarise the results, we found that snakes in a threatening posture attracted more attention than those in a relaxed, non-threatening posture which was manifested both in the number of fixations and the total dwell time. This was especially prominent in cobras, however, vipers in a threatening posture were also fixated more often than vipers in a non-threatening posture. Additionally, differences in dwell time revealed that Czechs dwelled on snakes in threatening posture longer no matter the snake morphotype. Contrarily, in Somalis, this held true only for the cobra morphotype; in the other two investigated morphotype groups (vipers and others) there was no difference in the attention paid to the snakes in threatening and non-threatening postures. Nonetheless, the only significant difference between Somalis and Czechs was in their reaction to vipers and specifically only in the dwell time, not the fixation count. Hence, we consider the overall pattern of reaction towards snake threatening postures relatively consistent across nationalities. We found no effect of gender.

In the last couple of years, researchers paid special attention to what features are crucial for the recognition of a snake. An early candidate in question was a curvilinear body shape. Whilst this feature can play an important role (Wolfe et al., 1992; LoBue, 2014), alone, it is most probably not sufficient as worms (of a similar curvilinear body shape) evoked a smaller reaction in an ERP study (Van Strien et al., 2016). Next, the effect of body posture was investigated. Lobue and DeLoache (2011) suggested that a snake’s coiled body shape is responsible for the faster detection of snakes in relation to distractors. However, Masataka et al. (2010) argued that the striking (vs. relaxed) body posture rather enhanced the speed of detection but was not its basis per se. This notion was supported by Etting and Isbell (2014) in their rhesus macaque behavioural study. Recently, the topic of recognition-relevant features of snakes was reviewed by Kawai (2019). In concordance with the most current results, he concluded that snake scales are the key characteristic (He et al., 2014; Van Strien et al., 2016; Isbell and Etting, 2017; Van Strien and Isbell, 2017; Kawai, 2019). Interestingly, the importance of snake scales could be illustrated already in hominids, as some palaeolithic engravings likely represent a snakeskin (Coss and Charles, 2021). Although we agree with the studies on the importance of snake scales, this feature cannot explain the results of this study since it is the key for snake identification only in the context of other animals. Therefore, we propose to take a step back and look at the bigger picture once more.

Above, we suggested that the primate visual system is not only adapted for faster detection of snakes (Isbell, 2006, 2009) but also fine-tuned for the snake’s threat display. Notably, some snake-typical features which were previously found to enhance the speed of snake detection seem to be exaggerated in snakes’ threat display. In the case of vipers, such feature is the coiled shape (Lobue and DeLoache, 2011; Etting and Isbell, 2014), which is very tight under the threat creating many loops, whilst in cobras, the risen body front exaggerates the snake’s curvilinear body shape (Wolfe et al., 1992; LoBue, 2014). These exaggerated features typical for threat display might be behind the larger spontaneous attention paid to snakes in threatening postures. Moreover, the erect posture (seen in threatening cobras) is very conspicuous, and many animal species take advantage of it when wanting to intimidate an opponent. Indeed, people perceive animals in an erect posture as more fear-eliciting (Prokop et al., 2021). This, however, opens a question of whether the attentional privilege of snakes in threatening posture is driven by visual or affective features. It has been previously shown that emotions can modulate attention towards a stimulus. Soares et al. (2009) reported that participants found the animal they were afraid of faster than non-feared but fear-relevant animals. Similar results were also found by Miltner et al. (2004) or Lipp and Waters (2007). The arousal might also affect visual attentional performance (reviewed in Zsido, 2023). For instance, Zsido et al. (2022) showed that stimulus arousal might be an important cue facilitating target recognition in a memory test. As for now, however, we do not have enough data to decide the extant visual and affective features of snakes in threatening posture modulate attention and it might even be that these two types of features are inseparable in nature.

Although the overall pattern of reaction towards snakes was similar for both Somalis and Czechs, we would also like to shortly address the found difference. Czechs dwelled longer at the snake in threatening posture no matter the morphotype, but the same was true only for cobras in Somalis. Could it be that Somalis differentiate (in a perceptual sense) between threatening and non-threatening postures only in cobras simply because the difference is the most conspicuous? This explanation would not fit well into the evolutionary framework and indeed, it is not the case. In a complementary study, individual photos of snakes very similar to photos used in this experiment were presented to Somalis and they were asked to order them according to elicited fear from the most to the least fear-eliciting (Frynta et al., in prep.). In this forced-choice experiment, six out of eight viper species were rated as significantly more fear-eliciting when in the threatening posture than in the non-threatening posture (Frynta et al., in prep.). We instead suggest that Somalis attribute the same level of threat to vipers no matter their body posture leading to the same observed dwell time. LoBue (2014) previously showed that knowledge or expectations can interact with low-level features of the stimuli in visual search tasks. In our case, the (communal) knowledge of the nature of cobra and viper attacks seems to be the key. Cobras are active in their defence; they either flee or display to the opponent and strike only afterwards. Vipers, on the other hand, are passive; they often rely on their cryptic coloration and motionless stance and strike when the opponent (usually unknowingly) comes too close. Even though the majority of Somali participants were university students now living in a city, most of them came from rural areas and pastoral families. We find it very likely that they personally knew someone that was bitten by a snake. When we asked local villagers, at least one person was willing to share their experiences in every village. No communal knowledge can be expected in Czechs since local snake fauna is not dangerous to humans and, moreover, participants were mostly from urban areas. This finding might be of importance for future studies since it illustrates that not all snakes are the same (see also Landová et al., 2018, 2020; Janovcová et al., 2019; Rádlová et al., 2019; Frynta et al., this issue).

To conclude, our results show that human attention is directed more towards cobras and also vipers in threatening postures. We hypothesise that it is a result of primate-snake coevolution during which not only snakes represented a danger to primates but also primates represented a danger to snakes. To be clear, we do not argue that snakes evolved specific threatening postures in response to predator pressure from primates. That is very unlikely because of multiple reasons including primates are not snake’s primary predators, and the threatening postures are not addressed uniquely to primates but to a variety of other potential predators, e.g., mongooses (Herpestidae) or birds (the secretarybird Sagittarius serpentarius, crested seriema Cariama cristata, or others). We simply argue that at some point in evolutionary history, the primate-snake relationship must have become less one-sided since this is the situation we witness today (Headland and Greene, 2011; Falótico et al., 2018; see also Harris et al., 2021). As a part of their defensive behaviour, snakes would be displaying threat and these threat signals would in turn become associated with intense danger leading to prioritised attention toward threatening displays. The features of the threat display could not be reliably used for the detection of snakes because the threatening posture is a defensive behaviour and hence is not displayed when hunting for prey. Nonetheless, prioritised attention is still adaptive as the threat display signals the immediate danger of a ready-to-strike snake. Although the prioritised attention towards snakes was previously tested in several different cultures, this is the first study investigating a population from Sub-Saharan Africa – a key region with regards to the evolutionary Snake detection theory.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

The studies involving humans were approved by The Institutional Review Board of Charles University, Faculty of Science and The Institutional Review Board of Amoud University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

DF and EL: conceptualization. IŠ: data curation. IŠ and DF: formal analysis, writing – original draft, and writing – review & editing. DF: funding acquisition and supervision. IŠ, MJ, VR, HE, KR, DK, DS, DB, PF, and DF: investigation. All authors contributed to the article and approved the submitted version.