Media representation of spiders may exacerbate arachnophobic sentiments by framing a distorted perception of risk

Spiders are able to arouse strong emotional reactions in humans. While spider bites are statistically rare events, our perception is skewed towards the potential harm spiders can cause to humans. We examined the human dimension of spiders through the lens of traditional media, by analysing more than 300 spider-related news published online in Italian newspapers between 2010 and 2020. We observed a recent, exponential increase in the frequency of the news, particularly those focused on medically important spiders – the Mediterranean black widow (Latrodectus tredecimguttatus) and the Mediterranean recluse (Loxosceles refescens). The news quality was generally poor: 70% contained different types of error, 32% were exaggerated, and in virtually none was an expert consulted. Overstated news referring to spider bites were significantly more shared on social media, thus contributing to frame a distorted perception of the risk associated with a spider bite and possibly reducing general public tolerance of spiders.


INTRODUCTION
Wildlife is an important emotional trigger in humans (Jacobs, 2009(Jacobs, , 2012Hicks & Stewart, 2018). Admiration and respect, surprise and excitement, transcendent feelings, but also fear and disgust are just a few examples illustrating the spectrum of emotions reported by people experiencing encounters with wildlife (Hicks & Stewart, 2018). An interesting aspect of the human dimension of wildlife is that sensitivity toward animals is largely conserved in most contemporary societies, even though wildlife no longer plays a central role in our every-day lives (Franklin & White, 2001). Studies suggest that emotional feelings toward wildlife are, indeed, in-born (Strommen, 1995;Davey et al., 1998;Prokop & Tunnicliffe, 2008;DeLoache, Pickard, & LoBue, 2010), often recurring with striking similarities across diverse cultural settings (Davey et al., 1998). As a direct consequence, animals-related emotions end up playing a key role in scientific and socio-political debates around both the management and conservation of wildlife (Jones, 2006;Singh, 2009;Frank, Johansson, & Flykt, 2015;Zainal Abidin & Jacobs, 2019;Drijfhout, Kendal, & Green, 2020;Straka, Miller, & Jacobs, 2020), and in the perception of risk (Knopff, Knopff, & St. Clair, 2016;Hathaway et al., 2017;Bombieri et al., 2018;Nanni et al., 2020).
Spiders are iconic examples of animals that can bring about strong emotional reactions in humans (Michalski & Michalski, 2010;Lemelin & Yen, 2015;Hauke & Herzig, 2017;Mammola, Michalik, Hebets, & Isaia, 2017), leading to a distorted perception of risk, especially when referring to spider bites. While less than 0.5% of spider species are capable of causing severe envenomation in humans (Hauke & Herzig, 2017), and no proven fatality due to spider bites had occurred in the past few decades (Nentwig & Kuhn-Nentwig, 2013;Nentwig, Gnädinger, Fuchs, & Ceschi, 2013;Stuber & Nentwig, 2016), the perception of the risk associated with spider bites remains skewed towards the potential harm spiders can cause in humans (Hauke & Herzig, 2017). These feelings seemingly find their psychological roots in our ancestral fear of venomous animals (Knight, 2008;Gerdes, Uhl, & Alpers, 2009), but might also have a cultural component (Davey, 1994;Merckelbach, Muris, & Schouten, 1996;Davey et al., 1998). As Cavell (2018, p. 2) nicely put it "… one of the most remarkable aspects of modern human-spider relations is the prevalence of arachnophobia in places with few or no highly dangerous spider species". Indeed, even though human-spiders encounters are frequent events because spiders are omnipresent in all terrestrial ecosystems (Turnbull, 1973), including indoor environments (Bertone et al., 2016), the objective risk of being bitten by a harmful spider is minimal in most areas of the world (Diaz & Leblanc, 2007). These considerations raise the questions of why such a skewed perception of risk persists in modern societies (Lemelin & Yen, 2015).
It is known that humans have the tendency to evaluate risk through feelings and emotions rather than objectively (Slovic & Peters, 2006), often overestimating the frequency of statistically rare events. For example, many people fear flying, even though the casualties associated with civil flights are estimated to be in the order of 0.07 deaths per billion passenger miles (Savage, 2013). The same line of reasoning can be applied to people's risk judgments of low probability events related to wildlife, such as being attacked by a large carnivore (Bombieri et al., 2018) or stung or bitten by a venomous animal (Langley, 2005).
Furthermore, a distorted perception of risk can be exacerbated by the way in which information is framed in the scientific literature (Bennett & Vetter, 2004;Stuber & Nentwig, 2016) or in traditional media sources (Gerber, Burton-Jeangros, & Dubied, 2011). As far as spiders are concerned, it has been demonstrated that there is a significant overdiagnosis of spider bites and envenomation in the medical literature (White, 2003;Bennett & Vetter, 2004;Vetter, 2004;Vetter et al., 2005;Vetter, Hinkle, & Ames, 2009;Stuber & Nentwig, 2016). A recent major role in spreading falsehoods about spiders could also be associated with traditional and social media, due to their high efficiency in conveying a message more directly and reaching a wider audience (Vosoughi, Roy, & Aral, 2018). It is understood how the media play an important role in the construction and circulation of risk images associated with animals, contributing to develop fears and ambivalence (Gerber et al., 2011). Yet, while spiders are the quintessential feared animals, there is still poor understanding of the role of the media in spreading (mis)information about them (Cushing & Markwell, 2010).
Here, we explored the human dimension of spiders in Italy through the lens of traditional and social media. We examined the media representations of human-spider encounters as published in Italian online newspapers over the past 10 years, in order to assess the accuracy, spreading, and sensationalistic content of news. We tackled the following questions: i) What is the content and quality of the information of each spider-related media report?
ii) What is the temporal distribution of spider-related news?
iii) Which factors determine the effective spreading of news on social media?
Our over-arching goal is to understand the potential role of online media in exacerbating arachnophobic sentiments and promoting a distorted perception of the risk associated with spider bites. This is important given that these negative sentiments may ultimately lead to lowering public tolerance towards spiders and reducing conservation efforts towards them (Knight, 2008;Simaika & Samways, 2018).

Media report search
We adapted the methodology of Bombieri et al. (2018) for retrieving media reports on humanspider encounters published in Italian online newspapers ( Figure 1a). We carried out online searches in Italian with Google news, choosing multiple keyword combinations. We first searched for the Italian words for bite ("morso"), followed by spider ("ragno") and one of the years between 2010 and 2020 (e.g., "morso ragno 2014"). We repeated the search using the word sting ("puntura") instead of bite, given that it is frequently used (incorrectly) by journalists (among others; see, e.g., Afshari, 2016). We then repeated the search, changing the noun "ragno" (spider) to the Latin and vernacular names of spider species generally perceived as dangerous in Italy: Cheiracantium punctorium ("Ragno dal sacco giallo"), Latrodectus tredecimguttatus ("Argia", "Malmignatta", "Vedova nera"), Loxosceles rufescens ("Reclusa", "Ragno eremita", "Ragno violino"), and Zoropsis spinimana ("Falsa licosa"). We compiled the list of species based on our experience in years of interaction with the staff of the Anti-poison Center in Milan (Centro Antiveleni) and the San Giovanni Molinette hospital in Turin, who regularly contacted us asking for expert opinions on spider identification (on average 4.6 requests/month in 2019). This search strategy led to a total of 260 searches: 2 actions ("morso" or "puntura") x 13 species names (the general words "ragno", 4 Latin, and 8 vernacular species names) x 10 years (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). For each unique keyword search, we checked news up to the final available page in Google news, collecting all the media reports referring to one or more encounters in Italy 5 between humans and spiders. We disregarded: i) media reports which did not mention a specific locality for the event; ii) media reports referring to spider bite events that occurred outside Italy (e.g., a report written in Italian but focusing on a spider bite that occurred in England); and iii) media reports not specifically reporting a spider-human encounter (e.g., news discussing best practices to deal with a spider bite).

Media report content
For each media report, we first extracted basic information: a) title, b) date of publication, c) journal name, and d) journal circulation ('Regional' or 'National"). We classified newspapers circulation as 'Regional' if their total circulation was below 50,000 copies and as 'National' if it was above 50,000 copies, using the 2017 Assessment for Press Circulation provided by the society Accertamenti Diffusione Stampa (ADS) srl. Whenever newspapers were not covered in this report, we used the information found on each newspaper' webpage.
Then we read the full article and scored the e) spider species as it was mentioned in the media report (even if the species attribution was incorrect based on indirect evidence), f) type of event ("encounter", "bite", or "deadly bite"), g) year of the event, h) location of the event, i) presence/absence of photographs of the spider, j) presence/absence of photographs of the bite, and k) possible mention of an expert-opinion (doctor, arachnologist, or general biologist). Since several media reports were discussing the same event, we created an identifier for each unique event ("Event_ID"), by combining location and year of the event (e.g., "Terni_2018"). We also derived WGS84 coordinates for each event location, by geo-referencing the nearest city on Google Earth.
Following Nanni et al. (2020), we expressed the success of each media report as its spreading on social media, using the number of total shares in Facebook. We chose Facebook, as it is one of the most used social media platforms in Social Science research (e.g., Wilson, Gosling, & Graham, 2012;Kramer, Guillory, & Hancock, 2014). We extracted Facebook shares using the API tool available on ShareCount webpage (www.sharedcount.com; accessed on 2 March 2020). When the number of shares exceeds 999, this tools returns a rounded number (e.g., 1K for number of shares between 1000 and 1999). In such cases, we used the lowest number (1000). Even though we compared the number of shares for media reports published in different years, we consider this a reliable approach (see Nanni et al., 2020). Indeed, the share of online news on social media typically reaches a stable plateau at 30 days after publication (Papworth et al., 2015).

Scientific quality of the media reports
We assessed the quality of each media report by checking for the presence/absence of four types of errors in text and figures: i) errors in photographs, when the photograph(s) of the species in the media report (if any) did not correspond to the species mentioned in the text, or when the attribution was not possible (e.g., blurry photographs); ii) errors in systematics and taxonomy, like the common mistake of considering spiders "insects" (Jambrina, Vacas, & Sánchez-Barbudo, 2010), but also subtle inaccuracies in term of Linnaean taxonomic ranks [e.g., Report_ID 271 (translated): "… the 'malmignatta', a genus of Italian spider belonging to the family of the species of the black widow"]; iii) errors in venom and other physiological or medical aspects or terminology [e.g., Report_ID 147 (translated): "… the venom sac was removed with surgery"]; and iv) errors in morphology and anatomy, such as the frequent "spider sting" instead of "spider bite" (Afshari, 2016).
Each error type was scored as present or absent, thus we did not counted cumulative errors of the same type in the same report.

Classification of Sensationalism
Three authors (MI, SM, and VN) independently evaluated the title, subheadings, and main text of each media report, and assessed it as overstated (sensationalistic) or not (neutral). We took the consensus between the three independent evaluations to minimize the effect of subjectivity.

Data analysis
We conducted all analyses in R (R Core Team, 2018). We graphically explored the content of media reports with barcharts and boxplots with 'ggplot2' (Wickham, 2016). For the two most abundant species, Latrodectus tredecimguttatus and Loxosceles rufescens, we explored temporal distribution of media reports using density plot, by computing a kernel density estimate with a 1.5 bandwidth adjustment for both the annual and monthly distribution of media reports (Wickham, 2016). For this and the following analysis, we excluded media reports published in 2020 given this year was covered only up to February.
We used generalized linear mixed models (GLMM) to explore the factors driving the share of news on Facebook. We followed Zuur & Ieno's (2016) protocol for presenting regression-type analyses, whereby we: i) conducted data exploration and identified the dependency structure in the data; ii) explained, fitted, and validated the regression models; and iii) interpreted the regression output and presented the main effect plots.
The data exploration revealed the presence of four outliers in the number of shares, namely media reports shared over 15,000 times on Facebook. We removed these four observation from the database. Furthermore, we observed that 39.5% of media reports were never shared on Facebook (Figure 2b). However, since these are "true zeros" (sensu Blasco-Moreno, Pérez-Casany, Puig, Morante, & Castells, 2019), we did not apply zero-inflated models.
We fitted GLMMs with 'glmmADMB' (Fournier et al., 2012), starting from an initial structure that included all covariates and random terms of interest: The random factor 'Newspaper' was introduced because reports published in the same newspaper usually share a similar language, style, and graphical elements. The random factor 'Event_ID' was introduced to take into account the fact that multiple reports in our dataset discussed the same events. We included the square of month (term month 2 ) to capture a possible seasonal response of the shares during the year (i.e., a quadratic relationship between shares and month).
The numbers of Facebook shares are counts, so we initially chose a Poisson distribution.
The Poisson GLMM was, however, highly over-dispersed (χ 2 : 227751553743; p<0.001) and so we switched to a negative binomial distribution. Once the initial model had been fitted, we performed a step-wise model selection in 'MuMIn' (Bartoń, 2019). We based the model reduction on Aikaike Information Criterion (AIC) and Aikaike weights [w i (AIC)] (Burnham & Anderson, 2004), in order to simplify the model and avoid overfitting (Hawkins, 2004).

Content of media reports
We collected and analysed 314 media reports published between 2010 and 2020, discussing 344 spider-related events attributable to 97 unique events (Figure 1b). The average (±s.d.) number of media reports discussing each event was 3.52 ± 6.72 (range 1-33). The two most discussed events were i) the story of a traffic warden from Terni who was supposedly bitten by a

Quality of media reports
One or more error types were present in 73% of media reports (Figure 3). The distribution of errors varied, however, depending on the species: most media reports referring to L. tredecimguttatus and other species contained no errors, whereas most reports on L. rufescens contained one or more errors (Figure 3a). The most frequent errors referred to spider morphology and anatomy (55.3%), species photographs (28.4%), and systematics and taxonomy (25.8%).
Errors referring to venom and other physiological aspects were present in 15% of media reports (Figure 3b-e).

Temporal distribution of media reports
We observed a strong temporal signal in the distribution of media reports between 2010 and 2019, with a recent increase in the number of news for both species, which was rather steadily increasing in L. tredecimguttatus and almost exponential in L. rufescens (Figure 4a). From a seasonal point of view (Figure 4b), we found that there was a clear summer peak, in July, in the frequency of reports for both species. This seasonal pattern was more evident for reports referring to L. tredecimguttatus.

Factors affecting the sharing of media reports on social media
The model that minimized AIC included year and sensationalism as fixed terms (Table 1).
We found a significant positive effect of the year of publication, with recent media reports being, on average, more frequently shared on social media (Figure 5a). Furthermore, media reports with sensationalistic content were, on average, more frequently shared on social media ( Figure 5b).
All other factors had no significant influence on sharing on social media, and were discarded during model selection (Table 1).  Table 1 for model selection and estimated regression parameters).

Content of media reports and temporal distribution
We found that the quality of online newspaper articles focusing on spiders in Italy is, in general, rather poor. Media report quality appears to be independent of the newspaper's circulation (national versus regional). Over 70% of media reports contained errors, 32% were characterized by an overstated content, and in virtually none of them was an expert consulted or interviewed.
The two most represented species in the media reports were Latrodectus tredecimguttatus and L.
rufescens, two species widely distributed in Italy (Pantini & Isaia, 2019). These two species belong to the only globally distributed genera responsible for medically important clinical syndromes, namely latrodectism and loxoscelism (Isbister & Fan, 2011). The fact that these species can deliver potentially harmful bites to humans seemingly explains why they are able to attract this great attention from the media.
Loxosceles rufescens is native to the Mediterranean basin (Planas, Saupe, Lima-Ribeiro, Peterson, & Ribera, 2014), but has been introduced to many areas of the world where it is considered an important invasive species (Nentwig, Pantini, & Vetter, 2017;Taucare-Rios, Nentwig, Bizama, & Bustamante, 2018). The Mediterranean recluse is a rather common inhabitant of natural and in-door habitats in Italy and thus, it seems likely that it has been coexisting with humans for centuries. Indeed, the species has been known in Italy since the second half of the XIX century, when the first catalogue on Italian spiders was published (Canestrini & Pavesi, 1868). According to scientific literature on Italian spiders ( Coincidentally, this increase also came after the publication in Italy of Quand sort la recluse, a crime novel by Fred Vargas, where Chief inspector Jean-Baptiste Adamsberg has to deal with a series of murders committed using the venom of L. rufescens. While there is probably no causal relationship between these events and the increase in number of reports, it is interesting to note that several recent media reports in our database referenced both sources. The quality of media reports referring to L. tredecimguttatus was better, and fewer reports had a sensationalistic content. Latrodectus tredecimgutattus was described based on specimens collected in Volterra (Tuscany). The species is distributed across a wide area in the Palearctic region, from the Mediterranean basin to Ukraine, Caucasus, Central Asia, and China (World Spider Catalog, 2020). In Italy, as well as in most in most other countries, the Mediterranean black widow is preferably found in ruderal areas of agricultural land and, just like L. rufescens, has been living close to humans for centuries. However, according to the data presented in scientific literature, its presence in strictly indoor habitats is infrequent -only 2 records out of the 23 available describing habitats (Pantini & Isaia, 2019). According to scientific literature on Italian spiders (Pantini & Isaia, 2019), most records of this species refer to natural or semi-natural (agricultural) habitats and only in one case (Pepe, 2005) has the species been reported in synanthropic habitats. The Mediterranean black widow began appearing in the media spotlight only in the past ten years (Figure 4a), with the highest number of media reports from late spring to early autumn, paralleling the period of highest activity of the species (Nentwig et al., 2020) and corresponding to the higher possibility of human-spider encounters. Given that most media reports on L. tredecimguttatus were in fact 'Encounters' (Figure 2a), namely reports of the species' presence as provided by readers of the different newspapers, the distribution of news may be somehow tracking the species' phenology, making it an unusual example of iEcology (Jarić et al., 2020).
We found that the risk scenario depicted by the media reports was unnecessarily alarmist, especially with regard to L. rufescens. First, no proven fatality due to a bite by L. rufescens has occurred globally (Nentwig et al., 2017). Second, overdiagnosis of spider bites is a rather common phenomenon for "popular" taxa such as Loxosceles and Latrodectus (Stuber & Nentwig, 2016). A conservative estimate would suggest that less than 10% of the bites reported in the media reports analysed here were delivered by the species described in the report (see Suchard, 2011). Third, in virtually none of the media reports is it written that the biting spider was brought to a hospital for identification, thus the causal attribution remains unconfirmed and merely suspected (Vetter & Isbister, 2008). Accordingly, the content of the majority of media reports analysed here has to be taken at best as anecdotic.
These considerations apply also to the three casualties associated with a bite by L.
rufescens reported in the media reports. The only scientifically supported fatality refers to a case of loxoscelism in a woman, 65, dating back to 2015. This event was discussed in the medical literature (Pezzi et al., 2016), and later began to be mentioned by some journalists (n=7 media reports). However, the validity of this medical report was readily questioned by Nentwig et al. (2017), because the identity of the spider biting the woman was not ascertained. Allegedly: "[The woman] was bitten the evening before hospitalization while cleaning the home cellar by a spider, which, from the description and place where the bite occurred, could probably be identified as the Loxosceles rufescens species" (Pezzi et al., 2016). Two other fatalities covered in the media reports -Cagliari (2017) and Aosta (2020) -are unverifiable, and most likely wrong, given that neither was the bite ascertained nor was the spider collected and identified.
The validity of these reports was even questioned in some newspapers, for example Report_ID 229 stating that "The story of the men who died due to a violin spider bite is probably fake news", or Report_ID 115 observing that "… he died three months after being stung [bitten] by a violin spider. But the cause of his death could be another" (titles literally translated).
The seasonal pattern in the distribution of news with a marked summer peak (Figure 4b), parallels what was found by Cushing & Markwell (2010) when analysing newspapers articles on the Australian endemic Sydney funnel-web spider (Atrax robustus). The higher prevalence of secondary news during the summer holidays is a well known trend in journalism whereby, in the absence of more relevant news, a secondary subject such as a spider bite is frequently able to make it to the front pages.

Social media amplification of sensationalistic contents
Social media have profoundly shaped the way information is produced and circulated. In recent years, social media platforms have become an important battlefield for political debates (Hall, Tinati, & Jennings, 2018), as well as the primary digital environments where people inform themselves and frame their perception of the world (Weeks, Ardèvol-Abreu, & Gil de Zúñiga, 2015). In parallel, social media have also become the preferential channel though which traditional news are disseminated and discussed (Lee & Ma, 2012), with most newspapers now actively using social media platforms such as Facebook and Twitter to spread their online contents more effectively (Ju, Jeong, & Chyi, 2014) In line with this, we found that the share of spider-related news on social media has increased significantly in recent years (Figure 5a). In the contemporary era, where "more iPhones are sold in a few days than there are tigers, elephants, and gorillas on the planet" (Chapron, Levrel, Meinard, & Courchamp, 2018: p. 651), this result did not come as a surprise.
However, not all news on spiders were shared with the same frequency online. While sensationalistic reports represent only about one third of the total media reports analysed in this survey, these were on average shared on Facebook two to three times more than neutral news ( Figure 5b). This results is in accordance with general studies demonstrating that newspaper articles with content evoking strong positive or negative emotions are more likely to become viral (Berger & Milkman, 2012). Being shared on social media, sensationalistic news will inevitably be more widely read. Due to their sensationalistic content, they are also more likely to remain imprinted in a reader's memory, especially in an arachnophobic reader's, since it has been demonstrated that arachnophobics recall spider-relevant information more effectively (Smith-Janik & Teachman, 2008). On top of this, social media platforms are a fertile ground for emotional contagion, the phenomenon whereby emotional states are rapidly transferred to others leading to massive-scale emotional homogenisation (Kramer et al., 2014). This may contribute to empowering a biased perception of risk (Gerber et al., 2011) and facilitate the persistence of arachnophobic sentiments.

Significance of results for spider conservation
Fear of spiders is one of the most prevalent animal-related phobias in humans  and thus, spider-related contents are an effective emotional trigger (Smith-Janik & Teachman, 2008). We have shown how some journalists are able to exploit arachnophobic sentiments to their advantage, framing sensationalistic news capable of attracting substantial online attention. Sensationalistic news that dramatize and overstate the frequency of spiders "attacks" on humans are also those which most attract social media. Through emotional contagion, this biased representation is spread online. Ultimately, this may result in lowering public tolerance for spiders and lead to lower willingness for conservation and management efforts.
As demonstrated by Knight (2008), aesthetic and positive/negative features of animals correlate to the protection each taxon receives. Accordingly, the main challenges facing invertebrate conservationists is to change the perceived negative connotations of invertebrates by the public , raising awareness about the importance of these often uncharismatic organisms for the correct functioning of ecosystems .
Traditional media have the potential to play an important role in changing the status quo, by offering the public unbiased representations of spiders. Thus, we urge journalists to renew their efforts toward objectivity and accuracy, which are best achieved by i) consulting and interviewing experts; ii) referring to scientific literature, as well as to modern online resources led by expert arachnologists (see, e.g., the "Recluse or Not?" project on Twitter; @RecluseOrNot); and iii) avoiding unmotivated sensationalism when describing biting events.
The traditional media arguably remain among the most powerful communication tools, capable of delivering their message effectively especially thanks to the aid of social media (Ju et al., 2014). If this potential is harnessed to the goal of delivering accurate information to the