Improper weapons are a neglected category of harmful objects

According to legislation, objects are typically classified as weapons if they are offensive per se (referred to here as proper) and if they are adapted for use as weapons or carried with the intent of causing injury (referred to here as improper), with specific regulations on their usage and possession in public spaces. However, little evidence exists on the validity of this distinction in psychology, despite a widespread recognition of the importance of psychological states and subjective perceptions in risk assessment. We conducted an online survey to evaluate hazard perceptions in relation to three dimensions (dangerousness, frequency of events, controllability) of three object categories: proper weapons, improper weapons, and everyday objects. The data from our 300 respondents reveal that the three categories of objects differ from one another on the three dimensions. Moreover, hazard perceptions differ between males and females for improper weapons but less so for proper weapons. These findings suggest that proper and improper weapons are two psychologically distinct categories, albeit with fuzzy boundaries. Investigations into their differential properties may thus help improve risk assessment in security contexts.


Results
Data were collected from 300 individuals in an online survey, with the main focus being on the investigation of danger perception in proper weapons, improper weapons, everyday objects (henceforth, Category). The questionnaire was divided in two blocks: the first block consisted of an evaluation task in which participants used a seven-point Likert scale to rate the degree of dangerousness, frequency of events (i.e. the frequency with which these items are involved in harmful events), and controllability (henceforth, Judgment) connected with each object; the second block consisted of a written naming (identification) task. Within each block, whose order www.nature.com/scientificreports/ was fixed across participants, items were presented randomly. For each participant, only ratings of correctly identified items were included in the analysis. A 2 × 3 × 3 mixed ANOVA was conducted to assess whether the judgements differed across the three categories and/or gender. Pearson's correlation was then utilized to evaluate the relationship between perceived dangerousness, frequency of events, and controllability for each category. Finally, two cluster analyses were performed to explore differences and similarities within the improper weapon category and between the proper and improper weapon categories.
ANOVA. Only 0.12% of the total observations (i.e. 48 out of 40,500 judgments, linked to 16 items across 13 participants overall) had to be excluded from the analyses due to incorrect identification. No significant deviations from normality were found (all ps > 0.05) but sphericity could not be assumed (all ps < 0.05). Therefore, results are reported after Greenhouse-Geisser correction. .050] was also found. A post-hoc power sensitivity analysis was performed using the software program G*Power 3.1 67 . The power analysis indicated that with 300 participants, an effect size of f = 0.23 with η p 2 = 0.05 and with an alpha = 0.02, we had power (equal to 1) for the three-way interaction between Gender, Category and Judgement. Bonferroni's post-hoc tests for the two-way interaction between Judgment and Category (mean differences, pairwise comparisons test statistic, and effect size are reported in Table 1) revealed that proper weapons were significantly perceived as more dangerous than both improper weapons (p < 0.001) and everyday objects (p < 0.001), and that improper weapons were perceived more dangerous than everyday objects (p < 0.001). Ratings for frequency of events were higher for proper weapons than improper weapons (p < 0.001) and everyday objects (p < 0.001), and were higher for improper weapons than everyday objects (p < 0.001). In addition, pairwise comparisons revealed that the perceived controllability for proper weapons was significantly lesser than for improper weapons (p < 0.001) and everyday objects (p < 0.001), and was lesser for improper weapons than everyday objects (p < 0.001).
Bonferroni's post-hoc tests for the three-way interaction revealed that a Gender effect was consistently present with improper weapons, as shown in Fig. 1. Means, standard deviations and test statistics are reported in Table 2. Compared to women, men judged improper weapons less dangerous (p < 0.001), less frequently involved in harmful events (p < 0.001), and more controllable (p < 0.001). For proper weapons, men had a higher sense of control over proper weapons (p < 0.001) but they did not differ from women in the perceived dangerousness (p = 1) and the perceived frequency of events (p = 0.22). For everyday objects, no gender difference emerged in any of the three dimensions of dangerousness, frequency, and controllability.
Correlations. The relationship among the dimensions of interest was then examined using Pearson's correlation coefficients. A Bonferroni correction was applied to the overall alpha threshold, which was thus lowered to 0.0022 for a single test. Results showed that proper weapons' dangerousness correlates positively with frequency of events, r(298) = 0.58, p < 0.001, and negatively with controllability, r(298) = − 0.42, p < 0.001. The same pattern was found for improper weapons, in which dangerousness correlates positively with frequency of events, r(298) = 0.62, p < 0.001, and negatively with controllability, r(298) = − 0.41, p < 0.001. The subjective perception of the frequency of events is negatively correlated with controllability both for proper, r(298) = − 0.29, p < 0.001,  Cluster analyses. Given the significant difference between proper and improper weapons for dangerousness, frequency of events, and controllability, two cluster analyses were performed to better characterize the categories under investigation 68 . Cluster analyses were carried out using R (4.0.3), with the "Cluster" (2.1.0), "Factoextra" (1.0.7) packages. A first cluster analysis aimed to identify any subsets within the improper weapons category. A second cluster analysis was performed to identify possible connections/overlaps between the two

Pairwise comparisons
Male

M (SD)
Female www.nature.com/scientificreports/ main categories of interest (proper weapons and improper weapons). Both analyses were based on dangerousness, frequency of events and controllability judgments, which served as input.

M (SD) Test statistics (t) Adjusted p value Cohen's d
For each cluster analysis, we first conducted agglomerative hierarchical cluster analyses using Ward's method of minimum variance with a squared Euclidean distance measure 69 . The best-distinguished cluster solution was then determined from the visual inspection of the resulting dendrogram, a tree diagram showing the arrangement of the clusters produced by hierarchical clustering. After identifying the number of distinct clusters, k-means clustering with the number of clusters as input for k was used to determine the content of the distinct groups.
The analysis for the improper weapons category yielded a four-cluster solution, with hammer and knife in cluster A, cutter, razor, saw, scissors, blowpipe, and firecrackers in cluster B, screwdriver, cooking thermometer, tube, baseball bat and wrench in cluster C, pliers and chisel in cluster D. The levels of dangerousness and frequency of events decrease as the degree of controllability increases, from cluster A to cluster D.
The analysis with both the proper weapons and the improper weapons categories yielded a three-cluster solution, with a peak of overlap between categories in the middle cluster, as shown in Fig. 2. Cluster A′ includes six proper weapons (gun, assault rifle, rifle, bomb, dynamite, dagger) and two improper weapons (hammer, knife); cluster B′ includes five proper weapons (sword, flare gun, puncher, bullets, taser) and six improper weapons (saw, cutter, scissors, razor, blowpipe, firecrackers); cluster C′ contains four proper weapons (mace, baton, detonator, Swiss knife) and eight improper weapons (baseball bat, tube, screwdriver, wrench, cooking thermometer, chisel and pliers).
Again, the levels of dangerousness and frequency of events decrease as the degree of controllability increases, from cluster A′ to cluster C′.

Discussion
Research on risk perception focuses on several areas (e.g. risk exposure, risk communication etc.). In this study we investigated the subjective feeling of whether different kinds of objects, namely proper weapons, potentially dangerous utensils (improper weapons), and non-dangerous everyday objects, may be a potential cause of harms. Studies on proper weapons are scarce and there appear to be even fewer studies on improper weaponry, which encompass work equipment or utensils with the potential to cause harm, despite the fact that they are not designed for it. To address this gap, we devised an online survey in which we collected judgments on dangerousness, as well as controllability and frequency of events, of improper weapons, proper weapons, and everyday objects.
The main finding of the study is that people's perceptions of hazard, of the frequency of criminal incidents involving them, and of their sense of controllability differ widely across proper weapons, improper weapons, and everyday objects. On all three dimensions, improper weapons differ both from proper weapons and from everyday objects: on the one hand, they are perceived as significantly less deadly, less frequently involved in harmful events and more controllable than proper weaponry; on the other hand, they are perceived as significantly deadlier, more frequently involved in harmful events, and less controllable than everyday objects. www.nature.com/scientificreports/ The difference we found supports the validity of the categorical distinction between proper and improper weapons in the Italian and international legislation, and attests to the need for an analogous distinction between these two categories in the risk assessment literature.
The observed disparity can be explained in light of studies on the subject of risk perception 38,46,52 , which map a wide range of risks in a two-dimensional factor space according to whether they were unknown and feared. The dimension of risks increases when the sensation of dread and the difficulty of control increase, along with a greater sensation of frequent involuntary exposure 46 . Thus, proper and improper weapons pose a higher risk than everyday objects because they are perceived as more dangerous and less controllable (that is, the severity of consequences), and more frequently involved in harmful events (that is, the probability of occurrence). Our assessment that there are two psychologically separate categories of weapons-proper and improper weaponsleads us to believe that the level of risk they pose is likewise distinct, and that people may act differently when confronted with either type of weapon, highlighting the need for additional research. Indeed, we found a positive correlation between dangerousness and frequency of events and a negative correlation between dangerousness and controllability and we hypothesise that frequency of events and controllability may have led to a higher degree of perceived dangerousness for both proper and improper weapons. In addition, we hypothesise that the lower degree of controllability of proper weapons, as well as the perception of a higher frequency of accidents involving them, may influence the perception of dangerousness of proper weapons compared to improper weapons; the same pattern can be found in the difference between improper weaponry and everyday objects.
Two cluster analyses indicated stratifications within the category of improper weapons and overlaps with the category of proper weapons. A first cluster analysis led to the identification of four clusters in the improper weapons category.
When looking at the composition of the clusters, knife and hammer appear at the opposite end of the spectrum compared to chisel and pliers. Sharp (e.g. the knife) and blunt things (e.g. the hammer) are the most commonly used improper weapons in criminal episodes (for Italy see 70 ; for the U.K. see 71 ; for South Korea see 72 ), while the chisel and pliers are much less used in harmful events 15 . The overall higher dangerousness of the knife and hammer might also be due to associated perceptions of higher severity of injuries caused by these objects 15,73 . The remaining objects are divided into two intermediate clusters and exhibit a wide range of physical characteristics (elongated, sharp, blunt shape, heavy or light).
Notably, the overall level of perceived dangerousness appears to decrease as an object's use becomes more specialized. Therefore, objects with more diverse uses (e.g., the knife, the hammer) are considered more dangerous than objects with more sectoral and specialized uses (e.g. the chisel). The second cluster analysis shows that the categories of proper and improper weapons have a graded structure just like natural categories do 74 . Objects in a category lie on a gradient of typicality, ranging from extremely good to very poor examples of the category 75 , with prototypical members serving as reference point to which other category members can be assessed 76 . Thus, categories vary for the degree of membership of their exemplars and they may even overlap other categories, as is the case with the categories of proper and improper weapons under examination.
As predicted, we found that gender influence risk perception of weapons; however, the pattern we found is rather articulated, suggesting that it may not be gender per se that drives the effect but rather everyday experience with the type of objects to be judged. To summarize, women judge improper weapons, but not proper weapons, riskier and more involved in events than men do, and they judge both improper and proper weapons less controllable than men do.
We suspect that the greater use of work tools by males than by women 77 may explain the gender difference found in the perception of risk in improper weapons and the controllability of proper weapons. In Italy at least, this may be indirectly linked to gender disparity in employment 78 . Among the workforce in the private sector where a wide unbalance between men and women is found 79 (Decree-Law n.142, October 16, 2020) figure craft workers, engineers, and installation and maintenance experts of electrical and electronic equipment for 2021 (95.1% male-female disparity). Intriguingly, in Brodeur et al. 61 , tools (including those that we classify as improper weapons) were the only inanimate stimuli showing a significant gender difference, males being more familiar with tools than females. This in turn may reflect gender differences in interests and exposure to experiences ingrained in Western culture that may be based only partly on biology 8 . Males' larger use and familiarity with improper weapons may influence them to associate more strongly these objects to their original function, a sort of functional fixedness effect 80 , leading men to perceive them as less threatening than women do. Individuals' familiarity with the way objects are typically used, according to Duncker 80 , hinders people from representing them under a new aspect or purpose. In this case, previous experiences that stress the conventional function 81,82 of hand tools might minimize males' connection of these objects with their alternative function as weapons.
Men and women differ widely also with respect to owning a proper weapon (for the United States, see 83 ) which may lead to the divarication of the sense of control about them between women and men. No data exist for civilian weapon ownership for Italy, but the imbalance between men and women in the armed forces 84 , and in hunting practices, and thus the differential in the authorized holding of proper weaponry, can serve as an indirect validation. Because proper weapons have a single function, the phenomenon of functional fixedness would be missing in this scenario, and the gender difference in the controllability could originate from women's poorer familiarity with weapons compared to males, as there are fewer women in possession of a weapon than men 83 .
By testing larger and more heterogeneous samples of participants it would be possible to ascertain the general character of the psychological distinction between proper and improper weaponry we have documented here. Additionally, the evaluation of extra features could help enrich our model and ponder on the wider implications of such distinction. Indeed, the disparities in risk perception of proper and improper weaponry that we report may have substantial implications for risk communication, in addition to their importance for risk assessment. Risk communication aims to bridge the knowledge gap between experts and non-experts, as naive people frequently lack proper awareness of the dangers they confront 46  www.nature.com/scientificreports/ than experts 33,37,38,41 . The increment of the relevant knowledge among people would result in a more informed and better calibrated judgments about the existence, nature, and/or severity of risks and hazards and more predictable and highly correlated (with domain-specific knowledge) risk perceptions 46,85 .

Methods
Stimuli collection and standardization. For the Italian legal system 1 (see also art. 1,2,4 of Italian law n. 110, April 18, 1975), the following are considered proper weapons: any sort of firearm (including ammunition); air-powered weapons; sidearms (such as swords, double-edged knives or switchblades, daggers, etc.); instruments for which there is an absolute prohibition (maces, punchers; batons are considered as proper weapons by decision no. 22314, July 8, 2019, of the Italian Supreme Court); bacteriological or chemical weapons (not included in this survey); all incendiary, exploding, or disrupting devices (bombs, Molotov cocktails). According to the same Italian legislation, improper weapons include blunt objects like clubs, tubes, chains, bolts, metal balls, hammers, and bars; cutting tools like kitchen knives or hatchets, are additional examples of improper weapons. Arches, blowguns and crossbows are also considered as improper weapons: they are recreational and hunting tools, hence should only be used for such purposes. For each of the classes of objects (apart for bacteriological and chemical weapons), at least two examples were chosen and included as items in the current study (i.e. two firearms, two cutting weapons, etc.).
For the purpose of this study, the distinction between everyday objects and weapons is based on the regulations of the Italian Civil Aviation Authority (available here: https:// www. enac. gov. it/ passe ggeri/ cosa-porta rebordo/ artic oli-vieta ti-in-cabina). Everyday objects were chosen among those allowed in a plane's cabin, while items from the other two categories are not allowed. Based on this distinction, clothing, accessories, toys, reading material (i.e. books, magazines), and plastic containers (i.e. glasses, bottles, plastic bags) are examples of everyday objects. Also for each of these subgroups at least two specimens were used.
Overall, 15 items for each of the three categories were selected. Item selection was determined by physical dimensions and frequency of use in the Italian language, so that all the objects selected have comparable physical dimensions that allow them to be carried in a hand luggage and relatively high frequency of use in the Italian language, avoiding regional, dialectal, obsolete and low-use terms (as found in the Nuovo Vocabolario di base della lingua italiana, NVdB, New Basic Vocabulary of the Italian Language, available here: dizionario.internazionale.it/nuovovocabolariodibase; see Supplementary Table S1 for details). As a result, the following items were included in our study: Proper weapons: mace, flare gun, bomb, dagger, puncher, taser, dynamite, bullets, baton, detonator, gun, sword, Swiss knife, rifle, assault rifle; Improper weapons: baseball bat, cooking thermometer, cutter, screwdriver, hammer, tube, firecrackers, pliers, chisel, razor, scissors, knife, blowpipe, wrench, saw; Everyday objects: glass, hat, jacket, ball, bottle, sock, shirt, book, cap, undershirt, bag, skirt, heel shoe, jeans, sneakers.
Photographs of these items were obtained via Google image search and selected among the files licensed with Creative Commons 1.0 Universal (CC0 1.0). Colored backgrounds or shadows, if present, were removed from the figures so that all items stood out against a white background, and a single object was shown per image. The images were resized (maintaining original proportions) and their dimension was standardized to 350 pixels (wide) × 500 pixels (high). As dangerous objects are seen as more hazardous when they are oriented towards participants than when they are directed away 86 , the orientation of proper and improper weapons was made uniform (from the participant's perspective, the weapon's handle-where present-is on the left side of the screen, and the offensive part is directed towards the right), in order to eliminate potential biases in hazard perception. Also, the orientation of everyday objects with a handle was made uniform and consistent with the criteria applied to weapons. For each image we calculated three standard visual complexity indices: edge density, feature congestion and subband entropy [87][88][89] . With a significance criterion of p < 0.02, a 3 × 3 repeated-measures ANOVA shows that the three categories do not differ from each other on these complexity measures. The values of the indices for each image are reported in Supplementary Table S1.
Participants. Participants were recruited via social networks and personal contacts at the Universities of Trento, Siena and Venezia (Italy). A total of 300 Caucasian adults (all of whom were over the age of 18) Italian speaking volunteers (120 males, M age = 32.06, SD age = 10.84; 180 females, M age = 30.64, SD age = 11.48) were included in the study. Prior to the experiment, all participants provided online informed consent and they received no compensation for their participation. All methods were approved by the University of Trento, Human Research Ethics Committee (Protocol 2020-019). The whole procedure was realized in accordance with the Helsinki Declaration.
Procedure. Participants were granted access to the Qualtrics survey platform via an open web link. First, they were provided with general information on the questionnaire. The rationale and the purpose of the study were made clear, with the main focus referred to as the investigation of danger perception in several categories of objects. It was specified that these objects could have been either weapons or everyday life objects. In addition, the following definitions of danger and weapon were given, in order to prevent misjudgments: Danger: "Danger is defined as an intrinsic trait or quality of a given factor that has the potential to inflict damage, as defined by art. 2, letter r, of Italian law n. 81, April 9, 2008. Danger is thus an intrinsic attribute (of the situation, object, substance, etc.) unrelated to external circumstances; it is a scenario, object, substance capable of hurting people, due to its properties or characteristics". www.nature.com/scientificreports/ Weapon: "Strictly from a technical standpoint, a weapon is any instrument capable of offending, either by its natural purpose or by the method in which it is used".
The questionnaire was divided in two blocks. The first block required evaluating the degree of dangerousness, controllability and frequency of events associated with each object; the second block consisted of a written naming task. The 45 pictures of the study items were shown once in the first block and once in the second block. The order of blocks was fixed across participants. Within each block, items were presented randomly.
In the evaluative task, the picture was presented at the top center of the page under which three questions appeared (Fig. 3).
Participants were asked to judge the extent to which the objects of the kind displayed could pose a danger (dangerousness; "How dangerous do you think it is?"), how common are harmful episodes involving the object (frequency of events "How much is this object implicated in harmful incidents?"), what level of control participants would be able to exert on that object in the context of an event (controllability; "How much control would you have over this object if you were in danger because of it?"). A seven-point Likert type scale (1 = Not at all to 7 = Extremely) was provided to answer each question. Participants were required to answer all questions in order to proceed to the next page. Once moved on to the next page, participants could not go back and amend their responses.
In the written naming task, each image was displayed in the top center of the page, with a text box below it where the participant could type the name of the item shown. Participants were asked to write the name of the item, even in case of uncertainty. After providing at least one response, they could then move on to the following page. The purpose of the written naming task was to ensure that participants could actually identify what they were judging; items for which an incorrect name was provided were excluded from the analyses.

Figure 3.
A page from the questionnaire relating to the perceived dangerousness, frequency of events, and controllability for the stimulus "cutter" is reported here. When reporting the questionnaire page in this figure, the gap between the questions was decreased, to optimize the use of space. www.nature.com/scientificreports/ In the final section of the questionnaire, participants were asked to provide demographic information about their age, gender and education. The IP address of the client computer was used to identify potential duplicate entries from the same user.
Normality and sphericity checks were carried out on the dependent variable, for any cell of the Judgement × Category design, using the Kolmogorov-Smirnov test and Mauchly's test, respectively.
In order to assess whether the judgements differed across the three categories and/or gender, a 2 × 3 × 3 mixed ANOVA was conducted, with Gender (Women, Men) as between-participants factor, Judgment (dangerousness, frequency of events, controllability) and Category (proper weapons, improper weapons, everyday objects) as within-participants factors. A significance criterion of p < 0.02 was used. In case of significant interactions, pairwise comparisons were conducted with Bonferroni's post-hoc test.
For each category, Pearson's correlation was used to assess the relation between the perceived dangerousness, frequency of events and controllability. The family-wise alpha threshold was 0.02 (lowered to 0.0022 per individual test after Bonferroni correction). Finally, for the proper and improper weapon categories, two cluster analyses were conducted for exploratory purposes. For both cluster analyses, an analysis of agglomerative hierarchical clusters using Ward's method of minimum variance with a squared Euclidean distance measure was performed first, followed by an analysis of k-means clustering.