Vehicle avoidance: The hierarchy of visual attention towards animals, plants, and vehicles

Chihiro Kioka; Kunihito Tobita

doi:10.1371/journal.pone.0330475

Abstract

The biophilia hypothesis posits that humans have an innate affinity for nature, with natural landscapes effortlessly capturing their attention, and a tendency to seek nature. The animate monitoring hypothesis suggests that humans have evolved to quickly detect and respond to animals for survival. The plant awareness disparity hypothesis argues that people notice plants less than animals due to perceptual biases and preferences. Based on these hypotheses, it was predicted that people’s visual attention would be superior towards animals, plants, and manufactured objects, in that order. This study investigated the hierarchy of visual attention towards animals (birds, mammals and humans), plants (fruit), and manufactured objects (vehicles) using a dot-probe task framework. The findings revealed no significant differences in reaction time or attentional bias for animal or plant stimuli. In contrast, perceptual processing was inhibited when viewing a vehicle and attentional avoidance occurred, resulting in slower reactions than to animals or plants. These findings offer partial support for the proposed hierarchy of visual attention, suggesting that while natural stimuli such as animals and plants receive comparable attention, some manufactured objects may elicit perceptual avoidance.

Citation: Kioka C, Tobita K (2025) Vehicle avoidance: The hierarchy of visual attention towards animals, plants, and vehicles. PLoS One 20(9): e0330475. https://doi.org/10.1371/journal.pone.0330475

Editor: Jie Wang, Education University of Hong Kong, HONG KONG

Received: June 4, 2025; Accepted: August 3, 2025; Published: September 22, 2025

Copyright: © 2025 Kioka, Tobita. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The datasets generated and analysed for this study can be found in the Open Science Framework (OSF) repository: https://doi.org/10.17605/OSF.IO/3RTSQ.

Funding: This work was supported by the Japan Society for the Promotion of Science (JSPS) through a grant awarded to KT [grant number 22K05710, https://kaken.nii.ac.jp/d/p/22K05710].

Competing interests: The authors have declared that no competing interests exist.

Introduction

Frequent visits to green spaces are associated with a lower risk of attention deficit hyperactivity disorder [1], Alzheimer’s disease [2], depression, and hypertension [3]. These examples illustrate the physical and mental health benefits people can experience from spending time in various natural environments. The biophilia hypothesis posits that humans possess a high innate affinity for nature, which is a trait modern humans share with our ancestors [4]. By gravitating towards nature, individuals may envision a future in which they can thrive [5], optimise their abilities, and enhance their overall well-being [6].

Evidence supporting this preference for nature can also be observed in visual attention bias. In a previous study using the dot-probe task, probes replacing natural landscapes dominated by vegetation were detected more quickly than those replacing urban landscapes [7]. Schiebel et al. [8] further confirmed participants’ inclination to approach nature and avoid urban environments using the dot-probe task, implicit association test, and approach-avoidance task. Natural landscapes capture more human attention than urban landscapes, effortlessly guiding one’s focus towards nature [9].

The animate monitoring hypothesis (AMH) posits that, within the realm of nature, animate objects possess an attentional advantage [10]. The group-foraging nature of early humans meant that rapid and frequent monitoring of both humans and animals was essential because their situations could change quickly. The AMH has been validated through a range of tasks. For example, in the change-detection task [10], changes involving humans and animals were found to be more frequently and swiftly detected than changes related to vehicles, buildings, plants, or tools, and were less prone to being overlooked. Furthermore, in the visual search task [11], animals were detected more rapidly than inanimate objects, while in the attentional blink task [12], the duration of undetectability of new stimuli following initial detection was shorter for animals than for non-animals. In addition, a comparison of the detection performance between artificial intelligence (AI) and humans revealed that while AI and humans performed equally well in detecting cars, humans outperformed AI in detecting humans [13]. These findings suggest that humans excel at recognising object categories of evolutionary significance. As noted above, accumulating evidence supports the AMH.

In contrast, the concept of plant awareness disparity (PAD) [14], formerly known as plant blindness, has been gaining wide recognition. Wandersee and Schussler [15] raised concerns that the United States displayed less interest in plants than animals, thereby negatively affecting the public’s scientific literacy. They initiated a campaign to enhance the public’s understanding of plants, attributing PAD not only to a preference for animals but also to visual perception mechanisms that make plants more likely to be overlooked. In practice, participants have been found to recall more animals than plants in recall tasks [16,17], and more readily detect animals compared with plants in the attentional blink task [18]. These studies underscore that PAD is reflected not only in people’s knowledge and interests but also in their behaviour.

Considering the three theories of biophilia, AMH, and PAD, visual attention likely prioritises animals, followed by plants and then non-living organisms. Previous studies have reported findings consistent with this hypothesis. Jackson and Calvillo [19] found that in a visual search task, humans and animals, body parts and fruits, and tools and vehicles were detected the quickest, in that order. They also suggested that evolutionary relatedness, which indicates the extent to which humans are exposed to relevant stimuli throughout the evolutionary process, may significantly affect visual information processing. Although differences in attentional tendencies among the three groups were noted, a comprehensive understanding of the underlying mechanisms is still needed. One perspective indicates that people may respond to the probe faster because it serves as an alerting stimulus. Conversely, the challenge of diverting attention from an alert target may delay responses to other stimuli [20]. Thus, the AMH can be interpreted in two ways: animal-related stimuli elicit faster responses because they inherently demand attention or responses to non-animal stimuli may be delayed because attention remains preoccupied with the alert animal target.

The purpose of this study was to confirm the hierarchy in visual attention towards animals, plants, and manufactured objects. This study further sought to elucidate whether this hierarchy is driven by facilitated attention or difficulty in disengaging. To achieve these aims, we used a dot-probe task capable of distinguishing between facilitated attention and attentional disengagement difficulty. Specifically, we tested three hypotheses: (1) responses to presented stimuli are faster in the order of animals, plants, and manufactured objects, (2) attentional facilitation occurs towards animals and plants, and (3) attentional facilitation is stronger for animals than for plants. If these hypotheses are supported, it would demonstrate that responses to animals are faster owing to the attraction of visual attention, and that the attracting power is more robust for animals than for plants.

Experiment 1

Methods

Participants.

The Graduate School of Sustainable System Sciences Ethics Committee at Osaka Metropolitan University approved all procedures (2022 (1) – 36). The experimental protocol adhered to the latest version of the Declaration of Helsinki. Informed consent was obtained through the participant recruitment screen on the crowdsourcing platform. Participants indicated their understanding of the experiment’s details displayed on the screen and their willingness to participate by checking a designated box. The participants received financial compensation for their involvement, and all surveys were conducted online. The recruitment period for this experiment was from May 14 to May 23, 2023.

An a priori power analysis conducted using the R package Superpower [21] determined that a sample size of 47 was needed to obtain a power of over.80 (see S1 Appendix for details). To account for potential data exclusions, 80 participants were recruited. Among the 80 participants engaged through CrowdWorks (https://crowdworks.jp), three were excluded because of errors in saving their experimental data. Thus, data were analysed from 74 participants (27 female, 46 male, 1 non-response; M_age = 41.11 years, SD = 8.97, range = 22–58 years) were analysed, after excluding two who failed the instructional manipulation check and one who self-reported that their experimental data should not be included in the study.

Stimuli.

Object categorisation involves distinct cognitive neural processes that are contingent on an object’s conceptual level [21,22]. Birds were categorised as animals, fruits as plants [23], and vehicles and tools as manufactured objects [10,19], aligning with the superordinate categorisation of Rosch et al. [24] to equalise category hierarchies. Birds, fruit, and vehicles were designated target stimuli, whereas tools were considered neutral stimuli because the previous study reported slower reaction times towards tools than the other three categories [10,19]. The stimuli for birds were selected from the class Aves in the Linnaean taxonomy [25]. The Standard Tables of Food Composition in Japan were used to select fruits [26]. Land vehicles were chosen from class 12 (Vehicles; apparatus for locomotion by land, air, or water) defined in the Nice classification of the World Intellectual Property Organization [27]. Tools were subjectively selected by the authors, ensuring no overlap with other categories. Sixteen images, all depicting tools, were used in the practice trials. The main trials included 112 images in total: 64 tools, 16 birds, 16 fruits, and 16 vehicles. There was no overlap between the images used in the practice and main trials. All stimuli were static images obtained from Pixabay (https://pixabay.com/) and were resized to 260 x 260 px.

The effect of colour information on object recognition was stronger for colour diagnostic objects (that is, objects such as fire engines and lemons that appear in a consistent colour) than for non-colour diagnostic objects (by contrast, objects such as cars and hammers that do not consistently appear in a characteristic colour) [28,29]. The influence of colour-based information on neural activity during object processing was found to occur for fruits but not for animals or tools [23]. Moreover, because colour captures attention merely by being present on the screen, regardless of stimulus category [30–33], line drawings that excluded the effects of colour were used in the experiments.

As an index of the objective visual complexity of the stimuli, the compressed image file size was used (see Supplementary Materials S1 Table). The JPEG compressed file size correlates with subjective evaluations of visual complexity, indicating that images with larger file sizes are more complex [34–37]. An analysis of variance was conducted to examine the effect of stimulus category on visual complexity, yielding significant results (see Supplementary Materials S2 Table). Post hoc t tests revealed that fruits were significantly more complex than birds, whereas there were no significant differences between vehicles and fruits, or vehicles and birds. Additionally, birds, fruits, and vehicles were found to be more complex than tools. For the post hoc t-tests, p-values were adjusted using the Benjamini and Hochberg method to control the false discovery rate.

Procedure.

This experiment used the version of the dot-probe task version developed by Koster et al. [20], which incorporates neutral trials as a baseline for reaction time and enables differentiation between facilitated attention and attentional disengagement difficulty. The experimental programme was constructed using PsychoPy ver. 2022.1.2 [38] and hosted on Pavlovia.org (https://pavlovia.org/). The total duration of the experiment was approximately 20 minutes.

Fig 1 illustrates the sequence of a single trial. A blank screen was presented for 1000 ms, followed by a fixation point at the centre of the screen for 1000 ms. Two images were simultaneously displayed side-by-side following the disappearance of the fixation point. Stimulus onset asynchrony (SOA) was set at 100 ms and 500 ms with identical stimulus durations to assess automatic initial attention allocation and controlled attention allocation [39]. Subsequent to the image disappearing, the probe emerged in one of the two positions where the images were presented and remained visible until the participant responded. Participants were instructed to press the ‘F’ key as quickly as possible using their index finger if the probe appeared on the left side of the screen or the ‘J’ key if it appeared on the right.

Download:

Fig 1. Flow of trials for the dot-probe task.

https://doi.org/10.1371/journal.pone.0330475.g001

The images were paired and featured either one target and one neutral stimulus or two neutral stimuli. Pairings of target and neutral stimuli were presented, and trials were considered congruent if the probe appeared in the position of the target stimulus. Trials with the probe on the opposite side were deemed incongruent. Each block comprised 56 trials arranged in a 7 (trial type: bird congruent/incongruent, fruit congruent/incongruent, vehicle congruent/incongruent, neutral) × 2 (stimulus position: left, right) × 2 (probe position: left, right) × 2 (SOA: 100, 500) design. Each image was presented once within a block and the order and pairing combinations were randomised.

Participants completed eight practice trials, followed by four blocks, totalling 224 trials. Between blocks, the participants were allowed an arbitrary rest period. An instructional manipulation check was administered after the second block to identify participants who had not earnestly engaged in the experiment [40,41]. After completing an experiment, participants were provided with the following instruction: ‘Last, it is vital to our study that we only include responses from participants who devoted their full attention to this study. Otherwise, years of effort (the researchers’ efforts and other participants’ time) could be wasted. You will receive credit for this study no matter how you answer the next item. We appreciate your honesty!’ Participants were then asked to express their preference with a simple ‘yes’ or ‘no’ response to the following question: ‘In your honest opinion, should we use your data in our analyses for this study?’ This question was designed to efficiently assess participants’ attentiveness to the experiment [42].

Statistical analysis.

The attentional bias index (ABI), attentional facilitation index (AFI), and disengagement index (DI) were computed based on reaction times [43,44]. The ABI was defined as the reaction time in the incongruent trials minus the reaction time in the congruent trials. A positive ABI value indicated that attention was directed towards the target stimulus (vigilance), whereas a negative value signified that attention was turned away from the target stimulus (avoidance). The AFI was calculated by subtracting the reaction time in the congruent trials from the reaction time in the neutral trials. A positive AFI indicated that attention was facilitated towards the target stimulus, whereas a negative AFI suggested inhibited processing of the target stimulus. The DI was obtained based on reaction time in the neutral trials minus the reaction time in the incongruent trials, with a negative DI indicating difficulty disengaging attention from the target stimulus [45].

A within-subjects three-factor analysis of variance (ANOVA) (factors: category, congruency, and SOA) was applied to the reaction times. For the ABI, AFI, and DI, a one-sample t-test was conducted with the null hypothesis set to zero. Additionally, a within-subjects two-factor ANOVA (factors: category and SOA) was used to analyse each indicator. Mendoza’s multi-sample sphericity test [46] was used before the ANOVA, and degrees of freedom were adjusted using Greenhouse–Geisser’s ε [47] for factors violating the sphericity assumption. For one-sample and post hoc t-tests, p-values were adjusted using the false discovery rate method proposed by Benjamini and Hochberg [48]. Analyses were conducted using R software version 4.1.3 [49] and the ‘anovakun’ function version 4.8.9 [50] for ANOVAs. The significance level was set at.05 for all analyses.

Split-half reliabilities for reaction time, ABI, AFI, and DI were obtained using the splithalf package [51] for R. Reliability estimates were based on 5000 random splits and corrected with the Spearman–Brown formula [52,53].

Results

Data preparation.

No participant exhibited a correct response rate below 80% (M = 99.7%, range = 98.2–100%) [54]. The number of incorrect trials for each participant ranged from 0 to 4 (M = 0.65, SD = 0.99) and the incorrect trials were subsequently excluded from the analysis. Trials with reaction times shorter than 200 ms or longer than 2000 ms were also excluded (M = 0.01, SD = 0.12, range = 0–1) [20]. Values more than three standard deviations from the mean in each condition for each participant were identified as outliers and excluded. The number of outliers per participant varied from 0 to 5 (M = 1.39, SD = 1.26), and overall, 104 outliers were excluded from the analysis. Based on these procedures, 0.92% of the total trials were excluded from the overall dataset.

Reaction time.

Table 1 shows descriptive statistics for reaction times at 100 ms and 500 ms SOA. S3 Table provides the detailed results of the ANOVAs performed on reaction times. A significant main effect of SOA was observed, indicating that reaction times were slower in the 100 ms SOA condition (M = 479.6 ms, 95% CI [468.8, 490.4], SD = 115.6) compared to the 500 ms condition (M = 449.5 ms, 95% CI [439.3, 459.6], SD = 109.1; F(1, 73) = 165.26, p < .001, η_p² = .694). Furthermore, a significant main effect of category was identified with adjustments made to the interaction between category and congruency. However, the main effect of congruency and the other interaction effects were not significant. For congruent trials, reaction times were slower for vehicles (M = 472.3 ms, 95% CI [453.7, 490.9], SD = 114.5) than for birds (M = 461.7 ms, 95% CI [443.1, 480.4], SD = 115.0; t(147) = −6.24, p < .001, d_z = −.092) or fruits (M = 462.0 ms, 95% CI [443.9, 480.0], SD = 111.2; t(147) = 5.15, p < .001, d_z = .091), with no significant difference between birds and fruits(t(147) = −0.13, p = .896, d_z = −.002). Reaction times for vehicles were slower in congruent trials (M = 472.3 ms, 95% CI [453.7, 490.9], SD = 114.5) than in incongruent trials (M = 463.3 ms, 95% CI [445.1, 481.5], SD = 112.1; t(147) = 4.83, p < .001, d_z = .079). The simple effect of category on incongruent trials and the simple effect of congruency on birds and fruits were not significant.

Download:

Table 1. Descriptive statistics for reaction times (ms) in Experiment 1.

https://doi.org/10.1371/journal.pone.0330475.t001

The permutation-based split-half reliability for reaction time was very high across all categories, regardless of SOA or congruency (Spearman–Brown reliability = 0.98, 95% CI [0.95, 0.99], see Supplementary Materials S4 Table for details).

Attention bias index.

Fig 2 presents the results of the ABI, AFI, and DI calculations. The detailed results of the one-sample t-test performed on each indicator are shown in S5 Table, and the detailed results of the ANOVA are shown in S6 Table. The one-sample t-test results revealed negative ABI values for vehicles in the 100 ms and 500 ms SOA conditions. No significant ABIs were observed for either birds or fruits. An ANOVA demonstrated a significant main effect of category, indicating that the ABI was smaller for vehicles (M = −9.0 ms, 95% CI [−12.7, −5.3], SD = 22.7) than for birds (M = 1.4 ms, 95% CI [−2.1, 5.0], SD = 21.8; t(147) = 4.43, p < .001, d_z = .469) or fruits (M = 2.8 ms, 95% CI [−0.2, 5.9], SD = 18.9; t(147) = −4.47, p < .001, d_z = −.568). However, no significant differences were found between birds and fruits (t(147) = −0.57, p = .569, d_z = −.069). The main effects of SOA and the interaction between category and SOA were not significant. These results suggest that participants’ attention was drawn towards neutral stimuli rather than vehicles. However, it remains unclear whether this was due to attention facilitation in the neutral condition or avoidance of vehicles. This will be further investigated in the subsequent analyses of AFI and DI.

Download:

Fig 2. Raincloud plots of the attentional bias index (ABI, left), attentional facilitation index (AFI, centre), and disengagement index (DI, right) in Experiment 1.

The results in the 100 ms SOA condition are in the top row, and the results in the 500 ms SOA condition are in the bottom row. Black dots in the cloud indicate mean values and error bars indicate 95% confidence intervals. SOA = stimulus onset asynchrony.

https://doi.org/10.1371/journal.pone.0330475.g002

The outcomes of a one-sample t-test showed a negative AFI for vehicles in both the 100 ms and 500 ms SOA conditions, whereas no significant AFI was observed for birds or fruits. An ANOVA indicated a significant main effect of category; however, the main effects of SOA and the interaction between category and SOA were insignificant. Post hoc comparisons revealed that vehicle processing (M = −8.5 ms, 95% CI [−12.4, −4.6], SD = 24.1) was more inhibited than bird (M = 2.1 ms, 95% CI [−1.8, 5.9], SD = 23.8; t(147) = 6.24, p < .001, d_z = .441) or fruit processing (M = 1.8 ms, 95% CI [−2.8, 6.4], SD = 28.4; t(147) = −5.15, p < .001, d_z = −.389), with no significant differences between birds and fruits (t(147) = 0.13, p = .896, d_z = .009).

DI was not subjected to an ANOVA because a one-sample t-test did not yield significant values for either condition. The Spearman–Brown reliability for ABI, AFI, and DI ranged from −0.52 to 0.23, indicating low internal consistency (see Supplementary Materials S7 Table for details).

Discussion

In Experiment 1, the dot-probe task was used to investigate attentional bias towards birds, fruits, and vehicles. In congruent trials, reaction times were longer for vehicles than for birds or fruits. Based on the ABI and AFI values, this discrepancy arose from the inhibition of vehicle processing and attentional avoidance away from vehicles. These findings were consistent across both the 100 ms and 500 ms SOA conditions, indicating that attentional avoidance occurred in automatic and controlled attention allocation.

In contrast, no discernible attentional bias was observed for birds or fruits, rendering the present results incongruent with both the AMH and PAD. Loucks et al. [55] noted that most AMH studies primarily employed mammals as stimuli, suggesting that animals bearing more remarkable similarities to humans are preferentially processed. Given that birds are represented in the brain at greater distances from humans than mammals [56], the AMH and PAD may not have been supported because of the absence of preferential processing for birds. Consequently, Experiment 2 was conducted using the same methodology but with mammals instead of birds to further explore this phenomenon.