Context-Dependent Modulations of Spatial Attention: A Simulated Driving Task

doi:10.21203/rs.3.rs-5010340/v1

Download PDF

Research Article

Context-Dependent Modulations of Spatial Attention: A Simulated Driving Task

https://doi.org/10.21203/rs.3.rs-5010340/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Human attention can be guided by semantic information of individual objects in the environment (e.g., traffic light). In the current study, we reported a novel phenomenon in which certain behavioural contexts could determine whether guidance of attention occurs. Using a driving simulator, we simulated a real-world driving task representing an over-learned behaviour. A peripheral cue-target paradigm was implemented where a roadside pedestrian avatar (target) appeared following a cylinder cue. Results revealed that, during simulated driving, participants (all with driver’s licenses) showed increased attentional facilitation when pedestrians were oriented toward the road compared to away. This orientation-specific selectivity was not seen if the 3-D context was removed, or the same visual scene was presented but participants’ viewpoint remained stationary, or an inanimate object served as a target during simulated driving. This context-specific attention modulation likely reflects drivers’ expertise in automaticallyattending to behaviourally relevant information in a context-dependent manner.

Psychology

Cognitive Neuroscience

context-specific behaviours

altercentric cognition

spatial cueing

experience-driven attention

driving

ecological validity

attentional learning

The present research highlights how our attention can be influenced by familiar real-world contexts. We found that drivers are more likely to focus on pedestrians oriented toward the road compared to those oriented away from the road, suggesting a learned defensive behaviour. This attentional bias only appeared in driving-like scenarios and when in the presence of pedestrians, indicating that our past experiences on the road shape how we pay attention to potential hazards. The modulation of attention likely takes place well before motor planning and subsequent overt motor behaviours (such as easing on gas and preparing to change lanes or brake) and can potentially offer a set of sensitive testing regimes for driving expertise. These insights are crucial for understanding how drivers perceive and respond to their environment, which can inform the design of safer road systems and driver training programs. Additionally, this research may have implications for understanding attentional differences in clinical populations, such as individuals with autism or anxiety disorders, who might process these cues differently.

Within the visual world, humans are constantly exposed to complex environments composed of a multitude of stimuli requiring varying levels of attention. Given our limited attentional resources, observers rapidly select which stimuli warrant attending. A recent surge in the literature has shown that the allocation of attention within dynamic visual scenes can be driven by factors such as behavioural relevance, scene semantics, or emotional significance (e.g., Beitner et al., 2021, Vater et al., 2022; Yu et al., 2023).

It has been well established that attention can be biased towards individual objects that convey inherent meaning. For example, a face with an angry expression attracts more attention compared to a face with a neutral expression (Okon-Singer et al., 2020). In addition, neutral, arbitrary objects without such inherent property that attract attention can become behaviourally relevant through learning in a specific context. An abstract experimental stimulus (e.g. a black square) that has been previously associated with a positive or negative outcome (e.g., monetary rewards, electric shock) can gain preferential attention in behavioural tasks when compared to objects that are not exposed to such association (Anderson et al., 2021; Failing & Theeuwes, 2018; Grégoire et al., 2021; Le Pelley & Newell, 2023; Schmidt et al., 2015).

Learning about the behavioural relevance of otherwise neutral objects can occur outside the laboratory, such as in naturalistic settings that occur through daily behaviour (Helbing et al., 2020; Kristjánsson & Draschkow 2021). Thus, to better capture realistic stimulus-context learning, researchers have demonstrated that real-world stimuli could elicit attentional guidance patterns that relate to previous experiences by participants in their lifetime. For example, researchers have presented participants with natural three-dimensional (3-D) scenes (e.g., a living room) and observed where participants fixated throughout the scene. Hayes and Henderson (2022) revealed that when freely viewing 3-D scenes, participants’ patterns of fixation were predominantly concentrated in locations of the scene that would be behaviourally relevant (e.g., a staircase) compared to less behaviourally relevant areas (e.g., empty corner of the room).

In experimental settings with more specific task requirements, it has been demonstrated that the learned selectivity of attention could be context-dependent. The most widely studied context is using action to induce a motor-specific context. When anticipating an upcoming action, attention to features relevant to the action is facilitated within intermediate attention tasks. For example, Fagioli and Hommel (2007) demonstrated that while preparing for either a grasping or pointing action, concurrent perceptual task performance was facilitated for target features associated with either action: better target size processing before grasping action and better target location processing before pointing action (also see Feldmann-Wüstefeld & Schubö, 2015; Han et al., 2020; Job et al., 2019).

In Fagioli and Hommel (2007), the action context is provided through overt behaviours (grasping and reaching) required in the experiment. It has also been shown that attention modulation can be achieved by visual stimuli implying motor action despite the possibility of the actual motor act being very low. This is due to internal models of behaviour being intrinsically activated when observing implied actions (e.g., seeing a posture indicative of a movement), oftentimes biasing attention toward locations or features that we expect to be the result of their implied action (Bach & Schenke, 2017; Bach et al., 2014). As an example, explicitly threatening body postures have been shown to reflexively guide attention as participants likely anticipate the outcome of certain postures (e.g., a kick to the left) despite never observing any actions (Azarian et al., 2016; Bannerman et al., 2010; Wang et al., 2018). These results highlight that the accumulation of experiences with explicit actions, regardless of the experimental context or the commitment/observation of any actions, can result in attentional guidance due to implicit behavioural ramifications.

In the studies described above, the modulation of attention was triggered by individual action-related stimuli alone (e.g., a body posture with implied action). In the current study, we intended to explore whether attention could be modulated by a broader task context while the target by itself does not offer motor implications without such context. In other words, the implied action is only present with a combination of the certain behavioural context and key stimulus feature that would otherwise not imply action.

We chose driving as the testing context as driving is an over-learned behaviour and experienced drivers would likely develop an “attention set” that would optimize the preparation for possible motor response in case there is a need (Crundall et al., 2012). In particular, this well-developed attention set would be activated automatically even in a simulated driving task. Moreover, such selectivity would only be apparent in certain contexts and not in other (non-driving) contexts. In addition, it is likely drivers would modulate their allocation of attention well before there is a need for overt motor response (e.g., change direction or speed of driving) even when the probability of the need for this motor action is very low.

In the current study, we used a modified spatial cueing paradigm where a peripheral spatial cue preceded the onset of the roadside target. In a typical version of this paradigm, participants respond to the onset of a target appearing in the periphery visual field following the onset of a spatially uninformative cue in the same or a different location as that of the target. A great number of studies have examined that, compared to a neutral stimulus, how stimulus with negative effects (e.g., angry faces, snakes, threatening words), acting as a cue or target, could influence target detection, localization and discrimination/categorization (e.g., Okon-Singer et al., 2020; Pérez-Dueñas et al., 2014).

Spatial cueing tasks are a more sensitive measure of attention across depth (see Britt & Sun, 2024; Chen & Wyble, 2018). For our task, a neutral cue (a cylinder) was presented before the target presentation in the same or opposite hemifield. In the experimental condition of the current study, licenced drivers drove in a driving simulator and moved forward along a straight road in a 3D virtual reality environment. The target was a pedestrian-like avatar appearing on the roadside some distance ahead of the driver’s viewpoint. We measured drivers’ performance (reaction time) in discriminating a target feature (pedestrian’s arm position) that was not directly relevant to driving.

The most important (task-irrelevant) manipulation in the current study was the orientation of the pedestrian, which could be either facing toward the road (inward) or facing away (outward) from the road. Note that although in the current experiment, the pedestrian never moved, in the real world, with the driver moving along the road combined with the pedestrian’s inward orientation, in theory, there is a possibility of an impending collision between the vehicle and the pedestrian, should the pedestrian step out from the roadside—pedestrians are a common roadside hazard while driving (Song et al., 2023). In contrast, the impending collision would be unlikely when the pedestrian assumed the outward orientation. Therefore, the performance difference when the orientation of the pedestrian was inward vs outward would be the indication of modulation of attention in a context-dependent manner. We anticipated that inward orientation would garner more attention as a safety precaution due to it being perceived as an implicit hazard.

In addition to the experimental condition implemented in each of the three experiments in the current study, we also implemented three sets of control conditions where the target orientation was also varied between inward and outward orientations, but the behavioural context was removed through three different ways: 2-D stationary context with an isolated pedestrian target (Experiment 1), 3-D stationary context with a pedestrian target (Experiment 2), and driving with an inanimate Light-Post target (Experiment 3). See Fig. 1.

The control condition in Experiment 1 intended to isolate whether the difference in visual appearance of the target alone (without 3-D context) between inward and outward orientation would generate a difference in processing speed. The control condition in Experiment 2 intended to test whether attention modulation could still take place with the same 3-D information as that in the experimental condition, except that the motor act of driving is absent. The control condition in Experiment 3 intended to examine whether the forward self-motion alone would lead to attention bias.

Participants

All participants in each experiment reported normal or corrected-to-normal visual acuity, and colour vision and were right-handed. All participants reported having a valid driver’s license and received course credit for their participation. Participants also completed an informed consent form before participating. The study was approved by the McMaster Research Ethics Board. For each experiment, power analysis (Lakens & Caldwell, 2021) revealed a minimum n = 24 for 90% power to examine a cue validity x condition x orientation interaction in RT based on previous cueing effect magnitudes across depth (Britt et al., 2024).

Experiments 1, 2, and 3 involved 28 (18–36 years of age; n = 22 female, n = 6 male), 24 (18–21 years of age; n = 18 female, n = 6 male), and 29 (17–28 years of age; n = 19 female, n = 10 male) undergraduate students, respectively.

Stimuli and apparatus

The stimuli were back-projected onto a screen (207.4 cm tall x 144.8 cm wide) positioned 150 centimeters (cm) in front of the participants. The entire setup was positioned inside a dark tent with minimal visual interference. The setup of a steering wheel device and gas/brake pedals were positioned like that in a regular car, and the participants sat at a comfortable distance from the device. The steering wheel is also equipped with response buttons.

This study presented depth information through monocular cues in a virtual 3-D environment. The 3-D virtual simulation was created using Vizard 7.0 virtual reality software. As shown in Fig. 1, the virtual environment consisted of a ground plane and two sets of upright parallel walls (2.75 vm height, 1 vm thickness, 15 vm length) oriented along the z-axis in the periphery. On both the left and right sides, there were multiple wall pieces with gaps of 6 vm between adjacent pieces. These wall pieces acted as left and right boundaries of a road on the ground and the cue and target objects appeared in the gaps.

Across all three experiments, the identical experimental condition was implemented. Participants in this condition were asked to use gas and brake pedals to undergo simulated driving (forward self-motion along the “road”) to follow a red lead vehicle (which also served as a point of fixation) and maintain a constant following distance (about 20 vm), creating a viewing angle of 4.57° in width, and 3.43° in height. The maximum driving speed of the red lead car and the participant’s car was set to be 10 m/s (acceleration = 4.5 m/s²).

In addition to the car-following task, in a given trial, a red peripheral cylinder cue (2.66° tall, 0.57° wide) first appeared on the roadside (inside the gap between wall pieces) at a horizontal eccentricity of 8.29° when the participants’ viewpoint was 43 vm in z-axis from the cue position. The pedestrian target (2.66° tall, 0.57° wide) then appeared at the same z-axis location (in world-centered coordinate) as the cue and with an eccentricity of 9.09° when the participants’ viewpoint was 30 vm (in the z-axis) away. The pedestrian avatar maintained a standing (not walking) post and appeared with a green shirt and beige pants. Importantly, the pedestrian’s body was oriented either toward the road or away from the road and could either have their arm (farthest one from the driver) up in the air or down at their side. Pedestrians were rotated around the y-axis 33° (inward) and 147° (outward), with 0° being facing perpendicular to the driving path, so that both pedestrian orientations appeared completely perpendicular to the participant's line of sight at 100 ms after the target onset. This was intended to create a mirror image of the target for the two orientations. The same experimental condition was implemented in each of the three experiments and compared to three different control conditions implemented in the three experiments respectively.

For Experiment 1, the 2-D control condition included only the cue and target implemented in the experimental condition and all the context for the 3-D environment was removed. The red fixation vehicle was replaced by a size-matched white box. In the periphery, one solid white box (width = 2.24°, height = 2.86°) appeared on either side of the fixation. The same cue and target as implemented in the experimental condition were presented with matching eccentricities within the peripheral white boxes.

In Experiment 2, the 3-D control condition included the central red fixation vehicle positioned 20 vm in front of the participant’s viewpoint. The participant’s viewpoint remained stationary at all times and was positioned 30 vm (along the z-axis) away from the cue and target locations. The peripheral cue in the shape of a red cylinder and peripheral pedestrian target appeared on either the left or right side of the display at identical horizontal eccentricities of 9.09°.

Lastly, in Experiment 3, the control condition was identical to the experimental condition except that the pedestrian target was replaced by an inanimate light-post target stimulus (2.66° tall, 0.71° wide). The light-post target contained two horizontal “arms”, positioned in the top and middle. A yellow square ‘light’ (0.28° tall, 0.28° wide) could appear on either arm at either the inward or outward positions. Overall, the four possible arm positions could match the configuration of the pedestrians’ arms.

Procedure

Before beginning the experimental trials, participants completed an 18-trial practice block containing their context(s) to become familiar with the environment and the response procedures. For the experimental condition, participants began a trial by fixating on the central red vehicle. Participants were instructed to press the gas pedal with their foot at the trial onset to follow the lead red car vehicle travelling at a max speed of 10 m/s. After approximately 1000 ms, a cue appeared on either the left or right side for 150 ms. Following a random stimulus onset asynchrony (SOA) between 1000–1400 ms, the target appeared on either the left or right side for 200 ms. Participants were instructed to respond (as fast and as accurately as possible) to the pedestrian's arm position with an ‘up’ or ‘down’ button by pressing one of the two buttons with their right hand on the steering wheel. In all conditions, participants were instructed not to make eye movements to the best of their ability and to remain focused in the center.

Design

Experiment 1 used a within-subjects design where each participant took part in both the 2-D stationary control and experimental conditions. Participants underwent 416 total formal testing trials (208 per condition). The experiment consisted of 26 blocks each containing 16 trials. Within each block, participants underwent an even number of left and right cue positions and target positions. The experiment was split in half so that the first 13 blocks involved the 2-D control while the last 13 involved the experimental condition, or vice versa. The order of conditions was randomized across participants.

Experiment 2 used a between-subjects design where each participant took part in either the 3-D stationary control or experimental conditions. The between-subject design was implemented in this experiment to ensure the strategy of attention allocation would not contaminate the response of the subsequent conditions should we implement a within-subject design. After all, the visual appearance of the environment in the two conditions was very similar. In either condition, participants underwent 416 total experimental trials. The experiment consisted of 26 blocks each containing 16 trials.

Experiment 3 used a within-subject design where both target types (Pedestrian, Light-Post) were blocked, and participants completed all trials with one type of target before the other. Participants underwent 416 total trials (208 per target). The number and order of blocks/trials were identical to Experiment 1.

RT data will be reported in order of experiments (Exp1, Exp2, Exp3). RTs longer than 1500 ms and shorter than 200 ms (2.05%, 3.25%, 2.96%), responses before target onset (0.14%, 0.17%, 0.25%), trials with SOAs outside the expected range (0.36%, 0.20%, 0.37%), as well as incorrect discrimination responses (3.02%, 1.56%, 1.53%), were excluded. Outlier data (3.75%, 3.93%, 3.85%) were also removed on either side of the RT distribution using a highly conservative threshold criterion of three times the mean absolute deviation of the median raw RT data per participant (Leys et al., 2013). Altogether, approximately 9.35%, 9.08%, and 8.96% of the trials were removed per experiment, respectively. Given error rates remained above 96% across all conditions it did not necessitate accuracy analyses.

Experiment 1: Experimental vs 2-D Stationary Control Condition

Reaction Times

Mean RT values were entered into a 2 (cue validity: valid vs. invalid) x 2 (orientation: inward vs outward) x 2 (condition: experimental vs 2-D control) type III univariate three-way repeated-measures ANOVA model. The ANOVA revealed a significant main effect of cue validity (F(1, 27) = 14.57, p < .001, η_p² = .35), a nonsignificant main effect of orientation (F(1, 27) = 3.15, p = .087, η_p² = .10), and a significant main effect of condition (F(1, 27) = 9.28, p = .005, η_p² = .26). In addition, significant interactions were found for cue validity x condition (F(1, 27) = 13.22, p = .001, η_p² = .33) and most importantly for three-way cue validity x orientation x condition (F(1, 27) = 4.68, p = .039, η_p² = .15). The cue validity x orientation (F(1, 27) = 3.06, p = .091, η_p² = .10) and orientation x condition (F(1, 27) = 1.87, p = .183, η_p² = .06) failed to reach significance¹.

¹Despite the insignificant orientation x condition interaction, we performed simple main effect analyses for the effect of pedestrian orientation in each condition and revealed a simple main effect of pedestrian orientation in the experimental condition (F(1, 27) = 6.21, p = .019, η_p² = .19) but not in the 2-D control condition F(1, 27) = 0.33, p = .571, η_p² = .01)—aligning with participants reacting quicker to inward-oriented pedestrians in the experimental condition only.

The main effect of cue validity revealed faster overall RT for validly cued trials (643.5ms) compared to invalid (668.2ms). But to further investigate the cue validity x condition interaction, a simple main effect analysis was conducted on cue validity within either condition. For the 2-D control condition, a one-way repeated measures ANOVA revealed no simple main effect of cue validity (F(1, 27) = 1.97, p = .172, η_p² = .06). However, the experimental condition revealed a significant simple main effect of cue validity (F(1, 27) = 24.1, p < .001, η_p² = .47). These results indicate that while the 2-D control condition revealed no significant facilitation effects, the experimental condition demonstrated the significant facilitation effect. To explore the significant three-way interaction, we completed the planned cueing effect analysis.

Cueing Effects

To explore the significant cue validity x orientation x condition interaction, we converted the RT measurements into their respective cueing effects. Cueing effects were calculated as the RT difference between invalidly cued trials and validly cued trials (RT_Invalid – RT_Valid). We then entered the cueing effects as the dependent variable to explore the simple main effect of pedestrian orientation within each experimental condition. Note, as previously mentioned the facilitation effects in the 2-D condition did fail to reach significance.

We then explored the simple main effect of pedestrian orientation (inward, outward) on the cueing effects in the experimental and 2-D control conditions. The ANOVAs revealed a significant simple main effect of pedestrian orientation in the experimental condition (F(1, 27) = 7.23, p = .012, η_p² = .21) but not in the 2-D control (F(1, 27) = 0.11, p = .741, η_p² = .004). While driving, participants had significantly greater facilitation effects when the pedestrian was facing inward (50.5ms) compared to outward (28.6ms) but this effect was not present in the 2-D condition (inward: 7.8ms, outward: 10.3ms).

Experiment 2: Experimental vs 3-D Stationary Control Condition

Reaction Times

Mean RT values were entered into a 2 (cue validity: valid vs. invalid) x 2 (orientation: inward, outward) x 2 (condition: experimental vs 3-D control conditions) type III three-way mixed analysis of variance (ANOVA) model. The ANOVA revealed a significant main effect of cue validity (F(1, 22) = 22.05, p < .001, η_p² = .50), a significant main effect of orientation (F(1, 22) = 17.93, p < .001, η_p² = .45), and a significant main effect of condition (F(1, 22) = 39.43, p < .001, η_p² = .64). In addition, significant interactions were found for cue validity x orientation (F(1, 22) = 8.57, p = .008, η_p² = .28) and most importantly for cue validity x orientation x condition (F(1, 22) = 4.87, p = .038, η_p² = .18). The orientation x condition (F(1, 22) = 1.39, p = .251, η_p² = .06) and cue validity x condition (F(1, 22) = 4.08, p = .056, η_p² = .15) failed to reach significance.

The main effect of the condition revealed significantly faster RT for the 3-D stationary control condition (627.4ms) compared to the experimental condition (792.0ms). In addition, the main effect of cue validity (and insignificant cue validity x condition interaction) revealed significantly faster overall RT for validly cued trials (control: 615.4ms, experimental: 763.8ms) compared to invalidly cued trials (control: 639.5ms, experimental: 820.8ms). Also, the main effect of orientation (and insignificant orientation x condition interaction) revealed significantly faster RTs for pedestrian stimuli facing inward (i.e., toward the road) (control: 618.4ms, experimental: 786.5ms) compared to facing outward (control: 636.5ms, experimental: 797.6ms).

Cueing Effects

Similar to Experiment 1, using one-way ANOVAs the analysis revealed a significant simple main effect of pedestrian orientation (inward vs outward) on the cueing effects in the experimental condition (F(1, 11) = 8.77, p = .013, η_p² = .44) but not in the 3-D stationary control condition (F(1, 11) = 0.52, p = .485, η_p² = .04). While driving, the participants had significantly greater facilitation effects when the pedestrian was facing inward (73.6ms) compared to facing outward (45.6ms), but not when stationary (inward: 25.7ms, outward: 21.8ms).

Experiment 3: Experimental vs Light-Post Target Control Condition

Reaction Times

Mean RT values were entered into a 2 (cue validity: valid vs. invalid) x 2 (orientation: inward, outward) x 2 (condition: experimental vs ‘Light-Post’ control condition) type III univariate three-way repeated-measures ANOVA model. The ANOVA revealed all main effects and interactions to be significant: cue validity (F(1, 28) = 12.47, p = .001, η_p² = .31), orientation (F(1, 28) = 15.0, p < .001, η_p² = .35), condition (F(1, 28) = 23.56, p < .001, η_p² = .46), cue validity x condition (F(1, 28) = 9.06, p = .005, η_p² = .24), cue validity x orientation (F(1, 28) = 9.50, p = .005, η_p² = .25), orientation x condition (F(1, 28) = 21.34, p < .001, η_p² = .43), and most importantly cue validity x orientation x condition (F(1, 28) = 9.99, p = .004, η_p² = .26).

The main effect of cue validity revealed faster overall RT for validly cued trials (655.9ms) compared to invalid (697.2ms). Further simple main effect analysis was conducted to observe if both conditions elicited significant facilitation effects. One-way repeated measures ANOVAs revealed simple main effects of cue validity in both the Light-Post control (F(1, 28) = 8.79, p = .006, η_p² = .24) and experimental conditions (F(1, 28) = 14.60, p < .001, η_p² = .34)—consistent with facilitation effects in both conditions.

Cueing Effects

As in Experiments 1 and 2, the cueing effect analysis revealed a significant simple main effect of pedestrian orientation (inward vs outward) in the experimental condition (F(1, 28) = 18.04, p < .001, η_p² = .39) but not with the light-post target (F(1, 28) = 0.04, p = .834, η_p² = .002). This again revealed that while driving, participants had significantly greater facilitation effects when the pedestrians were facing inward (64.4ms) compared to outward (35.4ms) but this effect was abolished when the pedestrian was replaced by the light-post (inward: 31.6ms, outward: 32.9ms).

Table 1

The descriptive statistics for Experiments 1, 2, and 3. Reaction times and cueing effects (facilitation) are represented in milliseconds. Standard error values are corrected for within-subject variability (Morey, 2008).
	Valid RT Invalid RT	Cueing Effect	Error Rate %	Valid RT Invalid RT	Cueing Effect	Error Rate %
Exp 1	2-D Stationary Control			Experimental Condition
Inward	632.4 (4.0) 640.2 (3.8)	7.8 (5.4)	3.36	645.7 (4.0) 696.2 (4.1)	50.5 (7.3)	1.99
Outward	634.9 (3.7) 645.2 (3.9)	10.3 (7.5)	3.88	671.3 (3.9) 699.9 (4.1)	28.6 (6.1)	2.78
Exp 2	3-D Stationary Control			Experimental Condition
Inward	606.4 (3.8) 632.1 (3.8)	25.7 (9.6)	0.68	752.2 (4.7) 825.8 (5.1)	73.6 (18.0)	2.16
Outward	626.3 (3.9) 648.1 (4.0)	21.8 (6.4)	0.84	776.3 (4.8) 821.9 (5.0)	45.6 (15.2)	2.52
Exp 3	Light-Post Target Control			Experimental Condition
Inward	637.9 (3.7) 669.5 (3.9)	31.6 (11.5)	1.62	654.7 (4.3) 719.1 (4.7)	64.4 (13.1)	2.79
Outward	667.9 (3.8) 700.8 (4.0)	32.9 (10.9)	1.78	666.7 (4.2) 702.1 (4.3)	35.4 (13.9)	2.80

Across all three experiments, our spatial cueing data aligned with our predictions: cueing effects were significantly higher for pedestrians oriented inward (i.e., toward the road) compared to outward, but only in the experimental condition. This pattern of facilitation was not present in any of the three control conditions in the absence of the combination of context and target feature. Collectively, consistent difference in cueing effect between the experimental and control conditions suggests that it was the contextual information that was able to evoke previously learned stimulus-context associations only when the context mimicked real-world driving conditions with roadside pedestrians.

In a driving situation, drivers are constantly exposed to many objects and events, and being able to strategically allocate attention to driving relevant events is very important. Attention modulation is the early process and perhaps always ongoing during driving. Overall, in the driving condition, expertise in modulating attention among visual events could be developed over years of driving experience and eventually, the allocation process becomes an automatic process.

The long-term learning of stimulus-context associations by licensed drivers that drove the selective attention focus on the pedestrian with inward orientation during driving context reveals behaviour consistent with models of attentional learning (Hall & Rodrigues, 2019; Kruschke, 2011; Le Pelley et al., 2016). These models suggest, just like our results, that attention can be selectively tuned to be predominantly distributed according to prior experiences (i.e., spatial or feature priors) that can be context-sensitive (Bar, 2004; Fischer et al., 2020; also see Zhou & Geng, 2024).

The nuance of our results is that the potential hazard of an inward-oriented pedestrian can bias attention when contextual information implies behavioural ramifications. Unlike the aforementioned literature showing planned actions (e.g., Fagioli & Hommel, 2007) or viewing action images (e.g., Schenke et al., 2016) can effectively bias attention toward action-congruent features or locations, the present results reveal a similar pattern, but only when both the stimulus and behavioural context are available. Here, based solely on the probabilities associated with learned characteristics of driving behaviour, the task-irrelevant orientation of the pedestrian target stimuli modulated attention as the experimental scenario suggested a potential roadside collision. Thus, we have furthered these theories of attentional learning by indicating a learned preparation effect based on conditional probability. Specifically, when inferring a potentially hazardous situation, previous memories from aversive experiences are easily accessible, prioritized over non-aversive memories, long-lasting, and can be retrieved by inherently neutral cues (Baczkowski et al., 2023; Bowen et al., 2018; Dunsmoor et al., 2015; Kalbe & Schwabe, 2021). We believe the combination of our inherently neutral pedestrian stimulus within a driving condition rapidly selected memory representations to effectively guide attention to act in a predominantly defensive manner.

Furthermore, a secondary finding of this study that warrants explanation was that the inward orientation resulted in significantly faster RTs for all conditions involving a pedestrian except the 2-D control condition (showing an insignificant simple main effect of orientation). We believe this evidence supports an altercentric account of attention via our social pedestrian targets (Kampis & Southgate, 2020). Altercentric attention would predict that the orientation of nearby social agents modulates the focus of attention of the observers. We believe the RT data accurately reflects this spatial relation of the participant and pedestrian, and the disappearance of this effect of orientation only in the 2-D control condition would reflect the disappearance of depth information, thus abolishing the congruent alignment of the pedestrian orientation and participants’ central fixation.

Overall, the difference in cueing effect between the experimental and control conditions found in our study suggests that contextual information could be able to evoke previously learned stimulus-context associations. From an applied perspective, it is perhaps worthwhile to develop a battery of testing scenarios like the one in this study for attention allocation during driving as this type of test could offer a more sensitive measure for driving expertise. This kind of testing would offer greater predictive value compared to examining drivers’ overt motor behaviours during driving.

Declarations and Research Transparency

Funding

Financial support for this study was provided by the Natural Science and Engineering Research Council of Canada (NSERC) grant 20021498.

Conflicts of interest/Competing interests

Not applicable.

Artificial Intelligence

No AI was used for the manuscript.

Ethics Approval and Consent

The McMaster University Research Ethics Board (MREB) approved the study. Participants received and completed an informed consent form before research participation.

Availability of data

The datasets are available at https://osf.io/3xpwz/?view_only=d0113c37d0f94509b2028633c75ff611 . The data will be made public upon publication.

Acknowledgements

We want to thank Brian Wong for his involvement during data collection, and Juan Lupiáñez and Bruce Milliken for discussions and suggestions.

Open Practices Statement

This study’s design and its analysis were not pre-registered.

Anderson, B. A., Kim, H., Kim, A. J., Liao, M.-R., Mrkonja, L., Clement, A., & Grégoire, L. (2021). The past, present, and future of selection history. Neuroscience & Biobehavioral Reviews, 130, 326–350. https://doi.org/10.1016/j.neubiorev.2021.09.004
Azarian, B., Esser, E. G., & Peterson, M. S. (2016a). Evidence from the eyes: Threatening postures hold attention. Psychonomic Bulletin & Review, 23(3), 764–770. https://doi.org/10.3758/s13423-015-0942-0
Bach, P., Nicholson, T., & Hudson, M. (2014). The affordance-matching hypothesis: How objects guide action understanding and prediction. Frontiers in Human Neuroscience, 8. https://doi.org/10.3389/fnhum.2014.00254
Bach, P., & Schenke, K. C. (2017). Predictive social perception: Towards a unifying framework from action observation to person knowledge. Social and Personality Psychology Compass, 11(7), e12312. https://doi.org/10.1111/spc3.12312
Baczkowski, B. M., Haaker, J., & Schwabe, L. (2023). Inferring danger with minimal aversive experience. Trends in Cognitive Sciences, 27(5), 456–467. https://doi.org/10.1016/j.tics.2023.02.005
Bannerman, R. L., Milders, M., & Sahraie, A. (2010). Attentional cueing: Fearful body postures capture attention with saccades. Journal of Vision, 10(5), 23–23. https://doi.org/10.1167/10.5.23
Bar, M. (2004). Visual objects in context. Nature Reviews Neuroscience, 5(8), 617–629. https://doi.org/10.1038/nrn1476
Beitner, J., Helbing, J., Draschkow, D., & Võ, M. L.-H. (2021). Get Your Guidance Going: Investigating the Activation of Spatial Priors for Efficient Search in Virtual Reality. Brain Sciences, 11(1), 44. https://doi.org/10.3390/brainsci11010044
Bowen, H. J., Kark, S. M., & Kensinger, E. A. (2018). NEVER forget: Negative emotional valence enhances recapitulation. Psychonomic Bulletin & Review, 25(3), 870–891. https://doi.org/10.3758/s13423-017-1313-9
Britt, N., & Sun, H. (2024). Spatial Attention in Three-Dimensional Space: A Meta-Analysis for the Near Advantage in Target Detection and Localization. Neuroscience & Biobehavioral Reviews, 105869. https://doi.org/10.1016/j.neubiorev.2024.105869
Britt, N., Haponenko, H., Cochrane, B., Milliken, B., & Sun, H.-J. (2024). Distribution of Attention in Three-Dimensional Space. https://doi.org/10.31234/osf.io/hkems
Chen, H., & Wyble, B. (2018). The neglected contribution of memory encoding in spatial cueing: A new theory of costs and benefits. Psychological Review, 125(6), 936–968. https://doi.org/10.1037/rev0000116
Crundall, D., Chapman, P., Trawley, S., Collins, L., van Loon, E., Andrews, B., & Underwood, G. (2012). Some hazards are more attractive than others: Drivers of varying experience respond differently to different types of hazard. Accident Analysis & Prevention, 45, 600–609. https://doi.org/10.1016/j.aap.2011.09.049
Dunsmoor, J. E., Murty, V. P., Davachi, L., & Phelps, E. A. (2015). Emotional learning selectively and retroactively strengthens memories for related events. Nature, 520(7547), 345–348. https://doi.org/10.1038/nature14106
Fagioli, S., Hommel, B., & Schubotz, R. I. (2007). Intentional control of attention: Action planning primes action-related stimulus dimensions. Psychological Research, 71(1), 22–29. https://doi.org/10.1007/s00426-005-0033-3
Failing, M., & Theeuwes, J. (2018). Selection history: How reward modulates selectivity of visual attention. Psychonomic Bulletin & Review, 25(2), 514–538. https://doi.org/10.3758/s13423-017-1380-y
Fischer, C., Czoschke, S., Peters, B., Rahm, B., Kaiser, J., & Bledowski, C. (2020). Context information supports serial dependence of multiple visual objects across memory episodes. Nature Communications, 11(1), 1932. https://doi.org/10.1038/s41467-020-15874-w
Grégoire, L., Kim, H., & Anderson, B. A. (2021). Punishment-modulated attentional capture is context specific. Motivation Science, 7(2), 165–175. https://doi.org/10.1037/mot0000211
Hall, G., & Rodríguez, G. (2019). Attention to perceive, to learn and to respond. Quarterly Journal of Experimental Psychology, 72(2), 335–345. https://doi.org/10.1080/17470218.2017.1339719
Han, S., Ji, E., Choe, E., Kim, D., & Kim, M.-S. (2020). Simple action planning can affect attentional allocation in subsequent visual search. Psychonomic Bulletin & Review, 27(5), 1014–1024. https://doi.org/10.3758/s13423-020-01758-z
Hayes, T. R., & Henderson, J. M. (2022). Scene inversion reveals distinct patterns of attention to semantically interpreted and uninterpreted features. Cognition, 229, 105231. https://doi.org/10.1016/j.cognition.2022.105231
Helbing, J., Draschkow, D., & Võ, M. L.-H. (2020). Search superiority: Goal-directed attentional allocation creates more reliable incidental identity and location memory than explicit encoding in naturalistic virtual environments. Cognition, 196, 104147. https://doi.org/10.1016/j.cognition.2019.104147
Job, X., Golemme, M., Bhattacharya, J., Cappelletti, M., De Fockert, J., & Van Velzen, J. (2019). The influence of motor preparation on the processing of action-relevant visual features. Scientific Reports, 9(1), 11084. https://doi.org/10.1038/s41598-019-47640-4
Kalbe, F., & Schwabe, L. (2022). On the search for a selective and retroactive strengthening of memory: Is there evidence for category-specific behavioral tagging? Journal of Experimental Psychology: General, 151(1), 263–284. https://doi.org/10.1037/xge0001075
Kampis, D., & Southgate, V. (2020). Altercentric Cognition: How Others Influence Our Cognitive Processing. Trends in Cognitive Sciences, 24(11), 945–959. https://doi.org/10.1016/j.tics.2020.09.003
Kristjánsson, Á., & Draschkow, D. (2021). Keeping it real: Looking beyond capacity limits in visual cognition. Attention, Perception, & Psychophysics, 83(4), 1375–1390. https://doi.org/10.3758/s13414-021-02256-7
Kruschke, J. K. (2011). Models of attentional learning. In E. M. Pothos & A. J. Wills (Eds.), Formal Approaches in Categorization (1st ed., pp. 120–152). Cambridge University Press. https://doi.org/10.1017/CBO9780511921322.006
Lakens, D., & Caldwell, A. R. (2021). Simulation-Based Power Analysis for Factorial Analysis of Variance Designs. Advances in Methods and Practices in Psychological Science, 4(1), 251524592095150. https://doi.org/10.1177/2515245920951503
Le Pelley, M. E., Mitchell, C. J., Beesley, T., George, D. N., & Wills, A. J. (2016). Attention and associative learning in humans: An integrative review. Psychological Bulletin, 142(10), 1111–1140. https://doi.org/10.1037/bul0000064
Le Pelley, M., & Newell, B. R. (2023). Reward-driven and memory-driven attentional biases automatically modulate rapid choice [Preprint]. PsyArXiv. https://doi.org/10.31234/osf.io/cgs98
Leys, C., Ley, C., Klein, O., Bernard, P., & Licata, L. (2013). Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 49(4), 764–766. https://doi.org/10.1016/j.jesp.2013.03.013
Morey, R. D. (2008). Confidence Intervals from Normalized Data: A correction to Cousineau (2005). Tutorials in Quantitative Methods for Psychology, 4(2), 61–64. https://doi.org/10.20982/tqmp.04.2.p061
Pérez-Dueñas, C., Acosta, A., & Lupiáñez, J. (2014). Reduced habituation to angry faces: Increased attentional capture as to override inhibition of return. Psychological Research, 78(2), 196–208. https://doi.org/10.1007/s00426-013-0493-9
Schenke, K. C., Wyer, N. A., & Bach, P. (2016). The Things You Do: Internal Models of Others’ Expected Behaviour Guide Action Observation. PLOS ONE, 11(7), e0158910. https://doi.org/10.1371/journal.pone.0158910
Schmidt, L. J., Belopolsky, A. V., & Theeuwes, J. (2015). Potential threat attracts attention and interferes with voluntary saccades. Emotion, 15(3), 329–338. https://doi.org/10.1037/emo0000041
Song, J., Kosovicheva, A., & Wolfe, B. (2023). Road Hazard Stimuli: Annotated naturalistic road videos for studying hazard detection and scene perception. Behavior Research Methods. https://doi.org/10.3758/s13428-023-02299-8
Vater, C., Wolfe, B., & Rosenholtz, R. (2022). Peripheral vision in real-world tasks: A systematic review. Psychonomic Bulletin & Review, 29(5), 1531–1557. https://doi.org/10.3758/s13423-022-02117-w
Wang, G., Wang, P., Luo, J., & Nan, W. (2018). Perception of Threatening Intention Modulates Brain Processes to Body Actions: Evidence From Event-Related Potentials. Frontiers in Psychology, 9, 2149. https://doi.org/10.3389/fpsyg.2018.02149
Yu, X., Zhou, Z., Becker, S. I., Boettcher, S. E. P., & Geng, J. J. (2023). Good-enough attentional guidance. Trends in Cognitive Sciences, 27(4), 391–403. https://doi.org/10.1016/j.tics.2023.01.007
Zhou, Z., & Geng, J. J. (2024). Learned associations serve as target proxies during difficult but not easy visual search. Cognition, 242, 105648. https://doi.org/10.1016/j.cognition.2023.105648

The authors declare no competing interests.

Download PDF

Version 1

posted

You are reading this latest preprint version

Context-Dependent Modulations of Spatial Attention: A Simulated Driving Task

Status:

Version 1

Abstract

Figures

Significance Statement

Introduction

Methods

Participants

Stimuli and apparatus

Procedure

Design

Results

Experiment 1: Experimental vs 2-D Stationary Control Condition

Reaction Times

Cueing Effects

Experiment 2: Experimental vs 3-D Stationary Control Condition

Reaction Times

Cueing Effects

Experiment 3: Experimental vs Light-Post Target Control Condition

Reaction Times

Cueing Effects

General Discussion

Conclusion

Declarations

References

Additional Declarations

Status:

Version 1