The effects of cognitive and visual workload on peripheral detection in the Detection Response Task

This is the peer reviewed and accepted version. Published in: Human Factors, 2018

The effects of cognitive and visual  workload on peripheral detection in the Detection Response Task

Wim van Winsum, Carnetsoft, Groningen, The Netherlands



Objective: The independent effects of cognitive and visual load on visual Detection Response Task (vDRT) reaction times were studied in a driving simulator by performing a backwards counting task and a simple driving task that required continuous focussed visual attention to the forward view of the road. The study aimed to unravel the attentional processes underlying the DRT effects.

Background: The claim of previous studies that performance degradation on the vDRT is due to a general interference instead of visual tunneling was challenged in this experiment.

Method: vDRT stimulus eccentricity and stimulus conspicuity were applied as within-subjects factors.

Results: Increased cognitive load and visual load both resulted in increased RTs on the vDRT. Cognitive load increased RT but revealed no task by stimulus eccentricity interaction. However, effects of visual load on RT showed a strong task by stimulus eccentricity interaction under conditions of low stimulus conspicuity. Also, more experienced drivers performed better on the vDRT while driving.

Conclusion: This was seen as evidence for a differential effect of cognitive and visual workload. The results supported the tunnel vision model for visual workload, where the sensitivity of the peripheral visual field reduced as a function of visual load. However, the results supported the general interference model for cognitive workload.

Application: This has implications for the diagnosticity of the vDRT: the pattern of results differentiated between visual task load and cognitive task load. It also has implications for theory development and workload measurement for different types of tasks.

Keywords: workload, focussed attention, Peripheral Detection Task, Detection Response Task, general interference model, tunnel vision model.

Précis: The effects of cognitive and visual load on visual Detection Response Task (vDRT) performance were studied in a driving simulator. Higher visual load resulted in poorer vDRT performance for more eccentric stimuli, while cognitive load did not reveal a statistical interaction with stimulus eccentricity on DRT RT.



The Detection Response Task (vDRT) has, since its inception in 1999 as the Peripheral Detection Task (PDT), been applied and studied in a large number of experiments. Originally it was presented as a simple method to assess the workload demands of both primary tasks (driving) and secondary tasks, such as in-vehicle information systems, or IVIS (Van Winsum, Martens & Herland, 1999). In the study of van Winsum et al. (1999), the PDT method was developed to unobtrusively assess short-term variations in mental workload. It appeared that variations in workload induced by both the primary task of driving and the secondary task of using an IVIS resulted in both higher reaction times and higher fractions of missed responses on the PDT.  The PDT stimuli were presented at angles ranging from 11 to 23 horizontal degrees (stimulus eccentricity). The method was initially based on the idea that the ability to process peripheral information decreases as foveal load increases (the tunnel vision model, see later). However, higher task load did not result in poorer PDT performance for more eccentric stimuli: no interaction between task load and stimulus eccentricity was found.

During the first years that followed, PDT proved to be sensititive to secondary tasks while driving (Olsson & Burns, 2000).  Patten, Kircher, Östlund, Nillson & Svenson (2006) found that inexperienced drivers had larger RT’s on the PDT compared to experienced drivers, and this was attributed to a higher workload in inexperienced drivers.  Although some other studies have found the PDT to be sensitive to the primary task demands of driving (Jahn, Oehme, Krems & Gelau, 2005; Bruyas & Dumont, 2013), most studies have focussed on the loading effects of secondary tasks. Törnros & Bolling (2005) found that performance on the PDT was impaired by dialling and conversation on both handsfree and handheld mobile phones. Patten, Kircher, Östlund & Nilsson (2004) found that PDT performance was affected by the complexity of the conversation using mobile phones. Harms & Patten (2003) found that PDT performance was sensitive to especially visual modes of presentation of navigation messages while driving. Although these studies have demonstrated that the PDT is sensitive to task demands, they have not identified which processes or functions are actually addressed.

Three lines of research have guided the developments towards the currently accepted idea that PDT mainly measures cognitive load and that stimulus eccentricity does not matter since there are no peripheral visual field effects of workload. These developments have resulted in a change of name from Peripheral Detection Task (PDT) to Detection Response Task (DRT). The first line of work concerns studies that have demonstrated that auditory and tactile DRT stimuli give similar results as visual stimuli. It was then reasoned that, since there’s nothing ‘peripheral’ in the task, DRT would be a more appropriate name. For example, Merat & Jamson (2008) examined the effect of two IVIS on visual, auditory and tactile versions of the task and found that all were equally sensitive to secondary task demands. They concluded that “the absense of a difference in signal detection by modality suggests that performance on these tasks relies on general attentional resources and is not modality specific” (p. 145). Conti, Dlugosch, Vilimek, Keinath & Bengler (2012) reported a study where it was found that tactile DRT, head mounted DRT and remote DRT were all sensitive to cognitive workload variations. Ranney, Baldwin, Smith, Mazzae & Pierce (2014) tested a head-mounted DRT, a remote DRT and a tactile DRT and concluded that all variants were sensitive to variations in task demand.

The second line of work concerns studies that have found no effects of stimulus eccentricity on DRT performance. Merat, Johansson, Engström, Chin, Nathan & Victor (2006) found no effects of stimulus eccentricity in a visual detection task, similar to the results of van Winsum et al. (1999). They concluded that stimulus eccentricity appears to have little influence on response time and that reduced detection performance is due to cognitive rather than perceptual interference. In a review, Victor, Engstrom & Harbluk (2009) stated: “there is little evidence that eccentricity has any major effect at all on detection of PDT-type stimuli (e.g., LEDs or graphical dots). …, it becomes clear that visual perceptual narrowing does not seem to be the main mechanism behind the effects found for the SDT (Signal Detection Task)” (p. 158).

The third line of work concerns studies that have demonstrated clear effects of cognitive loading tasks and those effects appear to be stronger compared to visually loading tasks. DRT performance is sensitive to increasing cognitive load as manipulated by working memory tasks, auditory-vocal speech tasks and mental arithmetic tasks (Conti-Kufner, 2017). For example Conti, Dlugosch & Bengler (2014) studied the effects of task instructions of DRT while performing n-back tasks and SuRT tasks with different levels of complexity. They found that DRT RT was sensitive to different levels of complexity of the n-back task. However, DRT was not sensitive to different levels of complexity of the SuRT task, which is a visual-motor task. This supported the idea that DRT performance mainly measures cognitive load. Bengler, Kohlman & Lange (2012) tested a visual and tactile DRT with listening and backwards counting tasks of different levels of difficulty. They found that, especially for the backwards counting tasks, DRTs were sensitive to task load.

These lines of research have resulted in support for the ‘general interference model’ that states there’s a general degradation that occurs equally for all extra-foveal stimuli regardless of their eccentricity from the point of fixation (Crundall, Underwood & Chapman, 2002). This model predicts main effects of task load and eccentricity, but no interaction between task load and stimulus eccentricity on DRT performance. The alternative ‘tunnel vision’ model states there’s a shrinking functional field of view where detection of more eccentric stimuli deteriorates stronger with increasing foveal load. This model predicts an interaction between task load and eccentricity, which can be interpreted as a reduction in the sensitivity of the peripheral visual field. More eccentric stimuli are then perceived less easily than less eccentric stimuli when task load increases.  For example, Miura (1987) reported an interaction of situational demand and target eccentricity on RT during a driving task which supported the ‘tunnel vision’ model. Similarly, Williams (1985) found that foveal load of a primary task interacted strongly with retinal eccentricity on detection performance of secondary task stimuli. However, concerning the DRT it was stated by Victor, Engstrom & Harbluk (2009): ‘performance degradation on the DRT is due to a general, amodal interference in attention selection rather than a modality-specific visual perceptual narrowing’ (p. 153). The measurement method, now named DRT, has recently been standardized as an applied measure of the attentional effects of cognitive load (ISO 17488:2016): Detection-response task (DRT) for assessing attentional effects of cognitive load in driving.

However, the absense of a task by eccentricity interaction in a number of studies does not constitute sufficient evidence for general interference. An important reason why no task by eccentricity interactions have been found for  vDRT stimuli may be that these stimuli have been too conspicuous. If a stimulus is highly conspicuous, which is normally the case with LEDs that are used in visual DRT tasks, a reduction of the sensitivity of the peripheral visual field may be difficult to measure. So, the reason why an interaction between task load and eccentricity of the stimulus has been difficult to find in previous DRT studies may be that the effect of stimulus conspicuity has not been examined before.

In the experiment described below, stimulus conspicuity and stimulus eccentricity are manipulated. vDRT is applied to both a cognitive task and a visual task, and it will be investigated whether the workload induced by those tasks can be measured separately and if the underlying processes can be unraveled. Also, it is examined whether support can be found for the hypothesis that higher visual load results in a reduction of the sensitivity of the peripheral visual field for more eccentric locations (tunnel vision instead of general interference).



The experiment was performed in a driving simulator of Carnetsoft. This desktop simulator consisted of a computer with a GeForce GTX 770 GPU, an i7-4790 CPU, windows 7 PRO Operating System and 16 GB RAM, with 4 monitors connected, each with a resolution of 1680×1050. Images were rendered over a 210 degrees horizontal field of view: 70 degrees for the forward view and 70 degrees each for the left and right out-of-the-windows views. A G29 steering wheel and pedal set of Logitech was used with a force feedback steering wheel and a rotation of -450 to +450 degrees, brake, clutch and accelerator pedals, a gear shifter and response buttons on the steering wheel, see Figure 1. The rotation angle of the steering wheel matches that of a real car, while the force feedback, using 2 electro engines, results in a realistic sensation of steering.



Figure 1. Carnetsoft desktop driving simulator setup.

The experiment was designed with the script language that is part of the system. The virtual environment consisted of a two-lane road with a lane width of 3.35 m. The road was semi-circular and continuous, delineated with broken center lines and continuous edge lines. It consisted of a sequence of curved left-turned and right-turning road segments, with radii of 150, 250, 288, 350, 400 and 500 m, and straight segments of 100 m length between curved segments. There were no visual objects, such a trees, positioned along the road, in order to make sure that the background on which the vDRT stimuli were displayed was the same for all stimuli. Also, in order not to mask the stimuli and make sure all were equally conspicuous, no oncoming or other traffic was created and a neutral lightblue sky with a thin layer of clouds was used.


Task instructions were read from a paper before the start of each task. During all tasks subjects were instructed to look at a white fixation cross drawn at the center of the middle display.

Backwards Counting Task (BC): Cognitive load. Subjects were instructed to count back silently from 200 in steps of one, paced by an auditory click once per second, to ensure that the task was the same for all subjects. The task took around 3 minutes. After the task was completed subjects were asked the final number they arrived at, to make sure the task was performed seriously. This was a cognitive task that addressed working memory and computational resources and affected cognitive workload. Bengler, Kohlman & Lange (2012) found that the complexity level of a counting task affected vDRT performance. Counting backwards was considered more difficult than counting forward, and backwards counting by one, as implemented here, is a relatively easy cognitive task. The task is a simple version of the Serial Sevens or the Serial Threes task discussed by Kennedy & Scholey (2000), that was designed as a cognitive task

Driving Task (DR): Visual load. Subjects were seated in front of the middle monitor, while their position was represented graphically as being seated in the left front seat of a car. In order to ensure that the task was the same for all subjects, the car pulled up in cruise control mode to 80 km/h with an acceleration of 1.5 m/s2, and subjects were instructed to steer the car within lane boundaries of the right lane and to give priority to this task (primary task) over other tasks. They were asked to look straight ahead at the white fixation cross throughout the task. At the end of the task, the car stopped automatically with a deceleration of 2.0 m/s2.  The road was curved in such a way that continuous steering was required to keep the car within the right lane boundaries. This driving task was chosen over other visual tasks because it required continuous visual attention and to motivate subjects to perform the task seriously: the consequences of driving off the road are serious in real driving. Although this driving task was not purely visual, because it contained cognitive and motor components as well, it clearly adds a visual load compared to the cognitive backwards counting task. It required continuous focussed visual attention to the forward view of the road.  Driving performance was measured with the standard deviation of lateral position (SDLP), which is a measure of swerving: a smaller SDLP is indicative of more accurate steering or better steering performance (see for example Brookhuis, 2014).

Visual Detection Response Task (vDRT). vDRT stimuli consisted of small red blocks of 0.5×0.5 cm that were presented on the middle monitor. Each session, 45 stimuli were presented 4.5 cm above the horizontal centerline of the middle monitor. On average every 4 s, with random variation between 3 and 5 s, a stimulus was presented at a horizontal angle of either 0, 10 or 20 degrees to the left of the line between the eyes of the subject and the centre of the middle monitor. The eyes of the subjects were 55 cm from the monitor.  Within each trial, 15 stimuli per eccentricity were presented semi-randomly. With 45 stimuli per trial the average duration of a vDRT measurement trial was 180 s.  Stimulus eccentricity was thus manipulated as a within-subjects factor within trials.

The vDRT stimuli were either High Conspicuity (HC), with an RGBA value of (1, 0, 0, 1), or Low Conspicuity (LC), with an RGBA value of (1, 0, 0, 0.3): the LC version was semi-transparant with an alpha value of 0.3, and thus less clear. Within each trial, conspicuity was either HC or LC. So stimulus conspicuity was manipulated as a within-subjects factor between trials.

At the center of the monitor, a small white fixation cross was drawn and subjects were instructed to look at the fixation cross instead of the stimuli and not turn their gaze at the stimuli and to press the response button on the steering wheel with the right-hand thumb as soon as the red square was detected. The right thumb was positioned on the response button during the task. Because peripheral instead of foveal detection was of main importance in this experiment, this fixation cross was included to make sure the subjects refrained from visual scanning in all conditions and focussed on a single area on the middle screen. A sound click was generated when the response button was pressed as a feedback cue. Stimulus duration depended on the response: in order to minimize the number of missed responses, the stimulus was removed immediately as soon as the response button was pressed, or when it was on-screen for more than 2.0 s: maximum stimulus duration was 2.0 s. See figure 2 for an overview of stimulus locations and stimulus characteristics.

Average reaction time (vDRT RT) was computed per trial for each eccentricity separately, if a response was measured within 2500 ms after stimulus onset (a ‘hit’).  If no response was measured within this period, this was coded as a missed response. Over these 15 stimuli hit rate was calculated as the number of hits divided by the total number of stimuli (15 per trial per eccentricity). So, dependent variables were vDRT RT and hit rate.

Several variants of the DRT have been used before in experiments, for example the tactile DRT, head mounted DRT and remote DRT, see for example Ranney, T. A., Baldwin, G., Smith, L. A., Mazzae, E., & Pierce, R. S. (2014). The version used in the current study was derived from the original PDT as defined in van Winsum et al. (1999). For clarity it will be referred to as visual DRT or vDRT.

The present version deviates from the ISO 17488:2016 standard in a few respects. A red block of which the alpha value (transparancy) was modified constitutes an easy and reliable way to manipulate the conspicuity of a stimulus. For experimental purposes this has important advantages over using a led light, as is a more general practice in vDRT research.  Also the contrast of the stimulus with the background can be controlled in a better way using computer image generation (a red block stimulus), the background of the DRT stimuli was a blue sky. In the ISO standard, stimulus duration was set to 1.0 second. However, if the sensitivity of the peripheral visual field is reduced by task load, which is an important hypothesis in this study, it may take longer before the stimulus is perceived by the subjects. For that reason, stimulus duration has been set to 2.0 seconds maximum in the present study.

Figure 2. Left picture: High Conspicuity (HC) stimulus. Right picture: Low Conspicuity (LC) stimulus. The vDRT stimulus is depicted at the largest eccentricity of 20° from the left of the center. The arrows point at the three possible stimulus eccentricities.


Subjects performed a practice trial for each task separately before the measurements started. First the vDRT was practiced for 45 stimuli, which was about 3 minutes. After that, the Backwards Counting Task was practiced for 100 s. Then the driving task was practiced for 5 minutes.

There were two measurement blocks for each subject. In total, each block took about 13 minutes to complete. These consisted of a fixed sequence of the following:

  • BL: Baseline vDRT measurement while seated behind the wheel with the engine off (stationary non-moving scenery). This took about 3 minutes.
  • BC: Backwards Counting task + vDRT measurement while seated behind the wheel with the engine off (stationary non-moving scenery). This also took about 3 minutes.
  • DR: Driving Task + vDRT measurement. The car pulled up to 80 km/h. When the car reached a speed of 80 km/h, the vDRT task started, and driving continued for about 3 minutes.
  • DR+BC: Driving Task + Backwards Counting Task + vDRT measurement. The car continued to drive at 80 km/h, while a new vDRT trial + Backward Counting started. After approximately 3 minutes of measurement, the car decelerated and stopped, and the measurement block was finished.

The two measurement blocks differed in stimulus conspicuity. One block with HC stimuli, and the other block with LC stimuli. The order of blocks was counterbalanced between subjects.


Sixteen young male subjects participated in the experiment. The average age was 18.7 years (SD=0.95, range 18 to 21). Eight subjects were licenced drivers (for an average of 1 year) while eight did not have a drivers licence. Although the differences in experience between these two groups were small, the effects of experience on vDRT performance were tested with EXPERIENCE as a between-subjects factor. The factor EXPERIENCE consisted of 2 levels: a group of subjects with no licence (Low Experience) vs a group of licenced drivers (High Experience). This research complied with the tenets of the Declaration of Helsinki. Informed consent was obtained from each participant.

Data Registration and Analysis

Average vDRT reaction time (RT) and vDRT hit rate were measured. A signal was coded as missed when, within 2.5 s after stimulus onset, no response was given. During each trial, average vDRT RT and number of missed signals were computed for each level of eccentricity separately. All data were analysed with repeated measurements analysis of variance within-subjects design (MANOVA) using SPSS. Because variances of reaction time data can easily violate the assumptions of equal variances, MANOVA was the preferred analysis technique. It gives more reliable results in these cases than ANOVA (Statistica, 2017. Quote: “MANOVA approach to repeated measures. To summarize, the problem of compound symmetry and sphericity pertains to the fact that multiple contrasts involved in testing repeated measures effects (with more than two levels) are not independent of each other. However, they do not need to be independent of each other if we use multivariate criteria to simultaneously test the statistical significance of the two or more repeated measures contrasts. This “insight” is the reason why MANOVA methods are increasingly applied to test the significance of univariate repeated measures factors with more than two levels. We wholeheartedly endorse this approach because it simply bypasses the assumption of compound symmetry and sphericity altogether.”).  The within-subjects factors were TASK, CONSPICUITY and ECCENTRICITY.

  • TASK: In the analyses, task performance was always compared with baseline (BL) performance,

– BC vs. BL to examine the effects of cognitive load (high vs. low cognitive load),

– DR vs. BL to examine the effects of visual load (high vs. low visual load) , and

– DR+BC vs. BL to examine the combined effects of cognitive and visual load (high combined load vs. low load).

In the BL task the levels of cognitive load and visual load were both low, so this represented a suitable baseline measurement for both the Backwards Counting and Driving Tasks.

  • CONSPICUITY consisted of two levels: HC (high conspicuity: clear vDRT stimulus) vs. LC (low conspicuity: transparant vDRT stimulus).
  • ECCENTRICITY consisted of three levels: 0°, 10° and 20° horizontal visual angle of the vDRT stimulus.

Driving performance (SDLP) was analyzed during the driving task (DR). The effects of EXPERIENCE on vDRT RT were examined by comparing DR vs. BL (TASK effect). This was done to evaluate whether more experienced drivers performed better on the vDRT while driving, as was found by Patten et al., (2006).

The confidence level for significance was set at p < .01. This α level was chosen over the more conventional .05 level because reaction time data often violate the homogenity of error variances and in such cases it has been recommended to chose a more conservative α level (Tabachnick and Fidell, 2001). Results with an  α level between .01 and .05 were referred to as ‘marginally significant’.


 Hit rate. Table 1 presents an overview of the hit rate for each factor. Most cells have an average of 1.00, which means that no signals were missed. Also in most cells (14 out of 24) the SD of hit rate was 0.0, which means there was no variance at all in those cells.  Because of this, hit rate could not be analysed statistically. Only the effects on vDRT RT are reported in the next paragraphs. Hit rates were very high and approximately the same for all tasks: BL, BC , DR and BC+DR.

TABLE 1. Averages of hit rate as a function of task, eccentricity and conspicuity, standard deviation between brackets. BL=Baseline, BC=Backwards Counting, DR=Driving, DR+BC=Driving+Backwards Counting

High ConspicuityLow Conspicuity
BL1.000 (0.000)1.000 (0.000)0.996 (0.017)0.996 (0.017)0.996 (0.017)0.996 (0.017)
BC1.000 (0.000)0.996 (0.017)0.992 (0.023)0.988 (0.027)1.000 (0.000)1.000 (0.000)
DR1.000 (0.000)1.000 (0.000)1.000 (0.000)1.000 (0.000)1.000 (0.000)0.988 (0.036)
DR+BC1.000 (0.000)1.000 (0.000)1.000 (0.000)1.000 (0.000)0.996 (0.017)0.975 (0.048)

Workload effects of tasks. To give a general idea of how the Backwards Counting and the Driving Task compared with respect to workload effects, table 2 presents the average vDRT RT for both the High Conspicuity and the Low Conspicuity conditions.

TABLE 2. Average vDRT RT (s) as a function of CONSPICUITY. Diff = Difference with BL.  BL=Baseline, BC=Backwards Counting, DR=Driving, DR+BC=Driving+Backwards Counting.

 High ConspicuityLow Conspicuity

The workload effects of the BC and the DR tasks, computed as the differences with BL, were comparable in size. Moreover, the workload effects were additive: the difference score with BL of the BC+DR task combination was approximately the sum of the individual BC and DR effects.

Effects of cognitive load: BC vs. BL.  Backwards Counting from 200 in steps of one affected cognitive load. The analyses were performed with the factors TASK, CONSPICUITY and ECCENTRICITY. Table 3 shows the statistical effects.

TABLE 3. Effects of TASK, CONSPICUITY (CONSP) and ECCENTRICITY (ECC) on RT. TASK levels: BC vs. BL.  ** = p<.001; * = p < .05.

TASK70.86< .001**15,1

The TASK effect was statistically significant at the p<.001 level. As can be seen in Figure 3, the lines of vDRT RT for the BL and BC tasks were approximately parallel: cognitive workload induced by Backwards Counting resulted in higher RT’s on the vDRT.

The main effect of CONSPICUITY was marginally significant at p<.05: clearer (high conspicuity) stimuli were detected a bit faster. As can be seen in Figure 3, RT’s were a bit higher for the low conspicuity stimuli, especially for the BC task (BL HC: 0.340, LC: 0.345; BC HC: 0.411, LC: 0.434).

The main effect of ECCENTRICITY was marginally significant at the p<.05 level. The values of vDRT RT and statistical effects are presented in Table 4. This effect was only significant in the baseline condition and was caused by consistently higher RT’s for all subjects to more eccentric (20°) stimuli compared to stimuli at 0° and 10° eccentricity. This was not the case in the BC condition, where vDRT was larger in the 20° eccentricity in 9 out of 16 subjects. So, the consistency of a very small difference caused a significant effect of ECCENTRICITY in BL while this was not the case in the BC condition. These small but consistent differences as a function of ECCENTRICITY were drowned in the larger differences between BL and BC, resulting in an absence of a significant TASKxECC interaction (see Table 3). As can be seen in Table 4, the size of the effect was small.

TABLE 4. Effect of ECCENTRICITY on RT for the BL and BC conditions: RTs in seconds. ** = p<.001.


Figure 3. BC vs. BL: vDRT RT as a function of  TASK, CONSPICUITY and ECCENTRICITY. SD indicated as vertical bars.

Effects of visual load: DR vs. BL. The Driving Task was assumed to increase visual workload compared to the baseline task. The analysis was performed with the factors TASK, CONSPICUITY and ECCENTRICITY. Table 5 shows the statistical effects. The results were very different from the effects of  cognitive workload.

TABLE 5. Effects of TASK, CONSPICUITY (CONSP) and ECCENTRICITY (ECC) on RT. TASK levels: DR vs. BL.  ** = p<.001.

TASK66.21< .001**15,1
CONSPICUITY28.20< .001**15,1
ECCENTRICITY54.38< .001**30,2
TASKxCONSP47.44< .001**15,1
TASKxECC26.44< .001**30,2
CONSPxECC34.67< .001**30,2
TASKxCONSPxECC27.97< .001**30,2

All main effects and interactions were highly significant with large F values. As can be seen in Figure 4a (left), the lines representing the Driving Task were not parallel with the lines of the BL task: especially in the low conspicuity condition, average RT’s while performing the Driving Task were higher for more eccentric stimuli, while the BL task showed a much smaller or no effect of stimulus eccentricity or stimulus conspicuity. This is the TASKxCONSPxECC effect: as visual workload increased, vDRT RT not only increased, it increased especially for more eccentric stimuli with a lower conspicuity.

Figure 4a. (Left) DR vs. BL:  vDRT RT as a function of  TASK, CONSPICUITY and ECCENTRICITY . SD indicated as vertical bars.

Figure 4b. (Right) DR+BC vs. BL: vDRT RT as a function of  TASK, CONSPICUITY and ECCENTRICITY. SD indicated as vertical bars.

Effects of cognitive load added to visual load: DR+BC vs. BL. Driving, while simultaneously performing the Backwards Counting Task (DR+BC), added an additional cognitive load to the increased visual load induced by the driving task. The analysis was performed with the factors TASK, CONSPICUITY and ECCENTRICITY. Table 6 shows the statistical effects. The effects of performing the Driving Task + Backwards Counting Task are similar as the effects of the Driving Task alone, except that, from inspection of Figure 4b (right), there was a simple additive effect of the cognitive loading BC task: vDRT RT’s were higher in the DR+BC task compared to the Driving task alone. Again, especially in the low conspicuity condition, average vDRT RT’s in the DR+BC task increased strongly for more eccentric stimuli.

TABLE 6. Effects of TASK, CONSPICUITY (CONSP) and ECCENTRICITY (ECC) on RT. TASK levels: DR+BC vs. BL. ** = p<.001.

TASK89.60< .001**15,1
CONSPICUITY22.20< .001**15,1
ECCENTRICITY56.46< .001**30,2
TASKxCONSP22.99< .001**15,1
TASKxECC39.51< .001**30,2
CONSPxECC15.33< .001**30,2
TASKxCONSPxECC12.55< .001**30,2

In order to test whether the effect of adding the BC task to driving was additive on the RT results, the effects of DR vs. DR+CB were tested. Table 7 shows the results.

TABLE 7. Effects of TASK, CONSPICUITY (CONSP) and ECCENTRICITY (ECC) on RT. TASK levels: DR vs. DR+BC. ** = p<.001.

TASK40.57< .001**15,1
CONSPICUITY45.73< .001**15,1
ECCENTRICITY78.59< .001**30,2
CONSPxECC44.35< .001**30,2

The main effect of TASK was significant, which means that adding the cognitive loading BC task to DR increased vDRT RT. None of the interactions with Task were statistically significant: adding a cognitive load to a visual task simply increased vDRT RT, but did not result in a stronger effect of eccentricity on RT for stimuli with a lower conspicuity. This is best seen by the absense of a statistically significant TASKxCONSPxECC interaction.

Effects of driving experience. It was tested whether the subjects with no licence differed from the subjects with a driver licence (factor EXPERIENCE) on driving performance, measured by SDLP, during the DR (driving) task. There was a statistically significant effect of EXPERIENCE on SDLP: F(13,1) = 9.68, p=.008. SDLP of subjects with no licence was 0.375 while SDLP of subjects with a driver licence was 0.280. So subjects with a bit more driving experience steered more accurately in the simulated driving task. Since the more experienced drivers performed the driving task better, the driving task may have been less demanding and workload may have been lower compared to the inexperienced group. Table 8 shows the statistical main effects and interactions of the factor EXPERIENCE.

TABLE 8. Effects of EXPERIENCE (EXP) on RT. TASK levels: DR vs. BL. ** = p<.001, * = p<.05.


Figure 5a shows vDRT RT as a function of TASK (BL. vs DR), CONSPICUITY and ECCENTRICITY for the low experience (no licence) group while Figure 5b does the same for the ‘high’ experience (licence) group.

Figure 5a. (Left) LOW EXPERIENCE: BL vs. DR: vDRT RT as a function of  TASK, CONSPICUITY and ECCENTRICITY . SD indicated as vertical bars.

Figure 5b. (Right) HIGH EXPERIENCE BL vs. DR: vDRT RT as a function of  TASK, CONSPICUITY and ECCENTRICITY. SD indicated as vertical bars.

Subjects with more driving experience responded faster (EXPxTASK effect) during the driving task, and RT was less affected by stimulus eccentricity (EXPxTASKxECC effect) and especially so in the low conspicuity condition (EXPxTASKxCONSPxECC effect). This suggest that driving generated a lower workload for experienced drivers and that RT of experienced drivers was less affected by stimulus eccentricity in the low conspicuity condition.


The effects of cognitive load and visual load were studied on vDRT reaction time performance. Cognitive load was manipulated with a forced-paced backwards counting task. Visual load was manipulated with a simple driving task, where subjects were required to steer the vehicle between the edge lines of the right lane on a road with left- and right turning curves. Speed was force-paced and set to 80 km/h, in order to have the same task complexity for all subjects. Both the cognitive loading task and the visual loading task resulted in increased vDRT reaction times. This confirms the sensitivity of the vDRT to both visual task load and cognitive load that was found in several previous studies (for example Van Winsum et al., 1999; Harms & Patten, 2003; Conti et al., 2012).

Earlier tests of task load x eccentricity effects on vDRT RT, discussed in the introduction, revealed no interactions of task load by stimulus eccentricity (van Winsum et al., 1999; Merat et al., 2006; Victor et al., 2009), and this was taken as evidence that the effects of workload on vDRT performance were not caused by a reduction of the sensitivity of the peripheral visual field. However, a clear, conspicuous, eccentric stimulus may be detected even when the sensitivity of the peripheral visual field is reduced. Instead, a less conspicuous stimulus may be detected less easily when the visual field’s sensitivity is reduced, and as a result reaction time may increase. This was tested in the present experiment by manipulating both stimulus eccentricity and stimulus conspicuity. A significant stimulus eccentricity by conspicuity interaction was found on vDRT RT as a function of visual load. However, cognitive load did not reveal an eccentricity by conspicuity interaction. These findings suggest that cognitive and visual load affect vDRT performance in a different way. Moreover, when both the cognitively loading task and the visually loading task were performed simultaneously, the effects of cognitive load were additive to the effects of visual load. This supports the conclusion that both types of load have different and separable effects on vDRT performance. So, an important conclusion is that only visual load, but not cognitive load,  affects the sensitivity of the peripheral visual field. The results of the present study suggest two attentional mechanisms that affect vDRT performance in different ways:

  • A higher cognitive load resulted in increased vDRT RT for all stimulus eccentricities equally: for cognitive load, the results matched the general interference model.
  • A higher visual load resulted in higher RT for stimuli with larger eccentricities which is a task x eccentricity interaction, mainly in conditions of low conspicuity. So, for visual load, the results matched the tunnel vision model.

Figure 6 shows the patterns of results as predicted by the general interference model and the tunnel vision model. The cognitive load effects (see Fig. 3) resemble the parallel lines that are characteristic of the general interference model. The visual load effects (see Fig. 4 for the low conspicuity condition) resemble the non-parallel lines of the tunnel vision model. However, as discussed in the introduction, most recent DRT research favours the ‘general interference model’ (Victor, Engstrom & Harbluk, 2009). The present results support this for the cognitive loading task, but not for the visually loading task.

Figure 6. Pattern of results (task load x eccentricity interactions) as predicted by the general interference (lower part) and the tunnel vision (upper part) models.

There are other patterns in the data that suggest which process is responsible for the tunnel vision effect. It was found that more experienced drivers not only performed the steering task better, they also performed better on the vDRT while driving. This confirms the results of Patten et al. (2006). So they probably experienced the driving task as less demanding or, phrased differently, they experienced less workload while driving compared to less experienced drivers. In addition, vDRT performance of inexperienced drivers was more affected by stimulus eccentricity compared to experienced drivers, so they seemed to experience more tunnel vision. This suggests that driving experience affects the process that is responsible for the peripheral visual field effect of visual load. A good candidate for that underlying process may be effortful focussed visual attention: inexperienced drivers may have been required to focus more foveally in order to perform the driving task adequately, and this attentional focus may have resulted in poorer detection of peripheral stimuli. This resembles Williams (1988),  who concluded that three things are necessary in order to induce tunnel vision instead of general interference: high foveal cognitive load, a focussed attention strategy and speed stress. Visual, effortful,  focussed attention may then be the source of  the tunnel vision effect induced by visual load in DRT performance.

There are also some limitations of the present study. No control mechanism, such as an eye tracker, was implemented to check if subjects did actually fixate their gaze on the fixation cross, apart from the explicit instructions.  However, Fotios, Uttley & Cheal (2016) concluded that instructing subjects to look at the fixation cross is sufficient. They used eye tracking to examine this issue and found that subjects maintained a high degree of foveal fixation at a fixation cross during peripheral detection tasks. So its likely that subjects behaved according to the instructions thoughout the DRT tasks. Another limitation of the present study is that there are other possible explanations for the interaction of conspicuity and stimulus eccentricity apart from visual load, such as the optical flow induced by driving or manual load (the requirement to steer while driving). This will be examined in a next experiment.

The use of both hit rate and reaction time to interpret the results in DRT experiments was recommended in the ISO 17488:2016 document to check for speed-accuracy tradeoffs. A problem with the hit rate as a dependent variable is that it often lacks sufficient variance for statistical analysis, which was also the case in the present study. A probable reason for the exceptionally high hit rates in the current experiment is the low vDRT RT’s: fast reactions are often accompanied by high hit rates (van Winsum et al., 1999) . The fast reactions in the current study, compared to most other studies, may be caused by the young age of the subjects. Strayer, Cooper, Turrill, Coleman, and Hopman (2015) found that younger subjects have lower DRT RT’s  and higher hit rates compared to older subjects.  It has been suggested by Ranney, T. A., Baldwin, G., Smith, L. A., Mazzae, E., & Pierce, R. S. (2014) that hit rates are lower in driving conditions compared to non-driving conditions. However, in the present study hit rates were equally high in both driving and non-driving (BC and BL) tasks.

The results of this study have practical implications for the use of the DRT in applied and theoretical studies. Since both cognitive and visual load resulted in increased RT,  visual load can be measured more precisely by adding eccentricity as a factor in the tests (stimuli presented at 10° and 20° horizontal eccentricity). To measure the effects of visual load it is also recommended to use stimuli with a low conspicuity. This also improves the sensitivity of the vDRT for effects of cognitive load. The effect of eccentricity on RT may be used to improve the diagnosticity of the test: a higher vDRT RT, compared with baseline measurements (task effect) accompanied by the absense of an interaction with eccentricity suggests cognitive load while a task x eccentricity effect suggests visual load.  For studies aimed at theory development into the ‘general interference’ model versus the ‘tunnel vision’ model it is recommended to include conspicuity of the peripheral stimulus as a factor because this improves the sensitivity to eccentricity effects.

The driving task in the present study differs from on-road driving in a number of respects. During the experiment, speed control and the cognitive task were externally-paced while the driver’s gaze was fixed. In contrast, during on-road driving these task aspects are often self-paced. This allows the driver to control visual and cognitive load to a larger extent during on-road driving. However, the results of the present study can be generalized to on-road driving during circumstances where the driver is subjected to high visual load that cannot be controlled easily by the driver, for example in prolonged night driving on a highway, driving in fog, or in high traffic density. In these circumstances peripheral visual field effects of visual load may occur.


  • Cognitive load resulted in higher vDRT reaction time, irrespective of stimulus eccentricity and conspicuity: for cognitive load the results matched the general interference model.
  • Visual load resulted in higher vDRT reaction time and a significant interaction of stimulus eccentricity and conspicuity was found: for visual load the results matched the tunnel vision model.
  • Cognitive and visual load revealed additive effects on vDRT reaction time.


Bengler, K., Kohlmann, M., & Lange, C. (2012). Assessment of cognitive workload of in-vehicle systems using a visual peripheral and tactile detection task setting. Work, 41(Suppl 1), 4919–4923.

Brookhuis, K.A. (2014). The role of traffic psychology in psychopharmacological research. Transportation Research Part F: Traffic Psychology and Behaviour, 25, 120-126.

Bruyas, M.-P., & Dumont, L. (2013). Sensitivity of detection response task (DRT) to the driving demand and task difficulty. In Proceedings of the seventh international driving symposium on human factors in driver assessment, training, and vehicle design (64–70). Iowa City: University of Iowa.

Conti, A. S., Dlugosch, C., Vilimek, R., Keinath, A., & Bengler, K. (2012). An assessment of cognitive workload using detection response tasks. In N. A. Stanton (Ed.), Human aspects of road and rail transportation (735–743). CRC Press.

Conti, A. S., Dlugosch, C., & Bengler, K. (2014). The effect of task set instruction on detection response task performance. In D. de Waard et al. (Eds.), Proceedings of the human factors and ergonomics society europe chapter 2013 annual conference (107 – 117).

Conti-Kufner, A.S. (2017). Measuring cognitive task load: An evaluation of the Detection Response Task and its implications for driver distraction assessment. Thesis. Technischen Universität München.

Crundall, D, Underwood, G. & Chapman, P. (2002). Attending to the Peripheral World While Driving. Applied Cognitive Psychology, 16, 459-475.

Dirkin, G.R. & Hancock. P.A.(1985). An attentional view of narrowing: The effect of noise and  signal bias on discrimination in the peripheral visual field. In I.D. Brown. R. Goldsmith. K. Coombes & M.A. Sinclair (Eds.) Proceedings of the Ninth Congress of the International Ergonomics Association, Bournemouth. London:Taylor and Francis.

Fotios, S., Uttley, J. & Cheal, C. (2016). Maintaining foveal fixation during a peripheral detection task. Lighting Research & Technology, 48(7), 898-909.

Harms, L., & Patten, C. (2003). Peripheral detection as a measure of driver distraction. A study of memory-based versus system-based navigation in a built-up area. Transportation Research Part F, 6, 23–36.

ISO 17488:2016, Road vehicles – Transport information and control systems – Detection-Response Task (DRT) for assessing attentional effects of cognitive load in driving. (2016). Geneva: International Organization for Standardization.

Jahn, G., Oehme, A., Krems, J. F., & Gelau, C. (2005). Peripheral detection as a workload measure in driving: Effects of traffic complexity and route guidance system use in a driving study. Transportation Research Part F: Traffic Psychology and Behaviour, 8(3), 255–275.

Kennedy, D.O. & Scholey, A.B. (2000). Glucose administration, heart rate and cognitive performance: effects of increasing mental effort. Psychopharmacology, 149(1), 63-71.

Merat, N., Johansson, E., Engström, J., Chin, E., Nathan, F., and Victor, T. (2006). Specification of a secondary task to be used in safety assessment of IVIS, AIDE Deliverable 2.2.3, European Commission, IST-1-507674-IP.

Merat, N., & Jamson, A. H. (2008). The effect of stimulus modality on signal detection: Implications for assessing the safety of in-vehicle technology. Human Factors, 50(1), 145–158.

Miura, T. (1987). Behaviour Oriented Vision: Functional Field of View and processing Resources. Eye Movements from Physiology to Cognition. Proceedings of the Third European Conference on Eye Movements, Dourdan, France, 563-572.

Olsson, S., & Burns, P. C. (2000). Measuring driver visual distraction with a peripheral detection task. In Nhtsa driver distraction internet forum. Electronic version:

Patten, C., Kircher, A., Östlund, J., & Nilsson, L. (2004). Using mobile telephones: Cognitive workload and attention resource allocation. Accident Analysis and Prevention , 36, 341-350.

Patten, C. J. D., Kircher, A., Östlund, J., Nilsson, L., & Svenson, O. (2006). Driver experience and cognitive workload in different traffic environments. Accident Analysis and Prevention38(5), 887–894.

Ranney, T. A., Baldwin, G., Smith, L. A., Mazzae, E., & Pierce, R. S. (2014). Detection response task evaluation for driver distraction measurement application. (Report No. DOT HS 812077), National Highway Traffic Safety Administration.

Statistica (2017). Assumptions and Effects of Violating Assumptions – Sphericity and Compound Symmetry. From: Overview/ AssumptionsSphericityandCompoundSymmetry.

Strayer, D. L., Cooper, J., Turrill, J., Coleman, J. R., & Hopman, R. J. (2015). Measuring cognitive distraction in the automobile III: A comparison of Ten 2015 in-vehicle information systems. Washington, DC: AAA Foundation for Traffic Safety.

Tabachnick B. G., Fidell L. S. (2001). Using Multivariate Statistics, 4th Edn. Needham Heights, MA: Allyn and Bacon.

Jan E.B.Törnros, Anne K.Bolling. (2005) Mobile phone use—Effects of handheld and handsfree phones on driving performance. Accident Analysis & Prevention, 37(5), 902-909.

van Winsum, W., Martens, M., & Herland, L. (1999). The effects of speech versus tactile driver support messages on workload, driver behaviour and user acceptance. Soesterberg: TNO Human Factors Research Institute (Report No. TM-99-C043).

Victor, T. W., Engström, J., & Harbluk, J. L. (2009). Distraction assessment methods based on visual behavior and event detection. In M. A. Regan, J. D. Lee & K. L. Young (Eds.), Driver Distraction: Theory, Effects, and Mitigation (135-165). Boca Raton, FL: CRC Press.

Williams, L.J., (1985).Tunnel vision induced by a foveal load manipulation, Human Factors, 27, 221–227.

Williams L. J., (1988). Tunnel vision or general interference? Cognitive load and attentional bias are both important. American Journal of Psychology, 101, 171-191

Biographies: Wim van Winsum has studied experimental psychology and received his PhD in 1996 at the University of Groningen, The Netherlands with his thesis “From adaptive control to adaptive driver behaviour.”  He has worked at the University of Groningen and at the Human Factors Research Institute TNO in Soesterberg, The Netherlands. He has done research in the fields of driving behaviour and human factors and driving simulator studies. Since 2000 he has worked as a developer of driving simulators for training and research.