Effect of target contrast and divided attention on the useful field of view

Abstract

Previous research has shown that there is a cost of dividing attention between the central and peripheral visual fields in a complex environment. However, it is not clear how stimulus factors, such as the contrast of the scene, affect the cost. The current study reports the results of two studies that address this question. In Experiment 1, temporal thresholds of the Useful Field of View (UFOV) tests were measured as a function of contrast and retinal eccentricity. The results showed that central-focused attention thresholds increased (i.e., performance decreased) as contrast decreased. Peripheral and divided attention task performance decreased as eccentricity increased. Surprisingly, peripheral and divided attention task performance were the best for medium rather than high contrast targets. The unexpected poorer performance under the high contrast condition might possibly be explained by the crowding effect. To test this possible explanation, in Experiment 2 the peripheral stimuli were simplified to minimize the potential crowding effect on peripheral target detection. The results showed that the unexpected effect of contrast on the cost of dividing attention could be accounted for by the crowding effect. When combined, the results from the two experiments suggest that the cost of dividing attention between central and peripheral targets is more pronounced for objects at greater eccentricity under lower contrast conditions, consistent with a tunnel effect. The implications of this finding are discussed in the paper.

1. Introduction

Many everyday visual tasks require observers to allocate their limited attentional resources to extract information necessary to complete an action successfully and safely. When performing these tasks, observers often need to attend to information distributed throughout their visual field. For instance, while operating a vehicle, a driver must not only attend to the center of the visual field to monitor the distance from a lead vehicle, but also simultaneously monitor the peripheral visual field to detect road signage, approaching vehicles, and pedestrians entering the intersection.

Due to the limited attentional resources (Kahneman, 1973), observers cannot attend to all the information in their visual field and thus must select what information to process. Various metaphors have been used to characterized the restriction of attention to select objects, including the attentional spot light and zoom lens (Eriksen and St. James, 1986; Eriksen & Yeh, 1985). According to this latter model, there is a limited area around fixation within which observers can extract useful visual information and that space has variable power. Also, research has suggested that the size of the attentional visual field is dependent upon the demands of the visual task, such as recognizing shapes in isolation or among distractors (LaBerge & Brown, 1989).

The useful field of view (UFOV) is described as the area from which an observer can extract information in a single glance (Ball & Owsley, 1993; Ball, Roenker, & Bruni, 1990), and has been used to characterize spatial attention. Tests for this construct generally evaluate central, divided, and selective attention (Ball, Beard, Roenker, Miller, & Griggs, 1988; Edwards et al., 2005; Owsley, Ball, & Keeton, 1995; Sekuler & Ball, 1986). The central test of attention requires participants to identify which of two similar appearing caricatures of a car or truck was presented. Detection performance is defined as the temporal duration of the target that supports a criterion level of performance. The divided attention task supplemented the central task with a peripheral stimulus target stimulus (i.e., caricature of a car). Participants were required to report both which target was presented in the central field and the visual meridian upon which the peripheral target was presented. The selective attention task differed from the divided attention task only in that the peripheral target was embedded among distractors consisting of inverted triangles. In each case, performance was assessed as the temporal duration of the target stimulus supporting a criterion level of performance. Ball and colleagues have shown that the UFOV is predictive of prospective and retrospective accident risk in older adults (Ball & Owsley, 1993; Ball et al., 2006; Owsley et al., 1998) while others have found it predictive of crashes in younger adults (McManus, Cox, Vance, & Stavrinos, 2015). Richards, Bennett, and Sekuler (2006) used a modified version of the UFOV test and calculated the cost of dividing attention by subtracting the score of the peripheral attention test from the score of the divided attention task. This approach allowed them to assess the added demands of a central identification task were to the peripheral localization. They found that the attention cost was greater for older adults than for younger adults.

In addition to differences in age groups, a number of other factors have been shown to increase the cost of dividing attention between foveal and peripheral targets. Two early dual-task studies reported converging results on the effect of stress and cognitive load on the detection of peripheral targets. Bahrick, Fitts, and Rankin (1952) found that participants who were given an incentive (i.e., money) to correctly complete the task were more likely to miss peripheral targets than central targets. Similarly, using temperature as a stressor, Bursill (1958) found that participants more frequently missed more peripheral targets than targets near the center of the visual field under high temperature conditions. The detection of peripheral targets is also affected by increased foveal task demands (Williams, 1989). For instance, increasing the central task difficulty by manipulating the complexity of the central target decreases the size of the UFOV (Ikeda & Takeuchi, 1975). This finding is consistent with limited capacity models of attention, which posit that attentional resources are finite and that the processing of peripheral targets or the performance of a secondary task is inversely related to the attentional demands of the primary or central task. Recent work by Ringer et al. (2016)suggests that a secondary task can cause both an overall loss in sensitivity and a greater decline in sensitivity for more peripheral stimuli (i.e., tunnel vision). They found that a secondary auditory n-back task only caused a general decline in sensitivity across the visual field but a secondary foveal visual task caused an effect more characteristic of tunnel vision where in sensitivity decline more steeply for more peripheral stimuli.

Another factor that might increase the cost of dividing attention is the contrast of the targets in the visual field. The contrast of targets could affect one’s ability to segregate objects from the background or other objects. Reduced contrast has been shown to impair many real-world tasks, including driving (Decina & Staplin, 1993; Emerson et al., 2012; Ginsburg, 2003). A number of environmental conditions common in driving, such as disability glare, fog, dust, and smoke, can create conditions with very low contrast (Boyce, 2009), as does most night-time driving (Leibowitz & Owens, 1977; Leibowitz, Owens, & Tyrrell, 1998). In reduced-visibility conditions, drivers have more difficulty with tasks such as object recognition (Evans & Ginsburg, 1985; Ni, Bian, Guindon, & Andersen, 2012; Wood & Owens, 2005) and steering control (Ni, Nguyen, & Zhuo, 2013). Therefore, understanding how low contrast affects dividing attention to multiple tasks is important for furthering knowledge about driving safety.

Although low contrast might be expected to negatively impact UFOV performance, it is less clear how the effect may manifest across the visual field. One possible outcome could be that the reduced contrast will result in generalized interference effect, causing an overall decline in performance in the divided attention task independent of eccentricity. Such an effect would be predicted based on the effortfulness theory (Rabbitt, 1968; Wingfield, Tun, & McCoy, 2005). The effortfulness theory suggests that there is a cost associated with the perception of a degraded stimulus, such as one with reduced contrast. According to this theory, degraded input requires more effortful processing, thus reducing the resources available for other tasks (e.g., detection of peripheral targets), the performance of a secondary task, or for downstream operations (e.g., encoding material into memory or other higher-level comprehension activities). Wood et al. (2009) used this theory to test the effect of degraded visual input (i.e., effects of simulated cataracts) on a battery of visual-based cognitive tests. They found found the speed of processing was slowed under reduced contrast conditions, even when participants were still able to perceive the stimuli. Therefore, the effortfulness theory predicts that reduced stimulus contrast would act as a form of generalized interference in a divided attention task, which we referred to as general interference hypothesis.

Alternatively, the effect of contrast may interact with eccentricity. In other words, the increased perceptual load associated with processing a low contrast central target could demand more cognitive resources differentially impacting sensitivity to more peripheral targets. As a result, the negative effect of reduced contrast would be more pronounced for more peripheral targets, therefore causing a tunnel effect (Ringer et al., 2016). This prediction is referred to as the tunnel effect hypothesis.

The primary goal of this study is to investigate the effect of contrast on the cost of dividing attention. In two experiments, we sought to test the two hypotheses described above. Specifically, Experiment 1a aims to test the general interference hypothesis and the tunnel effect hypothesis by measuring performance for three UFOV subtests at three different contrast levels and five eccentricity levels. Experiment 1b aims to replicate the finding in Experiment 1a using more contrast levels (i.e., increased from three to five). Experiment 2 attempts to control the potential crowding effect of annuli on detecting peripheral targets.

2. Experiment 1a

2.1. Participants

Twenty-one adults (12 males, 9 females) between the ages of 18 and 31 (M = 21.48, SD = 3.61) were recruited through the Wichita State University SONA Experiment Management System. All participants had normal or corrected to normal vision, except for one participant with an acuity of 20/60. Ocular dominance was determined through a variation of the Miles (1930) Test. More than half of the participants (n = 13) were right-eye dominant. Participants were compensated for their time with Psychology course credit for their participation in the study. All participants completed the entire experiment. All participants were naive to the purposes of the experiment and gave informed consent in accordance with a protocol approved by the Institutional Review Board panel of Wichita State University.

2.2. Design

The effect of peripheral target eccentricity and stimulus contrast was investigated in this experiment. Detection thresholds for peripheral targets was tested at five eccentricities (4°, 8°, 12°, 16°, and 20°) relative to the center of the display screen (Richards et al., 2006) and three contrast levels (low, medium and high). The low contrast condition was based on the participants’ contrast sensitivity threshold, which was measured monocularly with the dominant eye using a contrast detection task (Phan & Ni, 2011; Sowden, Rose, & Davies, 2002). The high contrast level of the light stimuli to the dark background was set at 0.29 Michelson contrast, in the same level used by Richards et al. (2006). The medium contrast was a logarithmic midpoint between high and low contrast level, as determined by the equation:

$$\text{medium}=low\times {10}^{\frac{\log \left(\frac{\text{high}}{\text{Iow}}\right)}{2}}$$

where “high” is the high contrast level and “low” is the contrast threshold for each participant. The contrast was adjusted by increasing or decreasing the luminance of the targets, annuli, and masks for the entire sequence; the background remained the same luminance for the duration of the experiment. The dependent variable was the display duration threshold, calculated using the best PEST procedure (Lieberman & Pentland, 1982; Pentland, 1980).

2.3. Apparatus and stimuli

All displays were generated by custom programs written in C++ and Open Graphics Library (OpenGL) in the Microsoft^® Visual Studio^® 2008 environment. The stimuli were presented on a 21′′ ViewSonic G810 CRT color monitor, at a resolution of 800 × 600, with a refresh rate of 60 Hz. Participants were seated with their chins in a chin-rest 50 cm away from the screen, and instructed to enter their responses via a modified keyboard. For the contrast detection tasks, the stimulus was viewed monocularly through a viewing hood, with a 7.6-cm aperture in the center. All participants viewed the computer displays in a dark room after at least 5 min of dark adaptation.

Contrast sensitivity was measured using a contrast detection task. The test stimulus was a Gaussian windowed sinusoidal grating (Gabor patch) displayed in the center of the screen with a spatial frequency of 6 cycles per degree and subtending 2.5° visual angle. The mean luminance of gray background was 50 cd/m². At each trial, two different beeps were sounded from the computer. During one of the two beeps, a Gabor patch appeared at the fixation point. Each Gabor patch was displayed for 100 ms, with an inter-stimulus interval of 1000 ms. Participants indicated during which beep they saw the Gabor patch by pressing either the “4” or “6” button on the keypad. Incorrect responses were indicated with another different beep. The threshold was estimated using a 3-down-1-up staircase method (Kooi, Toet, Tripathy, & Levi, 1994; Levi, 2008; Levitt, 1971). The program terminated when the participants had completed ten reversals. The mean Michelson contrast detection threshold in the fovea for all participants was 0.036, with a standard deviation of 0.0119; the maximum was 0.0693, and the minimum was 0.0188.

The UFOV was measured using an adaptation of the test used by Richards et al. (2006), which consisted of three tests: central-focused attention, peripheral-focused attention, and divided attention. Each trial in each test began with an initiation screen that showed the text “Ready” on a black background. Participants pressed the space bar to initiate the trial. A fixation cross appeared in the center of the screen after the key press for 500 ms. The central focused-attention task consisted of a central target, while the peripheral-focused attention task consisted of only peripheral targets. The divided attention task consisted of both central and peripheral targets. Central targets were one of four (“E”, “F”, “H”, or “L”) randomly selected letters and were displayed in the center of the screen and subtended approximately 1.0° visual angle. Peripheral stimuli were presented in one of 20 peripheral locations indicated by white circular annuli that subtended approximately 1.2° visual angle and formed an “X” pattern radiating from the central target (i.e., creating four radial arms). The target consisted of a white disk, which randomly appeared within one of the 20 circular annuli. The annuli and the target had the same luminance. The stimulius was displayed on a gray background (lumimance = 19.5 cd/m²). After the stimulus was presented, a checkerboard mask display was presented for 1000 ms covering all stimulus locations. Each checkerboard was the same size as the peripheral stimuli (i.e., subtending approximately 1.2° visual angle); the dark parts had the same luminance as the stimuli background, and the light parts had the same luminance as the targets. For the central identification task, only the central target appeared, and participants were prompted to report which of the four letters was presented. For the peripheral localization detection task, only the peripheral target appeared, and participants were prompted to report in which of the four radial arms the target was presented. The divided attention tasks included both the central and peripheral tasks. Participants first responded to the central letter task and then the peripheral target location task. A schematic representation of the three tasks can be found in Fig. 1, which shows the series of displays for each of the three tests. Fig. 2 shows the stimulus screen in more detail.

Fig. 1.

A schematic representation of the three UFOV tasks, showing the sequence of displays and their duration. Top: central focused attention; middle: peripheral-focused attention; bottom: divided attention.

Fig. 2.

An example of the stimulus screen of the divided attention task, highlighting the central and peripheral targets.

The threshold stimulus duration was measured using the best PEST procedure (Lieberman & Pentland, 1982; Pentland, 1980). There was only one threshold estimate for the central task. For the peripheral and divided attention tasks, a threshold was obtained for each of the five eccentricities. The minimum value was therefore the refresh rate of the monitor (16.6 ms). The highest value was set at 333.3 ms to reduce the likelihood of eye movements from the central target to the peripheral targets.

2.4. Procedure

Experimental sessions lasted approximately 75 min. Participants first completed the acuity and ocular dominance tests, and then were taken to the experimental computer, where they could adjust the chair and chin rest to suit their comfort.

Following five minutes of dark adaptation, participants performed the contrast sensitivity test monocularly with their dominant eye. Participants practiced the contrast detection task to ensure they understood the task before starting the actual test.

Upon completing the contrast detection task, participants completed a practice block for each UFOV task before each test block for each of the three UFOV tests. All participants completed the UFOV tests in a fixed order (central-focused, peripheral-focused, and then divided), while the order of three contrast levels was counterbalanced among participants. There were seven participants in each counterbalance group. Participants were instructed to take breaks between tests and they were allowed to take short breaks between trials to reduce fatigue.

2.5. Results

Threshold estimates were calculated for each participant under each condition for each task. Higher threshold values indicate worse performance. The attention costs were calculated by subtracting the divided attention threshold by the peripheral attention threshold for each condition and for each participant (Richards et al., 2006), and therefore this represents the overall cost of adding a peripheral-focused attention task to the central target identification task. Four analyses of variances (ANOVAs) were conducted on central-focused attention, peripheral-focused attention, divided attention, and attention cost. For each ANOVA, Mauchly’s test for sphericity was conducted and if the assumption was violated, an appropriate correction was applied to adjust the degrees of freedom (Greenhouse & Geisser, 1959).

2.5.1. Central-Focused attention

A one-way within-subjects ANOVA was conducted to evaluate the effect of contrast on central-focused attention. The effect of contrast was significant, F(1.09,21.81) = 38.40, p <.001, η_p² = 0.66, where threshold values increased as the contrast of the stimuli decreased. Fig. 3 shows the means and standard errors for this effect. Post-hoc comparison using the Bonferroni method to correct for type I error found that there was a significant difference (p <.001) between all contrast levels.

Fig. 3.

Effect of contrast level on central-focused attention threshold duration. Error bars indicate standard error.

2.5.2. Peripheral-focused attention

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the threshold duration (ms) of the peripheral attention test. The effect of contrast was significant, F(1.37,27.33) = 22.89, p <.001, η_p² = 0.53, where the medium contrast condition had the lowest threshold, followed by high and then low contrast conditions. The main effect of eccentricity was significant, F(1.90,38.00) = 32.94, p <.001, η_p² ₌ 0.62, where threshold duration increased as the target eccentricity increased. The interaction between contrast and eccentricity was significant, F(3.13,62.67) = 7.43, p <.001, η_p² = 0.27, as shown in Fig. 4.

Fig. 4.

Main effect of contrast level (A), main effort for retinal eccentricity (B), and Interaction of contrast level and retinal eccentricity (C) on peripheral-focused attention threshold duration. Error bars indicate standard error.

The non-monotonic main effect of contrast was unexpected, and may be an artifact of group averaging. In other words, the group average curve might not represent each individual participant. In order to rule out this possibility, the scores of each participants were examined individually across contrast levels, and the results are shown in Fig. 5. A similar effect was found for the majority of the participants. Since calculating the mean did not disguise the effect for individual participants, the averaged results are presented for the subsequent experiments.

Fig. 5.

Effect of contrast on individual participants’ threshold duration. Each line represents the data for one participant.

2.5.3. Divided attention

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on target thresholds for the divided attention UFOV test. The main effect of contrast was significant, F(1.53,30.67) = 56.725, p <.001, η_p² = 0.74. As with peripheral-focused attention, the medium contrast condition showed the lowest duration threshold, followed by the high and low contrast conditions. The main effect of eccentricity was significant, F(4,80) = 9.04, p <.001, η_p² = 0.31. The interaction between contrast and eccentricity was significant, F(4.35,87.04) = 3.89, p <.005, η_p² = 0.16, as shown in Fig. 6.

Fig. 6.

Main effect of contrast level (A), main effort for retinal eccentricity (B), and Interaction of contrast level and retinal eccentricity (C) on divided attention threshold duration. Error bars indicate standard error.

2.5.4. Attention cost

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the cost of dividing attention between central and peripheral targets. The dependent variable was the threshold duration difference (ms) between divided attention and peripheral attention tests. The main effect of contrast was significant, F(2,40) = 37.27, p <.001, η_p² = 0.65. The costs of dividing attention increased as the contrast of the stimuli decreased. The main effect of eccentricity was significant, F(4,80) = 2.84, p <.05, η_p² ₌ 0.12. The interaction between contrast and eccentricity was not significant, F(8,160) = 1.40, p =.20, η_p² ₌ 0.07.

Two post-hoc comparisons were conducted using the Bonferroni method to correct for type I error. For the main effect of contrast, the high level and the low level were significantly different (p <.001), but high contrast and medium contrast did not differ significantly (p =.10). None of the pairwise comparisons of the main effect of eccentricity were significant (p >.32). Fig. 7 shows the means and standard errors for the main effects of contrast and eccentricity.

Fig. 7.

Effect of contrast level (A) and retinal eccentricity (B) on the attention cost. Error bars indicate standard error.

3. Experiment 1b

The results from experiment 1a suggest that there is an interference effect of contrast on all three UFOV tests, and an effect of eccentricity on the focused-peripheral and the divided attention tests. These results, when combined, confirm the tunnel effect hypothesis rather than the general interference hypothesis.

What was unexpected was that the results of Experiment 1a that showed a non-monotonic effect of contrast on peripheral-focused attention and divided attention, with the best performance found under medium contrast levels. Threshold durations were higher for high and low contrast targets relative to medium contrast targets. These results demonstrate that the effects of stimulus contrast on attention to peripheral targets is non-monotonic. Experiment 1b aims to replicate Experiment 1a using a larger set of contrast levels to more precisely document the non-monotonic effect of contrast on performance.

3.1. Participants

Twenty adults (7 males, 13 females) between the ages of 18 and 31 (M = 20.95, SD = 3.35) were recruited through the Wichita State University SONA Experiment Management System. All participants have normal or corrected to normal vision. They were compensated for their time with Psychology course credit for their participation in the study. All participants completed the entire experiment.

3.2. Design

The design of experiment 1b is similar to that of Experiment 1a, except that performance was tested at three eccentricities (4°, 12°, and 20°) rather than five levels as were tested in Experiment 1a, and only five stimulus contrast levels (low, low-medium, medium, medium–high and high). High and low contrast levels were calculated the same as Experiment 1a, with the intermediate representing logarithmic quartiles between them.

3.3. Procedure

Experimental sessions lasted approximately 60 min. The procedure was the same as that in Experiment 1a with the following exceptions. All participants ran through the UFOV test in a fixed order (peripheral-focused, and then divided), while the order of five contrast levels was counterbalanced among participants. Thus, there were four participants in each counterbalance group. Participants were instructed to take breaks between the two UFOV tests and were allowed to take short breaks between trials to reduce fatigue.

3.4. Results

Threshold estimates were calculated for each participant under each condition for each task. Higher threshold values indicate worse performance. The values for the attention cost were obtained by subtracting the divided attention threshold by the peripheral attention threshold under each condition and for each participant. Three analyses of variances (ANOVAs) were conducted on peripheral-focused attention, divided attention, and attention cost. For each ANOVA, Mauchly’s test for sphericity was conducted, and if the assumption was violated, an appropriate correction was applied to adjust the degrees of freedom (Greenhouse & Geisser, 1959).

3.4.1. Peripheral-focused attention

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the display duration threshold (ms) of the peripheral attention test. The effect of contrast was significant, F(1.38,26.56) = 32.53, p <.001, η_p² = 0.63, where threshold scores decreased until the medium contrast level and then increased, showing a non-monotonic effect similar to that found in Experiment 1a. The main effect of eccentricity was also significant, F(1.40,26.68) = 67.17, p <.001, η_p² = 0.78, where threshold generally increased as the target was presented further from the center of the screen. The interaction of contrast × eccentricity was significant, F(3.46,65.72) = 8.79, p <.001, η_p² = 0.31.

3.4.2. Divided attention

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the display duration threshold of the divided attention UFOV test. We found a significant main effect of contrast, F(2.42,46.03) = 41.36, p <.001, η_p² = 0.69, and of eccentricity, F(2,38) = 33.86, p <.001, η_p² = 0.64. The interaction between contrast and eccentricity was not significant, F(4.41,83.73) = 2.35, p =.06, η_p² = 0.11. Fig. 9 shows the means and standard errors for the main effect of contrast and the main effect of eccentricity.

Fig. 9.

Main effect of contrast level (A) and retinal eccentricity (B) on divided attention threshold duration. Error bars indicate standard error.

Two post-hoc comparisons were conducted using the Bonferroni method to correct for type I error. For the main effect of contrast, there were the following significant pairwise comparisons (p <.005): high to low, medium–high to medium–low, medium–high to low, medium to medium–low, medium to low, and medium–low to low (Fig. 9A). For eccentricity, significant difference was found between 4 and 20, and between 12 and 20 (p <.05), but not between 4 and 12 (Fig. 9B).

3.4.3. Attention cost

A two-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the cost of dividing attention between central and peripheral targets. The main effect of contrast was significant, F(4,76) = 4.57, p <.005, η_p² = 0.19, as can be seen by Fig. 10. The costs of dividing attention generally increased as the contrast of the stimuli decreased. The main effect of eccentricity was not significant, F(2,38) = 3.04, p >.05, η_p² = 0.14. The interaction of contrast × eccentricity was significant, F(4.62,87.70) = 8.14, p <.001, η_p² = 0.30. Since attention cost was calculated by subtracting the divided attention threshold by the peripheral attention thresholds, a high threshold value (e.g., close to the maximum) in the peripheral attention would lead to a small attention cost. In this case, a small value might not represent the true attentional cost, but rather indicates a floor effect. However, this possibility did not occur in this experiment, as can be seen in Fig. 8.

Fig. 10.

Effect of contrast level on attention cost. Error bars indicate standard error.

Fig. 8.

Main effect of contrast level (A), main effort for retinal eccentricity (B), and Interaction of contrast level and retinal eccentricity (C) on peripheral-focused attention threshold duration. Error bars indicate standard error.

4. Experiment 2

While Experiments 1a and 1b provide supports to the tunnel effect hypothesis, the results showed an unexpected non-monotonic effect of contrast. This unexpected result could be due to the crowding effect of the presence of empty circles on targets in the periphery. The crowding effect refers to the deteriorated performance in detection targets in the peripheral visual field due to clutter (Whitney & Levi, 2011). Previous research investigating the effects of contrast on crowding found that lower contrast can facilitate the detection of a target (Chung, Levi, & Legge, 2001). Conversely, higher luminance levels have been shown to impede attentional blink tasks (Chua, 2005). In Experiments 1a and 1b, the annuli demarcating potential peripheral target locations were continuously (see Fig. 2) displayed, which might have influenced the detection of high contrast peripheral targets due to crowding. Experiment 2 tested this possible explanation for the non-monotonic effect of contrast on peripheral attention and divided attention.

4.1. Participants

Twenty-four adults (13 males, 11 females) between the ages of 18 and 36 (M = 21.59, SD = 4.07) were recruited through the Wichita State University SONA Experiment Management System. All participants have normal or corrected to normal vision. They were compensated for their time with Psychology course credit for their participation in the study. All participants completed the entire experiment.

4.2. Design

The design of Experiment 2 is similar to that of Experiments 1a, where display duration thresholds were tested at three contrast levels. These levels were determined the same way as in Experiment 1a. An additional variable was that of crowding, where the annuli were displayed (higher crowding) or not displayed (lower crowding) when the peripheral target appeared (Fig. 11).

Fig. 11.

Higher crowding condition (left) shows target disc with annuli in all possible target locations. The lower crowding condition (right) shows target disc without the annuli.

4.3. Apparatus and stimuli

The computerized display was generated using the same algorithm used in experiments 1a and 1b. Contrast sensitivity was again measured using a contrast detection task. The UFOV was measured using an adaptation of the test used by Richards et al. (2006), which consisted of two tests: peripheral-focused attention and divided attention.

4.4. Procedure

Experimental sessions lasted approximately 75 min. The procedure was the same as that in Experiment 1b with the following exceptions. All participants ran through the UFOV test in a fixed order (peripheral-focused, and then divided), while the order of three contrast levels and the two empty circle levels was counterbalanced among participants. Thus, there were six participants in each counterbalance group. Participants were instructed to take breaks between tests and were allowed to take short breaks between trials to reduce fatigue.

4.5. Results

Threshold estimates were calculated for each participant under each condition for each task. Higher threshold values indicate worse performance. The values for the attention cost were obtained by subtracting the divided attention threshold by the peripheral attention threshold under each condition and for each participant. Three analyses of variances (ANOVAs) were conducted on peripheral-focused attention, divided attention, and attention cost. For each ANOVA, Mauchly’s test for sphericity was conducted, and if the assumption was violated, an appropriate correction was applied to adjust the degrees of freedom (Greenhouse & Geisser, 1959).

4.5.1. Peripheral-focused attention

A three-way within-subjects ANOVA was conducted to evaluate the effect of contrast, stimulus eccentricity, and crowding on the display duration threshold of the peripheral attention test. Similar to the results in Experiments 1a and 1b, there were significant main effects of contrast (F(1.03,23.71) = 51.06, p <.001, η_p² = 0.69), eccentricity (F(2,46) = 75.86, p <.001, η_p² = 0.77) and significant interaction of contrast and eccentricity (F(2.14, 49.15) = 0.36.12, p <.001, η_p² = 0.61). Of particular interest to this study are the effect of crowding, and its interaction with contrast. The main effect of crowding was significant, F(1, 23) = 11.40, p <.005, η_p² = 0.31, where the high crowding condition produced higher mean scores (M = 91.01, SE = 5.11) than the low crowding condition (M = 79.67, SE = 5.91). Fig. 12 displays the mean and standard errors. The interaction of contrast and crowding was also significant, F(1.51, 34.80) = 9.06, p <.005, η_p² = 0.28. A simple main effect analysis showed that the crowded conditions resulted in significantly higher (p <.05) thresholds than the uncrowded conditions, but only at the highest contrast level. Fig. 13 shows the means and standard errors of the interaction. Additionally, the interaction of crowding, contrast, and eccentricity was not significant (F(2.18,50.04) = 2.90, p =.060, η_p² = 0.06).

Fig. 12.

The main effect of crowding on peripheral-focused attention threshold duration. Error bars indicate standard error.

Fig. 13.

The interaction of contrast and crowding on peripheral-focused attention threshold duration. Error bars indicate standard error.

4.5.2. Divided attention

A three-way within-subjects ANOVA was conducted to evaluate the effect of contrast, stimulus eccentricity, and crowding on the display duration threshold of the divided attention test. Similar to Experiments 1a and 1b, there were significant main effects of contrast (F(1.24, 25.84) = 94.91, p <.001, η_p² = 0.81), eccentricity (F(2,46) = 37.08, p <.001, η_p² = 0.62) and significant interaction of contrast and eccentricity (F(4,92) = 3.24, p <.05, η_p² = 0.12). Again, the aim of this study was the interaction between crowding and contrast. The main effect of crowding was significant, F(1, 23) = 22.91, p <.001, η_p² = 0.50, where the high crowding condition produced higher mean scores (M = 176.40, SE = 8.05) than the low crowding condition (M = 152.96, SE = 6.80) (Fig. 14). The interaction of contrast and crowding was also significant, F(2, 46) = 6.12, p <.005, η_p² ₌ 0.21. A simple main effects analysis showed a significant difference (p <.05) between crowded and uncrowded in high and low contrast conditions. Fig. 15 shows the means and standard errors of the interaction. No significant interactions were found for either crowding and eccentricity (F(2,46) = 0.85, p =.43, η_p² = 0.04) or crowding, contrast, and eccentricity (F(4,92) = 0.87, p =.48, η_p² = 0.04).

Fig. 14.

The main effect of crowding on divided attention threshold duration. Error bars indicate standard error.

Fig. 15.

The interaction of contrast and crowding on divided attention threshold duration. Error bars indicate standard error.

4.5.3. Attention cost

A three-way within-subjects ANOVA was conducted to evaluate the effect of contrast and stimulus eccentricity on the cost of dividing attention between central and peripheral targets. The main effect of crowding was significant, F(1,23) = 5.56, p <.05, η_p² = 0.20, where those in the high crowding condition (M = 85.38, SE = 6.24) showed higher attention cost than those in the low crowding condition (M = 73.27, SE = 4.78). The main effects of contrast (F(2,46) = 15.31, p <.001, η_p² = 0.40) and eccentricity (F(2,46) = 16.48, p <.001, η_p² = 0.42) were both significant. The attention cost increase as the target decreased in contrast and when it was it was displayed further from the central target. Fig. 16 shows the means and standard errors for the effects of contact and eccentricity. The interaction between contrast and eccentricity was significant, F(2.38, 53.76) =25.72, p <.001, η_p² = 0.53, but was likely due to floor effects similar to Experiment 1b. No significant interactions were found for crowding by contrast (F(2,46) = 0.34,p =.71, η_p² = 0.02), crowding by eccentricity (F(2,46) = 2.46, p =.10, η_p² = 0.01), and crowding by contrast by eccentricity (F(2.8,64.6) = 0.89, p =.47, η_p² = 0.04), suggesting the crowding effect could not account for the effect of eccentricity and contrast on the cost of dividing attention.

Fig. 16.

Main effect of contrast level (A) and retinal eccentricity (B) on the attention cost. Error bars indicate standard error.

5. General discussion

The current study examined: (1) the effect of contrast on the three UFOV tests; and (2) the effect of contrast and eccentricity on the cost of dividing attention between central and peripheral targets. Across all UFOV tests in all experiments, there was a significant effect of contrast on performance. As one would expect, for the central attention tasks, performance decreased with decreased contrast. Surprisingly, for both the peripheral-focused and the divided attention tasks, the performance was best (i.e., the threshold was lowest) under the medium contrast conditions as shown in Experiment 1a and Experimental 1b results. As explained above, this unexpected result could result from the crowding effect in the periphery caused by annuli demarcating locations where the peripheral target might appear. However, even with this unexpected effect, contrast did have an effect on the display duration threshold—both positively and negatively—across all UFOV tests. Additionally, the effect size for that main effect was fairly large for the UFOV tests (η_p² between 0.53 and 0.81), indicating that changing the contrast had a strong effect.

The results of Experiment 2 showed an interference effect wherein the presence of the annuli interacted with the contrast levels of the stimuli. For both peripheral-focused and divided attention, a significant interaction of crowding and contrast was found. While decreased performance was found when contrast changed from medium to low, consistent with the results of Experiment 1, the reduced crowding condition did not display the unanticipated decrease in threshold under the medium contrast condition. In fact, in the uncrowded condition, detection performance under high and medium contrast conditions were nearly the same for both peripheral-focused and divided attention tasks. This is consistent with the crowding effect explanation that the presence of the annuli interferes with the detection of peripheral targets at higher contrast.

As expected, there was a significant difference in the cost of dividing attention between contrast levels: as contrast decreased, there was a higher cost of dividing attention between central and peripheral targets. This effect was replicated in both experiments. A lower contrast not only delays processing speed (for the central task), but also makes dividing attention a more demanding task. The results were mixed in terms of the effect of eccentricity on attention cost. Experiments 1a and 2 found a significant effect of eccentricity on attentional cost, where Experiment 1b did not, the latter of which could be explained by a floor effect at the extreme (low contrast, furthest eccentricity) thresholds.

Specifically, in relation to the two hypotheses (i.e., general interference and tunnel effect hypotheses), these experiments offer greater support for the tunnel effect hypothesis. An interference effect was found when accounting for the presence of the annuli in Experiment 2. As the contrast decreased, performance decreased, particularly for the lower contrast levels. One thing to note is that all of the contrast levels were at or above each individual participant’s contrast detection threshold. As a result, the detrimental effect of lower contrast cannot be explained by the inability to detect the target. This detrimental effect of lowering the contrast is consistent with research on the effortfulness theory (Rabbitt, 1968; Wingfield et al., 2005), whereby degrading visual input makes early stages of visual processing more effortful. Such effect was also found for some cognitive tasks that simply decreased the contrast level of the displays through visual filters (Wood et al., 2009). However, when eccentricity is taken into account, performance declines under low contrast were more pronounced with increased eccentricity, providing support to the tunnel effect hypothesis. Therefore, the effortfulness hypothesis cannot completely explain the current results.

The tunnel effect hypothesis was further supported by the results that the attention cost between peripheral and divided attention was greater as contrast decreased and as eccentricity increased. A main effect of contrast was found in Experiment 1a, and an interaction of contrast and eccentricity was found in 1b. Additionally, in Experiment 2, the cost of dividing attention was also greater in the crowded condition. Taken together, this suggests that the demands added by the central task are greater when the contrast is lower and when the scene is more crowded. It is interesting to note that there was a more rapid decrease in attention cost than for the individual UFOV tests (i.e., peripheral attention and divided attention). Therefore, the cost of dividing attention may better capture the influence of external factors (e.g., contrast) on visual attention performance.

When combined, the results from the current experiments could have important implications for everyday tasks in which we have to divide attention across the field of view in low contrast situations, such as nighttime driving or driving under inclement weather conditions (e.g., in fog) or in individuals who have vision problems such as cataracts. In these situations, when an individual must divide his or her attention between multiple targets at very low contrast levels, he or she should be more cautious as significantly more time is needed for correct and accurate responses.

5.1. Limitations

For some of the more difficult conditions (i.e., lower contrast and larger retinal eccentricity) of the focused-peripheral and divided attention subtasks, participants were showing the maximum threshold possible (i.e., 333 ms) for the display duration. In Experiment 1a, 7 participants (out of 21) showed the maximum threshold under low contrast at eccentricity of 16° and 20°. In Experiment 1b, 6 participants (out of 20) showed the maximum threshold at eccentricity of 20° under low contrast condition. These results could indicate that the participants either were at that threshold, or could not adequately complete the task and thus would simply not do any worse (i.e., a floor effect). The latter would explain problems with attentional cost calculation where the difference between peripheral-focused and divided attention would be very small if the divided attention task were too difficult. While the display duration could have been extended to address this limitation, that may confound the measurement of the UFOV itself. Conceptually, the UFOV measures the area from which information can be extracted at a single glance (Ball & Owsley, 1993; Ball et al., 1990). According to this definition, we set the longest possible display duration as 333 ms to reduce the likelihood of eye movements from the fovea to the peripheral targets. Apparently, increasing the display duration would change the divided attention task to a visual search task. Furthermore, although this limitation might create a problem for interpreting the interaction, the direction of the main effects—particularly for contrast—should not be significantly affected. In case there was a ceiling effect (i.e., participants had a higher threshold in the divided-attention task), the effects of contrast and eccentric would have been larger than that shown in the results. Therefore, the current results of attention cost were conservative at most, which might reduce the power but should not change the direction of the relationship.

In addition to a potential floor effect in the divided-attention task, a ceiling effect was also possible in certain conditions in the peripheral-focused attention task (e.g., medium contrast and small eccentricity conditions). In this case, the minimum threshold of 16.7 ms was constrained by the temporal resolution (i.e., refresh rate) of the display monitor. Due to the technical limitations in the refresh rate, the threshold could not be lower than 6.25 ms for a 160-Hz monitor. Even so, the greatest gain is only about 10 ms when calculating the attention cost, which would not significantly change the direction of the relationship. Therefore, the monotonic effect of contrast on the attention cost found in the current study is a robust and reliable one.

Another possible concern is related to the way contrast sensitivity was measured and how the contrast was applied in the UFOV test. Contrast sensitivity was measured with a Gabor patch, not the solid circular targets in the experiment. The Gabor patch, consisting of sinusoidal gratings, is more primitive and has a higher spatial frequency than a solid disk (Palmer, 1999). Specifically, the spatial frequency and location of the target are different, and these two factors could influence contrast sensitivity (Boff & Lincoln, 1988). In this study, contrast sensitivity was measured at the fovea, which defined the low contrast level for peripheral targets at various retinal locations across the entire UFOV display (20° around the fixation point). Photoreceptor distribution varies as a function of retinal location (Østerberg, 1935), and contrast sensitivity generally declines with increased distance from the fovea (Harvey & Pöppel, 1972; Rijsdijk, Kroon, & Van der Wildt, 1980). Therefore, the “low” contrast level may have been below the participants’ contrast threshold at that particularly retinal location. This effect lessens as target size increases (Boff & Lincoln, 1998). Despite these potential limitations, participants did not exhibit a floor effect for low contrast targets at large retinal eccentricities. For example, Fig. 3 shows similar trends in the low and high contrast conditions. This would be impossible if the participants could not resolve the peripheral targets presented in the low contrast condition.

5.2. Future directions

The proposed explanation for the non-monotonic contrast effect found in Experiment 1 was supported by the results of Experiment 2. It is unclear how different properties of distractors may affect performance. In many UFOV tests, including the commercially available version (Ball, Owsley, Sloane, Roenker, & Bruni, 1993), there is a selective attention task with distractors present in the periphery. Since at high contrast the annuli acted at least in part as distractors, the interaction of different types of distractors and contrasts should be examined. This is of particular importance for researchers measuring divided attention instead of selective attention.

An additional direction for future research could be evaluating aging effects, as all participants in this study were younger. Older adults generally show declines in UFOV (Clay et al., 2005) and contrast sensitivity (Owsley, Sekuler, & Siemsen, 1983), and show much higher costs of dividing attention (Richards et al., 2006). Therefore, comparing the results of a younger and older sample could give more insights on the effect of age on visual functions, including the cost of dividing attention under low contrast conditions.