Effects of Reduced Contrast on the Perception and ... - SAGE Journals

Perception, 2010, volume 39, pages 1199 ^ 1215

doi:10.1068/p6558

Effects of reduced contrast on the perception and control of speed when driving D Alfred Owens

Whitely Psychology Laboratories, Franklin & Marshall College, Lancaster, PA 17604-3003, USA; e-mail: [email protected]

Joanne Wood, Trent Carberry

School of Optometry, Queensland University of Technology, Kelvin Grove, Queensland, QLD 4059, Australia Received 21 August 2009, in revised form 19 April 2010

Abstract. Misperception of speed under low-contrast conditions has been identified as a possible contributor to motor vehicle crashes in fog. To test this hypothesis, we investigated the effects of reduced contrast on drivers' perception and control of speed while driving under realworld conditions. Fourteen participants drove around a 2.85 km closed road course under three visual conditions: clear view and with two levels of reduced contrast created by diffusing filters on the windscreen and side windows. Three dependent measures were obtained, without view of the speedometer, on separate laps around the road course: verbal estimates of speed; adjustment of speed to instructed levels (25 to 70 km hÿ1); and estimation of minimum stopping distance. The results showed that drivers traveled more slowly under low-contrast conditions. Reduced contrast had little or no effect on either verbal judgments of speed or estimates of minimum stopping distance. Speed adjustments were significantly slower under low-contrast than clear conditions, indicating that, contrary to studies of object motion, drivers perceived themselves to be traveling faster under conditions of reduced contrast. Under real-world driving conditions, drivers' ability to perceive and control their speed was not adversely affected by large variations in the contrast of their surroundings. These findings suggest that perceptions of self-motion and object motion involve neural processes that are differentially affected by variations in stimulus contrast as encountered in fog.

1 Introduction Dense fog poses an extreme hazard to road safety. Although relatively rare, chain-reaction collisions in fog can be catastrophic. Crash statistics from the UK show that, compared to clear atmospheric conditions, collisions in fog cause 39% more casualties per accident, and they are seven times more likely to involve five or more vehicles (Sumner et al 1977). The US Transportation Research Board reported that in the early 1990s, 4 fogrelated collisions involved more than 240 vehicles, causing 21 fatalities and more than 90 injuries (Shepard 1996). The same report states that over the previous decade fog was a contributing factor to more than 6000 road fatalities in the US. Analyses of crashes in fog often emphasize the role of drivers' errors, particularly their failure to reduce speed appropriately to compensate for reduced visibility (eg Al-Ghamdi 2007). Why would this be the case in the face of such obvious danger? Perhaps drivers are not so much negligent as caught by surprise. Visibility can change quickly as drivers travel into a fog-shrouded environment, suddenly encountering the combined risks of (i) diminished visibility and (ii) the possibility of being struck from behind if they decelerate abruptly. Another possibility, and the primary impetus for this study, is that drivers might misperceive their speed when contrast is reduced by fog. In a call for research on ``motorway crashes in fog'' Brown (1970) hypothesized that drivers underestimate their speed in fog because their perception of distance is distorted. He reasoned that fog creates exaggerated aerial perspective (a monocular distance cue), which would cause

1200

D A Owens, J Wood, T Carberry

distant objects to seem farther away, and this distance illusion could result in drivers' feeling they are traveling more slowly. This hypothesis was partially supported by a study by Cavallo et al (2001) who found that the perceived distance of simulated taillights was overestimated by as much as 60% in a fog chamber; however, perceived speed was not assessed. Other psychophysical studies have shown that, independent of distance perception, the apparent speed of simple stimuli is directly affected by reduced contrast. Thompson's (1976) doctoral dissertation appears to be the first report of an illusory decrease of perceived speed when spatial contrast is reduced. Campbell and Maffei (1981) subsequently replicated the `Thompson effect', finding that the perceived speed of rotating gratings and dot patterns appears to decrease when image contrast is reduced, especially when the target is viewed in peripheral vision. Further studies by Thompson et al added evidence that the perceived speed of drifting sinusoidal gratings decreases with reduced contrast (eg Thompson 1982; Stone and Thompson 1992; Thompson et al 2006). Later work showed that the `Thompson effect' occurs with two- and three-dimensional patterns as well as one-dimensional gratings (Blakemore and Snowden 1999; Brooks 2001). Extrapolating from such findings, Campbell and Maffei (1981) suggested that `Ìn mist and fog there can be a dramatic decrease in the contrast of the objects around us, and this should slow down apparent motion in the peripheral field even further. This would lead to the illusion that our body was moving at a slower velocity than the true physical velocity. This could account for the tendency of inexperienced drivers to drive their car faster than is wise in conditions of low contrast ...'' (page 721).

Pursuing implications for driving, a much-cited study by Snowden et al (1998) used a desk-top driving simulator to assess the effects of reduced contrast. Their experiment involved two tasks. In one, participants adjusted the speed of a low-contrast (`foggy') scene to match that of a standard high-contrast (`clear') scene. In the second task, participants adjusted the apparent speed of travel to match target speeds under three levels of contrast (`clear', `misty', and `foggy'). Results from both measures showed that speed was underestimated by 20% in the foggy condition, leading the authors to conclude that `ìt appears that drivers think they are driving far more slowly than they actually are in foggy conditions, and therefore increase their speed'' (page 450). Similar results were reported in a conference paper by Distler and Bu«lthoff (1996). More recently, using projected video recordings of road scenes, Horswill and Plooy (2008) found that speeds appeared to be slower (by 5.8 ^ 8.3 km hÿ1) and more difficult to discriminate when the contrast of the video display was reduced. Other research indicates, however, that we should be cautious in generalizing results from studies of object motion to the perception of self-motion. Extensive research in neurology and neuroscience shows that visual perception and control of action are mediated by neural processes that are substantially distinct from those involved in perception of objects (eg Schneider 1967, 1969; Po«ppel et al 1973; Mishkin et al 1983; Weiskranz 1986; Goodale and Milner 1992; Kleinschmidt et al 2002; de Gelder 2008). Of particular relevance here, the phenomenon of vection (perception of self-motion induced by global optic flow) appears to be quite robust under impoverished visual conditions. Leibowitz et al (1979), for example, found that circular vection is not affected either by extreme blur (which reduces spatial contrast of the retinal image) or by reduction of luminance to scotopic levels. Similarly, studies with wide-screen driving simulators found that blur and reduced luminance have little or no effect on subjects' ability to steer along a curvy roadway (Owens and Tyrrell 1999; Brooks et al 2005), a finding confirmed in real-world driving (Higgins et al 1998; Higgins and Wood 2005; Owens et al 2007). These findings suggest that conditions known to impair foveal vision, such as blur and reduced luminance, have relatively little effect on the visual perception

Perception of speed when driving in reduced contrast

1201

and control of action. Leibowitz and his colleagues have proposed that such ``selective visual degradation'' may be useful for understanding driving behavior at night, when visual guidance remains efficient despite severe losses of focal vision (Leibowitz and Owens 1977; Leibowitz et al 1982; Owens and Tyrrell 1999). In the present context, the selective-degradation hypothesis predicts that, unlike perception of object motion, perception of self-motion may be relatively unaffected by reduced contrast. Here, we investigated the effects of reduced contrast on the perception and control of speed while driving in real-world conditions. Participants drove on a closed road course, without view of the speedometer, while contrast was manipulated by plastic diffusing filters attached to the interior of the vehicle's windows. Drivers' perception of speed was assessed with three dependent measures: verbal estimation, adjustment of speed to match target instructions, and estimation of the minimum stopping distance at various speeds of travel. We hypothesized that, if the `Thompson effect' applies to driving, as indicated by previous simulator studies (Distler and Bu«lthoff 1996; Snowden et al 1998; Horswill and Plooy 2008), participants would underestimate their speed of travel under conditions of reduced contrast. Alternatively, if vehicle control depends upon visual processes that are relatively unperturbed by reduced contrast, then performance would be minimally affected by manipulation of contrast. 2 Method 2.1 Participants Participants were fourteen volunteers (five women and nine men; mean age 35:8 14:8 years; range 20 ^ 57 years), who were licensed drivers from the general population that had driven regularly for at least 3 years. All participants had corrected binocular visual acuity of 6/6 (20/20) or better, and all reported that they drove regularly. Participants wore the spectacle correction that they normally used for driving. Two additional subjects were tested but were excluded from the study because of their unsafe driving behavior: one for excessive speed and the second for highly erratic driving behavior (requiring that the experimenters halted testing because of safety concerns). The study was conducted in accordance with the requirements of the Queensland University of Technology Human Research Ethics Committee. All participants were given a full explanation of the experimental procedures and written informed consent was obtained, with the option to withdraw from the study at any time. 2.2 Road course and experimental apparatus The experiment was conducted on a 2.85 km closed road course at the Mt Cotton Training Facility of the Queensland Department of Transportation, which has been used in previous studies of vision and driving performance (Wood and Troutbeck 1994). The circuit comprised an asphalt two- and three-lane road, situated in a rural `bush' setting, with multiple hills, bends, straight stretches and intersections, as well as standard road signs and road markings. The test circuit for the present experiment consisted of a full lap of the circuit, plus an additional half-lap connected by a curved side street (figure 1). Measures of speed and stopping distance were made at 5 test locations; 2 of these were passed twice, first on the full lap and again on the half-lap, making a total of 7 measurements for each test run. The measurement locations were selected to provide (i) a range of travel speeds from 25 km hÿ1 to at least 70 km hÿ1 and (ii) sufficient sight distance for the estimation of minimum stopping distance at each location. Each test location was marked by two orange traffic cones, positioned adjacent to an orange model wallaby (a small kangaroo commonly encountered on rural roads in Australia), which served as the simulated hazard for the estimation of minimum stopping distance.

1202


Start/Finish

1st loop 2nd loop 5

1

4,7 3,6

2

Figure 1. Schematic diagram of the closed road course: participants drove 1.5 laps as indicated by solid and dashed arrows. The positions of the 7 test locations are designated by the wallabies.

Tests were conducted in a right-hand drive Nissan Maxima sedan, equipped with automatic transmission. A foam-core occluder, mounted on the dashboard, obstructed the drivers' view of the speedometer but allowed an experimenter in the passenger's seat to observe and record travel speed. The contrast of the driving environment was reduced with diffusing filters, fabricated from plastic sheeting of two thicknesses (1 mil or 0.0254 mm, and 2 mil or 0.0508 mm), which were attached by Velcro fasteners around the interior margins of the windscreen and both front side windows. Care was taken to ensure that the filters were stretched when mounted so that wrinkles in the filters would not create additional distortion of vision. In order to assess the effects of these diffusing filters on spatial contrast, we measured the contrast sensitivity functions of two young observers with VisTech contrast sensitivity charts under laboratory conditions (figure 2). We also used Pelli ^ Robson and logMAR visual acuity charts to assess the contrast sensitivity and acuity of all participants in the driving environment while viewing through the different filter conditions from the driver's seat of the vehicle (table 1). Mean log contrast sensitivity

2.5

clear filter 1

2.0

filter 2

1.5 1.0 0.5 0.0

1

10 Spatial frequency=cycles degÿ1

100

Figure 2. Mean contrast sensitivity functions of two observers when viewing grating patterns on the VisTech Õ chart for the clear and the two diffusing filter conditions.

Table 1. Mean visual performance measures for three test conditions (SEM). Measure

Test condition clear view

Visual acuity (log MAR) Letter contrast (log sensitivity)

ÿ0.15 (0.03) 1.72 (0.03)

filter 1

filter 2

0.23 (0.04) 0.97 (0.04)

0.52 (0.03) 0.77 (0.05)

Stopping distance estimates were measured with a parallax-based distance measuring system that utilized two digital video cameras mounted a fixed distance apart on the roof of the test vehicle. This system recorded two overlapping images of the road scene ahead. Participants responded by tapping a large dash-mounted touch pad, which instantaneously labeled the video recordings via fiber optic links to both cameras.


1203

The stopping-distance responses were determined off-line through analysis of the parallax of the labeled video records. In addition to the advantage of minimal interference with the driving task, this technique has a high level of accuracy and validity (Jones et al 1998, 2010; Tyrrell et al 2004, 2009; Wood et al 2005). 2.3 Procedure ö driving measures Prior to testing, each participant drove one practice lap to become familiar with the test vehicle and road course. The practice lap followed the direction of travel opposite to the experimental test laps, with clear view (no filters) of the environment and the speedometer. Following the practice lap, participants drove around the test circuit nine times without view of the speedometer. This included three laps for each of the three contrast conditions. Each of these three laps was devoted to one of the three dependent measures, which was recorded at the seven test locations labeled in figure 1 (five on the large outer loop, plus two repeated on the smaller inner loop). Tasks included (i) verbal estimation of speed, (ii) adjustment of speed to match five different target speeds, and (iii) estimation of minimum stopping distance. The orders of the contrast conditions, and of the three tasks within each contrast condition, were counterbalanced across participants. During several test sessions, which were scheduled during morning or late afternoon hours when the sun was low, veiling glare on the plastic filters occasionally created severe restrictions of visibility at certain points along the test course. Participants were instructed in advance that, if such glare should occur, she/he should slow down immediately, and the experimenter would pull the plastic filter aside so that the driver could proceed beyond the sun's glare. Such glare problems were encountered during tests of five of the fourteen participants. Glare did not interfere with measures at the 7 test locations, with one exception where the problem was so troublesome for one participant that we were unable to complete tests under the denser low-contrast condition. Each test lap took approximately 3 ^ 4 min. The full test session, including vision tests and a debriefing, was completed in 1 h or less. For all testing, the drivers were instructed to follow the prescribed route, to drive as she/he normally would, and to undertake one of the three prescribed tasks. Instructions and procedural details for each task were as follows: Verbal estimates. For the verbal estimation runs, participants were instructed to travel at a comfortable speed throughout the trial and to estimate their speed in km hÿ1 when prompted at each of the 7 test locations. They were encouraged to use any number, odd or even, and to try to be accurate but not to worry if they felt uncertain. At the moment that the participants estimated their speed, the actual speed was recorded. Adjusted speed. For speed-adjustment measures, participants were asked to match their speed to a target level at each of the 7 locations, marked by wallabies in figure 1. The target speeds for the 7 test locations were 40, 50, 70, 25, 50, 60, and 25 km h ÿ1 , respectively. Target speeds were determined empirically in pilot tests in which three volunteers drove around the course at a comfortable speed. Participants were told to decline if the requested speed seemed too fast, and a different slower target speed would be used. When the participant stated that she/he had attained the target speed, the actual speed was recorded. Minimum stopping distance. For minimum stopping-distance measures, participants were instructed to report the ``last possible moment that you could stop'' before reaching a simulated wallaby, positioned at each of the 7 test locations. This is a relevant task at the Mt Cotton Training Facility, given that wallabies live in the area and, occasionally,

1204


run on to the road. We wanted to determine the last possible moment that the driver estimated that she/he could stop in case of such an event. Participants were also told there was no need to stop or slow down, but rather just to tap the touch pad, which was mounted on the dashboard, at the last possible moment she/he estimated it was possible to stop before reaching the simulated wallaby. Minimum braking distance was defined as the distance from the vehicle to the wallaby at the moment the touch pad was pressed. At each response, the actual speed was recorded. The response distance was later obtained off-line from the stereo-video measurement system. Throughout the test session, two experimenters were seated in the experimental vehicle (one in the front passenger seat and one in the back), and a third assisted in changing the diffusing filters between runs. The front-seat experimenter was responsible for instructing the participant, recording speed estimates, and ensuring the touch pad was pressed to record estimates of minimum braking distance; the back-seat experimenter ensured that the stereo-video measurement system was operating properly. 2.4 Vision testing and questionnaires After all the driving trials had been completed, vision testing was performed at a location along the road course that was neither in shade nor in direct sunlight, with the participant viewing the test charts through the windscreen of the vehicle from the driver's seat. We measured binocular visual acuity and letter contrast sensitivity for each contrast condition using test charts mounted on a movable post in front of the test vehicle, always testing with the denser (2 mil) filter first, followed by the 1 mil filter, and then clear view. Static high-contrast visual acuity was measured using two versions of a standard logMAR chart (Australian Vision Chart No 5) at a distance of 3 m, unless visual acuity was worse than the top line of the chart for a particular condition, in which case shorter viewing distances were used and the logMAR scores adjusted accordingly. Subjects were forced to guess letters, even when they were unsure, until a full line of letters was incorrectly read. Each letter seen was scored as ÿ0:02 log units. Contrast sensitivity was assessed with two versions of the Pelli ^ Robson Letter Contrast Sensitivity chart at a distance of 3 m. Participants were instructed to look at a line of letters and were instructed to guess the letter when they were not sure until a full line of letters was incorrectly read. Each letter was scored as 0.05 log units. After completion of the experiment, participants were asked a number of questions regarding their perceptions of the study and their own performance. This included ratings on a scale of 1 ^ 10 of how viewing through the diffusing filters compared to actual fog (1 not similar and 10 exactly the same), and how the filters affected their travel speed, speed adjustments, and their ability to estimate speed and braking distance compared to performance with clear view (where 1 much slower or less accurate and 10 much faster or more accurate). Participants were also asked to estimate their confidence in the speed and braking judgments as a percentage of that for the clear condition. 3 Results 3.1 Vision tests The effects of the diffusing filters on standard vision tests are illustrated in figure 2 and table 1. As expected, both filters served as low-pass optical filters degrading contrast sensitivity by 5 0:4 log units at spatial frequencies of 6 cycles degÿ1 and higher for `filter 1', and at spatial frequencies of 1.5 cycles degÿ1 and higher for `filter 2' (figure 2). The visual performance measures of participants taken under field conditions from the driver's seat confirmed that the diffusing filters caused significant reductions in


1205

both visual acuity (F1:68, 21:8 219:24, p 5 0:001, Z 2 0:944) and letter-contrast sensitivity (F1:51, 19:6 531:86, p 5 0:001, Z 2 0:976). 3.2 Overall speed A global estimate of overall speed was computed by averaging data collected at all 7 test locations during the verbal estimate trials, when participants were instructed to drive at a comfortable speed and asked to estimate that speed. As shown in figure 3, mean overall speed decreased significantly with reduced contrast, from an average of 58 km hÿ1 in the clear condition to 49 km hÿ1 in the filter 2 condition (F1:87, 24:29 24:27, p 5 0:001, Z 2 0:651). Pairwise comparisons showed significant differences between all three conditions ( p 5 0:02). Mean verbal estimates of speed during the same test runs exhibited a parallel significant decline with reduced contrast (F1:92, 24:99 7:95, p 0:002, Z 2 0:379), with mean values of 6 to 8 km hÿ1 lower than the mean actual speed, indicating that participants underestimated their actual speed in all viewing conditions, ie for the clear as well as the reduced-contrast conditions. Pairwise comparisons for the verbal estimates showed significant differences between the clear condition and both reduced-contrast conditions ( p 5 0:05), but no significant difference between the filter 1 and filter 2 conditions ( p 0:12). In summary, under reduced-contrast levels, participants were more comfortable driving at lower speeds compared to the clear condition, and they consistently underestimated their speed by 6 ^ 8 km hÿ1 regardless of the contrast level. 70 actual speed

Mean speed=km hÿ1

60

verbal estimate

50 40 30 20 10 0

clear view

filter 1 Contrast condition

filter 2

Figure 3. Mean speed of travel and corresponding mean verbal estimates of speed, averaged across 7 test locations during the free-flowing verbal estimation runs, for three contrast conditions. Error bars represent 95% confidence intervals.

3.3 Verbal estimates of speed Verbal estimates were examined more closely by analyzing changes in responses as a function of actual speed for each of the contrast conditions. One should recall that participants were instructed to travel `àt a comfortable speed''. Consequently, the actual speeds and corresponding verbal estimates varied from one participant to another. In order to compare the responses of all participants at a common set of actual speeds, each individual's data were fitted by linear regression separately for each of the three contrast conditions, which provided three speed-estimation functions for each participant (one for each contrast level). The speed-estimation equations of all participants are listed in the appendix. Mean goodness-of-fit (r) for these speed-estimation functions ranged from 0.89 to 0.94 (table 2). Individual participants' speed-estimation functions were used to derive comparable speed estimates using values of 25, 40, 50, 60, and 70 km hÿ1. The derived verbal estimates for all contrast conditions are presented in figure 4. Consistent with earlier studies (eg Evans 1970; Triggs and Berenyi 1982) and figure 3, the verbal responses underestimated actual speed for all contrast conditions. It is also evident in figure 4 that the mean verbal estimates increased linearly as a function of actual speed, and mean estimates

1206


Table 2. Mean Pearson product ^ moment correlations (r) indicating goodness-of-fit for linear regressions on data for individual participants (SDs). Test condition

Measure

Verbal speed estimates Minimum stopping distances

clear view

filter 1

filter 2

0.89 (0.10) 0.92 (0.04)

0.94 (0.04) 0.91 (0.06)

0.93 (0.05) 0.81 (0.16)

Mean estimated speed=km hÿ1

80

clear view

70

filter 1 filter 2 match

60 50 40 30 20 10 0

20

30

40 50 Actual speed=km hÿ1

60

70

Figure 4. Mean verbal estimates of speed for the 3 contrast conditions as a function of actual speed. The thin reference line (labeled `match') indicates responses that would match actual speeds. Values for fixed speeds were computed from linear regression of individuals' verbal estimates as functions of actual speed at the time of each response. Error bars represent 95% confidence intervals.

were similar for all contrast conditions. An ANOVA, with the Greenhouse ^ Geisser correction, showed a significant main effect of speed (F1, 14 231, p 5 0:001, Z 2 0:943) but no significant main effect of contrast (F1:35, 18:95 0:42, p 0:59, Z 2 0:03). The ANOVA also showed a significant interaction of contrast6speed (F1:47, 20:6 11:26, p 0:001, n 2 0:446). A posteriori analyses confirmed that the sources of the interaction were significant differences between the clear-view and the filter 2 conditions at travel speeds of 25 and 70 km hÿ1 (t14 2:558, p 0:023, and t14 2:334, p 0:035, respectively). Relative to the clear-view condition, estimates at the slowest speed (25 km hÿ1 ) were significantly lower in the filter 2 condition. The opposite effect occurred at the highest speed (70 km hÿ1), when relative to the clear-view condition, estimates were significantly faster in the filter 2 condition. There was no significant effect of contrast at the intermediate speeds (40, 50, or 60 km hÿ1). Thus, at the slowest speed, participants estimated that they were traveling slower in the lower-contrast condition (filter 2) than in the clear-view condition. This is similar to the `Thompson effect'. At the highest speed, however, they estimated speed as faster (contrary to the `Thompson effect') with the lowest-contrast filter as compared to clear-view condition. Contrast had no effect on speed estimates at intermediate speeds for any filter, and it had no effect for filter 1 at any speed. 3.4 Adjusted speed Figure 5 presents the mean adjusted speed as a function of requested speed for each contrast condition. Again, consistent with previous studies (eg Barch 1958), the mean adjusted speeds were generally faster than the requested speed, especially in the clearview condition, indicating that participants generally underestimated their actual speed.


1207

90 clear view Mean adjusted speed=km hÿ1

80

filter 1

70

filter 2

60

match

50 40 30 20

20

30

40 50 Requested target speed=km hÿ1

60

70

Figure 5. Mean adjusted speeds for three contrast conditions as functions of requested target speed. The thin reference line (labeled `match') indicates adjustments that would match requested target speeds. Error bars represent 95% confidence intervals.

Of greater interest for the present purposes, however, was the finding that the speed adjustments were consistently slower under reduced-contrast conditions, indicating that participants perceived they were traveling faster in the lower-contrast condition relative to the clear-view condition. An ANOVA, with the Greenhouse ^ Geisser correction, showed significant effects of contrast (F1:9, 24:67 38:19, p 5 0:001, Z 2 0:746), target speed (F2:04, 26:58 298:51, p 5 0:001, Z 2 0:958), and the interaction of contrast 6speed (F4:14, 53:77 2:92, p 0:03, Z 2 0:183). Pairwise comparisons showed significant differences between all three contrast conditions ( p 5 0:01). As seen in figure 5, the significant interaction resulted from a greater effect of reduced contrast at higher speeds; that is, relative to the clear-view condition, the reduction of adjusted speed in the filter 2 condition was greater at higher speeds. This implies that the participants felt they were traveling faster when contrast was lower, which is contrary to the `Thompson effect'. 3.5 Minimum stopping distance The third measure of speed perception was based on judgments of the last moment that the driver judged that she/he could stop before reaching the location of a model wallaby at each test location. The distance from each `last-moment stop' response to the wallaby was determined by analysis of the parallax-based video data. Actual speed was recorded simultaneously for each minimum stopping-distance response. Data for three participants were excluded from the analysis because they apparently misunderstood the task: one failed to respond until after passing the model wallaby; the other two responded at approximately the same distance regardless of actual speed. As in the verbal-estimation trials, participants were encouraged to travel at a `comfortable speed', and therefore, the actual speeds and stopping-distance responses varied across participants. In order to compare the responses of all participants at a common set of actual speeds, individuals' data for each of three contrast conditions were fitted by linear regression, providing three stopping-distance functions for each participant (one for each contrast level). The resulting stopping-distance equations of all participants are listed in the appendix. Goodness-of-fit for linear functions of the eleven participants was good, with mean r-values of 0.81 and 0.92 (table 2).

1208


The stopping-distance functions were used to derive comparable minimum stopping distances using the values of 25, 40, 50, 60, and 70 km hÿ1. As shown in figure 6, mean stopping-distance estimates increased as a function of increasing speed (F1, 10 88:04, p 5 0:001, Z 2 0:898). Although stopping distance estimates tended to be longer in the filter 1 condition, the main effect of contrast was not statistically significant (F1:52, 15:23 0:80, p 0:44, Z 2 0:074), nor was the interaction between speed6contrast (F1:94, 19:39 0:39, p 0:68, Z 2 0:038). These data indicate that most participants responded at increasing distances with increasing speed, and reduced contrast had no significant effect on their responses.

Mean estimated distance=m

90 80

clear view

70

filter 1 filter 2

60 50 40 30 20 10

20

30

40 50 Speed=km hÿ1

60

70

Figure 6. Mean minimum stopping distances for the three contrast conditions as a function of actual speed. Values for fixed speeds were computed from linear regression of individuals' stopping-distance responses as functions of actual travel speed at the time of each response. Error bars represent 95% confidence intervals.

3.6 Questionnaire Responses to the post-test questionnaire, which used rating scales of 1 to 10, indicated that: (i) Participants rated their view through the diffusing filters as somewhat similar to actual fog (mean 6:5 1:8, where `1' meant very different and `10' meant identical to natural fog), indicating that the appearance was moderately realistic. (ii) They rated their speed in lower-contrast conditions as generally slower than that with clear vision (2:9 1:2, where `1' meant much slower and `10' meant the same as in clear-view condition), indicating that, consistent with figure 3, they believed they traveled substantially more slowly under low-contrast conditions. (iii) They rated the accuracy of speed estimates with the filters as lower than that with clear vision (3:5 0:8, where `1' meant much less accurate and `10' meant equally accurate), indicating lower confidence for the low-contrast responses. (iv) They rated the accuracy of their stopping-distance estimates with filters as lower than their accuracy with clear vision (3:6 1:3, where `1' meant much less accurate and `10' meant equally accurate). Responses to additional questions, which requested confidence ratings on a percentage scale, indicated that participants were 65.7% (23:2) as confident of their speed judgments with the filters, and 61.1% (29:7) as confident of their stopping distance judgments. Finally, they estimated that they had traveled only 61.4% (23:4) as fast with the filters as in the clear-view condition.


1209

4 Discussion The object of this study was to investigate the effects of reduced contrast on the perception and control of speed while driving in real-world conditions. Previous research had shown that the perceived speed of external stimuli decreases with reduced contrast, suggesting that drivers might experience an illusion of reduced speed and, hence, might travel faster than they realized when driving in fog. Other studies had shown, however, that visual perception of self-motion (vection) and control of heading when driving (steering accuracy) are relatively insensitive to variations of blur and luminance, which raised the possibility that perception of self-motion is less susceptible than perception of object motion to the effects of reduced contrast (Leibowitz et al 1979; Owens and Tyrrell 1999; Brooks et al 2005; Owens et al 2007). The central question for the present study was: Do drivers experience an illusion of decreased speed when the contrast of the driving environment is decreased? The overall results, collapsed across all 7 test locations for the free-flowing verbal estimation trials, showed that mean travel speeds decreased progressively with reduced contrast, and the reduction in speed of travel was accompanied by a parallel decrease in verbal estimates of speed (figure 3). It seems plausible that slower speeds in the reduced-contrast trials reflect increasing caution under these conditions, which are similar to the effects seen for simulated (Wood and Troutbeck 1994, 1995) and true (Wood 2002; Wood and Carberry 2006) cataracts. With respect to absolute accuracy of participants' responses, our results are consistent with several earlier studies, which reported that, when tested on the road without view of a speedometer, participants generally underestimate speed and drive at higherthan-requested speeds (Barch 1958; Denton 1966; Evans 1970; Triggs and Berenyi 1982; Recarte and Nunes 1996). For the present purposes, however, differences among the three contrast conditions are of greater interest than absolute accuracy, because any effects of reduced contrast on the perception and control of speed should result in systematic differences between performance under reduced contrast as compared with clear-view conditions. Our investigation included three dependent measures to assess the effect of reduced contrast on speed perception: (i) verbal estimation of speed; (ii) adjustment of speed to match requested levels, and (iii) an indirect measure based on estimates of minimum stopping distance. Each was assessed at 5 speeds of travel. Of the 15 comparisons (3 DVs65 speeds), only oneöverbal estimates at the slowest speed (25 km hÿ1 , figure 4)öshowed an illusory underestimation of speed that resembles the `Thompson effect'. Eight comparisons showed no effect of reduced contrast överbal estimates at 40, 50, and 60 km hÿ1 (figure 4), and stopping distance estimates at all 5 speeds (figure 6). Six comparisons showed a significant overestimation of perceived speed, which is contrary to the `Thompson effect'överbal estimates at 70 km hÿ1 (figure 4) and speed adjustments for all 5 requested speeds (figure 5). We conclude that, contrary to many studies of perceived object motion, when drivers are tested under real-world conditions, their ability to perceive and control their own speed of travel is not adversely affected by large variations in the visual contrast of their visible surroundings. These findings weigh against the hypothesis that, under foggy conditions, drivers experience an illusory reduction of speed that causes them to travel faster than they realize. The present findings may seem surprising in view of the extensive literature documenting illusory reductions of perceived speed with reduced contrast. How did our participants avoid or compensate for such perceptual difficulties? We offer three possible answers: First, non-visual factors may have contributed to our participants' perception of speed. Evans (1970), for example, found that auditory information enhances the accuracy

1210


of speed estimates made by passengers in a moving motor vehicle. It also seems plausible that inertial forces from the vehicle's motion are important, as evidenced by qualitative differences in driving experience obtained with moving-base as compared with most fixed-base simulators. The present study manipulated only contrast, while non-visual sources of perceptual information were essentially constant across test conditions. It is possible, therefore, that non-visual factors negated or compensated for illusory visual effects. Further research is needed to explore these interesting questions about the interactions of visual and non-visual information while driving. Nevertheless, we assert that drivers' performance under the real-world conditions of the present study are more likely to represent what actually happens when one is driving in fog than do the previously reported laboratory-based simulations. Second, our manipulation of contrast may have been ineffective. The diffusing filters of the present experiment produced a uniform reduction of spatial contrast, rather than the progressive (exponential) reduction of contrast with increased distance that occurs in natural fog. While it is true that our low-contrast conditions differed from natural fog, they did, nevertheless, reduce the spatial contrast of visible surroundings by 0.5 to 1.5 log units (figure 2). Most previous studies also used uniform reductions of contrast (eg by manipulating the overall contrast of a video display) rather than a progressive diffusing medium like natural fog, and the magnitude of those contrast manipulations fell within those of the present experiment, although the stimuli in the earlier psychophysical experiments were generally much smaller and simpler than a full view of the natural environment (Snowden et al 1998; Horswill and Plooy 2008). Moreover, participants in the laboratory experiments were not moving, so their visual perception of self-motion could be questioned. Based on the earlier psychophysical research, it seems that our manipulation of contrast should have been sufficient to cause illusory reductions of perceived speed. But it did not. A third explanation of the difference between previous findings and the present study draws upon the distinction between àmbient' and `focal' vision (Schneider 1967, 1969; Held 1968, 1970; Trevarthen 1968; Norman 2002). A sizable literature indicates that functions related to visual recognition (focal vision) and guidance of action (ambient vision) involve different neural processes, which operate largely in parallel at multiple levels in the visual pathways, from parvo- and magno-type ganglion cells through the ventral and dorsal cortical pathways for recognition and guidance, respectively (eg Mishkin et al 1983; Goodale and Milner 1992; Broussaoud et al 1996; Kleinschmidt et al 2002). Evidence from studies of vection and vehicle control indicates that focal and ambient visual functions are differentially affected by optical blur and low luminance (eg Leibowitz et al 1979). For example, studies of night driving have shown that steering remains accurate with high levels of blur and very low (scotopic) luminance, when measures of visual recognition are severely degraded (Owens and Tyrrell 1999; Higgins et al 1998; Brooks 2005; Owens et al 2007). If the role of ambient vision in driving includes control of speed as well as heading, the present findings are fully consistent with research on driving performance under blurred and low luminance conditions. A fundamental difference between focal and ambient visual functions is field of view (FOV). Focal functions, like acuity and pattern recognition, depend primarily on the central (parafoveal) portion of the visual field, whereas ambient vision, especially the guidance of locomotion, utilizes information from a much wider area of the visual field. Research has shown, for example, that vection and steering accuracy are enhanced by a wide FOV (Leibowitz et al 1979; Owens and Tyrrell 1999). Following the pioneering work of James Gibson (1954, 1958), it has become clear that dynamic information from optic flow, which under most natural conditions fills the FOV, is important for visual perception and control of locomotion.


1211

The importance of optic flow for visually guided locomotion suggests that differences in FOV may provide a parsimonious explanation for the discrepancy between the present findings and earlier simulator studies. It appears that all experiments that have reported decreased perceived speed with reduced contrast have used central stimuli and a relatively narrow FOV. Most relevant are the simulator studies: Snowden et al (1998) used a desktop monitor of unspecified dimensions; Horswill and Plooy (2008) used a projected video display that was 37 deg wide. Both of these are narrower than the real-world conditions of the present study where drivers' horizontal FOV was unrestricted. It seems likely that displays with a narrow FOV would favor activation of focal processes that mediate perception of object motion, especially for observers who are seated in a stationary environment. Conversely, a wide FOV, filled with the optic flow that accompanies actual locomotion, would provide strong activation of ambient visual processes. From this viewpoint, the `Thompson effect' (decreased perceived speed with reduced contrast) is a phenomenon specific to focal perception of object motion; the perception and control of self-motion, however, depends on ambient visual processes which are relatively unaffected by reduced contrast. This interpretation gains added support from an unpublished doctoral dissertation by Pretto (2008). Using two driving simulators with wide FOVs (75 and 230 deg), he found that a uniform reduction of contrast had no effect on drivers' adjustment of speed. Moreover, in experiments that created an exponential density gradient analogous to natural fog, Pretto obtained speed adjustments that were slower in the fog than in clear conditions, a finding that appears to replicate the present results (figure 5). Pretto points out that, unlike uniform reductions of contrast, the progressive reduction of naturalistic fog has relatively little effect on the contrast of proximal areas of the optic flow field, which contain high angular velocities, while filtering out more distal areas of optic flow, which contain lower angular velocities. Because the optic flow field seen in natural fog comprises higher average velocity than the same field in clear conditions, drivers may misperceive their self-motion to be faster in fog than in clear conditions (see also conference abstracts of Dyre et al 2005 and Shrivastava et al 2005). More generally, the present findings highlight interesting challenges in generalizing evidence from simulator studies to driving in the real world. Simulators offer clear advantages in experimental control of visual stimulation and operational safety for participants. These advantages are often counterbalanced, however, by important limitations in the information available for perception (eg FOV) and in behavioral consequences of actual risks associated with driving. On the road, drivers may take good advantage of perceptual informationöboth visual and non-visualöthat is missing in most simulations, and they may automatically adjust their behavior to maintain a tolerable level of risk. On the other hand, simulators afford opportunities to dissect certain aspects of the perceptual environment, and they enable investigation of factors that could pose unacceptable hazards on the road. Thus, complementary purposes can be served by research on the road and in simulators. Ultimately, of course, we must balance exploratory benefits of simulation with validation of tests conducted under actual operational conditions. In summary, the present findings suggest that ambient visual functions are robust under adverse conditions of low contrast as well as low luminance and blur, when focal functions are degraded. The illusory reduction of perceived speed in low-contrast conditions, well documented by previous psychophysical studies, may be another characteristic of focal processes that does not generalize to ambient processes engaged in the visual control of action. Further studies are needed to compare effects of reduced contrast on perception of object motion as compared with self-motion in the same participants.

1212


Acknowledgments. This research was supported by grants from Franklin & Marshall College, Queensland University of Technology, and the Australian Research Council. We would like to express appreciation to Queensland Transport for allowing the use of the facilities at the Mt Cotton Driver Training Center and to the staff of the Mt Cotton Center for their generous cooperation and support. We are grateful to Herschel W Leibowitz, Johnell Brooks, Peter Thompson, Tom Bandon, and an anonymous reviewer for insightful comments on earlier drafts of the manuscript, and to Tabitha Carberry for assistance in data collection. References Al-Ghamdi A S, 2007 `Èxperimental evaluation of fog warning system'' Accident Analysis and Prevention 39 1065 ^ 1072 Barch A M, 1958 ``Judgments of vehicle speed on the open highway'' Journal of Applied Psychology 42 362 ^ 366 Blakemore M R, Snowden R J, 1999 ``The effect of contrast upon perceived speed: A general phenomenon?'' Perception 28 33 ^ 48 Boussaoud D, Pellegrino G di, Wise S P, 1996 ``Frontal lobe mechanisms subserving vision-foraction versus vision-for-perception'' Behavioural Brain Research 72 1 ^ 15 Brooks J O, 2005 The Actual and Estimated Ability of Younger and Older Drivers to See and Steer in Challenging Conditions: A Test of the Selective Degradation Hypothesis Doctoral Dissertation, Clemson University, Clemson, SC Brooks J O, Tyrrell R A, Frank T A, 2005 ``The effects of severe visual challenges on steering performance in visually healthy young drivers'' Optometry and Vision Science 82 689 ^ 697 Brooks K, 2001 ``Stereomotion speed perception is contrast dependent'' Perception 30 725 ^ 731 Brown I, 1970 ``Motorway crashes in fogöWho's to blame?'' New Scientist 48 544 ^ 545 Campbell F W, Maffei L, 1981 ``The influence of spatial frequency and contrast on the perception of moving patterns'' Vision Research 21 713 ^ 721 Cavallo V, Colomb M, Dore¨ J, 2001 ``Distance perception of vehicle rear lights in fog'' Human Factors 43 442 ^ 451 Denton G G, 1966 `À subjective scale of speed when driving a motor vehicle'' Ergonomics 9 203 ^ 210 Distler H, Bu«lthoff H H, 1996 ``Velocity perception in 3-D environments'' [Abstract] Perception 25 Supplement, 58 Dyre B P, Schaudt W A, Lew R T, 2005 ``Contrast gradients increase apparent egospeed while moving through simulated fog'' [Abstract] Journal of Vision 5 335 Evans L, 1970 ``Speed estimation from a moving automobile'' Ergonomics 13 219 ^ 230 Gibson J J, 1954 ``The visual perception of objective motion and subjective movement'' Psychological Review 61 304 ^ 314 Gibson J J, 1958 ``Visually controlled locomotion and visual orientation in animals'' British Journal of Psychology 49 182 ^ 194 Goodale M A, Milner A D, 1992 ``Separate visual pathways for perception and action'' Trends in Neurosciences 15 20 ^ 25 Held R, 1968 ``Dissociation of visual functions by deprivation and rearrangement'' Psychologische Forschung 31 338 ^ 348 Held R, 1970 ``Two modes of processing spatially distributed visual stimulation'', in The Neurosciences: Second Study Program Ed. F O Schmidt (New York: Rockefeller University Press) pp 317 ^ 324 Higgins K E, Wood J M, 2005 ``Predicting components of closed road driving performance from vision tests'' Optometry and Vision Science 82 647 ^ 656 Higgins K, Wood J, Tait A, 1998 ``Vision and driving: Selective effect of optical blur on different driving tasks'' Human Factors 41 224 ^ 232 Horswill M S, Plooy A M, 2008 ``Reducing contrast makes speed in a video-based driving simulator harder to discriminate as well as making them appear slower'' Perception 37 1269 ^ 1275 Jones K, Bentley B E, Wood J M, Woolf M I, 1998 `Àpplication of parallax for the measurement of visibility distances in the open-road environment'' International Archives of Photogrammetry and Remote Sensing 32(5) 74 ^ 79 Jones K, Wood J M, Woolf M I, Bentley B E, 2010 `À technique for on-road assessment of road sign visibility distances'', in Vision in Vehicles Ed. A Gale, in press (Publisher n.a.) Kleinschmidt A, Thilo K V, Bu«chel C, Gresty M A, Bronstein A M, Frackowiak R S J, 2002 ``Neural correlates of visual motion perception as object- or self-motion'' NeuroImage 16 873 ^ 882 Leibowitz H W, Owens D A, 1977 ``Nighttime driving accidents and selective visual degradation'' Science 197 422 ^ 423


1213

Leibowitz H W, Owens D A, Post R B, 1982, September ``Nighttime driving and visual degradation'' Society of Automotive Engineers Technical Paper Series No 820414 (Warrendale, PA: Society of Automotive Engineers) Leibowitz H W, Shupert C, Dichgans J, 1979 ``The independence of dynamic spatial orientation from luminance and refractive error'' Perception & Psychophysics 25 75 ^ 79 Mishkin M, Ungerleider L G, Macko K A, 1983 `Òbject vision and spatial vision: Two cortical pathways'' Trends in Neurosciences 6 414 ^ 417 Norman J, 2002 ``Two visual systems and two theories of perception: An attempt to reconcile the constructivist and ecological approaches'' Behavioral & Brain Sciences 25 73 ^ 144 Owens D A, Tyrrell R A, 1999 `Èffects of luminance, blur, and age on nighttime visual guidance: A test of the selective degradation hypothesis'' Journal of Experimental Psychology: Applied 5 115 ^ 128 Owens D A, Wood J M, Owens J M, 2007 `Èffects of age and illumination on night driving: A road test'' Human Factors 49 1115 ^ 1131 Po«ppel E, Held R, Frost D, 1973 ``Residual visual function after brain wounds involving the central visual pathways in man'' Nature 243 295 ^ 296 Pretto P, 2008 The Perception and Production of Speed During Self-motion: Evidence for Non-optimal Compensation Mechanisms'' Doctoral Dissertation, Universita Degli Studi di Padova, Padua, Italy Recarte M A, Nunes L, 1996 ``Perception of speed in an automobile: Estimation and production'' Journal of Experimental Psychology: Applied 2 291 ^ 304 Schneider G E, 1967 ``Contrasting visuomotor functions of tectum and cortex in the golden hamster'' Psychologische Forschung 31 52 ^ 62 Schneider G E, 1969 ``Two visual systems: Brain mechanisms for localization and discrimination are dissociated by tectal and cortical lesions'' Science 163 895 ^ 902 Shepard F D, 1996 ``Reduced visibility due to fog on the highway, NCHRP Synthesis Report No 228, National Cooperative Research Program, Transportation Research Board (Washington, DC: National Academy Press) Shrivastava A, Hayhoe M M, Pelz J B, Mruczek R, 2005 `Ìnfluence of optic flow field restrictions and fog on perception of speed in a virtual driving environment'' [Abstract] Journal of Vision 5 139 Snowden R J, Stimson N, Ruddle R A, 1998 ``Speed perception fogs up as visibility drops'' Nature 392 450 Stone L S, Thompson P, 1992 ``Human speed perception is contrast dependent'' Vision Research 32 1535 ^ 1549 Sumner R, Baguley C, Burton J, 1977 ``Driving in fog on the M4'' TRRL Supplementary Report 281, Transport and Road Research Laboratory (Crowthorne, Berks, UK: Department of Transport) Thompson P, 1976 Velocity Aftereffects and the Perception of Movement unpublished PhD Dissertation (Cambridge: University of Cambridge) Thompson P, 1982 ``Perceived rate of movement depends on contrast'' Vision Research 22 377 ^ 380 Thompson P, Brooks K, Hammett S T, 2006 ``Speed can go up as well as down at low contrast: Implications for models of motion perception'' Vision Research 46 782 ^ 786 Trevarthen C, 1968 ``Two mechanisms of vision in primates'' Psychologische Forschung 31 299 ^ 337 (in German) Triggs T J, Berenyi J S, 1982 `Èstimation of automobile speeds under day and night conditions'' Human Factors 24 111 ^ 114 Tyrrell R A,Wood J M, Carberry T P, 2004 `Òn-road measures of pedestrians' estimates of their own nighttime conspicuity'' Journal of Safety Research 35 483 ^ 490 Tyrrell R A, Wood J M, Chaparro A, Carberry T P, Chu B S, Marszalek R P, 2009 ``Seeing pedestrians at night: Visual clutter does not mask biological motion'' Accident Analysis & Prevention 41 506 ^ 512 Weiskranz L, 1986 Blindsight: A Case Study and Implications (New York: Oxford University Press) Wood J M, 2002 `Àge and visual impairment decrease driving performance as measured on a closedroad circuit'' Human Factors 44 482 ^ 494 Wood J M, Carberry T P, 2006 ``Bilateral cataract surgery and driving performance'' British Journal of Ophthalmology 90 1277 ^ 1280 Wood J M, Troutbeck R, 1994 `Èffect of visual impairment on driving'' Human Factors 36 476 ^ 487 Wood J M, Troutbeck R, 1995 `Èlderly drivers and simulated visual impairment'' Optometry and Visual Science 72 115 ^ 124 Wood J M, Tyrrell R A, Carberry T P, 2005 ``Limitations in drivers' ability to recognise pedestrians at night'' Human Factors 47 644 ^ 653

1214


Appendix Table A1. Verbal speed estimation. Linear regressions. Subject

Condition

Regression equation

r2

r

A

clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter

0:76x 4:79 0:97x ÿ 5:33 0:88x ÿ 1:14 0:99x ÿ 9:39 0:77x 1:28 1:26x ÿ 18:13 0:96x ÿ 2:16 1:05x ÿ 6:38 1:10x ÿ 13:39 0:52x 24:98 0:65x 19:40 0:50x 7:68 0:67x 3:71 0:67x 8:73 0:69x 5:75 1:29x ÿ 22:46 1:35x ÿ 25:35 1:31x ÿ 22:28 0:88x 2:52 0:97x ÿ 4:76 0:88x 0:33 0:37x 14:63 0:75x ÿ 1:28 1:01x ÿ 12:58 0:60x 24:40 0:85x 3:40 1:21x ÿ 9:01 1:07x ÿ 19:14 1:22x ÿ 16:48 1:36x ÿ 18:6 1:28x ÿ 16:38 1:29x ÿ 17:10 1:34x ÿ 27:26 1:04x ÿ 0:62 0:96x 1:53 1:72x ÿ 18:73 1:29x ÿ 26:77 1:41x ÿ 29:39 1:71x ÿ 34:99 0:95x ÿ 2:93 1:03x ÿ 3:31 1:52x ÿ 27:47 0:90x ÿ 8:24 1:12x ÿ 16:7 1:17x ÿ 18:5

0.79 0.93 0.81 0.73 0.81 0.99 0.95 0.92 0.92 0.77 0.68 0.61 0.87 0.97 0.92 0.91 0.94 0.93 0.67 0.74 0.77 0.21 0.87 0.82 0.82 0.95 0.92 0.93 0.92 0.97 0.98 0.90 0.89 0.86 0.91 0.97 0.83 0.87 0.84 0.93 0.88 0.88 0.94 0.89 0.91

0.89 0.97 0.90 0.85 0.90 0.99 0.98 0.96 0.96 0.88 0.83 0.78 0.93 0.98 0.96 0.95 0.97 0.96 0.82 0.86 0.88 0.46 0.93 0.90 0.91 0.97 0.96 0.96 0.96 0.99 0.99 0.95 0.94 0.93 0.95 0.98 0.91 0.93 0.92 0.96 0.94 0.94 0.97 0.94 0.95

B C D E F G H I J K L M N O

view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2


1215

Table A2. Minimum stopping distance. Linear regressions. Subject

Condition

Regression equation

r2

r

A

clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter filter clear filter

0:40x 6:72 0:39x 4:51 0:55x 1:82 0:63x 0:31 0:35x 12:97 0:65x 13:69 1:02x 7:756 0:56x 1:07 0:55x 3:78 0:98x ÿ 0:886 1:07x ÿ 3:53 1:12x ÿ 5:52 1:29x ÿ 2:11 1:26x ÿ 3:636 1:56x ÿ 15:72 1:24x ÿ 19:01 0:90x 7:45 0:80x 6:20 1:03x 5:22 1:86x 25:07 0:79x 25:53 0:72x 8:20 0:88x 4:88 0:97x 1:22 1:27x ÿ 27:13 1:06x ÿ 17:04 0:81x ÿ 4:508 0:95x 7:34 1:33x ÿ 17:86 1:19x ÿ 18:61 0:64x ÿ 0:228 0:78x ÿ 4:52

0.91 0.87 0.87 0.82 0.87 0.36 0.82 0.83 0.74 0.82 0.83 0.74 0.92 0.82 0.84 0.87 0.83 0.89 0.79 0.59 0.61 0.83 0.86 0.93 0.92 0.69 0.65 0.68 0.88 0.70 0.88 0.98

0.96 0.93 0.93 0.91 0.93 0.60 0.90 0.91 0.86 0.91 0.91 0.86 0.96 0.91 0.92 0.93 0.91 0.94 0.89 0.77 0.78 0.91 0.93 0.96 0.96 0.83 0.81 0.83 0.94 0.84 0.94 0.99

B C D E F H I J K L

view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1 2 view 1

ß 2010 a Pion publication

N:/psfiles/banners/ final-per.3d

ISSN 0301-0066 (print)

ISSN 1468-4233 (electronic)

www.perceptionweb.com

Conditions of use. This article may be downloaded from the Perception website for personal research by members of subscribing organisations. Authors are entitled to distribute their own article (in printed form or by e-mail) to up to 50 people. This PDF may not be placed on any website (or other online distribution system) without permission of the publisher.