Foggy perception slows us down

7/31/2019 Foggy perception slows us down

1/12

elife.elifesciences.org

Pretto et al. eLife 2012;1:e00031. DOI: 10.7554/eLife.00031 1 of 12

Foggy perception slows us downPaolo Pretto 1*, Jean-Pierre Bresciani 2,3, Gregor Rainer 3, Heinrich H Blthoff 1*

1Department o Human Perception, Cognition and Action, Max Planck Institute orBiological Cybernetics, Tbingen, Germany; 2Psychology and NeuroCognitionLaboratory, University Pierre Mends-France and CNRS, Grenoble, France;3Department o Medicine, University o Fribourg, Fribourg, Switzerland

Abstract Visual speed is believed to be underestimated at low contrast, which has beenproposed as an explanation o excessive driving speed in og. Combining psychophysicsmeasurements and driving simulation, we con rm that speed is underestimated when contrast isreduced uni ormly or all objects o the visual scene independently o their distance rom the viewer.However, we show that when contrast is reduced more or distant objects, as is the case in realog, visual speed is actually overestimated, prompting drivers to decelerate. Using an arti cialanti- ogthat is, og characterized by better visibility or distant than or close objects, wedemonstrate or the rst time that perceived speed depends on the spatial distribution o contrastover the visual scene rather than the global level o contrast per se. Our results cast new light onhow reduced visibility conditions a ect perceived speed, providing important insight into thehuman visual system.DOI: 10.7554/eLi e.00031.001

Introduction Visual contrast is usually re erred to as the di erence in brightness between an object and the back-

ground ( Hofstetter et al., 2000 ). Classical vision research experiments systematically investigatedhow visual contrast a ects objects motion perception ( Thompson, 1982 ; Stone and Thompson,1992 ; Blakemore and Snowden, 1999 ; Anstis, 2003 ). These studies have shown that the perceivedspeed o two-dimensional moving objects or example, plaid patterns on a computer screenisunderestimated when visual contrast is reduced. More recent studies based on driving scenarios havesuggested that the underestimation o visual speed at low contrast applies also to perceived sel -motion in three-dimensional environments ( Snowden et al., 1998 ; Horswill and Plooy, 2008 ; Owens et al., 2010 ). This nding was proposedand is still consideredas a possible explanation or exces-sive driving speed in og.

In the abovementioned studies, and more generally in all studies having assessed the e ect o visualcontrast on motion perception, contrast was reduced uni ormly or all objects o the visual scene, irre-spective o their distance rom the observer. While uni orm contrast reduction is a valid model to assess theperception o two-dimensional patterns moving on a computer screen, it is a poor model to investigate

how an atmospheric phenomenon like og a ects motion perception in three-dimensional environments.Speci cally, og alters visual contrast because tiny water droplets suspended in the air are interposedbetween the observer and the surrounding objects. The quantity o droplets increases along the line-o -sight as distance rom the observer increases. As a consequence, the contrast o the visual scenedecreases with distance, and visibility is better or close than or distant objects. There ore, the distance-dependent attenuation o contrast experienced in og is obviously very di erent rom the uni ormreduction traditionally adopted in vision research. Uni orm contrast reduction resembles more closelywhat one could experience when looking at the environment through a dirty glass or a oggy windshield.

To date, the e ects o distance-dependent contrast reduction on motion perception are unknown.In other words, we still do not know how og a ects perceived sel -motion. Here, we tested the

*For correspondence: [email protected] (PP);heinrich.bueltho @tuebingen.mpg.de (HHB)

These authors contributedequally to this work

Competing interests: The authorshave declared that no competinginterests exist

Funding: See page 11

Received: 12 July 2012Accepted: 05 September 2012Published: 30 October 2012

Reviewing editor : Jody CCulham, University o WesternOntario, Canada

Copyright Pretto et al. Thisarticle is distributed under theterms o the Creative CommonsAttribution License , whichpermits unrestricted use andredistribution provided that theoriginal author and source arecredited.

RESEARCH ARTICLE
http://elife.elifesciences.org/http://dx.doi.org/10.7554/eLife.00031http://dx.doi.org/10.7554/eLife.00031.001http://creativecommons.org/licenses/by/3.0/http://creativecommons.org/licenses/by/3.0/http://creativecommons.org/licenses/by/3.0/http://creativecommons.org/licenses/by/3.0/http://dx.doi.org/10.7554/eLife.00031.001http://dx.doi.org/10.7554/eLife.00031http://creativecommons.org/licenses/by/3.0/http://www.elifesciences.org/the-journal/open-access/http://elife.elifesciences.org/


2/12

Neuroscience


Research article

perceptual and behavioural e ects o distance-dependent contrast reductionthat is, ogon speedperception. We compared these e ects with those o the distance-independentthat is, uni ormcontrast reduction that has been used in previous studies to simulate og ( Snowden et al., 1998 Horswill and Plooy, 2008 ; Owens et al., 2010 ). To per orm these experiments, we used a state-o -the-art virtual reality setup allowing us to realistically simulate og in an ecological driving scenario(Loomis et al., 1999 ).

ResultsIn the rst experiment, we used a standard psychophysical procedure to test how contrast a ectsperceived visual speed. Twelve experienced drivers were presented with pairs o driving scenes andinstructed to estimate which scene moved aster ( Figure 1A ). One o the scenes (re erence) had clearvisibility and moved at one o three target speeds (40, 60, or 90 km/hr). The other scene (test) had

either clear or reduced visibility, and its speed was adjusted or each trial using a Bayesian adaptivemethod ( Kontsevich and Tyler, 1999 ). This method allowed us to determine the point o subjectiveequality (PSE) as well as the just-noticeable di erence (JND). The PSE corresponded to the speed atwhich the two scenes were perceived as moving equally ast. There ore, PSEs higher than the actualspeed o the re erence scene indicated speed underestimation, whereas PSEs lower than the speed o the re erence scene indicated speed overestimation. The JND corresponded to the smallest detect-able di erence between two di erent speeds. High JNDs indicated low discrimination sensitivity,whereas low JNDs indicated high discrimination sensitivity.

The contrast o the scene was reduced either in a distance-dependent manner, as would happenin natural og, or in a distance-independent manner, as has been done in previous experiments.

eLife digest The ways people respond to conditions o reduced visibility is a central topic invision research. Notably, it has been shown that people tend to underestimate speeds when visibilityis reduced equally at all distances, as or example, when driving with a ogged up windshield. Butwhat happens when the visibility decreases as you look urther into the distance, as happens whendriving in og? Fortunately, as new research reveals, people tend to overestimate their speed whendriving in og-like conditions, and show a natural tendency to drive at a slower pace.

Pretto et al. per ormed a series o experiments involving experienced drivers and high-qualityvirtual reality simulations. In one experiment, drivers were presented with two driving scenes andasked to guess which scene was moving aster. In the re erence scene, the car was driving at a xedspeed through a landscape under conditions o clear visibility; in the test scene, it was movingthrough the same landscape, again at a xed speed, but with the visibility reduced in di erent ways.The experiments showed that drivers overestimated speeds in og-like conditions, and theyunderestimated speeds when the reduction in visibility did not depend on distance. Furtherexperiments con rmed that these perceptions had an infuence on driving behaviour: driversrecorded an average speed o 85.1 km/hr when the visibility was good, and this dropped to 70.9km/hr in severe og. However, when visibility was reduced equally at all distances, as happens with aogged up windshield, the average driving speed increased to 101.3 km/hr.

Based on previous work, Pretto et al. developed the theory that the perception o speed isinfuenced by the relative speeds o the visible regions in the scene. When looking directly into the

og, visibility is strongly reduced in the distant regions, where the relative motion is slow, and ispreserved in the near regions, where the motion is ast. This visibility gradient would lead to speedoverestimation. To test this theory, the experiments were repeated with new drivers under threedi erent conditions: good visibility, og, and an arti cial situation called anti- og in which visibility ispoor in the near regions and improves as the driver looks urther into the distance. As predicted, theestimated speed was lower in anti- og than in clear visibility and og. Conversely, the driving speedwas 104.4 km/hr in anti- og compared with 67.9 km/hr in good visibility and 51.3 km/hr in og.

Overall, the results show that the perception o speed is infuenced by spatial variations invisibility, and they strongly suggest that this is due to the relative speed contrast between thevisible and covert areas within the scene.DOI: 10.7554/eLi e.00031.002
http://dx.doi.org/10.7554/eLife.00031http://dx.doi.org/10.7554/eLife.00031.002http://dx.doi.org/10.7554/eLife.00031.002http://dx.doi.org/10.7554/eLife.00031


3/12

Neuroscience


Research article

The di erence between these two types o contrast reduction is represented in Figure 2B and C Moderate and severe levels o reduction were implemented or each type o contrast alteration.

Importantly, or each level o reduction, the overall visual contrast was the same or distance-dependentand distance-independent alteration (Root mean square [RMS] contrast = 0.31 or moderate visibilityreduction and 0.19 or severe visibility reduction, vs 0.46 or clear visibility). In total, the experimentconsisted o ve visibility conditions: clear (no contrast reduction), moderate and severe og (distance-dependent contrast reduction), moderate and severe uni orm reduction (distance-independent contrastreduction).

Reducing the contrast o the visual scene altered speed perception [F (4,44) = 52.086, p


4/12

Neuroscience


Research article

Figure 2 . Visibility conditions. ( A) Clear weather conditions (clear visibility): contrast is unaltered and the visibility isoptimal in all directions (brown line). ( B) Distance-independent contrast reduction (uni orm contrast): visibility dropsequally or all objects o the visual scene, irrespective o their distance rom the observer (green lines). ( C) Distance-dependent contrast reduction ( og): visibility is good or close objects, and worsens as distance rom the observerincreases (red lines). ( D) Reversed distance-dependent contrast reduction (anti- og): visibility is poor or closeobjects, and improves as distance rom the observer increases (blue lines). ( E) Pictures o the actual setup: side view(le t) and drivers view (right).DOI: 10.7554/eLi e.00031.004
http://dx.doi.org/10.7554/eLife.00031http://dx.doi.org/10.7554/eLife.00031.004http://dx.doi.org/10.7554/eLife.00031.004http://dx.doi.org/10.7554/eLife.00031


5/12

Neuroscience


Research article

perceived speed under clear visibility (mean = 65 km/hr). This indicates that natural og led to an over-estimation o speed (equivalent to 76.4 and 94.3 km/hr or moderate and severe og, respectively;Figure 3A ). Conversely, when visibility was reduced in a distance-independent manner, higher speeds(PSE mean = 95.5 and 95.2 km/hr or moderate and severe uni orm reduction, respectively) werematched to the perceived speed under clear visibility. There ore, speed was underestimated with uni-orm contrast reduction (perceived speed equivalent to 41.4 and 41.5 km/hr or moderate and severereduction, respectively). Each condition di ered rom each o the others (p


6/12

Neuroscience


Research article

contrast is reduced in a distance-dependent manner. More speci cally, speed overestimation in og

could result rom the relative contrast between the central and peripheral areas o the visual eld.To test this hypothesis, we created an anti- og, that is, a distance-dependent contrast reductioncharacterized by an increase o visibility with distance (see Figure 2D ). With anti- og, visibility was bet-ter or distant than or close objects, that is, visibility was good or the portion o road situated at adistance ahead, and poor or the direct surroundings o the vehicle. In other words, normal og andanti- og resulted in opposite spatial distributions o contrast over the visual scene, although the globalcontrast reduction was the same or both og types. We hypothesized that i the relative contrastbetween central and peripheral areas o the visual eld underlies the speed overestimation observedwith og, then speed should be underestimated with the anti- og.

In the third experiment, we used the same psychophysical procedure as in the rst experiment tocompare the e ect o natural og and anti- og on perceived speed. Ten experienced drivers who hadnot participated in the rst two experiments were presented with pairs o driving scenes and instructedto estimate which scene moved aster (see methods o experiment 1). Three visibility conditions were

used: clear, og, and anti- og. Only one target speed was used (60 km/hr) as the rst two experimentsrevealed the same pattern o results or all three target speeds.

As shown in Figure 4A , perceived speed depended on visibility [F (2,18) = 65.64, p


7/12

Neuroscience


Research article

(mean = 0.248 and 0.225 km/hr or clear and og, respectively). This indicates that estimating drivingspeed was more di cult in the anti- og condition, which likely results rom the completely arti cialnature o this type o contrast reduction. Indeed, anti- og is a type o contrast alteration that neveroccurs in real li e. In that respect, it is interesting to mention that reduced speed discriminationper ormance was observed with both unusual types o contrast reduction, namely anti- og here anduni orm contrast reduction in the rst experiment.

The same 10 participants took part in a ourth experiment where they were instructed to drive attarget speeds with either clear or reduced visibility (same driving procedure as experiment 2). Thisexperiment aimed to compare the e ects o og and anti- og on produced driving speed. The visibilityand speed conditions were identical to those used in experiment 3. The order o experiments 3 and 4was counterbalanced between participants.

Driving speed was a ected by visibility [F (2,18) = 39.99, p


8/12

Neuroscience


Research article

and Thompson, 1992 ; Blakemore and Snowden, 1999 ; Anstis, 2003 ). In the present study, wereproduced this perceptual bias, showing that visual speed is indeed underestimated when contrast isreduced in a distance-independent manner. However, we show that this is only part o the wholepicture. In particular, we demonstrate here or the rst time that an identical global loss o visibility canevoke opposite percepts, depending on the nature o the underlying visual contrast reduction.There ore, contrarily to what has been consistently reported in previous studies, a global contrastreduction can also lead to an overestimation o visual speed. This is notably the case when contrast isnot reduced uni ormly or all objects o the visual scene but varies according to their distance rom theviewer. For instance, in og, contrast reduction is more important or distant than or close objects. Thisgenerates a distance-dependent visibility gradient between the peripheral and central area o thevisual eld. Our results show that in this situation, perceived speed is not determined by the globallevel o contrast per se, but rather by the spatial distribution o contrast over the visual scene. Morespeci cally, perceived speed is determined by the relative contrast between the central and peripheralareas o the visual eld. When visibility is better in the peripheral than in the central visual eld, as isthe case in og, speed is overestimated. Inverting the direction o the contrast gradient with anti- ogand thereby obscuring more the peripheral than the central region o the visual eld, inverts theperceptual bias such that speed is now underestimated. This highlights the critical role o the visibilitygradient in perceived speed, explaining why speed is unexpectedly overestimated in og despite aglobal reduction o visibility. Importantly, our results also evidence the direct relationship betweenperceived and produced speed. Speci cally, speed overestimation systematically prompted driversto drive slower, whereas speed underestimation led to aster driving paces. This demonstrates thatdriving speed is strongly a ected by perceived visual speed.

In real-li e driving, the roadsides usually include various objects and landmarks as trees, buildings,tra c signs, or pedestrians. Such objects next to the road can have a cognitive infuence on drivingspeed because they constitute potential obstacles that increase the risk o collision (e.g., tree or buildingi one drives out o the road and cars or pedestrians suddenly crossing the road). However, we wereinterested in the perceptual and not in the cognitive e ects o contrast reduction on driving speed.There ore, the roadsides o our driving scenario consisted o a grass texture clear o any object andlandmark. This prevented cognitive actors as those mentioned above to bias our results, but impor-tantly, it also prevented subjects o using these landmarks to estimate moving speed. Indeed, addinglandmarks or other higher-level cues would have given the subjects the possibility to use somecognitive strategies to estimate speed. For instance, subjects could have used relative size in ormationo landmarks to in er distances and counted passing them to assess speed. Undoubtedly, this would

have biased our results. Yet, one could argue that at a pure perceptual level, objects next to the roadcan also contribute to increase peripheral visual fow, thereby altering perceived speed. This would bemostly true with a driving environment that provides only little visual fow in ormation (e.g., verysmooth road pavement and roadsides). However, the scene we used consisted o a rough asphaltpavement and roughly textured grass roadsides (see Figure 2 ). These rough textures provided partici-pants with rich visual fow in ormation when driving, as attested by our results. Speci cally, in the clearvisibility condition, participants estimated visual speed with great accuracy. In addition, the variabilityo speed estimates was small (i.e., low JNDs) in all conditions, and even more with clear visibility. Thoseelements indicate that our driving scenario provided robust visual fow in ormation allowing or reli-able speed estimates.

Contrast-dependent modulation o neural activity has been reported at early stages o visualprocessing, as the retina ( Shapley and Victor, 1978 ), the lateral geniculate nucleus ( Solomon et al.,2002 ), and the primary visual cortex ( Levitt and Lund, 1997 ; Polat et al., 1998 ; Sceniak et al., 1999 )

It has also been reported in the middle temporal area (MT or V5), an area playing an important role inmotion processing ( Pack et al., 2005 ; Bartels et al., 2008 ) and speed detection ( Maunsell and VanEssen, 1983b ), notably via eed orward projections to the medial superior temporal area ( Maunsell and van Essen, 1983a ; Ungerleider and Desimone, 1986 ). Yet, linking these neurophysiological ndingswith the e ect o contrast on perceived speed in humans is not straight orward. For instance, theobservation that MT neurons tuned to high speeds are strongly activated by slow stimuli at lowcontrast ( Pack et al., 2005 ) would predict speed overestimation at low contrast. This is the oppositeo what has been usually observed with uni orm contrast reduction, including in the current study. Asstated by Pack et al. (2005) , this inconsistency might be resolved by assuming a bias towards slowspeeds when the total MT population activity is low. Such a bias towards slow speed is precisely what
http://dx.doi.org/10.7554/eLife.00031http://dx.doi.org/10.7554/eLife.00031


9/12

Neuroscience


Research article

is proposed by Bayesian models o speed perception ( Weiss et al., 2002 ; Stocker and Simoncelli,2006 ). Speci cally, these models rely on the assumption that speed measurements are intrinsicallynoisy, and that based on our everyday experience, slower motions are more likely to occur than asteronesresulting in an elevated prior or slow speeds. Whereas Bayesian models o speed perceptioncan accurately predict speed underestimation at low contrast ( Stocker and Simoncelli, 2006 ), speedoverestimation observed at higher speeds ( Thompson et al., 2006 ) and at low luminance ( Hammett et al., 2007 ) are more di cult to account or. Alternative models have proposed that speed is encodedas the ratio o the responses o physiologically plausible temporal lters ( Hammett et al., 2000

Hammett et al., 2005 ). These ratio models account or both under- and overestimation o speedat low contrast ( Thompson et al., 2006 ). However, none o the abovementioned models addressedsituations in which the amount o contrast reduction di ered or di erent areas o the moving visualscene, as is the case when driving in og. In that respect, these models do not explain the di er-ences in speed perception observed here with uni orm and distance-dependent contrast reduction.They do not account or opposite biases evoked with the same global contrast reduction at the sameactual speed.

Poor visibility conditions a ect millions o drivers around the world. Thousands o them die eachyear in a car accident. Excessive speed constitutes a major causal actor or these car accidents. Weshow here or the rst time how og biases speed perception, and we reveal the perceptual mecha-nisms underlying this bias, providing important insights into the human visual system. In particular, weshow that contrarily to what was previously believed, speed is overestimated in og because visibilityis poorer in the central than in the peripheral area o the visual eld. We also show that the behaviouralconsequence o this speed overestimation is a natural tendency to drive at a slower pace. There ore,drivers should probably listen to their visual system when it prompts them to decelerate.

Materials and methodsSubjectsThirty-two experienced drivers (23 males and 9 emales; aged 2135 years, mean = 25.3 years)participated voluntarily in the study (12 in experiment 1, 10 in experiment 2, and 10 in experiments3 and 4). All had normal or corrected-to-normal vision. They were paid, naive as to the purpose o the research, and gave their in ormed consent be ore taking part in the experiment. The study wasper ormed in accordance with the ethical standards laid down in the 1964 Declaration o Helsinki,in line with Max Planck Society policy and in compliance with all relevant German legislation. The

participants had the option to withdraw rom the study at any time without penalty and without havingto give a reason.

Experimental setupAll experiments were per ormed in an immersive virtual environment. For all experiments, the partici-pants were seated in a simpli ed vehicle mock-up equipped with steering wheel and pedals. Thesteering wheel haptic eedback was disabled so that no speed in ormation could be in erred rom thewheel eedback. Also, the steering wheel rotation amplitude was linearly mapped to the vehicle turningspeed and applied on the centre o mass o the vehicle. This way, a small rotation o the steering wheelresulted in a slow rotation o the vehicle around its vertical axis, whereas a large steering wheel rotationled to a ast turning, independently o the actual longitudinal speed o the vehicle. The mock-up waslocated at the centre o a large semi-spherical screen equipped with a multi-projection system. Thepanoramic screen surrounded the observer to provide an image that embraces almost the entire

human visual eld. More speci cally, a cylindrical screen with a curved extension onto the foor pro-vided a projection sur ace o 230 (horizontally) 125 (vertically). The resulting sur ace was entirelycovered by our LCD projectors, with a resolution o 1400 1050 pixels each. Overlapping regionswere blended by openWARP technology (Eyevis, Reutlingen, Germany). The geometry o the scenewas adjusted or an eye height o 0.8 m at a distance o 3 m rom the vertical screen. The virtual envi-ronment was created using 3DVIA Virtools 4.1 (Dassault Systemes, Vlizy-Villacoublay, France) behav-ioural engine, which was running distributed on a 5-PCs cluster, one or each o the projectors and asupervisor. The visual stimulus consisted o a virtual environment where a textured plane reproducinga straight single-lane road was scrolled at di erent speeds. The central sur ace o the plane was arough asphalt pavement, whereas the sides were covered with grass-like textures. The sky consisted


10/12

Neuroscience


Research article

o a homogenous grey texture with the same colour o the og and the plane used or the uni ormcontrast reduction (RGB = [128, 128, 128]).

Contrast reduction Visual contrast was reduced by blending the og colour (RGB = [128, 128, 128]) into the virtual scene,according to the alpha blending model C r = C o + C f (1 ), where, or each pixel o the projectedimage, C r is the resulting colour, C o is the original colour, C f is the colour o the og, and is the blendingactor. For the distance-dependent contrast reduction, the blending actor was determined by e fd

where f is the density o the og and d is the distance rom the observer to the depicted object. Thecolour o each pixel was converted into the corresponding brightness (the arithmetic mean o the red,green, and blue colour coordinates in the RGB colour space), and the scene luminance distributionwas computed based on the empirically determined unction between luminance and brightness. Theunction was determined in two phases: (i) a plane model with uni orm colour (brightness) was dis-played on the screen acing the observer, and luminance was measured on its sur ace with a MinoltaLS-100 photometer (Konica Minolta, Tokyo, Japan) or several brightness levels; (ii) the readings o thephotometer were plotted against the corresponding brightness and tted (R 2 = 0.999) by a quadraticunction that was then used to compute the luminance distribution o the scene or each visibility con-dition. The contrast o the scene was then computed as the RMS o the luminance (in cd/m 2) o thepixels in the virtual scene (similarly to what Snowden et al., 1998 did in their work). We set the ogdensity value to 0.1 or the medium visibility condition and 0.3 or the poor condition, in a range rom0 ( ull visibility) to 1 (no visibility). These values correspond to a meteorological visibility range (MVR) o 30 and 10 m, respectively. The MVR indicates the distance at which a white object appears with a 5%contrast ( Kovalev and Eichinger, 2004 ).

The distance-independent (uni orm) contrast reduction was implemented by inserting a transparentvirtual plane in ront o the scene. The plane brightness was composited with the brightness o thebackground image, according to the standard alpha blending model described previously. The opacityo the plane was adjusted to 0.28 and 0.52 in order to match the contrast o the moderate and severeog conditions, respectively.

The global attenuation o visibility was identical or both types o contrast reduction (RMS = 0.46or clear visibility, 0.31 or moderate contrast reduction, and 0.19 or severe contrast reduction). Anti-og (experiments 3 and 4) was obtained using a vertex shader technique ( Engel, 2005 ). The blendingactor was set to 1 e afd , where af is the density o anti- og, which was adjusted to match the overallscene contrast o the og condition (RMS = 0.24).

Design and data analysisFor the two-interval orced-choice (2IFC) procedure used in experiments 1 and 3, 80 trials per condi-tion were per ormed, or a total o 1200 randomly interleaved trials in experiment 1 (two sessions o six blocks each, total duration o 2.5 hr) and 240 in experiment 3 (two blocks, total duration o 30 min).

For experiments 2 and 4, each subject per ormed ve trials per condition, and mean driving speed wascomputed or each subject and condition. Experiment 2 consisted o 25 trials, per ormed in ve consecu-tive blocks, and lasted 2 hr in total. Experiment 4 lasted 1.5 hr, consisting o three blocks o ve trials each.

For all experiments, the order o presentation o the trials was ully randomized and di erent or allsubjects. During the 5-min breaks between two consecutive blocks and the 15-min break between twosessions, the lights o the experimental room were switched on and subjects could walk and relax.

Mean PSE and JND values in experiment 1 and mean driving speed values in experiment 2 wereanalysed using a 5 3 (contrast reduction [clear visibility, moderate og, severe og, moderate uni orm

reduction, and severe uni orm reduction], target speed [40, 60, and 90 km/hr]) repeated-measures(within subjects design) analysis o variance (ANOVA). Mean PSE and JND values in experiment 3 andmean driving speed values in experiment 4 were analysed using a 3 (contrast reduction [clear visibility,og, and anti- og]) repeated-measures ANOVA. The reported values are HuynhFeldt corrected, andpost hoc tests using the Holm adjustment method or multiple comparisons (p


11/12

Neuroscience


Research article

Additional information

FundingFunder Grant reference

numberAuthor

Max Planck Society Paolo Pretto,Jean-Pierre Bresciani

World Class University programunded by the Ministry o Education, Science andTechnology throughthe National ResearchFoundation o Korea

R31-10008 Heinrich H Bltho

The unders had no role in study design, data collection and interpretation, or the decision tosubmit the work or publication.

Author contributionsPP, Conception and design, Acquisition o data, Analysis and interpretation o data, Dra ting or revisingthe article; J-PB, Conception and design, Acquisition o data, Analysis and interpretation o data,Dra ting or revising the article; GR, Analysis and interpretation o data, Dra ting or revising the article;HHB, Analysis and interpretation o data, Dra ting or revising the article.

EthicsHuman subjects: The work was conducted in line with Max Planck Society policy and in compliancewith all relevant German legislation. The study did not require speci c approval by an ethics committee.However, the ethical aspects o the study were care ully discussed with local researchers not involvedin the experiments and prior to the study. The research did not involve serious risk or participants and didnot make use o invasive techniques. The participation in the study was voluntary and participantswere recruited using the Max-Planck Subjects Database. Be ore starting the experiments, an in ormedconsent orm was signed by participants, who were also in ormed that withdrawal rom the study waspossible at any time, without any penalty, and without having to give a reason. Participants dataprotection complied with the German Federal Data Protection Act.

References Anstis S . 2003. Moving objects appear to slow down at low contrasts. Neural Netw 16 :9338.Bartels A , Zeki S, Logothetis NK. 2008. Natural vision reveals regional specialization to local motion and to

contrast-invariant, global fow in the human brain. Cereb Cortex 18 :70517.Blakemore MR , Snowden RJ. 1999. The e ect o contrast upon perceived speed: a general phenomenon?

Perception 28 :3348.Engel W . 2005. SHADERX3: Advanced Rendering with DirectX and OpenGL: Charles River Media . Hingham, MA,

USA.Hammett ST , Thompson PG, Bedingham S. 2000. The dynamics o velocity adaptation in human vision. Curr Biol

10 :R882.Hammett ST , Champion RA, Morland AB, Thompson PG. 2005. A ratio model o perceived speed in the human

visual system. Proc Biol Sci 272 :23516.Hammett ST , Champion RA, Thompson PG, Morland AB. 2007. Perceptual distortions o speed at low luminance:

evidence inconsistent with a Bayesian account o speed encoding. Vision Res 47 :5648.Hofstetter HW , Gri n J, Berman M, Everson R. 2000. Dictionary of visual science and related clinical terms

(5th ed.). Butterworth: Heinemann.Horswill MS , Plooy AM. 2008. Reducing contrast makes speeds in a video-based driving simulator harder todiscriminate as well as making them appear slower. Perception 37 :126975.

Kontsevich LL , Tyler CW. 1999. Bayesian adaptive estimation o psychometric slope and threshold. Vision Res 39 :272937.

Kovalev VA , Eichinger WE. 2004. Elastic lidar: theory, practice, and analysis methods . John Wiley & Sons.Indianapolis, IN, USA.

Levitt JB , Lund JS. 1997. Contrast dependence o contextual e ects in primate visual cortex. Nature 387 :736.Loomis JM , Blascovich JJ, Beall AC. 1999. Immersive virtual environment technology as a basic research tool in

psychology. Behav Res Methods Instrum Comput 31 :55764.Maunsell JH , van Essen DC. 1983a. The connections o the middle temporal visual area (MT) and their

relationship to a cortical hierarchy in the macaque monkey. J Neurosci 3:256386.


12/12

Neuroscience


Research article

Maunsell JH , Van Essen DC. 1983b. Functional properties o neurons in middle temporal visual area o themacaque monkey. I. Selectivity or stimulus direction, speed, and orientation. J Neurophysiol 49 :112747.

Owens DA , Wood J, Carberry T. 2010. E ects o reduced contrast on the perception and control o speed whendriving. Perception 39 :1199215.

Pack CC , Hunter JN, Born RT. 2005. Contrast dependence o suppressive infuences in cortical area MT o alertmacaque. J Neurophysiol 93 :180915.

Polat U , Mizobe K, Pettet MW, Kasamatsu T, Norcia AM. 1998. Collinear stimuli regulate visual responsesdepending on cells contrast threshold. Nature 391 :5804.

Pretto P , Ogier M, Bltho HH, Bresciani J-P. 2009. Infuence o the size o the eld o view on motion percep-tion. Computers Graphics 33 :13946.

Sceniak MP , Ringach DL, Hawken MJ, Shapley R. 1999. Contrasts e ect on spatial summation by macaque V1 neurons. Nat Neurosci 2:7339.

Shapley RM , Victor JD. 1978. The e ect o contrast on the trans er properties o cat retinal ganglion cells.J Physiol 285 :27598.

Snowden RJ , Freeman TC. 2004. The visual perception o motion. Curr Biol 14 :R82831.Snowden RJ , Stimpson N, Ruddle RA. 1998. Speed perception ogs up as visibility drops. Nature 392 :450.Solomon SG , White AJ, Martin PR. 2002. Extraclassical receptive eld properties o parvocellular, magnocellular,

and koniocellular cells in the primate lateral geniculate nucleus. J Neurosci 22 :33849.Stocker AA , Simoncelli EP. 2006. Noise characteristics and prior expectations in human visual speed perception.

Nat Neurosci 9:57885.Stone LS , Thompson P. 1992. Human speed perception is contrast dependent. Vision Res 32 :153549.Thompson P . 1982. Perceived rate o movement depends on contrast. Vision Res 22 :37780.Thompson P , Brooks K, Hammett ST. 2006. Speed can go up as well as down at low contrast: implications or

models o motion perception. Vision Res 46 :7826.Ungerleider LG , Desimone R. 1986. Cortical connections o visual area MT in the macaque. J Comp Neurol

248 :190222. Weiss Y , Simoncelli EP, Adelson EH. 2002. Motion illusions as optimal percepts. Nat Neurosci 5:598604.

Documents

Foggy perception slows us down