Kits, colours, and confusion: A pilot study of vision and football

Vision and football

Association football (soccer) is a major global business involving huge sums of money,

much of it from television. But watching football is a complex, dynamic visual task:

who is moving where?öwho is offside?öwhere are the defenders?öhow many attackers

in the box? Hence, if there are visual factors that enhance or hinder perception by

viewers and by players, then these factors could be important to the success of the

sport and the business of football. Our repeated armchair experience has been that

some combinations of kit colours promote good `reading' of the game, perhaps by

enabling rapid visual search and visual segregation of one team from the other.

Other kit combinations seem to make matches confusing and almost unwatchable

by preventing easy search and grouping of players. Figure 1a suggests that, if the two

teams wear different but uniform colours (eg all-red vs all-white; Liverpool vs Real

Madrid), then the segregation and `readability' of the pattern of play is very good;

while figure 1b suggests that, if the two teams wear similar colours, albeit in opposite

combinations (eg red ^white vs white ^ red; Manchester United vs Southampton), then

the readability may be very poor. [See supplementary figure S1 for a real example

from the 2003 ^ 2004 season.]

The literature on visual search (mainly for single targets) tells us that simple colour

differences `pop-out' readily (Bauer et al 1996; Duncan 1989; D'Zmura 1991), while

conjunctions of colour and form typically do not (Treisman and Gelade 1980). One

general theory proposes that search is slow and difficult when targets are similar to

distractor items, or when distractors are dissimilar from each other, or both (Duncan

and Humphreys 1989). Thus, we suspect that some football matches (perhaps inadver-

tently) take advantage of `pop-out' and grouping by colour, while others create confusion.

Surprisingly, the Laws of the Game (http://www.fifa.com/refs/laws E.html) (Law 4) make

no prescription at all about kit colours, but different competitions and leagues have

a variety of rules about what shall be done when kit colours clash. The English FA

Last but not least

Perception, 2005, volume 34, pages 633 ^ 637

DOI:10.1068/p5302

(a) (b)

Figure 1. (a) Uniform colours. Liverpool vs Real Madrid (all-red vs all-white, on a green background).Verdict: Excellent. See online (http://www.perceptionweb.com/misc/p5302/) for colour versions of

all figures. (b) Crossed colours. Manchester United vs Southampton (red top and white shorts

vs white top and red shorts). This combination has been seen for at least the last three seasons.

Many others like this can be seen.Verdict: Very poor.

Cup rules, for example, state that ``Where the colours (shirts, shorts or stockings) of

the two competing Clubs are similar, both Clubs must change...'', but similarity appears

to be undefined. Many of the other kit rules are concerned with commercial rather

than psychophysical impact. In practice, referees may check that the shirt colours are

different, and that the shorts are different, and then suppose that all will be well.

From real matches and from schematic examples like figure 1b we doubt that this is a

sound strategy, but we need experimental evidence, rather than anecdote. We used

response time as a measure of difficulty in a visual-search and counting task, and removed

the possible influence of many extraneous variations such as camera angles, lighting

conditions, crowd effects and so on, by using schematic, still pictures.

Experiment: Count the players

The experimental task was based on a combination of visual search and counting:

simplified static images with coloured, shaded rectangles were used to represent the players

against a textured green background. On each trial, eight `players' were shown in random

locations without overlap (figure 1). The task was to indicate by a key-press whether

3, 4, or 5 of `your team' (indicated before each block of trials) were present. Selectively

gathering this kind of spatially distributed information is likely to be vital in `reading'

the match. Response time (RT) and accuracy were recorded. Instructions emphasised

both speed and accuracy. A `referee' and `ball' were present, but were to be ignored.

The key experimental factor was whether the teams wore different, uniform colours(eg red ^ red vs white ^ white) or crossed colours (eg red ^white vs white ^ red).

Images (figure 1) were generated as 8-bit TIFFs on a Macintosh computer inMatlab 5.2,

displayed on a 19-inch RGB monitor with PsychToolbox software (Brainard 1997). The

image was on until the subject responded with a key-press. The green `pitch' subtended

9.5 deg66.4 deg (about the same as a 25-inch TV set viewed from 10 feet) and was set

in a larger (full screen) green background. Player size was 30616 pixels (28.4 min ofarc

615.1 min of arc), with shirt length 19 pixels and shorts length 11 pixels. Chromaticity

and luminance of the test colours, measured with a Minolta CS-100A Chroma meter,

are given in table 1. Observers were allowed free, binocular viewing from a distance of

125 cm, with no fixation point.

The experiment had a 4-factor (3626362) repeated-measures design:

. number of target players [3, 4, or 5];

. type of colour combination [uniform vs crossed] (eg figure 1a vs figure 1b);

. colour pair [(blue, red), (blue, white), or (red, white)];

. target team [1 vs 2] (left or right on the instruction screen).

Different sets of images with different randomisations were used for each observer.

Each image was used just twice: once for target team A, and once for team B.

Number of targets varied within a block of trials, while other factors varied in random

order between blocks. Data for the ANOVA were median RTs from 8 repetitions of

each condition within a block. Subjects were five male, and five female volunteers

Table 1. Luminance and chromaticity of the kits, pitch, and surrounding screen.

Luminance=cd mÿ2 CIE chromaticity coordinates

x y

Red 19.3 0.521 0.336

White 96.4 0.286 0.325

Blue 14.4 0.163 0.121

Green pitch 52.9 0.284 0.477

Green surround 42.6 0.300 0.471

634 Last but not least

(age 20 ^ 35 years; median age 21 years). Short practice sessions were given before the

main blocks of trials.

Results

Crossed colours are bad. Figure 2a shows that when teams wore crossed colours

(eg red ^white vs white ^ red), it was much more difficult to count the players than

when kits were of uniform colour (eg red ^ red vs white ^ white). Average response time

to count the players in the target team was 750 ms slower with crossed colours than with

uniform colours (F1 9 . 49:4, p 5 0:0001). All observers showed the effect consistently.

Across individuals, the smallest difference was 340 ms and the largest was 1370 ms

(mean 752 ms; median 690 ms).

Not surprisingly, counting 3 targets was significantly quicker than counting 4 or 5,

by about 140 ms (F2 18 . 18:4, p 5 0:0001), irrespective of colour combination. There wasno significant interaction between these two factors (F2 18 . 2:00, p 4 0:1).

Figure 2b shows that the crossed-colour disadvantage was found for all 3 pairs of

colours tested, and that it was worst for the blue ^ red pairing where the crossed-colour

penalty rose to almost 1000 ms. This interaction between colour pair and type of combi-

nation was highly significant (F2 18 . 26:5, p 5 0:0001).

Unlike the RT measure, accuracy (% correct trials) was not a useful measure of

difficulty, because error rates were not high. Most effects in the ANOVA were not signif-

icant, although there was lower accuracy for crossed colours with 4 or 5 targets presentö

a significant interaction between number of targets and type of colour combination

(F2 18 . 8:01, p 5 0:005).

Discussion

Penalty!

We have found that it is much more difficult to `read' the spatial layout of an image

resembling a football match when the two teams wear crossed colours (the same

pair of colours but with opposite ordering of shirt and shorts) than when they wear

completely different, uniform colours. The response-time penalty averaged 0.75 s, and

was even greater (1.0 s) for the red ^ blue pairing. These are huge delays, implying great

difficulty in the crossed-colour condition.

,

,

,

,

,

(a) (b)

Response time=ms

2000

1800

1600

1000

1200

1000

800

600

400

3 4 5 blue ± red blue ± white red ± white

Number of target players Pairs of colours tested

uniform colour crossed colour

Figure 2. (a) Mean response times (1 SE from the ANOVA plotted against number of target

players, for the two kinds of colour combination (pooled over the three pairs of colours). (b) Mean

response times (1 SE) plotted for each pair of colours.

Last but not least 635

Such effects are consistent with the literature on visual search for single targets,

and with the factors that enable or impede texture segregation. The crossed-colour

condition probably prevents target `pop-out', and so impedes the localisation and

counting of target items. In a recent overview, Pashler (1999, chapter 3) concluded that

``targets defined by spatial arrangements of parts can yield strikingly difficult conjunc-

tion searches''. Heathcote and Mewhort (1993), for example, studied single-target search

in a situation analogous to ours. Target items were squares divided into black/white

or red/blue halves, while the seven distractor items were white/black or blue/red,

respectively, in different groups of observers. Response times to detect the target were

very slow, and averaged about 1 s (though they gradually improved with intensive

practice on the same condition). In our crossed-colour condition, the two teams also

differed only in the spatial ordering of their colours, and so the observer was presum-

ably forced to use an item-by-item search which can be painfully slow. The observer

probably also gets a much weaker sense of the pattern and spatial arrangement of the

players. Thus, with crossed colours, viewers (and players?) are effectively blind to

much of the information needed to make sense of the game.

We chose to compare the uniform-colour and crossed-colour conditions because

they may represent the best and worst cases, but other common combinations of kit

colours (eg blue ^white vs white ^ black; Everton vs Fulham) are likely to present anal-

ogous difficulties where the colour pairings defeat the processes underlying grouping

and pop-out.

Lack of realism might be an issue in our displays, and it would be interesting to

use real video clips instead of schematic stills. But there the introduction of cluttered

backgrounds, real human forms, overlapping and rapidly changing positions might

only make the task even more difficult. Supplementary figures S1 and S2 (see http://www.perceptionweb.com/misc/p5302/) are video stills from real matches, and the crossed-

colour case (figure S1) seems to us informally to be as difficult as the schematic images

used in the experiment. See also figure S3 on the same website.

Real football

We strongly suspect that the difficulties and delays seen in our experiment can also

exist when viewers watch real matches in which the choice of kits is poor. Watching

football involves much more than tracking individual players on the ball. One needs to

see the whole pattern of attack and defence, how these patterns are changing, where

the opportunities and threats lie, whether an attacker is offside, and so on. This kind

of global pattern recognition is needed rapidly, with rapid updating, if the viewer,

official, or player is to make sense of the match and make vital, quick decisions. Although

our experiment relates most directly to TV viewers, it seems likely that players and

officials might also be confused or delayed by impoverished information, and, if so,

the impact on the game itself could be substantial. Poor choice of kit colour combi-

nations, even if less severe than the crossed-colour case, is likely to be detrimental.

Games often turn on a few critical moments where players make the right or wrong

decision, and in top-level games players have to decide and act very quickly with no

time for scrutiny, so even a half-second delay could be disastrous.

To our surprise, we see many poor examples in real matches every season. This

suggests that the authorities and broadcasters do not have adequate procedures for

regulating the combination of colours in matches. As far as we can discover, it is the

referee who makes a decision about whether two kits clash. For example, Fifa World

Cup rules [http://fifaworldcup.yahoo.com/en/] state: ``If, in the opinion of the referee or

the match commissioner, the colours worn by two opposing teams might cause con-

fusion or prove unsuitable for television broadcasts, they shall be modified, using either

the entire reserve outfit or a combination of both''. The basis for the referee's opinion

636 Last but not least

is unclear, but it seems likely that this decision will be based mainly on the colour

and pattern of the shirts, viewed from close-up. In our view, it would be safer to

ensure that the two kits differ more fundamentally, in ways that are visible in periph-

eral vision, and at long distance, and are rapidly available to early visual processes

without slow, detailed scrutiny.Where the two teams wear a single colour from head to

toe (eg red vs blue; Liverpool vs Chelsea) all these requirements are met. Difficulties

arise when the teams have one or (as here) two colours in common. When teams

wear striped or patterned shirts, the spatial-frequency dependence of pattern and colour

vision will also have to be considered. For example, the black and white stripes of

Newcastle and the desaturated light blue of Manchester City may look very different

close to, but at a distance, in motion or in peripheral vision, the two may be quite

similar as Newcastle's stripes blur to a mid-grey. In short, then, better guidance could

be formulated by applying visual science to football.

Mark Georgeson, James Lampard

Neurosciences Research Institute, School of Life & Health Sciences, Aston University, Birmingham

B4 7ET, UK;

e-mail: m.a.georgeson@aston.ac.uk

Janet Georgeson

Centre for Socio-cultural and Activity Theory, School of Education, University of Birmingham,

Birmingham B15 2TT, UK

References

Bauer B, Jolicoeur P, Cowan W B, 1996 ``Visual search for colour targets that are or are not

linearly separable from distractors'' Vision Research 36 1439 ^ 1465 Brainard D H, 1997 ``The psychophysics toolbox'' Spatial Vision 10 433 ^ 436

Duncan J, 1989 ``Boundary conditions on parallel processing in human vision'' Perception 18

457 ^ 469

Duncan J, Humphreys G W, 1989 ``Visual search and stimulus similarity'' Psychological Review

96 433 ^ 458

D'Zmura M, 1991 ``Color in visual search'' Vision Research 31 951 ^ 966Heathcote A, Mewhort D J K, 1993 ``Representation and selection of relative position''

Journal

of Experimental Psychology: Human Perception and Performance 19 488 ^ 516Pashler H E, 1999

The Psychology of Attention (Cambridge, MA: MIT Press)

Treisman A M, Gelade G, 1980 ``A feature integration theory of attention'' Cognitive Psychology

12 97 ^ 136

ß 2005 a Pion publication

Last but not least 637

ISSN 0301-0066 (print)

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.

www.perceptionweb.com

ISSN 1468-4233 (electronic)