Perform ance and M ood Effects by
Jenniier A n n V eilch B.Sc., U niversity o f M anitoba, 1984 B.A., U niversi'y o f M anitoba, 1986
M A ., Q u een ’s U niversity, 1987
A f ’ P r p T P H A D issertation Subm itted in Partial Fulfillm ent o f the R equirem ents for the D eg ree o f
F A C U L T Y OF O/TADUATL STUDIES
C O C T O R O F P H IL O SO P H Y ...— in the D epartm ent o f Psychology
Ut AN
_________ W e accept this dissertation as conform ing to the required standard
D r. R. G ilford, Supervisor (D ep artm en t o f Psychology)
Dr. C. Porac, D epartm ental M em ber (D epartm ent o f Psychology)
f t S t . L. R o se n b lo jx ff D epartm ental M em ber (D ep artm en t o f Psychology)
MATE.
" Dr.
G. H ogya, O utside M em ber (D epartm ent o f Theatre)Dr.' G . E yans, E xternal Exam iner (Program in S ocial E cology, U niversity o f California, Irvine)
© JE N N IF E R A N N V E IT C H , 1991 U niversity o f V ictoria
A ll rights reserved. D issertation m ay not b e reproduced in w h ole or in part, by photocopying or other means,
Supervisor: D r. Robert G ifford
A B S T R A C T
Lighting research has produced a wealth o f k now ledge con cern in g the visual effects o f w orkplace lighting, but little understanding o f other behavioural con seq u en ces. O n e trend in current lighting practice is toward providing users w ith the opportunity to control their own w orkstation lighting, often through th e u se o f supplem entary task lighting. T he general assum ption is that personal control over lighting will lead to b etter perform ance and im proved m ood. T h e personal control literature in psychology is abundant and tends to support this b elief. E nvironm ental psychologists in particular have em braced th e id ea that th e provision o f ch oices in the physical environm ent w ill lead to desirable ou tcom es, su ch as feelin gs o f self-efficacy or com p eten ce.
T h e present study tested the n otion that personal con trol over lighting has beneficial effects, using a m odified 2 x 2 C ontrol X F reference factorial d esign w ith an additional control group to test for the possibility o f subject reactivity biases. T h e P referen ce variable w as included to test th e hypothesis that w orking under favoured conditions, regardless o f er.e’s ability to control them , beneficially affects perform ance and m ood. T h e d esign incorporated m easures o f m otivation and attention to attem pt to distinguish betw een tw o co m p etin g m echanism s that m ight underlie th e effects. M ale and fem ale undergraduate students w ere random ly assigned to o n e o f the five conditions in this laboratory experim ent for a 2-hr se ssio n during w hich they co m p leted a m ood questionnaire and intellectual tasks including br~'o-teaser puzzles, a creativity task, an arithm etic task, and a gramm ar w orksheet.
T h e m anipulation o f C ontrol as w ell as the m anipulation o f P reference (for the lighting at w hich on e w orked) affected ratings o f perceived control. Subjects in th e C ontrol condition reported higher feelings o f control than those in the N o C on trol C ondition; similarly, P referen ce G iven subjects rated their perceptions o f control m ore highly than P reference D e n ie d subjects.
C ontrary to ihe conventional w isdom am ong environm ental psychologists and designers, the results show ed that choices in th e physical environm ent are n o t always beneficial, at least w here lighting is concerned. C ontrol subjects perform ed m ore poorly on the intellectual
tasks and m ore slow ly o n the creativity task than N o C ontrol subjects. T h e ou tcom es are discussed as differential effe' is o f d ecision al versus cognitive control. D esig n applications o f the personal control construct await further research.
Exam iners:
D r. R . G ifford, Supervisor /(D ep artm en t o f Psychology)
^T5r. C. P orac, D epartm ental M em b er (D ep a rtm en t o f Psychology)
^ / i D r . L. Rosenbjj& cl, D ep artm en tal M em b er (D ep a rtm en t o f Psychology)
D r. G . H ogya, O utside M em ber (D ep a rtm en t o f T h eatre)
A B S T R A C T ...ii T A B L E O F C O N TE N T S ... iv LIST O F T A B L ES ... vii LIST O F F I G U R E S ... viii A C K N O W L E D G E M E N T S ...ix I n lr o d u c t io n ... 1 L ig h tin g ... 2 R eview o f R esearch ... 2 Lighting the O f f i c e ... 19 Personal C ontrol ... 22 T ypes o f C ontrol ... 22
C on seq u en ces o f P erson al C ontrol ... 27
T h e R esearch Q uestion ... 33 Explanatory M od els ... 34 R esearch D esign ... 36 H ypotheses ... 38 M ethod ... 44 S u b j e c t s ... 44
A pparatus and V ariables ... 44
S e t t i n g ... 44
Independent V ariables ... 47
M aterials and M easu res ... 50
P rocedure ... 52
R e s u l t s ... 54
A nalysis o f Variance ... 54
M anipulation C hecks ... 54
E xam in ation o f R e s i d u a l s ... 77
A N O V A s E xcluding H ypothesis- G u essers ... 77
A n a ly ses o f Covariance: D esired I n f lu e n c e ... 77
Pow er A n alysis ... 82
P ost h o c P ow er A n a ly s i s ... 82
A priori P ow er A nalysis ... 85
Path A n a l y s i s ... 87
Subjects’ L ighting P referen ces ... 88
Lighting T y p e A n a l y s i s ... 88
E xam ination o f E xclu d ed D a t a ... 70
D iscu ssion ... 75
T h eoretical I m p lic a t io n s ... 75
D esig n Im p lication s ... 78
D irectio n s for F uture R e s e a r c h ... 77
] d f e r e n c e s ... 102 A ppendix A ... 108 A ppendix B ... I l l A ppendix C ... 112 A ppendix D ...114 A ppendix E ... 115 A ppendix F ...110 A ppendix G ... 120 A ppendix H ...124 A ppendix I ...127 A ppendix J ... 128 A ppendix K ...152 A ppendix L ... 134
A ppendix N A ppendix O
LIST O F T A B ! ES
T able 1. Illum inance R ecom m endations c f N orth A m erica and Circa' Britain for Sim ilar A p p lication s (Ix) ... ... 14 T able 2. C om parison o f H yp oth eses under A ttention and M otivation M odels . . . . ... 42 T ab le 3 . C ontingency table o f hypothesis-guessing by c o n d i t i o n ... T ab ic 4 . M eans and standard deviations for m anipulation check item s by condition . . . 57 T able 5 . M arginal m eans and standard deviations for Preference effect on control ratings . 00 T able 6 . M arginal m eans and standard deviations for C ontrol effect on control ratings . . . . 01 T able 7 . M eans and standard deviations for task difficulty ratings and self-reported attention by condition ... ... <>3 T able 8 . U nivariate statistics for dependent variables ... 05 T ab le 9 . '■ om p on en t loadings for principal com ponents analysis o f perform ance m easures 71 T able 10. C ell m eans and standard deviations for dependent v a r ia b le s ... 75 T able 11. M arginal m eans and standard deviations for C ontrol effects ... 78 T ab le 12. C ell statistics for C on trol X Preference effect on incidental learning (hypothesis-gu essers excluded) ... HI T ab le 13. Sum m ary table for post hoc power analysis ... 83 T ab le 14. Sum m ary table for a priori power analysis ... 80 T ab le 15. Frequencies o f ranks o f lighting choices ... 8b T ab le 16. Lighting type freq u en cies o f naive-group s u b j e c t s ... 91 T ab le 17. Frequencies o f ranks o f lighting choices for excluded subjects ... 93 T ab le T \ D ep en d en t variable m eans for excluded and C ontrol-Preference D en ied subjects 94
LIST O F F IG U R E S
Figure 1. Sim ple and com p lex Landoll rings ... 5
Fiuure 2. Experim ental design ... 37
Figure 3 . A ttention m o d el o f the personal control -p erform an ce/m ood relationshi-p ... 39
Figure 4 . M otivation m odel o f the personal control -perl'orm ancc/m ood relationship ... 40
Fiuure 5 . H ypothesized C ontrol X Preference interaction, assum ing that a m otivational m echanism operates ... 43
Fiuure (>. Floor plan o f room . . . ... 45
Fiuure 7 . Lighting design in experim ental room ... 46
Fiuure S . Photographs ot experim ental setting ... 48
Fige _9. Histogram for perceptual reasoning sco res ... 68
Fiuure 10. H istogram for aggregate m easure o f intellectual perform ance ... 72
Fiuure 11. Scatterplot o f perform ance analysis residuals by sessio n ... 80
S o m any wonderful p eop le, so little space. Over the past few years I have b een privileged to w ork w ith and to learn from p eo p le o f integrity, and to all those p eop le I owe a debt 1 can never repay.
T o B ob Gifford, I am gratefui, first and forem ost, for providing an environm ent in w hich 1 could set my own direction through w hat seem ed at tim es an im penetrable m a/e. L om e R osen b lood challenged m e to excel, and kept m e honest. Clare Porac helped me to keep an accurate perception o f reality, and sh e has b een an admirable role m odel. G iles H ogya was invaluable in bringing a designer’s view point to my studies. I am thankful also to M ike H unter for his m any contributions to m y understanding o f research m ethodology and to this dissertation in particular. Stuart Kaye o f the U niversity o f M anitoba and D an Slokols o f the U niversity o f C alifornia, Irvine, have each shaped my understanding o f the field in important and w elcom e ways.
I fe e l privileged also to have had the opportunity to serve graduate students and, in the process, to work with so m any com m itted and sincere graduate students, faculty m em bers, end adm inistrators at U V ic. C ollectively, they helped m e to learn that a large institution need not be im personal nor inflexible. I am especially p leased to have been able to participate in this c o n nunity, and to see so m e o f our w ork result in visible changes. T he professionalism I observed has inspired m e to form n ew career aspirations; in this regard, I am especially indebted to D r. A n d y Farquharson, for his enthusiasm and encouragem ent o f my efforts as a beginning teacher.
Famil; and friends, o f course, provided the support netw ork to sustain me throughout m y graduate studies. M om and D ad , thank you for the phone calls, the pickles, and the unfailing love that m ade the gloom iest tim es endurable. Ronni, after years o f writing, w ords fail me; thanks, friend, for sharing w ith m e. T hanks also to G alia A rtzy, K en Bornn, Chip C litheroe, D e n ise C orrea, Steve E so, Lia F atels, O d ette G ould, D o n H Ine, Brenda Lowick, Julie
M acdonald, M ary Ann M ountain, Stacy N agel, D avid N ickoli, Shawna Pachal, M aithilec Pathak, Susan Polan, B eth Quinn, N in a R edding, T ed Scharf, D eb T h om pson, D ou g T olson, and E lain e
Vaughan for acts o f friendship to o n im erous to nam e. I w ould also like to acknow ledge tw o role m odels, D rs. Jennifer and T o m Shay, for inspiring m e 'o attem pt graduate studies w ith their enthusiastic com m itm ent to intellectual developm ent as a w ay o f life.
I am grateful also to th e N a tir a l Sciences and Engineering R esearch C ouncil o f C anada, the C anadian Scholarship Trust Foundation, the U niversity o f V ictoria, and the U niversity o f C alifornia, Irvine, for their g en ero u s finar. :ial support over the past four years. T h e Illum inating E ngineering Society R ichard K elly Grant and the A m erican Psychological A ssociation
D issertation R esearch Award assisted with the costs o f this rc..~arch project, which also d ep en d ed on the cooperation o f over 3()0 U niversity o f Victoria students as participants in th e various phases o f this study. A lon g th e way, the assistance o f L ief Bluck, R ichard Chadwick, G ab rielle Carev, Catherine Corey, L ance Grant, M orag M acneil, and M arianne M cD onald has b een invaluable.
T h e faith everyone has show n in m e is overw helm ing. P rom ise you’ll tell m e if m y future work d o esn ’t justify it?
Introduction
From its infancy, environm ental psychology has straddled the boundary betw een basic and applied research. E nvironm ertal psychologists attem pt tc understand basic psychological processes f om a particular perspective (cf. R ussell & W ard, 1982) w hile concurrently working to build a body o f k now ledge that can im prove the design o f environm ents for p eo p le (cf. Kaye, 1975). A lthough there is general acceptance that th e field has "come o f age" (Stokols, 1982; Stokols & A ltm an, 1987), there remain substantial gaps in our know ledge and few powerful theoretical fram eworks. O n e top ic that is poorly d evelop ed in this respect is the study o f the behavioural effects o f lighting (B oyce, 1987).
T h e scientific study o f lighting began with th e understanding o f the role ot lighting in m aking tasks m ore visible. E arly studies focused o n the lim its o f visual p erform ance (e.g., Blackw ell, 1946), and this work led to the establishm ent o f illum inance recom m endations and to im provem ents in industrial hygiene (B oyce, 1981). V isual perform ance, how ever, is only one possible process that lighting m ight affect: N onvisual effects o f lighting — that is, effects ihat occur d esp ite the adequacy o f the lighting given th e contrast and size o f the task elem ents and the age o f the view er — have received negligible research attention despite rep eated calls for further investigation (e.g ., B oyce, 1981; Sundstrom , 1986).
C urrent lighting design practice em phasizes installations that perm it individual adjustment o f w orkstation illum ination. D esigners (e.g., W otton, 1989) believe that providing personal control in this m anner w ill im prove workers’ perform ance and m ood. T his experim ent, therefore, exam ined the effects o f personal control over w orkplace lighting on in tellectu al task perform ance and m ood . T h e research literature generally supported the hypothesis that personal control over lighting is beneficial, and suggested two psychological m echanism s (attention and motivation) lor its action. T h e design o f the p resent study was an attem pt to test these m echanism s. Improving our understanding o f this type o f relationship can have im plications for design practitioners as w ell as for theorists. Thus, this investigation was representative o f the dual traditions o f environm ental psychology.
em pirical and theoretical foundations, which are review ed in the next few section s. W e begm with the lighting literature.
Lighting R eview o f R esearch
A s Boyce (19811 has pointed out, there have b e e n two approaches to lighting research: the practical, or field study, approach, and the laboratory approach. The latter has had by far th e greater effect on lighting practice because o f its greater role in th e establishm ent o f lighting standards and codes. Field studies generally lack controls to allow confident conclusions to b e m ade concerning th e causal relations b etw een illum ination and task perform ance.
T h e classic case that dem onstrates th e difficulty in conducting field studies involving illum ination is the H aw thorne studies (R oethlisberger & D ickson, 1939). T h e se w ell-know n experim ents conducted at an electrical assem bly plant system atically varied illum ination levels f illum inance'). In so m e experim ents, the light level w as increased, w hile in others, it was decreased. Paradoxically, productivity increased regardless o f th e direction o f the lighting increm ent. Even in the control group, which experienced no change in lighting, production im proved steadily. In further informal investigations productivity increased even w h en existing lam ps w ere replaced with identical lam ps instead o f brighter on es.
Gifford (1987) suggested that the H aw thorne studies m ight have d elayed the developm ent o f environm ental psychology by thirty years. C ertainly one conclusion draw n from th ese studies was that the physical environm ent has no direct effect o n human behaviour. This naive
determ inistic assum ption, how ever, ignored the apparent effect that em p lo y ee perceptions o f the lighting had on their work output. The H aw thorne studies w ere confounded, in any case, by the special characteristics o f the test room s that differentiated the experim ental subjects from the norm al work environm ent so that not only the lighting varied (S teele, 1973). In addition, the room layout perm itted m ore social interaction than th e regular workroom ; th ese special conditions might also have im proved productivity.
Field studies are o f lim ited use in developing lighting standards b eca u se each on e u ses a unique task having unspecified visual dem ands. It w ou ld be im practical to expect a field study to
identify th e optim al range o f lighting conditions for every pot sible occupation, and yet th e visual dem ands o f th e task are critical to the quantity and quality o f light that are required for its perform ance. T h e identification o f the m ost im portant o f these task d im ensions and their relationship *o illum ination has b een the approach taken by the laboratory studies in w hat Boyce (1981) called the analytic m odel.
T h e analytic m o d el. T his m od el o f lighting research uses sim ulations o f real w orld tasks under strict control in a laboratory setting. Various aspects o f th e task and the lighting can be m anipulated; ultim ately, researchers in this tradition h o p e to produce results that can b e extended to real w orkplaces.
It is im portant to recogn ize that the focus o f th e early lighting researchers w as the op tim ization o f visual perform ance. Their investigations concerned the characteristics o f lighting and tasks that m ade details easy to see and that w ould enable prolonged w ork sessio n s w ithout excessive fatigue. N onvisual asp ects o f the environm ent-behaviour relationship, such as m otivation, subjective im p ression s o f the illum ination, and lighting preferences, although not ignored, w ere clearly o f secondary interest (e.g., S im on son & Brozek, 1948).
S o m e o f the m ost su ccessfu l research in the analytic m o d el is W eston ’s (1962) series o f investigations o f task size (i.e., th e size o f the visual d etail) and contrast over a broad range o f illum inances. W esto n used a sim p le task that is primarily visual and that allow s easy
m anipulation o f size, contrast, and colour. This is th e L andoli fir g chart, a page o f closely sp aced circular rings with gaps oriented in one of th e eight cardinal directions. Subjects are asked to m ark all o f the rings o n the chart having gap s oriented in a sp ecified direction.
P erform an ce scores for this task consist o f the tim e taken per correctly can celled ring, corrected for th e m anual tim e n eeded to cancel a ring.
W esto n used gaps o f a b ro a d range o f sizes (visual angles o f 1.5 to 4.5 m in arc) and charts having contrasts [th e ratio o f background to task brightness (’lum inance’)] o f 0.28 to 0.97. T he illu m in an ces (the density o f light energy incident on th e task) ranged from alm ost 0 lux (lx) to nearly 10,000 lx (typical offices range from 500 - 1000 be). H is results sh ow that for any task, increasing the illum inance ca u ses an increase in perform ance that follow s a law o f dim inishing
rclurns. For each contrast and siz e com bination, eventually th ere is n o further perform ance increase associated with increases in illuminance.
A dditional con clu.ions c&-t b e drawn from W eston’s data. T h e p recise illum inance at which perform ance levels o ff dep en d s on the task size and contrast, and is low er for larger gap sizes and darker (i.e., larger) contrasts. Over illum inance ranges o f practical interest, o n e can im prove perform ance more by increasing task size or by increasing th e contrast than o n e can by increasing th e illum inance. Furtherm ore, a visually difficult task (i.e., sm all gap size w ith poor contrast) w ill always show poorer perform ance than a visually easy task, regardless o f the illum inance.
A n o th er type o f Landolt ring has been used to study th e effects o f task com plexity on visual perform ance. T h ese are show n in Figure 1. The visual dem ands o f this task can b e m ad e identical to the sim ple Landol' ring, but the com plex rings a re m o re difficult to search b eca u se gaps in the line must be distinguished from invaginations in th e circles. B o y ce (1974) com p ared the tw o typ es of Landolt rings at a number o f illum inances. H e found that th e basic form o f th e relationship b etw een illum inance and visual perform ance d ep en d s o n visual difficulty (con trast and size), but is m odified by task com plexity as a multiplying factor at each level o f illum ination. A t all lev els o f illum ination, perform ance w as slower for co m p lex rings than for sim ple rings.
Background lum inance, as w ell as task contrast and size, w as th e focu s o f an investigation by B oynton and B oss (1971) into visual search perform ance. T h e task required the u se o f b oth foveal ar.d peripheral vision, w hereas the typical Landolt ring chart u ses foveal vision a lo n e and therefore sim ulates different visual processes. A t m axim um contrast (background:task ratio =
l.(X)), perform ance varied little over a broad range o f background b righ tn esses (360 - 1370 c d /m J); th ere was no drop in perform ance until contrasts w ere reduced b elo w 0.10.
R elation sh ip s betw een lum inance and contrast at 50% target d etection threshold differed for sm all and large targets. Som e conditions o f lunr nance and contrast give easy perform ance, so that large changes in one dim ension (lum inance, contrast, and task difficulty) have little effect on target d etection , the authors concluded; com bining changes in various d im ensions, how ever, can m ake target detection im possible.
Sim ple an d com p lex Landolt rings
O O
o o
S ize and contrast have b e e n th e m ost extensively studied variables in th e illum inance-w ork relationship, but B o y ce (1981) p oin ted out that other factors o f potential im portance aw ait further investigation. A m o n g th ese
are
task m ovem ent, off-axis work, blurred target ed g es, and colou r contrast at low background lum inance.O n e non-task variable that is know n to have an effect is th e age o f th e view er. B oyce (1973) found that old er subjects pertorm cd consistently m ore poorly on th e L andolt ring chart than young subjects, at leasi at lo w illum inances (220 lx), although th e p erform an ce o f b o th groups w as the sa m e at high illum inance (1600 lx). R esu lts such as th ese sh ou ld lea d to a cautious approach to illum ination standards based o n perform ance data from a lim ited population.
T h e dep en d en t m easures con sid ered im portant by analytic m o d el research ers are n o t only perform ance data. Physiological in dices o f visual fatigue have lo n g b e e n sought, although n o single m easu re has proven to b e a reliable or a valid m easure o f the sta te o f th e visual system . A s B oyce (1981) said, "at best, th e literature on ocular fatigue can b e said to show' that so m e lighting con d ition s can produce ocular fatigue for so m e tasks som etim es" (p . 218).
T w o prom inent researchers in the N orth A m erican lighting d eb a tes o f th e m id-century were E rn est S im on son and J o s e f B rozek. T h ey questioned the light lev els reco m m en d ed by Luckiesh and M oss (1938), w hich w ere considerably higher than other recom m en d ation s. T inker (1939), for exam ple, found 31 be w as adequate for reading, but reco m m en d ed 100-150 be b e provided to give a "margin o f safety". L uckiesh and M oss recom m en d ed 20 0 - 500 be for
"ordinary reading" (p. 350). S im o n so n and B rozek (1948) sought to clarify this d ifferen ce using a sim ulated task perform ed over an extended period; th ey used a variety o f o u tco m e m easu res that included perform an ce, fatigue, and subjective im pressions.
T h e task resem b led a con veyor inspection job: Subjects cop ied letters appearing in a narrow slit. T he le tters w ere o n a band o f paper driven at a constant rate by a m otor. T h e work period lasted 2 hr; ea ch o f the six subjects participated b all con d ition s o f illu m b a n c e and lam p type, but w ere not inform ed o f th e purpose o f the study. B efore-and-afrer m easu res w ere m ade o f a variety o f visual functions including critical flicker fusion (C F F , the freq u en cy at w hich a
flickering light is seen as a continuous light) and a co m p lete ophthalm ic battery o f tests. D uring th e work period, perform ance w as evaluated at three tim es (after 5, 60, and 110 min o f w ork) and blink rate was also m easured three tim es. Subjective a ssessm en ts o f the various illum ination conditions w ere obtained separately, after th e w ork data had b een collected.
T h e researchers w ere careful to u se statistical procedures that they b elieved were appropriate to their data and to base their conclusions on significance test results instead o f graphical trends. T hey erred on the conservative sid e, how ever, b y analyzing their w ithin-subjects data as though it w ere betw een-subjects data. R e-an alysis is im p ossib le b ecau se they did not publish standard deviations. It is possib le, but unknow n, that statistically significant effects w ere overlooked.
T he results in S im on son and B rozek (1948) sh o w a p e n o r m a n e e optim um above 5 0 0 lx but b elo w 3000 be; it might b e in th e vicinity o f 1C90 be (they tested 20, 50, 150, 1000, and 3000 lx). P erform ance and subjective w ell-b ein g w ere optim al in th e 500-3000 range, and certain fatigue m easures (blink rate and recognition tim e for objects o f th reshold size) sh ow ed significant drops in this range. R egardless o f lam p type, 1000 be w as the preferred illum inance. Interestingly, Sim onson and B rozek found no effect o f illum inance on visual fatigue m easured with the standard ophthalm ological tests; fatigue w as d em onstrated after th e work sessio n by the C F F (low er values indicate greater fatigue), m axim um rate o f ey e m ovem ents, and subjective com plaints. T he authors noted that the optim um level they ob tain ed would equal the m inim um recom m en d ed level o f the then-current standards o f th e Illum inating E ngineering Society o f N orth A m erica (IE S N A ).
T h e sam e authors later review ed th e situation in th e w ork-illum ination field (S im o n so n & Brozek, 1952) and n o ted several difficulties with th e research; th ey disagreed w ith the
illum ination standards o f the day b ecau se o f th ese problem s. In theory, they said, illum ination levels should be set to m inim ize visual fatigue, w hich is a p hysiological problem ; how ever, the research did not show a system atic relationship b etw een fatigue trends in visual functions and illum inance. The work perform ance studies review ed w ere discredited beca se o f poor controls and inadequate procedures. S im on son and B rozek favoured laboratory study o f carefully
dem onstrated the existen ce o f an optim um illum inance, although not its value. T h e optim um should b e the value given in illum ination tables, but further research integrating perform ance, visual functions, and subject u . ic actions w as recom m en d ed as th e best rou te to im provem ents in illum ination standards.
S o m e contem porary research ers have continued this line o f work, including B o y ce (e.g., 1970), w h o found a perform ance optim um betw een 1888 and 4196 lx for a m eter-reading task. Others dispute the existen ce o f op tim al illum inances for task perform ance (e.g., Sm ith, 1978): In North A m erica, the dom inant m o d el during the past 35 years h as been th e visibility m od el o f lighting research b ased on th e work o f H . R . Blackw ell.
T h e visibility m o d el. B lack w ell began his research during the war years, w h en h e conducted an intensive investigation o f the lim its o f hum an vision (B lackw ell, 1946). This research -stablishcd the contrast thresholds for stim uli o f various sizes, contrasts, and exposure tim es a s a function o f adaptation lum inance. A s adaptation lum inance increased, contrast thresholds consistently decreased; that is, observers b ecam e m o re sensitive to low contrasts as they adapted to brighter light. T h is result m akes intuitive sense: W e k n ow from exp erience that brighter light m akes things ea sier to see.
T h e ability to detect d etails is th e key perform ance variable in the visibility m o d el. T h e m odel has reached a highly d efin ed state after d ecad es o f research by H . R . B lackw ell and others. The first detailed report w as p u b lish ed in 1959. It described th e operation o f the V isu a l T ask Evaluator, a device now know n as a visibility m eter. T his d ev ice m akes it p ossible to com pare the contrast o f actual tasks to th e threshold contrast o f a referen ce task at a particular adaptation lum inance. T he tw o contrasts are u sed in th e calculation o f a ratio know n as the visibility level. The d etailed procedure is as follow s.
T h e reference task is th e d etection o f a lum inous disc subtending 4 m in o f arc, view ed for 1 / 5 sec. By exposing subjects to a broad range o f background (adaptation) lum inances, it w as possible to obtain a standard visual perform ance curve relating threshold contrast to background lum inance (Blackw ell, 1959). T h e threshold standard is 99.9% accuracy, rather than th e m ore
com m on 50% d etectio n accuracy, b ecau se it w as felt that, for practical p u rp oses, near-perfect visual perform ance is required. T he standard curve used in the visibility m o d e l was constructed from a population o f 2 0 -30-year-olds w ith norm al or corrected-to-norm al vision, and the
referen ce illum ination is diffuse w hite light with colour tem perature o f 2850 K. Thus, for a given background lum inance, the curve estab lish es the contrast at w hich the lu m in o u s disc is just barely visible; this contrast is called C(L,-).
In calculating visibility level (V L ), the researcher places the target (read in g m aterial or other task m aterial) in to the visibility m eter and oets the background lu m in a n ce for the level under w hich the task is norm ally perform ed (L,.). T h e observer then adjusts a knob that controls a m echanism to cast a veiling lum inance (L y) over the task, like a cloud o f light, until it is just barely visible. T he m echanism is such that the total adaptation lu m inance is constant, because the background lum inance decreases exactly in step with the increasing v eilin g lum inance.
T h e target is th en rem oved from th e visibility m eter and the 4 -ir in d isc p resented. The veiling lum inance previously set is m aintained, w hile the background lu m in a n ce against which the disc is view ed is adjusted to th e visual task (L,.). T h e observer m an ip u lates th e contrast o f the disc until it is just v isib le to give the equivalent contrast (C (cq )) o f the target task in term s o f the standard referen ce task. V isibility level is calculated using th e formula:
C (eq )
VL =
____
C(Lr)
T h e V L thus calculated represents visibility under very special co n d itio n s that never exist outside a visibility m eter. T h e light is unpolarized, diffuse, o f a particular colou r tem perature, and p rod u ced w ith uniform lum inance at all parts o f the task. T he visibility m od el attem pts to correct for th ese sp ecia l conditions by th e use o f multiplying factors [C o m m issio n International d e l’E clairage (C IE ), 11,72].
T h e contrast renderin factor (C R F ) is an index o f th e effects o f th e polarization, spectral com p osition , and sp atial pattern o f the actual lighting installation on task visibility, in com parison to th e referen ce lighting conditions. T h e pattern o f lum inances in the task environm ent can
cither enhance or reduce visual perform ance potential in com parison w ith th e referen ce lighting; this quality o f the lighting installation is captured in th e disability glare factor (D G F ). In som e installations, a third factor becom es im portant, the transient adaptation factor (T A F ). Th o factor m easures the loss in contrast sensitivity that occurs w h en the lum inous environm ent is markedly nonuniform , so that the eye must frequently adapt to different levels o f b rightness as the gaze shifts. T h ese values, calculated from photom etric d a ta from th e actual lighting installation or from a visibility m eter, are used to calculate the effectiv e visibility level (V L ,,H ):
V L ,H = V L x C R F x D G F x T A F
T h e goal o f the visibility m odel researchers is th e creation o f a m o d e l that w ill allow the prediction o f visual perform ance from inform ation a b ou t the lighting installation, the task, and the observer. T h e m o d el is not yet at that stage o f d evelop m en t (B oyce, 1981); how ever, H . R. Blackw ell and O . M . B lackw ell (1968) reported an alyses com paring H . R . B lack w ell’s 1959 standard visual perform ance curve with visual acuity data ob tain ed at various lum inances. T he visual acuity data w ere obtained from studies that h ad used L andolt rings and letter recognition. A greem ent b etw een B lackw ell’s data and the visual acuity data w as good .
T h e C IE has reported that a com parison o f fou r studies o f visual w ork perform ance as a function o f VLj.w sh ow a consistent negatively-accelerating in crease b perform ance with increasing V Lj.,.t (C IE , 1972). T he value o f V L ^ b eyon d w h ich the in crease w as very sm all is approxim ately 8, and this is the visibility level set a s th e design standard by the IE S N A (1987).
O n e difficulty w ith the initial m od el w as its sta tic nature. It w as p ossib le to retroactively fit experim ental p erform ance data to V L ^ , but predicting actual p erform ance proved m ore
difficult. M ost tasks d o not involve on-axis, brief view ing, w h ich is the basis o f th e m odel; they dem and constant scanning. Later versions o f the m o d e l have included a factor that attem pts to correct (as yet im perfectly) for the degree to w hich th e target task d em an d s off-axis view ing (B oyce, 1981). Population data on off-axis sensitivity w ere ob tain ed by th e B lackw ells (H . R. Blackwell & O . M . Blackw ell, 1980).
T he original visibility m odel was constructed from a particular a g e group, and therefore (here w c .e difficulties in applying it to other view ers. W ide d ifferen ces in the standard visual
perform ance curve — the basis for th e entire m o d el — have b e e n reported for different age groups (O . M . B lackw ell & H . R . B lackw ell, 1971). Older adults have poorer contrast sensitivity than th e young adults o f the referen ce population. Thus, th e refined m o d el also incorporates a factor to adjust for the age o f th e view er. W ork continues o n th e determ ination o f the precise adjustm ents n eeded to account for th ese differences from th e reference visual perform ance function (B oyce, 1981).
R elative visual perform ance CRVP1 m od el The N ational R esearch C ouncil o f C anada w as the sou rce o f the m ost recent attem pt at m odelling visual perform ance. R ea (1986a, 1986b) used the philosophical foundations laid ou t by W eston and a m athem atical m o d el from
electrophysio' ;gy to m od el visual perform ance o n a num erical verification task. The
philosophical roots o f th e m o d el lie in th e use o f psychophysics to relate stim ulus dim ensions to quantifiable resp on se d im ensions (R e a , 1986a). Suprathreshold research in electrophysio ogy provided a m athem atical form ula dem onstrating that increasing stim ulation becom es progressively less effectiv e in eliciting greater sen sa tio n (R ea, 1986b); the data used to develop the m odel provided th e sp ecific estim ates for th e param eters relevant to the num erical verification task.
T h e R V P m o d el thus d escrib ed predicts visuai perform ance (resp o n se tim e corrected for m echanical response tim e) on an alphanum eric reading task from the task contrast and background lum inance. D a ta are for suprathreshold perform ance; thus, the RV P m odel avoids the difficulty o f the analytic m o d el in assum ing a linear relationship from threshold to
suprathreshold perform ance (s e e b e lo w concerning this difficulty v/ith IE SN A standards). B ased as it is o n a task that can occur in everyday life, th e m odel enjoys b etter external validity than the analytic m odel; how ever, w e do n ot yet know the extent to w hich its predictions might apply to reading-based visual tasks other than num erical verification.
R e a (1986b) has m ade no claim s that the R V P m odel is com p lete. O ther factors that need to b e incorporated into th e m od el include size o f critical details, view er age, and the lum inance distribution. T h e m od el d o es represent an advance in that it allows th e prediction o f visual p erform ance from task and lighting characteristics, and it predicts k ow perform ance might alter if either is changed. T his k n ow led ge m ight im prove design decisions; for instance, reducing
illum inance from 650 lx to 1M) Lx w ill reduce visual perform ance by only 2 % in young adults (R ea, 1986a). Such a loss m ight b e accep tab le in view o f the potential energy savings o f the lower illum inance. R e a advocated the use o f the R V P m odel in specifying lighting needs in cases where visual perform ance :s particularly im portant b ecau se it allow s this kind o f specificity about the effects o f Stimulus characteristics on visual perform ance.
T h e psychologist m ight n ote, how ever, that th e R V P m odel system atically excludes nonvisual aspects o f pcrform ai'ce. T he corrections applied to the perform ance m easure and the research design itse lf w ere created with p recisely this go a l in mind. R ea (1987) has criticized much o f th e earlier lighting research for b a sin g confounded visual and nonvisual perform ance in a variety o f ways. H e con sid ered betw een -su b jects d esign s to have confounded subject
differences with experim ental variables, ignoring the effects o f random ization; in his advocacy o f within-subjccts d esign s he appeared to d ism iss considerations such as hypothesis-guessing and H aw thorne-type effects. R ea strongly advocated the unam biguous relating o f visual responses and stim ulus con d ition s as the o n ly m eans to refining lighting recom m endations. T his approach, however, seem s unlikely to lead to m ore accurate recom m endations, b ecau se it excludes m any variables that are likely to affect lighting req uirem ents, including individual differences and preferences.
D e sp ite the uncertainties and gaps in our know ledge o f the relationship b etw een light and work, illum ination standards and codes are n eed ed . A variety o f recom m endations for
illum ination levels have b een m ade, each d ep en d in g o n the theoretical m o d el u sed. H aving discussed the principal m odels in current u se. w e turn now to their applications.
Standards and c o d e s. M ost countries subscribe to som e form o f lighting recom m endations for various installations. C ertain o f these regulations — for instance, safety lighting in stairw ells — have th e force o f law, but m ost exist in the form o f c o d e s written by o n e or another professional organization. T he sp ecific illum ination lev els that have been recom m en d ed for various tasks vary widely from country to country; E uropean standards a re low er than N orth A m erican on es, and individual exp ectation s for appropriate lighting vary accordingly (B elcher, 1985). T h e theoretical orientation that und erlies the recom m en d ation s provides a clue to the reason for this difference.
T h e IE S (U K ) C ode for Interior Lighting is th e British lighting standards guide. S o m e sam ples o f its recom m endations are given in T ab le 1 alon gsid e the parallel recom m endations from th e IF.SN A Lighting H andbook (IE S N A , 1987), the N o rth A m erican guide. The British recom m en d ation s, it can b e seen , are uniform ly low er than th e N orth A m erican
recom m en d ation s (table adapted from B oyce, 1981, p. 395).
T h e British code, first published in 1936, is b ased in part on W esto n ’s Landolt ring studies (S im on son & B rozek, 1952), although each su ccessive ed itio n has incorporate 1 changes that reflect n ew inform ation and new lighting technology. T h e IE S N A recom m endations a .e explicitly based upon B lackw ell’s visibility m odel; they set th e goal V L EFF= 8 for all lighting installations. A lthough th e choice o f this V L ^ w as based partly on research evid en ce, it is primarily an arbitrary level that .s believed to b e a sufficient "safety margin" to guarantee adequate illum ination.
T h ese tw o approaches are in g o o d agreem ent, according to an analysis o f visual acuity data (from the analytic m odel stu d ies) com pared to th e standard visual p erform ance curve (B lackw ell & B lackw ell, 1978). T his m eans that W eston’s data, the basis o f the British code, follow a sim ilar function for perform ance at various adaptation lum inances as B lackw ell’s threshold visibility for VLppp = 1. T h e N orth A m erican standards reach for a higher V L ^ - than exists in the research that is the b asis for the U .K . code, so it is not surprising that they should b e higher.
T h e N orth A m erican standards, although com p reh en sive [B oyce (1981) wrote that "the IES Lighting H an d b ook contains all there is to know about lighting and a bit more" (p. 394)], are op en to criticism . T he recom m en d ation s attem pt to ensure illum inances equivalent to VL,,,.,. • 8 a-- estab lish ed using reference lighting conditions; how ever, n o data exist to dem onstrate that any actual lighting conditions provide th e sam e visual perform ance curve as the reference conditions.
Furtherm ore, the lighting calculations that a designer w ould perform in order to convert from VL,.:FF to a task lum inance to b e supplied by the lighting system are based on a sim p le linear extrapolation o f the original standard visual perform ance curve to the higher visibility level goal. R o ss (1978) suggested that th e relationship betw een visibility level and perform ance is not the sa m e for su prathr.shold perform ance; if this is true, it throw s the entire fram ework o f the
Tabic 1
Illuminance Recommendations of North America and Great
Britain for Similar Applications (lx)
A pplication
IESNA
IES(UK)
Clothing factories
--sewing
5400
Glass works —
furnace rooms
320
Engineering
--medium machining
1100
General office
750
Drawing office
2200
Corridorr-'
220
1 0 0 0
150
500
500
750
150
IE S N A recom m en d ation s into doubt. O ther data discussed by R oss sh ow that, for equal visibility levels, th e sam e task presented in different printing to subjects o f various ages can have very different visual perform ance. T h is is contrary to th e m odel assum ption that visual perform ance should b e identical for all tasks o f equivalent V I ^ .-.
T h e IE S N A recom m en d ation s cover an enorm ous variety o f tasks and settings, and include adjustm ents for glare, task im portance, age o f view er, and surface reflectances. It is certain that the equivalent task contrast curves [C (eq)j have not b een obtained for every possible task; indeed, for m any tasks it w o u ld b e im p ossib le to determ ine. O ne serious drawback to the
recom m endations is th e lack o f any explanation o f th e way threshold visibility w as related to most tasks in th e ab sen ce o f data. T h e Lighting H andbook is a trem endous achievem ent, but its credibility as a scientifically derived specification system suffers because o f th ese problem s.
T h e reader o f lighting standards, B o y ce (1981) pointed out, quickly realizes the difficulty in writing su ch a d o cu m en t. Ideally, standards should rest on sound em pirical data; how ever, we do not yet have all th e answ ers. In the m eantim e, th e p eop le w ho design w orkplaces and hom es and schools w ant gu id an ce for their lighting specifications, and they cannot wait for scientists to form ulate th e perfect predictive m odel. A s a result o f this tension, w e do have sets o f
recom m endations, how ever im perfect, that can assist the lighting designer. It rem ains for other researchers to investigate the practical effects o f lighting installations o n behaviour; this is one m eans o f filling in th e gaps left b y the visual perform ance approach to lighting research.
I l lu m i n a n c e , and office w o rk . A lon g tradition o f b elief exists that eq u ates more, fight with better productivity, a b e lie f reflected in th e steady increase in the IE S N A recom m endations from 1942 to 1981 (Pansky, 1985). It is true that brighter fight im proves the resolu tion o f task details, as w e h ave seen; how ever, research evidence for an illum inance effect on paper-and-pencil office tasks is equivocal. T h e q u estion is not w hether it is possible to m ake the w orkplace sufficiently dark to destroy perform ance, b u t whether extra illum ination (over and above that necessary to resolve critical d eta ils) can im rxove perform ance.
A recent attem p t to integrate the literature concerning this issue w as a m eta-analysis conducted by G ifford, H ine, an d V eitch (in press). This quantitative integration o f the results o f
eleven articles found that, indeed, higher illum inances lead to greater perform ance; this sm all-to- m edium -sized effect a m id translate into a 19% im provem ent in perform ance if illum inance w ere increased from 450 to 2900 I.e. H ow ever, the m eta-analysis also su ggested that this effect dim inishes when p e o p le are given tim e to adapt to the conditions, although further research w ill he n eed ed to resolve this question.
O n e strength o f m eta-analysis is its ability to provide inform ation o f th e strength o f a relationship, as w ell as detecting its existence. T h e com bination o f m any experim ental results has • he added benefit o f providing a more stable estim ate o f the effect size. H ow ever, conclusions about causal relations remain lim ited by the quality o f th e studies included in th e m eta-analysis. R egardless o f the effect size estim ated in a m eta-analysis, it is im p ossib le to attribute that correlation to a causal relation betw een the variables if uncontrolled extraneous factors confounded the prim ary research. The discussion b elow describes so m e o f th e research on the illu m in an ce/office w ork issue and the reasons for the equivocality o f the evid en ce.
C lerical w ort has b een a popular choice for experim ental tasks in illu m in ation levels studies because sp e e d and accuracy are easily quantified for num ber verification and paired- com parison tasks. H ughes and M cN clis (1978) used num ber verification p erform an ce to com pare the effects o f three levels o f illum inance, 538, 1076, and 1614 be. G rou p s o f young and old fem ale clerical w orkers w ere hired to participate in the study, w hich u sed a w ithin-subjects design, in the first study, the participants adapted to on e o f the light levels, c o m p leted a sh eet o f num bers to be verified, and th en responded to a num ber o f questions about th e appearance and difficulty o f the task in that light condition. T he illum inance then w as changed, and the
procedure was rep eated . T w o levels o f illum inance constituted o n e sessio n . A ll p o ssib le
seq u en ces o f pairs o f the three illum inance levels w ere used; th ese w ere assign ed random ly to the 9 session s com p leted by each subject.
T he data indicate that processing tim e per sh eet decreased significantly as illum inance increased. O lder w orkers w ere consistently slow er than younger w orkers, regardless o f illum inance. A ccuracy in a secon d task, a variation o f the num ber verification task involving com parisons b etw een colum ns o f numbers, show ed steady im provem ent with increasing
illum inance.
T h e authors reported, in addition to speed and accuracy effects, significant differen ces in subjects’ ratings o f th e visibility, ease, and com fort o f the task, and o f th e brightness o f the lighting levels. H igh er illum inances w ere rated as m ore com fortable and m ore satisfactory; ratings o f the effort required, stim ulation, and distinctness o f th e task increased with higher illum inances as w ell (H u gh es & M cN elis, 1978). T hese effects w eaken the strength o f their conclusions: The procedure o f alternating the subjective ratings with the number verification task leaves o p en the possibility that th e participants becam e aware o f the m anipulation and the hyp oth eses and adjusted their w7ork accordingly. T he design is rem iniscent o f the original H aw thorne studies (R oeth lisb erger & D ickson, 1939), in which a sim iL ' process probably occurred.
B arnaby (1980) reported sim ilar findings, using them as the basis for a cost-benefit analysis o f the effects o f increasing illum inance in insurance offices. H e attem pted to dem onstrate that the co st o f the add ;d lum inaires and electricity w ould be m ore than offset by improved productivity. H is conclusions m ight n ot b e true today because electricity rates have increased substantially in the past d ecad e and are very variable across N orth A m erica; in any ca se, financial costs are not the on ly result o f increasing w orkplace lighting. T he environm ental cost o f the in creased energy production h a s led other writers to recom m end that excessively high
illum inances be avoided, particularly w hen the evidence to support their perform ance b enefits is w eak (e.g ., B oyce, 1981).
Sm ith and R e a (1978) stu d ied proofreading accuracy as a function o f age and illum inance. P erform an ce did im prove m on oton ically with illum inance, although the increase was less m arked for y o u n g than old subjects. In the 500-1600 lx range studied by H u gh es and M cN elis (1978), th e in crease in perform ance w as n egligible in com parison with the standard error. T h ese results cast doubt o n the n otion that raising illum inance levels will im prove office productivity.
C oncerning intellectual tasks o f th e sort done in m ost offices, the evidence for illum inance level effects on perform ance is lacking. N ull results have b een observed w ith a n u m ber o f tasks, each sim ulating a different a sp ect o f office work, as the follow ing discussion will dem onstrate.
O n e im portant dim ension o f nearly all office jo b s is reading com p reh en sion . Sm ith and R ea (1982) com pared perform ance on th e D avis R eading T est under illum inances o f 9.9, 125.9, 1922.2, and 4885 Ix, but found no significant effect o f illum inance on com p reh en sion or reading speed. T h e design w as a w ithin-subjects com parison with the illum inance conditions assigned to sessions using a Latin square. A lthough th e use o f this procedure con trolled for practice and fatigue effects, it did not com pensate for dem and characteristics associated with full k n ow led ge o f the purpose o f the study. A postexperim ental debriefing did occur, but th e authors (Sm ith & R ea, 1982) did not report w hether the subjects had b eliefs that m ight have altered their perforn.ance.
E arlier research (Tinker, 1959) had found that on a lest w ith m axim um perform ance accuracy, reading sp eed increases with increasing illum inance, to a m axim um around 1000 be ev en for poor contrast m aterials. T herefore, Sm ith and R ea (1982) w ere surprised by th e results th ey obtained and suggested that a procedural difference betw een th e studies m ight explain the discrepancy. T inker’s subjects had b een tested using an apparatus that controlled their posture and m aintained a constant view ing distance. T he participants in the experim ent b y Sm ith and R ea w ere perm itted to sit in any posture they chose T he latter strategy w as u sed to im prove th e external validity o f the findings; however, they observed that in dim light subjects ch o se to in clin e their heads closer to the test booklet. T h ey suggested that this m ight have com p en sated for w hat otherw ise w ould have b een perform ance deficits in poorer lighting (9.9 lx).
T h e nature o f reading itself is also a plausible reason for th e null findings o f Sm ith and R ea (19 8 2 ). T h ey pointed out that reading does n ot generally dem and th e discrim ination o f fin e details, but rather the recognition o f w h o le words. Perform ance in g en eral w ould not, th erefore, be exp ected to d eclin e until words could not be identified.
V eitch (1990) replicated the null results o f Sm ith and R e a (1982) w ithin th e 200-600 lx range. T h e study used a belw een -su b jects design so that subject expectations w ou ld b e
m inim ized. T he results, how ever, might have b een obscured by th e low internal reliability o f th e reading test. Posture w as unrestricted in this study, leaving o p en the possibility that subjects com p en sated for the lighting condition in their selection o f a com fortab le seatin g position, as had
b een observed b y Sm ith and R ea (1982).
C reativity is considered another im portant a sp ect o f o ffice work b ecause it dem ands com p lex in tellectu al processing. N elson, N ilsson, an d Johnson (1984) com pared illum inances o f 100 and 300 lx in a betw een-subjects experim ent o f illum inance and tem perature effects on the num ber o f w ords and th e num ber o f stories w ritten in a creative writing task. N o significant m ain effects w ere found o n the perform ance m easu res. M o o d questionnaires show ed that low illum inance led to less boredom than high illum inance.
N elso n , N ilsson , and Johnson (1984) also fou n d a curious interaction b etw een illum ination and tem p eratu re o n m ood: L ow tem perature (13°C ) and high illum inance (300 lx) produced "bad mood" and "sadness" ratings that w ere w orse than th e low tem p era tu re/lo w illum inance condition. T h e tren d w as th e reverse in the high tem perature (30°C ) conditions.
C reative w riting w as also th e task in a recent study by G ifford (1 9 8 8 ). T h e study used a b etw een-subjects design to com pare the effects o f 6 0 versus 900 lx on th e quantity o f written com m unication b etw een friends. T here w as n o statistically significant differen ce betw een illum inance levels in com m unication (G ifford called th e p = .06 for th e lighting effect "marginally significant"). T h ere w as, how ever, an interaction e ffe c t o f Illum ination X T im e, in which
participants in lo w light decreased com m unication o ver tim e, w hile th o se in bright light m aintained th e sam e level.
In sum , although o n e review (G ifford, H in e, and V eitch, in press) suggests that the relationship m ay b e real, design problem s in som e o f this research crea te doubt about the accuracy o f th e statem en t that m ore light m eans m o r e work. T h e v o lu m e o f equivocal results su ggests that future fighting research ought not to s e e k sim p le illum inance effects, but should strike o u t in n ew directions. C hanges have occurred in recent years in th e practice o f interior lighting, and th e se can point the way for interesting research program m es. W e turn now to the history and practice o f lighting design and its relation to fighting research.
L ighting the O ffice
In the first offices d esign ed for adm inistrative or clerical work, illum ination w as entirely p rovided by natural daylight (Sundstrom , 1986). T h e offices o f the m id-19th century were
contained in narrow buildings w ith large windows to take advantage o f all available light. Illum inance levels varied throughout the course o f th e day and with changing w eather conditions and seasons. T he n eed t(. have light penetrate to th e centre o f the building im p osed w idth limits o f 12-18 m because light cou ld penetrate only 7-10 m from the windows. A com m on design feature w as an interior courtyard that allow ed light to reach the centre o f the building. Interior partitions w ere usually m ad e o f glass to allow light to reach the clerical and reception areas from the private offices alon g the exterior walls.
It was not until the 1930s that artificial light b eca m e com m on in offices in N orth A m erica (Sundstrom , 1986). A s the advances in technology m a d e brighter interiors feasible, standards for interior light levels rose sharply (Pansky, 1985; Sundstrom , 1986). T he introduction o f fluorescent lam ps in the 1940s rem oved the practical difficulty o f overcom ing the heat production o f
incandescent lam ps and provided greater intensity o f light alon g with im proved energy efficiency. Soon, co o l white fluorescent lam ps b eca m e the standard for office lighting, and the illum inance level goal for office lighting w as set at 1000 lx.
C onventional office lighting consisted until recen tly o f uniform ly bright overhead
fluorescent lighting provided by evenly spaced lum inaires throughout the o ffice. H ow ever, poorly designed system s can create glare, and m on oton ou s designs are not aesthetically p leasin g (Ellis, 1986). T h e problem o f glare attracted increased atten tion follow ing the introduction o f video display term inals (V D T s) in m any office in the la te 1970s and early 1980s (e.g ., Shahnavaz & H edm an, 1984). V eilin g reflections (direct reflection s o f light sources) and glare on V D T screens w ere frequently associated w ith com plaints o f eyestrain and headache (e.g., Isen see & B ennett, 1983).
T he response from lighting designers w as to develop n ew lum inaires and to change the design convention (e.g., F loren ce, 1989). Typically, offices w ith V D T s today have a low er overall am bient illum inance su p plem ented by personal m o b ile task lighting that allow s individuals to direct the light w here it is m o st n eeded. T his d esig n represents a fundam ental shift from the assum ption that visual com fort is m axim ized by uniform lum inance across th e space to the b elief that variable lum inance is preferable b ecause it crea tes visual interest (cf. V ischer, 1989).