PhD theses

Cultural factors in Libya require female privacy to be maintained. Outside the home, female must wear clothing that reveals only their face and hands. When inside the home and located near windows, a similar degree of clothing cover or window treatment is required. This reduces exposure to natural daylight, with a resultant reduction in the health benefits of daylight. Females who wear hijab dresses when outside of the home expose only their hands and face, an exposed skin surface area of only 11.6% compared with 61% if wearing western-style summer clothing. Clothing restriction can be relaxed when in the home, but here the female privacy is maintained using window treatments and these also restrict access to daylight. Currently used window treatments in Libya (the roller blind and wooden shutter) make the interior space completely dark when closed, but when opened the interior space is exposed to the outside, which offers no privacy, and hence hijab style clothing must be worn when inside. This thesis explored the potential of window treatments to offer enough privacy so that females of some Muslim cultures might wear relaxed clothing when at home instead of needing to wear a high level of clothing.

The first stage in this study was a validation experiment where a novel pictorial clothing scale was created to allow females to state what level of clothing is needed to maintain privacy in different contexts. The result of the questionnaires inviting female participants from three nationalities (Saudi, Libyan and European) showed that variations in the cover and tightness of clothing affect the perceived level of privacy in different situations. For Libyan women, while a head scarf, and arms and legs fully covered by a jacket and trousers was the median expectation when inside the home but potentially visible to a stranger, this could be relaxed (to tighter-fitting clothing, greater degree of skin exposure) if visible only to members of the family.

The second stage was to explore the ability to provide sufficient visual privacy with two window treatments, horizontal blinds and frosted glass, varying the free area and degree of frosting respectively. The degree of privacy offered was operationalised by identification of the clothing level worn by a target behind the window treatment, the aim being to reduce identification to a chance level. Two actors were used, to consider the effect of skin tone, and two durations, to consider the effect of gaze behaviour. For observations of 0.3 s duration, only the extreme level of each treatment (horizontal blinds set to 3% free area and distortion level 20 for the frosted glass) led to chance levels of clothing identification, for both actors.

Road lighting in minor roads is designed primarily for the needs of pedestrians and cyclists. Pedestrians’ considerations include their safety and their feeling of safety after-dark. One factor that contributes to a pedestrian’s feeling of safety is the ability to evaluate other people (known as interpersonal evaluations); road lighting should enhance the ability to make the visual component of this evaluation. While past lighting studies have assumed the face is the critical target, and hence investigated the effect of lighting on facial recognition, this assumption has yet to be verified. Different personal features subtend targets of different sizes, colours, and contrasts, so a better understanding of the important visual cues is essential for appropriate research.

Three experiments were conducted to investigate the visual cues used by pedestrians in interpersonal evaluations. The first two used a subjective evaluation procedure, in which images portraying pedestrians in different situations were evaluated using either category rating or both category rating and paired comparisons. The assumption was that those situations rated (or chosen) as less safe would indicate critical visual cues. Experiment 1 compared the relative importance of gender, number of people, walking direction, direction of the fall of light, and the exposure of the face and hands: the results suggested that the exposure of face and hands had a greater effect than other cues, but did not distinguish between the face and hands. Experiment 2 was therefore conducted to further investigate exposure to view of the face and hands: the results suggested that the face is a more important cue than the hands. To test the robustness of these findings, Experiment 3 investigated the same question as Experiments 1 and 2 but used an objective measure, eye-tracking. The core assumption was that the most important visual cue in a scene is that which receives the first visual fixation and the longest duration of visual fixations: the results again suggested that the face is the most important visual cue in interpersonal evaluations.

It is therefore concluded that the face is the key visual cue used in pedestrians’ interpersonal evaluations. This supports the assumption of previous research investigating the effect of changes in lighting for pedestrians.

The lighting recommendations and guidelines for pedestrians propose that road lighting in residential roads mainly aims to enhance the walking safety after dark. However, the lighting standards may not be supported by sufficient empirical evidence.

The key visual tasks for pedestrians are obstacle detection and facial emotion recognition (FER). These have been studied in previous work but there are a number of limitations: FER studies have used 2D images and not 3D models; obstacle detection studies have used raised but not lowered trip hazards; these tasks were the sole focus of trials and hence were able to use a greater degree of cognitive resource than when in natural conditions. Further work was therefore conducted to investigate these limitations, and the implications for previous conclusions about how lighting changes affect the ability to detect peripheral objects and identify facial expressions.

Two pilot studies were conducted to test if 3D face models can be used for FER. The results confirmed that 3D face models could replace photographs by comparing the results with previous studies which were using photographs.

Three experiments were carried out. Experiment 1 compared obstacle detection performance when raised or lowered obstacles: no significant difference was found. Experiments 2 and 3 followed the methods used in previous obstacle detection and FER experiments but sought performance of these tasks in parallel rather than as separate experiments, thus to explore whether multi-tasking affects the performance of obstacle detection and FER. Experiment 2 used two illuminances; experiment 3 used a similar combination of obstacle locations, obstacle heights, emotion types and task conditions but expanded to five levels of illuminance. The results revealed a plateau-escarpment relationship between both obstacle detection and FER and light level.

To consider the impact of multitasking, these results were compared with the results of previous studies where obstacle detection and FER were performed in isolation. This comparison suggests that the performance of each task was impaired when conducting multi tasks.

It is concluded that the optimal horizontal illuminance for obstacle detection is 1.0 lux, even for multi-task condition. For FER, the optimal luminance was suggested to be 0.53 cd/m2 which was slightly lowered than proposed before. Further work is required to address the limitations of this research, including the impact of disability from glare, and variations in face orientation and lighting geometry.

Previous studies have demonstrated that a room with good view to the outside can provide its occupants with certain psychological benefits. However, the characteristics that constitute a good (or bad) window view have remained unclear. From literature review, it was hypothesised in this study that the quality of a window view is attributed to seven factors: proportion of greenery, number of visual layers, view elements, balance of view, diversity of view, openness of view and depth of view.

To test these hypotheses, 12 urban and sub-urban scenes were selected; 62 subjects were recruited to perform on-site viewing and evaluation of the selected scenes. The method of the view quality evaluation was based on real scenes viewed through “virtual windows” as defined by a portable viewing box, which was set up on site by the researcher. The viewing box enabled the observer to view the actual scenes as if viewing the same scenes through a physical window of 1.2 metres by 1.2 metres in size. Instead of the conventional “view satisfaction” level used in the previous studies, the rating scale for this experiment employed two different dimensions of affective quality– i.e., “pleasantness” of view (POV) and “excitingness” of view (EOV) as the basis for the verbal descriptors, which were anchored to a 4-point and a 10-point numeric scales.

The results of the first experiment were used to test the view quality predictions made using the seven view attributes. In addition, the experiment results were used to test whether there was a significant difference in the subjects’ evaluations of view quality between the 4-point and 10-point scale formats after both primary scale data were rescaled into a common 101-point scale. A second experiment was carried out to test the hypothesis that there is a significant difference in the perceived window view quality between actual-view and image-view modes. The second experiment was a systematic replication of the first: photographic images of the selected 12 window views were displayed on computer screen for a different group of 62 subjects to evaluate the view quality of the scenes using the same questionnaires for the first experiment.

Stepwise multiple regression and ordinal logistic regression analyses were conducted on the 10-point and 4-point scale data respectively to formulate prediction models of view quality. Results show that among the seven proposed view attributes, “view elements”, “balance of view” and “openness of view” were significant predictors of view quality in the linear model of POV. “Depth of view” appeared to be the poorest predictor of view quality – neither linear nor monotonic relationship could be established between this attribute and the view quality. “View elements” and “openness of view” were also significant predictors in the ordinal logistic model of POV. Validation of the proposed linear prediction model for POV was conducted using correlation analyses and one sample t-tests that compared the predicted view quality with a set of out-of-sample view evaluation data from a third experiment, which involved an independent group of 40 subjects.

The outcomes of analysis show that there is no significant difference in the mean POV (EOV) scores between the rescaled 4-point and rescaled 10-point ratings – whether the evaluation is carried out in actual or image viewing mode. In terms of scale reliability, the 4-point and 10-point scales in most cases showed moderate to excellent internal consistencies. Whether it is for actual or image view, 10-point scale appeared to have higher internal consistency and interrater reliability in most cases compared to the 4-point scale.

Overall, the results confirm the construct validity of the rating scales (either 4-point or 10-point scales) that were used in the assessment of actual or image view quality. The results suggest that 10-point scale is probably too fine for the purpose of evaluating window view quality, whilst 4-point scale is perhaps too coarse to achieve a sufficient discriminating power between the scale points. The optimum number of response categories on a rating scale for evaluating window view quality may be either 6 or 8. The study shows that there is no significant difference in the perceived view qualities between actual and image views. However, POV (EOV) ratings of the actual views generally have larger variances compared to that of the image views, probably because the subjects were affected by other visual cues when looking at the window views in real space, which contrasted with window views in pictorial space.

One reason for installing road lighting in subsidiary roads is to enhance pedestrian reassurance after dark. Low reassurance has been associated with poor mental health, social isolation and lower active walking. However, despite numerous studies, it remains unclear if there are optimal lighting characteristics for pedestrian reassurance.

Two field studies were carried out in the city of Sheffield in the UK. Field study 1 was designed to examine the day-dark approach proposed by Boyce et al. 2000, which uses evaluations of reassurance in the daytime as well as after dark, rather than after-dark only. Thus, this study had 24 participants, rating 10 test locations in daytime and after-dark, using a survey. It also considered the development of a composite evaluation item to characterise reassurance rather than rely on the response to a single question.

The results of field study 1 suggested that reassurance was better characterised by minimum illuminance and uniformity than by mean illuminance, the usually considered metric, but that was not an apriori hypothesis of field study 1. Therefore, Field study 2 was carried out with an expanded sample (35 participants) and a set of locations (16 roads) to test that hypothesis and also to consider the association between reassurance and three types of illuminances referred to in lighting guidance - horizontal, hemispherical, and semi-cylindrical. Results of Field Study 2 suggest the minimum horizontal illuminance and hemispherical mean illuminance are more relevant than horizontal mean illuminance for pedestrian reassurance.

Finally, some consideration to methodological matters is given, such as the impact in findings of asking participants to imagine after-dark settings and the validity of subjective assessments of lighting. Responses to an item regarding the perceived risk at night were analysed. These analyses suggested that asking participants to imagine an after-dark scenario might promote lower perceptions of safety. Also, the association of subjective evaluations of the lighting were analysed against the lighting metrics and reassurance appraisals resulting from study 1 and 2. Findings suggest that the perceived quality of lighting, in both studies, is associated with the recorded significant illuminances of each study.

A review of the literature suggests that current guidelines for road lighting lack a clear empirical basis. Where there is evidence, this tends to be based on motorists or pedestrians: there is little, if any, consideration given to the needs of cyclists.

This thesis presents an investigation of lighting for cycling after dark within an urban environment. Three empirical investigations were conducted. A field survey was conducted to investigate the influence of the ambient light level on the tendency to cycle. Mobile eye-tracking was used to investigate the gaze behaviour of cyclists in natural settings. This was done using two parallel measurements to reveal the critical of these fixations: performance on an audio dual-task and skin conductance response (SCR), and by that improved the ecological validity of previous similar research. A laboratory experiment was conducted to investigate obstacle detection under variations in the type, location, and level of lighting.

The field study revealed that cycling increases when the ambient light level is higher. This suggests that road lighting might be a tool to encourage more cycling. The eye-tracking study suggested that observing the path ahead is a critical task, reflecting a tendency to search for possible obstacles on the road.

Post hoc analysis of the eye-tracking data also suggested an influence of ambient light level on gaze towards aesthetic elements (architectural features) of the environment with such elements are suggested by the literature to be associated with positive cycling experience: this suggests that appropriate road lighting motivates the choice to cycle. The detection experiment revealed two significant effects: first; that road lighting and bicycle lighting may conflict. In other words, using bicycle lighting on a lit road may impair detection performance, not improve it. Second; that detection is improved when the front bicycle lamp is located on the wheel hub rather than the handlebar.

Collisions at intersections, which involve a car driver pulling out into the path of another road user (often a motorcycle), have generally been attributed to a failure in the driver’s visual search. These incidents have thus been described as being the result of ‘look but fail to see’ (LBFTS) errors. This explanation suggests that although the car driver directs their attention towards the approaching vehicle, they do not form a representation of this vehicle, indicating a perceptual error.

Previous theoretical frameworks used to understand these crashes have focussed on attentional and perceptual explanations, as well as the possible influence of top-down and bottom-up factors on drivers’ behaviour. While the investigation of the influence of top-down factors, such as experience and attitudes, has been continued in the current thesis by examining how these factors may affect drivers’ behaviour at junctions, the thesis also considers other potential explanations for the LBFTS error, by breaking down the previous framework into specific testable stages. The structure of the thesis and the theoretical basis for it is described in an extended introduction which is followed by six specific papers presented in the format in which they have been published or submitted for publication.

The first paper in the thesis describes an online survey completed by 579 motorcyclists and 102 car drivers exploring their opinions about junction crashes. The second paper describes a systematic review exploring the role of experience in drivers’ visual search strategies. After this there are a series of four papers describing seven different studies conducted in a high-fidelity driving simulator and a validation of one of these studies with drivers in an instrumented car driving real roads in Nottingham.

These papers describe the development of a methodology for exploring junction crossing behaviour in a simulator and revealed that drivers are prepared to accept risker gaps in front of approaching motorcycles compared to cars. One critical finding was that drivers’ representations for approaching vehicles were often surprisingly poor. There were occasions when drivers failed to report critical approaching vehicles, with these significantly more likely to be motorcycles than cars. These report failures were not predicted by how long the driver fixated on the approaching vehicle, but by drivers’ subsequent visual search after fixating on the vehicle.

One possible interpretation of these findings is that working memory may play a critical role in understanding junction crashes, with new information interfering with the retention of other safety critical information. A new framework for understanding the role of working memory in such situations is presented, along with proposals for practical interventions to prevent this crash.

Chloe Robbins on Researchgate

As a dynamic element revealing architectural space, daylight not only provides substantial illumination but may also influence how occupants interact with the space. This thesis investigates one aspect of interaction, whether there is an effect of daylight on seat choice behaviour.

Previous studies have provided limited evidence of an association between daylight and seating preferences of individuals, in part because each study employed different methods to measure and quantify seating preferences of individuals. This concern is compounded by the fact that previous research has tended to use a unique set of daylight metrics in addition to a unique set of measurement points in the test space. This raises the discussion as to the method by which daylighting conditions were evaluated and the procedure with which seating preferences were sought.

This study used two procedures to examine whether daylight affects seating preferences in an open plan room. The first was a stated preference approach in which individuals were asked to indicate the factors they perceived to influence their choice of seat location. Responses were sought from both those who were about to enter the room and those who were already seated in the room. Daylight was suggested to be the most important factor amongst those respondents already seated in the room, but was less important among those people who responded at the entrance.

The second was a revealed preference approach which draws inferences on seating preferences from the actual choices made by individuals in the test room. The data were collected using two methods. One was a snapshot method, recording actual seating behaviour of individuals at regular intervals and the other was a walk-through method, following individuals from the moment they entered the room until they chose a seat. The influence of daylight was investigated using a dynamic simulation modelling method to predict daylight illuminance in the test space. The method was to derive a set of daylight metrics for each individual seat over the observation period.

Results showed that higher illuminances led to increased seat occupancy, but only in close proximity to windows. It was found that using a questionnaire to ask people about their seat choice when already seated led to the suggestion that daylight had stronger influence than was found in the revealed preference approach.

Current guidelines for pedestrian road lighting are not based on empirical evidence. One approach to providing suitable evidence is to examine the effect of lighting on the visual tasks of pedestrians. This first requires an understanding of what these visual tasks are. An eye-tracking study was carried out in which pedestrians walked a real, outdoor route during the day and after-dark. A novel dual-task method was used to identify the critical visual tasks of the pedestrians. Reaction times to a concurrent audio response task were used to indicate instances when attention may have been diverted towards something significant in the visual environment. Analysis of the eye-tracking videos at these critical times found that the path and other people were the two most significant items looked at.

Observation of the path is important for detection and avoidance of obstacles and trip hazards. Good road lighting should therefore facilitate obstacle detection. An obstacle detection experiment was therefore carried out examining the effect of illuminance and Scotopic/Photopic (S/P) ratio on obstacle detection. The experiment improved the realism and ecological validity of previous research by introducing a dynamic fixation target, realistic apparatus scales and real walking (on a treadmill) whilst carrying out an obstacle detection task. Results showed that obstacle detection only improved with illuminance increases up to 2.0 lux. A higher S/P ratio (2.0) provided better detection performance than a low S/P ratio (1.2), but only at the lowest illuminance used of 0.2 lux.

The data is used to discuss optimal design criteria for pedestrian road lighting based on obstacle detection. However, other purposes of road lighting, such as creating a feeling of reassurance and enabling accurate interpersonal judgements to be carried out, should also be considered when designing pedestrian road lighting.

This thesis proposes that the spectral power distribution (SPD) of lighting can be modified to enhance spatial brightness. Energy saving is then possible by using SPD that allows illuminance to be reduced whilst maintaining the same level of spatial brightness. The Akashi and Boyce study demonstrates an energy saving of 33% by using lamps of higher correlated colour temperature but it is widely known that this is not a good metric for predicting spatial brightness.

The aim of this study was to identify a metric for predicting spatial brightness. The first approach followed the method of Cowan and Ware: use the results of past experiments to test potential metrics. 65 studies of spatial brightness and SPD were found. Initially, these lead to different conclusions as to whether SPD affects spatial brightness. The reasons for this are that they used different methodologies and hence review of method was used to screen the credible data from within these 65 studies: only 19 of them were considered to be credible.

This thesis focussed on the category rating procedure. The review of methods included an experiment comparing rating scales with different response ranges and a meta-analysis comparing results gained when either brightness or visual clarity were the objective of the experiment. Two potential metrics for spatial brightness are the scotopic to photopic (S/P) luminance ratio and the area of the colour gamut (GA). Results from the credible studies were used to test these models: while both models suggest a reasonable prediction, it was found that they were not independent for this set of data and it was therefore not possible to discriminate between them.

Hence an experiment was carried out to directly test these metrics. The experiment employed full field sequential evaluation of stimulus pairs, with matching and discrimination procedures. Three SPDs were compared, these chosen to isolate the S/P and GA effects. Following Berman et al, one pair had identical chromaticity but different S/P ratios: a second pair had identical S/P ratio but different gamut area; the third pair had different S/P and gamut area. The two procedures led to similar results: null condition trials confirmed that doubt about interval bias in the Berman et al data was unwarranted. It was found that lighting of higher S/P or higher GA enhance spatial brightness: it was also found that their effects appear to be additive.

When the final remodelling was done by adding the data points from the new experiment to the data set, the models of the difference of S/P ratio and the log ratio of GA had the best fits with spatial brightness. Their correlations were equally plausible with mean illuminance ratio of the data set.

This thesis demonstrates that SPD affects spatial brightness, allowing lower illuminances to be used when using lighting of higher S/P ratio and gamut area.

Lighting in residential roads is designed to enhance the visual ability to make interpersonal judgements, which is considered to be a critical task for pedestrians. There appears to be little empirical evidence supporting current standards and consistent conclusions cannot be derived from past studies based solely on facial recognition. This work extends investigation of the relationship between lighting and interpersonal judgements beyond the analysis of facial recognition. The results were used to explore how such data might be used to better estimate appropriate light levels for outdoor lighting.

Analysis of gaze behaviour using eye-tracking suggested that the effect of lighting on interpersonal judgements should be examined using the ‘desirable’ distance at 15 m and a duration of 500 ms: in past studies these have been arbitrary. Two pilot studies carried out to inform the experimental design suggested that (i) recognition of facial features is of particular interest, and (ii) standard facial expressions and body postures did not lead to consistent judgements of intent.

The first experiment collected forced choice judgements of emotion (from facial expression and body posture) and gaze direction after 1000ms exposure under 18 combinations of three luminances, two lamp types and three distances. Better performance was found with higher luminance and closer distance, but with diminishing returns according to a plateau-escarpment relationship. Effect of lamp type was not found in judgements of facial expression, but was found in judgements of body posture and gaze direction for some of the conditions lying on an apparent escarpment.

The second experiment provided further examination of facial expressions under 72 combinations of test conditions: six luminances, three lamp types, two distances and two durations. Luminance and distance were found having significant effect on expression recognition. The effect of lamp spectral power distribution (SPD) was not significant and the effect of duration was suggested to be significant only within the escarpment region of the performance versus luminance.

This thesis presents an investigation of disturbing reflections on display screens. Preliminary investigations of lighting for the Classrooms of the Future revealed problems with disturbing screen reflections, in particular on the interactive whiteboard, a new type of display screen not covered in current lighting guidance. There is evidence to suggest that the current guidance does not accurately predict user acceptability of visual conditions at display screens and does not accommodate rapid change in display technologies. A new system for predicting the user acceptability of display screen reflections is needed.

An experiment was carried out to evaluate acceptability of screen reflections using two psychophysical test methods: adjustment and category rating. Both methods identified the light source luminances at which screen reflections were just starting to be unacceptable to users. In addition to these subjective test methods, a reading task was carried out to provide an objective assessment of the effect of screen reflections on visual performance. All the tests were done using a range of screen types, chosen to represent those commonly found in ICT classrooms.

Results from the adjustment tests and the category rating tests are consistent with each other. They show that the current guidance for luminance limits to avoid disturbing reflections on display screens, in some cases specifies luminances that are too high leading to unacceptable reflections, and in other cases specifies luminances that are unnecessarily low. Analysis of the results from both subjective tests suggests significant effects of lighting-display parameters on user acceptability of screen reflections including screen type, angle of viewing, and size and luminance of the reflected light source. By contrast, the objective measure of task performance in the reading test was not responsive to reflections on the screens.

The findings of this study suggest that the acceptability of screen reflections can be predicted using knowledge of key lighting and display parameters including the size and luminance of the reflected light source, the screen specular reflectance, the effect from haze reflection, and the screen background luminance. A new model has been developed using these parameters to predict the acceptable luminaire luminances that can be exposed to display screens. And a new approach is proposed: rather than limit the lighting as is currently done, disturbing reflections can be avoided by means of minimum specifications for display screens.

Wayfinding tasks comprise decision points and interconnecting paths leading to a destination. Path choice at decision points is critical to the successful completion of wayfinding tasks. Research has found that signage is not the only influence on path choice and that influences vary depending on familiarity with an environment. People familiar with their surroundings have a cognitive map – a prior understanding of the environment – against which they can compare the environment as they experience it in order to orientate themselves. People unfamiliar with their surroundings, and therefore lacking a cognitive map of them, are found instead to rely upon wayfinding strategies to inform their path choice decisions.

This study investigates how aspects of the spatial design of buildings may assist unfamiliar users in finding the destination they are seeking within the building. Observations of people wayfinding in an unfamiliar building suggested that four aspects of spatial design affected route choices made at decision points. Four wayfinding strategies describe the behaviour observed:

1. Maintain a straight bearing through the building,

2. Avoid a change of level;

3. Walk towards a brighter space;

4. Choose the wider corridor.

Evidence supporting three of these was found in the literature. For the fourth – Choose the wider corridor – only limited evidence was available from the literature and hence further work was carried out to test the predictability of its influence on wayfinding behaviour.

An online experiment was conducted to investigate to what degree corridor width influences path choice and the interaction between the Choose the Wider corridor and Maintain a straight bearing wayfinding strategies. A means of categorisation, comprising two wayfinding principles, was devised for information in the environment and means of undertaking wayfinding tasks: reassurance principle – wayfinding strategies reassuring the wayfinder that they are taking the correct route - and Tools Principle - signage, maps, landmarks and other sources of information in and representing the environment, available to aid wayfinding decisions.

This thesis looks at strategies for wayfinding reassurance. It is proposed that unfamiliar users would find buildings more intuitive to wayfind within if they were designed with routes to likely public destinations that conform to the four wayfinding strategies. An applied test was conducted to confirm whether wayfinding ease could be predicted by analysing the routes within that building against the behaviours described by the wayfinding strategies. It was found that ratings of difficulty given by test participants matched predicted ratings based upon an analysis of the building’s conformance to the wayfinding strategies. It is suggested that if this analysis was conducted at the design stage it could limit potential wayfinding difficulties. Some possible designs as means of achieving this in new buildings and refurbishments are discussed.

Street lighting which operates at mesopic light levels has traditionally used low- and high-pressure sodium lamps because of their high luminous efficacies and long reliable lifespans. The characteristic orange light and poor colour rendering properties of these lamps now tend to be accepted less in the lighting of areas with significant levels of pedestrian activity.

Here the current trend is towards the use of lamps such as compact fluorescent and metal halide which by comparison with sodium lamps have a white colour appearance and good colour rendering. These effects are a function of lamp spectral power distribution (SPD) and, because SPD is understood to affect visual responses other than colour perception, a clearer understanding of its influence in mesopic conditions is required. Without an established relative spectral sensitivity function for mesopic conditions the matching of pedestrian lighting using standard V(λ)-based photometry does not guarantee equal levels of vision when different lamp SPDs are involved.

There is evidence to suggest that the replacement of sodium lamps with white lamps at the same illuminance can provide a brighter environment and increased visual performance. Confirmation of these effects would justify the revision of current lighting specifications for pedestrian areas. This study compared the performance of five different lamp SPDs in a series of psychophysical experiments designed to assess visual performance and brightness in mesopic conditions. Each experiment systematically isolated the effects of lamp SPD from those of the other lighting parameters.

The potential for illuminance reductions in pedestrian environments through the use of white lamps over sodium lamps was thus examined. Taken together, the current findings clearly demonstrate the importance of lamp SPD and consistent effects among younger and older pedestrians. As far as the maintenance of brightness levels is concerned, an illuminance reduction of approximately 30% is acceptable for white lamps. The consequences of this for the important pedestrian tasks of facial recognition and orientation are concluded to be small. The result of this study is a better understanding of the relative visual effects of the different lamp types used for outdoor lighting. The overall conclusion is that the potential exists for illuminance reductions with no loss of vision if white lamps replace sodium lamps in pedestrian environments.

It was confirmed by a literature search that the V(λ) curve, upon which photometry is based, ignores the contribution to brightness from the chromatic (opponent colour) mechanisms of visual processing. Therefore, standard photometry does not adequately represent the visual response to interiors illuminated by light sources of different colour property.

Experiments were conducted, comparing the response of human observers to seven discharge and tungsten lamps of different colour property illuminating adjacent booths. When the booths were matched for visual equality, it was found that lamps of good colour quality (full spectrum fluorescent) were set on average to a 20% lower illuminance than a reference source (warm white fluorescent), and lamps of poor colour quality (low pressure sodium) required over 120% more illuminance than the reference lamps.

It was also found that at visual equality the lamp of higher colour quality offered equal visual performance, and was preferred by the majority of observers for their workplace, despite being at a lower illuminance than the lamp of poorer colour quality. A number of models of visual response were reviewed and adapted to fit these new results, which include a wider range of lamp spectral quality than has been previously used.

Three models are proposed for further evaluation: (1) Cone Surface Area, a colour gamut model that accounts for both the colour appearance and colour rendering properties of a spectrum: (2) SWS-Lumens, a modification of Berman's Brightness Lumens in which the rod response is replaced by a SWS-cone response, and: (3) Chromatic Brightness, a two-channel opponent-colour model which excludes the V(λ) luminance response.