3 3. RESULTS AND DISCUSSION

Sixteen subjects completed both phases of the experiment. They were staff and students at the Lighting Research Center, but did not include faculty or others who had foreknowledge of the experiment. There were 6 males and 10 females.

3.1 3.1 ILLUMINANCE SETTINGS

Table 2 shows means and standard deviations for the illuminance settings made by the subjects in the test situation, and these data are illustrated in figures 10 and 11. In every case the illuminance in the comparison situation was preset to 50 lux, and it is apparent that subjects set the illuminance in the test situation to match the illuminance in the comparison situation.

Fig. 10.

Illuminance settings for low CCT

Fig. 11.

Illuminance settings for intermediate CCT

3.2 3.2 ASSESSMENTS OF APPEARANCE

Tables 3–8 show the distributions of the subjects' assessments of the appearance of the experimental condition in the test situation relative to the matching condition in the comparison situation. It can be seen that there are few instances of subjects reporting anything more than a slight difference between the appearances of the two rooms. In fact, all of the mean values correspond to fractional parts of a slight difference, and in many cases the reported difference of appearance between the three-band source and the comparison MR source was less than the reported difference between identical MR sources.

The brightness and clarity criteria attracted very few comments from the subjects. The color appearance of the Escher print illuminated by the three-band source attracted several comments. Some subjects who rated the difference favorably described the test situation as appearing “more white,” while others who rated it unfavorably criticized the nonuniformity of color appearance. Despite painstaking adjustment, the blending of the colored light sources in the test situation was imperfect, and this shortcoming was more evident when viewing the achromatic print than when viewing the chromatic artworks.

The color appearance of the Mondrian illuminated by the three-band source also attracted comments. The colors were described as “less vivid” or “faded,” particularly the red pigment. However, at the lower CCT there was some favorable comment that the blue pigment appeared brighter, and also unfavorable comment that there was an overall cooler appearance. These comments are understandable. Incandescent filament lighting enhances the apparent saturation of red colored surfaces, particularly at the lower CCT, and reduces the apparent saturation of blue surface colors.

The appearance of the Renoir painting under the three-band source attracted a variety of both favorable and unfavorable comments. At the lower CCT, favorable assessments (“colors seem to look better”) were matched by unfavorable assessments (colors “more washed out”). At the higher CCT, flesh tones were criticized.

3.3 3.3 DISCUSSION OF RESULTS

Subjects had been instructed to match the appearance of the test gallery to the comparison gallery. Thornton would have predicted that they would select a lower illuminance for the three-band source because, he claims, this source has greater visual clarity; but this was not the case. The subjects matched the illuminance in the test gallery to the illuminance in the comparison gallery in every case, indicating that illuminance effectively evaluated the overall sense of equality of appearance despite the very different spectral compositions of the two types of lighting.

The fact that the three-band source provides a given illuminance with substantially lower irradiance on the illuminated object offers a significant benefit for conservation. For the light sources used in the experiment, the irradiances in watts per square meter (W/m2) to provide 50 lux on the artworks are shown in table 9.

It is conventional to assess the exposure of illuminated museum exhibits in terms of lux hours per year (lx h/y), so that an object lit to 50 lux and exhibited for 3,000 hours per year is exposed to 150,000 lx h/y (IESNA 1996). This measure does not distinguish between the different irradiances of light sources at the same illuminance. If the light source in this example is a regular MR lamp, then changing to a three-band source at the same illuminance would reduce the exposure by 41% and would be equivalent to only 89,000 lx h/y of exposure to the regular MR lamp. Looked at another way, it would take 1.7 years of exposure to the low CCT three-band source to subject the object to the same effective exposure as would occur in one year with a regular MR lamp at the same illuminance. For the intermediate CCT sources, it would take 1.4 years.

Although the color temperatures of the MR and three-band sources had been carefully matched, it should not be supposed that their color rendering would be identical. The direction of the study was to assess whether, when matched for overall appearance, the three-band source would be an acceptable alternative to the MR lamp. As has been explained, incandescent lamps score high on the color rendering index scale because they have continuous spectral power distributions that closely match the reference sources that represent a CRI score of 100. Particularly at the lower color temperature, incandescent sources emit strongly at the long-wavelength end of the visible spectrum, which has the effect of increasing the vividness of red pigments and poses the question, what is meant by color distortion? Because the standard is a black-body reference source, for any low CCT source to achieve a high CRI rating it must have strong emission at the long-wavelength end of the visible spectrum. This emission will have the visible effect of enhancing the appearance of red pigments and making flesh tones appear warmer. This difference was recognized by the subjects and attracted a variety of comments concerning the appearance of the Mondrian and Renoir paintings. The comments were not necessarily unfavorable toward the three-band source, and whether it is appropriate to treat incandescent sources as the standard for color rendering is questionable, particularly at low color temperatures. At least we can be fairly confident that the artists did not paint these works under incandescent lighting.

The subjects' rating data are not suitable for elaborate statistical analysis because the scaling interval is arbitrary, that is to say, the interval between no difference (0) and slightly more (+1) is not necessarily equal to the difference between more (+2) and much more (+3). A simple comparison of the subjects' assessments in terms of average ratings assumes that a rating of −2 is canceled by another rating of +2. This is not an unreasonable approach to gaining an overall indication of acceptability, but examination of the data shows that there was a greater scatter of ratings for the three-band source. This difference is indicated by comparing the mean sum of squares for the experimental conditions, shown in tables 10 and 11. A rating of either −2 or +2 adds 4 to the sum of squares, from which the mean value for the data set is calculated. Table 10 shows that, for all artworks, there was little variation in the ratings of brightness difference, but for all of the other criteria the scatter of ratings was greater for the three-band source than for the MR lamp. Subjects had made the adjustment for equality of appearance, and the impression of brightness was not different for the two types of light source. However, subjects were conscious that the three-band source was different from the MR lamp. They rated it to be different for color appearance and for the sense of clarity and naturalness, but their comments did not indicate discrimination against the three-band source. Table 11 shows the mean sum of squares for all five criteria for each of the three artworks, and in every case the scatter of ratings was greater for the three-band source. A reason for the achromatic artwork's being assessed differently for color appearance has been mentioned, and it seems likely that it was less influential for the two chromatic artworks.

Moving on to consider the second research question, there are two approaches to developing practical light sources that could gain the conservation advantage of the three-band source. A new type of filter could be developed that would convert the continuous spectrum of a regular MR lamp into a three-band spectrum. While this filter would achieve high radiant luminous efficacy (lm/W(r)), the luminous efficacy of the lighting system in terms of lumens per watt of electrical power input would be poor. It must be recognized that whenever a filter is added to a lamp, it is necessary to increase the lamp wattage to maintain the illuminance. A more efficient approach would be to develop a new lamp type specifically for museum applications in which a three-band spectrum is generated by efficient conversion of electrical power. While this latter approach offers the prospect of a superior solution, the development costs are likely to be much higher.

It should not be presumed that the museum community will respond with enthusiasm to this initiative. There is a long history of museum directors, particularly art museum directors, insisting that natural light is the only true light for the museum experience. The distinctly unnatural spectral power distribution of the three-band source favored by this study is likely to be regarded with strong suspicion, whatever the visible difference in the lighting. A further study in a real art gallery directed toward gaining critical evaluations of museum professionals is recommended.