The experiments detailed in this study show three important results with respect to the use of printed colour charts to identify the colours of animal signals. Firstly, people vary with respect to the choices they make when asked to match one colour to the perceived closest match from a set of colours on a chart. This means that there may be differences in colour matches made by different individuals. These differences may, to some extent, be reduced by using high-quality charts with a greater range of colour matching options. Secondly, there were very large and significant differences in the matching choices made by individuals between charts produced from different printers. Different inks have significantly different spectral properties [30
]. This is a critical problem of 'self-made' charts, such as those made from RGB colour space. The aim of matching a colour signal to a specific section of a colour chart is that the chromatic content of the signal can be recorded, such as in terms of an RGB value. However, when charts with identical RGB values are produced from different printers these do not have the same properties, making comparisons to an RGB value irrelevant. Even if charts are printed from the same printer, with the same cartridge model and paper type, the exact values of the printed charts will still vary depending upon the toner levels (Cuthill & Stevens unpublished data). This means that complex printer calibrations are needed to ensure that the reproduction of more than one colour chart is accurate and invariable with respect to the chart properties. Furthermore, there still remains the problem that a linear increase in a colour value on a chart may not be linear in terms of the measured reflectance spectra and the perceived difference in colour, for many, if not all colour spaces. Thirdly, and perhaps least expected, there was the suggestion that colour matching results were significantly affected by the irradiant light conditions (non-significant for red charts, at P = 0.06, but in a significant interaction with printer type for the blue chart). For the red charts, the largest mean difference for judgements measured under different light sources was a difference of 14 pixel values on an 8-bit scale. For the blue-green charts, the effect of light varied between charts from different printers: for one printer the average difference in pixel values of best matches was 29, for another it was only 5. That the effects of lighting were modest was expected because humans possess colour constancy, where the visual system is capable of maintaining a constant appearance of colour quasi-independently of changes in the irradiant light. Presumably many other animals also show colour constancy [33
]. However, the adapting mechanisms are not perfect [38
] and changes in the irradiant light do have some effect on colour constancy. Different environments can vary significantly in their irradiant light characteristics [41
], and thus influence colour appearance.
Whilst in a natural situation, the perception of colours under different conditions will be about the same (i.e. a light red signal will always look light red), in terms of quantitative, or even qualitative scientific experiments, changes in perception may significantly impact upon results, sometimes by large degrees. Results in the field will be affected by the time of day, the weather, and the natural environment [6
], and will also differ under various laboratory lighting conditions. Finally, charts based on RGB colour space, even if used in studies of human vision, are incapable of reproducing the full range of colours perceptible to humans. The use of charts based on CIE data to estimate colours are better than the use of charts based on RGB colour space, but still unfortunately are based upon human subjective assessment.
The spectrophotometry analysis further supports the argument that printed charts could produce seriously inaccurate colour matching results. Firstly, different printers vary significantly in the reflectance properties of the colour charts that they produce. This means that comparisons between different printers will be unreliable, even discounting the effects of toner level. More seriously however, is the result that some printers do not show even gradations in reflectance between colour blocks with an even spacing in RGB colour space. Therefore, comparing the colours of animal signals between individuals (for example) via charts to obtain an R, G or B values could be seriously flawed – made worse when considering the non-linearity of RGB space in terms of visual perception.
Finally, as stated above, there are crucial differences between the visual perceptions of humans and non-human animals. The perception of a given colour signal to a human may be markedly different from that of the animal towards which the signal is directed. The fact that colour charts have numerous errors associated with them, especially self-made charts, in terms of human judgement, only emphasises the inadequacy of the method when used with respect to non-human animals. The fact remains that other animals' perceptions of a signal may drastically differ from our own.