When the light reflected from an object differs in spectral composition to the surrounding background these spectral differences are reflected in the excitation levels produced in each class of photoreceptor. The ability to see colours and to notice small colour differences is strongly affected by both the spectral composition and luminance level of the adapting light. Knowledge of the limits of colour detection is important in setting safety standards and guidelines in visually demanding workplaces, as varying conditions of illumination and chromatic adaptation are often encountered in different working environments. It is therefore of both fundamental and practical interest to be able to predict accurately how a human observer's chromatic detection performance changes with both light level and chromatic adaptation. The Colour Assessment and Diagnosis (CAD) test was employed to measure colour detection threshold ellipses under different states of chromatic adaptation and background light levels. The advantage of this new technique is that it isolates the use of colour signals by embedding the isoluminant chromatic stimulus in dynamic luminance noise. These measured threshold variations were analysed in terms of changes in L-, Mand S-cone excitation levels required for threshold in different colour directions. Models based on the measured chromatic threshold data are proposed that are capable of reconstructing entire detection ellipses. These models were based on experiments where observers had colour thresholds measured around a series of different chromatic adaptation points, over a range of light levels (typically from 0.3 to 31 cd m-2), and additionally away from the adaptation point. The findings reveal the independent adaptation states of individual cone classes on measured thresholds, i.e., the threshold in a given cone class depends only on the signal produced by the background in that cone class and is independent of the adaptation state of the other cone classes and hence independent of chromaticity and light level. The effect of adapting different areas of the peripheral retina when thresholds are measured foveally was also investigated. No long range retinal interactions were observed. The results show that the adaptation state of the periphery has no effect on colour detection thresholds made in central vision. Variations in L-, M- and S-cone contrasts curves were simulated to assess the influence that detection ellipse size and ellipse orientation have on them. This revealed a correlation between the L-cone contrast curve gradient and the corresponding ellipse orientation. This was additionally shown to correlate with the central 2.8o mean value of macular pigment optical density, hence providing a new method of estimating macular pigment level from colour detection ellipses. Steady state pupil sizes were analysed with the rod and cone excitations that produced them. These data indicate that when chromatically adapted, the steady state pupil size correlates strongest with the S-cone signal, and is independent of the actual chromaticity and luminance levels involved. Cone signal-to-noise ratios were extracted from repeated threshold measurements for a series of colour directions. Analysis of these revealed the existence of a constant signal-to-noise ratios over the full range of colour directions tested relative to a whitish background. The results show that as the cone contrast level increases in a particular cone class so does the associated noise.