Nature offers a bewildering diversity of flower colours. Understanding the ecology and evolution of this fantastic floral diversity requires knowledge about the visual systems of their natural observers, such as insect pollinators. The key question is how flower colour and pattern can be measured and represented to characterise the signals that are relevant to pollinators. A common way to interpret flower colours is using animal vision models that incorporate the spectral sensitivity of a focal observer (e.g. bees). These vision models provide a measure of colour contrast, which represents the perceived chromatic difference between two objects, such as a yellow flower against green leaves. Colour contrast is a behaviourally and physiologically validated proxy for relative conspicuousness of a stimulus. A growing number of studies attempt to interpret flower colouration through parameters that are grafted on to principles of human colour perception. A perpetuating measure to describe floral colours is via saturation, which is a metric in human perception describing a certain aspect of colourfulness and is, in pollination literature, often referred to as 'spectral purity'. We caution against the concept, calculation and biological interpretation of 'spectral purity' and similar measures that rest on an anthropocentric view, because it does not represent the diversity and complexity of animal visual systems that are the natural observers of flowers. We here discuss the strengths and weaknesses of common ways to interpret flower colouration and provide concrete suggestions for future colourful research.