Japanese-style gardens offer a wealth of spiritual and cultural value. In this study, we attempt to determine the optimum colour temperature for lighting Japanese-style gardens at night. We analyse the influence of a change in light source colour temperature on image recognition using digitalization and quantification with the semantic differential technique. In addition, we propose a new evaluation methodology for the semantic differential technique and examine the statistical significance of the results. We find that in summer, impressions of the optimum colour temperature for each element (vegetation, water, stone and structure) differ significantly, but in winter, because of the effects of snow, differences in the impression of each colour temperature are not as distinct. Moreover, the colour of the natural environment or overall season affects peoples’ preferred light source colour temperature.