We propose an experiment to rate the merit of four algorithms in achieving satisfactory tone-mapping. The appearance of a unique scene including luminance gradations and a wide distribution of luminance patches has been evaluated in both real and simulated situations by the same observers.The real scene consisted of a wide wall receiving controlled illumination. The test consisted of two horizontal gradations of grey with different gamma values, embedded in an achromatic noise background of high spatial frequency. Each observer was invited to choose the gradation he found “optimal”. The simulation was produced on a calibrated CRT display. Four tonemapping algorithms were implemented, three of which were linear, to render the simulated conditions.With the real scene, observers are able to judge accurately which gradation is the best representative of the optimal gamma. Under examination of the distribution of preferred choice around the optimal gamma, it seems that the rating of gamma values is about symmetric on a logarithmic gamma scale.The hypothesis is that a tone-mapping algorithm which performs well should yield the same optimal gamma as in the reality. After our experiments, it appears that the four algorithms which were tested fall in two classes, either under- or over-estimating the gamma values. Despite inter-observer variability, observers agree on their judgement.