Abstract
In my final year as a pre-clinical medical student at Cambridge in 1943 I became fascinated by the recently published paper by Hecht, Shlaer & Pirenne on the absolute threshold of human vision. It seemed to show, with very little room for doubt, that the human visual system was able to make very good decisions as to whether a weak flash of light had or had not just been delivered, and to do this it had to be using nearly all the statistical information available in the fluctuating messages from the retina to the brain. I did not understand some of the statistical arguments in this, so on William Rushton's advice I borrowed R.A. Fisher's Statistics for Research Workers from a friend. I told him how incredibly enlightening I found it, and he replied that he thought it was utterly useless and did not want to see the book ever again, so I still refer frequently to his now-tattered copy. My work was inspired by by Fisher's book, Hecht's paper, and also by Albert Rose's contemporaneous work on the quantum efficiency of vision and by Shannon's information thereoms. I have contributed facts and ideas demonstrating (I hope) that noise limits vision, whether the task is to detect photo-isomerisation of rhodopsin molecules against a background of thermal isomerisations, or to detect the autocorrelation resulting from Glass pairing in random dot patterns against the chance autocorrelation of spurious random pairs, and I shall talk about some of this work. Much of the genius of vision lies in the mechanisms that near-optimally collect relevant signals while keeping irrelevant noise to a minimum.