Currently there are two theories as to how Brunelleschi arrived at his rules of perspective, which deserve consideration. However, I have a third which I believe is preferable. David Hockney has demonstrated that Brunelleschi could have used a lens to throw the image of the Baptistry onto a surface on which it could then be traced. I could accept this theory if it were not for the evidence of Manetti, a friend of Brunelleschi's, who had “held the painting (of the Baptistery) in his hands and seen it many times”. Hockney's theory takes no account of this well-known text. My own theory starts and follows from Manetti's text.

The second theory is Prof. Martin Kemp's preferred theory. He actually lists six possibilities in his The Science of Art. (Hockney's was published after his book and was therefore not included). My two objections to Kemp are he cannot explain how it could be done and second he does not to take into account that there were fairly satisfactory versions of perspective in use at the time that Brunelleschi made his first demonstration. Therefore, Brunelleschi had to start from an image that would not be questioned: a mirror image. In fact, Brunellschi had to make a second demonstration years later to persuade people that his science was better than Giotto or Martini's observation.

Giotto and the ancients had a pragmatic means of conveying the space in a building and furniture within it based on it's appearance. This was done by paying particular attention to the two dimensional shape of the objects and of the spaces between the objects. Ambrogio Lorenzetti and Simone Martini refined this perspective around 1330 so that their complete picture space was remarkably logical. Kemp suggests that this perspective was used by Donatello as late as 1417 in his St. George relief. That is 16 years after Brunelleschi had made his first demonstration. What makes this more surprising is Donatello and Brunelleschi were good friends. Furthermore, Brunelleschi's new perspective did not gain traction in Siena til the 1450's. Clearly the first demonstration was too complicated to follow, or perhaps it smacked too much of wizardry. (A second mirror had to be introduced to right the reversal of his image painted on the silver mirror in order to compare it with the Baptistry.)

I proposed the solution that Kemp least favours of the six he published in a lecture at the Slade c. 1978. (I am not sure that I was the first to suggest it, though I found it independently). My theory is that Brunelleschi's first demonstration was painted on a burnished silver mirror. Manetti's description says that “Filippo put burnished silver in the sky” but the reason he gives for this strange behaviour “so that the clouds would be seen to be moved by the winds that blow” seems scarcely credible for a man of Brunelleschi's scientific bent. I suggest that the Baptistery and surrounds were painted over the silver but Brunelleschi left the sky unpainted either because the sky did not form a part of his idea or simply to explain his method. Brunelleschi started from an image of reality that all could accept. A mirror image is acceptable apart from its reversal. A construction with survey instruments - suggested by Prof. Kemp could not possibly be more persuasive and Kemp cannot explain how it might be done.

The use of silver as the mirror is easily explained: Brunelleschi was a goldsmith. Polished silver was the best reflector available and had the further advantage over glass that one could scribe directly on its surface with precision. The reflective surface could be touched without the thickness of glass intervening as in a glass mirror.

Kemp's main objection to this “simple” solution is that one's own image would blot out most of the Baptistery but this is not so. Once one has fixed the eye-piece and the mirror one is free to move, to use either eye, and to look through the eye-piece with one's head at whatever angle is convenient. By this method the size of the blot is no bigger than one centimetre in radius, that is the distance between the pupil of the eye and the outer skull as reflected. This would not have been the slightest impediment as it need not blot out the outer corners of the Baptistery which are the necessary points to fix the image. We can be certain that Brunelleschi would not have gone further with the work before analysing the geometry. The paint would have only been applied to clarify the image for others to appreciate how very like it was to the original. Naturally, it needed to be turned the right way round. This is why the peep-hole and second mirror were introduced. This method is not only the simplest as Kemp admits, it fits Manetti's description in all but the moving clouds. Neither Kemp nor Manetti seem to have understood the need for the second mirror in righting the reversed image traced on the silver mirror.

Kemp's further objections were:

1. That Manetti could not have been mistaken in thinking that the silver was applied to the sky area. But clearly if the rest had been painted over why should he not have made this mistake? On the occasion on which he held it in his hand we may presume he was preoccupied with seeing the image in the second mirror.
2. Kemp goes on - it does not justify Manetti's claim for Brunelleschi's geometry. But it does of course require analytical geometry to arrive at the rules of perspective from the scribed image.
3. The second of Brunelleschi's demonstrations did not require mirrors; adds Kemp. No, that was why its straightforwardness finally convinced the doubters. It was presumably arrived at through the perspective geometry derived from the first demonstration.

The strange way of viewing Brunelleschi's first experiment in perspective (with a second mirror) is easily explained by my film on YouTube, which follows this script.

(see A Documentary History of Art vol. I p 171-2, Anchor Books 1957 for Manetti's whole text in English)