QUOTE : It will be agreed that you can’t divide a cake up into its component crumbs and say ‘This crumb corresponds to the first word in the recipe, this crumb corresponds to the second word in the recipe’, etc. In this sense it will be agreed that the whole recipe maps onto the whole cake. But now suppose we change one word in the recipe; for instance, suppose ‘baking-powder’ is deleted or is changed to ‘yeast’. We bake 100 cakes according to the new version of the recipe, and 100 cakes according to the old version of the recipe. There is a key difference between the two sets of 100 cakes, and this difference is due to a one-word difference in the recipes. Although there is no one-to-one mapping from word to crumb of cake, there is one-to-one mapping from word difference to whole-cake difference. ‘Baking-powder’ does not correspond to any particular part of the cake: its influence affects the rising, and hence the final shape, of the whole cake. If ‘baking-powder’ is deleted, or replaced by ‘flour’, the cake will not rise. If it is replaced by ‘yeast’, the “cake will rise but it will taste more like bread. There will be a reliable, identifiable difference between cakes baked according to the original version and the ‘mutated’ versions of the recipe, even though there is no particular ‘bit’ of any cake that corresponds to the words in question. This is a good analogy for what happens when a gene mutates.