In a thoughtful discussion at The History Girls, A.L. Berridge ponders the difficulties in a writer's attributing words to her characters in historical fiction:
My current book is set during the Crimean War, and during a recent visit to Sevastopol I was introduced to a young man visiting the spot where his great, great grandfather had stood during the Battle of Inkerman. His pride in his ancestor was both obvious and justified (Colonel Henry Percy won the VC) and I [c]ouldn't help a slight qualm at the recollection I'd been putting dialogue into his mouth just two days before. I'd written nothing disrespectful, but coming face-to-face with a real flesh and blood descendant I had an uneasy sense of having taken a liberty.
But how could she not have taken that sort of liberty? I remember being troubled for a while, when I was reading Colm Tóibín's The Master, about what in the novel was securely based on fact and what - of the conversations and the thoughts ascribed by Tóibín to Henry James - was by way of probabilistic hypothesis. But it doesn't matter, I decided - provided that we remember it's a novel, that nothing of it can be taken as factual which can't be corroborated as being such from other, documentary, sources.