English undergraduate students, studying in London, consistently misspelled Gandhi as Ghandi despite intensive exposure to the correct spelling of the name. This is an exemplary misspelling in a number of ways that we detail in this paper. We show that statistical similarity to English predicts the misspelling but that statistical similarity to English cannot override 'rule-making' pressures in spelling pseudo-Indian names, for Ghandi-spellers also spelled the pseudo-name Ghalgi, while Gandhi-spellers did not, and this spelling is not predicted by statistical similarity alone. We conclude that, even with minuscule lexical knowledge of 'Indian' words and names, English readers use 'rules' in such tasks. This gives us hope that, once a correct spelling has been achieved it will be maintained, for new rules (presumably) replace old ones.