Text-to-speech systems traditionally use a pronouncing dictionary from which the phonemic representation of an input word is retrieved. In the case of a 'novel' word absent from the dictionary, letter-to-phoneme rules are invoked to produce a pronunciation. A proposal in the psychological literature, however, is that human readers pronounce novel words (or 'pseudowords' conforming to the spelling patterns of the language) not using explicit rules, but by analogy with letter/phoneme patterns for words they already know. We are developing a synthesis-by-analogy system which is, accordingly, also a model of novel-word pronunciation by humans. We investigate the computational methods of assessing the orthographic analogy module and the 'flexible' (context-independent) GPC rule module, which produces the set of plausible, candidate pronunciations phonemic analogy requires. We compare the resultant assessments across language, method of assessment, size and content of the lexical database. The investigations into these modules produced useful results for both British English and German. However, the best results for each of the two languages were obtained from rather different detailed implementations.