Most modern speech recognition uses probabilistic models to interpret a sequence of sounds. Hidden Markov models, in particular, are used to recognize words. The same techniques have been adapted to find domains in protein sequences of amino acids. To increase word accuracy in speech recognition, language models are used to capture the information that certain word combinations are more likely than others, thus improving detection based on context. However, to date, these context techniques have not been applied to protein domain discovery. Here we show that the application of statistical language modeling methods can significantly enhance domain recognition in protein sequences. As an example, we discover an unannotated Tf_Otx Pfam domain on the cone rod homeobox protein, which suggests a possible mechanism for how the V242M mutation on this protein causes cone-rod dystrophy.