Phoneme model construction

To account for response variance caused by the low-level phonemic content of the stories, a 39-parameter model that captures how often each of the 39 phonemes in English was spoken over time was constructed. The phoneme representations of the stories were used to construct this model: the lists of phoneme–time pairs (P, t) were rearranged into 39 lists, each of which contains only the times of a single phoneme. These lists of times were then down-sampled to the fMRI acquisition rate (2.0045 s).