Early automatic processing of lexical word accents: Valid words are stored holistically irrespective of stem tones

According to the dual route model of word processing (Pinker & Prince, 1991),
morphologically irregular and frequent words are stored in the lexical memory as full-form neural representations. In contrast, regular and infrequent words are decomposed into morphemes. So far, it has been unclear whether a tonal element―a lexical word accent―is treated by the brain as a decomposable component or an integral part of the word’s full form.
We recorded ERP responses in 17 native speakers of Swedish (6 males, mean age: 23.2 years, SD = 4.5) to Swedish words with valid and invalid combinations of stem tone and suffix. These stimuli consisted of four words―two valid and
two invalid combinations of stem tone and suffix (stem "krok", tone 1 and 2, suffixes "-en" and "-ar") presented as standards and deviants in four blocks, where validity and suffix varied orthogonally in a fully counterbalanced fashion.
The results indicate a full-form lexical retrieval of valid Swedish words with both word accents. This process appears to have two stages that engage different brain areas: first the right auditory cortex, where, likely, the pitch contour of the word accent is established, and then the left lexical memory area, where the memory trace of the appropriate (that is, bearing the correct word accent) lexical unit could be activated. Moreover, these results for the first time show that word accents are used in early automatic lexical retrieval.