Naho Ikuta, Jungho Kim, and Masatoshi Koizumi

The COE Program in Humanities, Tohoku University, Sendai

Scrambling (or word order alternation) in "free word order languages" such as Japanese and German has generated a considerable literature in theoretical linguistics and psycholinguistics (Karimi ed., 2003). However, few imaging studies have been performed of the processing of non-canonical (or scrambled) word order (Friederici et al., 2003). Using functional magnetic resonance imaging (fMRI), we investigated the effects of scrambling on brain activation by directly comparing the brain regions involved in the processing of Japanese transitive sentences with canonical (Subject-Object-Verb) and scrambled (OSV) word order.

In this study, we performed the following two experiments on canonical and scrambled sentences. In the first experiment, we examined brain activation involved in the processing of the whole sentence. In the next experiment, we examined the changes of activated regions as the sentence unfolded.

Experiment 1:

In Canonical and Scrambled conditions, grammatical sentences with canonical and non-canonical (scrambled) order were visually presented at the center of a screen phrase by phrase. In either condition, half of the sentences were semantically plausible, and the others were semantically anomalous. Participants were instructed to judge whether or not the sentences they just read made sense by pressing one of the two buttons ('Yes' and 'No') with their right hands.

In the Canonical and Scrambled conditions (each compared with the Rest condition), similar regions were significantly activated, including Broca's, Wernicke's, premotor and visual areas. This suggests that most cognitive processes involved in the comprehension of scrambled sentences are also responsible for the comprehension of canonical sentences. When these conditions were directly compared (Scrambled minus Canonical), we found activated areas in the left middle frontal gyrus and the left inferior frontal gyrus, which have been claimed to be selectively involved in syntactic processing (Hashimoto and Sakai, 2002).

Experiment 2:

Each SOV and OSV sentence was divided into the subject, the object and the transitive verb. They were presented one by one at regular intervals as visual stimuli. Words were also presented one by one at the same intervals as sentences. These words consisted of nouns that were the same nouns as those in subjects and objects and transitive verbs that were the same as transitive verbs in sentences. These words functioned as the control tasks. The participants read the presented sentences and words silently.

Subject of SOV sentences activated the same region as object of OSV sentences (left lingual gyrus). Object of SOV sentences activated the same region as subject of OSV sentences (left inferior frontal gyrus). SOV sentences and OSV sentences were different in regions activated when transitive verb was presented (SOV-left inferior temporal gyrus, OSV-left putamen and right middle frontal gyrus).

Therefore, the activated region common to the two experiments was left inferior frontal gyrus, including Broca’s area. This region was activated when reading the middle of a sentence in the experiment 2.

Based on the results of these experiments, we suggest that left inferior frontal gyrus is crucially involved in the processing of scrambled sentences. This finding is consistent with the observation that Broca's aphasics have difficulty comprehending scrambled sentences (Hagiwara and Caplan, 1990).