Tables 11, 12, 13, and 14 show the segmentation results of four types of compound stimuli: transparent compounds (TT), two types of partially transparent compounds (TO and OT), and opaque compounds (OO).
Then, we checked the segmentation results of these compounds to see if the transparent compounds have higher probability to be explicitly segmented into two units than the opaque compounds.
Larger MMN for high vs. low-frequency opaque compounds.
The larger N400 for transparent compared to opaque compounds replicates previous findings using German compounds (Koester et al., 2007), suggesting combinatorial processing for transparent but not opaque compounds.
The source of the difference between pseudo-compounds and low frequency opaque compounds was also bilateral.
For opaque compounds only, there was a larger MMN for high frequency than low frequency compounds.
Opaque compounds elicited an enhanced 'lexical' MMN, reflecting stronger lexical representations, to high- vs. low-frequency compounds.
Transparent compounds elicited a larger N400 than opaque compounds suggesting processing via a combinatorial mechanism (Koester et al., 2007).
By contrast, the meaning of opaque compounds cannot be derived from their constituents (e.g. framework, strawman) and thus may require dedicated whole-form lexical storage.
For opaque compounds, the direct access route is dominant (combinatorial processing will fail and access to individual constituents will be suppressed).
