Hello,

those are two different tasks. In lexical-semantic access you are f.e. showing your proband an object measuring how fast your proband is able to find a proper word for it. While phrase structure building you would rather test showing your proband a sentence and asking if the structure is right or wrong. They are not so much like components of one thing which fit each other.

Dear Adrià Rofes, 

the timing estimates for the processing of syntactic word category information in Friederici's various papers is largely based on her ELAN studies which have used problematic experimental designs and questionable approaches to data analysis. John Drury and I have addressed these issues in our 2012 paper (see reference below; I've also attached the pdf for your convenience).  There is an increasing number of (better controlled) studies from various labs (and a range of languages) showing that semantic integration is not generally delayed until the word category has been confirmed (as was most strongly claimed in Friederici, Gunter, Hahne and Mauth, 2004). For example, van den Brink and Hagoort (2004) demonstrated N400 effects that preceded any kind of syntactic word category violation (or phrase structure violation) effects in Dutch, (we have shown the same in an English study, both in reading and in auditory processing; e.g., Pauker & Steinhauer, 2009). Moreover, the "semantic blocking effect" (i.e., the absence of N400s on words that violate the phrase structure, which was interpreted as additional evidence for syntax-first models) turned out to be difficult to replicate even with the original German sentence materials. Although I am a co-author of the first ERP study reporting a 'semantic blocking' effect (Friederici, Steinhauer & Frisch, 1999),  I am not convinced of our original 1999 interpretation anymore.  A PhD student in my lab and I have recently analyzed ERP data from more than 150 German speakers for these violations, and virtually all of them elicited enhanced N400s for 'double violation conditions' (compared to the control condition).  Our current data suggest that the apparent 'semantic blocking' effect in the 1999 paper may have been partly due to the 'probe verification task' employed in that study.

In sum, the strong claim made by syntax-first models that phrase-structural analyses always precede semantic processing (including those processes reflected by N400s) is most likely wrong. This, however, does not necessarily imply that constraint-satisfaction models a la MacDonald et al. (1994) must be right.

I hope this helps. Please let me know if you need more details.

Karsten 

Steinhauer, K. & Drury, J.E. (2012). On the early left-anterior negativity (ELAN) in syntax studies. Brain and Language. 120 (2), 135-162.