Neural representation of minimal syntactic units

How and why humans can create and understand an infinite number of novel sentence remains a linguistic mystery, especially given the number and diversity of languages. Despite the apparent complexity of the problem, Generative linguists claim that the answer can be reduced to a single, simply defined and simply implemented function, Merge (Chomsky, 1995). Merge takes two syntactic objects (e.g., words) and joins them, forming a larger syntactic object, called a constituent. Merge operates iteratively, either applying to yet unmerged items or to the product of previous applications of Merge, building hierarchical, recursive structures (Chomsky, 2001). Merge is argued to be category-neutral, such that the derivation of noun phrases (NPs) is identical to the derivation of verb phrases (VPs), and so on. Little evidence comes from studies on language processing and brain function. Instead, studies of neural processing of syntax has focused on largescale sentential phenomena, involving several applications of Merge, and further processing requirements. Studies implicating Merge indicate that constituents of different sizes elicit activation in left inferior frontal gyrus (LIFG) during comprehension (Pallier, Devauchelle, & Dehaene, 2011) and production (Indefrey et al., 2001; Indefrey, Hellwig, Herzog, Seitz, & Hagoort, 2004). However, direct evidence for Merge is scarce: Bemis and Pylkkänen (2011) looked at effects of combining adjectives with nouns to form NPs, finding significant activation in left anterior temporal lobe (LATL). However, their design made it difficult to tease apart contributions of syntactic operations (i.e., Merge) and semantic composition (i.e., how meanings of the words combine). Zaccarella and Friederici (2015) report that Merge is processed in a small cluster within BA44. They use minimal compositions (i.e., two-word phrases) and two-word word lists, as well as pseudo-words to avoid effects of semantic composition (Bemis & Pylkkänen, 2011; Humphries, Binder, Medler, & Liebenthal, 2006; Pallier et al., 2011). However, they only investigated NPs. Here, we attempt to replicate Z&F’s findings across 3 further categories: Verb Phrases (VPs), Adjective Phrases (APs), and Prepositional Phrases (PPs) (:XPs). However, our aim is not just to see whether constituent XPs all engage the same (sub)regions of LIFG, but also to address whether the processing of each category is identical. That is, Merge is theoretically not category-specific, so we should not find category-specific patterns of activation. While we test Z&F’s claim that Merge is localized in Broca’s Area, our focus here is not to expand knowledge of the region, but a finer, neurological characterization of constituency distinctions and lexical categories.