Experimental evidence for the influence of structure and meaning on linear order in the noun phrase

hierarchical representations, encoding relations among linguistic sub-units, are a key feature of many theories of syntax and semantics. In the noun phrase, such representations have been argued to have a universal form, in which Adjectives are structurally closest to the Noun, then Numerals, then Demonstratives (Cinque 2005; Abels & Neeleman 2012). This hierarchy can explain (among other things) a striking asymmetry in the frequency of noun phrase word order patterns among the world’s languages: in the vast majority, Adjectives are linearly closer to the Noun, and Demonstratives farthest away (e.g., N-Adj-Num-Dem or Dem-Num-Adj-N). In other words, there appears to be a transparent mapping between the hypothesised structure and linear order in most languages. Building on previous research using artificial language learning experiments, here we aimed to provide robust experimental evidence that human learners are biased in favour of such transparent mappings. We showed that English speakers taught a new language in which modifiers were postnominal (unlike English), used their knowledge of the underlying structure to infer linear order. For example, they assumed that the language would have N-AdjDem order rather than N-Dem-Adj order. The latter is more similar to the surface order of English, but the former transparently reflects an underlying structure in which Adj is structurally closer to the Noun than Dem. Importantly, these findings provide stronger evidence than those reported in previous research (e.g., Culbertson & Adger 2014), in which a potential confound—use of a metalinguistic flipping strategy—could have explained learners’ preferences. We also found that the strength of the preferences we found in our experiments correlated with structural distance; learners’ preferences for the relative order of Adj and Dem were very strong, while for Adj and Num and Num and Dem they were relatively weaker. These results show that, when learning a new language, learners are sensitive to the detailed hierarchical structure of the noun phrase rather than the linear order of their native language. Our results also point to the need for future work exploring whether this structure is learned from experience with a particular language, or reflects a universal underlying representation that shapes noun phrase word order typology.