ABSTRACT
Can a human grammar exist without words? The fundamental question we address here is whether syntax, a core aspect of the highly integrated system of universal properties that form the grammar of any human language, is subserved by a language-specific neural organization or not (Hauser et al., 2002; Marcus et al., 2003; Tettamanti et al., 2002; Tettamanti and Weniger, 2006). In all natural languages the syntactic dependencies are established on the basis of the hierarchical phrase structure generated by recursive rules rather than by the linear order of words (Chomsky, 1957). The conclusion that syntactic dependencies are based on recursive phrase structure is corroborated by the fact that there is no human language where a certain word must occur at a fixed distance from another (Chomsky, 1956). Distances between words are specified in terms of relative position, and they can always be recursively expanded. For example, within noun phrases, the distance between an adjective and a noun can vary, such as in “a horrible flu,” “a horrible nasty flu,” “a horrible nasty bad flu.” Of course, this does not apply to idioms, such as in “kick the bucket” where “bucket” always follows “kick” after one word. But notice that, in this case, the complex “kick the bucket” is arguably not formed via a recursive procedure. In fact, it is invisible to transformations, such as passive: *“the bucket was kicked by John.” Thus, idioms can be considered on par with compounds, where the words forming the compound are displayed at a fixed order, such as in “blackboard.” Therefore, “non-rigid” syntactic dependencies (NRSD)— that is, syntactic rules established between words at varying positions— constitute the core type of dependencies found in the syntax of all natural languages. Their counterparts, “rigid” syntactic dependencies (RSD)—that is, syntactic rules established between words at fixed positions—are never found in human languages (Figure 16.1A). Syntactic dependencies with and without words. (A) There is no human language where a certain word must occur at a fixed distance from another. Distances between words are specified in terms of relative position and they can always be recursively expanded. So for example, when if occurs in the string, the sentence will contain the word then but the position of the two words cannot be fixed (cf. the underlined words in (i) versus (iii): RSD); rather, it must vary according to the structure and lexical choices of the intermediate words (cf. (i) versus (ii): NRSD). (B) Syntactic strings made of symbols were construed mimicking both RSD (symbol-RSD) and NRSD (symbol-NRSD), as highlighted here for the sake of clarity by the underlined symbols. Symbols are labeled here with their names (k1, k2,. k10) only for cross-reference with Section 2. Underlines and labels did not appear in the actual experiment. Different colors and different sizes (small vs large) allowed to establish agreement. Stimuli were symmetrically reproducing the contrasts in (A): strings of symbol-NRSD contained concordant elements at varying positions (example strings NRi, NRii, and NRiii following rule symbol-NRSD-1, left), while strings of symbol-RSD contained concordant elements at fixed positions (example strings Ri, Rii, and Riii following rule symbol-RSD-1, right). (C) fMRI task design: as exemplified here by rule symbol-NRSD-1, task sessions started with a sample block consisting of only correct (white boxes) strings, followed by a probe block consisting of randomly intermixed correct and incorrect (gray boxes) strings. As shown in the right frame, string presentation terminated by button press, after which a feedback indicating right or wrong answer appeared, followed by a variable time interval. https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9780203083246/fa648053-feeb-4da6-b12a-76dd7e75b888/content/fig16_01_C.jpg" xmlns:xlink="https://www.w3.org/1999/xlink"/>
