Experimental approaches to syntactic optionality

One of the most remarkable properties of humans is our complex language: a tool for communicating thoughts and ideas of a potentially endless richness and complexity. People around the world and from all cultures spend several hours each day engaged in linguistic activities, notably speaking to friends and family, with impressive ease and fluency. Levinson and Torreira (2015) estimate that we produce an average of 1200 utterances daily in conversation, with a speed of at least 10 speech sounds per second, organized into 1-2 words and several meaning-bearing units, called morphemes. This all seems to happen effortlessly; conversation analyses show that a speaker needs no more than 200 milliseconds to initiate an utterance after their conversation partner, easily stringing together 10-20 meaning components into coherent utterances. The core objective of linguistics is to understand how we can produce and comprehend complex linguistic structures at this speed. To do so, we need an explicit model of the speaker's mental lexicon — the storage of language building blocks (sounds, morphemes, words), and internal grammar -- the rules for combining these elements into larger utterances. Most importantly, we need a model of how the brain applies this grammar to the lexicon in real-time to derive meaning from utterances. In the ExSynOp project, we focus on one aspect of grammar: word order. The order of words is surprisingly rigid within any given language. For instance, consider the simple sentence (1) "Today I have already drunk three cups of coffee." This sentence contains nine words, yielding a total of 362880 (the factorial of 9) logically possible arrangements. Although there are alternative configurations that convey a similar message, altering the word order usually results in either slightly different meanings or ill-formed sentences, such as "Today drunk I have already three cups of coffee." While word order carries relatively little information load, listeners typically have no difficulty understanding sentences with non-standard order, as context and common sense guide interpretation. Despite this, languages strongly favor a deterministic mapping from meaning to word order. The goal of the ExSynOp project is to uncover the reasons behind this tendency: what drives the development of a strict mapping from meaning representations to linear orders? We begin by examining exceptions to this tendency, particularly instances of true optionality. For example, in Swedish, adult speakers alternate between two orders, such as "Igår tvättade bakaren sig efter jobbet" and "Igår tvättade sig bakaren efter jobbet" ("Yesterday, the baker washed {himself} after work"), concerning the placement of the reflexive object. This optionality is absent in the closely related language Danish. To understand the scarcity of optionality in human language, we compare closely related languages that have developed either strict or variable mappings from meaning to order. This allows us to explore whether language acquisition and processing are facilitated by strict mappings from meaning to form. In the first part of the project, we conducted studies targeting the processing of variation and found little to no effect of variability on production or comprehension. In the second part, we focused on populations exhibiting what we term non-target-like optionality — specifically, speakers likely to make errors in word order production. The two populations examined are young first language learners (children and teenagers) and adult second language learners. In both groups, we observe a learning process characterized by a general decline in optionality. For instance, in young speakers from Trøndelag, we track how the placement of syntactic subjects relative to verb particles becomes stricter over time. Initially, we find variation between the target-like "I går gikk Peter ut" and the non-target-like "I går gikk ut Peter" (lit. "Yesterday, John went out"). This variation persists until at least age 16, where non-target-like structures tend to disappear. In adult second-language learners of Norwegian, we observe no variation in subject placement regarding verb particles, yet find variability in other domains, such as reflexive object placement, mirroring Swedish variations. Again, this variability in word order gradually diminishes. Unlike second-language learners of Swedish, who may increase variability to match the target language. By studying both the development of these variations and their final outcomes, we can model how optionality decreases — showing how speakers become adept at finding the target word order among the 362880 possible arrangements for a nine-word sentence. This will increase our understanding of how word order restrictions are encoded within grammar, which is necessary for understanding the motivations behind rigid mappings from meaning to form in grammars.

The natural languages of the world are characterized by largely rigid, often highly idiosyncratic rule systems. Within any given language, it is hard to find instances of true optionality (i.e. a many-to-one mapping from form to meaning). The ExSynOp project sets out to explore this central puzzle: what drives the evolution of complex, rigid rule systems in natural languages? We do this by investigating the sources of regularisation, i.e., the reduction of variation in a language. Our focus on regularisation will bring new insights to one of the most debated issues in linguistics: are languages shaped predominantly by the usage patterns of adults (evident in processing, register/style choices, MacDonald 2013, Bybee 2015), or the learning preferences/limitations of children (Clark 1987, Newport 2005, Yang 2017, Chomsky 1986)? This project takes as its starting point a unique naturally-occurring case study in (apparent) optionality: word order variation in the closely related Mainland Scandinavian languages and varieties. We have identified four word order variables where at least one variety shows optionality, and another a strict rule: subject shift, particle shift, object shift and long object shift. By studying acquisition (L1 child and L2 adult), and processing (production/comprehension) of variable and non-variable grammars in the context of closely related languages, we can identify where preferences for regular systems arise. We will address three fundamental issues in the establishment of rigid grammars: (1) are there processing benefits (or costs) associated with categorical rules; (2) is the L1 language learner pre-disposed to categorical rules, or do categorical tendencies develop later and (3) is low-frequent syntactic variation in speakers conditioned by register/dialect? To address these issues, we will use novel large-scale experiments for speech elicitations, combined with methods for text-speech alignment, which will be developed within the project.