Is Structure Dependence Shaped for Efficient Communication?: A Case Study on Coordination

TL;DR

The paper tests whether structure dependence, a core syntactic property, can be explained by domain-general pressures for efficient communication. Using three artificial languages (no-reduction, structure-reduction, linear-reduction) derived from probabilistic grammars, and evaluating them with Recurrent Neural Network Grammars, it demonstrates that structure-dependent reduction yields higher communicative efficiency across a broad trade-off range between simplicity and informativeness. The findings imply that abstract syntactic properties like coordination may originate from efficiency-driven design principles, challenging the view that such properties are inherently domain-specific. This approach provides a framework for linking syntactic universals to functional pressures and suggests directions for extending the analysis to natural-language data and broader syntactic phenomena.

Abstract

Natural language exhibits various universal properties. But why do these universals exist? One explanation is that they arise from functional pressures to achieve efficient communication, a view which attributes cross-linguistic properties to domain-general cognitive abilities. This hypothesis has successfully addressed some syntactic universal properties such as compositionality and Greenbergian word order universals. However, more abstract syntactic universals have not been explored from the perspective of efficient communication. Among such universals, the most notable one is structure dependence, that is, the existence of grammar-internal operations that crucially depend on hierarchical representations. This property has traditionally been taken to be central to natural language and to involve domain-specific knowledge irreducible to communicative efficiency. In this paper, we challenge the conventional view by investigating whether structure dependence realizes efficient communication, focusing on coordinate structures. We design three types of artificial languages: (i) one with a structure-dependent reduction operation, which is similar to natural language, (ii) one without any reduction operations, and (iii) one with a linear (rather than structure-dependent) reduction operation. We quantify the communicative efficiency of these languages. The results demonstrate that the language with the structure-dependent reduction operation is significantly more communicatively efficient than the counterfactual languages. This suggests that the existence of structure-dependent properties can be explained from the perspective of efficient communication.

Figures (6)

• Figure 1: A coordinate structure (b) is derived by applying Conjunction Reduction, a structure-dependent reduction operation to a sentence-level coordinated kernel sentence (a).
• Figure 2: Examples of the three languages expressing the same meaning. The word order is set with all six switches being strictly head-final as in Japanese. For simplicity, information on number and tense has been omitted from the syntactic categories in these figures.
• Figure 3: Distribution of communicative efficiency for the three types of languages with 95% CI. The x-axis and y-axis represent the trade-off parameter $\lambda$ and communicative efficiency, respectively. Both predictability and parsability are z-transformed for an interpretation of $\lambda$. The structure-reduction languages are the most communicatively efficient under the parameter $\lambda\in[0.18, 0.93]$ for 95% CI.
• Figure 4: Distribution of predictability for the three types of languages. Error bars indicate 95% CI.
• Figure 5: Distribution of parsability for the three types of languages. Error bars indicate 95% CI.