Analogies in Spontaneous Discourse

Rachel ReichmanACL 1981 · DOI

Abstract

This paper presents a formal framework for analyzing analogies as strategic moves within natural, spontaneous conversation. By synthesizing Structure-Mapping Theory with Context Space Theory, the research demonstrates how analogies serve specific communicative goals—such as support, challenge, or topic shifting—and how their linguistic realization is governed by the hierarchical structure of discourse.

Introduction

Traditional research on analogy has largely focused on controlled experimental settings, where subjects map pre-defined domains in isolation. However, in spontaneous discourse, analogies are dynamic and often contentious. They are used to pivot between topics, argue points, and navigate complex social interactions. This paper addresses the gap between abstract cognitive models and the linguistic reality of natural dialogue, proposing a model that explains how speakers and listeners manage "context spaces" to maintain coherence during analogical reasoning.

Theoretical Framework

The proposed model integrates two primary components:

1. Context Space Theory

Discourse is modeled as a hierarchy of "context spaces," each representing a specific communicative goal (e.g., Support, Challenge, Further Development). Using an Augmented Transition Network (ATN), the framework tracks the state of these spaces—whether they are Active, Controlling, or Closed. This state determines how speakers use linguistic cues to signal shifts between a base domain and an analogous domain.

2. Integrated Structure-Mapping

Building on Dedre Gentner’s Structure-Mapping Theory, the author argues that analogies map relational structures rather than object attributes. Crucially, Reichman extends this by asserting that an analogy’s validity in discourse is tied to its "communicative goal identification." An analogy is deemed successful only if it serves the current objective of the conversation.

Case Study: The Ireland-Vietnam Analogy

The paper analyzes several transcripts of natural conversation to validate the model. A prominent example involves a political discussion regarding the history of turmoil in Ireland. In this exchange, a speaker draws a parallel between the British presence in Ireland and the United States' involvement in Vietnam.

Mappings between the Ireland and Vietnam contexts in Excerpt 2.

As shown in the mappings above, the analogy relies on relational identities (e.g., a foreign power intervening in a local conflict) rather than attribute identities. The speaker accepts the analogy not because the countries are inherently similar, but because the relational structure of the intervention matches the discourse's illustrative goal. Conversely, analogies are rejected when the underlying "cause" or "intent" of the relations fails to align, even if surface-level similarities exist.

Linguistic Markers and Deictics

A key finding is the strategic use of deictics (pointing words) to navigate context spaces. Speakers use "close" deictics (e.g., this, here) for the currently "Active" context space and "far" deictics (e.g., that, there) for spaces that are "Controlling" but not active. This mechanism allows listeners to disambiguate references across complex, nested analogies. For instance, a speaker might bypass a recently mentioned entity to refer to a more distant one if the latter resides in a still-active context space, while the former was part of a "closed" analogous sub-space.

Conclusion

By formalizing analogies as functional conversational moves, this research provides a bridge between cognitive science and linguistics. The model successfully predicts linguistic choices that traditional grammars might find anomalous. While the study is limited by its qualitative nature and small sample size, it lays the groundwork for a more robust computational understanding of how humans use relational reasoning to structure communication. Future work should focus on expanding the dataset and refining the semantic constraints that govern discourse generation.

Summary updated

BibTeX
@inproceedings{reichman1981analogies,
  author = {Rachel Reichman},
  title = {{Analogies in Spontaneous Discourse}},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  year = {1981},
  doi = {10.3115/981923.981942},
}
Figure