Referring as Requesting

Philip R. CohenACL 1984

TL;DR

This paper argues that the act of referring, often considered a "propositional act" within speech act theory, can be more effectively understood as an illocutionary act: a request for the listener to identify a referent. This reframing simplifies the theory of referring by eliminating the need for a separate propositional act and offers a more unified account of how we use language to point to things in the world.

What Problem Does This Paper Address?

Speech act theory, particularly as developed by Searle, distinguishes between different types of acts performed when we speak. One such act is "referring," which Searle categorized as a "propositional act." This means referring is seen as a component of a larger illocutionary act (like asserting or questioning) rather than a standalone communicative intention. However, this view struggles with utterances that only serve the purpose of referring, such as isolated noun phrases or existential sentences used to introduce a topic. The paper questions the necessity of Searle's "propositional act of referring" and seeks a more parsimonious and comprehensive explanation that can account for all uses of referring expressions.

Methods

The paper proposes an "Illocutionary Act Analysis" (IAA) where referring is framed as a request for the hearer to perform the act of "referent identification." This identification is initially understood in a perceptual sense: the hearer uses sensory input to pick out an object described by the speaker. The author revises Searle's notion by making identification an act performed by the hearer, not the speaker.

The core of the IAA is a plan-based reasoning framework, similar to those used for analyzing indirect speech acts. When a speaker utters something that implies a precondition for referent identification is met (e.g., an existential sentence like "there's a little yellow piece of rubber"), the hearer can infer that the speaker wants them to identify that referent. This inference is licensed by shared knowledge of speaker goals, hearer capabilities, and the cooperative principle.

The paper then demonstrates how this IAA can subsume Searle's original "propositional act analysis" (PAA). By correlating a specific complex propositional attitude with the use of definite determiners, the IAA can derive Searle's conditions for successful referring as a consequence of a request for referent identification.

Datasets

Not discussed in the paper. The paper relies on linguistic examples and an experiment involving telephone instructions, but no formal datasets are presented or used for training/evaluation.

Key Insights and Ideas

  • Referring as a Request: The most significant insight is reframing referring as a request for identification, rather than a distinct propositional act. This simplifies the theoretical landscape.
  • Hearer-Centric Identification: Shifting the focus of "identification" from the speaker's act to the hearer's act makes the request-based analysis more natural.
  • Plan-Based Reasoning: Utilizing plan-based reasoning, common in AI and linguistics, provides a robust mechanism for inferring the speaker's intention to request referent identification, even from seemingly simple utterances.
  • Unified Account: The IAA offers a unified way to handle both direct and indirect referring, as well as isolated referring expressions, which Searle's PAA struggled with.
  • Subsumption of PAA: The paper shows that Searle's conditions for referring can be derived from the IAA, suggesting the PAA is not independently necessary.

Key Results

The paper doesn't present quantitative results from experiments. Its key "results" are theoretical:

  • It demonstrates that utterances previously considered problematic for Searle's PAA (like isolated noun phrases or existential sentences used for reference) can be elegantly explained as requests for referent identification.
  • It shows that Searle's conditions for successful referring can be accounted for within the IAA framework, suggesting that the propositional act of referring is redundant.
  • The IAA is presented as a more general and independently motivated account that can subsume the PAA.

Limitations and Future Work

  • Perceptual Focus: The initial analysis of "identification" is limited to a perceptual sense. The paper acknowledges the need to extend this to Searle's broader concept of identification (establishing a coreferential link) to fully subsume the PAA.
  • Complexity of Speaker Intentions: The authors note that capturing the distinction between achieving effects on a hearer and doing so communicatively (in the Gricean sense) is complex, especially for non-referential uses of noun phrases.
  • Non-Communicative Effects: The paper briefly touches on the possibility of non-communicative effects in reference, which might challenge the Gricean requirement of intent recognition and could potentially provide grounds for a propositional act. However, the authors suspect Searle would not want to abandon Gricean principles.
  • Future Work: The authors suggest future work should explore how plan-based reasoning can explain cases where a hearer cannot identify a referent or where the speaker's intention is not what it initially appears (e.g., distinguishing referential from attributive uses of definite descriptions). They also aim to formalize the distinction between communicative and non-communicative effects.

Summary updated

BibTeX
@inproceedings{cohen1984referring,
  author = {Philip R. Cohen},
  title = {{Referring as Requesting}},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  year = {1984},
}
Figure