Is All Formulaic Language Created Equal? Unpacking the Processing Advantage for Different Types of Formulaic Sequences.

Gareth Carrol,Kathy Conklin

doi:10.1177/0023830918823230

Gareth Carrol, Kathy Conklin

Open Access

https://doi.org/10.1177/0023830918823230

Copy DOI

Abstract

Research into recurrent, highly conventionalized "formulaic" sequences has shown a processing advantage compared to "novel" (non-formulaic) language. Studies of individual types of formulaic sequence often acknowledge the contribution of specific factors, but little work exists to compare the processing of different types of phrases with fundamentally different properties. We use eye-tracking to compare the processing of three types of formulaic phrases-idioms, binomials, and collocations-and consider whether overall frequency can explain the advantage for all three, relative to control phrases. Results show an advantage, as evidenced through shorter reading times, for all three types. While overall phrase frequency contributes much of the processing advantage, different types of phrase do show additional effects according to the specific properties that are relevant to each type: frequency, familiarity, and decomposability for idioms; predictability and semantic association for binomials; and mutual information for collocations. We discuss how the results contribute to our understanding of the representation and processing of multiword lexical units more broadly.

Highlights

Cheetahs and Ferraris are two examples of members of the category ‘things that are fast’
While overall phrase frequency contributes much of the processing advantage, different types of phrase do show additional effects according to the specific properties that are relevant to each type: frequency, familiarity and decomposability for idioms; predictability and semantic association for binomials; and mutual information for collocations
Our results suggest that both Cloze probability and mutual information (MI) are both better explanatory variables for collocations than phrase frequency, all are derived from experience and support the broad conclusion that distributional factors are the primary drivers of collocational processing

Summary

Introduction

Cheetahs and Ferraris are two examples of members of the category ‘things that are fast’. Beyond this broad similarity, there is not much that makes the two comparable, and it is clear that the mechanisms that make each one fast are very different. Formulaic language is an example of something that has sometimes been defined just as broadly. It encompasses a broad range of multiword sequences that fulfil a number of communicative functions (Wray, 2002, 2008), and knowledge of such sequences is an important part of how we use language. Claims are made about the “holistic” nature of formulaic sequences, suggesting that all recurrent sequences are stored in the lexicon and retrieved directly, the nature of what is meant by this may be quite variable (c.f. Wray, 2012, p.234)

Methods

Results

Discussion

Conclusion