Understanding Formulaic Language

A Second Language Acquisition Perspective

Anna Siyanova-Chanturia, Ana Pellicer-Sánchez, Anna Siyanova-Chanturia, Ana Pellicer-Sánchez

• Publisher: Routledge
• Year: 2018
• ISBN:
  9781351797566
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

Understanding Formulaic Language: A Second Language Acquisition Perspective brings together leading scholars to provide a state-of-the-art, interdisciplinary account of the acquisition, processing, and use of formulaic language. Contributors present three distinct but complementary perspectives on the study of formulaic language – cognitive/psycholinguistic, socio-cultural/pragmatic, and pedagogical – to highlight new work as well as directions for future work. This book is an essential resource for established researchers and graduate students in second language acquisition and pedagogy, corpus and cognitive linguistics, psycholinguistics, sociolinguistics, and pragmatics.

Part I

Cognitive and Psycholinguistic Perspectives on Formulaic Language

1

Acquisition of Formulaic Language from a Usage-Based Perspective

Stefanie Wulff

Introduction

This chapter provides an overview of contemporary research on the second language (L2) acquisition of formulaic language (FL) from a usage-based perspective. Usage-based studies, in particular those based on large-scale and dense corpora of child-directed speech and learners’ productions that have become available in the past 20 years, have demonstrated how FL serves as a starting point for inductive construction learning that complements deductive, rule-based learning processes. This process, called bootstrapping, allows learners to move from lexically fully specified exemplars that are assigned a holistic communicative function to low-scope patterns that are only partially lexically fixed to fully schematized constructions (in the Goldbergian sense of the term; see Goldberg, 2006). The following section outlines the understanding of FL as one type of construction in usage-based linguistics. Given that the input learners receive is assigned a central role in the acquisition process, I then discuss the different relevant input characteristics, such as effects based on frequency and distribution, recency, functional prototypicality, and contingency of the form-function mappings in question (Ellis & Wulff, 2015a, b). I then turn to evidence in favour of constructions in the L2 before we look at the contributions of FL to the L2 acquisition process in more detail. The chapter closes with a few suggestions for future research.

This chapter is by far not the first overview of the role of FL in L2 acquisition, and so it draws substantially on previous publications. The interested reader may in particular want to consult Weinert (1995) for a comprehensive review of formulaic sequences (FSs) in non-native language up until the date of publication; Cadierno and Eskildsen (2015) and Eskildsen (2009) for much more in-depth descriptions of usage-based approaches to L2 acquisition in general; and Ellis, Simpson-Vlach, Römer, Brook O’Donnell, and Wulff (2015) for a recent overview of FL in L2 from a usage-based perspective.

Finally, a quick note regarding what this chapter covers and what it does not. Firstly, so as to avoid overlap with the other chapters in this volume, the reader should consult the respective chapters on processing of FL in the L2, influence of the first language (L1), and FL use in the context of foreign language learning, and specifically from a corpus-linguistics point of view, as this chapter deliberately only alludes to these (crucial) aspects in passing. Secondly, as the title suggests, this chapter only reviews research that falls under the label of usage-based approaches, or is at least compatible with the core assumptions of that set of linguistic frameworks, but in order to obtain a comprehensive understanding of state-of-the-art research on L2 FL use and acquisition, I strongly encourage the reader to look into formalist research such as Bardovi-Harlig and Stringer (2017), Myles (2012), and Myles and Cordier (2017), as this line of research has been advancing our understanding of the topic in important ways.

Formulaic Sequences as Constructions in Usage-Based Linguistics

A variety of linguistic theories fall under the label “usage-based linguistics” (UBL; see Barlow & Kemmer, 2000; Tummers, Heylen, & Geeraerts, 2005); minimally, they all share two working assumptions regarding (both first and) second language (L2) acquisition¹ (Ellis & Wulff, 2015b, p. 75):

1. (1) The linguistic input learners receive is the primary source for their second language (L2) acquisition.
2. (2) The cognitive mechanisms that learners employ in language learning are not exclusive to language learning, but are general cognitive mechanisms associated with learning of any kind.

In UBL, language knowledge is defined as a structured inventory of symbolic units or form-meaning pairs (Langacker, 1987, 2000) or constructions (Goldberg, 2006). In crucial opposition to rule-based approaches to language acquisition that define language acquisition as the acquisition of words and rules that specify how to combine them, UBL approaches discard with the traditional dichotomy of syntax and lexicon and instead advocate for a mental lexicon “in which abstract grammatical patterns and the lexical instantiations of those patterns are jointly included, and which may consist of many different levels of schematic abstraction” (Tummers, Heylen, & Geeraerts, 2005, pp. 228–229; see also Bates & Goodman, 1997). In construction grammar, these patterns are referred to as constructions (Goldberg, 2006). In keeping with Tummers et al.’s definition, constructions can be morphemes, words, phrases, and syntactic frames (see also Trousdale & Hoffmann, 2013). That is, simple morphemes such as -licious (meaning “delightful, attractive”) are constructions in the same way as simple words like raisin (meaning “partially dried grape”), collocations like raisin bran, idiomatic noun phrases like raisin ranch (meaning “retirement home”), and abstract syntactic frames like Subject-Verb-Object-Object (meaning that something is being transferred, as realized in sentences as diverse as Max gave Abigail a raisin, Max gave Abigail a hug, or Max baked Abigail a cake, where nuts, hugs, and cakes are being transferred, respectively). As the latter examples illustrate, not all constructions carry meaning in the traditional sense; many constructions rather serve a more functional purpose. The passive construction, for instance, serves the function of shifting the focus of attention in an utterance from the agent of the action to the patient undergoing the action (compare the passive A cake was baked for Abigail with its active counterpart Max baked Abigail a cake).

Given this definition of constructions, it follows that they have to be simultaneously stored in multiple forms that differ in their level of complexity and abstraction. To give a simple example of different levels of constructional complexity, the words raisin and the plural -s morpheme are simple constructions; both are stored also as constituent parts of the more complex construction raisins (“more than one raisin”). Different levels of constructional abstraction (also referred to as schematization) are evident, for example, in the fully lexicalized formula Thank you versus the partially schematized slot-and-frame greeting pattern [Good + (time of day)], which can be realized as lexicalized phrases like Good afternoon and Good evening; and the completely schematic [Noun Phrase + Noun Phrase] construction, which could be realized as raisin bran, cake pan, or soda can, to give but three examples. Likewise, sentences are not the product of applying a rule to a number of words to arrange them in a grammatical sequence, but rather a sentence is a specific combination of constructions – some simple, some complex, some lexically specific, some abstract. What did Max give Abigail, for instance, is a combination of the following constructions:

• Max, Abigail, give, what, do constructions
• VP, NP constructions
• Subject – Verb – Object – Object construction
• Subject – Auxiliary inversion construction

In summary, a speaker’s knowledge of their language(s) is a huge warehouse of constructions that vary in complexity and abstraction. More specifically, language knowledge means knowing the properties of constructions that determine what other constructions they can, or have to be, combined with. Typically, these properties are semantic or functional in nature such that any two constructions can only be combined if their meanings/functions are compatible, or if they can at least temporarily attain compatibility in a specific context or discourse situation (Goldberg, 2006).

Given this definition of constructions, UBL sees FL not as fundamentally different from genuinely productive, on-the-fly assemblies; rather, there is a continuum of constructions from simple to complex and from lexically specified to lexically open. In view of the fact that FL constitutes a significant portion of any competent speaker’s language knowledge – one estimate for spoken language is 28% (Biber, Johansson, Leech, Conrad, & Finegan, 1999) – placing FL at the centre of language knowledge rather than relegating it to the margins of what speakers do can be seen as a distinct advantage of UBL approaches over formalist theories.

Input Properties Contributing to the Second Language Acquisition of (Formulaic) Constructions

Frequency of Construction in the Input

In usage-based approaches, frequency is a driving force for language acquisition, with type and token frequency playing different roles. Token frequency is the frequency with which a particular construction occurs in the input; type frequency refers to the number of distinct realizations of a given construction. The English past tense morpheme -ed has a high token frequency – it occurs frequently – and it has a very high type frequency as it occurs with thousands of different verbs. In contrast, irregular past tense forms as in blew, sang, or rode have low type frequency as they occur with a much smaller number of different verbs. Type frequency drives the productivity of a construction since it gives the hearer varied opportunities to parse the construction in question, which strengthens the schematic representation of the form and makes it more available for re-use and novel uses alike (Bybee & Hopper, 2001). High token frequency likely has the opposite effect since it conserves specific realizations of a construction (see Bybee, 2006 for a detailed discussion of the conserving, form-reducing, and autonomy-stipulating effects of high token frequency).

The more often two constructions co-occur, the more entrenched that particular constructional arrangement becomes (Divjak & Caldwell-Harris, 2015). Ellis (1996, p. 107) refers to “the development of permanent sets of associative connections in long-term memory” as chunking, and he points out how chunks form the basis for automaticity and fluency in language use.

At the same time, it is important to point out that not all FL has to be particularly frequent to be acquired – think of rarer formulaic expressions such as needle in a haystack, red herring, or bite the dust. In contrast, the salience and thus learnability of these expressions is based on the salience they obtain by being unusual. Likewise, frequency is not a sufficient condition for status as an FS: n-grams like and of the and but it is are extremely frequent, yet they are neither psycholinguistically salient nor coherent (Schmitt, 2004). As evidence from psycholinguistic studies and corpus analyses suggests, speakers consider not only a construction’s frequency but also a variety of factors, including its sequential dependencies, prototypicality, and reliability of form-function mappings instantiated in the exemplar (see Ellis, 2002 for an overview).

Distribution of Construction in the Input

Acquisition is kick-started by initial exposure to massive, low-variance input that is centered around prototypical realizations (or exemplars) of the target construction (Elio & Anderson, 1981, 1984). Through this focused input, learners can identify the majority of the category members; through continued exposure to the full breadth of exemplar types, learners can later define the category boundaries (Nosofsky, 1988). Accordingly, both children’s input and output reflect Zipf’s Law (Zipf, 1935): the frequency of a word is inversely proportional to its rank in a frequency table. This means that the most frequent word occurs about twice as often as the second most frequent word, three times as often as the third most frequent word, etc. Goldberg, Casenhiser, and Sethuraman (2004) showed that Zipf’s Law holds when counting words in a given sample of authentic speech, and importantly, it also holds for verbs within a given construction. According to Goldberg et al., the Zipfian input distribution directs attention to specific typical verbs that are made salient by being extremely frequent in the input, which then can serve as the “path-breaking verbs” for category formation (see also Ninio, 1999, 2006). Ellis and Ferreira-Junior (2009a, 2009b) considered data from naturalistic L2 acquisition and likewise confirmed that the type/token ratio of the verbs in argument structure constructions is Zipfian. Furthermore, they demonstrated that akin to Tomasello’s (2003) claims for L1 acquisition, the most frequent and prototypical verbs seem to act as “verb islands” around which verb-argument constructions gradually emerge.

As Gries and Ellis (2015) point out, Zipfian distributions also point to another aspect of how constructions are distributed in the input, namely their dispersion. Constructions can be differently dispersed across the contexts they occur in: one construction may occur with about the same frequency across all its possible contexts (that is, it is evenly dispersed), while another construction clusters more frequently in some contexts than others (that is, it is unevenly dispersed). At least for initial stages of learning, it appears that the dense clusters of occurrences as provided by the most frequent elements in a construction with a Zipfian distribution facilitate their acquisition more than when a construction is more evenly dispersed: when Goldberg, Casenhiser, and Sethuraman (2004) compared two experimental conditions in their learning experiment, namely one in which participants heard five novel verbs 16 times in an evenly dispersed condition (4-4-4-2-2) versus one in which the 16 occurrences were presented less evenly dispersed (8-2-2-2-2), uptake was significantly better in the less evenly dispersed, more Zipfian condition.

Recency of Construction in the Input

Alongside frequency and context, we know from research in cognitive psychology that recency plays a crucial role as well (Anderson, 1989; Anderson & Schooler, 2000). Recency (also called priming or persistence) is an implicit memory effect such that exposure to a stimulus affects a response to a later stimulus. Recency has been shown to impact processing across linguistic levels from phonology to syntax (McDonough, 2006; McDonough & Mackey, 2006; McDonough & Trofimovich, 2008).

Salience, Redundancy, and Perception of Form of the Construction

Salience is the general perceived strength of a stimulus. According to the Rescorla-Wagner (1972) model of associative learning, the amount of learning induced from an experience of a cue-outcome association crucially hinges upon the salience of the cue and the importance of the outcome: low salience cues are less readily learned than high salience cues. Many grammatical structures in English have low salience in the input, for example, inflections like the third person singular -s morpheme. Consequently, these grammatical structures pose challenges even for advanced L2 learners.

Many grammatical constructions are not only low in salience but also are redundant in the listener’s understanding of an utterance in that they compete with more salient psychophysical forms, which further impedes acquisition. For example, third person singular -s encodes present tense, but adverbs like today are more salient in the input and effectively overshadow and block acquisition of the morpheme (Ellis, 2006, 2008b; Ellis & Sagarra, 2010b, Goldschneider & DeKeyser, 2001). Accordingly, L2 learners’ preference of adverbial over inflectional cues to tense has been well documented in longitudinal studies of naturalistic L2 acquisition (Bardovi-Harlig, 2000; Dietrich, Klein, & Noyau, 1995), training experiments (Ellis & Sagarra, 2010b, 2011), and studies of L2 language processing (Ellis & Sagarra, 2010a; VanPatten, 2006).

Prototypicality of Function

Some members are better...

Table of contents

• Cover
• Title
• Copyright
• Dedication
• Contents
• Notes on Contributors
• Formulaic Language: Setting the Scene
• Part I Cognitive and Psycholinguistic Perspectives on Formulaic Language
• Part II Socio-Cultural and Pragmatic Perspectives on Formulaic Language
• Part III Pedagogical Perspectives on Formulaic Language
• Index