Having assigned to the Hungarian sentence a binary predication structure, and having examined the properties of the logical subject of predication, or topic, we turn to the analysis of the predicate phrase. Categorially the predicate is a VP merged with morphosyntactic elements such as tense, mood, and agreement, and either extended into an aspectual phrase, or embedded in operator projections such as a focus phrase, a distributive quantifier phrase, and/or a negative phrase. The subject of this chapter is the minimal predicate, consisting of a VP, merged with morphosyntactic heads, and extended into an AspP, but not involving a focus, a distributive quantifier, or negation.

Argument order in the VP

The lexical core of the predicate of the Hungarian sentence is a verb phrase. It is assumed to be verb initial, with the arguments following the verb in an arbitrary order – as illustrated in (1). (What motivated the assumption of a verb-initial VP in the late 1970s was that the set of possible permutations of a verb and its complements could be derived most economically from a V-initial base. Later theoretical considerations – concerning the direction of theta-role assignment and Case assignment in Universal Grammar – also confirmed this view.)

Introduction

In first language acquisition research, it has long been recognized that quite lengthy strings, which would correspond to several adult words, can be treated as a single unit by the young child (e.g., Bolinger 1975:100; Crystal 1997:244; Plunkett 1993:44). In this chapter and the next, an attempt will be made to reconcile a number of observations made over the last three decades or more about these strings, and to accommodate them within a model of the child's language use and linguistic development. Peters (1983) characterizes the child's encounter with spoken language as follows:

It is that process of ‘capturing pieces’ that lies at the heart of understanding the role of formulaic sequences in first language acquisition. That children do store and use complex strings before mastering their internal makeup is generally agreed. However, researchers have varied in their views about how significant they are. Brown (1973) acknowledges that strings like What's that in the very young child are not a product of a grammar, but, rather, “must be generated by some simpler mechanism either as fixed routines or as simple frames in which a set of words could rotate” (p. 181).

Introduction

If there is a standard view of what formulaic language is (and the range of descriptions reviewed in this book must cast severe doubt on that), at its heart will be something about word strings which ‘break the rules’. They can break phonological rules, by displaying fewer stresses than expected and by being articulated faster and less clearly. They can break syntactic rules, by resisting pluralization, passivization and so on, and by containing constituents which do not take on their normal grammatical function. They can break lexical rules, by containing items which are archaic or have no independent existence. And they can break semantic rules, by combining to mean something other than they ought to, and by being more idiomatic than an equivalent nonformulaic combination.

Because of the focus on these various kinds of irregularity, formulaic language has customarily been viewed as exceptional, and has been relegated to a minor part of the lexicon. Although individual lines of research have noted its role in processing economy, interaction, language learning, and expression and recovery in aphasia, it has seemed sufficient to place formulaic word strings awkwardly at the edge of lexical models and, often, entirely outside of grammatical ones. In contrast, this book has drawn formulaic language from the edges of an account of linguistic processing to its very centre, by recognizing its many roles and by aligning the irregular minority with a regular majority of formulaic strings which are normally either overlooked entirely or treated with puzzlement.

In Chapter 1, we saw that the practical purpose of using formulaic sequences appears to be that they reduce processing effort. As Kuiper (1996) puts it:

By saving on processing, the speaker is able to focus on other kinds of concomitant activity. These may be, for instance, evaluating the ideas contained in the present conversation (e.g., Wray 1992), or engaging in another, unrelated, activity. Jaffe (1978) illustrates the difficulties of the overload which can occur when trying to pay attention to too many things at once:

Despite these observations, it will become clear in the course of this chapter that not all formulaic sequences can be satisfactorily explained in terms of their ability to reduce the processing load of the speaker.

Introduction

Aphasia is a term used to describe a number of different kinds of disruption to language following damage or surgery to the language-dominant hemisphere of the brain (usually the left). People with aphasia offer some insights into the nature of formulaic sequences which it would be difficult to obtain from those whose language-processing faculties are functioning normally. Formulaic, or automatic, language has long been recognized as a key feature of aphasia, and indeed it was being described in the medical literature long before it was identified by linguists as a significant component of normal language (see Benton & Joynt 1960, Critchley 1970 and Espir & Rose 1970:24ff for historical overviews). Case studies of aphasic patients date back several centuries, and in the twentieth much detailed work was made possible by the localized brain injuries of otherwise healthy young men returning from war (e.g., Goldstein 1948; Head 1926; Russell & Espir 1961). Accounts characteristically highlight as a curiosity the survival of some lengthy sequences even when very little other language remains. In one of the earliest published reports, from 1683, Peter Rommel wrote of an aphasic woman:

Introduction

In this chapter we shall explore the consequences of incorporating the formulaic features of aphasic language into a lexical model. In the models developed so far in this book, it has been possible to speak of a single lexicon containing units of different sizes, namely, morphemes, formulaic words and formulaic word strings. Now, however, it seems that it will be necessary to make finer distinctions, both because of what all aphasias have in common and because of the ways they differ.

As we have seen, formulaic word strings are strikingly resilient to most types of aphasia, but, as with single words (mono- and polymorphemic), not all types are equally spared, either within an individual's repertoire or from one patient to another. This seems to suggest that the lexicon needs to be divided in some way other than by unit complexity, so that subsets of words and longer strings can be lost, or spared, together. These different parts of the lexicon would need to be managed from different locations in the brain. Physical distance between operational centres does not necessarily need to be represented in terms of separate components in an abstract processing model. However, if such physical separation influences the pattern of deficits after localized brain damage, then it certainly does need to be modelled in some way.

Introduction

To know a language you must know not only its individual words, but also how they fit together. Part of this knowledge entails developing suitable rules to generate all the possible grammatical utterances of the language, but another crucial aspect is coming to know which of the feasible grammatical utterances are idiomatic and nativelike. Pawley and Syder (1983) point out that one of the most difficult tasks for even the most proficient non-native speaker is learning to select that subset of utterances that are customarily used by native speakers, from out of the much greater inventory of those that could be. The choices that result in such ‘customary’ turns of phrase as I'll be back in a tick/mo in preference to I'll return in a short while and let me have a go rather than I should like to try are, because the choices lie entirely within the grammatically possible, subtle to the point of slipperiness, and a non-native can only learn to prefer those which are the usual forms in a given speech community by observation and imitation (see, for example, Willis 1990:63–64). For some idiomatic expressions, the whole is not even a literal reflection of the meanings of the constituent words, and the functional associations can be quite arbitrary. For example, in some varieties of English, nice to meet you is both a greeting and a closer, while nice meeting you is only a closer (Schmidt 1983:152).

Introduction

Chapters 9 and 10 have revealed that the roles of formulaic sequences in second language production, comprehension and learning are various and complex. This chapter seeks to capture this complexity by developing a second language version of the first language model proposed in Chapter 7. Central to this will be the differences in use and extent of use of formulaic sequences in younger and older learners, and in naturalistic and taught learners. But first, we shall consider some puzzles which the model needs to be able to solve.

The Control and Use of Formulaic Sequences

The first puzzle is the one raised by Yorio's (1989) data. He found that the written English of a group of advanced ESL students in the United States contained a great many attempts at formulaic sequences, but that they were riddled with errors. For example, he found take advantages of; are to blamed for; those mention above; being taking care of; a friend of her; make a great job; on the meantime; with my own experience; put more attention to (pp. 62–63). Yorio's interpretation is that “these expressions are not simply memorized or taken in as wholes, but… are subject to whatever interlanguage rules the learner is operating under” (p. 62). The question, then, is: are they formulaic or aren't they? The examples seem close enough to their target not to be straightforward inventions. That means that the correct version has been encountered by the student, and that it has been recognized as formulaic.

Introduction

In the previous chapter, we saw how formulaic sequences are able to support the process of first language acquisition in a variety of ways. They appear to

• establish a culture of interaction with carers;

• supplement gesture and other nonlinguistic behaviours in conveying the most important manipulative messages before the production of rule-governed language is possible;

• represent the entry of the child into the group of those who know this or that rhyme or song and expect certain linguistic behaviour;

• provide the child with material for analysis; and

• reduce the child's processing load once novel construction is possible.

It was also evident that for many commentators the default status of formulaic sequences (apart from the fused utterances) is that they are only temporarily holistic. They have been stored whole because that is the only thing the child can do with them, but once analysis is possible, they are routinely broken down in order to identify regular configurations and useful recombinable units. However, there is a major problem with this assumption.

The stream of language which the child hears can be divided into different types. There are novel constructions that may never be heard again: these can be analyzed, but there is only one opportunity to catch them. There are strings which are formulaic for a particular speaker, because he or she has fused and stored them: these are not necessarily a chunk in anyone else's lexicon.

Introduction

Our basic and recurrent question in this chapter is whether the use of formulaic sequences by children learning an L2, as reported in the research literature, can be accounted for in terms of the agenda model presented in Figure 6.3. The full set of studies covered in this chapter is listed in Table 9.1, and the ages of the learners are represented in Figure 9.1. In all the cases of interest to us, the children had already established their L1 before they were exposed to the L2, and so are older than an equivalent L1 learner would be.

Very Young Children in L2 Daycare/Nursery

Because age is so tied into rapid physical, emotional and cognitive development in early childhood, we should expect the youngest learners to provide the closest match with the first language model, whether this be for neurological, cognitive or environmental reasons. We begin, then, with the youngest children. Virve and Karen were both under two years of age when their respective studies began, and were exposed to L2 in their daycare setting. Takahiro was two-and-a-half and attended an L2 medium nursery.

How Did the Learners Get Things Done?

We saw in Chapters 6 and 7 that ‘getting things done’ is an agenda of great importance to the infant, who is relatively helpless in meeting its own needs. Virve, Karen and Takahiro were not helpless infants: they could walk and feed themselves, and they already had a linguistic means of expressing their needs, their L1. On the other hand, they were still much less self-sufficient than an older child or adult. What strategies did they employ in trying to meet their needs through the agency of another?

This book began with a mystery. I had been reading about formulaic language in the context of language proficiency, and had been struck by three observations made in the literature. The first was that native speakers seem to find formulaic (that is, prefabricated) language an easy option in their processing and/or communication. The second was that in the early stages of first and second language acquisition, learners rely heavily on formulaic language to get themselves started. The third observation, however, seemed to fly in the face of the first two. For L2 learners of intermediate and advanced proficiency, the formulaic language was the biggest stumbling block to sounding nativelike. How could something that was so easy when you began with a language, and so easy when you were fully proficient in it, be so difficult in between?

I set myself the challenge of finding out, and focussed on two possibilities, both of which I now judge to be true. One was that the formulaic language described in the various areas of study was not quite the same thing in each case. The second was that there was some other key to understanding the nature of formulaic language, one which would be difficult to spot by looking only at the different types of data in isolation. The common link between formulaic language across different speakers might even not be linguistic at all.