This chapter discusses the role of frequency for Construction Grammar, especially concerning usage-based models of language, and offers definitions of different aspects of frequency, namely token frequency, type frequency, relative frequency, frequency of co-occurrence, and dispersion. It discusses how these aspects can be measured on the basis of corpus data, and how these measurements allow the observation of frequency effects that relate to phenomena such as entrenchment, ease of processing, productivity, phonological reduction, and resistance to regularization. These effects are illustrated by experimental and corpus-based analyses of lexical, morphological, and syntactic constructions. The chapter also addresses open questions regarding the role of frequency in constructionist research. Not only is the relation between corpus frequencies and theoretical notions such as entrenchment far from trivial, it is also important not to attribute effects to token frequency that can be explained by other, correlating variables. The chapter will also examine strategies that can reach beyond the use of frequency values in the future development of Construction Grammar.

Introduction

To most of you, it will not be immediately obvious why a textbook on English historical linguistics should contain a chapter on psycholinguistics. What do the two have to do with one another? Historical linguistics is the study of how languages change in the long term, over many generations of speakers, while psycholinguistics is concerned with the cognitive processes that take place when an individual speaker produces or processes language. It could thus be argued that historical linguists look at language at the level of entire linguistic communities, while psycholinguists try to understand how language works in the mind of a single human being. Against this dichotomy, this chapter will make the case that historical linguistics and psycholinguistics are much more deeply connected than it might initially appear. The centerpiece of that argument is the observation that language gradually changes in the long term simply because individual speakers in the here-and-now process it when they communicate. This idea was formulated in the nineteenth century by the historical linguist Hermann Paul:

Comments like these make Hermann Paul an intellectual grandfather of what is nowadays called usage-based linguistics (Bybee and Hopper 2001; Bybee 2010). As the name suggests, this framework assigns a fundamental role to language use, which is seen as the basis of speakers’ knowledge of language. As you experience language in use, you retain detailed memories of that experience and build up linguistic knowledge. This knowledge is flexible, and it is continually reshaped as you encounter new uses of language. This means that over time, your knowledge of language changes. With it changes your language production, which in turn influences the linguistic knowledge of the speakers around you. Through repeated cycles of small changes in usage, languages can change rather dramatically in the long term.

When language change is viewed in this way, it becomes apparent that it is closely intertwined with the psychological processes that shape language use in the here-and-now.

This chapter surveys the different sources of diachronic corpus data that are used in the present work and then moves on to the technical part of this book, namely, an elaboration of the corpus-linguistic methods that are applied. The methodological sections draw on ideas developed in Gries and Hilpert (2008, 2010), Hilpert and Gries (2009), and Hilpert (2011). In these references, a number of standard analytical procedures are adapted for the use in diachronic corpus linguistics, which imposes special restrictions on the handling and interpretation of quantitative data. Beyond applying these ideas to new kinds of data, the present chapter also explains why, on a theoretical level, these methods are particularly suitable for the study of constructional change.

Sources of data

This section offers a short overview of the data sources that are used in this book. While all sources represent mostly written English, they differ with regard to variety, register, and genre. The choice of these sources is partly motivated by necessity, as some research questions require corpora with grammatical or sociolinguistic annotation, corpora that are of sufficient size, or corpora that cover a certain time span. At the same time, the selection also shows how broad a spectrum of resources is available these days for diachronic corpus studies. With access to these resources, many different analytical approaches to constructional change are feasible that would not have been practicable only a few years ago.

The previous chapters have investigated constructional change at the levels of allomorphy and word formation; the present chapter turns to a higher level of grammatical organization and focuses on syntax. Whereas Croft (2007: 504) still remarks that “relatively few studies have taken a Construction Grammar approach to syntactic change,” recent years have seen an upsurge in constructional studies addressing the diachrony of syntactic patterns (de Vogelaer 2008, Fried 2008, Hilpert and Koops 2008, Traugott 2008, Trousdale 2008a, Noël and van der Auwera 2009, Koops and Hilpert 2010, Patten 2010, amongst many others). The enterprise of diachronic Construction Grammar is thus in full swing, especially with regard to the analysis of syntactic or morphosyntactic phenomena, and there are even a number of meta-theoretical reflections on the constructional approach to the diachrony of syntax (Noël 2007, Trousdale 2010, Fried in press).

In the studies mentioned above, the constructional approach is typically wedded to the perspective of grammaticalization theory (Hopper and Traugott 2003). The introduction of this book already made the point that the subject matter of grammaticalization studies and diachronic constructional approaches is nearly coextensive. In both strands of research the focus is on grammatical changes that affect individual form-meaning pairs, typically changes that represent standard grammaticalization processes such as uni-verbation, semantic bleaching, host-class expansion, reanalysis, and the like. In keeping with the approach that was taken in earlier chapters, the purpose of the present chapter is to illustrate and analyze constructional changes that are not readily subsumed under the heading of grammaticalization, either because they involve processes that are not part of the standard inventory of grammaticalization, or because they combine several standard types of change in ways that are not predicted, or easily accommodated, by grammaticalization theory.

This chapter turns to processes of word formation and to the question of how changes in derivational morphology can be analyzed from the perspective of Construction Grammar. Booij (2010) has recently developed a constructional approach to morphology that holds a number of important implications for the present work. Central to a constructional understanding of derivational morphology is the idea that word formation processes are represented as cognitive schemas. It is thus assumed that speakers generalize over sets of words such as baker, buyer, runner, seller, and speaker and arrive at the schema shown in (1), which allows them to produce new coinages.

(1) [[x]V er]N ‘one who Vs’ (Booij 2010: 2)

Morphological schemas are conceptually similar to word formation rules (Plag 2003: 30), but defining these schemas as constructions facilitates in several ways the description of their properties. First, word formation processes exhibit multiple restrictions in terms of form and function of their constituents; these can be inscribed directly in the constructional poles of structure and meaning. Having rich, frequency-sensitive information on form and meaning available within the schema can explain prototype effects, such as the intuition that runner is a better example of the schema in (1) than stander ‘one who stands.’ Second, re-casting word formation rules as constructions allows an analysis of the interrelations between different word formation processes, notably between schemas and their subschemas, between schemas that unify to create a new word formation process, and between paradigmatically related schemas (Booij 2010: 50). Each of these relations is briefly discussed below.

It was the aim of this study to come to terms with the notion of constructional change and to assess whether a constructional perspective on diachronic processes can usefully complement existing approaches to language change. Having discussed the developments of several grammatical constructions at different levels of structural organization, it is now the time to take stock of the empirical observations and to assess how the idea of constructional change can actually lead to a deeper understanding of diachronic linguistic processes. The following sections present arguments in support of this point and outline a few issues to be addressed by future research.

Constructional change and grammaticalization

A result that recurred in each of the three case studies is that many processes of change that constructions undergo fall outside the scope of what is profitably viewed as grammaticalization. In each case, some processes occur that are indeed commonly observed in grammaticalization, but at the same time, there are other processes that would not be expected.

In the development from mine to my, the gradual replacement of a form with an alternative that has less phonetic material and that spreads across more and more different contexts is fully consonant with a scenario of gram-maticalization. What is not as easily explained, however, is why two different forms, mine and thine, undergo developments that are parallel to such an extent that no behavioral differences between them can be statistically established. It is here that a constructional perspective provides helpful analytical tools. If the two forms develop in lock-step, this means that speakers have formed a generalization across both of them: both instantiate the same construction.

Constructions and language change

This book outlines an approach to language change from the perspective of Construction Grammar. The idea of ‘constructional change’ will be developed as a new way of looking at diachronic processes in language. The following chapters present several case studies from the history of English in order to illustrate what processes of change come together in the development of grammatical constructions, and how the adoption of a constructional perspective can deepen our understanding of grammatical change. Naturally, an undertaking of this kind has to justify why a new term – constructional change – is introduced to group together formal and functional processes of change that affect constructions. As will be discussed, many of these are in fact quite well understood individually, and some of them are already grouped together under the heading of grammaticalization (Hopper and Traugott 2003). In what cases should we speak of constructional change rather than use established terminology? This book is an attempt to formulate a comprehensive answer to this question and to offer example analyses in which the theoretical concept of a changing, developing construction yields a shift in perspective, thus revealing issues that were previously not amenable to analysis.

The general idea of constructional change is of course not new. In fact, it has gained fairly wide currency over the past years and work on its theoretical status, similar in spirit to the present study, is underway (Bybee 2010, Trousdale 2010). Much recent work on language change, especially research with a functional orientation, focuses on the level of grammatical constructions (Bergs and Diewald 2008, Traugott and Trousdale 2010).

The study of linguistic units below the structural level of the word has not received very much attention within the Construction Grammar enterprise. Despite the principal commitment “to account for the entirety of each language” (Kay and Fillmore 1999: 1), studying the structure and meaning of syntactic units has thus far been the main research objective. Due to the influential work of Goldberg (1995, 2006), the focus within syntax has furthermore been on argument structure constructions, which represent linguistic generalizations that are much more abstract than constructions that specify information at the level of actual sounds. Analyses that apply Construction Grammar to lower levels of structural organization have only recently begun to emerge. For instance, Booij (2010) offers a comprehensive approach to morphological constructions that will be of particular concern to the discussion in Chapter 4. In the present chapter, it will be explored whether a constructional perspective is also fruitful for the analysis of change at the level of word-internal structure.

The specific phenomenon of linguistic organization that will be tackled in this chapter is allomorphy, that is, the variation between different realizations of the same morpheme (Plag 2003: 27). Typically, but not necessarily, the choice between two allomorphs depends on the phonological context. Framed in a constructional perspective, allomorphy can be understood as phonemic variation at the sound pole of a construction. To illustrate, a standard example for allomorphy in English is the variation between the realizations of the indefinite determiner, that is, [ə], [ən], and the emphatic variant [eI].