This chapter begins with a general discussion of potential data types in variationist linguistics. Next, we present the two main data sources we use in the study: the International Corpus of English (ICE) and the Global Corpus of Web-Based English (GloWbE). The former comprises a set of parallel, balanced corpora representative of language usage across a wide range of standard national varieties. Each ICE corpus contains 500 texts of 2000 words each, sampled from twelve spoken and written genres/registers, totaling approx. 1 million words. GloWbE contains data collected from 1.8 million English language websites – both blogs and general web pages – from twenty different countries (approx. 1.8 billion words in all). Discussion of the corpora is followed by a detailed description of the data collection, identification, and annotation procedures for our three alternations. Here we carefully define the variable context for each alternation, and outline the methods for coding various linguistic constraints that are included in our analyses.

This chapter interrogates corpus data to analyze the three alternations subject to study in this book one by one using a battery of state-of-the-art analysis techniques in addition to customary descriptive statistics, Conditional Random Forest (CRF) modeling and mixed-effects logistic regression analysis. The goal of the chapter is to uncover qualitative generalizations: for example, we see that while effect directions of constraints on variation are generally stable across varieties of English, effects strengths can be significantly different.

This chapter is inspired by work in comparative sociolinguistics and quantitative dialectometry. We use a corpus-based method (Variation-Based Distance and Similarity Modeling – VADIS for short) to quantify the similarity between, and coherence across, the varieties of English under study as a function of the correspondence of the ways in which language users choose between different ways of saying the same thing. Key findings include the result that probabilistic grammars are remarkably stable across varieties but that coherence across alternations is not perfect.

This chapter surveys the literature on variation in general and on grammatical variables (a.k.a. "alternations'') in particular. Next, we review well-known grammatical variables/alternations in English as well as previous comparative investigations of grammatical alternations in English. Last but not least, we discuss in detail previous variationist work on the three alternations subject to study in this book: the genitive alternation, the dative alternation, and the particle placement alternation.

The present study explores the effect of speakers’ gender in the well-known dative alternation (e.g. Mary gives John an apple vs. Mary gives an apple to John) and weighs the impact of this language-external factor against language-internal factors such as length of the constituents or semantics of the verb. Following up on previous research that explored the dative alternation across nine varieties of English, the focus of the present work will be on Jamaican English, a variety where male and female speakers seem to use the two variants differently. 615 variable dative tokens of acrolectal Jamaican English speech were annotated for eleven language-internal and three language-external factors and subjected to conditional random forest and mixed-effects logistic regression analyses. The results of these analyses indicate that the predictor gender plays only a marginal role vis-à-vis other language-external and -internal constraints. At the same time, if only the two most important language-internal predictors are considered, gender turns out to significantly affect dative choice with male speakers preferring the prepositional variant more than female speakers. These results not only highlight the potential of syntactic alternations to serve as sociolinguistic variables but also point to possibly different social dynamics between male and female speakers in Jamaica.

There is an ongoing debate about the analysis of argument structure in a usage-based construction grammar. Some scholars have argued that argument structure is licensed by fully abstract schemas, but other researchers have claimed that argument structure is primarily determined by particular verbs. Chapter 7 argues that this controversy is easily resolved if we analyze argument structure in the framework of a network model in which verbs and constructions are interconnected by probabilistic links. For instance, the two constructions of the dative alternation occur with an overlapping set of verbs that are statistically biased to be used in one or the other constructions. The statistical biases can be analyzed as filler-slot associations that are shaped by two factors: (1) general conceptual processes of event semantics and (2) speakers’ experience with particular verbs and constructions. The analysis is supported by evidence from research on sentence processing and the extension of argument schemas to novel verbs in L1 acquisition and language change.

There is an ongoing debate about the analysis of argument structure in a usage-based construction grammar. Some scholars have argued that argument structure is licensed by fully abstract schemas, but other researchers have claimed that argument structure is primarily determined by particular verbs. Chapter 7 argues that this controversy is easily resolved if we analyze argument structure in the framework of a network model in which verbs and constructions are interconnected by probabilistic links. For instance, the two constructions of the dative alternation occur with an overlapping set of verbs that are statistically biased to be used in one or the other constructions. The statistical biases can be analyzed as filler-slot associations that are shaped by two factors: (1) general conceptual processes of event semantics and (2) speakers’ experience with particular verbs and constructions. The analysis is supported by evidence from research on sentence processing and the extension of argument schemas to novel verbs in L1 acquisition and language change.

Constituent structure is commonly represented in phrase structure trees, which one might analyze as some kind of network. However, while phrase structure graphs are useful to explain certain aspects of syntactic structure, they are not fully compatible with the dynamic network approach, as these graphs are built from primitive concepts. Challenging the traditional approach, Chapter 9 outlines a dynamic network analysis of constituency in which the hierarchical organization of linguistic structure is emergent from the interaction of several domain-general processes, including conceptualization, automatization, analogy and abstraction, which together account for both the syntactic generalizations that have been in the focus of formal syntax and the many idiosyncracies that reflect the influence of speakers’ experience with particular constructions. The proposed analysis casts a fresh light on syntactic constituency and has far-reaching implications for the theory of phrase structure and the analysis of word order correlations.