The aim of this study is to further contribute to the ongoing debate regarding the nature of “morphosyntactic lifespan change,” defined here as observable shifts in the grammatical choices individuals make between competing morphosyntactic structures. Through a quantitative case study of competition between two types of ing-nominals in seventeenth-century English, in which we factor in the grammatical contexts in which the variant structures can be used, we show that individuals vary in the extent to which they participate in the contextual diffusion of a new structure. We furthermore show that there is interindividual variability with respect to whether and what kind of lifespan change—frequency, constraint, and inventory change—is attested and highlight different patterns of intraindividual change: progressive, retrograde, and “mixed.”

A fundamental problem in research into language and cultural change is the difficulty of distinguishing processes of stochastic drift (also known as neutral evolution) from processes that are subject to selection pressures. In this article, we describe a new technique based on deep neural networks, in which we reformulate the detection of evolutionary forces in cultural change as a binary classification task. Using residual networks for time series trained on artificially generated samples of cultural change, we demonstrate that this technique is able to efficiently, accurately and consistently learn which aspects of the time series are distinctive for drift and selection, respectively. We compare the model with a recently proposed statistical test, the Frequency Increment Test, and show that the neural time series classification system provides a possible solution to some of the key problems associated with this test.

In one way or another, historical linguists have always been aware of the limitations inherent to working with linguistic data from bygone ages. One of the most substantial limitations, as Petré points out, is that all speakers of a historical variant of a language are unavailable for psycholinguistic study, essentially leaving researchers with their written records as the sole data source. As such, historical linguists often find themselves taking the role of corpus linguists, trying to understand the workings of a language ‘by studying aggregate data that pools the productions of many speakers and writers – often across different media, genres, registers, and even across different time periods’ (Arppe et al.2010: 3). As Petré points out, the practice of studying language on this aggregate level has dominated the methodologies in historical linguistic studies, and very little attention is paid to the individual level.

The present article offers a corpus-based analysis of the diachronic development of the usage profiles of three adverbial non-finite clauses in Modern English: the free adjunct, the verbal gerund and the absolute construction. By treating present-participial adverbial clauses and adverbial gerunds as part of a single adverbial ing-clause network, this article sheds new light on the different semantic and functional-pragmatic factors motivating the formal variation within the ing-clause network. By means of two mixed-model logistic regression analyses, we determine the relative impact of the independent variables of adverbial semantics, position, degree of coreference and length on the language user's choice in (i) whether or not to include augmentation (syndesis) and (ii) whether or not to include an overt subject in the adverbial ing-clause. The resulting picture is one of an emerging adverbial ing-clause network in which the internal variation is determined by principles of processing complexity.