International Journal of Corpus Linguistics - Volume 18, Issue 1, 2013

This study discusses an adverbial pattern which has so far been largely overlooked, namely ADV1 and ADV1, as in again and again, on and on and over and over. The paper is primarily based on the Corpus of Historical American English (COHA) and the Corpus of Contemporary American English (COCA). The data show that these patterns follow typical paths of change, such as a movement towards more abstract meanings (metaphorization; over and over increasingly referring to repetition rather than to physical motion), lexicalization (e.g. up and up being used as a noun with idiosyncratic meaning in on the up and up), subjectification (e.g. on and on expressing negative connotations), iconic variation (again and again and again referring to multiple repetitions), simplification (loss of again after over and over), and the development of discourse functions (and on and on meaning “and so on”).

This paper addresses relations between lexico-grammatical patterns and texts. Our focus is on a specific linguistic unit, the ‘suspended quotation’ (or ‘suspension’), which has received particular attention in Dickens studies. The suspended quotation refers to an interruption of a fictional character’s speech by the narrator with a sequence of at least five words. We show how corpus linguistic methods can help to systematically study suspensions in a corpus of Dickens’s novels: we investigate relationships between patterns of body language presentation and suspensions; we consider the distribution of suspensions across novels; and we illustrate how patterns in suspensions relate to meanings of reporting verbs. Overall, we argue that suspensions are discernible units that contribute to meaningful patterns in narrative prose.

The purpose of the present paper is primarily methodological. It provides evidence that a research question investigated in a corpus can produce different results according to the software adopted, queries the comparability of the results, and shows that the whole picture is more than one piece of software can reveal. The field of research is a corpus of Letters to Shareholders referring to 2008, the year the current financial crisis began, compared to identically constructed corpora from the same companies, referring to 2006 and 2010. The paper argues that the macro analysis of key words and key semantic domains needs to be complemented by close phraseological analyses, since it is in the phraseology of the texts that the nuances of a message are to be found.

Formulaic sequences are recognised as having important roles in language acquisition, processing, fluency, idiomaticity, and instruction. But there is little agreement over their definition and measurement, or on methods of corpus comparison. We argue that replicable research must be grounded upon operational definitions in statistical terms. We adopt an experimental design and apply four different corpus-analytic measures, variously based upon n-gram frequency (Frequency-grams), association (MI-grams), phrase-frames (P-frames), and native norm (items in the Academic Formulas List – AFL-grams), to samples of first and second language writing in order to examine and compare knowledge of formulas in first and second language acquisition as a function of proficiency and language background. We find that these different operationalizations produce different patterns of effect of expertise and L1/L2 status. We consider the implications for corpus design and methods of analysis.

While lexical bundles research identifies continuous sequences (e.g. the end of the, I don’t know if), researchers have also been interested in discontinuous sequences in which words form a ‘frame’ surrounding a variable slot (e.g. I don’t * to, it is * to). To date, most research has focused on a few intuitively-selected frames, or has begun with frequent continuous sequences and then analyzed those to identify associated frames. Few previous studies have attempted to directly identify the full set of discontinuous sequences in a corpus. In the present study, we work towards that goal, using a corpus-driven approach to identify the set of recurrent four-word continuous and discontinuous patterns in corpora of conversation and academic writing. This direct computational analysis of the corpora reveals a more complete set of frames than alternative approaches, resulting in the documentation of highly frequent frames that have not been identified in previous research.

This paper explores ways in which research into collocation should be improved. After a discussion of the parameters underlying the notion of collocation, the paper has three main parts. First, I argue that corpus linguistics would benefit from taking more seriously the understudied fact that collocations are not necessarily symmetric, as most association measures imply. Also, I introduce an association measure from the associative learning literature that can identify asymmetric collocations and show that it can also distinguish collocations with high and low association strengths well. Second, I summarize some advantages of this measure and brainstorm about ways in which it can help re-examine previous studies as well as support further applications. Finally, I adopt a broader perspective and discuss a variety of ways in which all association measures – directional or not – in corpus linguistics should be improved in order for us to obtain better and more reliable results.