Exploring the boundaries of formulaic sequences : a corpus-based study of lexical substitution and insertion in contemporary British English.

This thesis presents an investigation into formulaic sequences; namely multi-word
prefabricated phrases of either literal (e.g. good morning) or non-literal (e.g. kick the
bucket) reading. A property of such sequences is variation. Formulaic sequences can be
subject to varying degrees of lexical substitutions, grammatical variations, and insertions.
This thesis investigates the boundaries of variation: the limits of lexical substitution and
insertions for formulaic sequences, i.e. how much variation can occur before the sequence
stops being fixed and becomes context-dependent. The boundaries between one formulaic
sequence and another and the boundaries between a literal and non-literal reading are also
explored.
The formulaic sequences for investigation were chosen from The Longman Idioms
Dictionary (1998) and were explored using the British National Corpus (BNC). To
investigate the limits of variation, I developed and used a technique that I term the chaining
process. This is a systematic method of searching for sequences to find the maximum
lexical substitutes and insertions. The frequencies of variant forms found during the study
were recorded and analysed to highlight both so that both common and rare lexical
substitutions and insertions could be examined, and their limits explored.
A result of using the chaining process was that sequences could be seen to "link" together.
Formulaic sequences with the same underlying meaning and similar lexical set were found
to form groups. Use of the chaining process showed how different formulaic sequences
with similar meanings could link together in networks via common lexical substitutes, e.g.
flip your lid and blow your top link viaflip your wig - lose your wig -w lose your temper
and blow your temper. The use of the chaining process shows that formulaic sequences are
more similar than different in terms of semantics as well as construction. Sequences are not
autonomous; networks show that the boundaries of sequences are not as fixed as idiom
dictionaries may lead us to believe. These phrasal networks formed via the chaining process
provide a regular method of grouping formulaic sequences. This technique and analysis
contribute to lexicography and inform cognitive models of storing and organizing language.