Bringing the corpus back to life? Reflections on McCarthy on spoken grammar

Mike McCarthy gave the opening plenary at English UK Academic (North) conference last Saturday (Oct 4 2014), revisiting the topic of the grammar of spoken English. Spoken language, he pointed out, is in no way an imperfect, poorly realized version of the written form. Quite to the contrary, the written form is an imperfect attempt at abstracting from the actual, living, functioning reality of the spoken form. The wisdom of teaching the spoken form has a much older pedigree than most of us suspect, going back at least as far as latin teaching in the Tudor period. It was only during and after the reign of Henry VIII that language teaching became based on the prescriptive, written version of grammar which has held sway well into the 20th century and beyond.

He pointed out the ‘paratactic’ nature of spoken English, ie, the fact that it is made up as we go along in time, not planned in to conform with some kind of hierarchical structure as many grammarians would have it. It’s nature, in other words is linear and not some kind of tree. The consequences of this have not been easy to discern prior to the emergence of the spoken corpus, but now a wealth of insights have emerged:

• most language is co-constructed by two or more speakers in the course of conversation
• grammar based on written language is at odds with the facts of speech. For example, in speech, subordinate clauses are often free-standing, eg ‘If you’d like to wait here, please’.
• The tense/aspect system does not work as pedagogic grammars have suggested, for example, ‘We were wanting to book a trip to Italy’ is normal, not an anomaly.
• Headers and tails, considered bad form in writing, are normal in speech, for example, ‘Nice work, that is’.
• Ellipsis may be regarded as normal, and the non-elided form is an elaboration to make up for lack of context in the written form.

Reflections Insights from corpus linguistics such as the above are invaluable for ELT practitioners, and materials writers need to be especially aware of them. However, I don’t think the classroom implications are as direct as some would have us believe. For instance, must we really use only corpus-attested examples in the classroom? McCarthy beamed up two greeting exchanges, one from a coursebook audio script and one from real life. The coursebook script jumped out as being artificial and comical by comparison to the real world transcript – an easy target for ridicule. Yet it was far more useful classroom material for lower levels than the real-life example. The real-life example, for instance, included small talk referring to the presidential elections. That’s only useful if you happen to have just had a presidential election at the time of speaking.

An absurd example, I know, but the more general point is that it may be useful for learners to take on board a more generalizable sequence of ‘moves’ for a greetings sequence, stripped of the accidental features of a given, corpus-attested example.

A second point is that many features of spoken language emerge accidentally, as a consequence of a given set of contraints. For example, in speech, a person may start saying something and then have second thoughts, backtrack and begin again – a reformulation. While this may be something that it is very appropriate to point out in the context of listening, it would hardly be appropriate in the context of teaching speaking. You don’t need to teach people how to make slips of the tongue, reformulations, stutters and so on – these come for free, caused by the same time real-time processing contraints as caused their corpus-attested originals.

In some ways, a corpus is a corpse. It is the trace of many pieces of language in given instances of use. The learner’s job is not to resurrect this corpse, enacting pieces of corpus as if they were scripts. Rather, their job is to make their own use of the language – thus leaving behind their very own corpus, if anybody cares to record what they say. As teachers, we want to help them upgrade their use, but I don’t accept that exposing them to only corpus-attested examples is necessarily the most effective way of doing this.

See a response to this post from Mura Nava here.