Where do we backchannel?
. Material
is work makes use of material taken from the spoken British English mod-
ule of the CobuildDirect Corpus. e module is named “ukspok” and contains
9,272,579 words. e transcribers have painstakingly recorded not only all the
words but also all the grunts and sounds occurring in a conversation.3 Pauses oc-
curring in the conversation are also marked. For a study of the present kind it has
one great drawback, however: unlike e.g. the London-Lund Corpus (see Alten-
berg 1991) it has no prosodic annotation, which, if given, would have disambigu-
ated a number of occurrences. One such use would have been the indication of
upward intonation to show solicitation of signs of agreement (Scheglo 1982:80).
is is one reason for being cautious here. Another is, in the words of McCarthy
(2002:69), that “[s]poken corpora as a locus for research into human communi-
cation always run the risk that features of talk may be culture-bound, and it is
only in intervarietal and interlingual studies that one can nd safer ground for
generalisations.”
As the eld of backchannels is somewhat indeterminate, and, as McCarthy
(2002:52) points out, listener responses are characterised by their scalar nature,
there was never any question of trying to cover the whole eld. Six of the most fre-
quent English backchannels were selected for close study, viz. (in Cobuild trans-
literation) Mhm, Mm, Right,Uh huh,4Yeah ,Yes . One thousand occurrences of
each of those particles were randomly extracted from the Corpus. Some of them
include duplicated tokens, as in
(2) <M0X> I don’t think you’d be likely to say I don’t like that project.
<M0X> Right. Right.
<M0X> But <ZF1> it’s <ZF0> it’s used I’m sure <ZF1> in er <ZF0> in
reference.
Duplicated tokens were treated like non-duplicated ones. Not every instance of the
particles was a backchannel, so an operational denition was applied to the mate-
rial. ose cases which followed the pattern “Speaker 1: xxxx — Speaker 2: Mhm/
Mm/Right/Uh huh/Yeah/Yes. — Speaker 1: xxxx” were accepted as backchannels.
(4), further below, is such an example. Speaker <M08>’s Yes, Yeah and Uh huh in
the following example were thus not regarded as backchannels:
(3) <M01> […] and you cannot compare
<M08> Exactly yes. Yes <ZF1> that’s <ZF0> that’s. Yeah. Uh huh. Yes
exactly. Yeah the your rst thought tonight when […]
Such cases were called “Unanalysed” and not further dealt with. One of the ad-
vantages of dening backchannels in such a strict fashion is that the problem of