by Tony McEnery and Andrew Hardie; published by Cambridge University Press, 2012

Ethics and respondents

For corpora like the spoken component of the British National Corpus, we often draw a distinction between respondents – the people recruited by the corpus builders to collect data by going about their daily lives with tape recorders – and participants – which includes anyone whose spoke on the tapes as well as the respondents themselves.

The respondents carrying tape recorders around had to give informed consent – they had to understand that whatever they said when that tape recorder was running might well eventually be used in a corpus that would be widely available. They were sacrificing their privacy.

But another ethical issue relates to the people the respondents spoke to on tape. These other participants also sacrificed their privacy and it was therefore necessary for them to give informed consent. This consent was collected by the respondents themselves; so we see that in this case the process of data collection has actually imposed an ethical obligation on the respondents. Respondents are not researchers and cannot be expected to accept all the ethical obligations that are placed on researchers. Arguably, when spoken data collection imposes such obligations on respondents, it only makes the researchers’ responsibility of ethical oversight that much more critical. If the researchers have any doubt about whether consent procedures have been followed correctly by any respondent, the resulting data cannot ethically be included in a corpus.

A less obvious factor is the privacy of people talked about in the corpus. Their privacy is also arguably breached at times, and it is difficult or impossible to get their informed consent. So everyone’s privacy should be protected by changing all names to anonymise the data.

We can easily imagine serious consequences arising from a failure to take privacy seriously in the construction of a corpus. For example, distributing a spoken corpus in which accusations of criminal behaviour are made by one or more speakers might easily lead to a person being investigated by the police, if care is not taken with anonymisation.


This page was last modified on Monday 16 April 2012 at 10:09 am.

Tony McEnery Andrew Hardie

Department of Linguistics and English Language, Lancaster University, United Kingdom