I received a question from a researcher this week asking whether or not eliciting grammatical judgments for syntactic research required IRB review. IRB review is only required for projects that are “research” (which this collecting elicitations is) and that include “human subjects’ defined in U.S. federal research regulations. Here is the definition of human subject:
“45CFR46.102 (f) Human subject means a living individual about whom an investigator (whether professional or student) conducting research obtains
(1) Data through intervention or interaction with the individual, or
(2) Identifiable private information.
Intervention includes both physical procedures by which data are gathered (for example, venipuncture) and manipulations of the subject or the subject’s environment that are performed for research purposes. Interaction includes communication or interpersonal contact between investigator and subject. Private information includes information about behavior that occurs in a context in which an individual can reasonably expect that no observation or recording is taking place, and information which has been provided for specific purposes by an individual and which the individual can reasonably expect will not be made public (for example, a medical record). Private information must be individually identifiable (i.e., the identity of the subject is or may readily be ascertained by the investigator or associated with the information) in order for obtaining the information to constitute research involving human subjects.”
Based on this, would you conclude that eliciting grammatical judgments constitutes “human subjects” research? If no, how would you convince your IRB that this type of research does not need review? And then, how about collecting speech samples for phonetic analysis?
anggarrgoon
cahillm
drword563
ejfranci
jacrippen
Jack Martin
Jeff Good
jpboyle
posted by LSA Ethics Committee
natashawarner
speer2014
Sylvia Reed
tanyamatthews
LSA Ethics Discussion Blog 
I’ve come across two lines of argument here.
The first is that such proposals are subject to review because they involve human participants in an experiment designed to elicit generalizable information about human behavior. This makes grammaticality judgment tasks a type of psychological research.
The second is that these are tasks designed to get information about an abstract system where there is no personal identifiable information about the human participant; they are providing information about language, not undergoing any intervention.
In the first case, the IRB decides the proposal is subject to expedited review (or exemption following a brief review); in the second, it’s decided that the research does not fall within the bounds of IRB review.
It strikes me that either of these views could be argued for.
Grammatical judgements are typically presented in an anonymous fashion, and often the researcher doesn’t bother to keep track of which speaker a judgement came from. Thus they can’t constitute “private information” under this definition. That leaves only the “data through intervention or interaction with the individual” aspect to be argued.
You’re right about “intervention” not applying, so the only part of the definition that matters is “data through interaction with the individual”. Now, the way that data are obtained from individuals will vary from project to project, but from what I’ve seen the process is usually pretty dull and unthreatening. Usually just describing the manner of how the data are obtained is enough to convince the IRB that there’s no review required. The only thing that they might seriously be concerned about is what the reward for participation is, and how it’s conferred.
In either case, an IRB would be draconian if it required such a project to undergo anything more extensive than expedited review. If similar projects have been done by the department in the past, then it’s usually just a matter of attaching the IRB’s prior consideration and claiming that the new project is substantially similar. This oftentimes results in an exemption.
Now, as for speech samples, there’s a slight difference. It’s quite easy for people to identify speakers from a single short utterance. In one phonetics class of mine, it was instantly obvious that a particular young female speaker was the daughter of a professor in our department, whom I’d only met twice. Given this, the IRB is going to have to be more restrictive, and even if they aren’t the researcher still has some ethical issues to deal with.
The simplest solution is to restrict access to the recordings, however this is rather unsatisfying. If the research language has a large population, it’s easy enough for other researchers to replicate the data and so restricting access is not really a problem. But if the language isn’t widely spoken then data access is crucial for verifiable replication of measurement. So the data need to be public.
Interestingly, there’s an opposite balance to this issue. If the language is widely spoken then it may be more likely for an individual to be recognized by others, since the individual probably knows more people being from a large population. In contrast, if the language is not widely spoken then the individual is not likely to be known by more than a few people, although it’s much more likely that other speakers will know the individual.
Since recordings for phonetics are oftentimes rather innocuous, the issue of possible identification of anonymous participants is usually unimportant. Nobody particularly cares that the participant said “a merry melon a mile long” five times. The only people who might have serious concerns are those who are worried about their privacy, and if they are fully informed about access to the data then they’ll remove themselves from the project. So the important issue for the IRB then is that withdrawal from participation is made clear for the participants. The same is true for participants who speak languages with small populations.
There’s also the complicated situation where a researcher is working with a community who barely understands electricity, much less something like scientific research, the Internet, and so forth. But that’s a fundamentally different problem that I’ll leave to people who’ve worked in such situations.
[BTW, I’d really like it if there was a way to preview comments before submitting them…]
James brings up an interesting point about identification of language users. If the elicitation is transcribed, e.g., for publication or analysis, there is much less chance of speaker identification, but it still isn’t impossible. Suppose, for example, that I ask a language user to use a word in a sentence. If the resulting sentence is private information about the user’s family, then it might be easier in context to figure out who uttered the sentence. Making the actual data public is fraught with issues. And if you use video rather than just audio, the problem is compounded.