Researchers at Penn and Columbia were surprised to discover earlier this month that their study of racial bias had triggered a debate about another issue: When is deception appropriate in academic research?

Wharton's Katherine L. Milkman and Columbia's Modupe N. Akinola sent e-mails to 6,300 professors at top institutions. The messages, purporting to be from a student interested in their work, asked for a 10-minute meeting - half said quickly (to test an impulsive response), the other half said next week (to assess a thought-out response).

The students' names also varied to reflect sex, ethnicity, and race (Lamar Washington vs. Brad Anderson).

Those professors who responded (67 percent) got an immediate "debriefing" e-mail that came clean on the deception, apologized for any inconvenience, assured confidentiality, and explained the purposes of the research.

Not everyone appreciated the effort.

This study "belongs in the trash heap of ill-advised research projects," wrote Columbia statistics and political science professor Andrew Gelman on his blog, sparking a lively debate among academics. Gelman softened his words in later posts but still believes that subjects should have been asked beforehand whether they wanted to participate, and compensated (he suggests $10) if they did.

Milkman says that seeking prior consent - even with vague wording and well in advance, as Gelman proposes - would skew the findings based on who opted in. She never thought about using the departmental research money that funded the study to compensate professors for reading a short e-mail.

Unlike on Gelman's blog, the vast majority of responses to the authors were neutral or positive, she said. But she was surprised that nearly 30 percent were negative and a small number were angry; three used four-letter words.

Milkman has a knack for finding unusual research topics: While a Princeton undergraduate, she analyzed how the race and gender of New Yorker fiction editors affected the storylines and character demographics of what they published.

But she had never previously used deception, which can raise ethical concerns.

All studies involving human subjects must be approved by universities' Institutional Review Boards (IRBs) following detailed federal guidelines. Deception is allowed under certain circumstances. Among the requirements: that there is no other practicable way to conduct the study and that participants face no more than minimal risk of harm.

Deception tends to be used in studies of bias - whether employers are more likely to offer a job to a (fake) man, for example, than to a (fake) woman - and IRBs at Penn and Columbia both approved this study.

So why did it provoke controversy even before publication?

It may be that the authors of similar research never debriefed participants: They never knew they had been in a study, let alone duped.

And academics themselves generally have not been subjects of such bias studies. Milkman believes that will make the findings from this one - expected in several months - particularly significant.