Saturday, October 30, 2010

Marc Hauser: Plot Thickens, Solidifies, Cracks?

By now everyone's heard of the story of Marc Hauser, a well-known Harvard psychologist who's been suspended for research misconduct involving monkeys.

Until a few days ago, the situation seemed fairly clear: Hauser was guilty of several counts of misconduct, according to a Harvard committee, and there was no reason to doubt that judgement. Although the details of the investigation were never made public, it was generally believed that he'd fabricated the data from at least one experiment, as I explained previously.

However, all that was thrown into question on Monday by an article in the New York Times: Difficulties in Defining Errors in Case Against Harvard Researcher. The author, respected science journalist Nicholas Wade, writes that there's more to the story than first appeared, and specifically, that Hauser may not have fabricated data, instead being the victim of an innocent (if serious) mistake:
[A paper Hauser recently retracted], published in 2002, reported that rhesus monkeys can distinguish a novel string of sounds from a control sequence, an issue which has important bearing on their capacity for language. The novel and control sound sequences must be alternated... But the video of the experiment contains only novel sequences.

Critics like Dr. Altmann at first charged that the controls had never been done, and that since control conditions are reported in the paper, they must have been concocted. But Altmann... now says his earlier accusation was “heavily dependent on the knowledge that Harvard found Professor Hauser guilty of misconduct.” When he gave the issue further thought, he saw an alternative explanation.

In the experimental setup, the monkey is in a soundproof box. The researchers can see the computer is playing a sound but cannot hear it. What could have happened is that the computer, through a programming error, substituted a second test sound for the control sounds, and the researchers, unaware of the problem, wrote up their report assuming the control sounds had been played...

Even so, it is far from clear how the data on the video led to the reported results. This would be a devastating error, but not fraud. “It is conceivable that the data were not fabricated, but rather that the experiment was set up wrong, and that nobody realized this until after it was published,” Dr. Altmann wrote.

Wade also quoted two former students of Hauser's who praised his "unimpeachable scientific integrity” and who said his critics were “scholars known to be virulently opposed to his research program”, and quotes an anonymous Harvard academic as saying the investigation was "lawyer-driven", unnecessarily long, and unfair to Hauser.

But yesterday Gerry Altmann, the Editor of the journal Cognition which published the retracted paper, hit back against Wade in a blog post, saying that Wade "selectively quoted" him to give the impression that he'd backtracked from his earlier conclusion that Hauser falsified the data.
...there has been no stepping back. As I make very clear... the information I have received, when taken at face value, leads me to maintain my belief that the data that had been published in the journal Cognition was effectively a fiction - that is, there was no basis in the recorded data for those data. I concluded, and I continue to conclude, that the data were most likely fabricated...

It is true that I did write here that there existed an alternative explanation for what happened, based on a sequence of errors. However, for that interpretation to be correct ... the information I had been given, by Harvard’s Dean, would have to have been incorrect.
Essentially, Altmann says that while in theory Hauser could have made an innocent mistake, Harvard's investigation specifically ruled out this and concluded that no innocent explanation was possible.
So at the end of the day, it comes down to this: Do I believe what the Dean [of Harvard] told me were the results of a long, careful, and painstaking investigation, or do I simply make up a “Just So Story” instead?...

This entire saga is about the misrepresentation of truth. It is ironic that the journalists who profess to expose truth place such little value in it.
What are we to make of all this? The issue is extremely important - the "fabrication" of data in the Cognition paper was the most serious allegation against Hauser, and (to my knowledge) the only thing which proved that his misconduct was deliberate as opposed to sloppy.

The crucial question therefore is whether the Harvard investigation was right to rule out an innocent explanation of the Cognition data. Altmann correctly says that either Harvard are wrong, or Hauser falsified data.

But the problem is that the details of Harvard's judgement remain private. So we (including Altmann) seem to be left with a question of whether to trust Harvard University and their internal investigation.

No comments:

Post a Comment