This article was originally posted on Betsy Levy Paluck's website.
On his terrific blog, Professor Sanjay Srivastava points out that the current (vitriolic) debate about replication in psychology has been "salted with casually sexist language, and historically illiterate" arguments, on both sides. I agree, and thank him for pointing this out.
I'd like to add that I believe academics participating in this debate should be mindful of co-opting powerful terms like bullying and police (e.g., the "replication police") to describe the replication movement. Why? Bullying behavior describes repeated abuse from a person of higher power and influence. Likewise, many people in the US and throughout the world have a well-grounded terror of police abuse. The terror and power inequality that these terms connote is diminished when we use it to describe the experience of academics replicating one another's studies. Let's keep evocative language in reserve so that we can use it to name and change the experience of truly powerless and oppressed people.
Back to replication. Here is the thing: we all believe in the principle of replication. As scientists and as psychologists, we are all here because we wish to contribute to cumulative research that makes progress on important psychological questions. This desire unites us.
So what's up?
It seems to me that some people oppose the current wave of replication efforts because they do not like the tenor of the recent public discussions. As I already mentioned, neither do I. I'm bewildered by the vitriol. Just a few days ago, one of the most prominent modern economists, currently an internationally bestselling author, had his book called into question over alleged data errors in a spreadsheet that he made public. His response was cordial and curious; his colleagues followed up with care, equanimity, and respect.
Are we really being taught a lesson in manners from economists? Is that happening?
As one of my favorite TV characters said recently ...
If we don't like the tenor of the discussion about replication, registration, etc., let's change it.
In this spirit, I offer a brief description of what we are doing in my lab to try to make our social science rigorous, transparent, and replicable. It's one model for your consideration, and we are open to suggestions.
For the past few years we have registered analysis plans for every new project we start. (They can be found here on the EGAP website; this is a group to which I belong. EGAP has had great discussions in partnership with BITSS about transparency.) My lab's analysis registrations are accompanied by a codebook describing each variable in the dataset.
I am happy to say that we are just starting to get better at producing replication code and data & file organization that is sharing-ready as we do the research, rather than trying to reconstruct these things from our messy code files and Dropbox disaster areas following publication (for this, I thank my brilliant students, who surpass me with their coding skills and help me to keep things organized and in place. See also this). What a privilege and a learning experience to have graduate students, right? Note that they are listening to us have this debate.
Margaret Tankard, Rebecca Littman, Graeme Blair, Sherry Wu, Joan Ricart-Huguet, Andreana Kenrick (awesome grad students), and Robin Gomila and David Mackenzie (awesome lab managers) have all been writing analysis registrations, organizing files, checking data codebooks, and writing replication code for the experiments we've done in the past three years, and colleagues Hana Shepherd, Peter Aronow, Debbie Prentice, and Eldar Shafir are doing the same with me. Thank goodness for all these amazing and dedicated collaborators, because one reason I understand replication to be so difficult is that it is a huge challenge to reconstruct what you thought and did over a long period of time, without careful record keeping (note: analysis registration also serves that purpose for us!).
Previously, I posted data at Yale's ISPS archive, and for other datasets made them available on request if I thought I was going to work more on them. But in future we plan to post all published data plus the dataset's codebook. Economist and political scientists friends often post to their personal websites. Another possibility is posting in digital archives (like Yale's, but there are others: I follow @annthegreen for updates on digital archiving).
I'm interested in how we can be better. I'm listening to the constructive debates and to the suggestions out there. If anyone has questions about our current process, please leave a comment below! I'd be happy to answer questions, provide examples, and to take suggestions.
It costs nothing to do this--but it slows us down. Slowing down is not a bad thing for research (though I recognize that a bad heuristic of quantity = quality still dominates our discipline). During registration, we can stop to think-- are we sure we want to predict this? With this kind of measurement? Should we go back to the drawing board about this particular secondary prediction? I know that if I personally slow down, I can oversee everything more carefully. I'm learning how to say no to new and shiny projects.
I want to end on the following note. I am now tenured. If good health continues, I'll be on hiring committees for years to come. In a hiring capacity, I will appreciate applicants who, though they do not have a ton of publications, can link their projects to an online analysis registration, or have posted data and replication code. Why? I will infer that they were slowing down to do very careful work, that they are doing their best to build a cumulative science. I will also appreciate candidates who have conducted studies that "failed to replicate" and who responded to those replication results with follow up work and with thoughtful engagement and curiosity (I have read about Eugene Caruso's response and thought that he is a great model of this kind of response).
I say this because it's true, and also because some academics report that their graduate students are very nervous about how replication of their lab's studies might ruin their reputations on the job market (see Question 13). I think the concern is understandable, so it's important for those of us in these lucky positions to speak out about what we value and to allay fears of punishment over non-replication (see Funder: SERIOUSLY NOT OK).
In sum, I am excited by efforts to improve the transparency and cumulative power of our social science. I'll try them myself and support newer academics who engage in these practices. Of course, we need to have good ideas as well as good research practices (ugh--this business is not easy. Tell that to your friends who think that you've chosen grad school as a shelter from the bad job market).
I encourage all of my colleagues, and especially colleagues from diverse positions in academia and from underrepresented groups in science, to comment on what they are doing in their own research and how they are affected by these ideas and practices. Feel free to post below, post on (real) blogs, write letters to the editor, have conversations in your lab and department, or tweet. I am listening. Thanks for reading.
A collection of comments I've been reading about the replication debate, in case you haven't been keeping up. Please do post more links below, since this isn't comprehensive.