I clearly remember turning in my honours thesis for my undergraduate degree at university back in 2005, and the hours I had slaved over my Matlab script so that I could generate some sort of useful data.

I also remember feeling somewhat strange that submitting the code itself along with my thesis paper was not a requirement. Feeling it deserved some attention (and to make my paper look thicker), I attached a print out of my matlab code as an appendix.

Now I'm a volunteer at an EEG lab, and I'm constantly exposed to scripts written by various scientists in labs around the world. As I re-write many of these programs, I have started getting that familiar feeling - the actual code behind much of today's tech-dependent scientific research may be dangerously hidden from the spotlight.

Behavioral neuroscience is a particularly good example, for much of the data used in published papers is generated from a human subject's interactions with customized software. These published papers rarely contain the script used to code the software, and instead describe the task as it was intended by the researchers.

The reasons for the code's omission are logical - it costs a lot of money in working hours to generate these scripts, and any use of them should involve compensation to the authors.

The problem: What happens if these programs contain bugs? Software companies can deal with bugs by developing patches, or simply ignoring them if the experience for the end user is unchanged. In science though, a particular piece of software may be run once and never used again. If bugged, the published data might be used for further research by many different labs. Moreover, many labs share their code between researches with similar interests, which means the bugged code might very well "infect" other research.

Has the peer-review system adapted to the increasing dependence error-free programming? It only spells more work and cost for everyone, so I get the feeling nobody is in a rush to do it.

Submit Post