Skip to content

Policies#

Scientific conduct#

Thanks to the replication and reproducibility crises, and the ever-growing awareness of the mis(use) of statistics, the scientific conduct of researchers is more scrutinised than ever. These are our basic policies for conducting science:

Experimental data are sacrosanct#

We do not alter the content of experimental data. There should always be a complete, untouched set of any data files we obtain. Never load the only copy of any data set. Any changes to the formatting of data (e.g. changing from arrays to tables) are saved as copies.

Simulations should be replicable#

Every simulation in a paper should be replicable in principle: the same code run again with the same parameters should produce the same results. Hint: always save parameters from any simulation code that will be used for a paper. That includes the random number generator seed.

Code should be consistent and reviewable#

As all our work depends on code it is in our interests to adopt good coding practices. The lab code policy covers how to write code and the lab code reviews. Call a review to have the lab check any research-critical code.

Document your work#

Use whatever works for you, but document your work. That could be a lab book, whether paper or tablet, with notes on each analysis or each simulation. Or typing notes. Or exporting key plots to Powerpoint, Google Docs etc. Or using a Jupyter notebook or RMarkdown to encapsulate what you’ve done, if you’re comfortable with those things. Do something, so that you can give accurate answers to: what have you been working on? And: what have you found?

Every file is duplicated#

Your computer will die. All working files on your computer should be replicated elsewhere, at all times. At minimum, use an automated program to back up to an external hard-drive. Better, use cloud-based storage - this will also let you sync work across machines. The default cloud storage at Nottingham University is OneDrive.

A Just Culture#

As scientists we are always learning: about new ways of analysing data, about our own findings, about new models, about new published research. And learning is inherently about making mistakes and errors, in order to correct our skills and knowledge. So the lab operates a “Just Culture”, not a blame culture: mistakes and errors are expected; mistakes are not blamed on the individual; and mistakes are used as opportunities for learning by the lab: how to adjust our approaches to catch errors in the future.

An example of Mark’s mistakes

Chapters 4 and 5 of my PhD thesis contain a spiking neuron model of the feedback loop between the globus pallidus (GP) and the subthalamic nucleus (STN). As thrilling as it sounds. This is a negative feedback loop: the GP inhibits the STN; the STN excites the GP. However, in the midst of writing up Chapter 4, I discovered my code had a tiny but crucial error: the STN input to the GP had a minus sign in front of it. It was inhibiting, not exciting, GP. Already past the end of my funding, I had to ditch all the simulations that were to make up Chapters 4 and 5, and run them all again from scratch. The lesson for the future: when coding connections between neurons, don’t write the sign of the connection into the arithmetic (Input = A - w.B); always write the sign into the weights: (e.g. Input = A + w.B, and w = -1). That way, a quick glance down the parameter values at the top of the code will show up any errors.

Authorship#

Until we reach that bright future where papers are no longer the de facto currency of academia, and/or where author names are solely linked to contributions as they are in film credits, authorship will remain a tricky subject. Who gets their name on the paper, and where it goes on the list of authors, can cause problems. Our criteria are simple, and similar in spirit to the criteria of the ICMJE guidelines:

  1. Substantial intellectual contribution to the content of the paper, AND

  2. Contribute to the written content of the manuscript in a meaningful way [by contributing text and/or substantial editing; and approving the final manuscript]

The usual model is that the postdoc or PhD student leading the research project will be first author on publications from that project; and Mark will be listed as the senior author.

Authorship is open to discussion. Especially when we co-author with collaborators outside the lab, there will be discussions based on the balance of contributions, from e.g. the collection of data versus the analysis of data.

The badge of “corresponding author” can carry weight: it is the author taking ultimate responsibility for all the content of the paper, on behalf of all other authors. Typically this will be Mark. But this will be discussed if it is clear the intellectual contribution of the lead author merits being the corresponding author.

Work/life balance#

Working hours#

Scientific research is target-based, not hour based. If you are producing a constant, evident stream of work - dead-ends and all - then you’re doing it right. Core hours for the lab are 10am - 4pm. All lab meetings will be within those hours.

Email policy#

Turn it off. Keep your focus. We use Slack precisely so that you do not have to check email to keep intra-lab communications going. Check your email when you need to use it yourself; or when you have time to execute on whatever is there. Strongly advise that you do not have work email on your phone.

Slack policy#

Set notifications to only arrive during core work hours (10am - 4pm). Mark will occasionally send out of hours, as I sometimes have to shuffle email/Slack around other commitments; and when I remember stuff that needs sorting. There is no expectation to respond until the core work hours.

Holidays#

Take some. Take a day off to recharge after intense bouts of work. Take a week or two off to get away from research. Mark takes a non-negotiable two weeks off at Christmas, and turns down all requests for work, reviews, grants etc to do so. It’s awesome.

Behavioural conduct#

All lab members are expected to abide by the University of Nottingham’s policies for behavioural conduct. Briefly, these are: