Friday, September 26, 2008

Advice to a Young Scientist by Ron Weisman

I recently benefitted from the best "how to be a scientist" talk I've
heard since I hit graduate school (and started trying to be a scientist),
from the colourful, distinguished animal behaviour researcher Ron Weisman,
who I'd met before but never had much contact with. The advice that
follows is very specific to science, and some of it only applicable within
psych. To start, here are my notes and key excerpts from the paper he
published on the subject,

Weisman, R. G. (2008). Advice to young behavioral and cognitive
scientists. Behavioural Processes, 77(2), 142-148.

(he's working on a book)

- The most desirable approach is to begin with extensive observation of a
behavior in nature, move ahead to laboratory research, and then return to
nature to ensure that you have it right.
- My advice is to use as many tools as necessary to discover and explore
nature's secrets. [but think collaborators]
-You should not design your observations and experiments to test only one
hypothesis at a time. You should be testing as many alternative hypotheses
as there are ways for your experiments to come out. Think ahead about what
an experiment's possible outcomes might mean. If an experiment has only
two possible outcomes, one that renders it interesting and publishable and
one that renders it problematic and unpublishable, you have not designed a
good experiment. Anticipate the possible outcomes of an experiment and
tinker with the design until each of at least three possible outcomes has
a distinct and orderly explanation.
- Reviewers like to ask direct questions, such as, "Why did you do your
study in the way you describe?" For example, if your groups differed in
both pretreatment and treatment phases, a reviewer is very likely to ask
why you attribute the effect to the treatment phase. Reviewers also like
to ask about alternative explanations of your results. At the design
stage, you need to consider and sketch out possible outcomes so that you
can anticipate even an astute reviewer's alternative hypotheses about your
results. To conduct first-class research, you must consider and handle all
or at least most of the reviewers' possible questions well before
conducting the research.
- If you have waited eight or more weeks, you need to write the editor
politely requesting an action letter. Some authors have waited a year or
even two - don't join them. After 12 weeks, inform the first editor that
you are withdrawing your manuscript and submit your article to a second
journal.
- Once you have the editor's action letter and the reviews, read them
quickly, then put them away for a couple of days to let the heat go out of
them. No one likes criticism, and reviewers are critics, so the heat you
feel is unavoidable. However, don't wait more than a couple of days to get
back to your article as you have a lot of time and sweat in the
submission. The revision should be the highest priority item on your list.
Get the article and the reviews back out on your desk and begin to
organize the reviewers' comments under the topics the editor asked about.
Print out copies of the action letter and reviews, and work from the
copies to make an outline of your revisions and your cover letter for the
editor.
- Respond to reviewers' comments incisively and succinctly....Try to
answer in just a few words or at most in a sentence tucked into the
paragraph from which the comments arise. Beware of protesting too much:
never insert a paragraph, or worse yet an entire page, into an article to
answer a reviewer's comment. Sometimes it is best to paraphrase a
reviewer's comment, then either agree that it is an issue for further
research or point out how your research has already handled it - all this
in a line or two. If you must directly disagree with a reviewer, make it
as tactful and convincing as possible. Editors are rightfully biased in favor
of their reviewers. Bury your disagreements with reviewers deep in the
cover letter well after many instances of willing change.
- Every resubmitted article needs a cover letter; it should be highly
organized, carefully written, and keyed directly to the text of the
article by page and line number.
- Never make uninvited changes to your article
- You must be prepared to handle objections and I counsel you again to
give in as often as possible.

And some non-overlapping notes from the talk he gave, unfortunately not
capturing any of his very entertaining delivery. Love his approach to
drafting papers!

- Success in science requires a combination of luck and fire in the belly.
(if you can do a whole experiment without once looking at your data, you
might not have the fire)
- Not only must you have luck, but you must know when you've been lucky.
- Causation questions:
Proximate - mechanism
Phylogeny
Ontogeny
Ultimate - functional adaptation
You should be constantly thinking about all of them.
- Don't compare everything with everything. Base tests on the designs,
otherwise lose power.
- Stats secrets: if unequal variance, you can do two separate analyses. If
you can get invariance of p values over tranformations (e.g. to rank
order) you don't need normality.
- In-school proposals are for chumps and robots (robotic committees). Do
the shittiest job possible. The person who knows what to do well enough to
write a well-detailed proposal has already done the research. Proposals
for money are a different matter.
- Positive controls are critical: they let you falsify.
- Get and keep your participants' attention: prompts, rewards, whatever.
- How to write a paper: in this order: 1) methods, 2) figures, 3) results
to explain the figures or table, also at least outline the discussion. 4)
Intro (never use proposal intro) 5) finish discussion. Then revise the
whole article with linking and concluding sentences.
- Stick questions and summaries about your results in another file on
another monitor as you write, which becomes the discussion.
- Intro prepares the reader for the methods, results and discussion.
Prepare the reader all the way along for the conclusion.
- The writing scientists do is like brick laying, it's not like carving
marble.
- Memorize about 20 pages of Strunk & White.
- Make strong claims (not "it seems") - you will still get cited even if
you're wrong.
- Suck up to editors and reviewers at meetings and elsewhere. Helps if
they can put a smiling face to a name
- Pick a journal that publishes work like what you've done. Also pay
attention to impact factor. If rejected, try sending it to a better
journal.
- You must respond quickly and forcefully to the invitation to revise and
resubmit. Don't wait more than a couple of days to get back to your
article. Top priority.
- Revise & resubmit means they'll porbably take it.
- Cover letter to editor: "Reviewer 1 asked us to relate our work to
Jones's 2008 article, we now do so on line xx, page xx." Nothing more.
- Write a thesis as 3 or 4 papers stapled together.
- For postdoc, pick the best universities and the best supervisors, in
that order (for name recognition). Begin early and suck up big time. Write
professors directly and ask about their work. But think through your
questions and keep them simple. Get your supervisor to help. Remember:
it's impossible to overflatter academics.
- You go where the good job is (even the gates of hell)
- When you're hired, always bargain for more money.

2 comments:

A said...

Thanks for this piece...I'm in my last year of PhD and a part of this will be useful.

ronweisman said...

Gosh, I am Ron Weisman. I am pleased with your summary and your interest in my talk. You and a few other young scientist have now encouraged me to write that damn book. But as I have pointed out, Writing is hard work, if you think it's fun you are not doing it right. By that I mean thinking hard enough.

Thanks again,
Ron Weisman
ronald.weisman@queensu.ca