Writing for Research

Raewyn Connell

Summative writing

This is the writing that gets your findings into other people’s hands. It’s the subject of
most “Advice about Writing” texts, which teach you how to write journal articles, theses
or books. Other forms of summative writing exist, such as the old-fashioned
palaeontology monograph describing a particular species of fossil arthropod. (There’s a
very entertaining account of that genre in chapter 3 of Stephen Jay Gould’s book
Wonderful Life .) I suspect new forms of summative writing are emerging on the Internet.

For the time being, as I noted earlier, the journal article is queen. Therefore Part Two of
this booklet will detail how to write a journal article. Here, I want to raise a general issue
about summative writing.

Summative writing always involves selection; in fact, severe selection. Even with a long
book – I’ve written a few – the research produces far more material than the author can
include. Be kind to the reader! A journal usually forces selection by having a maximum
word length for articles. It is painful to leave out rich illustrations, amusing quotes, and
elegant arguments; but it has to be done. Get used to this pain, it will be with you
throughout a research career.

Selective interpretation 1966

(Image: Australian Liberal Party, from

The War Room, Nichols ed.)

A familiar point in the advice texts: good

writing is defined as much by what it

leaves out, as by what it puts in. The

problem in writing for research is to

select in a way that doesn’t distort the

truths established by the research.

Selective interpretation, as much as

stating falsehoods, is the rule in

propaganda.

Even without propaganda intent,

selection can mean distortion. Because

researchers and journals prefer

statistically significant results (there’s a

thrill in writing p<.01), many findings of

“no significant difference” go unreported

or unnoticed. I am convinced that this

pressure has distorted the whole field of

social-science and psychology “sex

difference” research. It really ought to be

called “sex similarity” research, because

p>.05 (i.e. no significant difference) is the

most common finding.