Structuring and writing academic papers

Here is a deliberately brief summary of guidance that some folk have found useful:

There two major models for structuring an academic paper: 1. the conventional model; and 2. the design model. Whatever method you use, it is important to set it up right, and plan your paper beforehand. You may find it useful to also check tips in my post on storyboarding research.

1. Conventional papers

The advice notes for this section all draw on Thomas Bassboll’s blog. He suggests

A paper should have 40 paragraphs, arranged in eight even-length sections:

· five paragraphs for Introduction and Background

establishes motivation and reader interest

· five paragraphs of Theory — wedded to, and focused for the empirics that you’ll present

· five paragraphs of Literature Review. Do not make this section too long!

· five paragraphs of Methods. This section tends to be ultra-conventional in many papers. Think about what details readers need to know: do not put more, nor less. People tend to really think about methods only when they see the data — it is important to mention here sufficient details about how data was obtained.

· Around 40% of your paper’s text should cover RESULTS. That would mean 15 paragraphs,maybe divided into two or three parts, e.g data empirics, analysis 1, analysis 2

Do not assume that information on the data or how to interpret them are obvious — discuss results. You need at least 5 paragraphs discussing the results.

· five paragraphs of Implications. These can be theoretical and practical — one per paragraph?

· Overall conclusion — at the start of this section mirror the last paragraph of your introduction but now substantively answering the questions posed, not asking them. Later on in the conclusion, open out to next stage research questions.

The beginning and the end of papers are the hardest to write. Ideas at the beginning and the end frame what has to be said. The beginning has to establish a commonality between you and the reader. Do not underestimate readers’ resistance to new work: establish readers’ interest (if you can) in the first paragraph.

This advice is based on an 8,000 words article. If your article has to be shorter, reduce the number of paragraphs on a pro-rata basis.

2. Designed papers

The focus here is on positively attracting readers to read, remember and cite your paper. There is a myth that “90% of papers published in academic journals are never cited… .” In fact, uncited papers published in academic journals are 12% in Medicine, 27% in the Natural sciences, 32% in the social sciences Social sciences, and 82% in the Humanities.

Meanwhile, non-elite journals are attracting more Google scholar citations — as this chart shows, people increasingly find what they want by searching online. They don’t just trawl the top 10 journals in their field anymore.

Source: arXiv:1410.2217v1 [cs.DL] 8 Oct 2014

Both these factors mean that it is well worth writing the best designed and most accessible paper that you can.

Choosing a designed sequence

Designed sequences are in stark contrast to “given structures” such as the meachanistic or conventional approaches discussed above. It is especially helpful to try to avoid structures that are just a ‘record of work done’ or the pure conventional structure still used in technical disciplines. Three basic designs are feasible:

1. The focus-down model. Perhaps 90% of work done still uses the focus down model.

2. The opening out model (useful in physical sciences)

3. The compromise model

These models differ a lot in how long readers have to wait before reaching ‘core’ material — the most original or value-added evidence or components. You have to decide what the core is from a reader’s point of view. The core is where you talk about new findings or show independent critical power.

The focus-down model

This model has a long literature review and a quick (usually too quick) a finish. The focus here often seems to be on the literature.

The opening-out model

This model uses a very quick set-up to move straight into the core. It can be hard to do this well. Some traditionalists in humanities and social sciences disciplines do not like it — because there is insufficient literature/ ancestor worship.

The compromise model

This approach starts by engaging attention, motivating readers and giving a brief, recent literature coverage only. It then continues by maintaining readers’ attention through a quick transition to the core (original or value-added) findings. Good signposting is a must. And there is enough space for a thorough analysis of findings, and a discussion of how they relate to the literature/previous findings at the end.

Integrating a paper

All the bits of your paper need to fit together — in an ‘industrial strength’ mode where every bit supports the whole. To check this, isolate out the parts below and see how they complement and reinforce (but don’t repeat or contradict) each other.

Make the title fit really closely with what you find out or announce as your conclusions . Put the story in the title. Check out the following example:

Starting a paper

• A high impact start engages readers’ attention (see above): followed by any framing or set-up text.
• Motivation material follows, designed to get readers to engage with the analysis to come.
• Signposts briefly point forward to the sequence of topics in the main sections to come

Finishing a paper

• Lead-out materials look back and draw out the conclusions of the analysis and their implications, revisiting each signpost in the inital sequence.
• Links outward and forward connect the findings or conclusions here to the rest of the relevant literature

Examples

Structuring chapters

Aim to divide the text relatively evenly into same-size main sections.

A common way of completely de-structuring a chapter or paper

Why doesn’t this work? First, almost all the text is in one section, 4.3. Second, the first-order headings confusingly signal sections of radically disparate lengths and significance. Third, two of the headings (Introduction and Conclusion) are purely formal ones, with no substantive content.

Patterns of explanation

DESCRIPTIVE PATTERN

The structure is set outside you. Examples include:

• guidebook/institutional patterns
• all chronologies or other narratives (unless periodized)
• a conventional listing or hierarchy of forces in the outside world
• ‘This is what I did in my thesis, at the start.., in the middle…., at the end’.

Actually this approach often works OK at whole-thesis or book length, despite the heavy fact load it demands.

· But it doesn’t work for articles, and

· It doesn’t personalize or distinctively organize the information being covered.

ANALYTIC PATTERN

ANALYTIC

Here the author compartmentalizes

· necessary or sufficient causes

· periodization/long-run or short-run influences

· conventional categories (social/political/econ)

· tracing out an algorithm..

This approach can work well with robust categories, because it

· selects more, structures more

· key danger is formalism/fruit cocktailing

· narratives become more complex this way

ARGUMENTATIVE

Focuses on controversy and conflicting schools of thought

· explicitly multi-theoretical

· but not just literature review sequences

This is difficult to do when

· your supervisor/department takes a strong, closed view (‘there is no alternative’)

· you are tackling new material — capturing overall arguments entails having a sophisticated, holistic view.

MATRIX PATTERNS

Generate more complex structures by combining two of the previous approaches. There are three main variants

- analytic (firstorder headings) plus descriptive (second order headings)

- descriptive (e.g. narrative) plus analytic

- argumentative plus analytic

Logistical issues — things to check when you have a draft

Length

- Look hard at the length of papers in your discipline — e.g. 3000 words medicine, but 8000 in social science, 10,000 words for special cases only

- Do not ever pass the limit: the rejection/ rewrite penalties for exceeding the opitmal length rise exponentially. Over-writing can make work un-publishable

- In writing a first draft text, aim for at most 80% of target. So 0.8*8K = 6,500 words

- Don’t write much less than the required length either , e.g. if norm is 6k, then anything under 6K looks too short

- Don’t despise short article forms, notes etc.

Assuring the core

You need to realistically identify what the core is, in other words what the value added is

· In a conventional paper, core materials = methods + analysis+ results

· In a designed paper, core materials = clear value-added materials — discovering new facts or showing independent critical power

Pre-core materials are: Lead-in, throat clearing — minimum 5 paragraphs

Post-core materials: also 5 paragraphs to set the core in context, integrate analysis with other papers, and lead out.

Coping with criticisms and ‘working up’ a paper that is not quite ‘there’ yet

Changing your text in response to reviews and comments

· Don’t ignore what reviewers say, however much you disagree

· Even wrong-headed comments have value — they show you where misunderstandings arise

· Generate a grid box of comments, your reactions, and the changes you’ve implemented.

· Always make some change for every comment

Five structural things to try — see this post for more guidance

- Do one thing well — cut out digressions, and secondary materials

- Apply the BBC test — does every element “build, blur, or corrode” the paper? Every piece in your writing does one of these three. Writing that builds moves things forward. Waffle or unclear or un-necessary writing needs to be corrected or deleted. Corrosive writing needs to be deleted.

- Flatten the structure. If you have too many headings without text in between, your structure is probably to complex. You need to flatten it, have fewer sections/building blocs (8 in conventional structure, maybe 4 in designed sequences)

- Say it once, say it right. Do not say argument in drips and drabs. Group similar points in one place

- Try paragraph re-planning

Four explanatory things to try

- Make the motivation clearer — never under-estimate the ‘strangeness’ of new work and resistance to (recognizing) originality

- Strengthen the argument tokens

- Improve the data and exhibits. Make it easy to understand. When possible, visualise

- Refresh the literature review . How ‘old’ is your current one? How new are the latest references?

Further resources

Producing and structuring papers is discussed in more depth in my book with Jane Tinkler, Maximizing the Impacts of Academic Research: How to grow the influence, practical application and public knowledge of science and scholarship (originally Macmillan, 2021, now Bloomsbury Press), Chapters 3 and 4.

You may also find useful some of Patrick Dunleavy, ‘Authoring a PhD’ (originally Palgrave, 2003, now Bloomsbury Press) which covers becoming a better author, stylist and self-editor.

And my @Write4Research Twitter feed pulls together a wide range of the latest writing advice.