How to write paragraphs

in research texts (articles, books and PhDs)


In English the core building blocks of any intellectual or research argument are paragraphs. Each paragraphs should be a single unit of thought, a discrete package of ideas composed of closely linked sentences. The most generally applicable sequence to follow is — Topic, Body, Tokens, Wrap.

  • The opening ‘topic’ sentence alerts readers to a change of subject and focus, and cues readers (in ‘signpost’ mode) about what the paragraph covers. It should never link backwards to material that came before (linkages are instead always made forward in ‘wrap’ sentences). So be wary of starting paragraphs with linking words (such as ‘However’, ‘Never the less’, ‘Furthermore’), lest they lead you into looking back. Instead topic sentences should clearly signal a new focus of attention. Yet they also need to be carefully written, to give readers the impression of a fluent, ‘natural’ progression of thought. Remember too that a signpost is just that — it is a very short cuing or naming prompt, not a mini-tour guide or a preview of the whole paragraph argument to come.
  • The main ‘body’ sentences give the core argument of the paragraph. In research work they need to clearly and carefully set out reasoning, describe results, develop implications, elucidate formulae, or elaborate and explain theoretical and thematic points. Body sentences constitute the mainstream of the paragraph, the core of the unit of thought.
  • Researchers normally must offer tokens to back up and support their core arguments. ‘Token’ sentences can be sprinkled across a paragraph amongst the body sentences, at apt points where they are most needed or useful. Typically token sentences are examples, references, quotations from other authors, supporting facts, or analysis of accompanying ‘attention points’, exhibits, tables, charts or diagrams. In some degree ‘token’ sentences are inherently digressive: they potentially lead away from the mainstream of the paragraph. Hence they need careful management, especially when two or more token sentences follow each other, without intervening ‘body’ sentences.
  • Finally the ‘wrap’ sentence serves to pull the paragraph argument together, to make clear to readers that a building block has been put in place. It should be constructive and substantive, adding value to the argument, not simply repeating early materials. It should also handle any link forward to the next paragraph that is needed.

Rational, skimming readers do not treat all parts of paragraphs in the same way. In search of the quickest possible appreciation of what is being said, they pay special attention to the beginning and ends of paragraphs, to the topic and wrap sentences — a technique commonly taught on ‘speed reading’ courses. When and if they look more closely inside the body of the paragraph, readers may also initially skip across token sentences. And they will normally put off digging into ‘hard’ formulae or tough exposition materials in search of a more intuitive (if approximate) understanding gleaned from the sentences that precede or follow them.

It follows that the beginning and endings of paragraphs should always be the most carefully written materials. Try to separate out these two sentences and look at them together. Check how they read, how substantive and informative they are, and how they might be improved.

Six common paragraph problems

Six things most commonly go wrong in writing paragraphs:

1 The author starts with a backward link to the previous paragraph, instead of a fresh topic sentence. Readers may conclude that this is simply ‘more of the same’ and so skip onwards to the next paragraph. Even those who persist may become confused — what is the paragraph really about? Is it the start sentence? Or the different point given in the now ‘submerged’ topic sentence that comes second?

2 The paragraph begins with a ‘throat-clearing’ sentence, or some formalism or other form of insubstantial sentence (or perhaps several such sentences). For instance authors might begin by discussing a caveat, a definition, a difficulty or a methods issue that form part of the provenance of the argument to be made. The effect is again to bury the real topic sentence one or two sentences deep in the paragraph. Readers may conclude on a quick look that the whole paragraph is just an insubstantial caveat, or navel-gazing of the familiar academic kind, and so skip forward, missing the change of focus completely. If they do persevere reading they may not correctly identify the now submerged topic sentence, and then find that the wrap sentence seems unjustified or tendentious, because it does not fit with the apparent topic.

3 The author starts the whole paragraph with another author’s name and reference, for instance: ‘Harding (2007: 593) argues …’ This is a beginning especially beloved of some PhDers and other unconfident authors, creeping forward with their argument propped up on the supports of other peoples’ work. Some postgrad students will construct whole sets of paragraphs in this manner, running over several pages, every one of which starts with another author’s name, especially in ‘literature review’ sections. They mistakenly believe that this way of proceeding will convince readers that they have closely read the literature. But when the first words of a paragraph are someone else’s name, the author is inadvertently signalling: ‘Here follows a completely derivative paragraph’ — or section if this pattern is repeated. So critical readers’ common reaction is to downgrade or skip the paragraph (or sequence of such paragraphs) and move on.

The easy solution to this problem begins by not thinking in terms of individual authors, but focusing instead on the schools of thought, or ‘sides’ in an empirical controversy, that the authors to be cited represent. Write a clear and free-standing topic sentence. Then explain the core ideas or propositions of one or more schools of thought involved in the body sentences. Relegate author names to the supporting references that come at the ends of sentences, where they belong.

4 A paragraph stops abruptly, usually because the author has become aware that it has got too too long. Commonly this occurs because token sentences have multiplied — perhaps because the planned brief exposition of an example or analysis of an exhibit have become unwieldy. Usually authors here make an enforced ‘emergency stop’, and then commonly write up what should have been the wrap sentence as the beginning of the next paragraph. The first paragraph then has a sequence of Topic, Body, Tokens but no wrap sentence. And the next paragraph 2 starts with the displaced wrap1 sentence, and has a buried topic2 sentence. Readers will get a bit lost at the end of paragraph 1 here, as a token or body sentence ends the paragraph with no form of recap. And they will read the displaced wrap sentence as signalling the topic of paragraph 2 (which it doesn’t). They may puzzle through paragraph 2, feeling that it was not what was promised at the start, or that it does too many things. Or again they may skip forward here, feeling that paragraph 2 only repeats.

5 Paragraphs get too long, extending beyond the acceptable research text range of 100-200 words to take up 300 words or more. Often this happens because tokens have multiplied or swollen outside the limits that can be handled easily. But because of their partly digressive character the author is reluctant to recognize the need to create separate paragraphs to handle them. Especially when they discuss attention points or exhibits that are complex and not designed to be self-contained and easily understood, body and token sentences may blur together, creating text where the mainstream argument becomes hard to distinguish.

The solution to very long paragraphs has to be brutal. Once a paragraph passes 250 words, it must be partitioned, usually as equally as feasible, and separate topic and wrap sentences provided for each part. If the problem arises from an overlong exposition of a token or an exhibit, then the author needs to find a solution that allows a partial digression to be smoothly handled. If a paragraph falls between 200 and 250 words this might be retainable, so long as the wrap sentence can still reconnect readers back to the (now rather distant) topic sentence.

6. A paragraph is too short. For a research text this occurs if it falls below 100 words, and especially if it consists of just one sentence or is less than 50 words. Normally, short, bitty paragraphs like this look terrible on the printed page of a journal or a research book, and they undermine the usefulness of paragraphs as argument building blocks. Short paragraphs happen because an author is unsure what to say, or has not properly thought through how a point or a set of points fit together or can be sequenced into the overall argument. Some reflect miscellanies of points that the author has not acknowledged as such. Other single sentence paragraphs are ‘orphan’ sentences that should be incorporated into longer nearby paragraphs but have not been — for example, in starting lists or sequences of connected paragraphs. Orphan sentences (and short paragraphs generally) should always be merged into their neighbours, so that they disappear.

Two groups of people need to take particular care adapting to this convention of English paragraphing at research level. First, Spanish speakers (and related languages) often write using multiple, very short or single-sentence paragraphs, organized in subtle thematic ways that English-speaking readers find very hard to follow. Such an audience will often see only a baffling multiplicity of paragraphs that is interpreted as disorganized thinking. Second, journalists, and now some academic bloggers also, use short paragraphs that look OK in newsprint or on narrow or spaced-out blog columns. All these types of author should aggregate their short paragraphs into longer ones of at least 100 to 200 words if they want to publish journal articles or research books in English. If you have ever seen a journalist’s writing transposed into book form without this change, you will appreciate that there are also strong aesthetic reasons to make this shift.

To follow up these ideas in more detail see my book: Patrick Dunleavy, ‘Authoring a PhD’ (Palgrave, 2003) or the Kindle edition, where Chapter 5 covers ‘Writing clearly’ and Chapter 6 ‘Developing as a Writer’.

There is also very useful advice on Rachael Cayley’s blog Explorations of Style.

And for new update materials see the LSE’s Impact blog and on Twitter @Write4Research



Writing For Research
Advice for authoring a PhD or academic book

Writing creative non-fiction at a research level is hard, skilled work, across all disciplines. Here Prof Patrick Dunleavy (LSE) collates some helpful resources