The Science Of Scientific Writing    Set F      The Discussion     Coherence: Speculative and Non-speculative Content      Exercise 1      Maps for Discussions    Exercise 2      Final Page .

Course Home

OVERVIEW: The way to well-written science

How to do the Course


PART I: Paragraphs and Sentences

SET A: Paragraphs: The Maps Behind Them

SET B: Paragraphs: Using Maps to Meet Readers' Expectations

SET C: Paragraphs with Something Extra: Points and Tails

SET D: The Generic Section: Expectations and Maps as Blueprints

SET E: Scientific Sections: The Methods and Results

SET F: Scientific Sections: The Discussion

SET G : Scientific Sections: The Introduction

SET H : Sentences

SET I : The Paper as a Whole



PART II: The Paper and its Sections


SET 1: Argument Parts

SET 2: Indicator Words

SET 3: Refining Claims

SET 4: Locating Arguments in Prose

SET 5: Rationale's Essay Planner

SET 6: Evidence in Arguments: Basis Boxes

SET 7: Assessing

SET 8: More on Assessing

SET 9: Analysis Maps

SET 10: Assessing Again

Synthesis 1: Position-Early Paragraphs

Synthesis 2: Position-Final Paragraphs

Synthesis 3: Writing a Discussion I

Synthesis 4: Writing a Discussion II

Set F: The Discussion

The Discussion is usually the third section to be written, following the Methods and Results. Overall it typically represents the greatest challenge of the paper because:

(1) it requires a deep understanding of the background literature if the author is to properly interpret the results.

(2) it often requires the author to use a variety of discourse modes, in particular argumentation and explanation. Argumentation is the least "natural" of the writing modes used in a paper. Scientists are generally much more comfortable with reporting and explaining, much less so with "debating".

(3) the variety of modes used in the Discussion means that extra care needs to be taken to maintain coherence. It is harder to ensure that one mode will predominate, and not be overwhelmed by the content of other modes.

Before considering issues of structure (Frame of Reference + Elaboration) and coherence, I want to first consider the various types of questions addressed in scientific papers. First, on this page, we will categorise them with respect to their tractability and specificity. On the next page, we will categorise them with respect to the primary mode discourse of the paper: descriptive or argumentative. I will introduce some new terms to help categorise these questions.

Tractability and Specificity of research problems

Research problems can be characterised in many ways. On this page we will consider two of the most important, and correlated, criteria: specificity and tractability (i.e. manageability, solvability). The correlation between the two is usually true when we consider probems of a related type: e.g. if a disease affects twenty species of mammals, it is easier to find a cure for one of those species than for all of them.

At some point, every research paper must address a tractable problem (e.g. "Which atmospheric gas, carbon dioxide or methane, absorbs more infrared radiation?"). But the authors will also typically connect this problem to a more general, less tractable one (e.g. "What causes global warming?). Heading in the "other" direction, the means by which the tractable problem is addressed may involve asking very specific, soemetimes quite obscure, questions (e.g. "At what frequency do the C-O and C-H bonds of carbon dioxide and methane vibrate?"). I will be introducing three simple terms to talk about questions at different levels of tractability and specificity. Question of the most general, and not necessarily tractable, type, I will call General Research Questions. Questions at the next level down, which must be tractable, and which are also typically more specific, I will call Specific Research Question/s. Most specific of all, and again, necessarily tractable are Research Objective/s. The variations in tractability also have mean that the two relationships (Research Objective-Specific Research Question and Specific Research Question-General Research Question) are of a different nature: only in the first pair will an answer to the more specific question have a major impact in resolving the more general question.

Undesrstanding the differenes between the three types of research question is important in structuring both the Introduction and the Discussion. To provide a more practical illustration of the three types of question, below we will see how they interact in a famous example of scientific research.


Pasteur's "disproof" of spontaneous generation

"Never will the doctrine of spontaneous generation recover from the mortal blow of this simple experiment". Louis Pasteur, 1864.

In 1859 Louis Pasteur performed a famous experiment using a swan-neck flask similar to the one below. Meat broth was placed in the main part of the flask and boiled. The long extension of the neck allowed normal room air to be in contact with the broth, but dust in the air never reached the broth because it was trapped by gravity at the lowest part of the downward curve. The meat broth remained clear for years, but if the settled dust was allowed to come in contact with the broth, the broth would turn cloudy within hours. Although this was never written up as a paper in itself, it was generally accepted by the scientific community of the time that this experiment had disproved the long-standing theory of "spontaneous generation", that is, the idea that life-forms could easily (and regularly did) arise from non-living matter.

In fact of course, his experiment did no such thing, because it only showed that in this very particular example, spontaneous generation does not occur. It is actually impossible to disprove spontaneous generation by any such experiment. It did however answer an interesting question that had arisen out of a previous meat-broth experiment by the Italian Lazzaro Spallanzani. Spallanzani had used a similar set-up to Pasteur, but his his flask was sealed, leading proponents of the the spontaneous generation theory to claim that the boiling of the air in the closed system had killed off some "vital principle" in the air, and that this was normally not the case in Nature. Thus their stance could be converted into the question: Could spontaneous generation occur with just two components: any organic matter, and air?

Pasteurs's experiment directly indicated that the answer to the question above was no, since some third component was required for the broth to spoil, something that was present on the dust in the untreated room air. The extension of the results of Pasteur's exeriment to the general believability of the theory of spontaneous generation can only be made indirectly, and requires an understanding of the scientific thinking of the time, and in particular, the growing belief in the germ theory of disease. Thus, in Pasteurs work, and in the way in which it was interpreted, one could argue that there is a four-tier series of questions, and the questions vary in their degree of specificity, in the way in which they "talk" to each other, and their scientific importance. This is summarised in the Table below..

Four levels of questions (or issues) in the Pasteur experiment

How the question relates to the question or issue above Scientific Significance
Research Area
What is the origin of life?
A general field of work in which there are many questions, some tractable, some intractable
Very high

General Research Question

Does life arise by spontaneous generation? The question is general and often intractable, and any answers to it indirectly address the Research Area.
Specific Research Question Can spontaneous generation occur with just two components: any organic matter, and pure air? The question is specific and tractable, and its answer indirectly addresses the General Research Question (above).
Research Objective Does boiled meat broth in a flask whose contents are in contact with room air, but not with the dust in the air, become cloudy? The question is very specific and tractable, and its answer directly addresses the Specific Research Question (above).


The further category of "significance" is based on ideas we will explore more when we cover the Introduction in Set G. More important for our purposes now is the way in which the questions (and their answers) relate to each other. Any Research Objective in a scientific study should yield a precise answer (in this case no), based on data (in this case, observations of cloudiness of the broth). The answer to the Research Objective should directly help us answer the Specific Research Question (in this case the answer is no). Notice that up to this point, the relationships have been ones of tractability and directness. But from this point on, things get murkier. The answer to the Specific Research Question addresses the General Research Question but only indirectly, and thus the answer to that question must remain imprecise.

In some research papers, the Discussion will only concern itself with relationships of high certainty (non-speculative), that is those between the Research Objective/s and the Specific Research Question/s. And indeed the only obligation of the authors is to do this. If however, the authors also choose to also address speculative ideas, then they need to give strong signals to the reader that they are doing so, either by:

(1) Using language that explicitly indicates that implications are being drawn, or explanations attempted. For example, the sentences below all indicate that the author is about to utilise the answers to the Research Objective or the Specific Research Question of the Pasteur experiment in some speculative way:

(a) Previously I have argued that the lack of cloudiness in broth exposed to dust-free air allows us to say that spontaneous generation (if it ever occurs) will require more than the presence of just organic matter and pure air. What implications does this finding have for our understanding of the long-standing theory of spontaneous generation?

(b) Next, how can we explain the rapid appearance of cloudiness in the broth when it was allowed to come into contact with the dust that had settled in the curve of the swan-neck?

(c) Given that, as argued above, dust in the air is a likely source of microbial contamination, what implications does this have for the large-scale preparation of food?

(d) Now that we suspect that microbes are present on dust in the air, what future experiments can we undertake to invesigate this further?

(2) Placing speculative content in structurally separate locations, for example :

(a) as one or more concluding paragraphs at the end of a Discussion sub-section. This might be appropriate when the explanations or implications are related only to the content of that sub-section.

(b) as one or more concluding paragraphs at the end of a Discussion as a whole. This might be appropriate when the explanations or implications are related to the content of several sub-sections of the Discussion. The speculative content might all be grouped together in a single "Conclusions" sub-section (with or without a sub-heading) or they may be grouped as a series of sub-sections (in which case sub-headings for each would be necessary).


The diagram below is a schematic of a Discussion fof a hypothesis- or model-driven paper. It has the following instructive features:

(1) A Frame of Reference paragraph that would provide a guide to the entire Elaboration part of the Discussion

(2) An Elaboration with multiple (six) sub-sections, each with its own sub-heading

(3) A mix of non-speculative and speculative paragraphs in the Elaboration (the latter marked "S"). Note that:

* the majority of the Elaboration is non-speculative.

* speculative content occurs both as a paragraph within a largely non-speculative sub-section, or as an separate, entirely speculative sub-section.

* speculative content is always placed after non-speculative content, either within a sub-section, or within the Discussion as a whole.