Use more tables

While almost every technical paper includes some figures, surprisingly few contain tables. It is true that, in some typesetting systems (e.g. LaTeX) tables are not so straightforward to create as they are in HTML, which I'm using to write this page. Nevertheless, tables present a golden opportunity for the author who is not a native speaker of English to present information concisely without the hassle of constructing full sentences.

The 'normal' use of tables, to present numerical data, doesn't need explaining, but even in this basic context tables could be used more often. And I will now suggest a couple of two ways to use tables more creatively.

Competitive analysis
Virtually every article in a consumer magazine such as Which? contains a table that rates different aspects of the products being surveyed. It is a hugely efficient way to present a comparison. And the author who is writing in a foreign language benefits because it reduces the need for grammatical constructions to a minimum. Some excellent technical authors have used 'comparative analysis' tables, but nevertheless they are still quite rare. Why? I don't really know.
I will show you what a table of this sort might look like, taken from the Introduction of a fictional paper:
MethodAuthor(s) PublishedSpeed RobustnessComments
Naive heuristic
(discussed in Ref. 3)
Infeasible: no implementation known
PredictiveHamm & Egg (2) 1995Slow on chi(ps) dataset (4) Fails on aligned data and singularitiesBasis of first usable implementations
CorrectiveKim & Chee (5) 1998Slow (authors' tests) Fails on aligned data 
Predictive-​corrective (P-C)Kartoffel-​Salat (4) 2002Under 20s on chi(ps) dataset Fails on aligned data
Modified predictiveHamm et al. (3) 2003Over 1000s on chi(p) dataset Remarkably reliable Uses interval arithmetic
Corrective-​predictive (C-P)
(current authors)
30s on chi(ps), but faster than P-C on larger datasets Fails on some singularitiesRequires time to create lookup

Table 1. Relative performance of recent approaches to the 'breakfast' problem.
Logically, a table of this sort could have another column to explain the various different methods. But in practice the amount of text required makes this infeasible. However, the explanations can still be presented in a structured way by itemizing them in a definition list. A large slice of your literature review can then be compressed into a short paragraph something like the following:
   Several approaches to the 'breakfast' problem have been published in the literature. The significant approaches are described below.
The itemized discussion of each technique follows here, and then the table is introduced:
   It has been difficult to achieve an acceptable combination of speed and robustness. The significant attributes of recently developed methods, and of the method reported in this paper, are compared in Table 1. (The chi(ps) dataset is due to Cook et al. [1]).
Since formulae and equations are (more or less) universal, they are friends to authors writing in a foreign language. But unfortunately equations still have to be explained. This can be quite awkward, leading to ugly sentences that start with a lower-case variable, many terms in brackets ("for all X" etc), and excessive use of 'arranging' words, such as "respectively".

These problems can be avoided by using a table to itemize and explain variables. Here it is applied to a very simple example, the equation of a circle.

Using text:
(x - a)2 + (y - b)2 - r2 = 0   (1)

The constants a and b are respectively the x and y coordinates of the centre of the circle, and r (r > 0) is its radius.

Using a table:
(x - a)2 + (y - b)2 - r2 = 0   (1)

The coefficients in this equation are explained in Table 2.

ax-coordinateof the centre
rradius(r > 0)

Table 2. Coefficients of the equation (1) of a circle.
This is a trivial example (HTML offers little support for typesetting equations) and the second version takes up substantially more room; nevertheless it is conceptually simpler. The advantages of using a table grow as the complexity of the equation increases. Furthermore, the 'cost' of a table can be spread over several equations which use similar sets of variables.

Back to the index page