experimental and theoretical work is done, it is important to write things up to make sure others
don't need to reinvent the wheel. Instead you will want them to acknowledge your genius. Hence
you have to write in a way to ensure that:
your readership will notice that your work is important to their work:
Provide a self-contained and clear
title and abstract.
readers who are familiar with the field but not with the particular problem treated by you can
still benefit from your work:
and useful theory. It should also
make clear why you consider the problem worthwhile working on. If there is a lot of theory, you
may consider a separate
readers who want to check your results or further develop your technique can do so:
section, describe all relevant facts that may have influenced your results
and all your operations during the experiment. Mention only experimental facts that relate to the
results you show. If you have changed or improved the setup before taking data, you do not
need to report your considerations that caused the change - somebody who plans to follow up on your
work needs to start where you have finished.
readers grasp your results at a glance:
Show plots of your
Results are the
immediate data you have recorded
as a function of the experimental parameters. In this section,
no derived quantities
should be displayed.
However, you can (and should) plot the data in a way that helps the analysis. For example, if
you want to determine the exponent of a power law, plot the data logarithmically. Generally, plots
are better than tables, and the same information should not be presented twice. It is important to
guide the reader through your results,
so write some text pointing to the important features
in the plot: the position of a maximum, the linewidth, signal-to-noise, change in slope - whatever
is important. But do not draw any conclusions at this stage. Quantify all
errors in primary measured results and parameters.
If you have to process your raw data (filtering, background
subtraction, smoothing etc.), describe it here, quantify the
errors introduced by the processing,
and show an example of how it affects your raw data.
State all your data to a precision commensurate with the error margin - only the final significant
digit should be affected by the stated error.
reproducibility (i.e. most of the time). Past tense is more appropriate when referring
to one-off events that cannot be repeated, such as an astronomical observation.
you convince your readers of the point you are making:
In a separate
derive qualitatively and quantitatively any properties or quantities from your data
using the formulae given in the introduction
or theory section. Refer to the
equations rather than repeating them. A Mathcad script is not acceptable - you must describe in words
what you do and why you do it. Analyse the
propagation of errors
quantitatively. The purpose of
error analysis is not to blame the kit/yourself/the technician but to provide a level of
confidence with the data. There are papers with horrendous error bars in the journals, but if
nobody else can measure it more accurately then this is perfectly acceptable! If you can/want to
derive a generalised model
from your results, do it in this section; and make it accessible by drawing a sketch
of its main features.
readers will remember the key points:
Provide a short
section. This is a summary without reference to literature,
figures and tables, and will not normally contain formulae.
people who have helped you will find their contribution acknowledged
Finally, give a short list of people (if any) who have helped you, and how. Only include those
who really did - there is no need to be over-polite.
Some further points you should keep in mind:
on an appropriate scale. Ensure that different data sets are clearly
distinguishable. If you plot lines to connect points, state in the caption whether these represent
theory functions or merely guides to the eye. In the latter case - do you really need them?
Fig.1 gives an example of well-presented data: Both data sets are shown on a scale which allows to
discern the detail of the pattern, and plotting them together in one graph allows to spot correlations.
Error bars help distinguish real fluctuations from random noise. Axes are labelled, and the data fill
the frame with little empty space.
Make sure that all figures and tables are enumerated and have a caption describing what is on
display and what the different markers stand for. Increase signal-to-noise by avoiding sentences
like "a graph to show" - people know that it's a graph; and graphs are usually meant to show something.
Use the figure numbers to refer to figures in the text. No figure should go without mentioning in
When referring to other people's work, you can use either of two formats, but be consistent with
it. Either enumerate
or refer to the papers by using the name
of the first author (Author 2002a). Then list the references at the end of the papers like
 A.N. Author, A.N. Other, J. Irrepr. Res. 23 (2002) 456
Author 2002a; A.N. Author, A.N. Other, J. Irrepr. Res. 23 (2002) 456
In the example, 23 is the volume, 456 the number of the first page, and 2002 the year of publication.
In the second format, distinguish different papers by the same author from the same year by a,b,...
after the year of publication. You can also cite books, other people's lab reports or scientific
 HC van de Hulst, Light Scattering by Small Particles, Dover Publ., New York 1981, p.123
 ER van Hoek, Gravel, Mud, and Avalanches, UWA group project report, 2001
 MY Colleague, unpublished
The latter makes sense e.g. if you want to make clear that a group who did the experiment earlier
has given you a vital clue. When citing from a book, always refer to the particular page or section
that is relevant in the context.
Use appendices wisely and sparely. Anything that is essential to your project should really be
in the report itself, where it can be referred to and explained properly. However, appendices can be
useful to add background information such as source code of programs you've written or data images of
similar samples where you have already shown and discussed a typical image in your report. There is
no need to reproduce data tables of every measurement you've taken during your project. In the majority
of cases, material considered for an appendix should either be in the report itself or could be left out
completely without detriment.
Before you submit your report or paper,
read it, sleep a night, and read it once again.
Incomplete sentences and misguided formatting can easily be avoided this way. These
really are important issues - if you make yourself easily understood people will be more receptive of
the point you're making. It's your work - be proud of it!
Here is a research paper from Phys Rev Lett. Because it is a short
paper, it doesn't have section headings like Introduction or Results etc, but it follows
the same structure. See if you can identify the different sections.