How To Write A Good Scientific Paper
How To Write A Good Scientific Paper
Chris Mack
Editorial
1 Writing an Abstract
The most common mistake in writing an abstract is to not pay much attention to it. Authors sometimes consider the abstract as an afterthought, something that can be thrown together after the real manuscript is written. Ive even seen abstracts that are nothing more than the first paragraph of the introduction. Needless to say, such a poor abstract is unlikely to encourage a potential reader (or a journal editor) to venture further. The abstract should be a concise, stand-alone summary of the paper, covering the following topics:3
background/motivation/context aim/objective(s)/problem statement approach/method(s)/procedure(s)/materials results conclusion(s)/implications *This is the first in a planned series of editorials covering all aspects of good science writing.
2 Structured Abstracts
For the past 25 years, structured abstracts have become required in most medical journals, though they are not very common in engineering and the physical sciences.4 I hope this will change, since I am a big fan of the structured abstract. Simply put, the structured abstract formalizes the five topical areas mentioned above by adding subheadings and subsections (the structure) into the abstract. While the exact structure can be modified to suit the topics of the journal (or even the specific paper), in engineering and physical sciences a five-structure format is probably best: background, aim, approach, results, conclusion. Each subsection should contain one to two sentences, answering the following questions: Background: What issues led to this work? What is the environment that makes this work interesting or important? Aim: What did you plan to achieve in this work? What gap is being filled?
AprJun 2012/Vol. 11(2)
020101-1
Editorial
Approach: How did you set about achieving your aims (e.g., experimental method, simulation approach, theoretical approach, combinations of these, etc.)? What did you actually do? Results: What were the main results of the study (including numbers, if appropriate)? Conclusions: What were your main conclusions? Why are the results important? Where will they lead? The benefit of the structured abstract is twofold: it forces the author to include information from all five categories, and it makes these five sections easy to find and access. But while it is logical that structured abstracts will be better than unstructured abstracts, there is in fact proof that this is so. The preeminent researcher into the efficacy of structured abstracts, James Hartley, reviewed some 31 studies that had been performed by 2004 and found that these studies demonstrated the superiority of structured abstracts.5 His review, as well as others,6 showed that structured abstracts
contain more information; are easier to read; are easier to search; facilitate peer review; and are preferred by readers and authors.
creates an effective surface inhibition over a depth into the resist of several correlation lengths. Conclusions: The differences between microscopic and macroscopic dissolution rate can have an important effect on how development rate models should be calibrated, depending on their use in continuum or stochastic lithography simulators. Then, deleting the subheadings and line breaks, a traditional abstract format is obtained. I added a transition clause at the front of the last sentence to make the abstract flow better, though this small change could easily have been left out. Photoresist development rate can be defined microscopically (the development rate at a point) or macroscopically (the propagation rate of an average resist height). In the presence of stochastic noise, these two rates will be different. In order to properly calibrate lithography simulators, the difference between these two definitions of development rate should be quantified. Using theoretical derivations and a stochastic (Monte Carlo) resist simulator, the propagation rate of a resist surface is characterized in the presence of stochastic variation in the resist deprotection concentration and a nonlinear development rate response. The resulting propagation rate can be more than an order of magnitude higher than for the case of no stochastic noise. Correlation in the development rate creates an effective surface inhibition over a depth into the resist of several correlation lengths. These results show that the differences between microscopic and macroscopic dissolution rate can have an important effect on how development rate models should be calibrated, depending on their use in continuum or stochastic lithography simulators.Note that while structured abstracts are typically longer than traditional ones, the 166-word length here is right on target for most journals. If anything, the approach and results sections could have been expanded slightly. Additionally, the structured method of abstract writing also helps to avoid useless but all-too-common phrases like in this paper and we report or will be discussed. The abstract should talk about the work, not about the paper; phrases like is discussed turn your abstract into a table of contents rather than a summary of the paper. Do not use the first person (I or we or the author). Also, there is rarely a need to use phrases like new or novel in the abstract, since it is only the novel results that should be mentioned. You know that you are finished writing the abstract when there are no words that could be taken out without changing the meaning.
This is all well and good, but JM3 does not use structured abstracts. Maybe it should, and maybe it will someday, but today it does not. The structured abstract is still important because it can be used in what I call the structure method of abstract writing. The method is quite simple. First, write a structured abstract. When you are finished and satisfied with the result, simply delete the subheadings and combine all the lines into one paragraph. Finally, reread this new abstract, and change sentence beginnings to increase readability and flow, if needed (though usually this will not be necessary). The result will be a well-written and effective abstract with most of the benefits of a structured abstract. To illustrate, here is an abstract for a paper that I am just now finishing. First, I wrote a structured abstract: Background: Photoresist development rate can be defined microscopically (the development rate at a point) or macroscopically (the propagation rate of an average resist height). In the presence of stochastic noise, these two rates will be different. Aim: In order to properly calibrate lithography simulators, the difference between these two definitions of development rate should be quantified. Approach: Using theoretical derivations and a stochastic (Monte Carlo) resist simulator, the propagation rate of a resist surface is characterized in the presence of stochastic variation in the resist deprotection concentration and a nonlinear development rate response. Results: The resulting propagation rate can be more than an order of magnitude higher than for the case of no stochastic noise. Correlation in the development rate
Editorial
errors, after just plain sloppiness, would be changes made to the paper after the abstract was written. Such errors and inconsistencies can largely be avoided by leaving the abstract-writing task till after the body of the manuscript is completely finished. The structured abstract can help make the abstract more informative, but it is still up to the diligence of the authors (and journal editors and reviewers) to make sure the abstract is accurate. There is a three-part test that should be applied to your abstract when you are finished:
It is probably impossible to define a universal procedure for creating a good titlethere is no equivalent structure method for writing a title. There are some basic guidelines, however, that make use of the structured abstract to guide the creation of the title. In general, the title should reflect the aim and approach of the work. Depending on the audience (and the specificity of the journal), some of the background may have to be included. Rarely are results and conclusions even hinted at in the title. Lets look at each of these items through the use of an example. Unlike the worlds of newspaper reporting and marketing press releases, the title of a scientific paper should describe the aim of the work, not the results. Thus, a good title might be Impact of temperature and pressure on the compositional uniformity of sputter-deposited aluminum alloys The following news-style title, on the other hand, is not appropriate: Optimizing temperature and pressure improves sputter-deposited aluminum alloy films Note that the good title is essentially a statement of the aim of the work. Often it is important to mention the approach used as well, though an experimental approach is generally assumed if it is not mentioned. If the study had been based on simulation (or some other approach), however, this would generally be included in the title: Impact of temperature and pressure on the simulated compositional uniformity of sputter-deposited aluminum alloys The title should be as specific as possible while still describing the full range of the work. For example, if only one aluminum alloy was being studied, that specific alloy should be mentioned in the title. If only aluminum alloys are studied, the title shouldnt say sputter-deposited metals or sputterdeposited alloys. On the other hand, the title shouldnt say aluminum alloys if gold was also included in the study. If the title had said uniformity rather than compositional uniformity, the reader could easily have believed that the paper was about thickness uniformity or some other parameter. And if only sputter deposition was studied, then leaving this information out would make the title insufficiently specific. A conflicting goal of the title is to be as short as possible (in 2011, JM3 titles ranged from 4 to 21 words in length, with an average length of 11.5). Specificity can often be improved through the use of more words, but a title that is too lengthy may not be read.8 Finding the best compromise between descriptiveness and brevity is where the art of authorship comes in to play. Going back to our example, here is a title that sacrifices too much specificity to obtain brevity: Impact of process parameters on the uniformity of aluminum alloys A good test for your title is to answer these questions: Does the title of your manuscript, seen in isolation, give a
4 Titles
When the abstract is written, it is now time to write the title. Unfortunately, it is against human nature to write the title last. Instead, the title is often the first thing written, at the top of that blank document that will soon become your manuscript. It is important to consider these first words as the working title. When the manuscript, and the abstract, are finished, it will almost surely be necessary to revise the title.
Editorial
full yet concise and specific indication of the work reported? Would someone interested in the exact topic of your paper, reading this title, be inclined to read the abstract? Avoid being overly clever with the titlea pun or a play on words may be great fun, but it is unlikely to help your article be found by a search engine (and can be easily misunderstood by an international audience). Titles should also be free of jargon unlikely to be understood by those not intimately familiar with the topic, and should not contain acronyms or trademarked terms. The overall goal should be a title that is clear and informative.
5 Keywords
This brings up the next topickeywords (also called subject terms). We are quickly passing out of the days when most people find your article by flipping through the print version of the journal. Today, your article is unlikely to be widely read unless it comes up relatively high on a Google or Google Scholar search-results list. The first and most important thing you can do to insure that your article is found by readers looking for it is to do a good job of writing the abstract and title. Following the advice given above should help. After that, you must decide on appropriate keywords. The important idea behind identifying the key words to be listed under the abstract as subject terms is simple: if you were looking for an article on exactly the topic of your manuscript, what key words would you type into a search engine in order to find it? Chances are, you would start with only two to four words or phrases. If that resulted in too many hits, or too many off-scope articles, then you would refine your search by adding one or two more phrases. These are the words or phrases (plus all of their common variants and synonyms) that should be included in the list of subject terms. Once you have a good list of keywords, go back and look at your title and abstract. Are these keywords found in the title and abstract? If not, someone searching for your article may easily miss it. The most important keywords should be found in the title, and in the abstract several times.
provide many of the benefits of a structured abstract for journals (like JM3) that do not (yet) use structured abstracts. This structure can also aid in the writing of the title, using information from the aim and approach subsections. To be sure, I have made just about every mistake described in this editorial. But I can certainly tell that as I have adopted the structure method of abstract writing my abstracts have become far more informative and useful. And if I keep practicing my own advice, I suspect that I will continue to improve. Someday, JM3 may adopt a formal structured abstract for all papers. Until then, I think the structure method of abstract writing described here might help you as much as it has helped me. Combined with a properly devised title, you will then have a paper that is off to a very good start.
References
1. R. G. Driggers, Editorial: how do you write a great abstract and why is it important?, Opt. Eng. 49(6), 060101 (2010). 2. W. Rhodes, Editorial: the abstract as a marketing tool, Opt. Eng. 49(7), 070101 (2010). 3. National Information Standards Organization, Guidelines for abstracts, ANSII/NISO Standard 239.141997 (1997). 4. R. N. Kostoff and J. Hartley, Open letter to technical journal editors regarding structured abstracts: this letter proposes that structured abstracts be required for all technical journal articles, J. Inform. Sci. 28(3), 257261 (2002). 5. J. Hartley, Current findings from research on structured abstracts, J. Med. Libr. Assoc. 92(3), 368371 (2004). 6. C. Zhang and X. Liu, Review of James Hartleys research on structured abstracts, J. Inform. Sci. 37(6), 570576 (2011). 7. R. M. Pitkin, M. A. Branagan, and L. F. Burmeister, Accuracy of data in abstracts of published research articles, J. Am. Med. Assoc. 281(12), 11101111 (1999). 8. H. R. Jamali and M. Nikzad, Article title type and its relation with the number of downloads and citations, Scientometrics 88(2), 653661 (2011).
6 Summary
A structured abstract is a proven way to give readers the information they need in an accessible and readable format. The structure method of abstract writing proposed here can