The Science of Scientific Writing

The Science of Scientific Writing (Gopen and Swan, 1990) is a nice paper for anyone who has to write anything technical.

Since we read English left to right, we come to expect context at the start of our informational blocks (sentence) and data at the end. Messing with this scheme tends to confuse the reader, making them work very hard to understand your point. Some general rules from the paper I thought I might write down:

  • Subjects and verbs should be together. Don't stretch out the sentence with extraneous information between what you are talking about and what you are saying about it. e.g.

    The cache, a 128 entry, dual ported, 4-way set associative design with on core tags but off core data, was found to be the major bottleneck of the system.

    The cache is the subject, and was found to be is the verb. They should not be separated by all the other stuff.

    The cache was a 128 entry, dual ported, 4-way set associative design, with on core tags but off core data. It was found to be the major bottleneck of the system.

  • Similarly, discard material which does not discern significant connections. As in the above example, it is quite easy to get into a habit of putting largely irrelevant information right in the middle of a sentence.

  • The stress position is the area around the end of a sentence. When reading, we tend to want the opening of the sentence to lead us in gently, reaching the big conclusion at the end. We then pause briefly, think about it, and move onto the next sentence. By making sure the stress position stresses the right information, we help understanding.

    10% of time was spent handling TLB refills in our flablah benchmark, which simulates a transaction processing workload.

    The above sentence hits us with the important information first, leaving us to finish the sentence and then wonder "what was that figure again". It could become

    Our flablah benchmark simulates a transaction processing workload. For this benchmark, we found that 10% of time was spent handling TLB refills.

    By keeping worthy material in stress positions we can make things much easier to read.

  • The first part of the sentence, the setup component, is the topic position. One option is to have the topic position linking backwards, so we know what the upcoming stress position is referring to. This is more or less what we did in the example above ("For this benchmark..."). Otherwise we want the topic to introduce the person, thing or concept which we are about to talk about; to quote the paper "Bees disperse pollen" and "pollen is dispersed by bees" are both very similar, but one suggests we are talking about bees and one pollen. You need to make sure you are setting up the sentence so that the emphasis of the stress position relates the foremost concept in the readers mind.

  • Be careful that topic positions keep telling a consistent story. Seems logical, but it is easy to get into situations where you introduce too many ideas without relating them to what has come before.

I'm interested in any other guides to writing people have come across.