(My paper “Can illness be bright? Metaphor comprehension depends on linguistic and embodied factors” has been accepted by the Cognitive Science Conference in July 2017 for a poster presentation. This post is to commemorate the first time I cut my 54-page lab report down to a 6-page manuscript.
The task sounds more daunting than it is. It certainly does not help if you realise that what takes you a month to do, your supervisor has done it in an afternoon (True story). My supervisor, Dermot Lynott, gave me two pieces of advice:
- Paste the paper into the CogSci template and start hacking pieces out.
- Better to cut whole paragraphs than individual sentences.
And also a caveat: Do not over-do it. There is no award for cutting it down to 5 pages only to realise it after submission (True story).
The truth is, I was a little disappointed by the advice (Dermot was nothing but lovable though), because I was hoping to know which bit to cut instead of how to cut. Now, I can answer my own question. Here is what I cut, what I wish I didn’t have to cut, and what I wish I did.
- Tables and figures. True, tables and figures make your results section clearer. They are easier to read, so your readers will thank you for saving their time. Unfortunately, they take up a lot of space. Each table for a mixed-effect linear regression could occupy about half a page. So, hack them off and working the stats into the text. You need to pay extra attention when you do this lest you might forget pieces of stats and your reviewer would wonder whether you actually did the test or not.
- Repetitions. When you have a report as long as 54 pages, you are bound to have a lot of repetitions. For example, my paper has two experiments. I would write down a brief summary of the research method and rehash the hypotheses at the beginning of each experiment, even though they have been covered in the Introduction. In this way, even if people did not take the time to read my whole 12-page introduction, nor did they bother reading into the Method, they were still able to figure out the crucial points of each study. Similarly, if you need to discuss theories at the end of the paper, you will have to recapitulate the theories you’ve covered in the Introduction, because your readers may not be an expert in the field as you are, and they will forget what “Career of metaphor” is at the end of the paper. That being said, however, if your manuscript is only 6 pages, you do not need to worry about the memory span of your readers as much. Also, your Discussion can be very brief about what your findings and contributions are, and you don’t have to dig too deep into the theories (at least that’s what I did).
- References and citations. This is a no brainer. Instead of having 10 references for each idea, stick to the most influential one.
- Full names. If the above three ways are not able to accomplish the task, you can also use abbreviations. In my paper, I had two variables, linguistic distributional frequency (LDF) and ease of simulation (EoS). I only used the full name in the abstract and the first time in Introduction. This didn’t seem to raise any eyebrows.
- Details of analysis – use with caution. I omitted lots of details of analysis. This was not an easy choice, and also turned out to be an unpopular choice. For instance, in my analyses, I used a technique called orthogonalisation. That is, I performed a principle components analysis (PCA) to make my two IVs independent from each other. The reason to do this was that I found evidence of net suppression in my mixed-effect linear regression model using the original variables. I expected linguistic distributional frequency to have a positive effect on response decision, i.e. the higher the linguistic distributional frequency was, the more likely a metaphor was to be judged as sensible. The zero-order relationship between linguistic distributional frequency and acceptance rate showed that. Trials with “yes” responses had higher LDF than trials with “no” responses. However, the regression model showed the opposite effect: as LDF increased, acceptance rate actually decreased. This was called net suppression, meaning that LDF actually accounted for the residuals of ease of simulation, thus enhancing the explanatory power of the model. The reason for net suppression could be that LDF and EoS correlated with each other. Thus, I used orthogonalisation to separate them into two independent variables each accounting an unique portion of variance of the response decision. PCA is not normally used for orthogonalisation but for dimensional reduction. That is to say, usually people use PCA to extra less variables than originally put in. However, for orthogonalisation, we extracted the same number of variables, so as to make sure that the extracted variables did not lose any information from the original variables. With varimax rotation, the two new variables became orthogonal to each other and each correlated highly with an original variable, so one new variable was called orthogonalised EoS and the other orthogonalised LDF. As you can see, explaining orthogonalisation in full will take up half a page, but without them the reviewers protested that my Results were too dense.
- * Background literature in Introduction. I should have cut this more heavily, but I’m still on the fence about it. In my paper, I spent the first half of Introduction to lay out theories of metaphor processing that are currently on the market and criticise their shortcomings. Then I brought in my framework. This might have done better for a long paper, but for a brief manuscript, it may be too much of a fanfare. One reviewer said I have spent the first page on what the paper isn’t about. It may work better to save them for the conclusion. I certainly see the point in this. But part of me also want to justify my choice as a personal style. Probably I should be less defensive about this.
That was pretty much it. For me, these tips did the trick. My paper was 5 pages and a half including references in the end. I got three different reviews. The Reviewer No.1 raved about how interesting and crucial the paper was. The Reviewer No.2 only gave two minor comments for clarification. The Reviewer No.3 criticised every bit of it and rejected it as a whole. (I side with the Reviewer No.1. ;D) Of course, I would hope my paper receive nothing but applause and praises. But as I read somewhere, science is not about what’s the best, but what will do.