What parts of scientific articles should be skipped and when?

[5U03N15]

I heard that only reading the abstracts of articles and then using them to make claims is bad. However, I often find that much of the information I read useless for me. For example, often the introduction gives me no useful information because I'm already introduced to the subject or the abstract gives no useful information because I already know the article is what I'm looking for. Much of the information in the methods section also seems unimportant to me as I want to learn about how reliable a study is (such as by looking at sample size of if it's a randomized controlled trial, correlational study, etc), and I'm unconcerned about the specifics of the experiment.

 

So, what parts of scientific articles should be skipped and when?

[hypervalent_iodine]

Basing your interpretation of an article from its abstract is often a bad idea. That being said, an abstract is a good way to screen papers (as in, figure out if they're worth reading or not). If you actually need to cite information from it, you would have to go a bit deeper into the results. Which parts you chose to read and which parts you skip I would think is something you determine on a case by case basis. It depends largely on why you're reading the paper in the first place and what you're hoping to extract from it.

 

It can be hard to know straight away how reliable a paper is without some amount detailed reading if the paper is new or is reporting something novel / controversial. Otherwise, I think that it can be a fairly simple process provided you already have a decent knowledge of the area and the work that has already been done in it. If the results of a paper are more or less in line with the literary consensus, then it should be okay. You don't need to read much beyond the abstract to determine this. The journal that a paper is published in can give some information about the quality of a paper, but this is not always true and I would be very cautious of using it as a definitive measure unless you know for sure that the journal is terrible. Bad papers can be published in good journals and vice versa.

 

If you want a more in-depth review of a paper's worth, it will require a bit more effort and reading. You would probably need to do this if the paper is making novel claims. An analysis of how good the statistical aspects of the paper are is maybe one indication, though it requires that you know how to interpret these parts of the paper (which you probably should) and of course, many papers don't need this sort of analysis. You can check subsequent articles that cite the paper, but that doesn't work if the paper is new or in an area that is not researched by many groups. In my experience in synthetic organic chemistry, both of these things can be a problem. Many papers have no statistics in them whatsoever and the discipline is both diverse and obscure enough that it is common for a good paper with well-executed research that is several years old to have barely any citations. In these cases, you have to use your own knowledge to pull apart their methods and look at their supporting information (spectra, etc.) to see if it makes sense or otherwise, just repeat the experiment yourself.

 

I'm sure there are other methods people much more familiar with this sort of thing use that I may be missing, so hopefully someone else will contribute their learned opinion on the matter.

[John Cuthber]

The ideal answer to "What parts of scientific articles should be skipped and when?"

is none and never.

 

The real world answer is it depends.

What are you trying to learn from the paper?

Are you just trying to copy some part of a procedure they used to see if it works for some other process?

If so you can probably just read that part of the experimental details.

Also, logically, you don't need to read any bits that you already know.

How do you tell which ones those are?

Well, unfortunately, you have to read them to check.

 

The least bad answer I can give is that you need to learn to skim read papers.

Of course, if someone has asked you too review the paper, you need to read the lot- probably twice or more.

[Klaynos]

Cite them, read them, it's not a terribly big ask...

 

But if you're just generally reading to find what you want to cite or if a particular question has been answered you'll learn over time what you can reasonably skip, that is something that is best learnt from doing rather than asking someone to tell you. You would often then read the whole thing of it's useful.

[ajb]

This will depend on the subject and what you want from a paper. For me, I start with the abstract and then have a scan at the references. Do they cite the people I would expect them to? If not why not? Then the introduction; you know your way round the literature, but do they? They may have missed something okay, but you need to know these people have a good idea of what they are doing. Then I may only read carefully the theorems etc that I need. The proofs I may read carefully if I need them, but quote often the result is more important to me than how it was obtained, on first reading anyway. I guess for subjects other than mathematics/mathematical physics etc substitute 'proof' with 'details of methodology'.

[CharonY]

John and ajb have the right approach. You do not utilize a paper efficiently by reading them front to back in a neutral way. You have to know what you want to get out of it (is it the method, the outcome, they hypothesis?) and read it accordingly. For example, if you have a specific claim and support it with a citation, you may either want to look at their conclusion if it is really what you claim, the discussion, where they at least hypothesize something like that or even in the results section, if your claim is based on their data.

[Dekan]

Does anyone actually read long scientific papers?

 

Or just glance at the introductory paragraphs, to get some idea what it's supposed to be about.

Then skim rapidly through the mass of verbiage, figures and diagrams in the middle.

 

Finishing by a look at the last paragraphs to see what the purported conclusion is. And give it a pinch of salt.

Edited October 24, 2014 by Dekan

[ajb]

Does anyone actually read long scientific papers?

How long is 'long'?

 

In my field 30 pages would be quite long, while I do see more sting theory articles near 100 pages.

 

I tend not to read a paper from page one to the final page in on go. I may skip sections that are not needed for what I want to do right now, but I will end up reading them at some stage.

[CharonY]

I would say it depends a lot of familiarity, too. If you are familiar with methodology, you can just glimpse at the graphs to assess the quality of data and what level of conclusions can be drawn from them. If the paper is from a different field, I am much more likely to read it completely and then add further paper to fill my knowledge gaps, before reading it again.

[imatfaal]

I have an important case report I have to read at present - it is about 75 pages of dense legalese. I will read head-notes (similar to abstract first), then I will check the statutes used and cases referred (parallel to sci citations), next I will skim through the case and flick across to read any important precedent cases mentioned, I will then pray that someone has written a nice brief case note on the judgment and read that/those, and then I will buckle-in, get some coffee, and read the case through properly making notes and ensuring I understand each section before moving on.

 

I would be very surprised if this form of approach is not repeated through many disciplines. Going straight to a deep reading is almost impossible for those without godlike knowledge - and just reading the abstract and the conclusions only serves to tell you whether you need to go back and read properly