37

I need to start learning about a new research topic, I've been collecting articles for about a year now. I collected +400 articles, spanning a range of ~40 years.

I haven't had the time to read them since I was working on finishing a few papers, I'm finished now so I want to get on it.

The question is: should I start from oldest to newest, or from newest to oldest?

I can see some pros and cons in both approaches:

Oldest to newest:

  • Pro: it feels right, since that is the way the research actually happened.
  • Con: It will take me a lot of time to become acquainted with the latest developments in the field, and I am not able to start working on it until I do.

Newest to oldest:

  • Pro: I will quickly develop some knowledge (not in depth though) into the actual state of the field. This allows me to start working, at least on minor things.
  • Pro: It will alert me of wrong paths taken previously in the field, so when I find them in the older literature I'll be already aware.
  • Con: It will be considerably more difficult to follow, specially at first, since I'll be diving into the analysis of the latest developments of an unknown field.
  • Con: I'll necessarily be jumping back to older literature anyway, risking going down the rabbit hole of research.
  • 11
    Are any of the papers in your collection reviews? Regardless of whether you read forward or back through time, beginning with reviews should help provide a wider picture which may make it easier to work through the rest of the collection – Ian_Fin Sep 7 '16 at 15:05
  • 24
    This question presents a false dichotomy; why assume that chronological order (or reverse chronological order) is the best way to go through these papers? – ff524 Sep 7 '16 at 15:11
  • 14
    I would suggest in the future to not do this again. You will likely find that many of those 400 articles were not worth collecting, which you would have determined rapidly if you had been reading things as you got them. Coming up to speed takes time, not just to read, but also for your brain to wrap itself around the topic. – Jon Custer Sep 7 '16 at 15:36
  • 4
    Are there any textbooks covering the field? – Patricia Shanahan Sep 8 '16 at 4:58
  • 7
    My guess is that you collected too many papers and you will likely not read whatever you leave for last. – Bitwise Sep 8 '16 at 9:40
44

Neither.

First, identify the "important" subset of your 400+ papers and read those. Second, try to identify the "good" subset of the papers and read those. Then, (if at all) read the other papers; perhaps you are doing all this with the intention of writing a review, in which case it makes sense to read oldest-to-newest to get a sense of the history.

The most obvious metrics which you can subsitute for "importance" are number of citations (these will likely include review articles) and the ranking (pick one) of the journal of publication. Note that "important" does not mean "good", "honest", "well-written", "reproducible", etc.

For a second sweep, focus more on what you actually do consider to be "good", based on authors and journals that you have found to be of consistent quality.

  • 4
    The issues I see with this answer are the "importance" metrics. The number of citations will be biased towards the older articles, and the ranking of the journal is of no use in my field (Astronomy) since there are 4-5 main journals which concentrate pretty much all research, and all of them have very similar IFs. – Gabriel Sep 7 '16 at 18:37
  • 3
    I very much agree that what I've suggested for "importance" are flawed metrics, but barring access to an expert in the field who can guide you, I can't think of better suggestions. Maybe you can weight your "number of citations" rating to take age into account? – Patrick Sanan Sep 8 '16 at 6:34
  • 4
    I completely agree with this answer, and would suggest using reviews to determine importance. It would seem to me that if you read a review of a literature and don't finish with a sense of what were the key papers in that literature then the author has done something wrong. – Ian_Fin Sep 8 '16 at 8:25
  • I agree, except for the part about finishing the article before clearing up the parts you don't understand. Never go on past a word you don't understand thinking, "Well, I'll catch that later." If there is a part of the article you don't understand, find out about it right then. You might look ahead in the text to see if it's explained, or you might check other sources, but the action of clarifying what you are reading or getting terms defined is separate from the action of simply reading or studying (even if it involves looking ahead in the current text for a definition.) – Wildcard Sep 9 '16 at 1:36
19

If the new reviews are too complex you might need to start with some older ones, but I would really suggest starting from the newest publications. People already spent years and years collecting all the old research, sorting through the mess and writing "concise" reviews about it. There's no need to do the reading to repeat that.

You would need to follow your nose / go down the rabbit hole sometimes, but at least you'd know there's a hole. If you'd start from general/random papers 30-40 years old you'd just read dirt most of the time, research that never went anywhere, or that turned out to be wrong.

Remember that you're studying some scientific field, not the history of the field.

7

Neither direction works well on its own. I have never profited from a literature review to learn a topic except in this way:

  1. Skim the Introduction, "Related work" and citation lists of current research. Find sources that introduced new ideas or that constitute the definitive treatment of them, as evidenced by the fact that they keep being cited. Possibly iterate this process once or twice.
  2. Read those older sources, in temporal order.
  3. Now read those current texts that seem relevant.

A scientific paper simply lacks the space to properly explain the concepts on which it builds, so inevitably you will have to achieve roughly the same experience and mind-state that the author had to understand it well. Reading what they read is hands-down the simplest way of achieving this.

  • 1
    +1 for highlighting the fact a scientific paper lacks the space to properly explain the concetps on which it builds. It takes years to achieve the same experience and mind state of an author – gansub Sep 8 '16 at 13:18
4

It depends on whether there's a lot of building on understanding certain concepts in order to understand some other things. If so, then you'll need to control the order you read things in rather carefully, so you don't end up rather lost.

If not, then follow your nose. Dive in where it seems most interesting, and then work your way around with your curiosity as your guide.

When this journey of discovery starts to slow down, then you can go through them more methodically. At this point you will be familiar enough with the area that you will be able to do some skimming (e.g. it looks like this one was a preliminary form of the idea that was later more fully fleshed out in such-and-so).

Make sure you make a guide for yourself, which you should update from time to time, of what your goals are, in other words, a list of what you are trying to accomplish and understand through this big reading project.

4

If you want to get an overview over a field, find an overview paper or a book.

If you want to understand a special theory/method/theorem, take the paper where it is explained and gather all the relevant references that the paper uses. In some cases you have to dig even deeper and get the second level of relevant references (relevant = a crucial part of the paper depends on it).

Then I would go back and forth between these papers: If you struggle to understand the material, go one level deeper, if you get bored by irrelevant stuff, go one level higher. Try to sketch the relevant parts and their relation in your own notebook. That would be the approach I prefer.

1

Conducting your reading from a purely chronological perspective seems incredibly arbitrary.

Jumping in at the deep end and reading published journal articles would unlikely be an effective way of learning - journal articles are typically aimed at an audience that already knows about the subject at had. Hence, they would not furnish you with the core concepts and fundamentals that you would need as a newcomer.

If learning an entirely new subject, my approach would be to read several recently published introductory texts, such as those aimed at undergraduates or collections of essay, ideally from a variety of different academic publishers or university presses.

I would also speak with colleagues, associates and subject librarians for advice and reading lists.

The aim is to get a feel for:

  1. Underlying concepts and theory of the subject
  2. Points of contention or disagreement
  3. Areas of uncertainty, current ground breaking research
  4. Seminal texts or thinkers

I would then read the seminal texts in the subject with the other three points in mind, together with recommended reading from the people mentioned above.

From that point on it would be clearer whether reading old articles would be more or less fruitful than starting with the later work (eg if something has been refuted it might not be a top priority to read or if a body of recent work has utilized an old article that might be where you should look).

1

I don't think it's productive to collect hundreds of articles before you even start reading. Until you've made yourself familiar with the subject, how can you know which articles are useful, and which are not?

I think the best thing to do is identify one or two well-written review articles. If the field is not too distant from your own, you should be able to read a review article from start to finish. Otherwise, you can at least read it to get a sense of what is going on, and from there identify a few more articles which will allow you to gain some more depth.

And my recommendation is to read review articles from "most recent" to "older" --- there is nothing scientific about naivete', in my opinion. Knowing more does not spoil the previous work.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.