To Chunk Or Not To Chunk: Not One, But Two Questions

In the spirit of conservation, I’m posing two questions as one.

It turns out that my use of the term “chunking” or “chunk” has been unintentionally imprecise. Thank you to Shaun for correcting me on this. I was using the term “chunk” to mean a sub-sequence within a larger sequence. So, at n-back 7 I break the overall sequence of seven into two sub-sequences of 4 and 3, or 5 and 2. Shaun pointed out that “chunking” really means identifying a meaningful sub-sequence to aid mnemonic memory and recall. So chicago is a chunk within the string of letters trhochicagopt.

OK. So that’s the first question out of the way. I’m not chunking, I’m “sub-sequencing,” which is, unfortunately, a much more clumsy and less handy term than chunking. But I’m sure we’ll better it.

So, rephrasing the remaining question: “To Sub-sequence or Not To Sub-sequence, that is the question.”

Rupert questioned the idea of rehearsal when I proposed it as part of a strategy for mastering n=5. Some might also question sub-sequencing, since it seems to undermine the development of working memory capacity. Will has pointed out that even with rehearsal and sub-sequencing the dual n-back test remains immensely challenging, and requires us to swap a whole bunch of stuff in and out of working memory quite quickly.

After the blog discussion with Shaun on Klingberg’s book (is it primarily focus or working memory we’re training beyond a certain point?), and after my session yesterday when I tried sub-sequencing n=10 with 6 & 4, I was thinking about this some more. It occurred to me that holding and swapping distinct sets of information in and out of working memory closely mirrors what we need to do when faced with demanding intellectual challenges in everyday life.

Rarely does a complex mental challenge require us to hold and manipulate just one set of data. Raw digit span capacity applies when transcribing a social security number, for instance, but when solving a multi-faceted problem we typically need to bring one or two parts of the problem into mind, hold them, query them, add another aspect, nudge things around, try an example, etc.

So, I would tend toward the conclusion that extending our single-sequence-recall limit as well as our combined-sub-sequence-recall limit both aid in enhancing out everyday intellectual skills.

Tags: , ,

3 Responses to “To Chunk Or Not To Chunk: Not One, But Two Questions”

  1. nanz says:

    Hmmm…I have tried both methods: dividing into smaller parts and of rehearsal.. neither worked very well for me. I may not come from a place of power here since I am only working on n=6…but here are my 2 cents. I seem to be most successful (so far) with remembering strings like a song. As I replay the last song in my head I am learning a new song. And the key here is I do not rehearse the song to learn it…I put in one sound at a time…A then B then C then D…not continually A, AB, ABC, ABCD…does that make sense? So I think this is another way to “remember”. Time will tell if this is a successful method for longer strings. Your thoughts…??

  2. Shaun Luttin says:

    My thoughts…

    Well, just a comment on what Martin talked about.

    First, thanks Martin for considering chunking further. I appreciate that.

    Second, perhaps another term like “trimming” could work instead of “sub-sequencing”.

    Third, if sub-sequencing doesn’t create meaningful chunks, then sub-sequencing isn’t using LT memory at all. When we are faced with n=11, rehearsing two sub-sequences of seven and four probably uses the same amount of working-memory as rehearsing the whole sequence of eleven does. If the sub-sequences are meaningless, then it’s all working memory.

    Warm regards,

  3. Will says:

    Going back to the Baddeley Hitch model of WM, it seems as if phonological processing requires rehearsing of information before it decays in memory storage. This in itself is not what is difficult about WM. It’s doing this AND paying attention to incoming information. I agree with Rupert that there’s value to improving short term storage in and of itself, but disagree that the dual n-back is the proper exercise to work on the storage aspect. I’d argue this is the difference between short term and working memory; that the latter requires rehearsal of the old amid new flows of information. As Martin points out, correctly I think, this is what high level cognition requires of us.

    Another question to ask: Why is digit span backwards so much more G loaded than Digit span forwards? Digit span forwards is a span task and digit span backwards is WM task. DS Backwards means you have to store and rehearse thereby requiring far more concentration rather than parrot the data back. I think dual n back mirrors DS backwards in this regard, and even more challenging given its dual nature.

Leave a Reply

You must be logged in to post a comment.