Friday, April 9, 2010

Optimal Learning & Training Schedules

An interesting question I have often considered has to do with the most efficient way to use time, in order to prepare for something, or to learn.  This is relevant in psychotherapy, in terms of helping therapeutic change progress at the most optimal pace.

To formalize the question, consider the following:

1) If you had 100 hours to learn something (e.g. to memorize a text; to learn a foreign language; to learn a musical instrument; to understand a set of complex ideas; to learn a new sports skill; or overcome a psychological symptom), how would you distribute these hours, so as to optimize the therapeutic change?  Would it be 10 hours per day, for 10 days in a row?  Or 1 hour per day, 100 days in a row?  Or 1 hour twice per day, for 50 days?  Or 1 hour per week, for 2 years (!?) 

2) Another set of constraints on this problem would be this -- if you had 10 weeks to learn something, a maximum of 10 hours per week to learn it, and a maximum of 10 hours on a single day to spend, what would be the best way to work?  Would it be 10 hours every Monday, for 10 weeks?  Or 2 hours every weekday? Or 1 hour twice a day on weekdays? 

It interests me to note that answers to this type of question come from different fields of research, from cognitive psychology to education to athletic training.

The most sophisticated piece of research I found regarding this issue is described in the following article:

Pavlik et al., "Using a model to compute the optimal schedule of practice," Journal of Experimental Psychology: Applied, v14 n2 p101-117 Jun 2008

The research shows that, in general, "spacing" is far superior to "blocking" in terms of time management or study scheduling.  That is, if you have 10 hours to learn something, it is better to split the time up into short blocks, with rest periods in-between, rather than spending all 10 hours at once.

Pavlik's article includes a much more sophisticated analysis:  for a memory task, items which were more difficult to remember were reviewed with a shorter interval, whereas easier or more well-learned items were reviewed with longer intervals.  As each item became more well-learned, the spacing increased gradually.  To review something too soon would not be using time well:  not only could that moment be used more efficiently to review something more difficult, it also does not develop the longer-term memory of the item as well.  It is most optimal to review something just as its memory is starting to decay.  These memory decays take place over a longer and longer time, the more you have learned something.    To review something with too long an interval between study trials would also be inefficient, as too much forgetting will have taken place, and an inefficient investment of time will need to be spent re-learning the same material.

Common practices in studying or practicing  include the following:
1) familiar or easy material is revisited too much:  it is often inefficient to review something you already know well, unless this causes you to  develop some new insight about it.
2) unfamiliar material is reviewed in large blocks of time (cramming) -- this is profoundly inefficient, and does not allow for long-term learning.

Pavlik's experiment also confirms that high levels of accuracy should be sought, right from the beginning, so as to maximize efficiency.

In summary, Pavlik's work shows that one should space learning efforts.  When just starting out, the spacing interval should be brief, with enough frequent review to master what you have just learned.  With the material mastered on a short-term time scale, the spacing interval can be extended, just enough to make the review slightly challenging.  This process continues, with gradual expansion of spacing intervals, until the material is permanently learned.   Once the spacing interval extends for days, weeks, or months, the learning will probably be permanent. 

The research is very incomplete on this matter, for a number of reasons:

1) the complexity of each individual learning task needs to be taken into account.  For example, if one is trying to solve a complex physics problem, or to comprehend a difficult concept in philosophy, it may be necessary to invest many solid, continuous hours of effort in a "block."  In this sense, each individual "trial" of learning takes place over many hours, rather than over seconds (as in memorizing a foreign-language word).  So, for more complex tasks, fragmenting one's study time could decrease efficiency.  But in a general sense, it will be extremely inefficient to try to "cram" in order to learn how to do complex physics problems.  The "spacing" needs to take place generously, but with each space over a period of days--allowing you to complete individual problems--rather than hours.

2) It remains true that action is required in order to learn.  If accuracy is valued so highly as a priority that action does not take place, than learning cannot occur.  So, for example, in order to learn a new language, one must practice speaking it, or using it.  If one is excessively meticulous about accuracy of vocabulary or grammar right from the beginning, and therefore one is silently contemplative in a conversational language class,  then the action cannot proceed, and instead a stifling self-critical process will inhibit learning and engagement. 

3) The existing research does not account for the powerful effects of "constraint-induced" neurologic change.  Immersive processes may permit the brain to develop new pathways much more efficiently -- anything less than immersion allows a continuing neural pathway of least resistance.  The Taubian ideas about stroke rehabilitation exemplify this phenomenon:  neurological recovery may be much more complete if the brain is not allowed to by-pass or compensate for the disabled body part:  in this way the brain's energy and capacity  and plasticity may be directed towards regaining lost function.  So, in this sense, a continuous "immersion" in a study process may be more effective than any sort of "spacing" regime.  The immersive experience would be a "block" lasting months at a time, continuously.  Of course, there could be smaller spacing effects within this.  Addiction recovery requires similar "immersion" in an abstinence process.  The neurological recovery from the addictive process could then proceed over months or years (typically a year being a significant milestone).   

4) Sometimes, large blocks of time can be useful.  Even though it is not the optimal schedule for using time, in terms of memory formation, it may be optimal on other levels, such as with developing the ability to maintain longer periods of attention in the subject matter, with developing deeper insights about patterns within the subject, or with developing a richer sense of community or identity around the activity.  Thus, a "weekend retreat" experience of something can be educationally powerful, even if the same number of hours spread over several weeks might be a more optimal use of time, if simple memory is the only consideration. 

Here are some references to other  research which addresses this question:
Extinction more effective if spaced rather than in a block of time.
Variable practice (involving several versions of a skill) has advantage over constant practice
Random training in basketball has better retention after 1 year
Contextual interference improves learning skill
Blocked practice better for immediate acquisition, random practice better for retention (long-term).
blocked practice better for acquisition, random practice better for retention (long-term) --pistol shooting
variable practice better in tennis
knowledge of results (KR) -- more is not necessarily better.  less KR improves results after a delay, especially if tested without KR
 shuffled practice of math problems vastly superior to standard blocked practice, when measured 1 wk later
spacing better, in general; but if the learner prefers a block strategy, then spacing less advantageous
1988 psychology article reviewing spacing as optimal memory strategy
1 day per week courses -- much inferior to 3 days per week

No comments: