Present, Practice, (Sleep?), Produce

Off-line learning? What, pray tell, is that?

Off-line learning, aka "slow learning", refers to performance improvements that don't show up until several hours (and sometimes several days) after practice or training. Sleep is often but not always required for off-line learning (OLL from here on) to occur.

Tell me more.

Typical experiments go something like this: A bunch of people are trained to perform some task and then their performance is tested. Eight or so hours later they're tested again and their performance is found to have improved. Often, the people who slept during that eight hours perform significantly better than those who did not.

How and why does this happen?

Hotly debated questions. The most widely held scenario seems to be this:

Performance in a given task can be enhanced by long-term brain changes including

the alteration of the electrochemical properties of certain synapses
the construction of new dendrites and synapses (What are they?)

These changes happen to some extent during training, but they happen more effectively when

the concentrations of certain chemicals in certain parts of the brain have been altered (For example...)
large populations of neurons are firing in a rhythmic, synchronized fashion

and these conditions come about during sleep, perhaps because they are not compatible with, say, eluding an enraged woolly mammoth.

So, during sleep, the neural firing patterns used during training are to some extent replayed and strengthened, or optimized, or...something. Thus, performance improves.

You speak of people being trained to perform tasks. What are these tasks you speak of?

I'll try to sum them up.

Common Tasks used to Investigate Off-line learning
Task Type	Example(s)
Motor procedure	Tapping fingers to thumb in a set sequence
Cognitive procedure	Solving sequential puzzles such as the Hanoi Towers puzzle
Visual perception	Detecting diagonal lines amid a field of horizontal lines
Auditory perception	Detecting changes in tone sequences; understanding systematically distorted words
Visuomotor (visual perception + a motor procedure)	Tracing images in the mirror;the tapping buttons in response to visual cues
Declarative recall	Recalling sets of paired words; recalling memorized texts

But the evidence for OLL isn't equally strong in all areas. For motor and visuo-motor procedural tasks and for visual discrimination tasks, it's very strong. For auditory discrimination tasks, results are largely positive, but there are only a handful of studies. Results for declarative memory tasks and cognitive procedural tasks have been mixed, and expert opinion is divided as to whether any significant OLL occurs.

And I should tell you that at least a few experts in the field don't believe OLL plays a significant role in human learning at all. ( See an abstract from the apparent leader of this faction.)

But even if we accept that OLL happens in the kinds of tasks used in these studies, can we generalize to second language learning? Isn't language learning an animal unto itself?

I don't believe it. See Nick Ellis' chapter in this book for a good summary of the position that language learning relies on the same basic processes used in other learning tasks.

OK, I will. In the meantime, could you specify how you think the tasks you mention above are related to language learning?

Sure. The process of learning to pronounce new phonemes, and the sequences of phonemes that make up words, and the sequences of words that make up high-frequency multi-word phrases, is not unlike the process of learning a finger tapping sequence. It is essentially a motor procedural task. And auditory discrimination (of phonemes, words, word groups) is certainly one requirement for listening comprehension. And the visual discrimination skills tested in the forementioned experiments could be relevant to the task of learning to recognize Roman characters and word shapes.

OK. So let's assume I believe in OLL, and I believe that it applies to processes relevant to language learning. What practical implications does that have for me as a language teacher?

Well, the temptation is to say something extreme: "No student should ever be expected to use any language item, be it a piece of vocabulary or a structural element, in a meaningful way on the same day that that student it is introduced to that language item. Present, practice, sleep, produce." But that would be silly. There are too many unknowns. For all we know, a tedious, halting struggle with a communicative (?produce?-stage) exercise may be the very best way to trigger OLL. Still, I think it would be reasonable for us teachers to allow the probable existence of OLL to influence us a bit-- to nudge us in the direction of stretching out our timeline. For example, suppose I'm planning to finish up a particular class period with a communicative activity that that lesson has been working towards. After that I plan to introduce some new vocabulary that will be used in tomorrow's lesson. Now suppose I notice that time is running out. There's no time for both of those activities. Which should be postponed? Common sense might say to postponed the new vocabulary; do the communicative activity so we'll had a nice tidy "present, practice, produce" lesson. With OLL in mind, however, I might choose to save the communicative activity for the next day and get started with the new vocabulary. That way, both the preparation for the communicative activity and the initial learning of the new vocabulary will have been "slept on" as we begin the next day's lesson.

Another practical implication of OLL in the classroom is that we might simply want to let students know about it. Let me explain why with a little analogy. Imagine Dan and Dave, two Olympic weight lifters. They both head into the gym one day, and, after warming up, they both try dead lifting 500 pounds. They both manage five repetitions. They both do four more sets of this exercise, each time managing five repetitions, except for the last, where fatigue limits them to four. As Dan leaves the gym he feels good. He knows he put in a good workout and he knows that over the next couple days, as long as he eats well and get plenty of sleep, his muscles will be developing. He'll be getting stronger. Dave, on the other hand, leaves the gym depressed. He started at five reps, made no progress, and wound up being able to do only four. He doesn't know that strength development is a delayed effect, so he can't feel good about his workout.

A student who doesn't know about OLL is a little bit like depressed Dave. It's not so extreme, because with language learning, unlike with muscle strengthening, real progress can indeed be made over the course of a single training session. However, if it's true that further progress is made after the training session, why not let students know about this? Why not give them that extra satisfaction that Dan feels as he leaves the gym?

Lets look at a particular example. If students are having trouble pronouncing "executive, legislative, and judicial" even after doing a lot of slow-motion work on those words, wouldn't it be nice to be able to say, "Well, I think you'll find it easier to pronounce those words tomorrow" rather than just, "Well, don't worry about it. We can work on it more tomorrow."

Do you say things like that?

Sometimes, but usually not without some hedging. As much as I believe that language learning depends on processes common to other types of learning, I would still like to see some studies that demonstrate OLL specifically for language learning before I start presenting it as something we can rely on.

Well, let's hope that such studies are done.

Yes, that would be nice.

Do you crave more detail? A good place to start is this article by Robert Stickgold. Unfortunately, you have to pay to get more than the abstract online.

And if you enjoy abstracts, by all means check out PubMed. For starters, you might try cutting and pasting the following search strings.

A narrow search:
Language[title] AND sleep[Title]

A broader search:
(Sleep[title] OR REM[Title] OR NREM[title] OR non-REM[Title] OR nap[Title] ) AND (learning[Title/Abstract] OR memory[Title/Abstract] OR memories[Title/Abstract] OR performance[Title/Abstract] OR knowledge[Title/Abstract] OR recall[Title/Abstract] OR "skill-learning"[Title/Abstract] OR skill[Title/Abstract] OR discrimination[Title/Abstract] OR Auditory[Title/Abstract] OR visual[Title/Abstract] OR perceptual[Title/Abstract] OR acquisition OR consolidation OR cognitive[Title/Abstract] OR declarative[Title/Abstract] OR non-declarative OR nondeclarative OR procedural[Title/Abstract] OR verbal[Title/Abstract] OR language[Title/Abstract] OR linguistic[Title/Abstract] OR phonological[Title/Abstract] OR episodic[Title/Abstract] OR motor[Title/Abstract]) NOT disorder NOT dysfunction NOT impairment NOT Epilepsy NOT apnea NOT disabled NOT drug NOT "false memories" NOT "mg per day " NOT dementia NOT pregnancy NOT neonatal NOT lesion NOT insomnia NOT driving NOT Alzheimer's NOT "Frontal lobe" NOT chronic NOT cancer NOT residents NOT resident NOT handedness NOT (dreams[Title] OR dreaming[Title]) NOT deprivation[Title] NOT Aging[Title] NOT Fear[Title] NOT disease NOT erection NOT erections NOT disabilities[Title] NOT shift[Title] NOT Impotence NOT Syndrome NOT Light[Title] NOT Cortex[Title] NOT Surgery[Title/Abstract] NOT Spaceflight NOT Astronauts NOT Respiratory[Title] NOT Deprived[Title] NOT deprivated[Title] NOT Schizophrenia[Title] NOT Inertia[Title] NOT Infant[Title] NOT Infants[Title] NOT Schizophrenic NOT Migraine NOT Adolescents[title] NOT Children[Title]NOT Child[Title] NOT "Sleep learning"[Title]