- Association
- In time
- an electronic professor

An " optimal " sample program: " the formula "


The most effective way to learn a language is to soak itself inside: to visit the country where one speaks it by avoiding the people who know your own language. You will have many troubles because of the misunderstandings, but you will learn quickly.

On the other hand, if you do not have this opportunity, or of limited time, one needs other means. Here are some elementary principles of empirical psychology, and the examples of their applications.

Warning: I am not a psychologist, and the experimental knowledges are simply those which I remember; -), without guarantee.


One do not know all the details of the mechanisms of the memory. But it is known that at a certain moment, there is information which does not exist in the mind, and which later, this information is well anchored in the mind.

Between the two points of time, information must be represented in a certain way in the brain, or between the brain and the elements external (what one hears, sees, etc). This supposes the idea materialist and reductionist according to whom that the mind is a phenomenon which occurs in the neural network of the brain and also in a sense in the " neural network " comprised of all the forms of society. I will restrict myself here with the memory which one wants to put in the brain such as it is independent of the external elements.

The brain has a highly parallel architecture. I.e., there is much activity which occurs in different " subsystems " at the same time. Sometimes these " subsystems " are separate spatially, but it is thought that often it between-overlap. These activities correspond sometimes to the conscious thought, but also to the " sub-conscious ", sometimes to the emotions, and sometimes to the activities of mechanical adjustment the such pressure of blood, the level of sugar, etc.


(1) a way of adding something with the memory easily is to associate already coded elements. Physically, this probably corresponds to the reinforcement of the synapses (of connected between the neurons), or to the creation of new synapses. (There is no creation of new neurons after the age of approximation five years.)

When one tries to include/understand something, it is a process of comparison in terms of already coded ideas, including, for example, in many intellectual traditions, in theory, codings of logic known as " first order classical logic ".

If one can make this process of integration with what one knows already, memorizing can be very strong.

(2) Another way is to add a " knowledge cérebrale " whose several elements are not coded. Mentally, you intuitively try to memorize, without the assistance of already known things. Physically, you make random activity between neurons, while waiting for that something " sticks ". Once that that " sticks ", you reinforce these synapses.

This method is much slower and less effective than the first, but when nothing is known, one has to begin somewhere. It is like that (partly) that the foetus and the baby learn.

Thus, it is easier to memorize by using already known elements, but one is not completely obliged.

In time

If one learn something during five minutes, one minute later, one will remember some easily. Ten minutes later, there will not be too much difficulty to remember some. But one year after (if nothing were done since the initial training), one will be astonished to remember of it, except if it were an exceptional event.

How does the memory in time disappear, if there are as few elements of association as possible?

(1) the experiments in laboratory were made while requiring of the subjects to study and learn from the lists of syllables entirely without meaning, therefore by avoiding easy association, except for association with the letters of the alphabet and their sounds. The subjects were tested immediately to check that they had learned well, and also after a certain time, of a few hours or several days.

It was found that their successes on the tests decrease exponentially (approximately) with the time chosen for the test of checking. At 24 hours after the initial training, one retains approximation 90% in means. What can be surprising, but if it is tested, by giving time for the ideas to remake surface, by writing what one " guesses " that one learned the previous day on a sheet of paper, and then checks, one will see...

Algebraically, one can write F = f0 exp(- T / 10) where F is the fraction selected and T is time in days. Thus, after ten days, one will retain of them only 30% according to this formula.

(2) Then, a new experiment was made. It was like the first, but immediately after the test to measure the fraction selected, one required about entirely relearning the list of syllables meaningless, as at the beginning. After a certain time (different for different subjects), one made a second test. The result was that if one makes a relearning after 24 hours, the rate of reduction is much slower. The formula above becomes something as F = f0 exp(- T / 70) where T is always in days.

It is necessary a week then before one goes down towards 90% of retention. So then, a new relearning is made, the rate at which memory is " lost " slows down still, and so on.

If one learns something at the day T, but one test and relearns it at the days T + 1, T + 7 and T + 28 one will arrive at F = f0 exp(- T / 3650) (approximately). I.e., that one will retain then approximation 90% one year later, while having made only one initial lesson and three revisions, and without the revisions being too difficult. Obviously, you could carry out a fourth revision after one year so that it is retained for a good fraction of your life.

These values probably correspond to the various physical events in the brain which code the memory, and it would be necessary to see the recent literature to see whether they are really the optimal values. But those go, at least in my own experiment, and according to what I recall of the basic texts of psychology.

Physical explanation: So that coding valid for a scale of time is valid, it is necessary that coding at the shorter time of scale was made. But it is enough that coding short term remains in the sub-conscious until the moment when one ratifies it by making the transfer towards a coding with the long term (relative). To bring back the idea in the sub-conscious before it is necessary to transfer towards a coding longer term is perhaps useless, and, in particular for the training of a list of ordinary words, is boring!

A professor who nicely corrects you with an infinite patience

When you study a list of word, or some element of grammar, to see which you have retained, to relearn that that you have forget, to them relearn, and of you retest some time until it that you of remember quickly of each word foreign in answer with a word of your own language, it be necessary a professor very patient.

One can use a paper sheet as professor, and retest several times with empty sheets on words which have not been learnt well. But one is likely to remember these words only according to the same sequence. If one test with a random order, the training is stronger.

Thus, here is a compilable software in f77 with pgplot (rather on Sun or Unix machine, but perhaps also on a PC).

To put together

A way of putting all that together is the following.

(1) the student uses the software to learn a list of words, which he or she prepares himself or herself in a file, that he or she names including a date such as 981008lecon.txt . He or she associates a more or less large " value " to mark the importance of the word in the file. (Usually, I advise values of 3 or 2.)

(2) By using this software, the first time that one meets the word one uses the principle of association. One creates a small story for the word, using anything which one finds in one's imagination, images, words of other languages, the emotions, etc.

For example, for this memory of
QN repeat it slowly please
REP a'p dzara' a'hista' bata'i'ye'
one could associate:
a'p: a harp (musical instrument), dzara': a Russian tsar, a'hista': _our_ own history, bata'i'ye': a battle with an Englishman who says " yeh " (yes).

To put it together, one can thus to imagine tsar Russian with toothing-stone in front of him (to remember that " a' p " is at the beginning), on the battle field with English which says " yeh ", and being noticed that Napoleon is there and thus actually it acts of " our " (" our ") own history all in the medium of the phrase/bataille. And it is necessary to add that the tsar (or Napoleon) request with the other to repeat slowly because it did not include/understand completely well.

It is stupid, but now that you know this small story, and provided that you have the good pronunciation, you will have evil to forget one of the most significant sentences for the beginner in hindi.

(3) Then, it is useless to relearn this list of words several days of continuation, or each two or three days. It is necessary to space relearnings in an optimal way in time, in particular that 24 hours.

For example, let us suppose that one has 16 sessions of two or three hours available during five weeks, on average three and one third " lessons " per week. With only one week " intensive " and in all 16 sessions, so that 200 words are retained during one year with 90%, a good formula, where T is the date in days, is then:

first week
T = 1 : to learn a list of 50 words, making turn the software, using association at the beginning, and continuing until (ideally) a score of 50, or if not up to a level which you consider satisfactory (1 session)

T = 2 : to revise (with the software) the list of the day précedent; and to learn (with the software) a new list of 50 words, in the same way (2 sessions: 1 revision + 1 training)

T = 3 : to revise only the second list; to learn a third list of 50 words (2 sessions: 1 revision + 1 training)

T = 4 : to revise only the third list; to learn a fourth list of 50 words (2 sessions: 1 revision + 1 training)

T = 5 : to revise only the fourth list (1 session)

the weekend: rest you without concern!

second week
T = 9 : to revise the first list (1 session)

T = 10 : to revise the second list (1 session)

T = 11 : to revise the third list (1 session)

T = 12 : to revise the fourth list (1 session)

third and fourth weeks

nothing to make

fifth week

T = 29 : to revise the first list (1 session)

T = 30 : to revise the second list (1 session)

T = 31 : to revise the third list (1 session)

T = 32 : to revise the fourth list (1 session)

And here are, 200 words learned well for the year. By naming the files with dates, it will be easy to revise with approximation the good dates. Useless to be maniac on the deadlines, an accuracy of 10% to 30% is probably sufficient. (A the limit, if the first revision is carried out two days after the initial training, it will be harder than if you had done it after only one day, but probably will give a reasonable result.)

Good training!
Home Page At IAP