Forgetting index

You can choose the speed of learning

Forgetting index is the proportion of elements that are not remembered at repetitions. It is usually expressed as a percentage. For example, if you choose the forgetting index of 10%, SuperMemo will try to make sure that you remember 90% of elements at repetition time. The higher the forgetting index the faster you learn but the less you remember! You can choose the default forgetting index with Tools : Options : Learning : Forgetting index.

Extreme values of the forgetting index can be frustrating

SuperMemo makes it possible to choose a forgetting index in the range from 3% to 20%. Very often users of SuperMemo ask why there is a limitation on this range. Many of you would be tempted to set the forgetting index to 1% or even 0%. This would only result in unnecessary waste of time. Forgetting index of 0.00% would mean that the intervals between repetitions should equal 0! If you decided to choose the forgetting index of 1%, the repetitions would be so frequent that you would probably be discouraged to ever come back to SuperMemo without ever discovering its power. In addition, the spacing effect would make your memories very weak. This would shorten intervals further and add more work and frustration.

If you do not have much experience with SuperMemo, you should set the forgetting index to 10%. This value is important for psychological reasons. If the forgetting index is too high, your repetitions will be stressful due to constant problems with recall. Your material will seem difficult to remember. This can be quite frustrating. On the other hand, if the forgetting index is too low, your repetitions will be annoyingly frequent. You will experience a sense of wasting your time on needless repetition.

If you feel you remember too little, reduce the forgetting index. If you feel you repeat too often, increase the forgetting index. In most cases, the value of 8%-13% will work best

Changing the forgetting index

  1. You can change the default forgetting index in your collection with Tools : Options : Learning : Forgetting index
  2. You can change the default forgetting index of a given category by setting its Forgetting index field (e.g. in Search : Categories). Use zero for the category forgetting index to make it reflect changes in the default forgetting index set in Options. Newly created categories use the default forgetting index set in Options
  3. You can also quickly change the forgetting index of individual elements with Ctrl+Shift+Up and Ctrl+Shift+Down. Note that Up increases priority and thus reduces the forgetting index. You can also change the forgetting index of elements with Element parameters dialog box (Ctrl+Shift+P) as well as with Priority dialog box (Alt+P)

Forgetting indexes used in SuperMemo

It is easy to confuse some terminology related to the forgetting index. Here is a short glossary:

  • requested forgetting index - forgetting index that you want to achieve in learning. You ask SuperMemo to plan repetitions in such a way that your measured forgetting index reaches a given value. The requested forgetting index may be set for individual items (see: individual forgetting index) or for many items at once. To change the forgetting index for many elements, you can use the default forgetting index (see below) or branch operations in the contents window (Contents : Process branch : Forgetting index)
  • measured forgetting index - forgetting index as it is actually measured at repetitions. It is displayed in the statistics window as Measured FI. Measured forgetting index is usually higher than the requested forgetting index, esp. if the forgetting index is less than 5%. This comes from the fact that SuperMemo imposes some limitations on the length of intervals that are to prevent excessively frequent repetitions. Secondly, if you delay repetitions, skip repetitions on a given day or use Postpone, you step away from the optimum learning process and inevitably increase the measured forgetting index. Finally, in an overloaded learning process, the measured forgetting index of low-priority items is higher by design (see: Tools : Statistics : Analysis : Graphs : Forgetting Index vs. Priority. You can restart the measurements of the forgetting index with Tools : Statistics : Reset parameters : Forgetting index record
  • default forgetting index - requested forgetting index set for all items with the default individual forgetting index by means of Tools : Options : Learning : Forgetting index
  • category forgetting index - requested forgetting index used in a given category. If it is set to zero, it is assumed to be the same as the default forgetting index
  • individual forgetting index - requested forgetting index set for individual items. It can be seen in the Element data window. It can be set with Element parameters dialog box (Ctrl+Shift+P) or Priority dialog box (Alt+P). You can also set forgetting indexes for the entire branch in the contents window with Process branch : Forgetting index
  • (optional) expected forgetting index - expected probability of failing to recall a given element if the repetition takes place at a given moment. Used internally by SupeMemo Algorithm
  • (optional) estimated forgetting index - value of the forgetting index derived from the scored grade on the basis of the correlation graph showing the relationship between the expected forgetting index and the average grade obtained at repetitions (G-FI graph). Used internally by SuperMemo Algorithm

Retention is higher than 100% minus the forgetting index!

If you set your forgetting index to 10%, you will remember 90% of the material at repetitions. This does not imply that your knowledge retention will be 90% only. Your average retention will be nearly 95%! This comes from the fact that 90% refers to the retention at repetitions, while the initial retention right after the repetition is theoretically 100%. During the inter-repetition interval, retention is decreasing from 100% to 90%. On average you roughly remember 95% of the material. The exact formula linking the forgetting index with the retention is as follows (source):

retention = -(forgetting index)/ln(1-(forgetting index))

Forgetting index


3% 98.49%
5% 97.47%
10% 94.91%
15% 92.29%
20% 89.62%

The reason that the retention is not equal to 1-0.5*(forgetting index) is that forgetting is approximately exponential in nature. Immediately after the repetition, forgetting proceeds at the highest rate.

Do you know that?

  • you can sort elements by the forgetting index by clicking the FI column in the browser (browsers are available from the View menu)
  • you can search for elements with a given requested forgetting index by means of View : Filter
  • you can see how SuperMemo plots the forgetting curve graphs and derives the length of the interval from the forgetting index. Try Tools : Statistics : Analysis : Curves
  • you can see correlations between the forgetting index and other learning parameters that have been measured in a group of 32 students
  • the concept of the forgetting index was introduced only in SuperMemo 6 for DOS in 1991. Earlier versions of SuperMemo worked as if the requested forgetting index was 10%. This value could not be changed 
  • individual forgetting indexes are kept in the file elinfo.dat, the default forgetting index is kept in <collection name>.kno file, and category forgetting indices are kept in the category registry (see: files used by SuperMemo)


(Tomasz Szynalski, Poland, Oct 18, 1998)
What value of the forgetting index ensures the optimum ratio of (retention)/(time spent per day)?

Paradoxically, the highest speed of learning can be accomplished ... without SuperMemo! In our daily life we pick up lots of facts that stay in our memory for long with few repetitions in lifetime! The problem is that these are usually not exactly the facts or rules that are critical to our goals. In other words, not the speed of acquiring new items counts but the speed of acquiring new items bearing a given content.

It is difficult to determine exactly what forgetting index brings the highest acquisition rate. Simulation experiments have consistently pointed to the value of 25-30%. You can even plot speed-vs.-forgetting graph using your own actual learning material in SuperMemo 98 using Tools : Statistics : Simulation. You will probably also arrive to similar results

As you perhaps know, SuperMemo disallows of the forgetting index above 20%. This comes from the fact that you should aim at achieving high speed of learning combined with high retention of the learned material. Setting the forgetting index above 20% would be like giving up SuperMemo altogether and coming back to remembering only that what is easy to remember. In highly interlinked material where new knowledge depends on the previously acquired knowledge, high forgetting rate can even be more harmful

Nevertheless, if you want to maximize the speed of learning with little control over what actually stays in your memory, set the forgetting index to 20%

(Tomasz Szynalski, Poland, Oct 18, 1998)
What retention can I obtain with the forgetting index set to 9%? What if I then change it to 12%?

For the forgetting index of 9%, the retention should be 95.4% (see the formula above). For 12%, the same figure will be 93.9%. Note that if your material is very difficult, your measured forgetting index may be higher than the requested forgetting index. This comes from the fact that SuperMemo imposes some boundary conditions on the increase of intervals. Elements that have been forgotten more than five times should be reformulated with a view to reducing their difficulty or increasing their mnemonic component.

If you initially set the forgetting index to 9% and later on increase it to 12%, you will probably start with retention of 94-95% which will later gradually decrease to 92-93% (after the change)

(Peter Cool, The Netherlands, Nov 6, 1998)
I started with SuperMemo 70 days ago (your French plus some words added by me; total 1000 words). In the first weeks I made a lot of mistakes so my measured forgetting index was 20%. Although I make very few mistakes now during repetitions the forgetting index decreases very slowly. Is this normal?
The measured forgetting index includes the record of all repetitions made since you started learning. That is why it changes at an ever decreasing rate. If your performance is good and you would like to more accurately check your current forgetting index, you might reset the forgetting index measurements with
Tools : Statistics : Reset parameters : Forgetting index record. It will not affect the learning process per se

(Manfred Kremer, Germany, Sep 7, 1998)
I noticed that frequently I get Optimum Interval in the element data window shorter than the last interval displayed as Interval. Is it a bug in SuperMemo?
No. If your forgetting index is very low, e.g. 3%, SuperMemo will often conclude that you will stand 97% chance of remembering a given element only if your next interval is shorter than the presently used one. In such cases, it will not accept the new value and the new interval will be at least 5% longer than the previous interval. Please note that the forgetting index equal 3% should only be used for selected high-priority items. Keeping the forgetting index at this level throughout the collection will make repetitions annoying frequent and ineffective

(David Mckenzie, New Zealand, Apr 8, 1998)
Why does not the first repetition after forgetting occur the next day after the unsuccessful repetition (this is advised by Tony Buzan and others)?

In SuperMemo, the length of the first interval is computed from the forgetting curve plotted in the course of repetitions. This is to make sure that a defined proportion of items is remembered (usually 80-97%). This proportion is programmed by means of the forgetting index. Depending on the forgetting index, the length of the first interval may range from 1 to 20 days, and is not set arbitrarily. It is computed from the record of repetitions and determined by the desired forgetting index (requested forgetting index is the proportion of items that are not remembered at repetitions). While Buzans recommendation is valid in many cases, you should not forget that SuperMemo computes intervals with a high degree of accuracy. Accurate intervals cannot easily be predicted without a computer 

(David Mckenzie, New Zealand, Apr 8, 1998)
Is there any point in keeping collections separate?

No. Once you master categories, templates, and subset operations, there is no point. You gain global search, global registries, global repetitions, global optimization, etc. This would not be advisable in SuperMemo 7 as item difficulty measure (E-factor) was dependent on the average difficulty of items in the collection. Presently, the item difficulty measure (A-Factor, or absolute difficulty factor) is absolute and does not depend on the context in which an item is placed. Only the length of the first interval will significantly be affected by the average difficulty of items in the collection. However, this shall not bear dramatically on the speed of learning. Especially that variable forgetting index for individual items makes it possible to set different first intervals for whole contents categories or branches of the knowledge tree

(Noel Clary, USA, Aug 17, 1998)
I have created my own database on plumbing and air conditioning. My forgetting index is quite high. Are there any tools in SuperMemo which could help me remedy this situation?

You might want to use View : Other : Leeches and locate the elements that cause most problems in learning. You must then go into your own mind to answer the question why these elements are hard to recall. Usually these are too complex, too long, too boring or too similar to other elements in the same collection. You can also send 3-4 most difficult items to us for review to receive some suggestions. Read more: Leeches, and 20 Rules of Formulating Knowledge

(Matt Cassidy, New Zealand, Sep 11, 1997)
Is it possible that with forgetting index equal to 3% I get the first interval equal to 6 days?
Yes. Especially if the material you work with is relatively easy. You should also remember about the random dispersion of intervals. In isolated cases, dispersion might produce intervals substantially longer (or shorter) than the optimum interval. For more read about SuperMemo Algorithm

Tony Buzan claims that 75% of information is lost if not reviewed in 24 hours. Does it not defeat the validity of SuperMemo in which the first interval is often longer than a week?

No. Buzan's claim may refer to textbook knowledge or complex knowledge structures (e.g. large mind maps). However, it does not seem accurate in reference to simple well-structured material in the light of results obtained with SuperMemo. In SuperMemo, if the student chooses the retention of 95%, the typical value of the first interval falls in the range 2-5 days depending on the student and the difficulty of the learned material. For retention 25%, the same interval might be as long as one month, though it cannot be verified experimentally with SuperMemo which limits the range of the forgetting index from 3-20%, which implies the overall retention in the range of 89-99%. For more see: Theoretical background of SuperMemo

I used SuperMemo 2 shareware, and was accustomed to repeating forgotten items on the next day. It is very irritating that in new SuperMemo, I do not have this possibility

SuperMemo will schedule forgotten items in intervals that are determined by the forgetting index. The greatest increase in the speed of learning in newer versions of SuperMemo as compared with SuperMemo 2 resulted from substantially increasing the length of the first interval. The student may be left with the feeling that he is likely to forget the item again if it is not repeated on the next day. Statistically, however, he will forget no more than the proportion defined by the forgetting index. By reducing the forgetting index to less than 5%, the length of the first interval is likely to drop to 1-2 days in most cases. Moreover, if you are particular about repeating a given item on the next day, you can choose Ctrl+M to memorize or rememorize an item with a selected first interval

I have an exam for a driver's license in 2 weeks. How can I best memorize the Traffic Regulations collection for SuperMemo? How can I increase the frequency of repetitions?

Although SuperMemo is not a cramming tool, and it would be much safer to start 2-3 months before the exam, the following shall work pretty well: (1) Set Tools : Options : Learning : Forgetting index at 3%. (2) When memorizing difficult items, choose Ctrl+M and provide the first interval value equal to one day. (3) Memorize the collection in equal portions in the period spanning from today to 2-4 days before the exam. Use Tools : Random review intensely over the last 2-4 days

Warning! Tools : Options : Learning : Forgetting index
will only affect the forgetting index in newly created categories and in element that use the default forgetting index (e.g. set in
Element Parameters as 0). If your collection has the forgetting index of individual elements fixed, use Forgetting index : Set forgetting index in the browser or in the contents window

You can easily change the forgetting index from the keyboard (#5737)
(J.P., Czech Republic, Sat, Jun 16, 2001 16:59)
I find that changing the forgetting index of an element is pretty time-consuming without any keyboard shortcut
You can use Ctrl+Shift+Up and Ctrl+Shift+Down to quickly modify the forgetting index of an element. You can use also use Ctrl+Shift+P to open the Element Parameters dialog box and Alt+F to get to the forgetting index field

Retention statistic is based on the assumption of regular repetitions and well-structured learning material (#6918)
(dansujp, Sun, Sep 16, 2001 3:07 PM)
When I returned from vacation, I expected the retention to be something like 80% because I have not done any repetitions for two weeks. But it was exactly the same as before I left
The Retention statistic is derived directly from the measured forgetting index on the assumption of a negatively exponential forgetting curve. This curve is only representative of well-structured learning material. In addition, the forgetting index measurements are averaged over all recorded cases. A break in repetitions will invalidate the statistic. Resuming repetitions is not a guarantee of accuracy as the large number of earlier repetitions will result in overestimating the retention on a small-sample measurement. The only valid estimation of retention after a break in learning is the one that follows resetting the past forgetting index record (Tools : Statistics : Reset parameters : Forgetting index record). This will result in gathering new data that will approach true retention for the sample tested with accuracy proportional to the number of repetitions made

Use the default forgetting index to globally change the forgetting rate (#6321)
(Jiri, Thursday, January 10, 2002 4:28 PM)
What is the purpose of the default forgetting index available in Options? As far as I know, all newly created elements have the forgetting index set according to their categories and there I cannot choose the default forgetting index
Tools : Options : Learning : Forgetting Index (default forgetting index) will be used in items that have their requested forgetting index set to Default (internally represented as zero). All items with the default setting will use the default forgetting index currently selected in Options. You can set the default forgetting index with contents and browser operations (Forgetting index : Default forgetting index)

Topics also use a forgetting index (#20768)
(Terrence T., Feb 14, 2004)
What is the meaning of forgetting index in topics? Should not this field be disabled?
The forgetting index field in topics has no impact on the way a topic is handled in the learning process. However, all cloze deletions generated from that topic will inherit its forgetting index. Also, all extracts will propagate the forgetting index to all its descendants. This way, you can reduce the forgetting index on very important articles to ensure better retention of the derived material. Lower forgetting index will also be used to protect your material from postpones when there is a heavy learning overload

A drop in the measured forgetting index is highly undesirable (#24805)
(mahabharatta, Wednesday, March 31, 2004 6:28 PM)
My measured forgetting index went up to 90%. I introduced a HUGE load of material over a short time period around 4 months ago. There was a 4 day burden of 900. In the last 3 months I've been introducing new materials sporadically, never more than 100 elements in one day, 300 in a week. During this period the measured forgetting index has been dropping. Today it is 29.9%. Is this normal?
Forgetting index of 90% tells you, you have introduced too much material at once. The time needed to recover will depend on the degree of overload and the complexity of the material. With forgetting index above 30%, your learning efficiency will not be much better from that achieved without SuperMemo. However, in the long run, if you make order in the learning process, this hectic period can still bring fruit.