Frequently asked questions about SuperMemo 7 for Windows

Q: Can a computer illiterate person use SuperMemo?

A: Yes. The entire knowledge needed to start is encompassed in the page long Minimum User's Guide which is present at the beginning of the manual, as well as on the installation disk (MINGUID7.TXT). Five simple operations needed to use SuperMemo can be learned in less than 10 minutes! No specialized knowledge is required.

Q: If I don't have a computer, can I use SuperMemo?

A: Yes. A limited, paper-and-pencil variant of SuperMemo exists, and is shortly described in the file SMPAP.TXT. A more detailed description you will find in the file SMPAP.DOC (Microsoft Word format).

Q: Can I use SuperMemo in a computer lab?

A: Yes, on condition that the access to computers in the laboratory is regular. Once you start working with SuperMemo, you must systematically continue the process of repetitions. Otherwise, the acquired knowledge will gradually be forgotten (e.g. if you stop after having used SuperMemo for 5 years, you will forget about 60% of the learned material in the first year!). In SuperMemo 7.4, you may find option Miscellaneous : Users particularly useful if you do not want one of your colleagues to confuse his database with yours.

Q: Do you have SuperMemo for platforms other than Windows?

A: Currently, SuperMemo is available for DOS, Windows, Mac and Amiga. Versions for Psion and OS/2 are in plans. Versions for Atari ST and Atari XL have been discontinued due to lack of interest among customers and are no longer available (please do not inquire).
To learn more about the Amiga version contact Twin Spark Soft, Poland, tel: (48) 12 444368, fax: (48) 12 477299.
To learn more about the Mac version contact AS Projekt, Poland, tel/fax: (48) 52 392759, e-mail: SOS@TELBANK.PL 

Q: What are the minimum hardware requirements of SuperMemo?

A: The minimum requirements are as follows:

SM2: IBM XT, DOS 3, 200 KB memory, no hard disk, any display

SM5: IBM XT, DOS 3, 300 KB memory, no hard disk, any display

SM6: IBM XT, DOS 5, 512 KB memory, no hard disk, EGA or better

SM7: IBM AT, Windows 3.1, 2 MB memory, 2-5 MB hard disk space, VGA or better

SM7 (for AudioVisual databases): 386, Windows 3.1, 2 MB memory, 10-50 MB hard disk space (depending on the size of databases)

SM for Amiga: Amiga, 1 MB memory, 2 MB hard disk space

SM for Mac: Apple MacIntosh, hard disk

Q: What are the basic differences between SuperMemo 2, 5, 6 and 7?

A: The full record of upgrades introduced to SuperMemo is listed in the file DEVELOP.TXT. In short, all versions of SuperMemo allow the learner to substantially increase the speed of learning. Figuratively, if traditional methods of learning are viewed as a pedestrian, SuperMemo 2 Public Domain may be compared to a bicycle, SuperMemo 5 Shareware to a Trabant, and SuperMemo 6 to a Mercedes. Learning with SuperMemo 7 is as fast as with SuperMemo 6, but introduction of audiovisual features adds some extra quality to learning. Each version includes some improvements to the user's interface, database management, fault tolerance, statistical analysis options, menu tree, keyboard shortcuts, parameter set, help, etc. Moreover, in SuperMemo versions above 5, the speed of learning may be regulated by the learner (in a trade-off with knowledge retention).

Q: What is the size of SuperMemo 2, 5, 6 and 7 in KB and databases?

A: The sizes are as follows: SM2 77 KB, SM5 108 KB, SM6 160 KB, and SM7 515 KB. The size of text databases is usually less than (5+0.1*Total) KB, where Total is the number of items in the database. Additionally, graphic files may take 20 KB per picture, and sound files take 10-20 KB per word.

Q: What are the basic differences between SuperMemo 7.0 thru 7.5?

A: SuperMemo 7.1 allows you to associate graphics and sound with SuperMemo items. SuperMemo 7.2 includes about 20 minor new options such as selectable fonts, copying and moving databases, word-wrapping, etc. SuperMemo 7.3 introduces over 20 major improvements such as full-screen database browsers, search browser (Find All), possibility of changing the order of items in the database, and the order of items in the intact queue, possibility of browsing intractable items, etc. SuperMemo 7.4 includes: integrated toolkit, child browsers, multi-user management, Spell-Pad for spelling drills, different fonts in questions and in answers, Mercy into future, CD-ROM audiovisual database access, etc. Finally, SuperMemo 7.5 adds the ability to use programmed DLL database modules that extend the applications of SuperMemo to all fields in which a training procedure can be programmed with the computer (e.g. touch-typing, intelligence tests, problem solving, playing musical instruments, etc.).

Q: Can I write my own SuperMemo database concerning topics of my interest?

A: Yes.

Q: If I prepare a database for SuperMemo, can I sell it to SuperMemo World?

A: Yes. Particularly if it is a good database, on a popular topic, and preferably written in English. Otherwise, your database may be included in the Database Bank, and you will be paid royalties from each copy sold (standardly 10% of the end-user price).

Q: What databases for SuperMemo 5 Shareware can I get from SuperMemo World?

A: Because of low profitability, SuperMemo World does not collect databases for SuperMemo 5. Contact shareware houses for information about SuperMemo 2 Public Domain and SuperMemo 5 Shareware databases.

Q: What databases are there in the SuperMemo World's Database Bank?

A: See the DBANK.TXT file for the most recent list of databases available from the Database Bank (use Miscellaneous : View text files).

Q: What is SuperMemo World e-mail?

A: SuperMemoMail

Q: Why does SuperMemo ask only those items which are difficult?

A: Just because they are difficult. You do not want to learn things you already know, do you? The purpose of SuperMemo optimization is to compute the best intervals between repetitions. Difficult items must be repeated more often; hence the impression that only the hardest items are considered in the process.

Q: If I have an item in my database, and I want to memorize it after memorizing some of the items that follow it, what should I do?

A: You can change the order of intact items by using Browse : Intact (see the Move menu). Alternatively you can press M (for Memorize) and R (for Reset) while inspecting the item. That will move the item in question to the end of the intact queue.

Q: How can I change the order of items in a database?

A: You can change the order of items by using Browse : All (see the Move menu). After moving a number of items to new positions, quit the browser (e.g. by pressing Esc). Respond with Yes to “Do you want to rebuild the database?”. Remember that the rebuilding process is quite time consuming and that it is quite error prone. Therefore, it is recommended that you run Tools : Recover before rebuilding a database, and that you make a back-up copy of the database files.

Q: How can I quickly transfer a single item from one database to another without losing its learning parameters?

A: Use Item : Transfer on the menu of editing windows, and choose the destination database to which the current item should be transferred.

Q: I found some errors and misspellings in your databases. How can I correct them?

A: When you see the wrong item on the screen, press E (for Edit) or click the mistaken field (question or answer). Edit the item using arrow keys, Backspace, etc.

Q: I left for vacation, and my learning process is in mess. I want to start all over again. What should I do?

A: Follow the steps:

1. Press Ctrl-W or choose Miscellaneous : Wipe to invoke the Wipe dialog box.

2. Set the maximum interval value to a period comparable to twice the length of your vacation.

3. Press Alt-R or choose the Reset-Wipe button.

4. Press Ctrl-M or choose Miscellaneous : Mercy to invoke the Mercy dialog box.

5. Set the Rescheduling period in which you want to catch up with outstanding repetitions (usually from 50% to 200% of the vacation period; depending on availability of time).

Though not without losses to what you have already learnt, your learning process shall become clear and regular. Remember, however, that despite the relief which Mercy and Wipe provide for irregular learners, the two options have also a potential to kill your progress. Use them only as a last resort!

In case you want to return to repetitions after a year or more, it may appear that it is more rational to start the learning process all over again. To start again, use Tools : Reset : Database to convert your database to the intact form.

Q: The User's Guide often mentions that the user should periodically back up his or her databases. What does it mean and how to do it?

A: To back up a file means to copy it to an archive diskette in case the original file was lost or damaged. The easiest way to back up your database is to use File : Copy. You can also back up your files by means of DOS commands such as COPY or XCOPY, as well as by means of tools such as Norton Commander, PC Tools or File Manager in Windows. Consider using archiving utilities like PKZIP, ARJ, LHARC, etc. This way you can save a great deal of disk space, esp. if you use audiovisual databases.

Q: Can SuperMemo work with databases compressed with DBLSPACE in DOS 6?

A: Yes

Q: How can I increase the frequency of repetitions before an exam?

A: SuperMemo is not supposed to be used as a cramming tool. It is rather designed for those who want to keep knowledge in their memory for a longer period of time (from one year to lifetime). To maximize the frequency of repetitions set Lapses to 3 (option Miscellaneous : Options) and run Approximate. The day before the exam you might try Random test or Demo, but this will certainly introduce a lot of noise into your long-term learning process.

Q: Are databases of SuperMemo 6 compatible with SuperMemo 7?

A: Databases of SuperMemo 6 are compatible with SuperMemo 7; however, ASCII characters below 32 and above 128 may be defined differently in DOS and Windows environments. Moreover, spaces (char #32) in standard non-fixed Windows fonts have lower width, which may corrupt some mathematical or chemical formulas. You can easily remedy the above problems by means of Tools : Replace strings or REPSTR.EXE included in the SuperMemo Toolkit. The easiest way to remedy DOS and Windows incompatibilities is to use the font Terminal.

Q: Why is it not possible to use databases of SuperMemo 7 with SuperMemo 6?

A: SuperMemo 7 was not designed as backward compatible with SuperMemo 6. If to neglect the fact that SuperMemo 6 does not have audiovisual capability, the main problem remains that SuperMemo 7 allows of longer text lines in its items. A partial solution is to convert the SM7 database to the SM6 format with Tools : Export to SuperMemo 6

Q: How can I run SuperMemo in a network?

A: Install the program on the server, and let particular learners keep their databases in their home directories. Each learner will have his or her own learning parameters (stored in INF files) and database parameters (stored in INI files). Each learner will have his default working directory and, optionally, password (defined by means of Miscellaneous : Users). However, all the users will share some parameters stored in the SM7.INI file (e.g. text file directory, help directory, etc.).

Q: I want to learn Russian. Can SuperMemo provide Cyrillic?

A: Yes. To get Cyrillic in your items do the following: (1) install Cyrillic in Windows (choose a commercial or shareware font, or define the font using any font editor), and (2) choose Cyrillic font for your RUSSIAN database (option Miscellaneous : Font). To keep Latin alphabet in questions and Cyrillic in answers, click the answer field and press Ctrl+F to choose Cyrillic for the answer field only.

Q: How can I learn mathematical formulas in SuperMemo 7?

A: You can choose one of the following:

* use TEX syntax if you know it

* use the options of the Image menu and represent you formulas using graphics

* define your own font with mathematical symbols or use a commercially available font that would satisfy your needs

Q: How can I improve the default sampling and resolution of recordings associated with SuperMemo items?

A: Create an empty wave file with the desired parameters and copy it in place of 0.WAV in the BIN subdirectory. Alternatively, use Audio : Import if you want to change only the parameters of a sound file associated with a single item.

Q: How can I change the default size and other parameters of bitmaps associated with SuperMemo items?

A: Create an empty bitmap file with the desired parameters and copy it in place of 0.BMP in the BIN subdirectory. Alternatively, use Image : Import if you want to change only the parameters of a bitmap associated with a single item.

Q: As I travel, I use SuperMemo 7 on a number of different computers. Do I always have to repeat the entire installation procedure?

A: No. Prepare a floppy containing the following files SM7., SM7.CI_, and BWCC.DLL (if you create new audiovisual items then add yet 0.BMP and 0.WAV). Copy all these files to a single directory and that shall suffice to run SuperMemo on an computer provided with Windows.

Q: For some time now, the Grade parameter hardly changes in the main window of SuperMemo. Is something wrong going on?

A: Parameters such as Grade, Lapses, Retention, and Mean time are, or are derived from, averages or trailing averages. Therefore, the longer you have been learning, the less they change. After a year or so, you will hardly see them change.

Q: Why is time updated only at repetitions?

A: The timer used in SuperMemo is used to estimate some parameters related to the learning process (Mean Time, Workload, etc.). Obviously, the time which the learner spends on editing the database, or even eating his/her lunch, is of little relevance to the method per se. The time spent for the entire learning session, if at all needed, can easily be computed by using an external timer, e.g. a separate Windows application.

Q: Why did you not implement Undo in SuperMemo?

A: A single-step undo in item controls is available in the Windows version by pressing Ctrl-Z or Alt-Backspace. Future versions will include multistep undo.

Q: What should I use Copy in the Edit window for? What for do I need the same item twice in the same database?

A: You can copy an item, if you want to add to the database another item which is only slightly different. This way you can spare some time by reediting the old item instead of typing the new one all over again.

Q: Can I define KBD files in SuperMemo 7 as it was possible in SuperMemo 6?

A: Yes. However, you can only make assignments to Alt-key combinations. To redefine the keyboard used when editing items do the following: (1) create a text file [database].KBD, (2) in the following lines of the KBD file put the key-character pairs separated by a space (e.g. ‘a ⤣146;,’u �;, etc.), (3) place the KBD file in your database directory. In the item editor, you will be able to use Alt-a to to get ⬼/p>

Q: Why do you use grades 0..5 with 5 standing for the best grade? This goes against standards used, for example, in the US?

A: As of version 7.1, grades are associated with names like 'good', 'fail', 'pass', etc.

Q: Why do you use grades likes 'null' or 'bright'? Numbers from 0 to 5 were much more intuitive (earlier versions)?

A: Depending on the country of origin, the learner may interpret the excellent grade as 5 or as 0. Short names make the grade scale more universal.

Q: When I back up my databases, do I have to back up SM7.EXE as well?

A: No.

Q: How can I do AND-search in SuperMemo 7?

A: Run Find : Search All with the first argument of the AND-search, and follow it with Search : Find all in the browser with the second argument of the AND-search.

Q: Upon undeleting a mistakenly deleted database, the search procedure started doing strange things. I could not remedy the problem with your recovery tools. What should I do?

A: Run Tools : Garbage. Remember that Tools : Recover (or RESCUE.EXE) does not recover from damage to ITM files, which store the text of items. The only way to recover from ITM file damage is to run Garbage with the option Compact, and try to rescue particular items manually.

Q: In my database I got a great deal of items with very-short intervals which seem to increase at an infinitely low pace. The E-distribution indicates that the database is rather well-structured. What do you advice?

A: Consider increasing the forgetting index (see Miscellaneous : Options : Lapses and Miscellaneous : Approximate).

Q: I learn to recognize opera passages, how can I hear the recording included with Audio : Import during repetitions before displaying the answer?

A: Press F10 or click the right mouse button in the Learn Window.

Q: How can a whole family use SuperMemo on one computer?

A: Install one copy of SuperMemo and duplicate these databases that are to be used by the members of the family. You can store databases under different names in the default database directory (e.g. databases MUM, DAD, CATE in the subdirectory C:\SM7\DB), or under the same name in different subdirectories (e.g. database ENGLISH in subdirectories C:\SM7\DB1, C:\SM7\DB2, C:\SM7\DB3, etc.). The first solution is somewhat better, as upon running SuperMemo you will automatically be positioned in the default database directory, and will have to specify the database name only at Open. Otherwise, the full path should be specified or located by browsing the directory tree. Optionally, you can also use Miscellaneous : Users to define user working directories and passwords for all members of the family.

Q: Why is the Edit window in SuperMemo 7 called EDIT 1, not just EDIT?

A: Because it is the first Edit window created. If you open more Edit windows, they will be marked by the number of similar windows created on your desktop (e.g. EDIT 2, EDIT 3, etc.).

Q: I would like to keep the definition of backgammon in my SuperMemo database, but I cannot cut the Webster wording to suit the 250 character limit, and yet make the definition understandable: "a board game played with dice and counters in which each player tries to move his counters along the board and at the same time to block or capture his opponent's counters ..."

A: If you do not know what backgammon is, you should not attempt to store the rules of the game in just one definition! You might choose one or two features that define backgammon unambiguously, and keep the rest of details in separate items, for example: "What do players try to do with the opponent's counters in backgammon?", "What is the size of the board used in backgammon?", "How many counters does a player have in backgammon?", etc. If you know what backgammon is, and you just want to make sure that you do not forget the name, you might just ask "What is the name of the game I played with Robert in summer 1993?". However, if you are just a learner of languages, it may appear sufficient to learn that backgammon is a kind of board game. In Advanced English, you will find the following item: Question: backgammon, Answer: kind of game.

Q : My computer was hit by a virus. SuperMemo reports integrity errors and I cannot work. What should I do?

A: Use Tools : Recover (or RESCUE.EXE) on the database in question and follow it with Tools : Garbage : Compact. Remember to back up your database beforehand.

Q: I neglected my repetitions for too long. I want to start again with the same database. What should I do?

A: Run Tools : Reset : Database on your database and start the learning process from scratch (remember to set the correct date in your computer).

Q: How to convert databases between different standards of national diacritical character codes?

A: Use Tools : Replace strings or REPSTR.EXE. Define your own FLT or CNV file or use FLT/CNV files supplied by SuperMemo World (if available). See the manual or TOOLKIT.TXT to learn how to define filter files (FLT).

Q: I have created two databases and would rather use one. How can I merge databases?

A: Use TRANSFER.EXE. For example to merge GEOG1 with GEOG2 run TRANSFER S=GEOG2 D=GEOG1 ALL

Q: I have created my own question and answer files in ASCII. I would like to convert them to SuperMemo databases. What is the easiest way?

A: Manually or programmatically, convert the files to the TXT format acceptable by Tools : Import text (all question lines starting with 'Q:' and all answer lines starting with 'A:'). Run Tools : Import text.

Q: Can SuperMemo work with other file formats, e.g., ASCII, DBase, Lotus, Word Perfect, Quattro Pro, ChiWriter, etc.?

A: No. However, if the files contain data in the form of questions and answers (e.g. word pairs), they can easily be converted to SuperMemo format. The easiest way is to convert them to the TXT format accepted by Tools : Import text.

Q: I create a database and use it together with my sister. How can we share files so that all correction introduced by me would automatically appear in her databases?

A: Sharing files is not allowed between the databases of two different users. The only thing you can do is to periodically use Tools : Transfer items (or TRANSFER.EXE) to transfer newly introduced items to your sister's database. However, you will not be able to edit both databases at the same time. All editing will have to take place before the act of transfer.

Q: I am creating a database for a group of students. How can we do the work simultaneously (creating the database and learning), without the need to type all the questions to each of the databases separately?

A: Use TRANSFER.EXE each time you add a new portion of items to your database. A batch file of the following format might be a good hint:

Syntax: MOVE <first_item> <last_item>






Example: MOVE 2001 MAX

Q: How to change the physical order of items in a database?

A: Choose Browse : All, use options on the Move menu to change the order of items, quit the browser, and respond with Yes to the question “Rebuild database?”. Make sure that: (1) your database is backed up before the process, (2) you preventively run Tools : Recover, and (3) you are ready to wait several minutes for the rebuilding process to complete (the larger the database, the longer it takes to rebuild the entire database).

Q: How to convert a SuperMemo database to text without losing its learning parameters and then to convert it back to a database?

A: If the name of your database is HISTORY, do the following:

1. run Tools : Export text and uncheck the option Slim

2. specify the name of the destination file (e.g. HISTORY.TXT)

3. do all the editing you need on HISTORY.TXT

4. create a new database

5. run Tools : Import text

6. specify the name of the edited source file (e.g. HISTORY.TXT)


1. run TRANSFER.EXE with the parameter TXT

2. in the file selection window choose HISTORY

3. press Esc when the program requests inputting the name of the range file

4. wait a while as items are transferred to a text file

5. press Esc when the program requests the name of the destination files (the text file will have the name HISTORY.TXT)

6. do all the editing you need on HISTORY.TXT

7. create an empty database HISTORY1

8. run TRANSFER.EXE again

9. instead of the source database choose HISTORY.TXT: in the file selection window (1) press Tab a few times until you get to the template window with the template *.INF, (2) change *.INF to *.TXT, (3) press Enter, and (4) select HISTORY.TXT

10. choose HISTORY1.INF as the destination database.

Note that this way you can also edit item information such as E-factors, intervals, etc.

HISTORY1 will not inherit optimization matrices from HISTORY. Only item information will be transferred. To inherit the OF and RF matrices, create HISTORY1 right after closing HISTORY. To reset the optimization matrices use Tools : Reset.

Q: How can I replace questions with answers in a word-pair database?

A: Press Ctrl+Tab to swap questions with answers in a single item, or run Tools : Swap or SWAP.EXE to swap the whole database.

Q: How can I print the content of a whole database?

A: (1). Use Tools : Export text or TO_TEXT.EXE with the option SLIM to convert a database to the text format, and (2) print the created text file. TO_TEXT.EXE is included in SuperMemo Toolkit.

Q: How can I manually install compressed databases?

A: Use the DOS command EXPAND. Remember to change the file name extensions which end with underscore. For example:



Q: Why are intervals in SuperMemo so long? In SuperMemo 2 they used to be much shorter.

A: They are as long as it takes to produce the desired proportion of items to be forgotten at repetitions. If intervals are irritatingly long, reduce the Lapses parameter in the Options dialog box, and follow it with Approximate.

Q: Why are some items repeated in intervals that increase just by one day between repetitions? I am certain that those intervals could be longer.

A: The intervals are as long as it takes to produce the desired proportion of items to be forgotten at repetitions (forgetting index). If repetitions are irritatingly frequent, increase the forgetting index (Miscellaneous : Options : Lapses), and follow it with Approximate.

Q: Why does not SuperMemo mix the order of items scheduled for repetitions? I have introduced some words in an alphabetical order and they are asked in the same order at repetitions. This is not optimal.

A: True. Please note, however, that items will be mixed as soon as their E-factors and intervals become different. The major advantage of not reshuffling items is the fact that at repetitions, short interval items are asked first. The learner has a general feeling of how long intervals are assigned to items that are currently being asked. This makes it possible to sharpen the accuracy of grading for long-interval items whose inaccurate placement in the repetition schedule has more severe consequences for the learning process.

Q: I think that lower grades, e.g. 3, should produce shorter intervals in comparison to higher grades, e.g. 5. It is not always so in SuperMemo.

A: In SuperMemo, lower grades may produce longer intervals because of the two following reasons:

1. Grade 3 (Pass) may result in the enhancement of the so-called spacing effect, which may be less visible for Grade 5 (Bright). The spacing effect says that longer intervals, and consequently a greater recall effort, produce more stable memory engrams. SuperMemo does not arbitrarily set the function of optimal intervals. It computes intervals which are most likely to result in the forgetting index defined in Miscellaneous : Options : Lapses; hence; the possibility of longer intervals for lower grades.

2. Irregularities in computing optimal intervals may also result from the fact that some entries of the RF and OF matrices can be computed more accurately than others, depending on the number of repetition cases recorded in the process. The real-time smoothing of the OF matrix is used to counteract this problem. Moreover, highly irregular matrices, e.g. resulting from longer breaks in learning, can be smoothed by means of the option Miscellaneous : Approximate.

Q: Why can I not see the correlation between intervals and the grades given in learning?

A: Your impression of no correlation between grades and intervals is quite common among those who begin their work with SuperMemo. It results from the fact that at memorizing new items, first intervals are randomly dispersed around a fixed value. This value comes from the model of an average learner, and can be modified only after the repetitions have shed some light on if the value should be increased, decreased or kept at the same level. Consequently, it will often happen that a lower grade will produce a longer interval and vice versa. At memorizing new items, grades cannot be used to estimate item difficulty because the program has no way of knowing if good grades come from easiness of items or from the fact that a given group of items has just been input to the database (inputting items is a form of repetition). You will start noticing the correlation between grades and intervals in a week or two.

Q: I used SuperMemo 2 Public Domain, and was accustomed to repeating forgotten items on the next day. It is very irritating that in SuperMemo 7 I do not have this possibility.

A: SuperMemo 7 will schedule forgotten items in intervals which are determined by the requested forgetting index. The greatest increase in the speed of learning with SM6/SM7 as compared to SM2 resulted from substantially increasing the length of the first interval! The learner 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 (specified in Miscellaneous : Options : Lapses). However, by reducing the forgetting index to less than 5%, you can make sure that the length of the first interval will drop to 1-2 days (this change may come to effect almost immediately if you choose Miscellaneous : Approximate).

Q: Why is the first interval after which the first repetition takes place not equal in all cases?

A: It is randomly modified to speed up computing its optimum value, as well as to produce smooth distribution of the number of repetitions scheduled over a period of a few days.

Q: I set Lapses to 10 and used Approximate but I do not know if this generated optimal OF and RF matrices?

A: You cannot do much damage with Approximate. Even if optimization slows down a bit, the program should recover fast by modifying the OF matrix to suit the optimum schedule. Moreover, note that Approximate does not affect the RF matrix.

Q: Why can I not see the effect of different grades during memorizing new items, and during the final drill? Are grades used there, and how?

A: At memorizing and at final drill, the interpretation of grades is limited to Pass (4 or more) and Fail (less than 4). None of the optimization parameters are affected. Here are the reasons:

- at memorizing new items, grades cannot be used to determine item difficulty, because the program has no way of knowing if a good grade resulted from easiness of the item or from the fact that the user has just input the given item to the database. Consequently, grades cannot affect optimization, and are only used to detect items that should be scheduled for final drill.

- Final drill is not used for optimization either. This is because of the fact that final drill relies on short-term memory, and is substantially more sensitive to inter-item interference as well as more dependent on the number of other items separating repetitions of the item in question. Final drill is used solely for the purpose of fixing the forgotten information in the user's brain. Grades are used only to eliminate from the final drill those items which score 4 or more.

Q: Why do not grades affect E-factors in memorizing new items?

A: At memorizing new items, grades cannot be used to estimate item difficulty because the program has no way of knowing if good grades come from easiness of items or from the fact that a given group of items has just been input to the database (inputting items is a form of repetition). E-factors can only be verifiably modified after the lapse of the first interval.

Q: Why do E-factors always equal 2.5 at the beginning of repetitions?

A: Historically, the value 2.5 comes from the development of SuperMemo; not from any specific property of memory. It could be 0.25 or 25 as soon as all the remaining optimization parameters were modified accordingly. You can look at E-factors as a very rough approximation of O-factors for the forgetting index equal to 15% and the repetition number greater than 2.

Q: What is the interpretation of E-factors? Why do they range from 1.2 to 3.5, and not just from 1 to 20?

A: Before SuperMemo 5, optimum intervals were computed using the formula:



OI(n) - optimum interval after the n-th repetition

EF - a factor expressing the difficulty of an item, now called the E-factor (E stands for easiness); the higher the E-factor, the easier the item

In SuperMemo 5, for simplicity, EFs became indices to the matrix of optimal factors. Obviously, E-factors might have values ranging from 1 to 20, but this would blur their intuitive interpretation as the approximate value by which subsequent intervals have to be multiplied. Moreover, the formulas for modifying E-factors were inherited almost literally from SuperMemo 2, which naturally suggests keeping their previous intuitive range.

Q: How can Repetition indicate 7 while Cases 31. Which number is true?

A: Repetition says how many times a given item has been repeated. Cases indicates how many repetition cases, of various items, have been used to compute entries of OF and RF matrices. In other words, Repetition (among item parameters) concerns a single item, while Cases (among repetition parameters) concerns many items with the same E-factor and Repetition values.

Q: How can the average interval (Interval in the main window) be longer than the number of days I have spent on learning (Day in the main window)?

A: It is possible. For example, imagine that you have just started learning and you have memorized only one item. If its New Interval is 4 days you will have the following in the main window:

Day 1

Interval 4.0

Q: In your materials I found a contradiction. On one hand you claim that once learned knowledge is constantly maintained in the learner's memory, on the other you say that after ceasing repetitions, I will gradually forget what I have learnt. Which is true?

A: Both facts are true. The term 'maintained' is understood as 'actively kept in memory by means of repetitions', not as 'remains in memory for ever'.

Q: I have been using your program for one week. In my estimation, I can learn no more than 50-100% faster than without the program. Do you claim that my estimation is conservative?

A: No. The power of SuperMemo can only fully be observed after a longer period of time. This comes from the close-to-constant knowledge acquisition rate in SuperMemo, while other forms of learning produce rapid decrease in the speed of learning in result of saturation of the learning schedule with randomly spaced repetitions. The estimation of the 50-fold increase in the speed of learning concerns the lifelong perspective. By using mnemonic techniques, you can easily beat SuperMemo in periods of up to one month. In other words, SuperMemo will not work for those who want to cram a great deal of data before an exam. On the other hand, it is of invaluable help for those who want to retain indispensable facts and figures for months and years.

Q: Your claim that SuperMemo allows me to learn with the speed that closely approaches the maximum speed with which memories can be formed does not sound convincing. What evidence do you have to substantiate the claim?

A: For a given collection of questions and answers SuperMemo computes a repetition schedule that cannot be much improved by algorithmic means. In this sense, it gives you almost the best possible repetition timing. Obviously, there is a great number of elements that can improve the learning process independently of SuperMemo. These are: (1) application of mnemonic techniques (the same item may be memorized in a number of ways, each of them providing different stability of memory traces), (2) repetition procedure and environment, (3) health and the status of the learner's mind, (4) pharmacological factors, etc. Moreover, the same knowledge may be represented in a different way (different order of items, different wording, etc.) which may greatly affect the speed of learning. As far as SuperMemo optimization algorithms are concerned, their accuracy cannot be greatly improved because of the environmental noise that interferes with the learning process. Each item may be more or less consciously repeated in common life situations; the process being beyond the control of SuperMemo. Shortly speaking, if you are healthy, in a good frame of mind, you use mnemonics, and you use well-structured databases under the supervision of SuperMemo, you may be quite sure that you are getting close to your maximum learning potential, and no natural method can help you speed up learning by a substantial factor.

Q: Tony Buzan claims that the first review (second repetition) of an item should take place in 24 hours. This is not so in SuperMemo. Why?

A: In SuperMemo, the length of the first interval is computed from the forgetting curve plotted in the course of repetitions to make sure that a defined proportion of items is remembered; usu 80-97% (this proportion is programmed by means of setting the forgetting index in the option Lapses). Depending on the forgetting index, the length of the first interval may range from 1 up to 20 days, and is not set arbitrarily. It is computed from the record of repetitions and determined by the requested forgetting index. (requested forgetting index is the proportion of items that are not remembered at repetitions)

Q: 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?

A: No. In the context of self-paced drop-out learning technique used by SuperMemo, Buzan's claim is certainly invalid. In SuperMemo, if the learner chooses the retention of 95%, the typical value of the first interval falls in the range 2-6 days depending on the learner 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%, and consequently the range of the overall retention in the range of 85-99%.

Q: How do you respond to the accusation that the whole SuperMemo theory is contradicted by the claim stated in the MegaMemory program of the American Memory Institute, which says that forming indelible memories is possible if suitable representation of the learned knowledge is used?

A: AMI is a commercial, not scientific institution. The claim that their program produces indelible memories can only be understood as part of the marketing strategy used to promote their otherwise very interesting program. The concept of permastore has been proposed in a limited number of publications in the field of psychology; however, the data collected in the research on molecular aspects of memory clearly refutes such a possibility. In SuperMemo, items whose optimal intervals reach beyond the biologically determined lifespan can be considered part of the permanent memory store. The easiest way to disprove the AMI's claim is to use their program to memorize a large body of intractable knowledge with and without help of SuperMemo (e.g. 1000 phone numbers would do). The knowledge retention after a 6-month-long period will stand at 85-98% with SuperMemo (depending on the forgetting index), and 5-15% without it (depending on individual capabilities)!

Q: How can you claim superiority of SuperMemo by comparing its retention levels with those obtained in the MegaMemory program? After all, in MegaMemory you do not have to spend time on repetitions! (compare the previous question)

A: The comparison was not supposed to show superiority of SuperMemo, it was only intended to falsify the claim of the existence of indelible memories. Mnemonic techniques and SuperMemo are complementary.

Q: Do you not think that the minimum information principle stands in conflict with the ages old rule that the learned knowledge should be highly associative in nature?

A: No. The minimum information principles concerns the representation of knowledge in SuperMemo databases, not in the learner's memory. The principle does not prevent great advantages coming from proper structuring of the learned material. In the optimum situation, the learner should first construct a cohesive model of the learned subject, and only then, apply SuperMemo to make sure that the learned knowledge is sustained in memory as a whole. The knowledge may be highly associative, but strictly targeted neural stimulation, achieved by means of granular representation of knowledge in SuperMemo, is necessary to effectively induce molecular processes responsible for memory formation.

Q: On what basis do you ground your claim that SuperMemo increases the speed of learning from 10-50 times?

A: For knowledge retention of 95%, it can be computed that the number of repetitions in an average learning lifetime (i.e. about 60 years) is 50 times greater for equally spaced repetitions than for progressive repetitions (as used in SuperMemo). For repetitions with no regular spacing scheme, this number may even be greater (e.g. traditional read-and-review textbook learning). Moreover, the greater the required knowledge retention, the greater the increase in the knowledge acquisition rate (classical forms of learning almost never reach knowledge retention above 10% in periods longer than a year!). In practise, users of SuperMemo claim that it increases their speed of learning from 50% to 2000%. These values are, however, highly subjective, as they do not account for so-called intractable items, which are practically unmemorizable without SuperMemo. In other words, learners tend to underestimate the fact that they can reach knowledge retention from 90 to 99%, which would hardly be possible by using any other method.

Q: How can you assume that traditional learning involves equally spaced repetitions? What length of intervals do you use to compare equally-spaced and progressive schedules? (compare the previous question)

A: The comparison did not imply that traditional learning is based on equally spaced repetitions. The term traditional learning is used throughout this book to denote learning with no regular pattern of repetition timing (e.g. 2-4 repetitions in a school year and no repetitions later on). As the level of knowledge retention decreases here gradually from 95-100% to 0-5%, there is no comparable retention figure. In other words, traditional learning cannot compete with SuperMemo in retention categories over 60%, because this retention is not possible without a system for repetition spacing. Therefore, equally-spaced learning was used as a substitute, in which a permanent retention of over 90% is feasible and comparable with SuperMemo. The length of the interval in an equally-spaced schedule is determined by the level of knowledge retention used in the comparison. Obviously, this will be exactly the same value as the length of the first interval used by SuperMemo for the same level of knowledge retention (e.g. 2-20 days for retention 95%, depending on the difficulty of the learning material).

Q: How can you claim that retention of over 10% is hardly achievable by other means than scheduling repetitions? Do you mean then that as a highly skilled expert I know only 10% of what I have learned in my field?

A: A greatly talented reader, fluent in mnemonic techniques, might be able to recall up to 100% of a recently read book. However, this level of knowledge retention cannot be maintained without repetitions for longer than 1-5 days. The 10% figure concerns the periods on the order of years and decades. As a highly skilled expert you certainly remember 100% of what you have learned, but rather not much more than 10% of what you have read and wanted to learn. Please recall the long line of hundreds of textbooks and dozens of courses you have passed before getting your present expert knowledge. Pick one of the books, open it at random and check how much you remember. If you could write your expert knowledge in the form SuperMemo items, you would be surprised to see that it contains only 5-30 thousand items, which you could have memorized in just a year. The books you have read and courses you have passed must have contained an equivalent of 500,000-10,000,000 items.

Q: When an infant burns its fingers, it will not easily forget the experience. How does it relate to your claim that nothing learned once can be remembered for ever without repetition?

A: Some low-level memories may differ in nature from memories typically involved in high-level learning. This results from the existence of specialized nervous circuitry involved in storing inborn memories, which may combine with newly acquired experience. This way, a number of reflexes does not have to be ever learned, and remains imprinted in memory for lifetime. However, in the above case, one should not overlook the fact that repetition of an unpleasant experience such as burning one's fingers, does not have to be based on repeating the experience itself. The mere fact of visual contact with flames, may invoke unpleasant memories and serve as a repetition. This way, each subsequent contact with fire, hot plate, etc. will effectively function as a repetition, without a need to experience the pain.

Q: Can SuperMemo be used to forget things?

A: Forgetting is a molecular process that cannot easily be induced by natural methods. The more so, there are no sensitive methods to induce selective forgetting, though lesion to some parts of the cerebral cortex may produce roughly localized amnesia. However, there is a component of forgetting that may be influenced. This component is interference. Whenever we learn new things, they always interfere with previously learned material. The interference may enhance some memories while obliterating others. This fact can be used to employ SuperMemo in forgetting, by formulating and memorizing a large number of contradictory items that strongly interfere with remembered facts that are to be forgotten. However, you should not expect the effectiveness of such a procedure to be anything but poor.

Q: Can I use SuperMemo to improve intelligence?

A: Yes. Let us define intelligence as the capability to process information. Your brain is like a computer: the better software it runs, the better it performs. You can identify the elements of knowledge that, for example, will make you perform better in intelligence tests. Those elements, if representable in audiovisual or textual form, can be used to learn, and consequently, improve your intelligence. Yet much greater scope for improvement comes with programmable SuperMemo in which, you can write your own DLL libraries of repetitions, or use libraries written by somebody else. With programmable SuperMemo you can learn everything that does not require specialized hardware (if the appropriate programmed database is available, or if you write it by yourself)!

Q: Soon computers will surpass human beings in their intelligence. Is it not better to invest in artificial intelligence rather than to try to force the forgetful memory to accomplish things with ever-disappointing performance?

A: Without disputing the time table for the soon it is worth noting that the soon may become sooner if we intelligently use the human intellectual resources which are indispensable for the progress in artificial intelligence.