Incremental reading (Advanced level)

Contents


Additional skills

Recognizing unsuitable texts

Best articles for incremental reading are fact-rich and context-rich. You need to develop your own rules for selecting quality reading material. Nothing can substitute for your own experience. You will learn to identify texts that are hard to process and yield lower reading efficiency.

Let's consider two extremes:

Many articles fall in between Wikipedia and fiction extremes. Where Wikipedia would say "the http protocol", a typical article might just say "the protocol" (if "the" is clear from preceding passages). Some text are peppered with "as I mentioned in the previous chapter", or "go to the next section", or "the three points explained in the previous chapter", etc. Abuse of working memory in text writing makes incremental reading difficult. This means that if authors use a lot of "referential ambiguity", the texts are not good for incremental reading. Example ambiguity keywords: "the", "they", "he", "it", etc.

Many articles are also filled with irrelevant chaff (supermemo.com isn't impervious to that weakness). Too much beating about the bush without clearly stating the conclusions that are most important to the reader. "Speculative philosophy" might be a good inspirational read, but probably not for incremental reading. You need to decide.

Some narratives should just best be read passively. They may be a compilation of facts that are generally obvious. In such cases you can just read and dismiss. Or you can read and schedule another review in a month or in a year (if you worry you miss something important). Or you can try to write, in your own words, a sentence or two on what new things you have learned from the narrative. Your sentence would shortly extract the quintessence from an otherwise lengthy passage. If it is meaningful and quintessential, you shall find little trouble with locating keywords suitable for clozing.

You need to develop your own rules for deciding which articles are good for reading. The chief rule might be: import anything that looks interesting, start reading, and if you recognize tell-tale problems, justdeprioritize or delete or fish for a few highlights and then delete.

Remember that you need to differentiate between unsuitable texts and difficult texts.

Handling incomprehensible articles

In incremental reading, you will often encounter material that is difficult to understand. You will need to develop analytical skills that will help you identify the reasons for the difficulties. If the culprit is the author, delete the article. If you need to digest other pieces of your collection first, delay the article. If you need more knowledge, delay the article and import more knowledge that will be needed to boost understanding. Do not forget that some texts make an inherently poor material for incremental reading (e.g. descriptions of scientific experiments, mathematical derivations, programming examples in source code, case studies, etc.). In such cases, use traditional methods of thorough analysis, summarize results of your analysis, and use SuperMemo to keep track of your own summaries. See: Recognizing unsuitable texts (incl. example).

This is how you can approach complexity in incremental reading:

  1. Start reading the article from the top. Once you find a difficult fragment, analyze it, and diagnose the reasons for your comprehension problems
  2. If the rest of the article does not depend much on the difficult fragment, extract it, and keep on reading
  3. If the rest of the article cannot be understood without understanding the difficult fragment choose one of the following:
    • if you need more knowledge to understand the fragment: postpone the article (Learning : Reschedule or Ctrl+J)
    • if the fragment is hopelessly intricate and leaves no hope for the future (e.g. because of wrong grammar, wording, formulation, logic, etc.), delete the article, and search for alternative material
  4. If you decide to postpone the article with Ctrl+J, decide what new knowledge you will need before getting back to the difficult fragment. List dictionary entries, encyclopedia articles, articles on the net that you will need to process before going any further. Schedule the search for materials as separate topics or try to search for new knowledge instantly
  5. Estimate the earliest time when you hope you will be able to understand the difficult article and use the appropriate interval with Ctrl+J. If the article includes high priority knowledge, it is always better to err on the safe side and provide a too early review

Randomizing repetitions

You can execute outstanding repetitions in a subset. If you would like to use a random sequence, follow these steps:

  1. open the subset
  2. choose Random : Randomize browser on the browser menu (Shift+Ctrl+F11)
  3. choose Learn on the browser menu (Ctrl+L)

One memory, one action

In incremental reading, you achieve highest efficiency if your process knowledge in small steps separated in time. This way you can accomplish a good memory effect at little processing effort. However, many usersfall into traps of inefficiency where too little or too much work is done on a subject in a single review.

Futile review

Futile review is an example of insufficient work done (one action, zero memory). Futile review is born in this mental scenario: you see a topic and think: I am in no mood for this material now. Let's execute Next repetition. This is wrong! You must take action or you will loop into wasting time and learning little! It is a cardinal sin to execute processing operations without actually learning anything. When a topic arrives, you may have a dozen of excuses: I do not like this one. Let's do it tomorrow. Or I am too sleepy for this one. Or this one will take too much time. If you find yourself in a loop and constantly rescheduling the same topic, or spending time rescheduling a number of topics, you are hurting the efficiency of learning! This is the time that could be spent on more productive steps.

If you do not want incremental reading to become a waste of time, you must always take some action when you see a topic. For example:

In other words, either make tiny inroads into the text, or mark it clearly as low priority or to-be-done later. Consider also Delete or Done!

Item perfectionism

Another facet of the same problem is taking too many actions on a single piece of information. It is highly inefficient to work on colors, fonts, pictures, priorities, etc. during a single repetition. All those actions can be spread over time. Naturally, setting the right priority is one of the most important steps. Perhaps a piece of information is not important enough to ever squeeze through your crowded learning. If so, you will save tons of time on not doing colors and styles.

Here also you should remember: one memory, one action. Each time you review a piece of information, you are allowed to do an edit, font change, template change, category change, etc. However, unless you can do all your actions in a single burst of machine gun keyboard strikes, or unless some actions are associated with learning new things, you should rather limit your actions to a single step per single repetition.

One memory, one action rule demands that every operation in incremental learning should leave a trace in your memory. It also says that one operation on a piece of data is better than two operations at the same time

Example: incremental item structuring

The following item may look like a violation of the 20 rules:

Question: Inflammation is produced by eicosanoids and cytokines. Eicosanoids include (1) prostaglandins fever and vasodilation, and (2) [...] that attract certain white blood cells
Answer: leukotrienes

However, the 20 rules were written in 1999 for classical SuperMemo. Incremental learning is incremental across the spectrum of rules and principles. In particular, formulating items and building understanding are incremental too.

According to the one memory, one action principle, the presented item will assume its final shape some time in the undefined future (or never, if its priority is not high enough). It must be processed incrementally due to the following factors:

  1. incremental build up of comprehension, and
  2. incremental reformulation that requires time.

The 20 rules say "'Do not memorize until you understand", however, understanding is also an incremental process. Converting this item to plain "What eicosanoids attract white blood cells?" might make sense only if the student fully understands and remembers the hierarchy of inflammation factors, and involved eicosanoids. If this is notthe case, carrying the context in the shape of this complex item is a form of transitional stage between a topic and an item. The item still asks the question vital for active recall. However, it also makes sure that full context is provided until the rest of the knowledge structure is firmly established in student's mind. In incremental reading, the order of reading is often chaotic, the understanding is incremental, and the effort to build a solid knowledge structure is gradual too.

In addition to the incremental buildup of comprehension, extensive edits of items are costly (esp. a total rewrite of the item to a plain question). In fact, one of the main advantages of incremental reading is the minimum need for typing. This is why we use electronic sources in the first place (instead of just books that offer no disadvantage when entire items are typed in anyway). This is why an important efficiency principle in incremental learning is to minimize edits by complying with the one memory, one action principle.

Let us see how the presented item might evolve in successive repetitions. Note that all edit steps my proceed only with solidification of related knowledge (i.e. a single repetition may actually bring no edits at all). The execution of those steps will also be somewhat dependent on item priority. High priority items will receive more exposure, more processing and will demand better formulation quality.

Take 1: original complex item

SuperMemo: A cloze deletion whose formulation is improved incrementally over many repetitions in accordance with the principle: "one memory, one action"

Figure: A cloze deletion related to inflammation, with a formulation that seems to violate the 20 rules. This cloze will be improved incrementally over many repetitions along the principle: "one memory, one action"
Take 2: moving a clue to the answer field
Question: Inflammation is produced by eicosanoids and cytokines. Eicosanoids include (1) prostaglandins that produce fever and vasodilation, and (2) [...]
Answer: leukotrienes (that attract white blood cells)
Take 3: removing the prelude
Question: Inflammation: Eicosanoids include (1) prostaglandins that produce fever and vasodilation, and (2) [...]
Answer: leukotrienes (that attract white blood cells)
Take 4: bare bones item
Question: Inflammation: Eicosanoids include (1) prostaglandins, and (2) [...]
Answer: leukotrienes

Learning lists

Lists and sets are difficult to remember. It is hard to remember the whole set of countries that belong to the European Union. When learning lists, you should rather decompose the problem into smaller subproblems.

Let us consider an example in which you want to memorize the entire sequence of letters in the alphabet. It won't be very effective if you use the following item:

Question: What is the sequence of letters in the alphabet?
Answer: A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, Z

You will notice that you frequently stumble on parts of the sequence and need to stop repetitions just to exercise the entire sequence in the traditional way (like we all learn poems by rote).

However, you can approach this in a way that guarantees quick effects:

Question: What is the sequence of letters in the alphabet between A and E?
Answer: A, B, C, D, E
Question: What is the sequence of letters in the alphabet between D and H?
Answer: D, E, F, G, H
Question: What is the sequence of letters in the alphabet between G and K?
Answer: G, H, I, J, K
etc. etc.

After 2-3 weeks of repetitions, you may take on an extra task of recalling the whole sequence after each repetition of these simplified items. This will make sure you can recite the entire alphabet quickly. You will also frequently rehearse that parts of the sequence that are harder for your memory (e.g. V, W, X, Y, Z) as opposed to those that are much simpler (e.g. A, B, C, D, E).

Practical problems with memorizing lists

You may develop a good methodology for memorizing lists in SuperMemo. However, you may later discover that the memorized lists are not very useful in real life, or worse, your memory may fail you when you undergo a baptism of fire. You already know that list should not form an answer to a question. In a list A, B, C, SuperMemo needs to separately understand your difficulties with linking A and B, B and C, etc. You can use cloze deletion to learn lists using multiple items. For example:

Parent template topic/extract:

A B C

Items generated with cloze deletion:

Item 1:

Question: [...] B C
Answer: A

Item 2:

Question: A [...] C
Answer: B

Item 3:

Question: A B [...]
Answer: C

It is important to know that Item 1 above may make you fail to answer with A to the question C if you only learn to answer Item 1 by understanding the association of B with A. In such cases, you will need even more work by formulating items: A-B (where A is the question and B is the answer), A-C, B-A, B-C, C-A, and C-B. Although you will get six items instead of one, you knowledge is likely to be more solid and you may actually spend less time on repetitions of those multiple items than on repetitions of the conglomerate A-B-C item

Lists can often be ignored

In a majority of cases, we do not need to learn lists at all. The perception that a list is worth memorizing is often a reflection of a bad habit we bring from school where lists are a frequent feature at exams.

A user of SuperMemo asked: "Could you please help me with extracting items from the following text? I am really not sure where to mark the boundaries of extracts and how to use cloze deletion?":

Changing Rates of Mental Illness

Mental illness is becoming an increasing problem for two reasons. First, increases in life expectancy have brought increased numbers of certain chronic mental illnesses. For example, because more people are living into old age, more people are suffering from dementia. Second, a number of studies provide evidence that rates of depression are rising throughout the world

This fragment is difficult to process because it is an enumeration (a list) that forms one large logical structure. However, for understanding the subject, you do not really need to remember how many factors affect mental illness. You primarily need to remember the relationship between the cause and the effect. If you ignore the enumeration, you can simply produce the following topics that will each be easy to process further:

  • Mental illness is becoming an increasing problem
  • Increases in life expectancy have brought increased numbers of certain chronic mental illnesses
  • Because more people are living into old age, more people are suffering from dementia
  • Rates of depression are rising throughout the world

If you believe that you cannot live without the enumeration, you can first extract the facts listed above, and then simplify the enumeration by deleting all superfluous information:

Mental illness is increasing for 2 reasons:
  1. increases in life expectancy have brought increased chronic illnesses.
  2. rates of depression are rising

Using Decompose

Little known but very useful function in SuperMemo can help you tackle lists.

If you are learning the names of fruits, you might see a text like this:

Examples of fruits are apples, oranges, pears, cherries, banana...

You do not want to memorize the whole list of fruits, esp. that the longer it gets the more likely you are to fail. There is no definite or complete list of fruits that would make a good memorization target. All you really want to know is that an apple is a fruit, or an orange is a fruit, etc. The list itself is rather useless. You can keep it as a reference in SuperMemo, but memorizing it would be a waste of time.

You can use Decompose to quickly achieve your learning goals in reference to fruits. Start from modifying the topic:

Apples, oranges, pears, cherries, banana ... are fruits.

Generate a template cloze:

Question: Apples, oranges, pears, cherries, banana are [fruits/vegetables]
Answer: fruits

Convert that cloze to a version that can easily be decomposed by placing list members into braces using slashes as separators:

Question: {apples/oranges/pears/cherries/banana} are [fruits/vegetables]
Answer: fruits

You can now treat the question field with Reading : Decompose on the component menu to get a series of clozes like these:

Question: apples are [fruits/vegetables]
Answer: fruits
Question: oranges are [fruits/vegetables]
Answer: fruits
Question: pears are [fruits/vegetables]
Answer: fruits
etc.

SuperMemo: An exemplary cloze ready for Decompose

Figure: An exemplary cloze ready for Decompose.

Thorough preview of articles

It is hard to formulate a simple algorithm for deciding when to preview the whole article before reading, and/or when to extract the most important fragments during a preview. After all, the article can also be read incrementally in a linear sequence. This is a case a multi-criterial optimization where many factors must be taken into account and the ultimate decision will depend on your own preferences.

Here are some criteria:

In summary, these are the most important incentives for the whole-article preview:

Unified store of knowledge

SuperMemo should conglomerate all your knowledge and your learning material. Ideally, you should just keep a single body of knowledge in a single collection. That main collection should be enriched daily by newimports of articles and multimedia from various sources. Keeping your knowledge in a single collection is vital for subset review, searching for knowledge and references, searching for pictures, statistics, progress monitoring, import defaults, etc. SuperMemo provides a rich set of tools for handling disparate areas of knowledge in a single continuous learning process. However, in certain situations, you may create separate collections for various purposes. There are two main reasons for keeping collections separate:

  1. collections that are not part of your lifelong learning, and
  2. collections that need special treatment due to various technical limitations imposed by SuperMemo or current technology.

Here are some examples of situations that justify creating collections separate from your main body of knowledge:

Flow of knowledge

The figure below roughly illustrates the flow of knowledge in time depending on knowledge difficulty:

SuperMemo: Time flow of knowledge in the learning process

Figure: The horizontal axis corresponds with the repetition number and the vertical axis represents intervals (logarithmic scale). Despite a popular belief, the semi-log scale does not produce a linear graph here. Clearly the increase in the length of intervals slows down with successive repetitions. Moreover, the graph corresponding with zero lapses (red curve), results from the superposition of items with lower and faster increase in intervals (determined by difficulty). The bell-shaped curve is determined by all contributing items (below repetition number 10) and then only by difficult items or items with low forgetting index for which the increase in the length of intervals is significantly slower (above repetition 10). To see the above graph in your own collection, use Tools : Repetitions graph on the browser menu

Incremental reading of PDF articles

PDF is a proprietary format. SuperMemo does not support PDF natively. This has always made PDF materials harder to process than ordinary HTML text imported from the web.

There are 4 approaches that are most often used to process PDF incrementally. You will need to see which one is best for your particular material. It may happen that you will need to resort to mixed strategies and use different approaches to different texts. The 4 options are:

Conversion to HTML is most convenient and least expensive. However, some converters and/or some PDFs produce HTML that is quite different than the original, and/or difficult to process in SuperMemo (e.g. requiring extra filtering, or extra manual formatting). Page snapshots are a fast way to read and import pages that are difficult to convert or are read only (e.g. manuals that require a specific page layout). Copy and paste approach is best for articles that can easily be selected in their entirety and which do not contain too many pictures. Finally, the incremental approach is most natural for SuperMemo, however, instead of using read-points, the student needs to make a note where he or she stopped reading the text.

For more solutions, see: PDF at SuperMemopedia.

PDF to HTML converters

As SuperMemo cannot host PDF files natively for incremental reading, the most promising long-term solution is the use of PDF to HTML converters. However, all converters have their limitations. The languages defined by HTML and PDF are not equivalent. Not all expressions of PDF can be expressed in HTML. This is why you need to check a few converters and see which one is best at processing your type of learning material.

Conversion

Here is a short list of converters as compiled at SuperMemoPedia:

If you are not sure which one to choose, you might use this one: PDF online (it keeps a limit of 2 MB on the size of files to convert).

Processing

Once texts are imported to SuperMemo, they may need further processing:

Using MS Word

Some students use MS Word as an intermediary stage in conversion to HTML. However, as much as the conversion of PDF to HTML is never complete and accurate, there are differences between PDF and MS Word that make lifedifficult. At the moment of writing, there aren't any tools that would reliably convert PDF to MS Word format. Pasting from Adobe Reader to MS Word yields poor results (e.g. texts in columns lose or mix portions of texts). If you own MS Word, you can give it a try, however, using converters or other methods will probably work better.

Using OCR

Some users swear by OCR. See: Using OCR to convert PDF for SuperMemo.

PDF with visual learning

You can convert PDF to pictures and employ the tools of visual learning.

Here are the steps:

  1. convert PDF to pictures either by saving as JPEG, or by using page snapshots with Print Screen
  2. paste pictures to SuperMemo or import pictures from a local drive
  3. as pictures are not searchable, it is important to copy&paste a part of article (if possible) that could later be used in search and review. For example, abstracts of scholarly articles with necessary reference tags (#Title, #Author, #Source, etc.)
  4. use visual learning to process the imported materials. Pictures will need to be trimmed (e.g. (1) Alt+click, (2) marquee selection of the interesting part, (3) press Esc, and (4) choose Permanently cut/crop the zoomed/trimmed image file). Use visual extracts (Alt+X) to focus on smaller portions of imported articles
  5. annotate the pictures or write down summaries to further process most important portions with incremental reading. Without the use of cloze deletion, imported material may quickly be forgotten.

This method of reading PDF is analogous to incremental reading of paper materials.

See more at SuperMemoPedia: PDF and Visual Learning.

PDF copy and paste

If convertion of PDF to HTML is not satisfactory, you may need to resort to old plain copy&paste. Some PDF texts paste whole chapters pretty well with little extra text/code (e.g. page headings). Some PDFs respond well to Ctrl+A (for selecting all text) and Copy with Formatting (right click). Usually, Adobe Reader will not let you do multi-page selections that would paste to SuperMemo fast and nicely. You may even get annoyed with your attempts to copy a multi-column portion of a single page. You will get one column selected, middle of the second, and the ending of the third. Other double-column texts will paste as a mix of both columns per line. Sometimes pasting multi-column texts is not possible. This means that you will often need to go from page to page, or even column to column, and repeat copy&paste. That can take ages. Similarly, pictures do not copy as part of HTML. So you need to copy&paste pictures separately. Here again, when Adobe Reader will not let you copy pictures easily, you can just use Print Screen, paste to SuperMemo and trim&crop in SuperMemo (see: Visual learning). Tables are usually best processed in the form of pictures as they usually import poorly to HTML. Some PDF documents do not even allow of selecting texts. In those cases you will need to resort to PDF visual learning too. Sometimes you may just give up and read the article straight without bothering to do it incrementally, or look for alternative texts, or just give up reading altogether. You can also read the article using traditional methods, and copy to SuperMemo only the portions that you would normally extract within SuperMemo.

After pasting text from Adobe Reader to SuperMemo, it will often be jagged, disrupted, wrongly formatted, etc. Converting short passages to plain text often dramatically improves readability. The shortcut is Ctrl+Shift+F12.

For more, see: PDF Copy and Paste

Incremental PDF

If you want to work incrementally with PDF articles, you can use the following strategy:

  1. Save your PDF files to a dedicated folder
  2. Import them regularly with File : Import : Files and folders
  3. Define a PDF template with a large HTML component and a tiny binary component for holding PDF (use that template as the default for the import category, or apply it after import to all PDF articles)
  4. When reading is scheduled for the PDF element, click the binary component and jump to the page where you last finished reading (to automatically go back to the last view pages, in Adobe Reader, check Edit : Preferences : Documents : Restore last view settings when reopening documents)
  5. Paste all important fragments of the PDF document to the HTML component (if texts cannot be selected, use the Print Screen option, paste the picture, and use visual learning tools to extracts individual fragments)
  6. Paste figures to SuperMemo by clicking a selected picture, right-clicking, and choosing Copy Image in Adobe Reader or Print Screen. Use visual learning tools to process larger pictures (trimming, zooming, extracting, etc.)
  7. Process the pasted texts in parallel with reading the PDF document (esp. if they are very important). Alternative, start incremental reading in HTML only when you finish reading the document in PDF

Incremental reading of paper articles

If you have many notes taken from paper journals, or you must read paper articles, you can use a few methods to employ incremental reading in that process. Working with paper will never be as effective as working with electronic material. However, you can still triple your performance with the benefits of incremental learning.

Here is an exemplary algorithm for processing paper notes:

  1. before getting down to processing, search the net! Many noteworthy articles have already been published online. That saves lots of time
  2. if the text is of general nature, you might find a better equivalent (e.g. at Wikipedia)
  3. create a dedicated PaperNotes.kno collection for easy backup. It might grow to a huge size! You can use a dedicated processing time slot for your paper notes
  4. very short notes or very important notes, you can type by hand. The advantage of typing is that it can shorten the texts, give them more meaning, and have them instantly searchable in the collection
  5. the rest of texts you can import as pictures or process with OCR (e.g. use a pen scanner to pick the most valuable pieces from the book)
  6. use a digital camera to quickly snap paper pages. A scanner yields better quality but is too slow
  7. copy the pics to the hard disk
  8. use File : Import : Files and Folders to import all pictures at once
  9. use visual learning and the tools of incremental learning to prioritize and process the imported material (the extract option is particularly useful)

A user of SuperMemo wrote a few words about his experience with OCR in Studying Law with SuperMemo.

See also:

Philosophy of incremental reading

Incremental learning is not for everyone

See: Incremental learning is not for everyone

Big picture in incremental learning

A frequent reservation voiced by skeptical observers is that dismembering texts into little units will result in an inevitable loss of the ability to see things from a distance in their entirety. The question is: Does incremental learning produce a loss of the big picture? What the skeptics fail to appreciate is the power of spaced repetition that stands behind SuperMemo. The SuperMemo method ensures high retention of once-mastered knowledge. This means that there is minimum disintegration of the coherence of knowledge once it is understood and well modeled in student's mind.

The main advantage of SuperMemo is that you convert lots of disparate pieces of information into a solid model of reality that lives in your memory. All these pieces can be dispersed randomly in your collection like pieces of a jigsaw puzzle, however, they fit into a coherent entirety that stays firmly intact in your mind. In other words, incremental learning is reductionist at the level of knowledge processing, but is holistic at the level of memories stored in your brain.

The big picture worry comes from the fact that in traditional education, students rely heavily on their short-term memory. They cram a single subject intensely before an exam only to forget most of that knowledge in the following months unless the subject is reviewed in later courses. It is true that short-term memory can act as a glue that holds the big picture in memory. However, it is the short-term memory and the weakest long-term memories that are gone first. With that in mind, it is only natural that you might worry that if you start intermingling more than a few courses, you will end up with chaos and confusion. It is not the case with incremental learning. New memories, once established, are gradually reinforced to permanently enter the long-term storage. All incremental learning does is to help you focus on a small portion of the material at any given moment.

In traditional (unspaced) education, the big picture is maintained with the glue of short-term memories. As such it is volatile, and subject to forgetting. In incremental learning, the big picture has a lasting value!

Worrying about the big picture in incremental reading is as if you worried that you might forget the structure of your family just because you meet too many members at the same time at a big family occasion, or that you might forget your name when focusing too much on spelling its individual letters. Once the big picture resides in your long-term memory, you can play with details to your heart's content. Today, tomorrow, or in a couple of years.

Incremental learning can make you smarter

See: Incremental learning can make you smarter

Knowledge that makes you smarter

Smart and dumb learning

Not all knowledge is of great value. Memorizing all tributaries of Amazon make take precious time from other areas of learning. Unless you are an expert on Amazon, rain forest, geography, Indian tribes or related areas, you would rather not want to begin your incremental learning experience from that Amazon exercise.

Your learning efforts must be based on high applicability of newly acquired skills and knowledge. If you memorize the whole phone book, your problem solving ability will increase only slightly (mostly through the beneficial effect of memory training on the health of your brain). On the other hand, a simple formula for expected payoff may affect all decisions you make in problem solving and in life in general. It can, for example, save you years of wasted investment in lottery tickets. Millions of people are enticed with huge lottery jackpots, yet they would never agree to give up their whole income for life in order to get it back at retirement in one-off payment, which is a frequent probabilistic payoff equivalent of taking part in lotteries. Using the terminology defined above, you will find most benefit in mastering and understanding highly abstract rules of logical thinking and decision making.

To accomplish smart learning, you will need to constantly pay utmost attention to what material you decide to study. You must avoid short-term gratification at the cost of long-term learning. It may be great fun to learn all Roman emperors and details of their interesting lives and rule. However, unless you study with a big picture in mind (e.g. in an attempt to understand why civilizations thrive or fall), you may benefit less than by struggling through less entertaining but highly applicable formulas of operation research, which might, for example help you optimize your diet, investment, daily schedule, etc. In other words, you cannot be guided by just the fun of learning but by your goals and needs. In time, you will learn to see the link between long-term learning and long-term benefits. You will simply condition yourself to love beneficial learning. Hard study material can still provide instant gratification.

While you focus on your goals, you cannot forget about the overall context of human life. You cannot dig solely into studying car engines only because this happens to be your profession. This would put you at risk of developing a tunnel vision. You might spend years improving a liquid fuel engine efficiency while others would leap years by getting involved in lithium batteries or hydrogen engines. One of the main reasons for which companies go bankrupt is that their leadership fails to spot the change. As corporate darwinism eliminates short-sighted teams, future society will witness more and more intellectual darwinism. To understand the trends and the future, you need to study human nature, economics, sociology, history, neurophysiology, mathematics, computing sciences, and more. The more you lick, the stronger your predictive powers and your problem solving capacity and the creative strength.

Should we learn "trivia" such as "Which year was the Internet born?". The concept of trivia is highly relative! To a child in a kindergarten, the birth of the Internet is rather meaningless. At this stage of development, the child may find it difficult to grasp the concept of the Internet itself. Many parents will wait until the primary school before showing a web browser to their child. The value of putting the date on the birth of the Internet probably develops only in the context of an effort to understand the history of technological development. In this context, 1969 may be as important as the name of Gutenberg. Only when multiple events of the 1960s and the 1970s dovetail together, the commissioning of ARPANET becomes meaningful. When we figure out that we landed the man on the moon before making the first connection via the net, 1969 looms larger. If we dig deeper, we may find it inspiring to know that when Charley Kline tried to log in on October 29, 1969, the network crashed as he typed the letter G. Imagine, you work on commissioning a major installation that you have worked on for several years. You know that the installation implements revolutionary concepts yet it keeps on crashing. You are about to lose heart. This may not necessarily be an emotional event, after all you also need to apply probability to deciding when to give up blind-alley pursuits even after years of investment. The juxtaposition of the small letter G and the groundbreaking concept of the interconnected world will help you see the big picture. If your concept is great enough, you will go on through another 100 crashes in hope of diagnosing the reason. That perseverance may seem to have little to do with "being smart", but it will make you into a winner (given a bit of luck).

Listen to other people's advice and valuations. The younger you are the more you should listen. In the end though, it must be you who determines the criteria for sifting golden knowledge from trivia. Only you can measure the value of knowledge in the light of your own goals.

Remember that not all knowledge can easily be formulated in a declarative manner. Remember then to use the power of your own neural networks: solve problems, practice your skills, compute, abstract, associate, etc. You and others may not be able to see or verbalize some rules but your brain will extract them in the course of practice. Once the rules have been developed, try to formulate them and write them down. This can be of benefit to you and others.

Knowledge that assists problem solving

Problem solving in artificial intelligence can be used to mimic those mental faculties of a human being that we associate with intelligence and creativity. There are many approaches to problem solving such as deduction, induction, abduction, reasoning by analogy or probabilistic methods, neural networks approach, searching in state space, and more. And even though some of these can generate false conclusions or uncertain responses, they can all be expressed by means of deductive methods such as those used in theorem proving in mathematics. Deduction proceeds from axioms or premises towards true formulas or assertions through logical derivation using valid inference rules. The better the selection of assertions and the selection of inference rules, the broader the reasoning capacity of an expert system or human brain. Rich inferential knowledge stored in our memory makes us fast in thinking, creative, intelligent, efficient in problem solving, etc. Yet we need factual knowledge as the raw processing material for derivation. Eventhe creative vein and inspiration can be expressed in terms of multithreaded derivation and backtracking well formalized in logic programming. No magic fluids or reflections of the soul are involved here. Just a plain network of firing neurons recovering the encoded patterns of facts and rules. See also: Roots of creativity and genius.

To boost your problem solving capacity you need to master a great deal of inferential knowledge. If you learn about specific classes of problems to solve, you may need to choose between memorizing the outcome of inference or its individual steps. Both equip you with a formula for transforming knowledge into new quality. Learning derivation steps is costly. At the same time it can equip you with highly abstract derivation rules. You need to balance cost and benefit. We cannot blindly assume that it is better to derive answers to problems than to simply memorize the answers. In problems with multiple instances, we will tend to learn the derivation. For example, it is easier to learn how to derive the third power of X than to memorize all useful pairs: argument-result. On the other hand, most of people memorize the multiplication table at the time when they could still derive the result from bit-operations on binary numbers. Memorizing some 40 or so combinations seems more practicable, esp. that derivation always takes time and we make use of the multiplication table thousands of times in a lifetime. Here memorization saves hours and days in the lifetime perspective. Brain armed with rich factual and inferential knowledge will associate the most remote ideas and derive the most unexpected conclusions. Through a conscious control over this rich reasoning process, we have built the present civilization.

Knowledge that assists creativity

Creativity is based on association of ideas. It is helpful to be born with a creative brain that keeps spewing new ideas in all circumstances. However, creativity can also be assisted. When you master a great deal of facts and rules in areas adjacent to the problem that you try to creatively solve or elaborate upon, your creativity gets enhanced. If you employ incremental learning, the association of random ideas will be a norm throughout the process.

All innovation in the history of science and technology is based on association. Forgetful memory can be helped with external sources of knowledge and each major scientific breakthrough is based on a series of smaller steps, many of which will be based on hours of search, trial-and-error and experimentation. The brain armed with extensive knowledge is more likely to come up with the needed associations. As in problem solving, inferential knowledge is also particularly valuable in creative efforts. The rules that you know, help you derive new truths, and associate these with all your current knowledge.

Two ideas do not come together to produce a great invention unless they sit in the same head. They either have to be called up at the same time or one has to come unexpectedly from the outside:

SuperMemo can help you be creative not only by combining various pieces of knowledge in your mind. It can help you generate new ideas while making repetitions! Creative associations do not come from the mere existence of two pieces of knowledge in your brain. Those two pieces of knowledge have to light up at the same time. Only this way can your brain make a connection. Strangely, a repetition related to genetically modified tomato can light up memories related to tomato juice, cucumbers, genetic disorders, take-home dinner, flu or even the silent Mars Polar Lander. If your collection combines knowledge pertaining to different subject domains, the stream of new ideas and unexpected associations coming to your mind may surprise you!

Incremental learning vs. human progress

Incremental learning has a potential to boost your knowledge and boost mankind's potential to solve major problems of the day. However, in the long perspective, the role of human learning will be lessened by the development of artificial intelligence.

Knowledge and history

Throughout the ages, knowledge was the cornerstone of human progress. From Stone Age to Information Age, in pain, we have built a tiny oasis of civilization in ruthless expanses of the evolving universe. The history of the mankind is made of billions of individual lives that keep on sparking and fading. Born of self-preservation instincts imprinted by evolution, history books paint a picture of a constant string of wars, conflict of interest, loss and gain of influences, lust for power and submission to weaknesses of human nature. On the other hand, the greatest achievement of the evolution, the rational mind, kept on contributing to new findings, discoveries, technologies and philosophies. Progress has always hinged on discovering new truths and preserving them for posterity in form of stories, solutions, tools, books, and other carriers of information. Knowledge is the basis of human power, yet it constantly struggles with two forces that regularly diminish it: death and forgetting. We can preserve knowledge in books and other forms of information storage. However, this knowledge translates to value only then when it is used by the creative power of the human brain. The limitations of the human brain will remain a bottleneck of progress for many years to come. We will develop artificially intelligent knowledge processors not earlier than in a decade or two.

Knowledge and death

Death poses an ageless challenge to educating new generations. Years of hard work needed to gain knowledge on professorial level are obliterated in a single act of death. Newborns need to go through years of education before they are able to access, read, and comprehend this text. They all have to struggle with basic literacy skills, lessons of safe sex and teen pregnancy, lessons on superiority of altruism over egoism, the difference between wise and not-so-wise choices, existential questions, etc. Although constant reeducation may contribute to gaining a fresh perspective in each generation, it is also painfully wasteful. As yet, there is no efficient remedy to the death of knowledge. All we can do is to attach more weight to healthy lifestyle and health research. Those two promote longevity of knowledge in a single generation.

Knowledge and forgetting

Forgetting is a natural process that makes it possible to efficiently use the limited memory space of the brain. We forget to dispose of knowledge deemed less important in order to make space for knowledge of higher importance. Currently we have only a limited control over what we remember and what we forget. Today, the most important tool that we can use to prevent forgetting is practice. We can minimize time needed for practice by using spaced repetition (i.e. learning technique based on computing optimum intervals between repetitions). Spaced repetition is the key to maximizing knowledge within a single human lifetime.

Immortal knowledge

Artificial intelligence is our best hope for approaching immortal knowledge. It can nearly eliminate the problem of death (except for the heat death of the universe or factors that we cannot foresee today, e.g. the collapse of space, and the like). Artificial intelligence can also eliminate the problem of forgetting (at least within the bounds of the available storage). Some forgetting is needed to shape crisp associative knowledge, but that is a welcome phenomenon that leads to more applicable knowledge. Despite great hopes we might have about artificial intelligence, the best path towards immortal knowledge must still rely today on the use of the human brain. This is why we believe that incremental learning is so vital for further human progress.

Incremental reading is an extension of traditional book reading

Someone remarked: "If Gutenberg was a blessing then incremental reading might be a curse!". Is incremental reading an attack on traditional books? If you read in pieces and with endless interruption, does it not destroy the storyline?

Whether incremental reading is a curse or a blessing depends on the way it is employed. There is no sharp transition between traditional reading and incremental reading. In the simplest case, you can use incremental reading exactly in the same way as you would read a book. Partitioning of texts and interruptions are not compulsory. You can read the entire text from top to the bottom without a single interruption. This would make sense if you needed a storyline for context, and did not want to bother with committing it to long-term memory. If you do take breaks or skip portions of texts or change the natural sequence of reading, it all happens in situations that have their counterparts in the world of books:

In other words, in extreme cases, there may be no difference between traditional and incremental reading. Gutenberg's blessing is safe. If you believe interruptions or multiplicity of subjects are beneficial, you can employ them at greater ease that it is the case with book reading. At the other extreme, you may wish to take on thousands of independent articles, make interruptions a norm, focus your reading only on portions that you deem most important, etc.

A rule of the thumb is: use traditional reading when you read stories or you read for enjoyment. Use incremental reading to process learning material, textbooks, notes, or scientific literature that you need to remember for life (or at least for many months).

Knowledge acquisition: Areas of optimization

There are seven main areas where the learning process can be enhanced. Incremental learning plays a role in each, however, the first two areas will benefit only through your growing experience, and ability to quantify your progress:

  1. access to knowledge: with the advent of the world wide web, we all have information literally at our fingertips. Searching for knowledge with Google is fast, easy and accurate. Wikipedia is the biggest encyclopedia in human history. Access to knowledge is no longer a bottleneck of human learning. It virtually became a non issue! Web's quality and role will for continue increasing exponentially as more and more people appreciate its potential and contribute to its growth. There are still many complementary sources of information that compete successfully with the Internet. However, it is only a question of time before you will be safely able to rely on the Internet as your sole source of information. The great benefit of reading the web, as opposed to reading the books, is that the hypertext nature of the web enforces a very compact and usually self-explaining nature of individual articles. A jump to a randomly selected page in an average book will leave you confused due to the context-dependence of the material. On the other hand, it is less likely the same confusion will trouble you in a random jump to a selected page of an equivalent material placed on the web. Web authors usually put more effort to add contexts to the page (at least in the form of hyperlinks). In other words, it is easier to build quality knowledge by reading single pages of the web than by reading single pages of paper books. We are getting closer to the ideals of incremental life-long learning as opposed to thorough-review learning which for many ends with the end of school years. In the busy days of modern society, few can afford a thorough review of their rusty knowledge in individual fields. It is much easier to fix the gaps incrementally: today an article on the structure of the atom, tomorrow an article on a healthy diet, etc. And all that strictly adjusted to individual's interests and professional priorities
  2. selecting knowledge: you will face the need to fill the gaps in your knowledge in many more areas than your time permits or your memory makes possible. You can ask SuperMemo to help you scrupulously note down and prioritize all areas of knowledge that need an enhancement! In SuperMemo, you are the master of what you learn and what you neglect. Your ability to select valuable information will grow in proportion to the acquired knowledge. Incremental learning helps you quantify your knowledge and extrapolate into the future. This will help you be more selective in choosing your learning materials.
  3. reading: this is the first stage where knowledge makes an actual intimate contact with the brain. Traditionally, it is streamed into memory in a more or less linear manner (i.e. paragraph after paragraph). Incremental learning helps you delinearize this process and optimize reading by enhancing knowledge selection and prioritization concurrently with reading. For example, you should be able to say "This paragraph can be processed later" or "This paragraph requires utmost attention right now" or "This paragraph can be skipped for good even if I decide to read the article again" or "I want to read this paragraph again in three days and in more detail" or "I want to mark this paragraph with lower priority and come to it only after all higher priority paragraphs have been processed", etc.
  4. representing knowledge: the way in which you present knowledge will affect comprehension and retention (i.e. how well you remember). Things that are simple are easier to understand. Things that are simple are also easier to remember. Many people do not realize the degree to which simplicity can affect learning. Many people doubt that even the most complex material can be presented in a very simple way. Einstein noticed that "it should be possible to explain the laws of physics to a barmaid"
  5. remembering knowledge: SuperMemo helps you eliminate the problem of forgetting. You will base all learning on spaced repetition that SuperMemo pioneered a quarter of a century ago. SuperMemo produces immense savings in time by scheduling review of the studied material only then when the review is necessary (see: Introduction to SuperMemo)
  6. life cycle of knowledge: knowledge in your collection and in your memory keeps on evolving and maturing. This will involve continual rewording, reprioritizing, and re-associating pieces of knowledge. You will often give up portions of knowledge that become outdated or lose their high-priority status. You will apply the rules of knowledge representation that will make knowledge easy to remember. Your knowledge will also become more associative in time. In other words, it will become a more suitable ground for making intelligent choices
  7. using knowledge: knowledge translates to value when it is properly used. In the long-run, skills discussed in this article will indirectly help you become more creative and skillful in using your own knowledge. Not surprisingly, your skills needed to efficiently use knowledge are also part of knowledge itself and tend to grow spontaneously as your knowledge increases

Cost-Benefit Analysis

Cost of knowledge

Each piece of knowledge stored in SuperMemo adds to the total cost of learning in terms of time. If knowledge is well-formulated, we can estimate that a single element will be repeated from 8-20 times in a lifetime. If a single repetition time falls in the range from 3 seconds to 15 seconds, we could conclude that the expected lifetime cost of a single item ranges from 24 seconds to 5 minutes. SuperMemo has a statistical measure called Cost. It measures the total time cost of a single memorized element. In a well-managed long-term learning process, this cost is usually estimated at 2-3 minutes/item in a 10-year bracket. This is then more than the theoretical prediction due to the fact that each collection contains a small subset of so-called leeches which dramatically increase the average cost/item (note that leeches can easily be detected and removed from the learning process).The rational criterion for deciding if a piece of knowledge should be memorized by means of SuperMemo is to judge the benefits of having the given piece constantly available in memory. If the benefits do not seem to add up to more than 10 minutes, the student might simply not add an item to SuperMemo. With a dose of practice, this analysis becomes a semi-automatic process, and should painlessly blend with your life

The cost-benefit criterion is: if costs of not knowing a piece of knowledge is greater than the cost of repetitions in a given period of time, add this piece of knowledge to SuperMemo (otherwise do not addit, or relegate it to lower priority).

Cost of items

Cost of a single well-structured item in lifetime ranges roughly from 24 seconds to 5 minutes. At any given moment it may be approximated for a 30-year bracket, by multiplying Future repetitions (in element data) by Avg time (in learning statistics). SuperMemo displays this value in the element data window. If your item is ill-structured (i.e. difficult to remember), this cost may bloat! To eliminate ill-structured items use leech analysis.

Benefit

Each piece of knowledge in your collections should be associated with a tangible benefit. Only you can accurately guess the value. For example, the value of memorizing the opening hours of your gym in a given time might be approximated by multiplying (1) the probability you will choose wrong hours by (2) the time-cost of missing the gym. For example, if you believe that the probability is 25% and the cost of choosing wrong hours is 40 minutes, the cost of knowing the opening hours is around 10 minutes. In such a case, memorizing the opening hours will be cheaper than missing the gym. However, if the probability is sufficiently low or the time loss sufficiently small, you should not add opening hours to SuperMemo. For example, if the probability is 10% and time loss is 3 minutes, you are not likely to recover your learning investment! There are naturally less clear-cut cases in-between. If you are not sure if you should add a piece of knowledge to SuperMemo, add it with sufficiently low priority. With overflow tools (e.g. auto-postpone, auto-sort, etc.), you can drag along less valuable pieces of knowledge at little cost.

Priorities

You will quickly realize that there are awfully many pieces of information that pass the cost-benefit criterion. You cannot ever hope to learn all this information. This is actually optimistic. This means that with a good selection of knowledge, you can gain far more than just a few minutes on memorizing the gym hours. You may gain untold hours by learning things that transform your life. Your best tool in making sure you always focus on the most beneficial material is to use the priority queue. If you have too many items to remember, those of lower priority will get a less meticulous treatment and will carry a higher probability of forgetting. However, this also means that they will benefit from the spacing effect and you will learn more of such items at higher speed (despite lower retention).

Selecting knowledge

The Library of Congress holds 10 terabytes of printed materials. Global knowledge resources can be measured in petabytes (10005 bytes). All digital information has already reached zettabytes (10007 bytes). Only a micro-fraction of those resources can be mastered by an individual in a single human lifetime. Even a single copy of the Encyclopaedia Britannica goes in detail far beyond what a single human being can encompass in a lifetime! The actual speed of learning and lifetime learning limits can be measured with SuperMemo (see: Theoretical aspects of SuperMemo).

The microscopic capacity of the human brain has not prevented it from building the present civilization as we know it. The human power comes from:

  1. collective effort - a billion heads is more than one
  2. specialization of labor - all collective tasks are subject to top-down functional decomposition and a single brain usually only needs to process a fraction of information at a time
  3. knowledge selection skills - the associative power of the human brain combined with the selective nature of forgetting help us retain memories that are actually most useful in problem solving

Incremental learning is a great way to combat forgetting. However, forgetting plays an important role in our lives. It runs a valuable garbage collection on knowledge we acquire daily. If the power of forgetting is diminished (as it is the case in incremental learning), your responsibilities in the area of selecting knowledge increase manifold!

SuperMemo will help you eliminate forgetting! At the same time, it will increase your responsibility for selecting knowledge that is truly important and applicable. If used without care and attention, SuperMemo may actually waste your time by helping you remember reels of garbage trivia

A piece of information that occupies just several bytes of your hard disk may carry a relative value that my translate to a net gain of millions of dollars as well as a net loss of millions of dollars. It may also carry no value whatsoever. For example, a sentence written in French "SuperMemo vous aide a mémoriser et apprendre diverses informations comme une langue, des chiffres, etc." may be of nearly zero value for someone who does not know French. At the same time, an item related to a Heimlich maneuver can save the life of a family member. We know that the expected payoff equals the value of the payoff multiplied by its probability. Therefore, the low probability of a family member choking and the probability of actual successful application of the maneuver make the value of "Heimlich item" a fraction of the value of the human life. At the same time, even minor errors in medical knowledge of a physician can actually cost somebody's life and carry substantial negative value!

Frequently, you will find more benefit in memorizing the three best things you have learnt today than in memorizing a whole monothematic article to the last detail!

Readers are leaders

You must have heard it from your parents or teachers that "Readers are leaders". Harry Truman added a pessimistic twist to that claim: "Not every reader is a leader, but every leader is a reader". That qualification tells you that reading is not a panacea, and that making reading smart is vital. With the advent of the web, Wikipedia, blogosphere, and other sources of knowledge, access to information is easier than ever. This also means that information becomes more overwhelming. There are always too many things to read. And it is hard to select from a myriad of essential sources. This is no longer a world with a hundred books of which you needed to pick five. Now you literally have thousands of menu options for a single meal. Old bookworm dilemmas are now even more painful. Search the web for solutions for book selection, and you will literally find dozens of blogs with advice. However, most of the advice is based on volatile skills that are hard to master. You need to become an expert in book selection, in speed-reading, in skimming, in skim-or-read decisions, in highlight-and-review strategies, in margin notes and in underlines. For traditional massive reading mastery (without SuperMemo), you would need to go for the extreme: reading in the toilet, reading on the train, reading on your mobile phone, making use of your dead time, etc. You would need to ruthlessly control youremotions with "not to read" decisions, or "not to finish" decisions, or "not to re-read" decisions, etc.

If you master incremental learning, you will not solve the problem of information overload. However, you will systematically address all issues that pester massive readers. Once you embark on the only rational reading road, you will free your mind from all the dilemmas and stresses involved in reading. Incremental learning helps you focus on learning itself!

As for the bathroom time, incremental learning can free that for creative thinking. After all, you also want to ponder your own ideas, not only process other people's output.

Incremental reading solves all major dilemmas of massive readers! Incremental readers are leaders.

Does incremental reading slow down mastering complex subjects?

Could incremental reading slow down or hamper learning complex subjects such as quantum mechanics? Is complex knowledge based on just "remembering passages" or does it go well beyond it. Any form of interruption or delay while working on many things in parallel can potentially slow down this process. Many people believe that incremental reading is good for cramming, but not for serious learning, esp. when complex subject matter is concerned.

Example Opinion 1: tackling quantum mechanics

Here is an example opinion of someone who never tried incremental learning:

I believe that incremental reading (IR) may slow down or hamper learning complex subjects that require deeper understanding -- as opposed to superficial remembering. From what I'm told, in IR, the student spends but a few seconds on a passage related to one subject, and then jumps to another, completely unrelated subject. Such switching among disparate subjects makes it difficult, if not impossible, to engage in deeper thoughts and thus to discover deeper truths beyond the superficial remembering of passages. When I read, I look for concepts and methods that underlie the necessarily descriptive language used to convey them, and I try to imagine their applicability in my other pursuits. In other words, discovering deeper truths requires deeper thoughts. The latter are difficult to reach if only a few seconds are spent on a subject at a time. As a practical aside, technical papers in hard sciences and engineering are peppered with notational inconsistencies among different authors, different research groups, and different journals. The same symbols may mean different things in different scientific papers. Similarly, different authors may use different symbols to denote the same quantity. Sometimes the differences are gross and easy to spot, and sometimes they are only subtle. When reading one paper at a time, the reader familiarizes him/herself with the terminology and notation, and can follow the line of reasoning while recognizing the subtleties. Having a few dozen (or hundreds!) articles read in parallel, frequently switching among them, creates a nightmare scenario of confusion where the meaning is lost in the jungle of various notations and subtle terminology differences. This makes it impossible to follow the line of reasoning in any one of the papers. Below are a couple of (imperfect) metaphors that may convey the sense of loss when trying to apply IR to learning complex subjects and to discovering deeper truths:
  1. If you are a wine connoisseur, imagine that you are only allowed to smell the wine, but that you are not allowed to drink it, or even to put it in your mouth. How good would your knowledge of wine be? With practice, you would probably learn to distinguish various wineries, age, etc. You may even be more efficient at it, and be able to do it in record time. Yet, I venture a guess that you would feel something missing, that you would feel somehow cheated, that you would feel that you don't know all that there is to know about wine. And rightfully so. Same with IR. You just 'smell' the knowledge, but you never 'taste' it. It may be enough if your goal is to learn a foreign language. But some areas require that you dig deeper if you want the real insight, the real truth.
  2. Physical exercise. A well-rounded routine has both cardiovascular and resistance training elements to it. IR is like having only the high-intensity cardio. Sure, if you stick with cardio, after a while you will be able to deliver a small package across town in record time -- and that may be all you aspire to. However, it will not give you the upper body strength to help move a 400-kg optics bench. Same with IR: you may impress your audience with the breadth of your knowledge, but you will have to leave the heavy intellectual lifting to others.

The above reasoning can easily be proven false. We can imagine two extremes:

It easy to show that both are flawed. For example, traditional learning is deprived of spaced repetition, while pure incremental learning may be seen as deprived of conversation or lab practise. Clearly, the optimum falls somewhere in the mix between the two extremes. Incremental learning is a supplementary tool in well-rounded education. It is not supposed to monopolize your day and your thinking. It should not be seen as a replacement, but as an enhancement. As such, it is up to you to find your optimum.

Incremental reading is all you want it to be. It can be speed-reading, cram-reading, or mass-reading. It all depends on the priority criteria which you choose. For that reasons, it would be best described as a reading management technique. As such it is indispensable independent of the complexity of the studied matter. On one hand, you can speed-read articles faster than in conventional speed-reading and yet leave vital paragraphs for future review. On the other hand, you can meticulously dismantle individual paragraphs and convert them into classical questions-answer knowledge that will stay with your for ever. In addition, you can freely manipulate the volume of the material flowing into the reading/learning process. You can focus on a hundred most important articles or you can opt for thousands. Naturally, in the latter case, your time allocation for individual articles will be minute. For example, if you import 10,000 articles to SuperMemo, you might end up with 50,000 to 100,000 extracts within a year of 1-hour daily reading. In such circumstances, low priority articles will indeed linger for months in the process. Naturally, this is exactly the purpose of incremental reading: focus on what is important without neglecting anything that falls within your area of interest. If your focus changes, you can use search and navigation tools to speed up the review of most important portions of your reading material.

Incremental reading is universal in textbook learning. Whatever complex concepts you need to analyze, and whatever computations you need to make on the margins, you can do in parallel with incremental reading of the textbook (assuming you have an electronic version). An old rule says: whatever you need to remember for life (or at least months), process incrementally (to improve memory, boost understanding, and save time in the long term). Whatever you need at the moment, for the sake of understanding the subject, process in bulk (non-incrementally). By combining the two, you can get the most of your learning.

Metaphorically, a pencil is a useful tool that can enhance your life. You will not want to replace your computer with a pencil, or use a pencil while cooking. However, you can still enhance your life by having the pencil handy. Needless to say, we believe, incremental reading is far more useful than a pencil.

The presented reasoning is not only wrong, it is also based on a few misconceptions about incremental learning:

Example Opinion 2: struggling with basic physics

Here is another example opinion of someone who tried incremental learning but found it difficult:

I think incremental reading is either very difficult or impossible to use when learning some concepts of physics. For example, I have the following text about the Earth and the Sun, how would you handle this with incremental reading?

The Earth is moving very very slowly away from the Sun. This happens for two reasons. The first is that the Sun is constantly losing mass because of the solar wind. As the mass of the Sun decreases its pull on the Earth decreases and so the Earth moves slightly further away. The second reason is to do with tidal forces. In exactly the same way that the Moon is slowly moving away from the Earth, the Earth is very slowly moving away from the Sun. In the Earth-Moon case the Moon pulls on the Earth creating tides and slowing the Earth’s rotation very slightly, making the day longer. This action has a reaction - the Moon's orbit is speeded up. If something travels faster it must move outwards to remain in an orbit and so the Moon slowly drifts away from us at a rate of 3.8 centimeters per year. The same situation happens with the Sun but the Earth’s influence on the Sun is much smaller than the Moon’s influence on the Earth. The result is the Earth’s tiny tiny drift away from the Sun

The example isn't much harder to process than other pieces of knowledge suitable for incremental reading. For success in similar cases one needs an encyclopedic text or a degree of editorial effort to dismantle portions of more elaborate prose. The presented example poses two obstacles:

  1. Implicit enumeration. The text mentions two reasons why the Earth moves away from the Sun but it does not name them in an explicit sentence. One needs to read the entire passage to find out the second reason.
  2. Explaining by analogy. The effect of tidal forces on the Sun is explained by describing similar forces created by the Moon. You cannot extract the "second reason" without including and understanding the "Moon example context".

Here is how the presented text would be handled with incremental reading (note the editorial effort as well as the need to entirely rephrase one of the sentences):

Some authors make incremental reading very difficult by assuming a great deal of knowledge on the part of the reader or, as it is the case here, loading student's working memory with a great deal of data rather than building knowledge gradually (i.e. from the ground up). In the discussed example of quantum mechanics, the basic vocabulary needed to process any meaningful text is pretty extensive and capitalizes on manybranches of physics and mathematics. In addition, good understanding of all concepts requires a great deal of mathematical practice that goes beyond linear reading.

Incrementalism and interruption are not compulsory

If you love to go through texts from cover to cover without a stop, and always find it hard to stop reading a book before sleep, you may still benefit from incremental reading.

There are no incremental readers who did not begin from misgivings about interruption. Paradoxically, the stronger your misgivings, the better candidate for a good incremental reader you might be!

A popular misconception is that there are impatient people who are predisposed to be incremental readers - let's call them sippers - and those who love to devour knowledge in large chunks - let's call them gulpers. The truth is that all creative individuals are of a gulper nature. Incrementalism is both a skill and a habit all gulpers may learn over time.

Nobody loves SuperMemo as of the first day. It may take a few weeks to notice its power. And still, as we do not have sensors of the speed of forgetting, you need a dose of rational mathematical appreciation of what SuperMemo does to your brain. You cannot easily sense the power of knowledge and how fast it is being undermined by forgetting.

Incremental reading takes far longer to be appreciated than SuperMemo itself. To employ SuperMemo, you need to learn only two operations (Add new and Learn). For incremental reading, you need a toolset that keeps growing and improving over the years of use. Yes! Even after a few years of learning, you will discover new ways you can speed up your own learning with incremental reading. It may take a year before you might notice first signs of addiction to incremental reading. A benign form of addiction, mind you, with few negative side effects.

The paradoxical tendency of "gulpers" to become good incremental readers comes from their hunger for knowledge. The fact that you cannot stop reading is a powerful expression of this hunger and it is the primary driving force that will help you become an addictive reader. What you might be missing still is the understanding of the power of incremental reading and the hunger to "switch for more". Incremental reading will help you develop a hunger for maximizing the value of information you are processing at any given moment.

You can begin incremental reading today without ever having to stop reading an article that you find fascinating.

In incremental reading, interrupted reading is a norm, but is NOT compulsory!

You can read all articles from front to back and only use incremental reading tools for prioritizing articles and extracting most important sentences and converting them to clozes. In other words, you do not need incrementalism to achieve solid retention of knowledge. An ordinary web surfer has only two alternatives when encountering an article that seems worth reading: (A) Fascinating, let's read, and (B) Not fascinating enough. Perhaps I will read some other time. In contrast, an incremental reader can determine the priority of the article and always read only the articles from the top of the current priority list (perhaps with a user-defined degree of randomization). Moreover, at any time, he or she can say: Interesting, but not as much as I thought. Let's downgrade the priority and come back later (if ever).

A gulper is driven by a natural neural mechanism that underlies all human progress: curiosity. The same mechanism can be used to magnify incrementalism: curiosity of what article or paragraph comes next. Once you develop a healthy incremental reading process, you will add another natural neural mechanism: impatience. Impatience is also a buttress of progress. We do not like long stretches of low efficiency. We like instant gratification of success and the bigger the success the better. In incremental reading, you are constantly driven by curiosity and yet you itch-to-switch as soon as the text you are reading does not bring sufficient value-per-time. The healthier your incremental reading process, the more value per second you can extract. You will develop a sense of the realistic average value stream, and each time you fall below that expectation, you will add up to the incremental nature of reading (even if the fault is yours, not the text author's, e.g. when the gaps in your knowledge produce poor comprehension). By combining curiosity with impatience, you can convert from a gulper to a sipper. And still you will be able to read top-quality articles top-to-bottom without interruption. Incremental reading helps you prioritize by content instead of reacting to transient evaluative impressions.

You will notice that incremental attitude is a habit you grow as your technical and parsing skills improve. Rarely will you delete lower quality articles, but these will fade in priority and may indefinitely linger in the process. As a result, you will maximize the educational effects of every precious second you spent on learning.

As an incremental reader, you might gradually develop a dislike of old-style books (as opposed to importable e-books and articles). If you choose to read a book, it is as if you said: "this is the most important reading material in the whole world". Then the whole series of paragraphs in the book are considered the most important paragraphs to read in their precise sequence as they appear in the book. You give the author of a book God-like powers to stream information into your brain in a flawless, omniscient, and omnipotent way.

Gulpers and sippers are not biologically different! The conversion from one to the other goes via the understanding of incremental learning, mastering SuperMemo toolset, honing the skills, self-control, rationalization of the learning process, and gradually pumping up the average value of knowledge streamed into one's memory.

Incremental learning for perfectionists

If you are a perfectionist, you may initially have problems with accepting the chaos of incremental learning. You may wonder why you should leave your cloze deletions unfinished before they look perfectly adorned with fonts, stylesheets, and pictures.

If you give incremental learning a more determined try, you will understand that the opposite is true: perfectionists should love incremental learning! Your perfectionist nature should accept the overriding rule: maximum quality knowledge in minimum time. It is not the beauty of clozes in your collection that counts, but the beauty of knowledge in your mind. For a skillful student, incremental learning is based on a set of perfectly-formed strict and rigid rules that guarantee the maximum speed of knowledge acquisition. It is true that some of these rules can make you uneasy at first. If you see a sentence that qualifies for a cloze, the rule is: execute the cloze deletion and defer worrying about its exact formulation to its first repetition. The mere choice of the cloze keyword will leave sufficient traces in your memory to qualify as a repetition. In such circumstances, perfecting the formulation of the cloze will become art for art's stake. A higher level rule is: minimum work for maximum memory effect. Therefore, you will improve the formulation of the cloze as soon as you proceed with the first repetition. And again, you will do only as much work as it is necessary to successfully complete a single repetition act. Again you give up details and frills. Ultimately, your cloze will become perfectly formulated, perfectly prioritized, and perfectly placed in your knowledge tree. Alternatively, it will be deleted or left lingering in your "to do" subset. It is the perfect rules of incremental learning and the perfect learning results that should feed your perfectionist needs, not the perfect "look" of your learning material.

Many people tend to hold the world wide web in contempt calling it the "human information garbage dump". This attitude makes it hard to utilize the web as the "goldmine of human knowledge". Tim Berners-Lee created "perfect rules" (HTML, HTTP) for knowledge dissemination by the populace. We can adapt our own "perfect rules" for mining the web. Incremental learning uses "perfect rules" to convert web data into golden knowledge. As a perfectionist, you should not worry about the chaos of the web or chaos of your collection. What really matters is the perfect golden end-result: wisdom!

Finally, if you still cannot live with imperfectly formulated clozes, nothing prevents formulating them perfectly. The formulations may be more satisfying to your perception, but you will, naturally, learn at a slower rate.

Incremental reading is not an attention destroyer

In incremental learning, it is very important to make the right choice of the learning material. Many texts or videos are unsuitable for incremental processing. SuperMemo user who never tried incremental learningwrote:

It is possible that article structure and quality do matter that much in incremental reading because the real learning bottleneck is the human brain (speed of cortex plasticity)? What if you can only absorb X items of knowledge each day -- if you try to do more, you do something bad to your brain? Here's one idea: If you overload your brain with new stuff, it won't have the time to form meaningful connections between the things you know, so your knowledge may be reduced to the ability to answer gameshow-like questions.
Solving any significant problem requires periods of prolonged concentration. I fear that an information addiction (200 new tidbits per day) leads to attention deficit. Your brain is used to getting something shiny and new every 15 seconds (a new tweet, new funny pic, new headline, etc.), so when you tell it to work on one thing for 4 hours, it doesn't listen. I think I was able to concentrate more on one thing in times before Web surfing.

So, to put a tabloid spin on it, incremental reading could be the ultimate attention destroyer!!!

There is a lot of truth in the above reasoning in reference to attention, memory bottlenecks, "meaningful connections", etc. However, comparing SuperMemo to Twitter or Facebook as used by Internet junkies is very inaccurate! The reward in incremental reading is based on quality learning, not "something shiny or something funny". Naturally, nothing prevents "shiny/funny" things to be imported to SuperMemo. This is why the nature of the ultimate reward will also depend on one's personality and self-discipline.

When employed along the recommended rules: incremental reading should dramatically increase the attention (as explained in Advantages of incremental learning).

Learning speed bottleneck

Cortex plasticity is indeed the bottleneck in the learning process. All speed-reading and speed-learning efforts may go to naught if you do not employ spaced repetition, which ultimately determines the speed of establishing long-term memories. Remember that in incremental reading, the volume of the material may be very high, however, the ultimate number of items entering the learning process, in the ideal case, is relatively small (usu. 10-20 per day, not 200!). It simply takes lots of time to fish for golden knowledge that will bring best value in the long term.

Overloading memory and the role of sleep

The existence of the memory bottleneck is the direct consequence of the overload concern. You can overload your learning process with excess information, however, you are unlikely to "overload your long-term memory". The processes of forgetting and garbage collection executed in sleep have evolved precisely to prevent this problem. However much you try to learn excess facts, forgetting will clean up the excess, and memory optimization in sleep will ensure you develop all necessary "meaningful connections". Naturally, this will happen only if you get all sleep that you need (i.e. avoid using alarm clock, sleeping pills, staying up late, etc.).

For more see: Neural optimization in sleep

Learning vs. problem solving

It is true that solving problems requires high concentration. However, in the ideal world, you should devote separate time slots to (1) learning and (2) problem solving. In Covey's terminology, your learning boosts the Production Capacity, while your problem solving time is your Production time. Naturally, you can marry the two slots when problem solving occurs in conditions of information deficit. Incremental reading is an ideal tool for such situations. You can combine the inflow of new information with creative efforts and problem solving while retaining maximum focus on the problem at hand. This is explained inAdvantages of incremental learning: Creativity. You can optimize the degree of monothematic focus by using various tools of incremental reading, esp. Search&Review as well as branch review.

See also: Incremental problem solving.

Internet addiction

Internet distractions can be focus destroyers when working with SuperMemo. However, this is more a matter of self-discipline than an inherent problem associated with SuperMemo. It is up to you to decide if you wish to stray to Facebook or Twitter. Incremental reading may encourage a degree of straying (e.g. to import supplementary material from Wikipedia and/or dictionaries). The whole concept of the priority queue was developed precisely to counteract the cost of such straying. In incremental reading you stray, import, prioritize and... forget (about the excursion to the web). You may visit 20 pages, but instead of wasting time, you import and prioritize. You are back onto your focused path in minutes. The whole process can be under your rational control and the web may become an ally rather than an enemy. Metaphorically speaking, an Internet junkie is constantly distracted with shiny titbits, while incremental reader focuses on fishing for golden knowledge.

Incremental reading boosts attention

Incremental reading increases attention by letting you focus on a manageable portions of knowledge without feeling overwhelmed, without straying, without getting stuck on harder material, without worrying that you might miss important pieces when speed-reading, etc. Your best way to experience that improvement is to try incremental reading. However, you should know that the effects will not be instant. You will need to invest a lot of time in learning the tools, and then even more time in honing your strategies and learning about your own memory and reasoning. Few incremental readers become truly enthusiastic in their first months of learning.

Article quality matters in learning

In incremental reading, you will quickly develop skills needed to instantly differentiate between high quality articles and articles that are full of fluff and wasteful prose. You will indeed fish for catchy headlines, meaningful sections, minimum off-topic commentary, etc. Article quality will determine your ability to employ speed-reading, and to quickly prioritize your material. This has nothing to do with instant gratification obtained from social media, instant news, and other net distractions.

Can you really read thousands of articles at the same time?

A visitor to supermemo.com commented on incremental learning claims:

A friend recommended incremental reading to me. However, I was instantly turned off by your claim: read thousands of articles at the same time. This is not only impossible, but also sounds like overhyped marketing. All you do in incremental reading is split articles into chunks and read those separately. With this method you lose the big picture. I am afraid you are selling snake oil.

That skepticism is understandable, and yet we stick with the original claim: you can indeed read thousands of articles at the same time. This is because reading can be understood as a process or as anact. No one can make a sane claim of multiple reading acts at the same time. In incremental reading, only one article receives a laser focus at any given moment of time. Few students read more than 50 articles in a single day, and they rarely read them all in their entirety. However, they may easily read or skim a few thousand articles in a month, and keep hundreds of thousands of articles in the incremental learningprocess. In that sense reading thousands of articles at the same time is an accurate if somewhat enticing description.

To understand incremental reading, you need to understand SuperMemo (or repetition spacing in general), as well as the effects of intermittent reading on memory. Reading in chunks without the help of the underlying learning process based on spaced repetition indeed makes little sense. However, once you master and understand the techniques of incremental reading, you will get the effect opposite to the one expected: not only will you not lose the track of the big picture, you will keep a lasting, durable and coherent memory image of whatever you found important in the processed articles.

In a well-managed incremental learning process, the big picture is retained in a state of better coherence than it is the case in traditional learning.

Incremental reading vs. memory interference

You may have read about interference in learning. When students learn two things one after another, they perform worse than in cases where they focus on one thing. This might sound like a reason to disqualify incremental reading as an effective learning method.

It is true that interference can ruin learning. If you read about a subject without fully understanding it and follow it with another subject that is confusingly similar in nature, you will indeed perform worse intests. However, this effect is much less pronounced if the first subject is studied with solid comprehension. Incremental reading makes it possible to read only as much as you understand. Then it encourages long-term retention by producing cloze deletions. Finally, it periodically rediscovers weaknesses in the learning process and fills the gap. When well executed, incremental reading produces an opposite effect. It minimizes interference by forcing you to resolve contradiction in your material. It ruthlessly punishes all cases of incomplete understanding. In classroom conditions, you can get a foggy pass at subject A, then worsen the fog by digging into subject B. In incremental reading, SuperMemo will force you to jump from A to B and back to A until the two form a harmonious body of knowledge withminimum interference and maximum connectivity. Note that the same research on interference produces diametrically different results when the interfering topics are subject to continual re-reading. Re-reading is frequent in SuperMemo and multiple active repetition of cloze deletions is a norm. The outcome of the experiment may also be obscured by adding a degree of novelty to old reviews which greatly improves attention. Better learning follows in the wake.

Why Wikipedia is better than Encyclopedia Britannica?

Wikipedia is better

Britannica is hard

Due to its style, Britannica is often not suitable for incremental reading. It is pleasant to read, it explains things, it digresses, it provides examples, however, it is not fact-packed and context-rich like crowdsourced Wikipedia.

Try working the following text with incremental reading. The text was taken from an article about sleep from Britannica:

In addition to the behavioral and physiological criteria already mentioned, subjective experience (in the case of the self) and verbal reports of such experience (in the case of others) are used at the human level to define sleep. Upon being alerted, one may feel or say, "I was asleep just then," and such judgments ordinarily are accepted as evidence for identifying a prearousal state as sleep. Such subjective evidence, however, can be at variance with both behavioral classifications and sleep electrophysiology, raising interesting questions about how to define the true measure of sleep. Is sleep determined by objective or subjective evidence alone, or is it determined by some combination of the two? And what is the best way to measure such evidence? More generally, problems in defining sleep arise when evidence for one or more of the several criteria of sleep is lacking or when the evidence generated by available criteria is inconsistent. Do all animals sleep?

This piece begins with an ominous context-buster: "already mentioned". It poses interesting questions. However, it will take a while for an incremental reader to find cloze keywords that would answer the most essential question of long term learning: What do I really want to remember for life from the passage I have just read?

Simple English Wikipedia is dangerous

Many users believe that Simple English Wikipedia is better for understanding problems than Wikipedia itself. However, simplified terminology leads to terminological imprecision. Even though the language is simpler, the problems presented may get a superficial or misleading treatment. Simple English Wikipedia might be an advantage if you want a short story of the Chinese Empire as opposed to a lengthy Wikipedia dissertation. However, if you try to understand complex economics or quantum physics, relying on simpler texts may backfire. Very often, you should rather struggle with complex terminology by importing supplementary material. Going the easy way may take you astray.

Incremental reading vs. the books

Incremental reading of electronic materials is superior to reading books or reading in the browser (given the same quality of the study material). All basic reading, bookmarks, and highlight methods start falling apart once the volume of books or articles increases beyond a certain level.

Without SuperMemo, you won't easily prioritize, sort, organize, schedule, re-prioritize, search&review, etc. The whole SuperMemo "engine" in the background is the most important component in the process. For a larger volume of material, when reading in the browser, you will probably be just 5-10% as effective as with incremental reading (in the long run).

In incremental reading you can use the following tools:

Incremental reading is likely to reduce your tolerance for:

Incremental reading will sharpen your skills in:

Incremental learning vs. the news

You will rather not want to use SuperMemo to process news!

Donald Rumsfeld claims to live by the rule: "First reports are always wrong".

News carries knowledge that is usually valid in a short term only. Incremental learning is useful only for long-term memories. The greatest benefits of trying to process news with incremental learning is to realize how transient news is, and to moderate the craving for news by making it "old news" through the incremental approach.

These are the main reasons why news is a poor candidate for your learning material:

News may have more value for you at younger ages when you still need to learn a lot about the world, or when you are trying to learn a new language and want to understand the news broadcast from another country. In such cases, use news mostly as the source of indications of what supplementary material should enter your learning process for you to comprehend the news better.

Unless you study journalism (and need to understand the process), or political sciences (incl. the impact of news on the public), or you just live and die by being up-to-date, you should limit your craving for news and focus on knowledge with long-term applicability.

Example: how breaking news breaks the news

Consider this funny episode from CNN that illustrates the obsession with "breaking news". On a sunny Sunday of September 22, 2013, Fareed Zakaria spoke about Angela's Merkel impact on Germany, and Europe:

Merkel has taken important steps to help Europe's struggling economies, spending tens of billions of dollars directly and indirectly on them. On the other hand, Merkel has imposed austerity on much of Europe, which has been excessive and counterproductive. Her argument is that it is the only way to get governments like Greece and Italy to become more competitive. Now here's the irony...

Now here is a real irony. The intelligent commentary was interrupted by the breaking news related to Angela Merkel herself. The viewer did not get a chance to learn about the historic role of the German leader. Instead he or she got a 2 min. "deal" to watch Angela Merkel vote in a federal election. Three full minutes of relaxing silence and a chance to admire Angela's dress and body language in this historic moment.

Ah yes, if you cannot stand similar interruptions and want to learn about the historic role of Angela, see Zakaria's transcripts at FAREED ZAKARIA GPS: Not the Time for Big Sticks; Interview with Bill Clinton; Will Germany's Merkel Practice What She Preaches?

Minimum information principle

Minimum information principle says that your items should be as simple as possible and always ask a single simple question. Conglomerating information in [[Glossary:Spaced_repetition|spaced repetition] results in slower learning. For example, you may be tempted to formulate items like this:

Question: What was decided at the Council of Trent, beginning in 1545, and how long did the Council go on?
Answer: The basic beliefs of the Catholic Church; 18 years.

Two separate memories should be separated in SuperMemo due to the fact that they nearly always will require different timing of repetitions. If you can always activate the same mental pathway in thinking about the Council of Trent ("neurons firing together" in the same pattern), your particular item has a good chance of surviving long in the process without a memory lapse. However, once you build a large database of similar items, and you review your sizeable material under the pressure of time, your review will always tend to strip redundant pieces of information. Overtime, your nice item will be reduced to the bare bones of information that will often fail its primary test: applicability in real life. It may happen, that despite zero memory lapses, in 2-3 years, someone will ask you about a Council of Trent in a new context and you will be amazed that you won't be able to reasonably answer the question despite having all the necessary pieces of information included in your item. Two memories of different difficulty might be compared to two different planes of different flying characteristics. The difficult piece (e.g. 18 year duration of the Council) might be compared to a slow flying plane. The easy piece (here the reference to the Catholic Church) might be compared to a modern jet. Review of the conglomerated item might be compared to flying both planes at the same speed. In an extreme case, this might be impossible. The compromise speed might be too high for a slow plane, which might disintegrate beyond a certain speed limit, while the faster plane cannot slow down enough without stalling. In our memory, forgetting is equivalent to forgetting, while stalling is caused by the spacing effect. By doing complex and repeatable reasoning at each repetition, you might act as if handling both planes using remote control. However, this is always difficult and requires lots of focus and deliberation at repetitions. Your brain has natural defenses against such "enforced repetitive reasoning". It is designed to be "intellectually lazy" and thus energetically efficient. Practice shows that incremental reading produces many more items. However, those items are usually much easier to remember. In the end, you spend less time on reviewing 5-10 items than you would spend on an item that would conglomerate information and suffer repeated memory lapses or very short intervals.

In the course of the evolution, the brain developed strategies for abstracting away from the details and retaining only the most essential, useful and frequently used information. Those strategies are great forsurvival, but aren't as good in reaching our educational goals. Council of Trent is a typical example of knowledge we wish to have, but that is pretty expensive. This is because, for most people, it does not get reinforced in run-of-the-mill conversations, TV shows, daily applicability, or at water cooler at work. The situation might differ if you, in particular, read a lot on the subject matter. This might help the memory establish itself in an efficient manner. Incremental reading makes it possible to root such difficult-to-retain knowledge firmly in the context, and still make sure that individual repetitions focus on a very specific and cheap-to-retain memories.

This is how the same paragraph might be processed with incremental reading, and paradoxically cause a significant saving in time in the long run:

Question: The Council of [...], which began in 1545 and lasted for 18 years, made decisions about the basic beliefs of the Catholic Church
Answer: Trent
Question: The Council of Trent, which began in [...](year) and lasted for 18 years, made decisions about the basic beliefs of the Catholic Church
Answer: 1545
Question: The Council of Trent, which began in 1545 and lasted for [...] years, made decisions about the basic beliefs of the Catholic Church
Answer: 18
Question: The Council of Trent, which began in 1545 and lasted for 18 years, made decisions about the basic beliefs of [...]
Answer: the Catholic Church
Question: The Council of Trent, which began in 1545 and lasted for 18 years, made decisions about [...]
Answer: (the) beliefs of the Catholic Church
Question: [...], which began in 1545 and lasted for 18 years, made decisions about the basic beliefs of the Catholic Church
Answer: The Council of Trent

In the end, if you are sure this item works for you, check its performance in the course of the next few years. If you pass the interval of 2 years without a lapse, you can say that this particular item indeed works for you. In that case, there is no disagreement between you and the 20 rules. It is just that for most people, this item is pretty likely to generate a lapse within 2 years even if reviewed at correct timing. Depending on the item difficulty, the number of repetitions in the first 2 years might be as low as 3 or well above 20. If your default forgetting index is 10%, this translates to a span from 70% chance of retaining the item to the totally unacceptable 90% chance of forgetting! This last number is little understood and little realized by the users of SuperMemo, and should always make you think a lot about the rules of efficient formulation of knowledge.

Examples: Incremental reading in action

The following examples have been collected from various articles published at supermemo.com. They illustrate how electronic texts can be converted to cloze deletions, and how formulation problems are solved at the level of articles, extracts and cloze deletions.

Example: Cloze deletion

Passively processed ideas in the form of sentences rarely leave a durable trace in your memory even if they are reviewed regularly. Very often, as soon as after 2-3 months, you will notice that at review time, you actually do not seem able to recall that you have ever had a given sentence in your collection. You will quickly discover that you need active recall in order to remember. Active recall is a process in which you must answer questions. For example, you may be presented with a picture of Charles Darwin and be asked to recognize his face. In the long run, you need to replace passive review with active recall. Otherwise, your memory of the fact will not be permanently consolidated.

The fastest way of converting simple sentences into active recall material is to use a cloze deletion. Using cloze deletion, you work with simple declarative sentences like:

WW1 was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in 1914

Those sentences are converted into question-answer pairs that can be used in actively stimulating your memory for best recall:

Question: WW1 was precipitated by the assassination of Archduke Francis Ferdinand of [...](country/empire) by a Serbian nationalist in 1914
Answer: Austria-Hungary
Question: WW1 was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in [...](year)
Answer: 1914
Question: [...](war) was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in 1914
Answer: WW1

Example: Extracts and deletions

Example text (submitted by a student)

After the discovery of Pluto, it was quickly determined that Pluto was too small to account for the discrepancies in the orbits of the other planets. The search for Planet X continued but nothing was found. Nor is it likely that it ever will be: the discrepancies vanish if the mass of Neptune determined from the Voyager 2 encounter with Neptune is used

Example processing

There are many ways in which the processing of texts can be done in incremental reading. The example below is just one of the ways.

Extract 1 and 4 clozes: Pluto and orbit discrepancies
Pluto is too small to account for the discrepancies in the orbits of the other planets
  1. Question: [...](planet) is too small to account for the discrepancies in the orbits of the other planets
    Answer: Pluto
  2. Question: Pluto is too [...] to account for the discrepancies in the orbits of the other planets
    Answer: small
  3. Question: Pluto is too small to account for the [...] of the other planets
    Answer: discrepancies in the orbits
  4. Question: Pluto is too small to account for the discrepancies in the orbits of [...]
    Answer: planets
Extract 2 and 1 cloze: Planet X
Pluto was too small to account for the discrepancies in the orbits of the other planets. The search for Planet X continued but nothing was found
  1. Question: Pluto was too small. The search for Planet X continued and [...] was found
    Answer: nothing was
Extract 3 and 5 clozes: Voyager and Neptune
Pluto was too small to account for the discrepancies in the orbits of the other planets. The discrepancies vanish if the mass of Neptune determined from the Voyager 2 encounter with Neptune is used
  1. Question: Pluto was too small. The discrepancies [...] if the mass of Neptune determined from the Voyager 2 encounter is used
    Answer: vanish
  2. Question: Pluto was too small. The discrepancies vanish if the new [...] of Neptune is used
    Answer: mass
  3. Question: Pluto was too small. The discrepancies vanish if the mass of [...] determined by the Voyager 2 is used
    Answer: Neptune
  4. Question: Pluto was too small. The discrepancies vanish if the mass of Neptune determined from the [...] encounter with Neptune is used
    Answer: Voyager 2
  5. Question: Pluto was too small. The discrepancies vanish if the mass determined from the Voyager 2 encounter with [...] is used
    Answer: Neptune

Example: Processing an article

Reading the article

Let us have a look at an example of a very short, self-containing article, posted in April 1999 on the CNN website. This short article can be read in minutes and can serve as a positive incentive towards adjustments in your diet.

Antioxidants may slow aging process, study says April 5, 1999 Web posted at: 9:39 p.m. EDT (0139 GMT) From Correspondent Linda Ciampa -- Research at Tufts University indicates that a healthy diet fortified with certain fruits and vegetables may slow down and even reverse the aging process. Foods rich in antioxidants -- such as blueberries, strawberries, spinach and broccoli -- have what doctors call high ORAC (Oxygen Radical Absorption Capacity) levels. Middle-aged rats who were fed a high ORAC diet in the USDA-sponsored study experienced less memory loss than those given a normal diet. Some of the older, slower rats became as lively as their younger peers after taking antioxidants. "We prevented both some brain and some behavioral changes that one normally sees in these rats when they hit 15 months of age," said USDA researcher Jim Joseph. Antioxidants are effective in destroying free radicals -- cell-damaging compounds that can help cause cancer and heart disease and speed the aging process. "It's pretty well accepted that aging is due to the production of free radicals. So anything we can do nutritionally to provide additional antioxidants is likely to protect us in the process of aging," Joseph said. That fact already has prompted many to eat a diet rich in antioxidants. "I look at it as sort of a savings account. I'm benefiting today from eating right, but I'm also going to have it in the future," said 30-year-old Cori Alcock. "As I age and grow older, I'll have good health as well."

#Title: Antioxidants may slow aging process
#Author: Linda Ciampa
#Date: April 5, 1999
#Source: CNN

Extracting the essence from the article

In the course of reading, you should select the most important sections of the article. The article introduces some facts related to healthy diet and adds a lot of redundant explanations. For your review, you are only likely to need the core message which usually makes up a fraction of the entire text. Please have a look again at the same text with four most critical sections emphasized (numbering is not needed and is used only for your convenience for further reference):

Research indicates that a healthy (1) diet fortified with certain fruits and vegetables may slow down and even reverse the aging process. (2) Foods rich in antioxidants -- such as blueberries, strawberries, spinach and broccoli -- have high ORAC (Oxygen Radical Absorption Capacity) levels. Middle-aged rats who were fed a high ORAC diet in the USDA-sponsored study experienced less memory loss than those given a normal diet. Some of the older, slower rats became as lively as their younger peers after taking antioxidants. "We prevented both some brain and some behavioral changes that one normally sees in these rats when they hit 15 months of age," said USDA researcher Jim Joseph. (3) Antioxidants are effective in destroying free radicals -- cell-damagingcompounds that can help cause cancer and heart disease and speed the aging process. (4) "It's well accepted that aging is due to the production of free radicals. So anything we can do nutritionally to provide additional antioxidants is likely to protect us in the process of aging," Joseph said. That fact already has prompted many to eat a diet rich in antioxidants. "I look at it as sort of a savings account. I'm benefiting today from eating right, but I'm also going to have it in the future," said 30-year-old Cori Alcock

Improving the wording of extracts

Once you extract important fragments from an article, you may need to reformulate individual fragments to make sure they are fully context independent, free of redundant information, easy to read, and formulated in such a way that the beginning of the fragment serves as the introduction to the latter phrases and not vice versa. Please have a look at the example from the healthy diet article. We selected four important fragments and these fragments (presented on the left in the table) were reformulated to become fully-independent pieces of information (on the right). Please note that two fragments have generated more than one reworded fragment and that one fragment was deleted as it appeared to be redundant upon closer analysis.

The original fragment pasted without change from the CNN articleModified fragment: shorter and easier to read (sometimes split into more than one part)
Extract 1
  1. diet fortified with certain fruits and vegetables may slow down and even reverse the aging process
  1. Diet of fruits and vegetables may reverse aging
Extract 2
  1. Foods rich in antioxidants -- such as blueberries, strawberries, spinach and broccoli -- have high ORAC (Oxygen Radical Absorption Capacity) levels
  1. Examples of foods rich in antioxidants: blueberries, strawberries, spinach and broccoli
  1. Foods rich in antioxidants have high levels of Oxygen Radical Absorption Capacity (ORAC)
Extract 3
  1. Antioxidants are effective in destroying free radicals -- cell-damaging compounds that can help cause cancer and heart disease and speed the aging process
  1. Antioxidants destroy free radicals
  1. Free radicals are cell-damaging compounds that cause cancer, heart disease and aging
Extract 4
  1. It's well accepted that aging is due to the production of free radicals
After a closer scrutiny, the fragment on the left seems to be redundant when compared with the one listed above. We can delete it from the set

Generating cloze deletions

We can now convert items generated earlier into active recall items based on cloze deletion:

Original extractCloze deletions generated from the extract
Extract a
  1. Diet of fruits and vegetables may reverse aging
  1. Question: Diet of [...] and vegetables may reverse aging
    Answer: fruits
  1. Question: Diet of fruits and [...] may reverse aging
    Answer: vegetables
  1. Question: Diet of fruits and vegetables may reverse [...]
    Answer: aging
Extract b
  1. Examples of foods rich in antioxidants: blueberries, strawberries, spinach and broccoli
  1. Question: Examples of [...] rich in antioxidants: blueberries, strawberries, spinach and broccoli
    Answer: foods
  1. Question: Examples of foods rich in [...]: blueberries, strawberries, spinach and broccoli
    Answer: antioxidants
  1. Question: Examples of foods rich in antioxidants: [...]
    Answer: blueberries, strawberries, spinach and broccoli
Extract c
  1. Foods rich in antioxidants have high levels of Oxygen Radical Absorption Capacity (ORAC)
  1. Question: [...] rich in antioxidants have high levels of Oxygen Radical Absorption Capacity (ORAC)
    Answer: Foods
  1. Question: Foods (rich/poor) in antioxidants have high levels of Oxygen Radical Absorption Capacity (ORAC)
    Answer: rich
  1. Question: Foods rich in [...] have high levels of Oxygen Radical Absorption Capacity (ORAC)
    Answer: antioxidants
  1. Question: Foods rich in antioxidants have (high/low) levels of Oxygen Radical Absorption Capacity (ORAC)
    Answer: high
  1. Question: Foods rich in antioxidants have high [...] of Oxygen Radical Absorption Capacity (ORAC)
    Answer: levels
  1. Question: Foods rich in antioxidants have high levels of [...] (ORAC)
    Answer: Oxygen Radical Absorption Capacity
  1. Question: Foods rich in antioxidants have high levels of Oxygen Radical Absorption Capacity ([...](abbreviation))
    Answer: ORAC
Extract d
  1. Antioxidants destroy free radicals
  1. Question: [...](food component) destroy free radicals
    Answer: Antioxidants
  1. Question: Antioxidants (destroy/create) free radicals
    Answer: destroy
  1. Question: Antioxidants destroy [...](harmful compounds)
    Answer: free radicals
Extract e
  1. Free radicals are cell-damaging compounds that cause cancer, heart disease and aging
  1. [...] are cell-damaging compounds that cause cancer, heart disease and aging
    Answer: Free radicals
  1. Free radicals are [...]-damaging compounds that cause cancer, heart disease and aging
    Answer: cell
  1. Free radicals are cell-(building/damaging) compounds that cause cancer, heart disease and aging
    Answer: damaging
  1. Free radicals are cell-damaging compounds that cause [...](health problems)
    Answer: cancer, heart disease and aging

Although we have generated 20 cloze deletions from the original 5 extracts, it is important to stress that reviewing this much of the learning material will ultimately cost you less time and the memory effect will be better! Note that cloze deletions meticulously test your knowledge of all important semantic aspects of the learned article.

Converting cloze deletions to plain questions

After you extract fragments and formulate active recall questions, you should continue to constantly reevaluate the importance of individual pieces of information, their wording, delete less important pieces and move them for later review, etc. Examples of reformulated cloze deletions can be found below. Note that Clozes 4-6 and Cloze 20 have been split further to eliminate set enumeration (it is easier to independently associate cancer or aging with free radicals than to list all health problems caused by them)

Original cloze deletionReformulated active recall item
Clozes 4-6
  1. Question: Examples of [...] rich in antioxidants: blueberries, strawberries, spinach and broccoli
    Answer: foods
  2. Question: Examples of foods rich in [...]: blueberries, strawberries, spinach and broccoli
    Answer: antioxidants
  3. Question: Examples of foods rich in antioxidants: [...]
    Answer: blueberries, strawberries, spinach and broccoli
  • Question: Are blueberries rich in antioxidants?
    Answer: yes
  • Question: Are strawberries rich in antioxidants?
    Answer: yes
  • Question: Is spinach rich in antioxidants?
    Answer: yes
  • Question: Is broccoli rich in antioxidants?
    Answer: yes
Cloze 12
  1. Question: Foods rich in antioxidants have high levels of [...] (ORAC)
    Answer: Oxygen Radical Absorption Capacity
  • Question: What does ORAC stand for?
    Answer: Oxygen Radical Absorption Capacity
Cloze 17
  1. Question: [...] are cell-damaging compounds that cause cancer, heart disease and aging
    Answer: Free radicals
  • Question: What is the name of cell-damaging compounds that cause cancer? (cleaned off with antioxidants)
    Answer: Free radicals
  • Question: What are free radicals?
    Answer: cell-damaging compounds (causing cancer, heart disease and aging)
Cloze 20
  1. Question: Free radicals are cell-damaging compounds that cause [...](health problems)
    Answer: cancer, heart disease and aging
  • Question: Do free radicals cause cancer?
    Answer: yes
  • Question: Do free radicals contribute to heart disease?
    Answer: yes
  • Question: Do free radicals cause aging?
    Answer: yes

Example: Rewording texts

Wordy articles may require some rewording of sentences before cloze deletions can be generated.

For example, the following texts puzzled a user as it appeared to be hard to process incrementally:

In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease, even after being passed through a porcelain filter known to retain all bacteria, contained an agent that could infect other tobacco plants.

In 1900 a similarly filterable agent was reported for foot-and-mouth disease of cattle.

Before you begin learning, you can save lots of time by looking for articles that are properly structured and written in a concise language that will help you save lots of time. For example, Wikipedia is an excellent source. As it is edited by many people in an incremental manner, it is highly context-independent. In comparison, Britannica is wordy, full of pronouns, definite clauses, and various context references.

Where Britannica might say (fictitious example): "Over the next five years, he struggled to obtain a patent for his invention", Wikipedia might say explicitly "In the years 1883-1889, Edison struggled to obtain a patent for a filament of carbon of high resistance". This context-independent style can save you hours of parsing and re-editing.

In the presented example, the first sentence is causing trouble because the author tried to tell you far more than you might wish to process in one go.

One strategy is to start with monster clozes, and simplify them incrementally while learning. However, you could save lots of time with another strategy, in which you split the sentences into more manageable portions. Unfortunately, in this case, some editing will be necessary in the beginning. You will also need to carefully parse the meaning of the passage. You could, for example, separate who and what components of the sentence

who: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller than bacteria.

what: In 1892, Iwanowski showed that the sap from tobacco infected with mosaic disease even after being passed through a porcelain filter known to retain all bacteria, contained an agent that could infect other tobacco plants.

From those two mini-topics, you can generate several clozes that will cover the essence of the passage:

Question: In [...](year) the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller than bacteria
Answer: 1892
Question: In 1892 the [...](nationality) botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller than bacteria
Answer: Russian
Question: In 1892 the Russian [...](specialty) Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller thanbacteria
Answer: botanist
Question: In 1892 the Russian botanist [...](name) showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller than bacteria
Answer: Dimitri Iwanowski
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that [...](what?) from tobacco plants infected with mosaic disease contained an infectious agent smaller than bacteria
Answer: sap
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from [...](type) plants infected with mosaic disease contained an infectious agent smaller than bacteria
Answer: tobacco
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with [...](disease) contained an infectious agent smaller than bacteria
Answer: mosaic disease
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained [...] smaller than bacteria
Answer: an infectious agent
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent smaller than [...]
Answer: bacteria
Question: In 1892 the Russian botanist Dimitri Iwanowski showed that the sap from tobacco plants infected with mosaic disease contained an infectious agent [...] than bacteria
Answer: smaller
Question: In 1892, Iwanowski showed that the sap from tobacco infected with mosaic disease even after being passed through a [...](type) filter, contained an agent that could infect other tobacco plants
Answer: porcelain
Question: In 1892, Iwanowski showed that the sap from tobacco infected with mosaic disease even after being passed through a porcelain filter known to [...](property), contained an agent that could infect other tobacco plants
Answer: retain all bacteria
Question: In 1892, Iwanowski showed that the sap from tobacco infected with mosaic disease even after being passed through a porcelain filter, contained an agent that [...](property)
Answer: could infect other tobacco plants

The above questions are only a rough beginning. Only during learning will you be able to identify holes in these items. You will see where they cause trouble, why they may be hard to remember or what questions are imprecise or confusing. You will fix those deficiencies incrementally while learning.

Example: Conglomerating information

A frequent sin committed by new users of SuperMemo is to create "monster items" that include many bits of information. Those items should tackled by treating all bits individually.

Example of a monster item

Question: a rod-and-tube element temperature sensor consists of:
Answer: a high expansion metal tube containing a low expansion rod. The rod& tube are attached on one end. The tube changes length with changes in temperature, causing the free end of the rod to move

Suggestions

This is a typical case of combining a number of items in one with a detriment to the ability to recall the combined item. The suggestion here is to split the item into a number of simpler items that reproduce the same information in student's memory:

Question: What are the two parts of a rod-and-tube temperature sensor?
Answer: rod and tube
Question: What is the expandability of the tube in rod-and-tube sensor?
Answer: high
Question: What is the expandability of the rod in rod-and-tube sensor?
Answer: low
Question: How is temperature indicated in the rod-and-tube sensor?
Answer: tube moves relative to the rod
Question: Where are rod and tube connected?
Answer: On one end
etc. etc.

Example: Evolution of knowledge

Changes to individual pieces of knowledge will take place in steps upon successive reviews. Here are exemplary steps that show a complete evolution of a single article into a finished item based on active recall:

  1. Imagine that you find an article on the net, e.g. The criticism of global capitalism, and you decide to read it and remember it for ever
  2. You import the article to SuperMemo
  3. You read the article (e.g. once its turn comes up in incremental reading)
  4. While reading, you extract most important paragraphs. One of these, let us say, refers to Kuznets hypothesis
  5. The extracted paragraphs will live separate lives in SuperMemo and will be scheduled for separate review, i.e. independent of the review of the parent article. The extracted paragraphs in the parent article will be marked as processed. Once all paragraphs in the parent article are processed, you will terminate the review of the parent article and keep on reviewing only its components (e.g. selected paragraphs)
  6. Upon the first review, usually after a few days, you read the extracted paragraph again and analyze it as to how it should be processed further. You may decide to postpone it, remove it from the learning process, shorten it or extract the most important sentences that you want to remember
  7. If you decide to extract a single statement in reference to Kuznets hypothesis it will again be marked as processed in the original extract and will assume a separate review cycle in SuperMemo
  8. Upon the first review of the extracted sentence, you make further decisions as to its further life in SuperMemo. Let us say, this is the wording of the Kuznets sentence:
    Acc to Kuznets hypothesis, growth (from the low income levels associated with predominantly agrarian societies) would first lead to an increase, and then to a decrease in income inequality
  9. In order to capture the essence, you would probably decide to shorten the above sentence to the following form:
    Acc to Kuznets hypothesis, growth would first lead to an increase, and then to a decrease in income inequality
  10. At the same time, other parts of the same parent article might establish a memory trace that would say that Kuznets hypothesis has been based on relatively weak empirical data. Moreover, recent research clearly indicates that the hypothesis is false (growth actually seems to equally benefit both the poor and the rich). You could then enhance the extract with words controversial or even recently falsified. For example:
    Recently falsified Kuznets hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality
  11. Upon the next review of the same sentence, you may decide to convert it into a number of cloze deletions. This conversion will be incremental, i.e. you may decide to first create a cloze deletion asking about the name of the controversial hypothesis and only later ask about its actual meaning (the meaning is relatively easier to remember and shall survive longer in your memory without active recall). Your cloze deletion could then look like this:
    Question: Recently falsified [...](name) hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality
    Answer: Kuznets

    This cloze deletion would again assume a separate life from the original sentence in which the keyword Kuznets will again be marked as processed. This is the original Kuznets sentence with one keyword marked as processed:

    Recently falsified Kuznets hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality
  12. The same sentence will generate a few separate cloze deletions that will be processed independently. Upon the first review of the cloze deletion created in the previous point, you may decide to simplify it in accordance with the rules of formulating knowledge in learning:
    Question: Recently falsified [...](name) hypothesis claimed that growth would first lead to an increase in income inequality
    Answer: Kuznets
  13. Upon the next review, you can, but you do not have to, convert the cloze deletion into a standard question-answer item:
    Question: What is the name of the hypothesis that falsely claims that income inequality initially increases with growth?
    Answer: Kuznets hypothesis
  14. The above question-answer pair is probably as simple as it can only be. Certainly, it is simple enough to be relatively easy to remember. This item will be repeated in intervals determined by SuperMemo. You can decide how well you want to remember it. By default, it will be remembered with 95% probability of recall and require 5-15 repetitions in lifetime. The establishment of durable memory traces in your memory, completes the life cycle of this particular piece of knowledge. The only thing that remains is the memory-sustaining review in intervals ranging from months to years (as determined by SuperMemo)
  15. Once you convert all important keywords from the Kuznets hypothesis into separate cloze deletions, you will remove the parenting paragraph from the review process. You will no longer passively review the original declarative hypothesis. You will continue repeating individual clozes and that will ensure your perfect recall of the hypothesis for as long as you deem necessary

Example: Building comprehension incrementally

It is not unusual to generate a cloze that will keep causing problems. It is not the subject in question, but the complexity of the sentence that seems to send the brain into a panic mode. You may keep re-reading bad clozes and get the impression that at each re-reading the understanding decreases (instead of growing).

Here is an example of a bad cloze taken from a real learning process. It caused 5 lapses in succession, and needed special treatment to make it palatable and useful.

Question: Companies or other groups are issued emission permits and are required to hold an equivalent number of [...] (or credits) which represent the right to emit a specificamount
Answer: allowances

With some experience, you will quickly notice that the problem resides in the poor quality of the chosen text. Your first red flag should come with "companies or other groups". Unless you are an insider to the subject, you will instantly wonder what "other groups" means. If you look at the essence of your question, you will notice that "other groups" does not add to the core of the question and can easily be skipped. After some analysis, you can make an effort to formulate a straightforward question, however, it is easier and more cost-effective to take an incremental approach (as long as you keep understanding the question, which may not be the case with you).

As soon as you notice that "other groups" are out of place, you can simplify the cloze:

Question: Companies are issued emission permits and are required to hold an equivalent number of [...] (or credits) which represent the right to emit a specific amount
Answer: allowances

You may also notice that allowances and credits are synonymous (acc. to that particular question), and you may not need to tax your memory with both terms. You can therefore make the question easier:

Question: Companies are issued emission permits and are required to hold an equivalent number of [...] which represent the right to emit a specific amount
Answer: allowances/credits

During the next repetition you may wish to simplify the question further. You can now start feeling confident that the question will stick to your memory. If that makes you feel better, you will have made a big step towards better recall:</p>

Question: Companies hold a number of [...] which represent the right to emit a specific amount
Answer: allowances/credits (emission permits)

Some time in the future, you might take yet another incremental step:

Question: [...] represent the right to emit a specific amount (by companies)
Answer: allowances/credits (emission permits)

Finally, you may realize that you question is actually a definition of a term (credit/allowance). As such, it does not even need the cloze. You may also notice that the definition misses an important context point. The original question spoke of carbon emissions (which is probably indicated by the Wikipedia references that provide the context of the question).

Question: (What is the name of) the right to emit a specific amount of CO2 (by companies)?
Answer: (carbon) permits/allowances/credits

With the above definition, learning should finally be easy and fun. Remember that it is a frequently used standard to separate synonyms or equivalent answers with a slash ("/"). This means that any answer will do permits, allowances, or credits. You do not need to list them all and you do not need to remember about semi-obvious carbon prefix.

Interestingly, it can easily be found that the text comes from May 7, 2010 version of Wikipedia. It was red-flagged with "clarification needed", which instantly tells you that you are not the only one who got an issue with the paragraph. In such cases blame the author and search for better texts. In this particular example, the Wikipedia text was improved just 2 weeks later, and the current version of Wikipedia holds a better version (Aug 2013).

Example: Unsuitable texts

Not all texts are suitable or easy to process with incremental reading. You will not want to process a literary novel with incremental reading. You may still prefer to read it on paper in a bathtub. Examples of texts that are difficult to process are: flowery materials, materials rich in explanations and metaphors, programming code, case studies, mathematical derivations, experimental research documentation, etc. Incremental reading is easiest for encyclopedic materials. Materials that are not suitable will often include a valuable message; however, you will often be better off by phrasing it on your own and processing your summary with incremental reading. For example, you would not want to memorize the Linux source code. However, you could find some specific facts or regularities in the code, describe them shortly and then learn the description incrementally (perhaps with snippet code illustrations).

Example 1: Texts that are too general

Here is an example paragraph that caused learning problem to a user of SuperMemo:

A good enterprise architect should enable the right balance between the needs of the organization for an integrated IT strategy, permitting the closest possible synergy across the extended enterprise, and allowing individual business units to innovate safely in their pursuit of competitive advantage.

The user wrote:

I have no problem understanding this phrase (synergy across business units must be balanced against freedom to innovate within business units), and I constructed two cloze deletions from it for either side of the balance, and when presented with either question I can fill in the blanks. Yet, I was in a discussion recently defending a synergy position, not realizing that it might jeopardize innovation, not even realizing I had this SuperMemo question pointing it out to me. In other words, even though I can answer the question in SuperMemo, it is not something that stuck in my memory, i.e. the synergy is not associated with freedom-to-innovate and vice versa (maybe it will after this mail). So, I suspect that I am only able to answer the question based on recognition of the question or some such, and not on recognition of the association. I figure that a better way to associate the two would be to ask something like "what must be balanced against each other" but this question would be so general in nature that it would create serious interference with other questions that deal with other aspects of enterprise architecture that need to be balanced against each other. Or I would have to make it more specific, again risking to give away the answer in the question, which would also not cause the association to form.

Can you identify a problem here?

In similar cases, you need to pause to ponder what kind of questions you want to be able to answer having read the passage. If the questions are too general or too obvious, you just need to trust your own intelligence and creativity to be able to answer them on the basis of your experience and more specific questions in the given area of knowledge.

In this case, keywords such as "synergy" or "innovate" might provide a hazy way to capture the meaning of the passage. However, very general texts are not suitable for treatment with cloze deletions. You may waste unnecessary time on re-reading the entire passage at question time, or waste time on simplifying the passage to capture the essence. In school jargon, you might call similar passages "waffle". They may carry an important message, they may help the flow of text, they might be explanatory, but they do not yield material suitable for memorization. In the extreme case you can juxtapose Wikipedia-like "IT = Information Technology" with obvious "waffle" that cannot be clozed: We should be nice for other people.

In all sorts of exams, you will always need to tackle lots of "waffle". You will also meet teachers who demand fluent "waffle" performance. However, this is not the type of knowledge that will make you a better expert or a better person. If you meet "waffle", pause to think if there are questions that truly flow from the text, or if the text is too general to be handled with SuperMemo. In your case, you might do better by perhaps adding some meatier passages on enterprise synergy or constraints on innovation or... Actually, you are the best person to find supplementary material that will help you better understand the underlying issues.

If waffle bothers you, try to find a Wikipedia equivalent. Due to the nature of crowdsourcing, Wikipedia lends itself perfectly to incremental processing. Once you get the hang and feel the benefit, you will quickly learn to spot text and passages that are less suitable and provide less benefit when processed with incremental reading.

Example 2: Unsuitable text from supermemo.com

Here is another example that comes from supermemo.com. The text itself is not bad, but it resorts to metaphors that should serve as explanations, not as learning material suitable for generating cloze deletions.

Intelligence as processing power: the raw nimbleness and agility of the human mind. When you see a smart student quickly learn new things, think logically, solve puzzles and show uncanny wit, you may say: This guy is really intelligent! See how fast his brain reacts! The student has a fast processor installed and his RAM has a lightning access time. He may though still need a couple of years to "build" good software through years of study. IQ tests attempt to measure this sort of intelligence in abstraction of knowledge. The difficulty of improving processing power by training comes for similar reasons as the fact that programming cannot speed up the processor

The above text is metaphorical. It reiterates the same message a few times using different words in an attempt to find a metaphor that will strike a cord with the reader. Consequently, it is enough you extract only the core message. For example:

Intelligence as processing power: IQ tests attempt to measure this sort of intelligence in abstraction of knowledge

You could also add:

Intelligence as processing power: The difficulty of improving processing power by training is similar to the fact that programming cannot speed up the processor

Once you learn the above 6 cloze deletions (marked by 6 clozed keywords), you will most likely be able to recall that it should be very difficult to train for an improved score in an ideally designed IQ test.

What do people say about incremental reading?

Incremental learning has its followers and its hardline critics. The following two sections present both sides of the story.

Incremental reading in the eyes of its users

Here are a few excerpts from blogs that speak of incremental reading.

Dealing with information overflow

At bobo23.net you learn the following:

You want to read your favorite blogs, you get e-mail newsletters every day, you have websites you check regularly, newsgroups, mailing lists, forums, interesting Wikipedia articles – a lot of digital input you want to keep up with. But unless you make reading on the computer your full time job – you can’t. So how to select the really important stuff out of it?

[...]

This is where the concept of Incremental Reading comes into place. Sometimes you only want to read about a specific topic, sometimes you just want to read a bit of a complicated article or just read about anything randomly to build new connections / enhance creativity. You can all do that with incremental reading and do not have to worry to miss something. Sooner or later (you can influence that) it will appear in your incremental reading process.

[...]

You collect all the information you want to process and store them in one place. Then you review all the articles (or any other kind of information) randomly or by category. You can highlight important parts, set a reading point (bookmark), extract fragments and generate Question-Answer items for later repetitions.

To read the entire entry, visit bobo23.net: Incremental Reading - Dealing with Information Overflow

Len Budney on incremental reading

Len Budney wrote the following:

Incremental reading is another amazing innovation of Supermemo. I haven't seen or heard about such an idea anywhere else!

[...]

You are gradually digesting the article into smaller and smaller pieces.

When those pieces have become truly bite-size, you can convert them to real flash-cards. With a single click you can create "fill in the blank" type flash-cards, or you can create a normal question-and-answer flash card. When the nugget of information is converted to a flash card, you can "Forget" that piece of text: you've digested it all the way down into facts that you won't be forgetting!

I use the incremental reading feature for reading long, complicated documents, and it's world's better than any other method I've tried. (And with a PhD in Mathematics, I've certainly studied my share of documents!) It forces you to read the important bits very carefully--but still saves you time by helping you to ignore the less important bits. I can't recommend it highly enough. In fact, I'd give my left arm to get a Palm application which supports incremental reading!

Visit Len Budney's blog to read the article: Incremental reading

How to supercharge your language learning

In his blog tmwbuckley writes:

Incremental reading is a fairly misunderstood feature of SuperMemo. Incremental reading entails importing your reading material into a flashcard. This can be a huge Wikipedia article for instance, or an entire eBook. Whilst reading, if you encounter a word you’re unfamiliar with, you can highlight the sentence or paragraph it sits in and make it into an entirely new flashcard automatically, ready for you to review later on. By highlighting, pressing Alt+X to extract the segment – you will see the section become blue, meaning this part has been extracted to another flashcard. [...] Incremental refers to reading the content in small chunks over time, until you’ve read all the material and extracted the information you want from it. You can mark where you’ve reached in the reading material up to using the ‘set the read point’ tool on the Read toolbar, so next time you see the article/chapter/whatever, SuperMemo will display it visibly instead of you scrolling down. [...] Once you’ve finished reading the flashcard, and extracted all of the juicy goodness from it, you can delete it using Shift-Ctrl-Enter. This command is different to pure delete, because it will retain all of the flashcards you created from it, whereas delete will cause you to lose these. Which would be a pain in the derriere. SuperMemo, using it’s super-duper algorithm, will schedule the review of these items for you. [...] Incremental reading represents a powerful chimera of extensive and intensive reading. And this can be used with anything – news stories, blog posts, Wikipedia articles.

For the entire article, see: Incremental Reading – How to supercharge your language learning

Taking note: Incremental Reading

In a blog devoted to note taking, MK is skeptical when writing about incremental reading.

MK introduces the process with these words:

The student extracts the most important fragments of individual articles for further review. Extracted fragments are then converted into questions and answers. These in turn become subject to systematic review and repetition that maximizes the long-term recall of the processed texts. The review process is handled by the proven repetition spacing algorithm known as the SuperMemo method.

MK then voices the reservations:

The basic idea is that you "break down" a larger document into smaller chunks in order to learn them by heart. Nothing wrong with that, but this method will not allow you to read thousands of articles at the same time. Nor will it necessarily lead you to understand the big picture.

Here is the blog entry from MK: Incremental reading

Big picture

The issue of the big picture has been raised over and over by observers who are not intimate to the incremental learning process. See: Big picture in incremental reading

Reading thousands of articles at the same time

As for the phrase "reading thousands of articles at the same time", we will continue using that catchy phrase. It should be pretty obvious that we do not mean the same "moment of time" but the same "period of time". We do not envisage a user with a thousand of monitors, nor a user photoreading thousands of articles flashing in front of his eyes in a second. We, naturally, mean having thousands of articles in the learning process, which, in heavily overloaded collection, may mean that an article will never actually gets read due to scoring too low in priority. For more see: Incremental reading marketed dishonestly

The way of the ronin: About incremental reading

The way of the ronin blog says:

What have I been doing instead of updating this blog? Mainly modifying and improving the system which I use for studying and starting a new project based in the incremental reading technique.

[...]

What I want to speak about it’s my user’s experience after using Anki for a whole year in a regular way and SuperMemo for half year (still doing it) also in a regular way.

[...]

SuperMemo has a very complex interface which most of the times makes potential users think twice before buying it.

[...]

After one week or two of using it I got used to it’s interface. In fact, I realized It was quite intuitive.

[...]

What about schools then? Don’t they use the same methods over and over again for years and decades? If all this was true wouldn’t it mean that they should adapt their system to each of their students -which of course they don’t-?

Read more in Anki vs. SuperMemo

How to incrementally read anything

LittleFish in his blog writes:

When I first started using Supermemo, it appeared that Incremental Reading was a valuable feature, but I never really used it. Once I began experimenting with Incremental Reading (Successfully) and adopting the Incremental Reading mindset when I look at learning material, I am now fully convinced of its superiority for nearly all of my intellectual needs. It is light years ahead of its time, and like the core concepts that power Supermemo, eventually it will likely be utilized on almost every level of the educational system (However many years that takes).

[...]

Visit the blog: Supermemo Adventures: How to Incrementally Read Anything

Luis Gustavo Neves da Silva on memorization

Luis Gustavo writes about cramming and wisdom:

'When I learned how to use SuperMemo, one of my first reactions was to show it to everybody around. I was very enthusiastic about the program and its underlying principles. However, I was very disappointed to find out how many people would say: "This is a cramming tool", "This is just memorizing", "This has nothing to do with reasoning or intelligence".

[...]

I believe that in our culture we have even developed a sort of fear of memory as a human weakness that would allow us to be manipulated, and even dehumanized. Just take a look at our pop culture. How many movie protagonists' memories are being erased, changed or manipulated.

[...]

Teachers seems to see no limit in piling up new material in front of the kids and if someone's memory fails ... its all his or her own fault or negligence.

[...]

When asked to explain the reason for believing that memorizing is harmful, the supporters of "reasoning" would come up with a confusing and conflicting mix of words such as memorizing, data, information, knowledge and even wisdom. Looking up these words in a dictionary, you will find a lot of circular references, each word being related to or defined by the use of the others.

Read more: It's more than just memorizing

Incremental reading diagram

bababan.net writes:

I have produced this simple diagram of the incremental reading process. The explanation on the SuperMemo web was too long and too scientific for me, I needed something simple & handy. Hopefully it will also help you to jumpstart into incremental reading with SuperMemo. It only includes the core functionality, for all the bells and whistles, more explanation

Note that the diagram uses old icons, old terminology and old keywords.

Criticism of incremental reading

Incremental reading is useless

Critics of incremental reading are not too numerous or vocal. The credit probably goes to the fact that this technology is relatively new and not too popular beyond SuperMemo insider circles. Nearly all criticism comes from those who never used incremental reading themselves. As such, it hardly ever addresses the substance of the technology. Some of the criticism seem to address the technology in general (e.g. the Internet, the web, computer software, etc).

For example, Emil K. wrote the following criticism of incremental reading:

  1. Information on the internet is often unreliable
  2. It will take a lot of time to find a 100% reliable information on the web
  3. The slow internet connection augments the amount of squandered time dramatically
  4. Software maintenance wastes more time
  5. It can be used only when there is a computer available - mostly at home.

I believe that my method is better than what you propose:

  1. Read a book that has been proven excellent by time and by ratings of others who have spent THEIR time exploring the book.
  2. Instead of some software and internet, use a pen and a notebook to scan the book for important information and copy down for later examination/memorization.

Also, copy down the source and the date so that when creativity comes, the sources are reliable - it will be a disaster to waste plethora of time on a research only to find out in the end that the sources did not have any statistical significance. Magazines like Nature and Science already publish the worthiest of all research articles. Books by Feynman, Borh, Dirac, and Einstein are already the best in the field. Thus, my letter is a cogent polemic that refutes SuperMemo.

Observer's Opinion: Why I would not use incremental learning?

Here is an exemplary opinion of an observer who is focused on problem solving and short-term goals. For this observer, lifelong learning is understood as "I learn what I need right now". Naturally, this approach is missing an important factor: learning in anticipation.

Massive learning

I do not care about massive learning. I solve problems. I learn as much as I need to solve the problem. After the problem is solved, I go to another. I do not care about long-term knowledge. I am technical. Knowledge in my field goes out-of-date in 2-3 years.

Lifetime memories

I do not care about lifetime memories. I learn for short-term projects. My memory is good. I am a good problem solver.

High retention

As I learn for short-term, my retention is always high (within the scope of my needs).

Comprehension

I understand my texts well enough. If something is unclear, I look for better texts.

Uniform progress

I do not care about learning "nothing about everything". I learn stuff that I need here and now. Why waste time and memory space on memorizing British monarchs. That's good for kids at school, not for a real problem solver.

Creativity

I am creative enough. Thank you. I do not need to mix up domains of knowledge in hope of some elusive Nobel-winning discovery. I solve problems. I keep all my knowledge in my head. I put the pieces together when I need them. I do not have problems with creativity.

Consistency

I instantly recognize contradictions when reading about the materials of interest. I work for short-term projects, keep all my knowledge in memory, and instantly know if something fits or does not fit the model. This is a non-issue!

Stresslessness

I love my job. I do not consider it stressful. Just the opposite. I think struggling with complexities of SuperMemo might give me a heart attack!

Attention

I am passionate about my job. I do not force myself to learn boring stuff. My attention is excellent.

Consolidation

I do not work on projects that drag for years. I do not need to consolidate. I learn what I need, solve a problem, and move on.

Prioritization

I prioritize by what I need at the moment. I need A, I read A. I need B, I read B. In problem solving, rarely do you need to read dozens of articles at the same frame of time. You read what you need and fix things.

Speed (of reading)

I read fast enough. I have no idea how interrupted reading could be faster than normal reading or skimming.

Speed (of formulating items)

I do not need items.

Meticulousness

I am very meticulous. I do not need help in that department.

Training

The training in skills that one is supposed to develop with incremental learning is irrelevant. The skills are either of little value for me, or I get them in the course of my work anyway.

  • Recognizing suitable texts - all texts are suitable if they lead me to a goal. I Google, I read, I solve. Period.
  • Formulating knowledge - why waste time on formulating anything? I read and work. I need no intermediary time waster.
  • Mnemonic skills - all those mind maps sound like a monumental waste of time.
  • Speed-reading skills - my speed of reading is fast enough for my needs.
  • Semantic skills - I develop these in the course of normal reading too. I agree. The more you read, the better you get at understanding what you read.
  • Prioritization skills - I do not prioritize my reading. I read and execute.
  • Editing and SuperMemo skills - SuperMemo is one thing I really do not need.

Fun

I am not sure how adding complexity to your life with the concept of incremental learning can be more fun than just doing great things, doing them well, and having people praise your for the good you have accomplished.

SuperMemo user's opinion: Why I do not use incremental reading?

Here is an exemplary opinion of a user of SuperMemo who understands incremental reading, and still has many reservations:

I have been using SuperMemo for 19 years now (with occasional breaks). I understand the concept of incremental reading pretty well. I read about it a lot. Many times I thought of trying it. However, I am still not convinced that I should embark on this journey. Please see my story. I will gladly hear your opinion. Am I mistaken or am I different, you think? Or is the incremental reading picture a bit too rosy?

Learned read-only mentality

I started making repetitions years back with a goal to improving my English. To do that, I relied on learning with a ready-made Advanced English collection containing 40,000+ words and phrases. The learning material was comprehensive (far exceeding my needs), and the default learning process efficient enough to quickly propel me into good command of the language. Encouraged by the results I kept exploring the program but rarely took advantage of the ever-expanding gamut of SuperMemo's new advanced options. It seemed as if I hadn't really really needed them to get the results I wanted (at least in the English language department). All I did was make outstanding repetitions, memorize my daily quota of new words and phrases, and, occasionally, add a piece of vocabulary or grammar which wasn’t originally in the collection.

Incompatibility of my profession with incremental approach

As for my other learning needs, they start with my professional life which happens to be computer programming. And it seems like an area of knowledge which is not particularly well-suited for declarative learning. Furthermore, Dr Wozniak himself admits learning materials for programming are not suitable for processing with incremental reading. In other words, I seem to be unable to realize the promise of incremental reading in my life’s vocation thus leaving me less determined to explore it for other domains.

Stress factor

I believe that I have achieved a pretty good mastery of SuperMemo over the years. It does not mean though that SuperMemo still isn’t able to throw a curve ball every now and then. Take the Spread priorities feature for instance. When I first checked it, I was greeted with 0!!! against the yellow background in one of the fields: Step. I tried to change the value but no matter what I did I kept coming back to square one. As it turned out, it was SuperMemo’s way of telling me there aren’t enough priority positions for the subset I selected. SuperMemo may lack in the implementation department, even though it tries to address this, many times, at great length, in the documentation. In other words, with enough persistence you can find your way around the program. This journey of discovery may be stressful but nowhere near as much as the whole idea of incremental reading if you give it a careful thought. In fact, the more I read about it (at an attempt to dispel my doubts), the more stressful it becomes. Particularly, the entire prioritization stage seems overwhelming (to the point of nausea). The advantages you so diligently describe may all be true (although some are far from obvious and/or instant). Most of those advantage will hold only as long as you set the priorities right. But the entire process seems like a wild goose chase to me (not only because the priority queue is relative in its design). It would be interesting to know how much the student actually spends on priorities in the course of a daily learning session. For example, take a priority scale. In its relative variation (i.e. when expressed as percentage), it ranges from 1 to 100 which is simply too wide (the creators of the ABC analysis recognized that long time ago). I remember reading somewhere SuperMemo novices find it hard to grade yourself on the 6-piece scale. This now seems like a breeze as compared with giving a priority to a piece of knowledge. I understand the intention for more granular control but it is simply mind-boggling. To make matters worse, there are a number of processes (both manual/conscious and (semi-)automatic/unconscious) that erode your painstakingly set priority hierarchy. I do not mind when it actually changes but what if I accidentally stumble against an item whose priority I set as important, it still remains important, but somehow ended up in the later stages of the priority queue (and, from all what I have read, it is more than a remote possibility). When I read the article, I am often reminded of the following quote by Margaret Thatcher: "I do not know anyone who has got to the top without hard work. That is the recipe. It will not always get you to the top, but should get you pretty near." It holds a great promise but it is far from guarantee. I believe, I recognize really well how complex incremental reading actually is, and having that clearly in mind, I conclude your advantages section offers even less in the form of guarantee. You can invest a whole lot of time, and in the end SuperMemo will only make you realize how dismal you are at managing it all (because, make no mistake, you need to carefully manage it). The toolset may be rich but is it natural? Furthermore, it all depends on the correct input. Your advantages include uniform progress. Yeah, right. Developing tunnel vision when you import and process a large number of articles related to a single topic is as much if not (considering our human deficiencies) more likely.

Comment: You need to try incremental approach to get a good feel

In the section Incremental learning is not for everyone, you can read about personality factors that may conspire to make you fail with incremental learning. It is hard to make a prediction on the basis of your text. The fact that you studied incremental reading is a very good predictor of success. On the other hand, the fact that the mere concept of priority queue can generate stress, tells you a lot about your personality. You might just be one of those who excel at theory and fail in practical situations for a number of reasons. You should definitely give incremental learning a try and see how it works. If frustrations keep multiplying beyond the first two months of use, your chances may drop. It is vital to start from simplest concepts: import, extract and cloze. All the remaining tools should wait for their turn until you start enjoying the process. For the same reason, you should start from learning about things you love (not about algebra or economics). Your hobby, your pet, your favorite sport, health, your favorite music band, your favorite actor or actress, etc. Use Wikipedia as its texts are well-suited for incremental reading. Only after 2-3 months, when you start sensing the new quality of knowledge emerge in your mind, will you be able to anchor emotionally in the process and find sufficient motivation to overcome successive obstacles.

Short comments on the rest of your text: