When do we start forgetting?
If I read a chapter about the function of the liver, when do I start forgetting? When does the forgetting process begin exactly? When should I come back to that chapter for review?
Forgetting begins immediately. Optimum review time is hard to determine, but can be resolved with the use of a computer.
Your memories about the liver are like a sand castle. The weathering will begin immediately. The surface of the castle will start crumbling first. The wet interior of the castle with survive longer. This is an uneven process and there is no optimum time for reviewing the chapter about the liver.
However, we can address individual memories with spaced repetition. In SuperMemo, you tackle all components of the castle independently. It is easy to remember that liver is located in the stomach. It is harder to remember details of cholesterol synthesis in the liver. Those separate memories need to be addressed separately and on a different schedule. You probably never need to review the location of the liver. Each time you review its biochemistry, you will recall the organ that rests below the diaphragm. The location of the river is an example of a well-connected well-integrated memory. It has so many associations and links with things we know that it will safely get reviewed over and over on its own. The location of the liver memory is like the wet interior of the sand castle. For knowledge coherence, self-learning is vastly superior to learning at school. Many students waste years of life on cramming that results in virtually no long-term recall. Their knowledge never gets a chance to crystallize.
If you are a student of medicine, you would need incremental reading to best tackle the subject and all individual memories. You might import a chapter about the liver from your e-book or from Wikipedia. In the next step, you would extract all texts related to individual memories: location of the liver, cholesterol, biliary flow, etc. Each sub-text would receive its priority. For example, location of the liver in the abdomen would receive low priority or you would just delete it from your learning materials collection.
For each individual notion related to the liver, forgetting begins immediately as well. All memories are like microscopic sand castles. You will hardly ever be able to reduce your castles to grains of sand (i.e. individual synapses between nerve cells).
You will find many articles about spaced repetition that misinform you about the forgetting process. Most of writers and lecturers will mention the exponential forgetting process. It is easy to look at the forgetting curve and think forgetting is a ruthless gradual process. Consequently, you will hear a claim that "you need to review at the time when you start forgetting", or "you need to review when the memory is on the verge of forgetting". These are false notions. The exponential nature of forgetting comes from the fact that each granular memory has the same probability of forgetting today and in a decade (assuming the memory survives that long). This is very counter-intuitive if you recall that retrievability is defined as the probability of recall. That probability keep dropping over time. If you remember radioactive decay from physics, natural forgetting is very similar. Radioactive decay helps explain the constant probability of forgetting:
We can forget things fast or never depending on random chance (like in radioactive decay). However, the rate of forgetting will depend on the probability of forgetting in unit time, which in turn depends on the memory stability.
For mathematical reasons, the two things are inseparable: exponential decay and constant probability of forgetting. The illusion of changes in the forgetting rate come from the fact that the forgetting curve represents a mass of memories (e.g. words of English memorized with SuperMemo). Granular memories are like radioactive isotope atoms. Exponential forgetting shows up only (1) if memories are maximally simplified, and (2) if memories are of the same memory strength (stability). A simple memory may have a form of a cloze deletion. For example, the sentence common bile duct is the union of the common hepatic duct and the cystic duct, may be used to ask a simple question: common [...] duct is the union of the common hepatic duct and the cystic duct.
Note that memories may differ in nature. For example, learning about the liver will form declarative memories. Learning to bike will form procedural memories. Procedural memories do not get forgotten that fast. A child will always forget faster because of the birth of new brain cells that interfere with new memories.
When we speak of probability of recall, we mean a mass of similar memories in the past. When we speak of probability of forgetting, we speak of a specific singular memory here and now.
Figure: The first review forgetting curve for newly learned knowledge collected with SuperMemo. Power approximation is used in this case due to the heterogeneity of the learning material freshly introduced in the learning process. Lack of separation by memory complexity results in superposition of exponential forgetting with different decay constants. On a semi-log graph, the power regression curve is logarithmic (in yellow), and appearing almost straight. The curve shows that in the presented case recall drops merely to 58% in four years, which can be explained by a high reuse of memorized knowledge in real life. The first optimum review interval for retrievability of 90% is 3.96 days. The forgetting curve can be described with the formula R=0.9907*power(interval,-0.07), where 0.9907 is the recall after one day, while -0.07 is the decay constant. In this is case, the formula yields 90% recall after 4 days. 80,399 repetition cases were used to plot the presented graph. Steeper drop in recall will occur if the material contains a higher proportion of difficult knowledge (esp. poorly formulated knowledge), or in new students with lesser mnemonic skills. Curve irregularity at intervals 15-20 comes from a smaller sample of repetitions (later interval categories on a log scale encompass a wider range of intervals)