Confidence-Based Repetition (CBR)

Modified on by Andrew Cohen

Confidence based repetition explained

We have all experienced that “tip-of-the-tongue” feeling.  You know, the one where you know you know something, but you just can’t quite remember it at the moment?  The reason you experienced the feeling is probably that it had been a bit too long since you had thought of the given fact (perhaps because it was a very rare word or name).  Although the feeling can sometimes be quite disconcerting, it does have a very positive benefit:  Once you are finally reminded of the correct answer, you will probably never forget it again!

This substantial memory-retention benefit is the central premise behind confidence-based repetition (CBR), a term coined by the learning experts at Brainscape.  CBR uses your own confidence ratings to space each exposure in the optimal interval of time: research shows that the optimal interval of spaced repetition is the longest amount of time before you would have otherwise forgotten the concept (Bahrick & Phelps, 1987).  Using CBR to determine the repetition interval for each concept results in the most efficient use of study time for a given subject.

The Emergence of Confidence-Based Repetition

Several learning software programs have applied CBR under different names over the years.  SuperMemo, created by Polish psychologists in the 1970s, was the first widespread application that people could purchase and install onto their PC.  Since the advent of the internet, several dozen similar “flashcard” programs have proliferated, the most notable of which being Mnemosyne, Anki, and most recently Brainscape.  The majority of such programs allow users to create and share their own personal content as well as study content from a public library.  Some programs even allow users to export their own content to a mobile phone.

As for the methods of assessing users’ confidence in each concept, though, there is wide variation between existing programs.  Some programs simply ask “Were you right?” after each answer is displayed, giving the user a choice of Yes or No.  Other programs allow you to “flag” hard concepts to come back to later.  Finally, programs like Brainscape request that users rate their confidence on a numeric scale.

We at Brainscape believe that assessing users’ confidence on a numeric scale, and using that information to determine when the concept is displayed again, is the most effective application of study time.  This method of CBR requires the user to actively recall the concept rather than simply recognizing it on a list of multiple choices, and it determines the pattern of repetition for the user rather than requiring the user to remember when to access their bank of “flagged” items.  Brainscape’s use of a 1-5 confidence scale is particularly effective because it conforms to most people’s level of metacognitive precision (Likert, 1932; Son, 2004). As more learning institutions discover the science and effectiveness of CBR, we expect the method to proliferate rapidly over the coming decade.  Advances in web and mobile technologies will continue to make it more accessible in any learning setting and for nearly any type of factual content.  From schools, to corporations, to language learning, to test prep, modern culture simply requires too much learning for us to use anything less than the most efficient learning tools.

Want to learn more? Check out our Confidence-Based Repetition white paper.


Bahrick, H.P., & Phelps, E. (1987). Retention of Spanish vocabulary over 8 years. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13(2), 344-349.

Likert, R. (1932). A Technique for the Measurement of Attitudes. Archives of Psychology, 140, 1–55.

Son, L.K. (2004). Journal of Experimental Psychology: Learning, Memory, and Cognition, 30(3), 601-604.

Brainscape is a web & mobile education platform that helps you learn anything faster, using cognitive science. Join the millions of students, teachers, language learners, test-takers, and corporate trainees who are doubling their learning results. Visit or find us on the App Store .


Jamie 9 years ago

Surely confidence based repetition opens a huge margin for error. In order for me to accurately know how confident I am with a given learning objective I need to have seriously well developed meta-cognitive processes. Also the subjective nature of self evaluation means that a whole host of factors (mood, tiredness, context) can interfere, meaning that you're not defining the repetition spacing according to something that's consistent and therefore trustworthy. To me this is another example of someone coming up with a sexy name for something that's actually pretty flawed.

Andrew Cohen 9 years ago

Good points, Jamie, and we actually cover these concepts in our white paper (  The important thing to remember is that your confidence ratings are *relative* rather than absolute.  Even if you have systematically given overconfident ratings to every flashcard across the board, whichever ones have the lowest confidence ratings will still come up more often.  Using Brainscape also helps you refine your metacognitive self-evaluation so you will become an even more effective studier over time.

LittleFish 11 years ago

"Finally, Anki, Brainscape, and the latest version of SuperMemo request that users rate their confidence on a numeric scale."

A correction: Since the time Supermemo was first created in paper form in 1985, it has always used a scaled self-grading system to remember content. It began as 1-5, but has now evolved into the present from of 0-5. 0 = Null, 1 = Bad, 2 = Fail, 3 = Pass, 4 = Good, 5 = Bright. Since most recent Spaced Repetition Programs are based on older versions of Supermemo, I think a scaling grade system is not a recent innovation in those programs either.

Andrew Cohen 11 years ago

Thank you for the clarification! I hadn't been able to get my hands on earlier versions of Supermemo, so I was referring to the latest version to be safe. Supermemo is indeed a pioneer in the field of spaced-repetition, and I am glad to have the opportunity to work on a product that helps make its methodologies accessible in a different medium and interface.

Ayesha Nicole 8 years ago

Which version? The most recent now is SM15.

Margot 10 years ago

This learning model is unique in that it allows the user to be intrinsically motivated. She is not answering the questions in order to get a particular score (of correct/incorrect answers) after a round of studying. The only measure of success is how confident she is in what she's been studying, which takes the pressure off of her to perform. This allows the user to be more honest with herself; thus making the study session optimally beneficial, without the added anxiety of extrinsic evaluation.

comments powered by Disqus