It’s Okay to Fail: How Productive Failure Facilitates STEM Learning and Transfer in Comparison to Direct Instruction

by Ashley Chen

What role does failure play in learning? The main goal of instruction is to promote long-term learning such that knowledge can be retained and applied to novel contexts (Soderstrom & Bjork, 2015). In other words, educators strive to equip learners with knowledge that is both durable and flexible. Considering that classes tend to build upon information previously learned, it is important that students are able to successfully apply prior knowledge and skills in new problem-solving scenarios. This process may be challenging for students who are likely to experience failure without instructor support (Kapur, 2008). While students tend to be aversive to failure, it plays a natural and important role in the learning process. In this blog post, I will describe instructional strategies that educators implement in the classroom and the role that failure plays in facilitating learning and transfer.

Direct Instruction versus Productive Failure

Instruction is typically provided in the form of lectures. In science, technology, engineering, and mathematics (STEM) domains, these lectures are complemented by problem-solving activities, during which students are asked to apply learned concepts to familiar and new scenarios. Two instructional strategies that educators use in the classroom are direct instruction and productive failure, the main difference between the two being the timing of instruction. Notably, with direct instruction, instruction is provided prior to problem-solving, whereas with productive failure, problem-solving precedes instruction.

With direct instruction, educators teach students the correct procedures to solve problems, removing any opportunity for students to encode errors (Klahr & Nigam, 2004). It is argued that without such instruction, students are unable to gain the knowledge needed to derive solutions, especially for problems that require multi-step procedures (e.g., algebra, computer programming). In addition, working memory resources needed for new learning are preserved under direct instruction. When students lack the prior knowledge needed to solve a problem, they expend cognitive resources to search for solutions via trial and error. By directly showing students how to solve the problems, the burden on working memory is lessened (Kapur, 2014). However, it is argued that this burden is exactly what students need to better prepare themselves to learn from subsequent instruction.

Productive failure appears to be counterproductive because it has students solve ill-structured problems (i.e., problems beyond the current skill set and abilities of the learner) before instruction. Essentially, students are unable to generate canonical solutions since their knowledge is incomplete. However, according to the theory of impasse-driven learning, this experience of failure may be a prerequisite for later learning (VanLehn, 1988). Doing problem-solving first is beneficial in that students are encouraged to retrieve prior knowledge to generate original, but incorrect or subpar solutions, a process which equips students to revise their current understanding of the to-be-learned concept (Schwartz & Martin, 2004). Furthermore, having students generate their own solutions helps them identify inconsistencies between their answers and the canonical solution, allowing them to attend to and encode critical information, somewhat similar to the pre-testing effect (testing prior to study influences learning of subsequent information; Kapur, 2015; Kornell et al., 2009). Research on productive failure has shown that it can be more effective than direct instruction.

Productive Failure and Learning

The effectiveness of direct instruction versus productive failure as an instructional method has been compared across STEM domains, notably in math learning. In Kapur (2014), 75 ninth graders were randomly assigned to learn about standard deviation via direct instruction or productive failure. A final test consisting of multiple-choice and free-response questions targeted procedural knowledge (do students know the basic procedure to compute standard deviation), conceptual understanding (do students understand the critical features of standard deviation), and transfer (can students use prior knowledge to solve novel problems about normalization). Results revealed that while productive failure students perceived greater mental effort during problem-solving and instruction, they significantly outperformed students who received direct instruction on conceptual understanding and transfer problems and performed equally on procedural learning problems. A follow-up study by the same researchers indicated that it is necessary to have students generate their own solutions, as simply studying and evaluating peer-solutions during problem-solving does not lead to optimal learning in comparison to productive failure (Kapur, 2014). Therefore, even though students feel mentally challenged, struggle and failure can be valuable for one’s learning.

Instructional content plays an important role as well. 240 tenth graders also learning about standard deviation were taught using direct instruction and productive failure. However, during the instruction phase, in addition to being shown the canonical solution, students were randomly assigned to receive instruction that compared and contrasted typical student solutions to the canonical solution or not receive such instruction. On an immediate and one-week delay free-response test, those who were taught using productive failure with additional instruction scored higher than those who received standard instruction (i.e., no compare and contrast) on conceptual knowledge problems, whereas there were no differences in performance for productive failure and direct instruction students when they both received standard instruction (Loibl & Rummel, 2014). Thus, productive failure is effective when students are able to identify knowledge gaps and inconsistencies between their answers and the canonical solution.

Educational Implications and Future Directions

While most classrooms tend to use direct instruction, educators should consider implementing productive failure to better support students’ conceptual understanding and transfer performance. Notably, instruction could be structured in a manner that allows students to compare common student solutions to the canonical one. This can increase students’ global awareness of knowledge gaps and curiosity to learn. Instructors should also encourage collaboration during problem-solving. Research has shown that students benefit from collaborative learning as it allows them to generate more solutions and have thoughtful discussions as to why proposed solutions will or will not work (Kerrigan et al., 2021). In other words, when students work together to accomplish a shared learning goal, they are likely to have higher achievement and greater productivity.

However, while the benefits of productive failure have been observed across various studies, findings have been limited to STEM learning in elementary through high school educational contexts. Future studies should explore if productive failure is beneficial in less-structured domains such as in the humanities or arts and in higher education. Collecting data surveying how curious and motivated students are to learn the concept and related material before and after instruction and problem-solving may be valuable as well. In fact, it has been demonstrated that problem-solving before instruction enhances students’ intellectual need and curiosity, the mnemonic benefits of which contribute to productive failure’s appeal as an instructional strategy (Fandakova & Gruber, 2021; Loibl & Rummel, 2014). In addition, when encountering failure, students lose motivation to persist, while others who are able to recover from error experience enhancements in self-efficacy, which motivates them as they continue learning (Tawfik et al., 2015). While productive failure can be a useful instructional method, its efficacy beyond STEM domains and its motivational aspects warrant further study.

Productive failure may be undesirable in that it slows learning and introduces more opportunities for error during instruction. However, it has been shown that conditions that impede the learning process can lead to better long-term outcomes (Bjork & Bjork, 2011). Educators should acknowledge that it is natural to make mistakes, and even encourage it, as students are bound to experience failure at some point in their lives. Productive failure can prepare students to be successful during real-world problem-solving, and its potential as an instructional strategy is underappreciated. Failure is a fundamental aspect of the learning process, and we need to accept that it is okay to fail.


Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. Psychology and the Real World: Essays Illustrating Fundamental Contributions to Society, 2(59-68).

Kapur, M. (2008). Productive failure. Cognition and Instruction, 26(3), 379-424.

Kapur, M. (2014). Productive failure in learning math. Cognitive Science, 38(5), 1008-1022.

Kapur, M. (2015). Learning from productive failure. Learning: Research and Practice, 1(1), 51-65.

Kerrigan, J., Weber, K., & Chinn, C. (2021). Effective collaboration in the productive failure process. The Journal of Mathematical Behavior, 63, 100892.

Klahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661-667.

Kornell, N., Hays, M. J., & Bjork, R. A. (2009). Unsuccessful retrieval attempts enhance subsequent learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4), 989-998.

Loibl, K., & Rummel, N. (2014). Knowing what you don’t know makes failure productive. Learning and Instruction, 34, 74-85.

Schwartz, D. L., & Martin, T. (2004). Inventing to prepare for future learning: The hidden efficiency of encouraging original student production in statistics instruction. Cognition and Instruction, 22(2), 129-184.

Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2), 176-199.

Tawfik, A. A., Rong, H., & Choi, I. (2015). Failing to learn: towards a unified design approach for failure-based learning. Educational Technology Research and Development, 63(6), 975-994.

VanLehn, K. (1988). Toward a theory of impasse-driven learning. Learning Issues for Intelligent Tutoring Systems, 19-41.