What's better – discovery learning or direct instruction? It depends on learner's characteristics (too).

Student Learning: What Has Instruction Got to Do With It? Hee Seung Lee and John R. Anderson  Annu. Rev. Psychol. 2013.64:445-469  http://www.usc-dr-edens.org/uploads/7/2/5/3/7253252/annurev-psych-student_learning__and_instruction.pdf

Discovery learning is a learning style in which students receive minimal instruction and construct their own knowledge.  Direct instruction is the traditional learning style of teaching the material and then practicing it.  Which style is better?

Although this question is interesting and relevant for every study domain, most research cited in this interesting paper was done in math.  I hope the conclusions can be generalized to other domains.

Learning conditions that introduce certain difficulties during instruction, as happens in discovery learning, appear to slow the rate of learning but often lead to better long-term retention and transfer than learning conditions with less difficulty.

This argument was assessed in a study in which students were trained to decipher cryptograms with different forms of instructional methods. The researchers compared students who were given explicit rules followed by problem practice with students who just tried to solve the problems and had to discover the rules. The discovery students did better on transfer problems that required new rules.

In another study,   children in discovery classrooms  used a variety of physical manipulatives and worked in pairs to solve mathematical problems. After working in pairs, a teacher led a whole-class discussion, and the children talked about their interpretations and solutions. After a while, students were given a standardized achievement test. The results showed that students in discovery and regular classrooms were not different in terms of the level of computational performance. However, the students in the discovery classrooms demonstrated higher levels of conceptual understanding than those in the regular classrooms.

In another study, children were traced for three years to assess understanding of concepts and procedures on multidigit addition and subtraction. The study compared students who used an invented strategy with students who used a standard algorithm. Students who invented a strategy were able to use not only their own invented strategy (if asked to do so), but also the standard algorithm after they learned that. Invention students also showed better understanding of base-ten number concepts and better performance in a transfer task. On the other hand, the algorithm group showed significantly more buggy algorithms in their problem solving than did the invented-strategy group,   implying that they depended on the use of learned procedures and lacked a deep conceptual understanding about the computation procedures.

 Discovery learning is believed to increase students’ positive attitudes toward learning. Learning through exploration allows students to have more control in a task, and this in turn fosters more intrinsic motivation. In addition, it is argued that discovery learning enables students to learn additional facts about the target domain.

The discovery learning approach appears to be effective only with high levels of practice and more time for the learning phase than allocated in direct instruction. In the early phases of learning, students who are learning through discovery make more errors and understand the material less.  Only when they have enough time to learn and practice they enjoy the advantages of discovery learning.

Another instruction method that works well is worked examples. Worked examples   provide an expert’s solution that students can emulate. Students  are typically given step by-step solution steps, and a final answer to the problem. Worked examples are usually alternated with problems to be solved.  Worked examples are very effective in the early phase of learning.  Students who are prompted to generate their own explanations for worked examples show  greater learning gains than those who are prompted to paraphrase provided explanations for the same example.

In another study, 7^th grade students solved multistep equations.  One student group studied sets of two differently solved solutions to the same problem.  The students were asked to compare the two solutions and to contrast them.  The two solutions were written on the same page, and each step was named.  Another student group studied the same sets of two worked examples, but each was presented on a different page.  The students were not asked to compare and contrast them but rather to think about each solution.  After two days of this intervention, the student's conceptual knowledge,  procedural knowledge  and procedural flexibility were tested.  The students who compared between the two worked examples gained more procedural knowledge and more flexibility, and had a better transfer ability than students who learned each example by itself.  There was no difference between the groups in conceptual knowledge.

Often it's best to integrate discovery learning and direct instruction. In one study, students learned the concept of density. In the direct instruction condition, student were told the relevant concepts and formulas on density and then practiced with contrasting cases. In the discovery condition, students had to invent formulas with the same contrasting cases first, and then formulas were provided only after they completed all the inventing tasks. Both groups of students showed a similar level of proficiency at applying a density formula on a word problem; however, the invention students showed better performance on the transfer tests that also required an understanding of ratio concepts but had semantically unrelated topics.   The direct instruction students did not have a chance to find the deep structure because they simply focused on what they had been told and practiced applying the learned formulas. The inventing activity appeared to serve as preparation for future learning, and thus when the expert solutions were provided later, these students could appreciate the expert solutions better than those who were not prepared. Even though most students fail to generate valid methods on their own during the invention phase, this failure experience actually helps students become prepared to learn better in the following learning phase by activating students’ prior knowledge and having students attend to critical features of the learned concepts. 

Discovery learning is not always better than direct instruction.  It depends on learner characteristics. Experienced learners (those who have prior knowledge of the material) benefit more from minimal instruction (that is, discovery learning).  Students with high levels of prior knowledge benefit more from comparing and contrasting two worked examples.  Students with low levels of prior  knowledge learn less well this way.  Comparing and contrasting only overloads them.  High ability students benefit more from discovery learning than low ability students.   When they have difficulties, high ability students tend to lean more on previously studied examples than low ability students.  High ability students spend more time studying worked examples than low ability students.  They also have more solution ideas, they can explain the solutions better, and they can identify their misunderstandings better than low ability students.  Students with low ability gain more from direct instruction.