Training and Transfer of Complex Cognitive Skills: Effects of Worked Examples and Conventional Problem-Solving

Thirty six senior students in chemical engineering were randomly assigned to three treatment groups in an experimental study that examined the impact of different instructional strategies for troubleshooting malfunctions in a computer-based simulation of a chemical processing plant. In two groups, different types of worked examples, process-oriented and product-oriented, were given to participants as instructional strategies for troubleshooting four plant malfunctions. The third group was given a conventional problem solving strategy for the same four problems. The results of participants’ performance on solving a set of eight near-transfer problems indicated no significant transfer differences among the treatments. Neither did a far transfer task result in any significant differences. The findings of the current study supported the notion of the “expertise reversal effect” (Kalyuga, Ayres, Chandler, & Sweller, 2003), which argues that presenting new information to learners with pre-existing schemata in a given domain does not improve transfer and may induce extraneous cognitive load. Given the prior knowledge of the participants, these findings were also consistent with Sweller’s (2004) thesis on the “central executive function” and his description of the “redundancy effect.” In his analogy between evolution by natural selection and human cognitive architecture, Sweller (2004) lists the assumptions of cognitive load theory. The theory assumes that the purpose of instruction is to build knowledge by making small incremental changes in long-term memory. He argued that, similar to the way a potentially injurious drastic alteration in the human genome is usually prevented by the process of natural selection in species, a sweeping change in an individual’s long-term memory is prevented by the severe limitation of working memory when assimilating unfamiliar information. Such assimilation when no schema exists for organizing new information is performed by searching and testing the fit of random combinations of elements in the new material against premises derived from established assumptions retrieved from long-term memory. The demand on working memory is raised exponentially as the number of unfamiliar interacting elements of information is increased. The random search is essential to the human cognitive architecture when learners face completely unfamiliar information and a central executive is absent. However, when facing familiar information as opposed to unfamiliar information, a highly effective central executive function becomes available. As opposed to the human genome, this function is not a general biological structure, but a specific learned structure retrieved from long-term memory. In other words, cognitive processes conditioned by domain-specific knowledge act as the central executive when sufficient elements of instructional material are familiar. On the other hand, when the learner lacks a central executive because the information is unfamiliar, the use of worked examples in the design of instructional material can provide a surrogate central executive that constrains the problem space and the number of interacting elements to be randomly searched. Based on Sweller’s (2004) argument, when instruction is properly designed, effective changes to longterm memory structures are orderly and occur in small increments. Cognitive load theory provides guidelines for designing instruction. Investigating and expanding upon these guidelines, researchers have identified instructional strategies that can facilitate incremental changes in long-term memory. The use of worked examples is one of those instructional strategies that promotes efficient and effective learning by reducing extraneous cognitive load through the introduction of schemata and by acting as the “instructional central executive” (p. 21) and accommodating the limited capacity of the working memory of novices. When teaching complex cognitive skills to novices, the instructional strategy of presenting a set of worked examples for learners to study has been repeatedly found more effective than the conventional problem solving strategy in which they are provided problems to solve immediately after presentation of information in the domain (e.g., Cooper & Sweller, 1987; Paas, 1992; Paas & van Merriënboer, 1994a; Sweller & Cooper, 1985; for a review, see Atkinson, Derry, Renkl, & Wortham, 2000). Two types of worked examples, process-