Minimally guided teaching strategies have become popular over the past several decades. Minimally guided methods include a variety of techniques that involve students working on problems or discussing issues without specific instruction on how to solve the problems, and sometimes even without instruction about what they are supposed to accomplish. In contrast, direct instruction involves teachers providing students with all the information they need to learn (Clark, Kirschner, & Sweller, 2012). Minimally guided techniques are based in constructivist learning philosophy, proponents of which assume that learning involves construction of knowledge and that people learn most effectively when they engage in the discovery of knowledge rather than having knowledge imparted to them. Many authors advocate for the effectiveness of constructivist teaching methods, and many assume the superiority of these methods over direct instruction (Steffe & Gale, 1995; Mayer, 2004). Such methods are quite popular and are a central component of the Common Core educational standards that have been adopted by nearly all states in the United States.
Bruner’s (1961) article is often cited as the origination point of minimally guided instruction – often referred to as discovery learning – although the philosophical roots extend back at least to the work of Dewey (1897; 1902). Bruner distinguished between methods that are essentially teacher-centered, wherein the teacher sets the pace of instruction and decides on the mode of teaching, and those that are student-centered, wherein teachers and students are more collaborative and students take an active role in the learning process. He hypothesized that the process of students learning by discovery might have unique benefits over traditional instructional processes, speculating that “discovering for oneself teaches one to acquire information in a way that makes that information more readily viable in problem-solving” (p. 26).
Around the same time that Bruner (1961) expressed his hypotheses regarding the effectiveness of discovery as a learning method, Piaget (1965) summarized trends he had observed in education over the preceding 30 years. He noted a growing emphasis on the idea that students should discover knowledge and learn concepts independently rather than having someone else impose that knowledge. He further emphasized the importance of active learning, although he noted that classroom activities are not ends in themselves, but instead become valuable to the extent that they promote cognitive activities such as reflection and abstract thinking. Piaget was critical of methods where students are provided with concepts rather than discovering them, arguing that direct instruction methods are less likely to lead to learning that transfers to broader contexts.
Over the subsequent decades, the hypotheses of Bruner (1961) and Piaget (1965) formed the backbone of various constructivist teaching approaches. Jonassen (1991; 1998) describes constructivist perspectives of learning as based on the assumption that knowledge cannot be transmitted to learners, but rather must be constructed by them based on individual experiences with the world. Jonassen contrasts constructivist views, where reality is assumed to exist primarily in individuals’ minds, to objectivist views that assume an external reality that students can come to understand through direct instruction.
The literature on minimally guided instruction is particularly challenging to summarize because many terms are used inconsistently which makes it difficult to clearly understand what teachers actually do when they apply various techniques. Conducting research on minimally guided learning is likewise difficult because there is no consistently applied definition of what it is (Klahr & Nigam, 2004; Alfieri, Brooks, Aldrich, & Tenenbaum, 2011). As a consequence, there is surprisingly little direct experimental research comparing fully guided and minimally guided instruction. Many claims concerning the effectiveness of different methods are based on interpretations of related educational research rather than on direct experimental comparisons. Moreover, there is much disagreement among educational researchers concerning what broad techniques even qualify as minimally guided. Since teachers virtually never provide students with a problem to solve or a topic to discuss without any guidance at all (Brunstein, Betts, & Anderson, 2009), the distinctions between differing levels of guidance quickly become blurred.
One of the few well-controlled direct comparisons of discovery learning with direct instruction was conducted by Klahr and Nigam (2004), who compared the techniques in terms of both initial learning and transfer of learning to other tasks. The researchers randomly assigned third and fourth grade students to learn to design experiments via either direct instruction or discovery learning. In both conditions, students engaged in active learning by developing experiments to test scientific questions pertaining to the motion of balls of different materials rolling down ramps varying in steepness, length, and roughness. In the direct instruction condition, the teacher presented and explained examples of good and bad experimental designs; in the discovery condition, students spent the same amount of time designing experiments, but without any explanations or examples. Immediately after this phase, students each designed four new experiments; a week later, they each evaluated science fair posters produced by sixth grade students from a different school. The data were collected by an experimenter who was blind to the experimental conditions, and the students’ responses were coded by independent evaluators.
Klahr and Nigam (2004) found that students in the direct instruction condition improved far more in their ability to design quality experiments – ultimately performing twice as well as students in the discovery condition. The majority of students (77%) in the direct instruction condition achieved mastery of the learning task, whereas only 23% of the students in the discovery condition achieved mastery. Direct instruction led to much greater success for students at all levels of initial performance. Moreover, students’ ability to critique science posters did not vary as a function of the learning method. Mastery of the learning task was the critical variable determining students’ ability to evaluate the posters – whether students had achieved that mastery through direct instruction or discovery learning. In sum, direct instruction led to greater learning overall, and discovery learning did not lead to greater transfer of learning to another context.
In an interesting partial replication of Klahr’s and Nigam’s (2004) study, Dean and Kuhn (2006) compared discovery with direct instruction over a longer time frame. They randomly assigned fourth grade classes to one of three conditions with the same learning objective used by Klahr and Nigam. Students in a practice condition worked on computerized problems without teacher guidance for 12 sessions over 10 weeks. Students in a second condition engaged in the same 12 practice sessions preceded by a single direct instruction session at the outset where a teacher presented two good and two bad examples of experiments along with explanations. The length of this direct instruction session was not reported, but it appears from the procedure that it was quite brief. Students in a third condition received only the single direct instruction session with no subsequent practice. When students were tested seven weeks after the last session, those who had received both practice and direct instruction were no better off than students who had only practiced. The authors concluded that, at least in this case, direct instruction was not a necessary part of the learning process.
There are several reasons to be wary of this conclusion. First, the single direct instruction session was apparently very brief and was followed by 12 practice sessions for both groups. It is perhaps not surprising that the effect of the direct instruction session would be diluted after students’ participation in 12 practice sessions. Second, students in the practice only condition still received guidance in the form of an introductory session where the learning activities were explained, as well as reminders during some practice sessions of information from earlier sessions. This would further dilute any direct instruction effect since students in all groups were in fact receiving guidance. Finally, the researchers did not compare equivalent amounts of unguided practice and direct instruction. The finding that very minimal direct instruction increased learning only slightly beyond a great deal of practice is of uncertain value.
In another test of learning with minimal guidance, Rittle-Johnson (2006) assigned children in grades three through five to two conditions in which the students solved math problems. In the instruction condition, the experimenter directly taught students a specific strategy for solving the problems; in the discovery condition, the experimenter provided no instruction and told students to think of a way to solve the problems. Students then solved a series of problems, each of which was followed by feedback and the correct answer. They then took an immediate post-test and another post-test two weeks later. When solving the initial problems, students who received direct instruction averaged nearly twice as many correct answers as those in the discovery condition – despite the fact that all students received feedback on their performance as they worked through the problems. Students who had received direct instruction also did better on both the immediate and delayed tests, which included problems that were analogous to those they had seen previously, as well as problems that had some similarities but were different from what the students had seen before. Rittle-Johnson concluded that direct instruction was a more reliable way of teaching students the correct way to solve problems. She also noted that many students in the discovery condition – but no students in the instruction condition – used incorrect strategies, and she found no evidence that the discovery method led to greater transfer of learning.
Alfieri and colleagues’ (2011) meta-analysis provides a broad perspective on the relative effectiveness of discovery learning versus guided instruction. These researchers emphasized that a host of varying techniques have been included under the heading of discovery learning, but that the term is most commonly used to refer to methods in which students must acquire knowledge and understanding independently rather than having it provided to them. Alfieri and colleagues integrated the findings from more than 100 studies comparing minimally guided discovery methods with explicitly guided techniques. Overall, guided instruction led to greater learning, although the effects varied widely in strength due to the wide variety of samples, learning objectives, and research methods. Interestingly, studies published in top-tier journals demonstrated larger benefits of direct instruction over unguided instruction than studies published in lower-tier journals. Although the superiority of guided instruction was observed across learning domains, it was stronger when students learned verbal and social skills than when they learned math and science topics. Some types of explicit instruction were more effective than others, but all were more effective than unguided discovery. The researchers concluded that unguided discovery is not an effective strategy in terms of student learning, and that techniques such as direct teaching, having students do worked examples where they have access to all steps in the problem-solving process, and providing students with explicit feedback on their performance are much more effective. They further concluded that direct instruction is usually necessary to teach students problem-solving approaches before discovery methods can be employed, and that optimal teaching methods all involve some kind of meaningful guidance, instruction, or feedback.
There is evidence that student learning can sometimes be enhanced under conditions of minimal guidance. Brunstein et al. (2009) argue that discovery learning and direct instruction represent ends of a continuum ranging from one extreme at which students are told what to learn, to the other extreme at which they must figure out what to learn. However, Brunstein and colleagues also point out that all discovery learning methods involve some guidance and that pure discovery can therefore be distinguished from guided discovery. They randomly assigned undergraduate students to use a computerized tutoring program to solve complex math problems under one of four conditions: receiving verbal instructions, receiving a direct demonstration of how to solve the problems, receiving both verbal instructions and a demonstration, and receiving neither instruction nor a demonstration – which was intended as a discovery learning condition. The researchers reported that students in the minimally guided condition were ultimately most successful in that they completed a lengthy problem set more quickly. However, Brunstein and colleagues noted that in fact their discovery condition did provide guidance by providing general instructions on the purpose of the problems to be solved, limiting the options that students had for exploring possible solutions, and providing feedback on student performance. They speculated that additional instruction throughout the process would likely have led to similar success and even greater efficiency. The researchers concluded that students may sometimes be able to learn with minimal guidance but only when they search within a limited number of possible solutions, and engage in extended practice, on tasks that are not too cognitively demanding.
There are several reasons why instruction with minimal guidance tends not to have the dramatic advantages that advocates expect. First, the emergence of minimally guided instructional philosophy and techniques occurred prior to the publication of much of the existing research on human cognitive processing, and the techniques are in some ways inconsistent with that research (Kirschner, Sweller, & Clark, 2006; Sweller, 2009). Perhaps most relevant are the now well-known limitations of human short-term memory (Miller, 1956), which are particularly striking when a person is trying to actively work with the information in his or her immediate consciousness rather than merely maintaining it in memory (see Cowan, 2000, for a review). Sweller argues that educational methods that require teachers to deliberately withhold knowledge from students so that the students can discover it themselves are problematic because they ignore the limitations of working memory.
In his review of cognitive load theory, Artino (2008) similarly concluded that students will learn less effectively when teaching methods place excessive demands on their working memory. He cites the perils of teaching approaches characterized by extraneous cognitive load, which he defines as the extra load placed on working memory when students have to engage in cognitive tasks unrelated to the material or skills to be mastered. He argues that good teaching should minimize such extraneous demands on working memory, and that teachers can accomplish this by directly providing students with the background information they need to solve a particular problem, and then giving students the opportunity to solve increasingly complex problems. He questions the wisdom of employing minimally guided techniques with students learning novel information, because such techniques are likely to overwhelm students’ working memory. Other researchers agree that direct instruction places fewer demands on working memory, because students can focus their attention on relevant information and need only understand the problem and how to solve it rather than having to expend cognitive energy searching for the problem itself (Clark et al., 2012). Artino asserts that any teaching method must be adapted to the current knowledge level of the students involved. When students are learning truly novel material, they will likely need direct instruction; as their knowledge of the topic deepens, they will be able to perform more complex tasks with greater autonomy. This perspective is consistent with Bjork’s and Bjork’s (2011) description of what they term desirable difficulties relevant to learning. They explain that engaging in tasks that make learning seem more difficult – such as varying the methods used to learn – can lead to more effective learning, but only if the learner has “the background knowledge or skills to respond to them successfully” (p. 58). In the absence of such knowledge, they state, the same methods create difficulty that is detrimental to learning.
The full relevance of this progression from novice to expert, and its implications for instructional methods, is evident when one considers the cognitive skills necessary to demonstrate true expertise in any particular domain. Kirschner et al. (2006) cite evidence that people who become very effective at solving problems in a specific domain are successful because they have a great deal of relevant past experience stored in long-term memory. Novices have access only to the information immediately available to them, and their ability to process that information is greatly restricted due to the limitations of working memory. In contrast, experts can draw on the vast experience stored in their long-term memory, making working memory limitations far less pertinent. Experts can therefore quickly determine how to solve a problem because they can recognize similarities to past problems. Kirschner and colleagues assert that minimally guided teaching approaches require novice students to tax their working memory capacity by trying to identify relevant information and an effective problem-solving strategy, thereby leaving students with less working memory capacity to transfer information to long-term memory. Without relevant knowledge, students can only “blindly search for possible solution steps” (Clark et al., 2012: 10).
Neglecting to consider the limitations of working memory may have especially dire consequences for low-achieving students. Clark (1982) reviewed nine studies examining the interaction between teaching method and student ability in determining student achievement. In the majority of studies, lower ability students learned more effectively from less cognitively demanding methods characterized by greater structure and guidance. High ability students tended to learn more from methods that provided less guidance and therefore imposed greater cognitive demands. Paradoxically, students reported greater enjoyment of teaching methods from which they learned less – although they were not aware that they were learning less. High ability students tended to report greater enjoyment of structured methods characterized by lower cognitive load, whereas low ability students preferred unstructured methods that imposed greater cognitive load. Three decades after Clark’s review, he and his colleagues continued to express concern that minimally guided approaches pose the risk that only high-achieving students will discover knowledge, and other students will become frustrated and disinterested (Clark et al., 2012). Even Bruner (1961), an early advocate of discovery learning, asserted that “Discovery … favors the well-prepared mind” (p. 22).
A second factor that may limit the effectiveness of minimally guided approaches is the nature of the material taught as part of formal education. Sweller (2009) cites a number of constructivist arguments regarding the human ability to learn many skills outside formal educational environments and with little or no conscious effort. For example, most children learn to speak and engage in meaningful social interaction simply by being around others who are engaging in those behaviors. However, there are important distinctions between these type of skills and the skills that teachers typically want students to learn. Geary (2012) argues that skills that were relevant to daily survival throughout human evolutionary history develop with little effort on the part of learners. He considers skills such as speaking and social interaction to be examples of such biologically primary knowledge. In contrast, Geary categorizes skills such as reading, writing, solving math problems, and critically evaluating evidence as biologically secondary knowledge because these skills have existed only for a tiny portion of human evolutionary history. Developing these skills requires conscious effort and explicit instruction. Geary asserts that children are inherently motivated to learn primary knowledge but not secondary knowledge, stating “We would not need modern schooling, or at least not 12 or more years of it, if children found the activities that promote secondary learning as engaging as they find interacting with friends, and secondary learning as effortless as native-language learning” (p. 613).
Finally, minimally guided approaches often inappropriately equate cognitive activity with behavioral activity. In his critique of minimally guided techniques, Mayer (2004) agrees with other critics (Kirschner et al., 2006; Sweller, 2009) that the constructivist educational philosophy indeed has merit due to its emphasis on active student involvement, student construction of knowledge, and practical application of knowledge. However, Mayer takes issue with the assumption that active learning requires actual behavioral activity and unguided discovery. He questions constructivist assumptions that teaching methods such as interactive activities and group discussions are effective, whereas supposedly passive methods such as reading books and listening to lectures are ineffective. He refers to this assumption as the constructivist teaching fallacy because it “equates active learning with active teaching,” and notes that discussing a problem guarantees neither active cognitive processing, nor a solution (p. 15). His critique parallels Piaget’s (1965) observations from four decades earlier – that cognitive activity is a far more important factor in the learning process than behavioral activity. Mayer acknowledges that it is difficult to develop strategies that treat behavioral activity as a means to facilitate cognitive activity rather than an end in itself. Nonetheless, he emphasizes that thinking, rather than behavioral activity, is the most important factor in the learning process and that “guidance, structure, and focused goals” are critical for promoting thinking (p. 17).
It is important to note that there is much disagreement among researchers regarding what specific teaching methods are appropriately defined as minimally guided. For example, Kirschner et al. (2006) place discovery learning, inquiry learning, constructivist learning, and problem-based learning all under the heading of minimally guided techniques, calling them “essentially pedagogically equivalent approaches” (p. 75). Artino (2008), likewise, refers to all these techniques as similar to one another in their withholding of direct instruction. Other researchers disagree that techniques such as problem-based learning are in fact minimally guided (Hmelo-Silver, Duncan, & Chinn, 2007; Schmidt, Loyens, Van Gog, & Paas, 2007). Schmidt and colleagues agree that minimally guided instruction is not effective for teaching students novel information. However, they argue that problem-based learning – where students work in small groups on problems that include a description of an event that students must attempt to understand – is actually characterized by flexible and adaptive guidance. They note that group members usually study relevant background material on their own between discussions, and that discussions are often guided by a tutor who may at times provide information to assist the process. Hmelo-Silver and colleagues agree, arguing that in both problem-based learning and inquiry learning, teachers sometimes provide necessary information – although only after students’ need to have such information becomes apparent.
Kirschner and his colleagues responded to these advocates by citing evidence that problem-based learning is by definition self-directed because students must search for possible solutions rather than being directly taught how to reach solutions (Sweller, Kirschner, & Clark, 2007). They assert that requiring students to search for solutions strains working memory resources and inhibits learning in ways that would not occur if students were simply provided with the necessary information. Interestingly, they also argue that collaboration itself taxes working memory resources because students must cope with the cognitive demands of group interaction. Hmelo-Silver and colleagues (2007) claim that scaffolding in the form of providing information to students when necessary can increase the effectiveness of problem-based and inquiry learning. Kirschner and colleagues criticize this view as ignoring the most obvious and direct form of scaffolding which begins with providing students with a problem and teaching them how to solve it.
Based on the educational research to date, as well as related research on human memory and cognitive processing, there appears to be little basis for claims that minimally guided teaching methods are more effective than guided approaches. Importantly, many critics of withholding guidance do not find fault with discovery-based learning when such learning includes direct instruction. Mayer (2004) cites several studies dating to the 1950s suggesting that guided discovery is more effective than unguided discovery. Moreover, Alfieri and colleagues (2011) provided evidence through their meta-analysis that guided discovery may in fact lead to greater student learning than either unguided discovery or explicit instruction. Mayer concludes that students will benefit most from a balance of appropriate guidance to help them identify what must be learned, and appropriate freedom to become engaged with the learning process and to actively work to make sense of what they are learning. Other scholars agree that such a balance is desirable, but note that it is challenging to implement because it is difficult to know when it is best to teach students a solution and when it is best to let them discover a solution (Brown & Campione, 1994).
Withholding guidance from students may have negative consequences beyond ineffective learning. Researchers have cited evidence that minimally guided approaches can lead students to discover false information, which results in confusion and the establishment of misconceptions (Brown & Campione, 1994; Kirschner et al., 2006; Clark et al., 2012). Students may also become frustrated when they are unsure what to do. Brunstein and colleagues (2009) reported that in one of their studies, half of the discovery learning students – but none of the direct instruction students – quit the experiment because “they felt totally lost and did not want to continue” (p. 798). Minimally guided methods may also lead to expanded achievement gaps between groups of students. Von Secker and Lissitz (1999) cited US National Education Standards that advocate student-centered learning, assert that active learning must involve interaction between students, and deemphasize the importance of teachers presenting information in favor of students discovering knowledge. To investigate the wisdom of these standards they studied US Department of Education data on a sample of more than 2,000 tenth graders who were representative of a much larger US student sample and found that a shift away from teacher-centered instruction did not lead to any significant improvement in average student achievement, and actually increased gender and ethnicity achievement gaps. The researchers concluded that using student-centered methods rather than teacher-centered methods will not improve student learning unless students first acquire some basic knowledge so that self-directed activities and group collaboration can be effective.
It is interesting to note that one of the most frequently cited articles ostensibly advocating minimally guided methods actually contains a more balanced perspective than secondary sources imply. Bruner (1961) presented many of his ideas in the form of hypotheses and did not make broad claims about the effectiveness of discovery methods. For example, as noted earlier, he stated that “discovering for oneself teaches one to acquire information in a way that makes that information more readily viable in problem-solving” (p. 26), but he followed immediately with the infrequently cited statement, “So goes the hypothesis. It is still in need of testing” (p. 26). It is also apparent from Bruner’s article that he, like other scholars, had difficulty describing exactly what various proposed discovery methods would look like in practice.
Critics of minimally guided instruction generally do not question the potential value of having students engage in independent work to practice skills they have been taught, but rather the belief that knowledge students gain via minimally guided approaches is somehow more valuable or more useful than knowledge that teachers present in a direct fashion (Sweller, 2009; Clark et al., 2012). Indeed, it remains to be empirically demonstrated that teachers can enhance student learning by deliberately withholding information from students. Sweller et al. (2007) make this point effectively, and point out that the superiority of discovery methods over direct instruction has been a dominant assumption in education for many years, and has proven to be “sufficiently attractive to be impervious to the near total lack of supporting evidence from randomized, controlled experiments” (p. 120). There may be conditions under which specific minimally guided techniques are superior to direct instruction for particular learning objectives assuming particular student characteristics, but such parameters have not been adequately identified and tested. Consequently, there is little evidence to conclude that minimal guidance in general is superior to direct teaching methods.