This chapter examines the development of knowledge as a primary outcome of learning and how learning is affected by accumulating knowledge and expertise. HPL I1 emphasized these topics as well, but subsequent research has refined and extended understandings in a variety of learning domains. The first section of this chapter describes the problem of knowledge integration from the perspective of learning scientists and illustrates with research findings how people integrate their knowledge at different points in their development and in different learning situations. The second section describes what is known about the effects of accumulated knowledge and expertise on learning. The second half of the chapter discusses strategies for supporting learning. The committee has drawn on both laboratory- and classroom-based research for this chapter.
HPL I noted that the mind works actively to both store and recall information by imposing structure on new perceptions and experiences (National Research Council, 2000). A central focus of HPL I was how experts structure their knowledge of a domain in ways that allow them to readily categorize new information and determine its relevance to what they already know. Because novices lack these frameworks, they have more difficulty assimilating and later recalling new information they encounter. This chapter expands on these themes from HPL I, citing relevant research reported since that study.
Knowledge integration is a process through which learners put together different sorts of information and experiences, identifying and establishing relationships and expanding frameworks for connecting them. Learners must not only accumulate knowledge from individual episodes of experience but also integrate the knowledge they gain across time, location, circumstances, and the various formats in which knowledge appears (Esposito and Bauer, 2017). How knowledge acquired in discrete episodes is integrated has been debated for decades (Karmiloff-Smith, 1986, 1990; Mandler, 1988; Nelson, 1974). Some researchers have suggested that infants are born with foundational knowledge that provides the elements necessary for learning and reasoning about their experiences (Spelke, 2004; Spelke and Kinzler, 2007) or that infants can build from basic inborn reflexes to actively engage with the world and gradually build skills and knowledge (Fischer and Bidell, 2006). Others have argued that all knowledge is generated through an individual’s direct experience with the world (Greeno et al., 1996; Packer, 1985).
More recent work suggests that the integration of knowledge is a natural byproduct of the formation and consolidation of episodic memories (Bauer, 2009; Bauer et al., 2012). As described in Chapter 4, when a memory is consolidated, the learner associates representations of the elements of the experience (e.g., sights, sounds, tactile sensations) and these associations serve to help stabilize that memory. At the same time, these representations may also be linked with older memories from previous experiences that have already been stored in long-term memory (Zola and Squire, 2000). The fact that old and new memory traces can be integrated shows that these traces are not fixed. Instead, elements common to the new and stored memory traces reactivate the old memory and, as the new memory is consolidated, the old memory may be reconstructed and undergo consolidation again (Nader, 2003). When information from either learning episode is later retrieved, elements of both memory traces will be reactivated and will be simultaneously available for reintegration. As memory traces with common elements are simultaneously activated and linked, knowledge is expanded and memories are iteratively reworked. Figure 5-1 illustrates how this happens.
These linked traces may then be integrated with additional new information that comes to the learner later, and another new memory trace undergoes consolidation. Interestingly, it is exactly this process of integration of information from different episodes that may explain why people are sometimes unable to explain when and where they gained particular knowledge. Because the information generated by memory integration was not actually experienced as a single event, the information was not tagged with its origin (Bauer and Jackson, 2015).
The studies of knowledge acquisition in children and college students presented in Box 5-1 illustrate the capacity to integrate unconnected infor-
mation and retain this knowledge starting at a very young age. These studies underscore the active role of the learner; that is, even young children do not simply accrue knowledge from what they have experienced directly but build knowledge from the many things that they have figured out on their own, which, over time, they can do with less repetition and external support.
As discussed in Chapter 2, adequate sleep is important for integration and learning. The brain continues the work of encoding and consolidation during sleep and facilitates generalizations across learning episodes (Coutanche et al., 2013; Van Kesteren et al., 2010). Specifically, activation of the hippocampus (which plays a key role in memory integration) during sleep seems to allow connections between memory traces to be formed across the cortex. This process promotes the integration of new information into existing memory traces, allows for abstraction across episodes (Lewis and Durant, 2011), and leads to the possibility of building novel connections, which may be both creative and insightful or may be bizarre (Diekelmann and Born, 2010).
When people repeatedly engage with similar situations or topics, they develop mental representations that connect disparate facts and actions into more effective mental structures for acting in the world. For example, when people first move to a new neighborhood, they may learn a set of discrete routes for traveling between pairwise locations, such as from home to school and from home to the grocery store. Over time, people naturally develop a mental representation of spatial relationships, or mental map, that stitches these discrete routes together. Even if they have never traveled between the school and the grocery store, they can figure out the most efficient route by consulting their mental map (Thorndyke and Hayes-Roth, 1982). The observation that experts in a domain have developed frameworks of information and understanding through long experiences in a particular area was a central focus of HPL I. In this section, we briefly describe some of the benefits of expert knowledge (a more detailed discussion of the benefits of expertise appears in HPL I) and then discuss the knowledge-related biases that may come with expertise.
One of the most well-documented benefits of the acquisition of knowledge is an increase in the speed and accuracy with which people can complete recurrent tasks: remembering a solution is faster than problem solving. Another benefit is that people who develop expertise can handle increasingly complex problems. One way this occurs is that people master substeps, so that each substep becomes a chunk of knowledge that does not require attention (e.g., Gobet et al., 2001). People also learn to handle complexity by developing mental representations that make specific tasks easier to complete. When Hatano and Osawa (1983) studied abacus masters, they found that even without an abacus in front of them, the masters had prodigious memories for numbers and could carry out addition problems with very large numbers because they had developed a mental representation of an abacus, which they manipulated virtually. These abacus masters did not show similarly superior ability to remember or keep track of letters or fruits—tasks that were not aided by manipulating a virtual abacus.
A third benefit is an increase in the ability to extract relevant information from the environment. Experts not only have better-developed knowledge representations than novices have but also can perceive more information that is relevant to those representations. For example, radiologists are able to see telling patterns in an x-ray that appear merely as shadows to a novice (Myles-Worsley et al., 1988). The ability to discern more precise information complements a more-differentiated mental representation of those phenomena.
An implication of this ability is that students need to learn to see the relevant information in the environment to help differentiate concepts, such as the difference between a positive and a negative curvilinear slope (Kellman et al., 2010).
A fourth benefit of acquiring expert knowledge is that it helps people use their environment as a resource. Using what is known as distributed cognition, people can offload some of the cognitive demands of a task onto their environment or other people (Hollan et al., 2000). For instance, a major goal of learning is to develop knowledge of where to look for resources and help, and this is still important in the digital age. Experts typically know which tools are available and who in their network has specialized expertise they can call upon.
Finally, acquiring knowledge helps people gain more knowledge by making it easier to learn new and related information. Although some cognitive abilities related to learning novel information decline, on average, with age, these declines are offset by increases in knowledge accumulated through the life span, which empowers new learning. For example, in a study of young adults and older adults (in their 70s) who listened to a broadcast of a baseball game, the older adults who knew a lot about baseball recalled more of the broadcast than the young adults who knew less about baseball. This occurred despite the fact that the younger adults had superior executive functioning (Hambrick and Engle, 2002).
As people’s knowledge develops, their thinking also becomes biased. But the biases may be either useful or detrimental to learning. The word “bias” often has negative connotations, but bias as understood by psychologists is a natural side effect of knowledge acquisition. Learning biases are often implicit and unknown to the individuals who hold them. They appear relatively early in knowledge acquisition, as people begin to form schemas (conceptual frameworks) for how the world operates and their place within it. These schemas help individuals know what to expect and what to attend to in particular situations (e.g., in a doctor’s office versus at a friend’s party) and help them develop a sense of cultural fluency—that is, to know how things work “around here” (Mourey et al., 2015).
Psychologists distinguish two types of bias: one is intrinsic to learning and primarily useful and empowering to the learner; the second occurs when prior experiences or beliefs undermine the acquisition of new knowledge and skills.
An aphorism from the context of medical diagnosis illustrates the two types of bias: “When you hear hoof-beats, think of horses not zebras.” In the United States, horses are much more common than zebras so one is much more likely to encounter the common “horses” than the rare “zebras.” Of course, one should modify assumptions in light of additional evidence: if the
large mammal from which the hoof-beats emanate has black and white stripes, it is much more likely to be a zebra than a horse. Thus, if one sees a striped animal in a zoo but insists that it is a horse and not a zebra, this resistance to new information is a strong form of the limiting effects of bias on learning. A person may fail even to notice the zebra at the zoo because he was so strongly expecting to see a horse instead and was attuned to notice only that kind of animal.
Making matters even more complicated, two people who have different prior levels of expertise, or different beliefs, might legitimately have different interpretations when initially presented with the same information. But if sufficient additional information suggests a particular interpretation, they should converge on an answer, especially if the higher level of expertise is brought to bear.
Beliefs about human-caused global climate change are a good example of the biases that blind individuals to new evidence. Despite nearly universal consensus among climate scientists that global climate change is taking place and that this change is induced by humans’ behavior, a considerable proportion of adults in the United States do not accept these interpretations of the evidence. One might expect that higher levels of science literacy would be associated with greater agreement with the scientific consensus. However, Kahan and colleagues (2012) found that it is among the individuals with the highest levels of science literacy that the most stark polarization is apparent. Those who only seek out and attend to information consistent with their prior beliefs will create an “echo-chamber” that further biases their learning. Often this echo-chamber effect is socially reinforced, as individuals prefer to discuss the topic in question with others whom they know hold beliefs similar to their own.
Stereotypes perpetuate themselves through learned bias, but not all learning biases are considered to have negative consequences. For example, some positive biases promote well-being and mental health (Taylor and Brown, 1988), some may promote accuracy in perceptions of other people (Funder, 1995), and others may be adaptive behaviors—for example, selective attention and action in situations in which errors have a high cost (Haselton and Buss, 2000; Haselton and Funder, 2006). Hahn and Harris (2014) have written a useful historical overview of research on bias in human cognition.
Still other biases refine perception and serve to blur distinctions within categories that are not meaningful while highlighting subtle cross-category distinctions that may be important. For example, very young infants respond equally to phonological contrasts that matter in their language (e.g., “r” and “l” if the baby lives in an English-speaking context) and those that do not matter (e.g., “r” and “l” in a Japanese-speaking context). Over time, infants lose this discriminatory capability. This loss is actually a benefit, reflecting the baby’s increasing efficiency in processing his own language context, and is a mark of
learning (Kuhl et al., 1992). In the other direction, dermatologists may learn from experience and formal training to distinguish subtle features of moles and skin growths that signal malignancy, features that to an untrained eye are indistinguishable from those of benign growths.
Biases affect the noncognitive aspects of learning as well. In a variable world, highly stable task environments are not guaranteed and so training to high efficiency may actually create a mindset that makes new learning more difficult, impeding motivation and interest in continuous growth and development. For instance, a person who has learned how to organize her schedule using a specific tool may be reluctant to learn a new tool because of the perception that it will take too much time to learn to use it, even though it may be more efficient in the long run. In this example, it is not that the person is unable to learn the new tool; rather, her beliefs about the amount of effort required affect her motivation and interest in learning. This kind of self-attribution, or prior knowledge of oneself, can have a large influence on how people approach future learning opportunities, which in turn influences what they will learn (Blackwell et al., 2007).
We have seen that building a knowledge base requires doing three things: accumulating information (in part by noticing what matters in a situation and is therefore worth attending to); tagging this information as relevant or not; and integrating it across separate episodes. These three activities can happen relatively quickly and automatically, or they can happen slowly through deliberate reflection. However, these processes alone are not sufficient for integrating and extending knowledge. Learners of all ages know many things that were not explicitly taught or directly experienced. They routinely generate their own novel understanding of the information they are accumulating and productively extend their knowledge.
Inferential reasoning refers to making logical connections between pieces of information in order to organize knowledge for understanding and to drawing conclusions through deductive reasoning, inductive reasoning, and abductive reasoning (Seel, 2012). Inferential thinking is needed for such processes as generalizing, categorizing, and comprehending. The act of reading a text is a good example. To comprehend a text, readers are required to make inferences regarding information that is only implied in the text (see, e.g., Cain and Oakhill, 1999; Graesser et al., 1994; Paris and Upton, 1976). Some types of inferences help readers track the meaning of a text by integrating different information it supplies, for example by recognizing anaphoric
references (words in a text that require the reader to refer back to other ideas in the text for their meaning). Other types of inferences allow a reader to fill in gaps in the text by recruiting information from beyond it (i.e., background knowledge), in order to understand information within the text. Though these types of inferences are essential for understanding, they are thought to survive in working memory only long enough to aid comprehension (McKoon and Ratcliff, 1992).
Other inferences that learners make survive beyond the bounds of working memory and become incorporated into their knowledge base. For example, a person who knows both that liquids expand with heat and that thermometers contain liquid may integrate these two pieces of information and infer that thermometers work because liquid expands as heat increases. In this way, the learner generates understanding through a productive extension of prior learning episodes.
Effective problem solving typically requires retrieved knowledge to be adapted and transformed to fit new situations; therefore, memory retrieval must be coordinated with other cognitive processes. One way to help people realize that something they have learned before is relevant to their current task is to explicitly give them a hint that it is relevant (Gick and Holyoak, 1980). For example, such hints might be embedded in text, provided by a teacher, or incorporated into virtual learning platforms. Another strategy for helping people realize that they already know something useful is to ask people to compare related problems in order to highlight exactly what they have in common, increasing the likelihood that they will recall previously acquired knowledge with similar properties (Alfieri et al., 2013; Gentner et al., 2009).
Kolodner et al. (2003) gives the example of an architect trying to build an office building with a naturally lit atrium. She realizes that a familiar library’s design, which includes an exterior wall of glass, could be reused for the office building, but would fit the building’s needs better if translucent glass bricks were used instead of a clear, glass pane. This kind of design-based reasoning is incorporated into problem-based learning (Hmelo-Silver, 2004) activities. Problem-based learning emphasizes that memories are not simply stored to allow future reminiscing, but are formed so that they can be used, reshaped, and flexibly adapted to serve broad reasoning needs. The goal of problem-based learning is to instill in learners flexible knowledge use, effective problem-solving skills, self-directed learning, collaboration, and intrinsic motivation. These goals are in line with several of the goals identified in other contexts as important for success in life and work (National Research Council, 2012b).
People’s learning benefits from a steady increase, over many decades, in the accumulation of world knowledge (e.g., Craik and Salthouse, 2008;
Hedden and Gabrieli, 2004). This accumulation makes it easier for older adults not only to retrieve vocabulary and facts about the world (Cavanagh and Blanchard-Fields, 2002) but also to acquire new information in domains related to their expertise. For example, physicians acquire medical expertise, which enables them to comprehend and remember more information from medical texts than novices can (Patel et al., 1986). It is also thought that older adults can compensate for declines in some abilities by using their extensive world knowledge. For instance, medical experts depend less on working memory because they can draw on their expertise to reconstruct only those facts from long-term memory that are relevant to a current need (e.g., Patel and Groen, 1991).
The knowledge learners accumulate throughout the life span is the growing product of the processes of both learning new information from direct experience and generating new information based on reasoning and imagining (Salthouse, 2010). These two cognitive assets together—accumulated knowledge and reasoning ability—are particularly relevant to healthy aging. Reasoning and knowledge abilities tend to be correlated. That is, people who have comparatively higher reasoning capacity are likely to acquire correspondingly more knowledge over the life span than their peers (Ackerman and Beier, 2006; Beier and Ackerman, 2005). Reasoning ability is a major determinant of learning throughout life, and it is through reasoning, especially in contexts that allow people to pursue their interests, that people develop knowledge throughout their life span (Ackerman, 1996; Cattell, 1987).
On average, however, the trajectories of reasoning and knowledge acquisition are different across the life span. A number of research studies have described the general trajectories of age-related changes in ability, using a variety of measures and research designs (cross-sectional and longitudinal), and have shown a fairly consistent trend in which the development of knowledge remains steady as reasoning capacity (the ability to quickly and accurately manipulate multiple distinct pieces of factual information to make inferences) drops off (Salthouse, 2010). However, there is considerable individual variability in the trajectories, which reflect individual health and other characteristics, as well as educational and experiential opportunities and even social engagement. Yet, even though there is an average decline in inferential reasoning capacity through adulthood, there is not a corresponding decline in the ability to make good decisions—a more colloquial use of the word “reasoning.” In other words, the research does not suggest that the average 14-year-old reasons better about what to do in a complex or emotional real-world situation than would an average 50-year-old. Instead, it describes the 14-year-old’s stronger ability to quickly manipulate multiple distinct pieces of factual information to make logical and combinatorial inferences.
The growth or decline of abilities can be expected to vary not only between individuals but also within the same person over time (Hertzog et al.,
2008). Two 50-year-olds may have extremely different cognitive profiles, such that one may generally have the same ability profile as an average 30-year-old and the other may more closely resemble an average 70-year-old. Within the same person, abilities will decline or grow at varying rates as a function of that individual’s continuing use of some skills and intellectual development in particular domains; losses and declines are associated with disuse of other skills. (Factors that influence cognitive aging are discussed in Chapter 9.) As mentioned, new learning depends on both reasoning ability and knowledge acquisition (Ackerman and Beier, 2006; Beier and Ackerman, 2005). Even though reasoning abilities decline with age, knowledge accumulated throughout the life span facilitates new learning, as long as the information to be learned is aligned with existing domain knowledge. When people select environments for education, work, and hobbies that capitalize on their already-established knowledge and skills as they age, their selectivity allows them to capitalize on their repertoire of knowledge and expertise for learning new information (Baltes and Baltes, 1990).
Cognitive abilities change throughout the life span in a variety of ways that may affect a person’s ability to learn new things (see Hartshorne and Germine, 2015, for discussion). For instance, as people age, learning may rely more on knowledge and less on reasoning and quick manipulation of factual information. However, examining peoples’ cognitive abilities and learning becomes increasingly complex as people develop past the age of formal education. One reason is that the ways in which people learn become increasingly idiosyncratic outside of a standardized educational curriculum, and understanding this process requires assessing knowledge gained through a wide variety of adult experiences that different individuals amass over a lifetime (Lubinski, 2000). The unique complexities of adult learning and development are discussed in Chapter 8.
As described in Chapter 2, learning is inherently cultural, given that a person’s experiences in a culture affect biological processes that support learning, perception, and cognition. In the area of reasoning, for example, researchers have explored fundamental differences in peoples’ reasoning about three basic domains of life: physical events (naïve physics), biological events (naïve biology), and social or psychological events (naïve psychology) (see e.g., Carey, 1985, 2009; Goswami, 2002; Hirschfeld and Gelman, 1994; Spelke and Kinzler, 2007; also see Ojalehto and Medin, 2015c, for a review). These distinctions are compelling in the sense that each reflects a set of intuitive principles and inferences. That is, each domain is defined by entities having the same kind of causal properties. These might be marked, for example, by the way they move: physical entities are set into motion by external forces,
while biological entities may propel themselves. These domains are important for understanding cognition because researchers have suggested that whereas the perception of physical causality is universal, causal reasoning in the biological and psychological domains is culturally variable.
Two studies illustrate ways to examine these issues. Morris and Peng (1994) presented two types of animated displays to American and Chinese participants. One set of displays depicted physical interactions (of geometrical shapes), whereas the other set depicted social interactions (among fish). The participants’ answers to questions about what they had seen suggested differences in attention to internal and external causes across the groups, but those differences depended on the domain (social or physical). The authors concluded that attribution of causality in the social domain is susceptible to cultural influences but that causality in the physical domain is not.
Beller and colleagues (2009) asked German, Chinese, and Tongan participants to indicate which entity they regarded as causally most relevant for statements such as “The fact that wood floats on water is basically due to . . . ”. Ratings varied by the cultural background of respondents and also by the phenomena participants were considering. In general, the German and Chinese participants, but not the Tongan participants, considered a carrier’s capability for buoyancy only when the floater was a solid object, such as wood, but not when it was a fluid, such as oil (Beller et al., 2009; see also Bender et al., 2017). This is an area of research that has barely been explored, but results to date suggest that the perception of physical causality may in fact not be universal and may be learned in culturally mediated ways.
People are naturally interested in strengthening their ability to acquire and retain knowledge and in ways to improve learning performance. Researchers have explored a variety of strategies to support learning and memory. They have identified several principles for structuring practice and engaging with information to be learned to improve memory, to make sense of new information, and to develop new knowledge.
Several scholars have looked across the research on the effectiveness of specific strategies for supporting learning (Benassi et al., 2014; Dunlosky et al., 2013; Pashler et al., 2007). The authors of these three studies looked for strategies that (1) have been examined in several studies, using authentic educational materials in classroom settings; (2) show effects that can be generalized across learner characteristics and types of materials; (3) promote learning that is long-lasting; and (4) support comprehension, knowledge application, and problem solving in addition to recall of factual material. These three analyses identified five learning strategies as promising:
- retrieval practice;
- spaced practice;
- interleaved and varied practice;
- summarizing and drawing; and
- explanations: elaborative interrogation, self-explanation, and teaching.
The first three strategies are ways of structuring practice that are particularly useful for increasing knowledge retention.
Some evidence shows that the act of retrieval itself enhances learning and that when learners practice retrieval during an initial learning activity, their ability to retrieve and use knowledge again in the future is enhanced (Karpicke, 2016; Roediger and Karpicke, 2006b). The benefits of retrieval practice in general have been shown to generalize across individual differences in learners, variations in materials, and different assessments of learning. For example, researchers have found effects across learner characteristics in children (Lipko-Speed et al., 2014; Marsh et al., 2012). Studies have also suggested that retrieval practice can be a useful memory remediation method among older adults (Balota et al., 2006; Meyer and Logan, 2013; also see Dunlosky et al., 2013, for a review of effective learning techniques). However, most of this research has addressed retrieval of relatively simple information (e.g., vocabulary), rather than deep understanding.
Research has also demonstrated the effects of retrieval practice on recall of texts and other information related to school subjects. For example, Roediger and Karpicke (2006a) had students read brief educational texts and practice recalling them. Students in one condition read the texts four times; students in a second group read three times and recalled the texts once by writing down as much as they could remember; and students in a third group read the material once and then recalled it during three retrieval practice periods. On a final test given 1 week after the initial learning session, students who practiced retrieval one time recalled more of the material than students who only read the texts, and the students who repeatedly retrieved the material performed the best. The results suggest that actively retrieving the material soon after studying it is more productive than spending the same amount of time repeatedly reading.
Attempting retrieval but failing has also been shown to promote learning. Failed retrievals provide feedback signals to learners, signaling that they may not know the information well and should adjust how they encode the material the next time they study it (Pyc and Rawson, 2010). The act of failing to retrieve may thus enhance subsequent encoding (Kornell, 2014).
Such studies suggest that self-testing can be an effective way for students to practice retrieval. However, evidence from surveys of students’ learning strategies and from experiments in which learners are given control over when and how often they can test themselves suggests that students may not test themselves often or effectively enough (Karpicke et al., 2009; Kornell and Son, 2009). Many students do not engage in self-testing at all, and when students do test themselves, they often do so as a “knowledge check” to see whether they can or cannot remember what they are learning. While this is an important use of self-testing, few learners self-test because they view the act of retrieval as part of the process of learning. Instead, they are likely to retrieve something once and then, believing they have learned it for the long term, drop the item from further practice.
Researchers who have compared spaced and massed practice have shown that the way that learners schedule practice can have an impact on learning (Carpenter et al., 2012; Kang, 2016). Massed practice concentrates all of the practice sessions in a short period of time (such as cramming for a test), whereas spaced practice distributes learning events over longer periods of time. Results show greater effects for spacing than for massed practice across learning materials (e.g., vocabulary learning, grammatical rules, history facts, pictures, motor skills) (Carpenter et al., 2012; Dempster, 1996), stimulus formats (e.g., audiovisual, text) (Janiszewski et al., 2003), and for both intentional and incidental learning (Challis, 1993; Toppino et al., 2002). Studies have shown benefits of spaced practice for learners of ages 4 through 76 (Balota et al., 1989; Rea and Modigliani, 1987; Simone et al., 2012; Toppino, 1991). Cepeda and colleagues (2006) found that spaced practice led to greater recall than massed practice regardless of the size of the lag between practice and recall.
There are many possible reasons why spaced practice might be more effective than massed practice. When an item, concept, or procedure is repeated after a spaced interval, learners have to fully engage in the mental operations they performed the first time because of forgetting that has occurred. But when repetitions are immediate and massed together, learners do not fully engage during repetitions. In the case of reading, one possible reason why massed re-readings do not promote learning is that when people reread immediately, they do not attend to the most informative and meaningful portions of the material during the second reading, as illustrated by Dunlosky and Rawson (2005) in a study of self-paced reading.
A few researchers have attempted to identify the spacing intervals that promote the most memory—a “sweet spot” where spaced practice confers benefits before too much forgetting has occurred (Cepeda et al., 2008; Pavlik and Anderson, 2008). For example, a study of vocabulary learning among fifth
graders suggested that a 2-week interval showed the best results (Sobel et al., 2011). Another classroom-based study of spacing effects focused on first-grade children learning to associate letters and sounds during phonics instruction (Seabrook et al., 2005). The children who received spaced practice during the 2-week period significantly outperformed the children who received a single massed practice session each day.
In general, the literature on spaced practice suggests that separating learning episodes by at least 1 day, rather than focusing the learning into a single session, maximizes long-term retention of the material. However, it is important to note that wider spacing is not necessarily always better. The optimal distribution of learning sessions depends at least in part on how long the material needs to be retained in memory (i.e., when the material will be recalled or tested). For example, if the learner will be tested 1 month or more after the last learning session, then the learning should be distributed over weeks or months.
Interleaved and Variable Practice
The way information is presented can significantly affect both what is learned (Schyns et al., 1998) and how well it is learned (Goldstone, 1996). Variable learning generally refers to practicing skills in different ways, while interleaving refers to mixing in different activities. Varying or interleaving different skills, activities, or problems within a learning session—as opposed to focusing on one skill, activity, or problem throughout (called blocked learning)—may better promote learning. Both strategies may also involve spaced practice, and both also present learners with a variety of useful challenges, or “desirable difficulties.” Researchers have identified potential benefits of variable and interleaved practice learning, but they have also found a few benefits for blocked practice.
Several studies have shown benefits for blocking, at least for category learning (Carpenter and Mueller, 2013; Goldstone, 1996; Higgins and Ross, 2011). Moreover, when given the option, a majority of learners preferred to block their study (Carvalho et al., 2014; Tauber et al., 2013). Interleaving can boost learning of the structure of categories; that is, learning that some objects or ideas belong to the same category and others do not (Birnbaum et al., 2013; Carvalho and Goldstone, 2014a, 2014b; Kornell and Bjork; 2008). Other researchers have examined interleaved practice in mathematical problem-solving domains (Rohrer, 2012; Rohrer et al., 2015).
Carvalho and Goldstone (2014a) found that the effectiveness of the presentation methods (interleaved or blocked) depended on whether the participant engaged in active or passive study. They also found that interleaving concepts improved students’ capacity to discriminate among different categories, while blocked practice emphasized similarities within each category. These results
suggest that interleaved study improves learning of highly similar categories (by facilitating between-category comparisons), whereas blocked study improves learning of low-similarity categories (by facilitating within-category comparisons).
Interleaved study naturally includes delays between learning blocks and thus easily allows for spaced practice, which has the potential benefits for long-term memory discussed above. However, it may be beneficial because it helps learners to make comparisons among categories, not because it allows time to elapse between learning blocks (Carvalho and Goldstone, 2014b). The mechanisms that underlie the benefits of either interleaved or blocked study (e.g, possible effects on attentional processes) are ongoing topics of research. As with other strategies, the optimal way to present material—interleaved or blocked—and the mechanisms most heavily involved will likely depend on the nature of the study task.
The other two strategies for which there is strong evidence—summarizing and drawing and developing explanations—draw on inferential processes that research shows to be effective for organizing and integrating information for learning.
Summarizing and Drawing
Summarizing and drawing are two common strategies for elaborating on what has been learned. To summarize is to create a verbal description that distills the most important information from a set of materials. Similarly, when learners create drawings, they use graphic strategies to portray important concepts and relationships. In both activities, learners must take the material they are learning and transform it into a different representation. There are differences between them, but both activities involve identifying important terms and concepts, organizing the information, and using prior knowledge to create verbal or pictorial representations.
Both summarization and drawing have been shown to benefit learning in school-age children (Gobert and Clement, 1999; Van Meter, 2001; Van Meter and Garner, 2005). Literature reviews by Dunlosky and colleagues (2013) and Fiorella and Mayer (2015a, 2015b) have identified factors that appear to contribute to the effectiveness of summarization and drawing activities.
A few studies have suggested that the quality of students’ summaries and drawings is directly related to how much they learn from the activities and that learners do these activities more effectively when they are trained and guided (Bednall and Kehoe, 2011; Brown et al., 1983; Schmeck et al., 2014). For example, the effectiveness of drawing activities is enhanced when learners
compare their drawings to author-generated pictures (Van Meter et al., 2006). Similarly, providing learners with a list of relevant elements to be included in drawings and partial drawings helps learners create more complete drawings and bolsters learning (Schwamborn et al., 2010).
A group of researchers compared summarization and drawing and suggested that their effectiveness depends on the nature of the learning materials. For example, Leopold and Leutner (2012) asked high school students who were studying a science text about water molecules, which contained descriptions of several spatial relations, to either draw diagrams, write a summary of the text, or to re-read the text (the control condition). Those who created drawings performed better on a comprehension test than those who re-read the texts. However, those who created written summaries performed worse than those who re-read. The authors concluded that the drawing was more effective in this case because the learning involved spatial relations.
Note-taking, either writing by hand or typing on a laptop, is a form of summarizing that has also been studied. For example, Mueller and Oppenheimer (2014) found that students who hand-wrote notes learn more than those who typed notes using a laptop computer. The researchers asked students to take notes in these two ways and then tested their recall of factual details, conceptual understanding, and ability to synthesize and generalize the information. They found that students who typed took more voluminous notes than those who wrote by hand, but the hand-writers had a stronger conceptual understanding of the material and were more successful in applying and integrating the material than the typers. The researchers suggested that because writing notes by hand is slower, students doing this cannot take notes verbatim but must listen, digest, and summarize the material, capturing the main points. Students who type notes can do so quickly and without processing the information.
Mueller and Oppenheimer (2014) also examined the contents of notes taken by college students in these two ways across a number of disciplines. They found that the typed notes—which were closer to verbatim transcriptions—were associated with lower retention of the lecture material. Even when study participants using laptops were instructed to think about the information and type the notes in their own words, they were no better at synthesizing material than students who were not given the warning. The authors concluded that typing notes does not promote understanding or application of the information; they suggested that notes in the students’ own words and handwriting may serve as more effective memory prompts by recreating context (e.g., thought processes, conclusions) and content from the original lecture.
Encouraging learners to create explanations of what they are learning is a promising method of supporting understanding. Three techniques for doing this have been studied: elaborative interrogation, self-explanation, and teaching.
Elaborative interrogation is a strategy in which learners are asked, or are prompted to ask themselves, questions that invite deep reasoning, such as why, how, what-if, and what-if not (as opposed to shallow questions such as who, what, when, and where) (Gholson et al., 2009). A curious student who applies intelligent elaborative interrogation asks deep-reasoning questions as she strives to comprehend difficult material and solve problems. However, elaborative interrogation does not come naturally to most children and adults; training people to use this skill—and particularly training in asking deep questions—has been shown to have a positive impact on comprehension, learning, and memory (Gholson et al., 2009; Graesser and Lehman, 2012; Graesser and Olde, 2003; Rosenshine et al., 1996). For example, in an early study, people were asked either to provide “why” explanations for several unrelated sentences or to read and study the sentences. Both groups were then tested on their memory of the sentences. Those who asked questions performed better than the group that just studied the sentences (Pressley et al., 1987). Studies with children have also shown benefits of elaborative interrogation (Woloshyn et al., 1994), and the benefits of elaborative interrogation can persist over time (e.g., 1 or 2 weeks after learning), though few studies have examined effects of elaborative interrogation on long-term retention.
Most studies conducted by researchers in experimental psychology have used isolated facts as materials in studying the effects of elaboration and have assessed verbatim retention, but researchers in educational psychology have also looked at more complex text content and assessed inference making (Dornisch and Sperling, 2006; Ozgungor and Guthrie, 2004). For example, McDaniel and Donnelly (1996) asked college students to study short descriptions of physics concepts, such as the conservation of angular momentum, and then answer a why question about the concept (e.g., “Why does an object speed up as its radius get smaller, as in conservation of angular momentum?”). A final assessment involved both factual questions and inference questions that tapped into deeper levels of comprehension. The authors found benefits of elaborative interrogation for complex materials and assessments and also found that those who engaged in elaborative interrogation outperformed learners who produced labeled diagrams of the concepts in each brief text.
Self-explanation is a strategy in which learners produce explanations of material or of their thought processes while they are reading, answering questions, or solving problems. In the most general case, learners may simply be asked to explain each step they take as they solve a problem (Chi et al., 1989b; McNamara, 2004) or explain a text sentence-by-sentence as they read it (Chi
et al., 1994). Self-explanation involves more open-ended prompts than the specific “why” questions used in elaborative interrogation, but both strategies encourage learners to elaborate on the material by generating explanations. Other examples of this work include self-explanations of physics.
An early study of self-explanation was carried out by Chi and colleagues (1994). Eighth-grade students learned about the circulatory system by reading an expository text. While one group just read the text, a second group of students produced explanations for each sentence in the text. The students who self-explained showed larger gains in comprehension of concepts in the text. A subsequent study showed similar results (Wylie and Chi, 2014). Self-explanation has now been explored in a wide range of contexts, including comprehension of science texts in a classroom setting (McNamara, 2004), learning of chess moves (de Bruin et al., 2007), learning of mathematics concepts (Rittle-Johnson, 2006), and learning from worked examples on problems that require reasoning (Nokes-Malach et al., 2013). Self-explanation prompts have been included in intelligent tutoring systems (Aleven and Koedinger, 2002) and systems with game components (Jackson and McNamara, 2013; Mayer and Johnson, 2010). However, relatively few studies have examined the effects of self-explanation on long-term retention or explored the question of how much self-explanation is needed to produce notable results (Jackson and McNamara, 2013).
A few studies have explored the relationship between self-explanation and prior knowledge in learning (Williams and Lombrozo, 2013). For example, Ionas and colleagues (2012) investigated whether self-explanation was beneficial to college students who were asked to do chemistry problems. They found that prior knowledge moderated the effectiveness of self-explanation and that the more prior knowledge of chemistry the students reported having, the more self-explanation appeared to help them learn. Moreover, for students who had just a little prior knowledge, using self-explanation seemed to impede rather than support performance. The researchers suggested that learners search for concepts or processes in their prior knowledge to make sense of new material; when the prior knowledge is weak, the entire process fails. They concluded that educators should thoroughly assess the learners’ prior knowledge and use other cognitive support tools and methods during the early stages of the learning process, as learners strengthen their knowledge base.
Finally, teaching others can be an effective learning experience. When learners prepare to teach they must construct explanations, just as they do in elaborative interrogation and self-explanation activities. However, elaborative interrogation and self-explanation both require that the learner receive fairly specific prompts, whereas the act of preparing to teach can be more open-ended. Teaching others is often an excellent opportunity to hone one’s own knowledge (Biswas et al., 2005; Palincsar and Brown, 1984), and learners in this kind of interaction are likely to feel empowered and responsible in a
way that they do not feel when they are the passive recipients of knowledge (Scardamalia and Bereiter, 1993). Peers may be able to express themselves to each other in ways that are particularly relevant, immediate, and informative. Although peer learning and teaching are often quite effective, teachers and instructors typically come closer to injunctive norms and provide better models to observe.
A foundational study of the effects of teaching on learning by Bargh and Schul (1980) has served as a template for subsequent studies. Bargh and Schul asked participants to study a set of materials and either prepare to teach the material to a peer or simply study it for an upcoming test. Both groups were tested on the material without teaching it; only the expectation to teach had been manipulated. Students who prepared to teach others performed better on the assessment than students who simply read and studied the material. Effects of preparing to teach have been replicated in studies since Bargh and Schul’s foundational work (e.g., Fiorella and Mayer, 2014).
The benefits of teaching are evident in other contexts. For example, research on tutoring has shown that while students certainly learn by being tutored, the tutors themselves learn from the experience (see Roscoe and Chi, 2007). Reciprocal teaching is another strategy, used primarily in improving students’ reading comprehension (Palincsar, 2013;Palincsar and Brown, 1984). In reciprocal teaching, students learn by taking turns teaching material to each other. The students are given guidance: training in four strategies to help them recognize and react to signs of comprehension breakdown (questioning, clarifying, summarizing, and predicting) (Palincsar, 2013).
The research suggests several possible reasons why teaching may benefit learners. Preparing to teach requires elaborative processing because learners need to generate, organize, and integrate knowledge. Also, as mentioned, the explanations that people create may promote learning in the same way that elaborative interrogation and self-explanations promote learning. The process of explaining to others is active and generative, and it encourages learners to focus on deeper questions and levels of comprehension. Explaining in a teaching context also involves retrieval practice, as the teacher actively engages in retrieving knowledge in order to explain instructional content and answer questions. Although researchers have documented benefits of explanation, there are cautions to bear in mind. For example, a few researchers in this area have noted that in developing explanations learners may tend to make broad generalizations at the expense of significant specifics (Lombrozo, 2012; Williams and Lombrozo, 2010; Williams et al., 2013). Children tend to prefer a single explanation for two different phenomena (e.g., a toy that both lights up and spins), even when there are two independent causes (Bonawitz and Lombrozo, 2012). Likewise, when diagnosing diseases based on observable symptoms, adults tend to attribute the two symptoms to a single disease, even when it is more likely that there are two separate diseases (Lombrozo, 2007;
Pacer and Lombrozo, 2017). The tendency to prefer simple, broad explanations over more complex ones may affect what people learn and the inferences they draw. For each of the different types of explanation strategies, researchers have noted reasons for educators to plan carefully when and how they can be used most effectively.
Learners identify and establish relationships among pieces of information and develop increasingly complex structures for using and categorizing what they have learned. Accumulating bodies of knowledge, structuring that knowledge, and developing the capacity to reason about the knowledge one has are key cognitive assets throughout the life span.
Strategies for supporting learning include those that focus on retention and retrieval of knowledge as well as those that support development of deeper and more sophisticated understanding of what is learned. The strategies that have shown promise for promoting learning help learners to develop the mental models they need to retain knowledge so they can use it adaptively and flexibly in making inferences and solving new problems.
CONCLUSION 5-1: Prior knowledge can reduce the attentional demands associated with engaging in well-learned activities, and it can facilitate new learning. However, prior knowledge can also lead to bias by causing people to not attend to new information and to rely on existing schema to solve new problems. These biases can be overcome but only through conscious effort.
CONCLUSION 5-2: Learners routinely generate their own novel understanding of the information they are accumulating and productively extend their knowledge by making logical connections between pieces of information. This capacity to generate novel understanding allows learners to use their knowledge to generalize, categorize, and solve problems.
CONCLUSION 5-3: The learning strategies for which there is evidence of effectiveness include ways to help students retrieve information and encourage them to summarize and explain material they are learning, as well as ways to space and structure the presentation of material. Effective strategies to create organized and distinctive knowledge structures encourage learners to go beyond the explicit material by elaborating
and to enrich their mental representation of information by calling up and applying it in various contexts.
CONCLUSION 5-4: The effectiveness of learning strategies is influenced by such contextual factors as the learner’s existing skills and prior knowledge, the nature of the material, and the goals for learning. Applying these approaches effectively therefore requires careful thought about how their specific mechanisms could be beneficial for particular learners, settings, and learning objectives.