Understanding and Promoting Thinking About Knowledge

Philosophical Epistemology

In many of the most cited sources in our database, the authors drew from philosophical epistemology to define epistemic cognition and determine its scope (Chinn, Buckland, & Samarapungavan, 2011; Greene, Azevedo, & Torney-Purta, 2008; Hofer & Pintrich, 1997; Muis, Bendixen, & Haerle, 2006). In the first sentence of their landmark review of epistemic cognition research, Hofer and Pintrich (1997) defined epistemology as “an area of philosophy concerned with the nature and justification of human knowledge” (p. 88). This definition guided their conceptualization of epistemic cognition, and many others as well (e.g., Greene et al., 2008; Muis et al., 2006). Chinn et al.’s (2011) recent argument for an expansion of the scope of epistemic cognition research was based on their review of contemporary philosophical epistemology, in which they found that “epistemologists view their field much more broadly” (p. 145) than previous education researchers had suggested. Clearly, philosophical epistemology has influenced, and continues to influence, education research on epistemic cognition.

It is far beyond the scope of this chapter to comprehensively review epistemology, a field with roots stretching back to Plato. ¹ Instead, we describe how ideas from “traditional” epistemology (Rysiew, 2016) have informed the scope and foci of epistemic cognition research. Then, we discuss how more recent “naturalized” epistemology scholarship has pushed researchers to consider more social and situated views of epistemic cognition.

Education and Psychology

Education has always been concerned with how people come to understand the nature of knowledge and develop strategies to justify belief. Plato’s The Republic (1991) outlined an epistemology of ideal knowledge and an educational model to obtain it; the Socratic dialogues articulated a view on how knowledge could be obtained and justified. In the modern era, Dewey (1910) expounded on the value of developing “scientific” habits of mind, and modern psychologists have routinely explored how insights from the study of knowledge generation and application are relevant in educational settings. Here we take a historical approach to highlight salient models, findings, and trends in broad strands of research related to epistemic cognition: personal epistemology research as it has been conducted within educational and developmental psychology, scholarship on epistemic aspects of cognitive development from developmental psychology, and disciplinary education research.

Disciplinary Education

Education in the disciplines of math, science, and history has been concerned with the epistemological aspects of these disciplines for decades. Historically, educational psychology research in school subjects has focused on epistemic beliefs and their relation to learning processes and academic outcomes (e.g., Depaepe, De Corte, & Verschaffel, 2016; Qian & Alvermann, 1995; Schommer et al., 1992). Researchers in the learning sciences have focused instead on relations between engagement in disciplinary practices (e.g., scientific inquiry, historical argument) and understanding of disciplinary knowledge and disciplinary epistemologies (e.g., Rosebery, Warren, & Conant, 1992; Schoenfeld, 1992; Wineburg, 1991). There are a number of recent reviews of research related to epistemic cognition in each of these disciplines, including the long-standing research on conceptions of the nature of science (Elby, Macrander, & Hammer, 2016; Lederman, 2007), and on epistemic cognition in mathematics (Depaepe et al., 2016; Muis, 2004) and history (VanSledright & Maggioni, 2016). Here we summarize trends from these disciplinary lines of research as they apply to the broader study of epistemic cognition.

One feature of epistemic cognition research in the disciplines is that it has been framed within discipline-specific epistemological concerns. Sometimes these overlap with the more general conceptions from personal epistemology research; for example, the concern in science education that students understand the tentativeness of science knowledge maps onto the general dimension of certainty of knowledge. Yet, within disciplinary education, the epistemologies of the disciplines are fleshed out more deeply and specifically than is typically the case in the psychology or personal epistemology literatures. Science educators link tentativeness to the theory-laden nature of observation in science (Lederman, 2007), while asserting that some claims are more certain than others (Osborne, Collins, Ratcliffe, Millar, & Duschl, 2003). Historians locate uncertainty in the distance between the present and the past and the ambiguity of historical objects (VanSledright & Maggioni, 2016). In mathematics, the issue of certainty presents as whether one believes mathematical procedures are given or constructed (Depaepe et al., 2016).

The later half of the 20th century saw in all these academic disciplines a growing recognition that students need to learn the epistemology of a discipline in addition to its theories, concepts, and facts (Duschl, 2008; Schoenfeld, 1992; Wineburg, 1991). This concern has driven a range of efforts to assess students’ understanding of specific epistemological conceptions within the disciplines, as well as to examine how constructs from personal epistemology relate to learning in the disciplines. At the same time, researchers have become increasingly interested in characterizing how conceptions of knowledge and knowing in the disciplines are related to particular sense-making practices in those disciplines. A range of studies have shed light on beliefs and thinking about knowledge and knowing as situated within disciplinary activities and, thus, peculiar to particular disciplines (e.g., Cobb, Stephan, McClain, & Gravemeijer, 2001; De Corte, Op’t Eynde, Depaepe, & Verschaffel, 2010; diSessa, Elby, & Hammer, 2002; Greene & Yu, 2014; Hammer, 1994; Maggioni, VanSledright, & Alexander, 2009; Samarapungavan et al., 2006; Sandoval & Çam, 2011; Wineburg, 1991).

With respect to broader debates about the generality or specificity of epistemic cognition, research in the disciplines has made a strong case for disciplinary specificity tied directly to the obviously different epistemological perspectives among the disciplines. Empirical research on learning across disciplines is starting to show that students appreciate, for example, that evidence in science and history looks different (Herrenkohl & Cornelius, 2013). Likewise, across disciplines, elementary and high school students ask different types of questions when seeking to understand texts (Portnoy & Rabinowitz, 2014). For epistemic cognition research broadly, it is crucial to recognize that disciplines have developed their own epistemologies in response to the particular questions and problems of knowledge development they have faced. The likelihood of epistemic cognition predicting academic performance in these fields depends on the degree to which individuals can adopt and use discipline-specific epistemic norms and practices (Goldman, 2011; Greene, 2016).

The State of Epistemic Cognition Research

The past half-century has seen research on epistemic cognition grow immensely in a number of fields. The disparate roots of research into how people generate knowledge, interpret knowledge, and think about the nature of knowledge have begun, in the past 10 to 20 years, to converge in some ways. Current models are explicitly informed by relevant work in the field of philosophical epistemology (Chinn et al., 2011; Greene et al., 2008; Muis et al., 2006). Contemporary models of epistemic cognition also pay more attention to the antecedents and early forms of epistemic cognition suggested by work in developmental psychology (Burr & Hofer, 2002; Hofer & Pintrich, 1997). Earlier debates about the specificity or generality of epistemic beliefs have mostly given way to a recognition that epistemic cognition has contextual and disciplinary aspects that are only somewhat understood (Sandoval, 2012), creating a need to better explain both the extent and mechanisms of generalization of epistemic cognition. The historical trends in these fields are converging to create a broader conceptualization of epistemic cognition, albeit not without some growing pains (Chinn et al., 2011; Hofer & Bendixen, 2012; Sandoval, 2012). Conceptual expansion of the idea of epistemic cognition has brought with it questions about the appropriate boundaries of the concept (e.g., Alexander, 2016). We believe these questions are important. To answer them, researchers must begin to identify and wrestle with the differences in how they conceptualize, measure, and intervene in people’s engagement with knowledge and knowing.

Conceptualizations of Knowledge and Cognition

One of the major intellectual shifts in the social sciences over the past decades of the 20th century and into the 21st has been the cultural turn (Bruner, 1996). In psychology, education, and philosophy, there has been a shift from viewing thinking and knowing as fundamentally individual activities to viewing them as fundamentally social (e.g., naturalized epistemology; Goldman, 1994; Quine, 1969). This shift, however, has been taken up to different extents, and differences in where scholars, and even fields, see the locus of explanation of epistemic cognition on this individual–social continuum creates a tangle of empirical results and interpretations that have yet to be clarified.

Piaget’s (1972) genetic epistemology and stage theory of development directly influenced epistemic cognition research from the start. This view of knowledge construction as an individual achievement continues to underlie much of the psychology research pertinent to epistemic cognition. The term personal epistemology as a moniker for this research itself conveys the assumption that persons construct epistemologies as part of their efforts to interpret the world around them. The constructivist abandons the idea that knowledge is an object in the world. Instead, “knowledge refers to conceptual structures that epistemic agents, given the range of present experience within their tradition of thought and language, consider viable” (Glasersfeld, 1989, p. 124). That is, while cultural traditions of thought and language may shape experience, individual epistemic agents build conceptual structures that can viably be considered knowledge. This is not to say that constructivists necessarily abandon the idea of an objective reality (see Phillips, 1995, for a discussion of variants of constructivism on this point), only that our knowledge of reality is constructed. The cognitive constructivism of Piaget locates that knowledge as a construct of individual minds.

Several models of epistemic cognition maintain this cognitive constructivism as the underlying definition of knowledge. Piagetian (e.g., King & Kitchener, 1994; Perry, 1970) and neo-Piagetian (e.g., Kuhn et al., 2000) models propose progressive trajectories that vary mainly in the generality they posit for the stages along the way. Belief systems models (e.g., Hofer, 2004; Schommer-Aikins, 2004) articulate dimensions of belief and their possible influences and interactions. All of these models share an assumption that cognitive representations (i.e., knowledge and belief) are constructed by individuals, via some level of reflection on experience, and then applied to subsequent experiences.

Sociocultural theories of cognition and development derived from Vygotsky (1962, 1978) locate the individual knower within a specific cultural and historical context, thereby framing knowledge as a cultural product and knowing as a cultural activity. Models of epistemic cognition that have been influenced by sociocultural perspectives vary in where they locate knowledge and knowing. What we can call a cognitive contextualist view (see especially the next section), attempts to account for how social factors influence individual knowledge construction, including individuals’ ideas about knowledge and knowing. Bendixen and colleagues (Bendixen & Rule, 2004; Muis et al., 2006), for example, have proposed models that attempt to account for how environmental (i.e., social) factors and features of specific domains influence individual epistemic cognition. Muis et al. (2006) have proposed that general epistemic beliefs develop within the social and cultural context in which people grow up. Hammer and Elby (2002) have suggested that participation in culturally shaped patterns of activity lead to the construction of epistemic “resources,” fine-grained cognitive structures tied closely to contexts of use. These contexts of use become bound into epistemological “frames” that organize the conditional activation of epistemic resources (Elby & Hammer, 2010). That is, in contrast to Muis et al. (2006), the epistemic resources model suggests that generalization is an outcome of repeated use.

A stronger social perspective on knowledge and knowing derives from situative perspectives on cognition (Engeström, 1987; Lave & Wenger, 1991), allied with the social epistemologists discussed above (Fuller, 1988; Goldman, 2011), who see knowledge as fundamentally a social construct and epistemology as a social phenomenon. The situative perspective takes as its base unit of analysis the individual-in-interaction, emphasizing features of interaction external to individuals and their minds. This interactional view on cognition sees knowledge as distributed among social and material resources and deployed within cultural activity. Thus, individuals’ and groups’ interpretations of the purpose of an activity are seen as shaping the epistemic cognition that might occur. Chinn and colleagues’ (Chinn et al., 2011; Chinn et al., 2014) model thus highlights processes of epistemic cognition as tied to the epistemic aims an individual might pursue during activity, how those aims relate to nonepistemic aims, and how aims are related to reliable processes of knowledge development. These reliable processes are social in nature: They are social practices developed within particular communities to solve specific epistemic problems (Chinn & Rinehart, 2016). The situative perspective on epistemic cognition thus sees knowledge as reflected in practice—in how people accomplish activity—rather than as a mental representation of some aspect of reality that is then applied (Kelly, 2016).

In sum, the divergent perspectives on knowledge and cognition generally are reflected in current models of epistemic cognition, not surprisingly. The crucial difference among these perspectives with respect to epistemic cognition research is that they theorize knowledge itself, its development, and related concepts such as belief and the role of context, rather differently. Such differences have implications for the unit of analysis in epistemic cognition research (e.g., individuals’ construction of knowledge propositions vs. the interplay of cultural forms of knowledge and individuals’ practice) as well as for perspectives on how epistemic cognition develops (e.g., a focus on changes in individuals’ conceptions of knowledge vs. changes in a community’s norms and practices and individual participation in those). The parallels between personal epistemology and disciplinary education models of epistemic cognition, and traditional and naturalized epistemology, are clear.

Whether such differences are incommensurate is perhaps an ideological question. Packer and Goicoechea (2000) argued that cognitive constructivism and the situative perspective are ontologically incommensurate, whereas Greeno (2015) suggested they address different levels of explanation. It is clear, however, that broad differences in how knowledge and knowing are theorized produce differences in how issues of contextualization and generality are addressed in research on epistemic cognition. They lead to different explanations of the nature of developmental pathways for epistemic cognition. They also, naturally, produce differences in research methods for identifying, defining, and studying phenomena of epistemic cognition.

Generality and Context in Epistemic Cognition

Hofer and Pintrich (1997) noted that the developmental and belief system models they critiqued paid only marginal attention to the issue of domain specificity versus domain generality in personal epistemology, with an underlying assumption that beliefs and thoughts about knowledge and knowing were largely consistent across domains and disciplines. Hofer and Pintrich (1997) discussed the possibility that both domain-specific and domain-general beliefs about knowledge and knowing existed, while suggesting there might be specific ways of thinking about knowledge and knowing within academic disciplines that made general models of personal epistemology less applicable. The idea that epistemic cognition occurs at different levels of specificity, as well as the idea that there are discipline-specific forms of epistemic cognition, has gained empirical support and been reflected in models of personal epistemology and epistemic cognition. Articles on domain or disciplinary differences in epistemic cognition were frequent among the most cited sources in our data set (e.g., Buehl & Alexander, 2001; Hofer, 2000; Kuhn et al., 2000; Muis et al., 2006).

To study generality and context, researchers working within the belief system paradigm of personal epistemology conducted between-subjects investigations (e.g., Jehng, Johnson, & Anderson, 1993; Lonka & Lindblom-Ylänne, 1996), where students majoring in different domains were compared with respect to their epistemic beliefs. They also conducted within-subject investigations (e.g., Buehl et al., 2002; Hofer, 2000), where the same students were asked about their epistemic beliefs in different domains. In general, this work supported the idea that epistemic beliefs varied as a function of academic domain and, moreover, that such variation was related to perceptions of domain structuredness. That is, researchers found that students viewed knowledge as more certain and integrated, and that they more readily accepted experts as sources of knowledge, in well-structured domains, such as mathematics, than in ill-structured domains, such as history. At the same time, however, this line of research seemed to indicate that students’ epistemic beliefs were not solely domain-specific, meaning that students could simultaneously hold both domain-specific and more domain-general or overarching beliefs about knowledge and knowing (Buehl & Alexander, 2001; Muis et al., 2006).

Recent research has started to target epistemic beliefs at a topic-specific level, that is, as beliefs about more delimited subject areas (e.g., World War II) that can be subsumed under an academic domain (e.g., history). This research has led to the use of topic-specific questionnaires (Bråten, Gil, Strømsø, & Vidal-Abarca, 2009; Stahl & Bromme, 2007; Trautwein & Lüdtke, 2007) or think-aloud methodology (Ferguson, Bråten, & Strømsø, 2012; Hofer, 2004; Mason, Ariasi, & Boldrin, 2011; Mason & Boldrin, 2008; Mason, Boldrin, & Ariasi, 2010) to assess students’ beliefs about specific topics along different belief dimensions (e.g., the certainty and simplicity of knowledge). For example, such topic-specific epistemic beliefs have been shown to uniquely predict comprehension when students read texts about the same topics (Bråten & Strømsø, 2010; Bråten, Strømsø, & Samuelstuen, 2008; Strømsø, Bråten, & Samuelstuen, 2008).

In the past decade, the empirical work in personal epistemology summarized above has been reflected in theoretical frameworks that take the multilevel nature of epistemic beliefs into account. For example, in their theory of integrated domains in epistemology framework, Muis et al. (2006) posited that beliefs concerning the certainty, simplicity, and source of knowledge, as well as the justification for knowing, may exist at the level of general epistemic beliefs that develop in nonacademic contexts, domain-general academic beliefs that develop in educational contexts, and domain-specific academic beliefs that reflect the various domains of knowledge that students experience. Muis, Trevors, Duffy, Ranellucci, and Foy (2016) have acknowledged that an additional level of specificity may be required beyond the domain-specific level, such as topic-specific beliefs about knowledge and knowing. In a more parsimonious model, at least with respect to levels of specificity, Greene et al. (2008) proposed that beliefs regarding the simplicity and certainty of knowledge and the justification for knowing vary at a level between domain specificity and domain generality, mainly differentiated between well-structured (e.g., mathematics and physics) and ill-structured (e.g., history and literature) domains. In more recent work, Greene and Yu (2014) used qualitative interview data from novices and experts to critique prior conceptualizations of domain specificity in personal epistemology research; they called for greater attention to differences in knowledge and knowing across disciplines.

Although research has indicated that students’ beliefs about knowledge and knowing are contextualized, in the sense that they differ across domains and topics within domains, this work is limited by the fact that it has essentially used the same questionnaires targeting the same belief dimensions across domains and topics. As a result, mean differences between scores on those dimensions across domains have been used to indicate domain specificity in epistemic beliefs, and correlations between scores on those dimensions across domains have been used to indicate domain generality in such beliefs (e.g., Buehl et al., 2002; Hofer, 2000; Muis et al., 2016). Also, most research using think-aloud methodology to assess a topic-specific level of epistemic cognition (e.g., Ferguson et al., 2012; Mason et al., 2011) has coded the resulting verbal protocols in terms of dimensions from epistemic belief system models, such as the Hofer and Pintrich (1997) multidimensional framework. Given that those measures presumed, a priori, the existence of a common set of dimensions or beliefs across domains, it could be argued that what has been captured within such research is not really domain- or discipline-specific epistemic beliefs and thinking but, rather, the extent to which scores on general belief dimensions vary across domains and disciplines (Elby et al., 2016; Hofer, 2006). To better understand patterns of epistemic beliefs and thinking within different disciplines, other approaches may be needed.

As indicated above, perspectives from disciplinary education suggest contextual differences both between and within disciplines. Researchers representing such perspectives argue that students’ epistemic beliefs and thinking should be understood as situated within individual and collective disciplinary learning activities (Sandoval, 2012, 2015). Some research suggests that variations in epistemic aims and virtues may also be discipline specific or even situation specific. In an intriguing study in history, Gottlieb and Wineburg (2012) showed that historians’ epistemic cognition, specifically the criteria they used for evaluating claims and evidence, depended on the relationships between the documentary material under consideration and the values and commitments of the historians. The same historians could switch their thinking between academically and religiously based forms of epistemic cognition, even within the same text. It seems highly unlikely that such disciplinary and situational variation can be captured by a small set of generally applicable dimensions, an assumption on which much research on the specificity-versus-generality issue within personal epistemology seems to rest (Greene & Yu, 2014).

Arguably, the theoretical framework best aligned with the discipline-specific, situated view discussed above is the contextualist resource framework of Hammer and Elby (2002; Elby & Hammer, 2010), which focuses on variations in epistemic cognition in response to situational demands. This means that learners are conceived of as responding adaptively or maladaptively to specific tasks by activating and deactivating existing fine-grained epistemic resources. According to Elby and Hammer (2010), the activation of fine-grained epistemic resources in a specific context over time may form a stable network with belief-like qualities, an “epistemological frame” that is typically activated in that context and sometimes even generalized across contexts.

Likewise, Chinn and colleagues (Chinn et al., 2011; Chinn et al., 2014; Chinn & Rinehart, 2016) have argued for a fine-grained, context-specific approach to epistemic cognition, where epistemic cognition varies across situations and may predict situation-specific patterns of learning. At the same time, Chinn et al.’s framework is a multicomponential model that expands on previous multidimensional personal epistemology models, with a fixed set of epistemic cognition components seen as applicable across disciplines and situations. How epistemic cognition plays out within those components is viewed as largely dependent on contextual factors, however, allowing for essential variability across and within specific disciplines.

Finally, Barzilai and Zohar’s (2014) multifaceted framework, distinguishing between cognitive and metacognitive aspects of epistemic thinking, can be said to reflect recent developments toward further contextualization and situativity in epistemic cognition research. In particular, they proposed that the cognitive aspect of epistemic thinking, concerning specific information, knowledge claims, sources, and epistemic strategies, may be highly specific in the sense that it varies with task contexts. The metacognitive aspect of epistemic thinking, however, is viewed as having more pronounced domain-general qualities within this framework. Epistemic metacognition involves, among other things, knowledge about what makes a source reliable, but epistemic cognition involves using that knowledge to evaluate a particular source (see also Barzilai & Zohar, 2012).

In sum, a disciplinary view of epistemic cognition implies that general models of personal epistemology, as well as models that operate with a fixed set of dimensions across domains and disciplines, do not sufficiently capture important differences in the epistemic norms and practices across the disciplines. Such critiques imply the need for potentially extensive model revision, which can be interpreted as another fault line between the fields. To build theory bottom-up from such research, a first step could be to produce more local models applicable to a limited range of learners and contexts. In turn, such models might be combined to produce more overarching frameworks for epistemic cognition that are better grounded in actual knowledge construction and evaluation within the disciplines (Elby et al., 2016). Such models may then be used to identify commonalities or convergences in epistemic cognition across contexts.

Developmental Pathways

Despite the consistency of proposed developmental trajectories in models of epistemic cognition described earlier, evidence for a clear developmental pathway has been inconsistent, and multiple lines of evidence undermine the notion of omnibus stagelike epistemic worldviews (Hofer & Bendixen, 2012; Sandoval, 2012). This has led to proposals that people develop domain-specific epistemic theories (Hofer, 2006) that follow the general trend identified by Kuhn et al. (2000). This line of theorizing argues that such domain-specific theories develop at variable rates, dependent on personal experiences. Within domains, however, scholars such as Hofer, Kuhn, and others (e.g., Hallett, Chandler, & Krettenauer, 2002) have argued for recapitulation of the same developmental trend from absolutist to evaluativist.

A problem for developmental models is that “evidence for just about every conceivable sort of epistemic development has been shown to characterize persons of just about every conceivable age” (Chandler, Hallett, & Sokol, 2002, p. 161). This weight of competing evidence and interpretations about epistemic development led Chandler et al. (2002) to argue that epistemic development is recursive, that people pass through epistemic levels more than once within and across domains of epistemic cognition. At the moment, it is not clear what evidence can determine the viability of the developmental recursion hypothesis. Elby et al. (2016) argued that the recursion proposal is similar to the proposal of planetary “epicycles” as an effort to rescue Ptolemaic theories of celestial motion. They suggested that their epistemic resources model provides a more parsimonious account, arguing that development is nothing more or less than the construction and coordination of epistemic resources through experience in the contexts of activity in which their use is productive.

At issue is that what appears to be a broad developmental trend may be an artifact of researchers’ own assumptions and efforts to document that trend. There is a need to account for a broad range of evidence that is at least potentially inconsistent with the reigning developmental hypothesis. Notions of epistemic development from Perry (1970) forward have been rooted in a distinction that sees development as a slow, broad, maturational process and learning as a more rapid, specific, enduring adaptation to environmental influences. This distinction is problematic for epistemic development for several reasons. First, epistemic change, whether characterized as development or not, may depend on systematic, organized schooling, which can hardly be said to be a natural maturational process. Second, describing epistemic change as independent of maturation seems to overlook the possibility that interventions to change epistemic cognition may be more or less successful depending on more general developing patterns of epistemic thinking in individuals. Finally, specific changes in epistemic cognition from particular interventions may generalize to influence epistemic developmental trajectories as described by Perry (1970) and other developmentalists (e.g., Baxter Magolda, 1992; King & Kitchener, 1994; Kuhn et al., 2000).

Much more conceptual and empirical work is needed to distinguish models and mechanisms of epistemic change from epistemic development, to the extent that scholars continue to see them as separate, albeit related, processes. Mechanisms of change have been proposed, from Hammer and Elby’s model of activation of epistemic resources and their stabilization into epistemological frames from repeated activation in context (Elby & Hammer, 2010), to Bendixen and colleagues’ constructs of epistemic doubt, volition, and resolution strategies (Bendixen & Rule, 2004; Muis et al., 2006). More work is needed to clarify proposed mechanisms of change, including how components of such mechanisms are involved in specific instances of epistemic change, as well as the ways in which changes in epistemic cognition may generalize. Research is needed that looks at epistemic cognition across multiple contexts over longer periods of time, to connect microgenetic processes to ontogenetic ones (Sandoval, 2014).

Research Design and Methods

Research methods and designs for capturing epistemic cognition vary a great deal across the fields engaged in this work, but can be roughly differentiated based on the assessment methods used, the contexts in which the work is done, and the kinds of epistemic cognition that are targeted. The most controversial type of assessment methods used are those that rely on self-report. The majority of self-report instruments require participants to rate their degree of agreement or disagreement with various statements by using Likert-type response scales. The first such instrument was Schommer’s (1990) EQ, which was followed by various iterations and variations of that instrument (e.g., Epistemic Belief Inventory; Schraw, Bendixen, & Dunkle, 2002), as well as instruments that require participants to make semantic differential judgments, such as Connotative Aspects of Epistemological Beliefs (CAEB; Stahl & Bromme, 2007). These instruments, most commonly used by adherents of dimensional models of epistemic cognition (e.g., Hofer & Pintrich, 1997; Schommer-Aikins, 2004), have been criticized for their failure to replicate posited relations between items and proposed factors (Clarebout, Elen, Luyten, & Bamps, 2001; DeBacker et al., 2008) and for their reliance on response formats (i.e., Likert-type response scales) better suited for the measurement of attitudes than beliefs or practices (Carifio & Perla, 2007; Hofer & Sinatra, 2010).

Aside from psychometric concerns, self-report assessment methods have been criticized as inauthentic, requiring respondents to construct or make explicit aspects of their epistemic cognition that many scholars believe to be largely tacit, complex, and contextual (Chinn & Rinehart, 2016; Sandoval, 2012; Sinatra & Chinn, 2011). Finally, despite attempts to use self-report instruments to examine within-person differences in epistemic cognition across disciplines (e.g., Buehl et al., 2002; Greene et al., 2008), there are many researchers, predominantly those who study epistemic cognition within disciplinary education, who claim these instruments require assumptions that epistemic cognition is traitlike rather than a function of context (Elby & Hammer, 2010). To some degree, this claim about the inherent assumptions regarding the use of self-report instruments is borne out in practice: The vast majority of studies conducted with self-report instruments like the EQ or the Epistemic Belief Inventory involve designs where participants complete self-report instruments in artificial contexts (e.g., completing a survey individually, without being asked to engage in a learning task), as opposed to more authentic contexts that actually invoke epistemic cognition (e.g., arguing about the meaning of experimental data; Sandoval, 2012).

Various forms of semistructured interviews have been used to measure epistemic cognition, beginning with Perry’s (1970) seminal work and continuing through many of the studies on developmental models of epistemic cognition (Mason, 2016). For example, such interviews were used extensively over the course of King and Kitchener’s (2004) scholarship on their reflective judgment model; they involved presenting participants with ill-structured problems about which “reasonable people reasonably disagree” (p. 5), such as whether news reporting is accurate or chemical additives to food are safe. These interviews were concurrent, meaning that a scenario or example was presented to participants, who then had to respond to interview probes about it, often “without practice or even much time to collect their thoughts” (King & Kitchener, 2002, p. 57). Like the majority of research using self-report instruments, studies involving concurrent semistructured interviews tend to be conducted outside actual learning or practice contexts.

Recently, there has been an increase in the number of studies designed to capture aspects of developmental and dimensional models of epistemic cognition while in the context of thinking about everyday problems or scenarios. For example, there has been an increase in the use of concurrent think-aloud protocols (Ericsson & Simon, 1993) to surface participants’ epistemic cognition as they engage in everyday epistemic tasks such as searching the Internet for information (e.g., Barzilai & Zohar, 2012; Ferguson et al., 2012; Greene, Yu, & Copeland, 2014; Hofer, 2004; Mason et al., 2010, 2011). Cognitive interviews have also been used to assess the ways in which people understand the items on survey instruments (Greene, Torney-Purta, Azevedo, & Robertson, 2010), as well as the ways in which their thinking might challenge assumptions of dimensional and developmental models of epistemic cognition (Greene & Yu, 2014; Muis, Duffy, Trevors, Ranellucci, & Foy, 2014). Importantly, however, the majority of this research involving concurrent measures of epistemic cognition has been conducted outside of formal learning (i.e., classrooms) or practice contexts.

On the other hand, many authors from a disciplinary education background have argued that epistemic cognition should be studied in authentic practice contexts, such as when students conduct experiments or construct models (Elby & Hammer, 2010; Kelly, 2016; Sandoval, 2005). Observation and discourse analysis are two predominant nonintrusive assessment methods that allow data collection while students engage in authentic tasks (Kelly, 2016). The data that arise from these techniques are often context-specific and interpreted solely within a single model of epistemic cognition, which makes it difficult to make inferences from them to other developmental or dimensional models (Sandoval, 2005).

Observation and discourse analysis can be done nonintrusively, but some researchers have argued that the addition of retrospective interviews to these designs can lead to more informed understanding of people’s epistemic cognition (Chinn et al., 2011; Sandoval, 2005). For example, Pluta, Chinn, and Duncan (2011) assessed epistemic cognition by first asking participants to engage in scientific inquiry and then prompting them to reflect on their epistemic practices and ideals. Data from these retrospective interviews can be triangulated with observation and discourse analysis data gathered during practice to understand both enacted epistemic practices and the epistemic aims and ideals that students bring to those practices (e.g., Chinn & Rinehart, 2016; Sandoval, 2012). Hammer and Elby have studied their epistemological resources model by triangulating across observational studies of activity (Rosenberg, Hammer, & Phelan, 2006) and cognitive interview studies (diSessa et al., 2002).

There are a number of scholars who examine how people think about knowledge and knowing across and outside of formal learning contexts (Bricker & Bell, 2016). While these researchers may not identify their work as epistemic cognition research, this area of scholarship makes an important, unique contribution to the field. The models of epistemic cognition providing the foundation for this area of scholarship vary, from developmental models (e.g., Weinstock, 2011) to epistemological resources models (e.g., Bang & Medin, 2010). As would be expected given the research focus, these scholars study epistemic cognition in a wide variety of contexts, including courtrooms (Weinstock, 2016), Internet browsing (Strømsø & Bråten, 2010), and museums (Bell, Bricker, Reeve, Zimmerman, & Tzou, 2012). Capturing epistemic cognition in situ often requires open-ended assessment methods and nonintrusive research designs, such as observations and case studies, although some of this research includes interviews either during or after observing epistemic cognition in practice. Research across contexts has revealed how people enact epistemic cognition differently in response to these contexts (Goldman, 2011; Gottlieb & Wineburg, 2012). Greene (2016) argued that within-person differences in enacted epistemic cognition may be due, in part, to people’s perceptions of the salient epistemic system, or systems, across contexts (Goldman, 2011). Students who are hesitant to enact epistemic practices in the classroom may do so eagerly in a museum, and their perceptions of the epistemic systems at play may hold the key to understanding these differences in performance across contexts. Scholarship on these contextual differences illustrates how multiple assessment methods are needed to capture not only how epistemic cognition happens but also how features of context are perceived as relevant (Bricker & Bell, 2016).

Our review of research designs and methods in epistemic cognition has revealed a number of trends concerning the assessment methods used, the contexts in which research is conducted, and the kinds of epistemic cognition under study. Continued psychometric problems with self-report methods of assessing epistemic cognition, coupled with the often-decontextualized or inauthentic contexts in which these data are gathered, and their presumptions of domain consistency or traitlike aspects of epistemic cognition, cast doubt on research findings from these studies. The growing acceptance that epistemic cognition is largely tacit, and that people’s enactment of epistemic cognition can often be different from their expressed reconstruction of that cognition, has led an increasing number of researchers to pursue concurrent assessment methods (e.g., observations, discourse analysis, think-aloud protocols) that can be implemented in contexts where participants are asked to think about knowledge and knowing (e.g., inquiry activities, searching the Internet for information on controversial issues). Researchers’ models of epistemic cognition seem to predict their choice of context, with adherents to dimensional or developmental models often investigating how people enact epistemic cognition outside of practice settings, and adherents to situative models of epistemic cognition more often using various kinds of inquiry contexts as research settings (Mason, 2016). An important question being studied by researchers of everyday epistemic cognition is the degree to which research in one of these contexts (e.g., classrooms, museums, courtrooms) can connect to, or inform, research in other contexts. Regardless of method, research design context, or underlying model, it is apparent to us that there is a clear need for research involving multiple, mixed methods of gathering data, across a variety of contexts, to understand how epistemic cognition is acquired, enacted, and changed (Chinn et al., 2011). Fault lines across methodology seem to be the most amenable to a “more is more” approach.

Integrating Perspectives on Epistemic Cognition

A basic issue to resolve, from our reading of the many fields that now comprise epistemic cognition research, is the relation between traditional cognitive accounts focused on individual cognitive structures and processes and situative accounts of persons functioning within systems of cultural activity. Greeno (2015) argued for a notion of “integrative pluralism” that accepts that cognition can be seen as occurring, and thus modeled, at multiple levels. He asserted that models focused at different levels of aggregation can each advance understanding of important phenomena. Beyond mere pluralism, however, is the goal of integrating alternative conceptualizations. Greeno suggested that a single, unifying theory that accomplishes such integration is obviously desirable but not currently feasible. He argued instead for case studies that can integrate explanations for some aspects of functioning across cognitive and situative levels. Given the scope of contexts where epistemic cognition is now seen to occur and the various perspectives on how it happens, integration of findings from these myriad contexts should be a major focus of research in the coming century.

Progress toward an integrative pluralism of accounts of epistemic cognition can occur through an expansion of current trends in epistemic cognition research. Obviously, to integrate theoretical accounts across multiple levels of analysis requires an articulation of the levels. Greeno (2015) suggested there are at least three levels relevant to cognition generally: the individual, the individual-in-interaction, and the activity system(s) in which the interaction occurs. Much of the research comprising the data set for this review focuses on the individual, on characterizing the structure of epistemic beliefs and their function. Some of it attends more closely to individuals-in-interaction from a situative perspective, by attending to how material and social resources available within a given interaction mediate the epistemic cognition engaged by individuals. We think that relatively little epistemic cognition research to date focuses at the level of activity systems, in other words, at the level of communities of people and the practices collectively used to solve epistemic problems. There are studies of episodes of collective activity and its mediation (e.g., Kelly, 2016; Rosenberg et al., 2006). Muis and Duffy’s (2013) construct of epistemic climate is an attempt to characterize an aspect of an intact activity system, a classroom, in relation to its influence on the epistemic cognition of actors in that activity system. Greene’s (2016) discussion of interacting epistemic systems, and how they affect individuals in those systems, is another way of conceptualizing effects among different levels. We encourage more work at the activity system level as an avenue to increase our understanding of how particular communities of practice might structure epistemic cognition. Kelly (2016) points to sociological studies of science as a helpful analog for such work.

Integrating across levels of analysis further requires that individual researchers, or teams, are at least aware of, if not specifically attentive to, work and models at levels that are not familiar. For example, researchers focused on individual cognitive structures and their assessment should be mindful of how particular research settings may both shape the goals of participants and structure the social and material resources that might mediate epistemic cognition in that setting. Researchers looking at the participation of individuals in collective activity could consider how to trace the individual cognitive consequences from that participation. We see such efforts beginning to happen, and likely to continue, across sites of intervention and measurement of epistemic cognition.

Bridging Schooled and Unschooled Contexts

In the past decades, epistemic cognition researchers have made considerable headway in understanding how more adaptive epistemic cognition can be promoted at different educational levels (Bråten, 2016). In elementary classrooms, intervention work in science (e.g., Metz, 2011; Ryu & Sandoval, 2012), mathematics (e.g., Mason & Scrivani, 2004; Verschaffel et al., 1999), history (Nokes, 2014; VanSledright, 2002), and language arts (Reznitskaya et al., 2012) suggests how engagement in scaffolded disciplinary practices can change students’ beliefs and thinking about knowledge and knowing in these disciplines. Moreover, many studies have tried to induce epistemic change in secondary and postsecondary classrooms, focusing on different disciplines (e.g., Bell, Matkins, & Gansneder, 2011; Hynd-Shanahan, Holschuh, & Hubbard, 2004; Muis & Duffy, 2013). Whether any changes brought about by such interventions affect people’s beliefs and thinking about knowledge and knowing outside formal educational contexts is currently not known, however (see Bendixen, 2016; Muis, Trevors, & Chevrier, 2016).

One important aspect of epistemic cognition that is highly relevant outside school is how people evaluate conflicting knowledge claims about complex scientific and socioscientific issues (Bromme & Goldman, 2014; Tabak, 2016), for example, when searching for information on the Internet (Strømsø & Kammerer, 2016). In such instances, people’s critical-analytic evaluation of sources becomes a key competence, yet this is a competence that many researchers, educators, and policymakers around the world feel is not adequately developed through schooling, not even at the secondary and postsecondary levels (Bråten & Braasch, in press). Some recent studies (e.g., Braasch et al., 2013; Kammerer, Amann, & Gerjets, 2015; Mason, Junyent, & Tornatora, 2014; Stadtler, Scharrer, Macedo-Rouet, Rouet, & Bromme, 2016) suggest that people’s ability to cope with conflicting knowledge claims about scientific and socioscientific issues can be improved through brief source evaluation interventions. However, next to nothing is known about whether or how epistemic change resulting from participation in disciplinary practices in school can be brought to bear on students’ judgments of disciplinary issues in everyday contexts, highlighting the issue of transfer across schooled and unschooled contexts. There is thus a clear need for future research that investigates and seeks to promote connections between epistemic cognition embedded within disciplinary practices in school and thinking about knowledge and knowing outside formal educational contexts. Such research should not be limited to the discipline of science, however.

For example, VanSledright and Maggioni (2016) suggested that citizens will have difficulties participating in genuine democratic discourse and making balanced decisions about the future without developing adaptive epistemic cognition within the discipline of history, involving the weighing of evidentiary support for conflicting accounts and using interpretive decision criteria (e.g., corroboration). To what extent promoting adaptive epistemic cognition through history education will actually contribute to critical, deliberate thinking about controversial social and political issues outside school (e.g., helping people avoid polarization) is an open question for future epistemic cognition researchers to address.

Until recently, epistemic cognition change has been seen as a progression from naïve to sophisticated general views on knowledge. It is now coming to be seen as appropriate adaptiveness to context. A reliance on authoritative sources, for example, is no longer seen as inherently inferior to other ways of knowing, given that much of what we learn comes from the testimony of others (Chinn et al., 2011). Teaching for adaptive epistemic cognition in a way that not only promotes disciplinary learning but also has consequences for how students engage and cope with controversial scientific and social issues outside school requires teachers who know the epistemologies of the disciplines they teach, the applicability of those disciplines to students’ everyday lives, and pedagogies that promote connections between epistemic cognition in the discipline and in everyday life. One issue in need of much further research is the potential lack of calibration between teachers’ own epistemic cognition and the epistemic aims (e.g., deep understanding) and reliable processes (e.g., reasoned argumentation; Chinn et al., 2014) they try to implement in their classrooms. This is an important issue because less adaptive epistemic cognition on the part of the teachers may interfere with teaching for adaptive epistemic cognition in students, highlighting the need to support the development of epistemic cognition among teachers participating in epistemic cognition interventions (Brownlee & Schraw, in press). An important goal for epistemic cognition researchers should be to develop empirically based, testable instructional models that can guide interventions for epistemic change (Bråten, 2016). Preferably, such instructional models should also take into consideration the need to help teachers develop their own epistemic cognition and describe how this can be effectively done.

Finally, more intervention work is needed that spans individual and social levels of intervention and coordinates between them. For example, research is needed regarding how to integrate reflection on one’s own epistemic beliefs regarding particular disciplinary issues with collective reflection and argumentation as part of social disciplinary practices dealing with those issues. This implies that the effectiveness of epistemic cognition interventions would have to be evaluated quite broadly by attending to changes in disciplinary practices at the interpersonal level, in addition to individual belief change and potential interactions between those loci of intervention (e.g., Ryu & Sandoval, 2012).

Sites of Measurement

Many of the most cited publications we reviewed, from Hofer and Pintrich’s (1997) article through more recent reviews and conceptual revisions of epistemic cognition (e.g., Barzilai & Zohar, 2014; Chinn et al., 2011), have lamented the poor reliability and lack of construct validity for scores from Likert-type scale instruments. Numerous studies have shown that the most frequently used self-report measures of epistemic cognition do not meet even minimal psychometric criteria, particularly in terms of configural invariance across samples and contexts (e.g., DeBacker et al., 2008). There appears to be a growing consensus that the ease of administering Likert-type scale assessments of epistemic cognition comes with concerns about the authenticity of whatever it is that those instruments purport to measure (Hofer & Sinatra, 2010; Sinatra, 2016). The field should no longer tolerate publications that utilize instruments with low reliability, poor data-model fit, or even worse, no investigation of construct validity at all. There is nothing innately flawed about quantitative methods of measuring epistemic cognition, but developing, testing, and refining such measures requires effort, time, patience, and persistence. Researchers interested in developing such measures should endeavor to understand and meet epistemic ideals for psychometric validity (e.g., Bandalos & Finney, 2010; Knapp & Mueller, 2010) and should fully report the evidence regarding such efforts in their empirical work. Editors, reviewers, and publishers should be critical of manuscripts that do not meet these ideals.

From our vantage point, epistemic cognition research needs a “right tool for the right job” perspective. People can self-report best about thoughts and attitudes that are explicit in their minds and that require little construction in the moment. Whether self-report instruments should use Likert-type scales or other forms of response (e.g., semantic differentials; Stahl & Bromme, 2007) is an open question. On the one hand, epistemic cognition phenomena that may be amenable to self-report assessments include epistemic metacognitive knowledge and epistemic aims (Barzilai & Zohar, 2014; Chinn et al., 2011). On the other hand, collecting data about what people do when enacting epistemic cognition requires that measurement occur in contexts where people actually enact epistemic cognition (Kelly, 2016). Analyzing data as people engage in epistemic practices, including observations of the practices used, as well as of the discourse between people (e.g., Herrenkohl & Cornelius, 2013; Ryu & Sandoval, 2012; Samarapungavan et al., 2006), can be a powerful way to understand not only people’s epistemic cognition but also their perceptions of larger epistemic systems such as science or history (Goldman, 2011; Greene, 2016), as well as how those systems enculturate people to normative practices and knowledge (Kelly, 2014).

One benefit to observations is that they can be used to collect data without interfering in people’s epistemic cognition. However, the bridge from observation data to research findings is inference, which often can be difficult to justify. The challenges of deriving and justifying inferences increase as researchers move from studying what people do (e.g., epistemic practices), to studying what they think but often do not verbalize (e.g., epistemic vices). Even more difficult are inferences about what people have learned but rarely reflect on (e.g., epistemic metacognitive skills, epistemic ideals), and what they feel but rarely think about (e.g., epistemic metacognitive experiences; Barzilai & Zohar, 2014; Chinn & Rinehart, 2016). Epistemic cognition phenomena that are rarely consciously invoked may be best measured by prompting individuals to surface and reflect on the otherwise nonconscious aspects of their epistemic cognition (e.g., “How did you feel when you realized that your views on evolution were being challenged?”). Regardless of whether researchers intervene into participants’ epistemic cognition or not, these data can and should be usefully triangulated with data from semistructured interviews (e.g., Feucht & Bendixen, 2010), cognitive interviewing (e.g., Barzilai & Weinstock, 2015; Greene, Torney-Purta, Azevedo, et al., 2010; Muis et al., 2014), or cued reflection (e.g., Berland & Crucet, 2016; Ferguson, Bråten, Strømsø, & Anmarkrud, 2013), as well as data regarding the context in which the participants engaged in epistemic cognition (e.g., Kelly & Crawford, 1997).

Of course, the more unusual the context, the more the findings are encapsulated. Completing a survey before a class begins is not something that students do very often; therefore, it is not surprising that data from this context have shown only modest-to-small relationships with data drawn from more familiar contexts, such as grades in a class. What is needed is more description of the contexts in which epistemic cognition is measured, and more research that systematically varies context in investigations of the same epistemic cognition phenomena. Such studies would reveal how perceptions of contextual factors, such as epistemic climate (Feucht, 2010; Muis & Duffy, 2013), influence enacted epistemic cognition. Likewise, drawing from psychometrics research, there is a need for studies that assess the same epistemic cognition phenomena across contexts designed to elicit typical performance (e.g., casual browsing of the Internet for a question of curiosity) and maximal performance (e.g., heavily scaffolded or supported investigations of questions with serious implications, such as reviewing literature for a dissertation).

In sum, the field’s growing recognition of the need for more attention to context and tools of measurement is not a sign of methodological or philosophical splintering, nor does it warrant calls for nihilism or methodological relativism. Rather, it represents the field’s better understanding of the problem space and of the complexity of the phenomena under study. To borrow from Perry (1970), we suspect that in the coming years, the field will make commitments within methodological relativism, where different combinations of epistemic cognition phenomena and contexts will be addressed with different measurement tools, which will be seen as more or less useful depending on the circumstances.

Transdisciplinary Possibilities

Academic disciplines tend to fragment themselves in order to focus on specific questions and problems, and psychology and education are no different in this regard. Furthermore, education is a huge field informed by multiple disciplinary perspectives besides psychology, including anthropology, sociology, history, and more. These perspectives often clash around theories, methodologies, and ideologies. Nevertheless, it is clear that scholars working on issues of epistemic cognition recognize the need to bridge disciplinary boundaries, and this trend will, and should, continue. As we have shown in this review, perspectives from disciplinary education research have reframed questions of the general versus specific character of epistemic cognition. This has come from now well-established lines of communication between educational psychologists, disciplinary education researchers, and learning scientists working within those disciplines. The next step to take in this dialogue is to try to synthesize empirical results across subfields to scrutinize available models of epistemic cognition. A recent example of this effort includes a broad synthesis of research across the elementary grades in math, science, and social studies as those areas of teaching might bear on the theory of integrated domains in epistemology model of epistemic cognition (Bendixen, 2016), an effort notable for its attention to substantive differences among disciplines and their significance to modeling epistemic cognition broadly. More syntheses like these ought to be conducted in the coming century.

Extending this sort of synthesis requires several things. Fundamentally, it requires researchers to be knowledgeable about the many fields of research that bear on specific questions of epistemic cognition. This demands both training across disciplines that would seek to develop integrative specialists around particular aspects of epistemic cognition, and research collaborations that cross specializations and can produce the sorts of case studies that Greeno (2015) suggested. It may also be the case that epistemic cognition research needs particular lines of communication to become more open. For example, the links between relevant text comprehension research (e.g., Bråten, Strømsø, et al., 2009) and disciplinary education research are relatively weak, and the same is likely to be true for other subfields of epistemic cognition research. Finally, it is critical that integrative work explicitly tests predictions of various models of epistemic cognition against one another. Epistemic cognition research over the past century has seen only a profusion of possible models, with no real pruning of the contenders. While a single theory of epistemic cognition and its development seems infeasible now, the work of comparing the value of alternative models should be done.