Theories (as a Form of Knowledge)

THEORIES AS MENTAL REPRESENTATIONS

In the 1980s psychologists and educators who were interested in the nature of mental representation began to draw attention to theory-like properties of mental representations. Although the specific forms of mental representation proposed such as schemas (Rumelhart, 1980) mental models (Gentner & Stevens, 1983; Johnson-Laird, 1983) näıve theories (Carey & Spelke, 1994; Gopnik & Wellman, 1994), and explanatory frameworks (Samara-pungavan & Wiers, 1997) vary greatly, they share certain features. These mental representations are molar or larger in scale and scope, and form structures that coordinate a variety of conceptual elements in complex and multi-faceted relationships. Such representations serve important explanatory functions, allowing people to organize, predict, and control their experiences of the world. For example, Rumelhart (1980, p. 37) says of schemas, “it is useful to think of a schema as a kind of informal, private, unarticulated theory about the nature of events.”

The most explicit analogy between everyday mental representations and theories is drawn in the research on “näıve theories” (Gopnik & Wellman, 1994). “Novice-as-theorist” accounts assume that people are theory generators. Carey and Spelke (1994) have suggested that some conceptual frameworks, such as a näıve physics of object identity and motion, appear in early infancy and may be innate. A key issue for näıve-theoretic accounts of knowledge acquisition has to do with general structural and qualitative properties of näıve belief systems or their degree of “theoreticity.” Three important aspects of theories are:

1. Structure and content: Theories comprise of a core of interrelated conceptual elements which constitute the explanatory principles for a target domain. The core explanatory principles are not mere empirical generalizations from experience but rather are abstract interpretations of experience. This is reflected in the fact that competing theoretical principles can provide different interpretations for the same body of experiential data. Theories form conceptual topographies. For example, in addition to domain-specific explanatory principles (e.g., natural selection in Darwinian theory), they contain ontological principles (Chi, Slotta, & de Leeuw, 1994) and epistemological principles or beliefs about what the central problems of the domain are, what sorts of data or evidence can be brought to bear in solving these problems, and how the explanatory fit of concepts can be evaluated (Samarapungavan & Wiers, 1997).
2. Function: Theories are explanatory devices. The core beliefs of a theory provide causal explanations for the phenomena it circumscribes. Frameworks also allow for generative predictions;
3. Corrigibility through epistemic processes: Theories can be revised as a result of various epistemic processes of knowledge evaluation (Laudan et al., 1986).

The näıve theory approach has borrowed heavily from the work of historian and philosopher of science Thomas Kuhn (1962). Kuhn claimed that major historical changes in scientific theories are accompanied by changes in the explanatory core and boundaries of a discipline, whereby notions of what counts as data and which problems a theory should address change as well. Using similar ideas to explain why students have particular difficulty with the scientific concept of heat, Wiser (1988) suggests that novices have intensive rather than extensive concepts of heat that resemble a historical precursor of the current scientific theory, one in which heat and temperature were not differentiated. Thus, while students correctly predict that on identical burners, a big vessel of water must take longer to boil than a small one, they also make the “wrong” prediction that it would not take more heat to boil water in the big vessel because their concept of heat is intensional. In other words, in the latter situation the students are using the term “heat” in the sense of “temperature” and their predictions are correct with regard to temperature. Wiser concludes that like their historical counterparts, students' theories of heat are both different from and incommensurate with the expert theory.

Chi and her associates (Chi, Slotta, & de Leeuw, 1994) have proposed an interesting version of the incommensurability argument. They suggest that novices have ontological theories or beliefs about what “kinds” of things exist and what sorts of ontological properties each class or subclass in an ontological hierarchy of “kinds” can possess. Specifically, they propose that peoples' ontological knowledge is organized into at least three “trees” or taxonomies— matter, processes, and mental states. The three trees are defined by mutually exclusive ontological attributes. For example, objects in the category of matter (e.g., water, cars, and dogs) have ontological attributes such as having volume and mass. Similarly, processes have ontological attributes such as “occurring over time.” Chi and colleagues argue that novices have difficulty with the acquisition of scientific concepts because these concepts require a restructuring of their ontologies. For example, while novices typically classify heat as a substance belonging to the ontological tree of matter, scientists see heat as a constraint-based interaction belonging to the ontological tree of processes.

THE DEBATE OVER NOVICE KNOWLEDGE

There is a vast body of research that shows that novice ideas about the natural world differ in conceptual content from those of scientists (Vosniadou & Brewer, 1992, 1994; Wiser, 1988). However, researchers disagree profoundly about the qualitative nature of such lay concepts. For example, based on research about näıve concepts in the domain of motion and force, diSessa and his colleagues (diSessa, 1993; diSessa, Gillespie, & Esterly, 2004) describe novice knowledge as a weakly organized system of beliefs that lacks internal coherence, is unstable over time and problem context, and malleable in the face of anomalous evidence. In a reinterpretation of the work in intuitive physics, diSessa and associates suggest that novices do not have anything like a naive “theory” of motion or force. According to diSessa (1993), what novices have is a fragmented, unstable, and malleable collection of beliefs which are low-level abstractions of everyday experience. diSessa refers to such beliefs as p-prims, short for phenomenological primitives. He argues that novice beliefs are unlike scientific theories because they are not constrained by epistemic requirements of coherence or “systematicity.”

Research in several domains suggests that people's belief systems about some aspects of the natural world are theory-like, at least in some important respects. For example, children and lay adults appear to construct coherent and robust biological constructs based on principles of biological essentialism across a variety of cultures and task contexts (Ahn et al., 2001). Samarapungavan & Wiers, (1997) describe novice beliefs about species and speciation in terms of explanatory frameworks which they describe as a small set of explanatory principles that constrain but do not fully pre-specify the mental models that children construct when presented with novel biological problems about the nature of species and speci-ation. Part of the evidence for theoreticity comes from the observation that at least some of the core beliefs of novice biological frameworks cannot be induced directly from experience. For example, it is hard to conceive of any direct phenomenal experience that would lead elementary school children to a belief in the spontaneous generation of complex species (Samarapungavan & Wiers, 1997).

Vosniadou and Brewer and colleagues have conducted research on children's conceptual development in the domain observational astronomy. Their research shows that across a variety of cultures many elementary school children construct scientifically inaccurate but coherent and explanatory mental models of the earth's shape and the day-night cycle (Samarapungavan, Vosniadou & Brewer, 1996; Vosniadou & Brewer, 1992, Vosniadou, Skopeliti, & Ikospentaki, 2004). These representations allow people to generate explanations for phenomena such as the seasons, eclipses, and the day/night cycle and to make and evaluate predictions about what will happen in novel scenarios. Additionally, their research shows that initial theory-like representations are corrigible because children revise their mental models over time to integrate new information presented in formal schooling.

Some researchers claim to have obtained empirical results that contradict Vosniadou and Brewer's findings in the domain of astronomy (Nobes et al., 2003). However, Brewer and Vosniadou and colleagues (Brewer, 2008; Vosniadou, Skopeliti, & Ikospentaki, 2004) point out serious methodological flaws in these studies.

THEORIES AS FORMS OF KNOWLEDGE: OPEN QUESTIONS

The debates about the utility of viewing knowledge as being organized in theory-like structures are hard to resolve definitively because of the great variability in the empirical studies on both sides of the issue. For one thing, findings about the degree of coherence and explanatory power in novice knowledge vary by domain. For example, Nakhleh and Samarapungavan (1999) found that elementary school children's ideas about the nature of matter cohered loosely at an ontological level but were not sufficiently coherent in terms of the specific explanations generated for phenomena such as phase transitions to be called explanatory frameworks. It may be that it is easier for people to form coherent knowledge structures in some domains than in others.

A second difficulty in resolving the debates about novice knowledge is the methodological variation in the studies. The studies draw from different populations; vary dramatically in sample size, in the nature and variety of tasks used to elicit data, and the methods for analyzing and aggregating data. One sub-domain in which two studies employed fairly similar methodologies but resulted in radically different conclusions about novice knowledge is that of force and motion. Ioannides and Vosniadou (2002) conducted a cross-sectional study with 105 Greek children (preschoolers through ninth graders) to investigate the development of concepts of force and motion and concluded that children used a small set of coherent constructs to explain phenomena in this domain. diSessa, Gillespie, and Esterly (2004) conducted a quasi-replication of the Ioannides and Vosniadou study with 30 American children across a similar age range. diSessa and colleagues found that the American children invoked a greater number of constructs to explain the phenomena of force and motion, and their use of these constructs varied contextually, showing a lack of coherence.

Although the two studies were relatively similar in the methods used to elicit data, there were nonetheless important differences. One difference was linguistic; the Greek children were tested in the Greek language while the American children were tested in English. Therefore, it is possible that at least some of the questions were not equivalent in translation between the two studies. A second, more important difference was that the American sample was much smaller, less than a quarter of the size of the Greek sample. Given the wide range of ages involved, the sub-samples at each age point may have been too small to be truly representative of the population at large. Thirdly and most importantly, although the procedures for data collection were similar in the two studies, the procedures for data analysis were not. Ioannides and Vosniadou (2002) scored children's explanations of their initial answers to each question while diSessa, Gillespie, and Esterly (2004) noted that they were unable to score children's explanations. Thus, despite the similarities in data collection procedures, the actual data set that was analyzed across the two studies was likely to be quite different.

Differences of the kind described above make it hard to render definitive judgments about the utility of regarding lay or everyday representations of the world as theorylike. In general, even if people represent some aspects of the natural world in theory-like ways, it would be implausible to suggest that all knowledge is organized in theories. Even highly regarded scientists such as Charles Darwin, Albert Einstein, or Michael Faraday, probably did not have good theories for every realm of experience. The extent to which knowledge representations are theory-like probably depends on a number of factors including the nature of domain phenomena, the individual's sustained interest and curiosity in the domain, as well as educational and cultural factors. For example, for American children, the dominant cultural model for observational astronomy corresponds to the scientific model. In other domains such as biology, there are often salient competing cultural models to the scientific theory of evolution for important phenomena such as speciation. The availability and salience of such competing cultural models, and the epistemic authority that students accord to various sources of information, are likely to affect both the content and the quality of students' representations.

Given the diversity of theoretical perspectives and empirical findings described above, it appears that while both children and adults may spontaneous construct theory-like representations of the world in some domains of knowledge, they are unlikely to do so uniformly across domains. Additionally, scientific theories are the products of the public institutions of science and their coherence stems in part from their development and modification under processes of rigorous evaluation and critique by the community of scientific practice. In contrast, individual knowledge representations used in everyday life are rarely subjected to such scrutiny and evaluation. Consequently, when compared with their scientific counterparts, näıve theories are invariably likely to be far less cohesive and consistent.