Igniting Scientific Thought Process In Students

Contributed by:
Scientific thinking is a particular form of human problem-solving that
involves mental representations of
(1) hypotheses about the structure and processes of the natural world and
(2) various methods of inquiry used to determine the extent to which those hypotheses are consistent with phenomena.
1. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 67 [67–99] 5.11.2018 11:50AM
4 Improving Students’ Scientific
David Klahr, Corinne Zimmerman, and Bryan J. Matlen
The production of a scientifically literate population is a fundamental goal of our
educational system. The justifications for that goal, and descriptions of paths toward
it, have been reiterated many times in recent decades, as exemplified by major policy
statements and specific recommendations from prestigious organizations ranging
from “Benchmarks of Scientific Literacy” (AAAS, 1993) to the recent “Framework
for K-12 Science Education” (NRC, 2012). Consequently, substantial effort has been
devoted to determining how to increase the likelihood that, as students progress
through school, they will acquire at least a rudimentary understanding of funda-
mental domain-general scientific concepts and procedures, as well as a nontrivial
amount of domain-specific concepts. However, given the vast number of those
procedures and concepts, it is not surprising that the full science curriculum pre-
sented to students from pre-school through high school has often been characterized
as “a mile wide and an inch deep” (Li, Klahr, & Siler, 2006; Santau et al., 2014).*
Thus, the challenge facing researchers interested in improving science education
is to enhance the quality and generality of the answers to two related questions: What
is scientific thinking? and How can it be taught? In this chapter, we attempt to answer
the first question by presenting a brief summary of a broad framework that char-
acterizes the essential aspects of scientific thinking and reviewing the developmental
origins of scientific thinking. We answer the second question by describing a few
representative examples of research on teaching science in specific domains, such as
physics, biology, and earth sciences – organized according to the framework – and
selected from the extensive literature on different ways to improve children’s basic
ability to think scientifically.
What Is Scientific Thinking?
Scientific thinking is a particular form of human problem-solving that
involves mental representations of (1) hypotheses about the structure and processes
of the natural world and (2) various methods of inquiry used to determine the extent
* Thanks to Audrey Russo for her painstaking proofreading and editing of the references. Preparation of
this chapter was supported in part by grants from the Institute of Education Sciences to Carnegie
Mellon University (R305A100404 and R305A170176), and from the National Science Foundation to
the University of Chicago (1548292). Opinions expressed do not represent the views of the US
Department of Education or the National Science Foundation.
2. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 68 [67–99] 5.11.2018 11:50AM
68 david klahr, corinne zimmerman, and bryan j. matlen
to which those hypotheses are consistent with phenomena. Scientific discovery is a
type of problem-solving that involves a high-level search through two complemen-
tary problem spaces (Newell & Simon, 1972). One such space is the Hypothesis
Space, and the other is the Experiment Space. The two spaces are linked through the
process of Evaluating Evidence. This dual-space framework, dubbed SDDS (for
Scientific Discovery as Dual Search) by Klahr and Dunbar (1988), is used to frame
the literature reviewed in this chapter.
The three cognitive processes from the SDDS model are listed in Table 4.1, along
with corresponding science practices that have been elucidated in recent policy
statements about science education (NRC, 2012). Hypothesis Space Search involves
the formulation and refinement of hypotheses. The specific practices that support this
search are (1) asking questions and (2) developing and using analogies and models.
Experiment Space Search involves planning and carrying out investigations. Finally,
Evaluating Evidence requires (1) analyzing and interpreting data and then (2)
constructing explanations. Moreover, each of these science practices can vary
along a dimension from being highly domain-specific to domain-general. In Table
4.1, we have listed some representative publications that focus on specific cells in the
overall taxonomy of scientific thinking and we will describe some of them in more
detail below.
The particular studies summarized in this chapter were chosen because they
exemplify specific psychological processes subsumed in the aspects of scientific
thinking that have been used to organize the rows in Table 4.1. We have further
divided the studies into those focusing on children’s domain-specific knowledge and
those focusing on domain-general knowledge, even though many of those we have
classified as domain-general are – of necessity – situated in specific domains. For
example, we do not focus on children’s developing knowledge about such concepts
as the periodic table of the elements, or friction, or the definition of absolute zero.
Nor do we focus on how children learn about particular domain-specific processes
(e.g., alternating current, photosynthesis, Newton’s Laws, or chemical equilibrium).
Instead, we approach the topic of scientific thinking from the perspective of cogni-
tive and developmental psychology. Using this perspective, we describe some highly
general aspects of what it means to “think like a scientist” or to exhibit “scientific
thinking” about some domain or problem, and we comment on how different
instructional approaches and pedagogical strategies can facilitate children’s acquisi-
tion and mastery of this kind of knowledge.
What Are the Developmental Origins of Scientific Thinking?
Beginning in infancy, children learn about the natural world in ways that will
influence their later scientific thinking. A notable achievement in the early years of
life is the ability to think representationally – that is, being able to think of an object
both as an object in and of itself and as a representation of something else. Even three-
year-old children are able to use scale models to guide their search for hidden objects
in novel locations by applying the spatial relations in the model to the corresponding
3. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 69 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 69
Table 4.1 A taxonomy for categorizing psychological investigations of aspects of science
education, with representative examples of each type
Type of knowledge
Cognitive processes Science practices Domain-specific Domain-general
Forming and Asking questions A1 D1
Refining Samarapungavan, Chouinard (2007)
Hypotheses Mantzicopoulos, and Patrick Jirout and Klahr (2012)
(Hypothesis Space (2008)
Search) Kuhn and Dean (2005)
King (1991)
Developing and A2 D2
using analogies Christie and Gentner (2010) Raghavan and Glaser (1995)
and models
Matlen et al. (2011)
Clement (1993, 1982, 2000)
Vendetti et al. (2015)
Lehrer and Schauble (2004)
Investigation Skills Planning and B E
(Experiment Space carrying out Metz (1997) Chen and Klahr (1999)
Search) investigations
Schwichow, Zimmerman et Sodian, Zaitchik, and Carey
al. (2016) (1991)
Siler and Klahr (2012)
Zimmerman and Croker
Evaluating Analyzing and C1 F1
Evidence interpreting Amsel and Brock (1996) Masnick and Morris (2008)
Penner and Klahr (1996a, b)
Kuhn (2011)
Masnick, Klahr, and
Knowles (2016)
Constructing C2 F2
explanations Inagaki and Hatano (2008) ynatt, Doherty, and Tweney
Lehrer and Schauble (2004) (1977, 1978)
Kelemen et al. (2014)
Note: The structure of this table is taken from Zimmerman and Klahr (2018). The cognitive processes categories
(row headings) and knowledge types (column headings) are adapted from Klahr and Dunbar (1988), Klahr
(1994), and Klahr and Carver (1995). The science practice row subheadings are from the Framework for K-12
Science Education (NRC, 2012). Cell contents cite a few studies that exemplify the row and column headings for
each cell. The corresponding text in the chapter cites additional examples.
4. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 70 [67–99] 5.11.2018 11:50AM
70 david klahr, corinne zimmerman, and bryan j. matlen
relations in the real world (DeLoache, 1987), and four- to five-year-olds develop the
ability to engage in pretend play, where one (usually mundane) object can represent
another (often exotic) object (Hopkins, Dore, & Lillard, 2015; Sutherland &
Friedman, 2013), such as when a stick becomes a magic wand or a sword. This
fundamental ability lies at the heart of important scientific skills such as analogical
reasoning and lays the groundwork for later scientific tasks such as interpreting and
reasoning with models, maps, and diagrams (Uttal, Fisher, & Taylor, 2006).
The nature of knowledge acquisition in young children is a source of intense
debate in developmental science. Some developmental theories espouse that chil-
dren’s acquisition of knowledge is similar, in many respects, to scientists’ – that is,
guided by top-down processes and theory-level explanations (e.g., Gelman & Coley,
1990; Gopnik & Sobel, 2000; Keil et al., 1998). Other theories espouse that knowl-
edge is acquired through lower-level perceptual, attentional, and memory-based
processes (e.g., Rakison, Lupyan, & Oakes, 2008; Sloutsky & Fisher, 2004; Smith,
Jones, & Landau, 1996). Regardless of the way in which children acquire knowl-
edge, it is clear that by the time they enter formal instruction they have strong ideas
about the way in which the world works, that is, about what causes what. However,
many of those ideas are partially or entirely incorrect (Vosniadou, 2013). In the
science education literature, these mistaken, distorted, or partially correct notions,
ideas, and beliefs about the world that children often bring to the science classroom
are sometimes referred to as “preconceptions” and other times as “misconceptions.”
As Horton (2007) put it:
“Misconceptions” seems excessively judgmental in view of the tentative nature of
science and the fact that many of these conceptions have been useful to the students
in the past. “Preconceptions” glosses over the fact that many of these conceptions
arise during the course of instruction. (Horton, 2007, p. 4, emphasis added)
In this chapter we use both terms, roughly according to the context as described by
Children’s preconceptions have been shown to influence learning in a variety of
scientific domains including physics (Clement, 1982), thermodynamics (Lewis &
Linn, 1994), astronomy (Vosniadou & Brewer, 1994), biology (Inagaki & Hatano,
2008; Opfer & Seigler, 2004), geoscience (Gobert & Clement, 1999), and chemistry
(Wiser & Smith, 2008). Moreover, preconceptions can exist not only for domain-
specific knowledge but also for domain-general science concepts, such as learning
the principles of experimental design (Chen & Klahr, 1999; Lorch et al., 2010; Siler,
Klahr, & Matlen, 2013) or understanding the purpose of scientific models
(Grosslight, Unger, & Jay, 1991; Treagust, Chittleborough, & Mamiala, 2002).
One particularly interesting and widespread preconception was first reported in
Vosniadou and Brewer’s (1992, 1994) classic studies of children’s mechanistic
explanations regarding the day and night cycle. Prior to formal schooling, many
children have ideas about how and why day turns to night, ideas that are deeply
rooted in children’s everyday concrete experiences (and which parallel the history of
early prescientific notions). For instance, many children initially believe that the Earth
is a flat plane and the sun travels from above to below the Earth to create day and
5. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 71 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 71
night. Such preconceptions often continue to exist even after children receive formal
instruction. For example, on being instructed that the Earth is in fact round, many
children refine their conception of the Earth to being a flat disk-shaped object in order
to account for what they are told (i.e., the Earth is round) and what they observe (i.e.,
the Earth is flat). Vosniadou and Brewer (1992) identified several similar morphed or
synthetic understandings, such as the belief that the Earth is a hollow sphere with a
flat bottom on which we live. One challenge to science education is the fact that such
preconceptions are particularly resistant to formal instruction. Several studies have
identified preconceptions that conflict with scientific conceptions even after semester-
long or university-level courses (Clement, 1982; Lewis & Linn, 1994; Wiser &
Smith, 2008; Zimmerman & Cuddington, 2007).
Chi (2013) has distinguished between two types of learning based on preconcep-
tions. The first is knowledge acquisition, or gap-filling, where students’ prior knowl-
edge is incomplete and they need to acquire new knowledge. The second is where
students have prior knowledge that is in conflict with to-be-learned knowledge (e.g.,
scientific concepts). For example, preschool-age children’s tendency to believe that
plants are inanimate objects conflicts with the scientific perspective that plants are
living organisms. When children begin to conceive of plants as living organisms, it is
said that they have undergone a recategorization of plants into a new ontological
category – a conceptual change.
In sum, children come to school with the necessary prerequisites for scientific
thinking, but they also bring with them strong beliefs about how the natural world
works and these “theories” are often deeply flawed. As a result, science educators face
three important challenges. First, although all educators are responsible for teaching
that can be described as “gap-filling” (because of incomplete knowledge), they also
have to deal with the many documented scientific misconceptions that are robust and
continue into adulthood (Zimmerman & Cuddington, 2007). Second, children’s under-
standing of the procedures used by scientists to investigate the world, such as the
design of unconfounded experiments, is subject to conceptual difficulties and mis-
conceptions (Chen & Klahr, 2008; Matlen & Klahr, 2013). Finally, learning to conduct
scientific investigations involves the coordination of many science practices (see Table
4.1), which require extended instruction and refinement over the school years.
How Can Children’s Scientific Thinking Be Taught?
Not only does scientific thinking require the development of several funda-
mental precursors, such as the ability to think representationally and to develop and
revise causal theories about the world, but it is also highly culturally and education-
ally mediated (Zimmerman & Croker, 2014). Thus, the essential goal of science
education can be characterized as having three primary aspects.
i. Fostering conceptual thinking in science: The aim here is to teach children
something about scientific knowledge, that is, what has been learned from tens
of thousands of years of human efforts to better understand the natural world.
6. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 72 [67–99] 5.11.2018 11:50AM
72 david klahr, corinne zimmerman, and bryan j. matlen
This is the product of science and it includes a vast knowledge base. As noted at
the beginning of the chapter, one of the challenges of science education is the fact
that rather than coming to school with no knowledge of specific science con-
cepts, children come with deep and often robust misconceptions (e.g., children –
and many adults, alas! –often fail to distinguish between mass and weight, or
they believe that the sun circles the Earth, or that energy and force are the same
thing). These misconceptions must be detected and remediated based on an
understanding of how misconceptions can be changed in the face of new
ii. Fostering procedural thinking in science: To teach children some of the ways that
humans have devised to acquire that knowledge. That is, to teach them some of
the fundamental processes of science. Here, too, children often have deep
misconceptions. For example, with respect to experimental procedures, young
children, even as late as their middle school years, do not know how to distin-
guish confounded experiments from unconfounded experiments (Chen & Klahr,
1999; Kuhn et al., 1995).
iii. Fostering the ability to apply “School Science Knowledge” to “Everyday
Scientific Thinking”: That is, to encourage and facilitate children’s ability to
use what they have learned about scientific products and scientific processes to
extend and enrich both of them during their day-to-day engagement with the
world around them – i.e., to teach children how to use what they know about
science to do even more science. Of course this last goal has a wide range of
objectives, from simply being able to approach everyday problems by proposing
simple hypotheses and being able to identify causal factors (“Why is my base-
ment wet?: Is the downspout at the front of the house clogged?”; “Why am I so
jumpy this evening?: Is it because I ate a lot of candy at lunch?”) to creating a
research project for submission to the International Science and Engineering
Fostering Conceptual Thinking
Science educators are responsible for ensuring that students understand both the
concepts (i.e., the “Disciplinary Core Ideas”; NRC, 2012) and the processes of
science. With respect to understanding the vast number of domain-specific science
concepts, the literature is difficult to adequately summarize. As of 2009, Reinders
Duit’s bibliography of research studies on conceptual change in science had more
than 8,400 entries.2 To illustrate, consider a single chapter on children’s under-
standing of physical science concepts. Hadzigeorgiou (2015) reviews studies of
children’s ideas about matter, heat, temperature, evaporation, condensation, the
water cycle, forces, motion, floating, sinking, electricity, and light. Each of these
topics can be further unpacked to constituent subcomponents (e.g., electricity con-
cepts include current, voltage, charge, electrons, resistance, and circuits, to name a
For the most recent update, see http://archiv.ipn.uni-kiel.de/stcse/
7. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 73 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 73
few). Given this daunting proliferation of specific topics, in this section, we briefly
focus on key educational strategies that have been used across different conceptual
domains to support scientific thinking.
As mentioned above, preconceptions significantly influence how children think
about and learn scientific concepts. Therefore, facilitating scientific knowledge
acquisition requires an understanding of both the preconceptions children bring
and how likely that prior knowledge is to change in the face of new information.
Work by Kelemen and colleagues (2014) has demonstrated that children’s deep
scientific misconceptions (in this case, about natural selection) can be remediated
by relatively brief engagement in well-designed explicit instruction that directly
targets the misconception. Other research has shown that sharpening children’s
observational skills is important because prior belief may influence what is
“observed” (Chinn & Malhotra, 2002; Echevarria, 2003). For example, many chil-
dren (and adults) assume that a heavy object falls faster than a light object; because it
is difficult to observe both objects simultaneously, this ambiguous situation results in
expectations of what will be observed, which in turn influences interpretation,
generalization, and retention. Chinn and Malhotra (2002) concluded that belief
change based on unexpected evidence is possible but making the correct observa-
tions is key. Instructional interventions with scaffolding were successful in promot-
ing conceptual change such that children learned how to make observations unbiased
by their initial conceptions.
One of the more effective strategies for effecting conceptual change is to (1)
present information that conflicts with children’s preconceptions and then (2) gra-
dually support the adoption of scientifically accurate knowledge. For instance, a
correct understanding of physical forces is a common source of difficulty for
students. Many students believe that objects resting on top of a surface (e.g., books
resting on a table) exhibit only a downward force (gravity) but not an upward force
(the table). To confront this preconception, Minstrell (1992) developed an ingenious
intervention to induce cognitive conflict. Specifically, Minstrell prompted students to
hold a book in the outstretched palm of one hand, and then asked them to explain the
forces that were acting on the object. Most students initially asserted that there was
only a downward force (i.e., gravity). Then Minstrell started piling an increasing
number of books onto the students’ hands until they acknowledged that their hands
were exerting an upward force in order to counteract the gravitational force. Using
this revised conceptual model of compensating forces, students were able to general-
ize this new knowledge to new related instances (Minstrell, 1992). A similar strategy
was used by Clement (1993), who attempted to anchor students’ understanding of the
same concept with a series of close analogies. Clement’s first example consisted of a
spring: Few students disagree that the spring exhibits an upward force if weight is
placed on it. Before moving to the example of books on a table, Clement used an
intermediate example – books resting on a foam base. The intermediate example
acted as a bridge between the spring and the table examples, inducing students to
recognize the similarity between the cases. Cognitive conflict has also been used to
address procedural misconceptions that students have about how to conduct con-
trolled experiments (Schwichow, Zimmerman, et al., 2016). This study will be
8. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 74 [67–99] 5.11.2018 11:50AM
74 david klahr, corinne zimmerman, and bryan j. matlen
described in more detail in the section below on “Early Experimentation Skills.”
These cases illustrate that students are most likely to undergo conceptual change
when new information is progressively sequenced, starting from students’ precon-
ceptions and moving in well-designed instructional steps toward the resolution of the
misconception. We return to related strategies in later sections.
Other strategies for fostering conceptual change include interventions designed to
facilitate the process of making accurate observations and measurements, including
the need to teach students that observations can be biased and measurements include
error (Chinn & Malhotra, 2002). Koerber, Osterhaus, and Sodian (2017) demon-
strated that scaffolds in the form of diagrams and explanations were effective in
remediating misconceptions in kindergarteners and 2nd graders. Schuster and col-
leagues (2017) compared two different epistemic approaches to teaching
Disciplinary Core Ideas. They found that both the “guided” and the “direct” instruc-
tional approaches, which both involved active student engagement, were effective at
promoting conceptual learning gains.
Conceptual change is clearly essential for scientific thinking. Here, we focused on
describing just a few educational strategies that may be used across scientific
concepts. Shtulman (2017) argues that replacing the intuitive ideas that children
have about the world is a key challenge of science education. Shtulman describes the
conceptual challenges for a variety of physical and biological science topics (e.g.,
energy, motion, inheritance, illness). Intuitive theories, whether about growth or the
cosmos, have many properties in common (e.g., they tend to be rooted in perceptual
features). However, Shtulman (2017) argues that helping students develop more
scientifically accurate theories requires educators to analyze individual concepts in
depth to ascertain the most effective way to challenge intuitive theories: “Instruction
that neglects the domain-specific nature of intuitive theories and their scientific
counterparts has about as much chance of working as the chance of a nuclear
physicist making an important discovery in immunology” (p. 249).
Fostering Procedural Thinking
As noted previously, the three key cognitive processes involved in scientific thinking
that correspond with various science practices (see Table 4.1) can be applied to any
domain of science. As children learn to engage in inquiry, these investigation skills
can be fostered individually and in concert with other skills. For these scientific
practices, we review some of what is known about children’s developing abilities
along with illustrative research on fostering these types of procedural thinking in
Forming and Refining Hypotheses
Of the three cognitive processes of SDDS, search in the hypothesis space has the
most in common with conceptual thinking in science, as it typically involves a search
of relevant domain-specific knowledge as represented in the hypothesis space. When
one is engaged in inquiry or investigation activities, however, hypothesis-space
9. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 75 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 75
search is instantiated in the service of the scientific practices of asking questions and
developing or using models (NRC, 2012).
Asking questions and curiosity. Asking questions is one of the foundational
process skills of scientific practice (NRC, 2012). However, rather than viewing
science as a process of posing and then finding answers to questions, students
often believe that the goal of science is to demonstrate what is already known
(Kuhn, 2005), or to see if something “works” or to invent things (Carey et al.,
1989). However, asking questions for which the answer is not yet known is a crucial
element of inquiry that students must learn (Kuhn & Dean, 2005). Students must
learn that question-asking is a defining feature of science. An essential precursor to
asking good questions is curiosity (Klahr, Zimmerman, & Jirout, 2011). The funda-
mental importance of curiosity in science education is indicated by its nearly
universal inclusion as a desired “habit of mind” across a variety of influential science
curricula, educational standards, and assessment goals (AAAS, 1993; NEGP, 1993,
1995; NAEYC, 2012; NRC, 2000). For example, the National Science Teachers
Association’s official position statement on early childhood science education3
recommends that teachers
recognize the value and importance of nurturing young children’s curiosity and
provide experiences in the early years that focus on the content and practices of
science with an understanding of how these experiences connect to the science
content defined in the Next Generation Science Standards. (NSTA, 2014, p. 3)
Nevertheless, although “curiosity” is acknowledged to be an essential part of science
at all ages and levels of sophistication, it remains a notoriously elusive psychological
construct in both the adult (Lowenstein, 1994) and the child (Jirout & Klahr, 2012)
Simple problem-solving tasks that require question-asking have been used to
investigate children’s ability to recognize specific instances of uncertainty and to
evaluate information. Chouinard (2007) and others have demonstrated not only that
young children can determine which questions to ask to address uncertainty but also
that they can use information yielded by the answers to their questions to resolve it.
Other research has examined children’s abilities to ask questions in particular
domains in order to investigate their understanding of various phenomena. For
example, Greif and colleagues (2006) investigated young children’s ability to ask
domain-specific questions on a structured task. Children were instructed to ask
questions about unfamiliar objects and animals, which they were able to do –
averaging twenty-six questions asked across twelve pictures. Many questions were
quite general, such as “What is it?” Other questions, however, showed that children
recognized and understood that different questions should be asked of the different
categories (i.e., objects and animals).
Several studies have gone beyond simply demonstrating young children’s nascent
ability to ask “good” questions about the natural world and have created procedures
aimed at increasing the efficacy of such questions. For example, King (1991)
10. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 76 [67–99] 5.11.2018 11:50AM
76 david klahr, corinne zimmerman, and bryan j. matlen
demonstrated that 5th graders could be trained to use strategic questions to guide
their cognitive and metacognitive activity during problem-solving with partners, and
that when they did use such “good” questions, they learned more about the system
they were investigating.
Analogical Thinking and Use of Scientific Models. Analogical reasoning is a
very powerful way to form and refine hypotheses and to scaffold scientific under-
standing. It involves aligning representations based on their shared relations
(Gentner, 1983, 2010). When one of the representations is better understood than
the other, information from the familiar case (i.e., by convention, termed the
“source”) can be used to inform the scientist’s understanding of the unfamiliar
case (i.e., by convention, termed the “target”). For example, an important concept
in molecular biology is how enzymes and substrates interact. This concept is easier to
understand when it is compared to a lock and key – the key acts as an unlatching
mechanism, fitting into the lock to open it, just as the enzyme fits into the substrate to
break it apart. By putting these domains into correspondence based on their shared
relations, further inferences might be drawn: for example, a specific key only fits a
specific lock, therefore, enzymes may only react with specific substrates. This
example illustrates the inferential power that can be derived from analogies, even
where there is limited knowledge of the target domain.
Scientists frequently use analogies to generate hypotheses and explain scientific
phenomena and to interpret and construct scientific models (Dunbar, 1997). Thus, it is
not surprising that analogies have played a role in many scientific discoveries. A fitting
example is that of Johannes Kepler (Gentner et al., 1997), who observed that the
planets farther away from the sun moved in an elliptical path that was slower than that
of the planets closer to the sun. To explain this phenomenon, Kepler drew an analogy
to light – he reasoned that, just as light is weakened when viewed from a distance, the
sun might dispense a moving power (vis motrix – an early predecessor to gravity) that
becomes weakened when objects are farther away (Gentner et al., 1997). This analogy
in turn contributed to his discovery of the laws of planetary motion.
In addition to the “classical” analogies underlying some of the great scientific
discoveries, contemporary scientists use analogical reasoning in their everyday
practice, such as the design and interpretation of experiments. Dunbar (1995), in
his groundbreaking investigations of the thinking processes involved by scientists in
several contemporary molecular biology laboratories, discovered that there are two
general forms of analogies used by scientists in discussing and explaining their work.
Within lab group discussions, the analogies tend to be fairly “local,” as when one
particularly well-known specific process in the lab is used as the base for analogi-
cally interpreting a recent empirical result. In contrast, when describing their work to
outsiders (e.g., science reporters in the media), the scientists use “distant” analogies,
in which the base domain is fairly familiar to the public and it is used to describe
some important features of a new discovery in the lab.
Even early school-age children have at least rudimentary analogical reasoning
abilities. However, owing in part to limited domain knowledge (Bulloch & Opfer,
2009; Goswami, 1991, 2001) and a less than fully developed prefrontal cortex
11. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 77 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 77
Crust Skin
Mantle Inner
Inner Pit
Figure 4.1 Visual representation of the “Earth is a like a peach” analogy (after
Matlen et al., 2011)
In this analogy, the peach, whose structure is known to the learner, serves as the
source and the to-be-learned inner structure of the Earth serves as the target.
(Vendetti et al., 2015; Wright et al., 2008), young children are easily distracted by
perceptual features of analogies at the expense of overlooking their relational
structure, making them more prone to irrelevant encodings and conceptual misalign-
ments. Despite young children’s tendency to overlook relational information,
researchers have identified strategies that children can be trained to use to support
their scientific reasoning. One strategy is to prompt comparison-making between
source and target domains (Christie & Gentner, 2010; Gick & Holyoak, 1983).
Having been directly instructed to compare two domains, children are more likely
to look beyond superficial features of problems and identify their common relational
The process of comparison-making is facilitated when children’s attention is
directed toward relevant relationships (Vendetti et al., 2015). Visually representing
both the source and the target of the analogy is one way in which comparison-making
can be augmented (Gadgil, Chi, & Nokes, 2013; Richland & McDonough, 2010). As
an illustration, Matlen and colleagues (2011) aimed to teach elementary-age students
geological concepts, such as how mountains and volcanoes are formed. Matlen and
colleagues (2011) presented children with analogy-enhanced text passages that were
accompanied by either a visualization of both the source and the target (see Figure
4.1) of the analogy or a visualization of just the target (the latter of which is the most
common form of presentation in elementary science texts). Matlen and colleagues
(2011) found that children were more likely to learn and retain geoscience concepts
when text passages were accompanied by visualizations of both the source and the
target. Simultaneous visual presentation prompted students to compare the two
domains and reduced the cognitive effort of having to recall each representation.
Further, the impact of visual aids is enhanced when the to-be-aligned components are
made perceptually similar (Jee et al., 2013) or when they are spatially aligned
12. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 78 [67–99] 5.11.2018 11:50AM
78 david klahr, corinne zimmerman, and bryan j. matlen
(Matlen, Gentner, & Franconeri, 2014; Kurtz & Gentner, 2013), such that attention is
directed to the corresponding components.
Scientific models also rely on the process of analogical reasoning and are fre-
quently used in science. The source of the analogy constitutes a familiar case, which
helps the scientist understand the phenomenon under study. Developing and using
models constitute a “signature practice of the sciences” (Quellmalz et al., 2012, p.
366) that is related to the search for hypotheses or explanations. Models are com-
monly used in science and engineering to support theory-building, argumentation,
and explanation (Nersessian, 2008). For example, Watson and Crick’s physical
model of the structure of DNA, drawings, and schematic representations (e.g.,
Faraday’s sketches of electromagnetic tori or Darwin’s tree of life) are cases of
models that not only are used by practicing scientists but also provide powerful
pedagogical value in teaching these concepts to students.
Because models are of central importance in scientific practice, they are now
widely emphasized in science education and science assessment (Clement, 2000;
Lehrer & Schauble, 2000, 2012; NRC, 2012). In both science and science education,
the ability to develop and use models is becoming increasingly sophisticated due to
the scaffolding provided by computers and computer simulations. Simulation mod-
els can be used to learn about and investigate phenomena that are “too large, too
small, too fast, or too dangerous to study in classrooms” (Quellmalz et al., 2012, p.
367). Science education includes numerous domain-general and domain-specific
examples of the instantiation of such model-based practices. Particular curricula
are designed around the importance of models, such as the Model-Based Reasoning
in Science (MARS) curriculum (Ragavan & Glaser, 1995; Zimmerman, Raghavan,
& Sartoris, 2003). Domain-general examples include learning about variability
(Lehrer & Schauble, 2004) and decomposition (Ero-Tolliver, Lucas, & Schauble,
2013) and domain-specific examples include evolution in elementary school
(Keleman et al., 2014; Lehrer & Schauble, 2012), ecosystems in 6th grade (Lehrer,
Schauble, & Lucas, 2008), and biomechanics of the human elbow (Penner et al.,
1997; Penner, Lehrer, & Schauble, 1998).
It is important to recognize that although analogies and models have the potential
to be powerful teaching tools, teachers must help students to differentiate between
perceptual and conceptual similarities between the model and what is being mod-
eled. In learning about multicomponent dynamic systems, young children tend to
focus initially on perceptual features rather than on relations among components.
However, with appropriate instruction, in which a model’s relational structure rather
than its perceptual features is reinforced, children can learn how to engage in
relatively sophisticated reasoning. For example, Penner and colleagues (1997)
asked 1st grade children to construct models of their elbow. Although children’s
models retained many superficial similarities to the arm (e.g., children insisted that
their models include a hand with five fingers, represented by a foam ball and popsicle
sticks), children were eventually able to construct models of their elbow that retained
functional characteristics (e.g., incorporating the constraint that the elbow is unable
to rotate 360 degrees), and were also more likely than a nonmodeling peer group to
ignore superficial distractors when identifying functional models. With sustained
13. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 79 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 79
practice and scaffolding, children can overcome the tendency to attend to superficial
similarities and can begin to reason with more abstract models that retain mostly
relational structure, such as graphing the relationship between plant growth and time
(Lehrer & Schauble, 2004) or modeling variability in nature through a coin flip
(Lehrer & Schauble, 2000). Early on, perceptual features serve as an invitation for
children to compare the cases at a relational level (Gentner, 2010). Gradually
weaning away these irrelevant perceptual features has proven to be an effective
way to scaffold understanding with analogies and models (e.g., Clement 1993;
Kotovsky & Gentner, 1996).
Investigation Skills: Searching the Experiment Space
Science and engineering practices, such as designing fair tests and interpreting
evidence generated from controlled experiments, are included at every grade level
from kindergarten through Grade 12 in the Next Generation Science Standards
(NGSS Lead States, 2013). The design of an experiment to answer a question or to
test a hypothesis can be construed as a problem to be solved, via search in a space of
experiments (Klahr, 2000; Newell & Simon, 1972). Of course, experimentation is
just one of several types of legitimate scientific inquiry processes that are involved in
planning and carrying out investigations (see Lehrer, Schauble, & Petrosino, 2001),
but here we focus on the substantial body of literature on ways to improve children’s
experimentation skills. In the following sections, we first describe research on the
developmental precursors of experimentation skills, followed by studies in which
participants are engaged in the full cycle of experimentation.
Early experimentation skills. Science education for young children tends to focus
on investigation skills such as observing, describing, comparing, and exploring
(NAEYC, 2012; NSTA, 2007). Until fairly recently, Piaget’s stage theory (e.g.,
Inhelder & Piaget 1958; Piaget, 1970) was used to justify waiting until adolescence
before attempting to teach science process skills (French & Woodring, 2013; Metz,
1995, 1997). However, an accumulation of evidence about human learning (e.g.,
NRC, 2000) has resulted in a more nuanced story about the developmental course of
experimentation and investigation skills and the extent to which well-designed
instruction can accelerate that development (NRC, 2007). Learning to conduct
experiments involves the coordination of several component processes such as
identifying and manipulating variables and observing and measuring outcomes.
Not until the later school years, after extended instruction, scaffolding, and practice
can children successfully coordinate all of these steps (e.g., Kuhn et al., 2000).
Several studies have examined the precursors of the later ensemble of experimenta-
tion skills.
One of the fundamental skills in experimental design is the ability to construct a
situation in which causal factors can be unambiguously identified. In her classic
study, Tschirgi (1980) presented children and adults with a variety of everyday
problem-solving situations (e.g., baking cakes, making paper airplanes) that
involved a positive or negative outcome and several potential causal variables such
14. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 80 [67–99] 5.11.2018 11:50AM
80 david klahr, corinne zimmerman, and bryan j. matlen
as “John baked a cake using honey, white flour, and butter, and it turned out terrible”
or “Susan made a paper airplane and it turned out great.” The character would
propose a hypothesis about a variable that may have caused the outcome (e.g.,
“John thinks that the honey made it taste bad” in the cake story). The participant in
the study would then be asked to select one of three options in order to help the
character (John) test the hypothesis. In the vary-one-thing-at-a-time (VOTAT)
option, the proposed variable was changed, but the others were kept the same (e.g.,
bake another cake with everything the same except the sweetener: use sugar instead
of honey). This strategy would produce an unconfounded experiment. In the hold-
one-thing-at-a-time (HOTAT) option, the hypothesized variable was kept the same
but the other variables were changed (e.g., bake another cake with the same sweet-
ener but change the type of flour and shortening). The change-all (CA) option
consisted of changing all of the variables (bake a cake with sugar, wholewheat
flour, and margarine). All participants were more likely to select the HOTAT strategy
when the outcome was positive. That is, the presumed causal variable was not
changed, while all the other variables were changed (thus producing a confounded
experiment) in the hope of maintaining the positive outcome. For a negative out-
come, the logically correct VOTAT strategy (consistent with a controlled experi-
ment) was chosen more frequently than HOTAT or CA, suggesting that participants
were searching for the one variable to change in order to eliminate the negative
outcome. Although 2nd and 4th graders were more likely to select the CA strategy
for the negative outcomes (hoping to eliminate all possible offending variables at
once), all participants were influenced by the desire to reproduce good effects and
eliminate bad effects by choosing a strategy based on pragmatic outcomes (rather
than logical grounds).
Croker and Buchanan (2011) used a task similar to Tschirgi’s, but included
contexts for which three-and-a-half–year-olds to eleven-year-olds held strong prior
beliefs (e.g., the effect of cola vs. milk on dental health). For all age groups, there was
an interaction of prior belief and outcome type. The logically correct VOTAT
strategy was more likely to be selected under two conditions: (1) when the outcome
was positive (i.e., healthy teeth) and consistent with prior belief or (2) when the
outcome was negative (i.e., unhealthy teeth) and inconsistent with prior belief. Even
the youngest children were influenced by the context and the plausibility of the
domain-specific content of the situations that they were reasoning about.
In what has become a classic example of an ingenious study of young children’s
scientific reasoning, Sodian, Zaitchik, and Carey (1991) presented 1st and 2nd grade
children with the challenge of designing a simple experiment to distinguish between
two possible causal factors. Children were told that they had to figure out whether
their home contained a large mouse or a small mouse. Children were shown “mouse
houses” in which they could put some food that mice like. One house had a door
through which either a large or a small mouse could pass. The other house had a door
that only a small mouse could traverse. In the “find out” condition, the children were
asked to decide which house should be used to determine the size of the mouse (i.e.,
to test a hypothesis). Of course, if the house with the small door is used, and the food
is gone in the morning, then only a small mouse could have taken the food. If the food
15. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 81 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 81
remains, they have a large (and now hungry!) mouse. Importantly, Sodian and
colleagues had a second condition, the “feed” condition, in which children were
asked what house to use if they wanted to make sure that the mouse would get fed no
matter what his size. If a child can distinguish between the goals of testing a
hypothesis with an experiment versus generating an effect (i.e., feeding the
mouse), then he or she should select the small house in the find out condition and
the large house in the feed condition. Sodian and colleagues found that children as
young as six could distinguish between a conclusive and inconclusive experimental
test of a simple hypothesis when provided with the two mutually exclusive and
exhaustive hypotheses or experiments. Piekny and Maehler (2013) used the mouse
house task with preschoolers (four- and five-year-olds) and school children (seven-,
nine-, and eleven-year-olds). It was not until age nine that children scored signifi-
cantly above chance, and not until age seven (a year later than in the Sodian et al.
study) that children showed a recognition of, and justification for, conclusive or
inconclusive tests of a hypothesis.
Klahr, Fay, and Dunbar (1993) investigated developmental differences in adults’
and 3rd and 6th grade children’s experimentation skills by presenting them with a
programmable toy robot, in which participants first mastered most of the basic
commands (see Figure 4.2). They were then challenged to find out how a “mystery
key” worked by writing and then running programs that included the mystery key. In
order to constrain the “hypothesis space” participants were provided with various
hypotheses about the mystery key (only one of which was correct). Some examples
of what the mystery key might do include (1) repeat the whole program N times, (2)
repeat the last step N times, and (3) repeat the last N steps once. Some of these
hypotheses were deemed highly plausible (i.e., likely to be correct) and others were
deemed implausible. When presented with a hypothesis that was plausible, all
participants set up experiments to demonstrate the correctness of the hypothesis
(e.g., Experiment 1 in Figure 4.2).
When given an implausible hypothesis to test, adults and some 6th graders
proposed a plausible rival hypothesis and set up an experiment that would discrimi-
nate between the two. The 3rd graders also proposed a plausible rival hypothesis but
got sidetracked in the attempt to demonstrate that the rival plausible hypothesis was
correct. Klahr, Fay, and Dunbar (1993) identified two useful heuristics that partici-
pants used: (1) design experiments that produce informative and interpretable results
and (2) attend to one feature at a time. The 3rd and 6th grade children were far less
likely than adult participants to restrict the search of possible experiments to those
that were informative.
Bullock and Ziegler (1999) collected longitudinal data on participants, starting
when they were age eight and following them through to age twelve. They examined
the process skills required for experimentation, using separate assessments to tease
apart an understanding of experimentation from the ability to produce controlled
experiments. When the children were eight-year-olds they were able to recognize a
controlled experimental test. The ability to produce a controlled experiment at levels
comparable to adults did not occur until the children were in the 6th grade. This study
16. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 82 [67–99] 5.11.2018 11:50AM
82 david klahr, corinne zimmerman, and bryan j. matlen
Figure 4.2 The control panel and two sample programs for discovering how the
“mystery key” (labeled “RPT”) works on a simulated robot (shown in its “home”
position in the center of the screenshot) (after Klahr, Fay, & Dunbar, 1993)
Hypothesis X (Hx) is that if a number N is appended to the RPT key, then the robot
will repeat the entire preceding program N times. Hypothesis Y (Hy) is that the
robot will repeat only the Nth step once.
For the first experiment (E1), the participant runs a very short program with only
one step preceding the RPT key, that instructs the robot to go forward one unit,
and then to do whatever RPT does, one time. Because the program is so short,
both hypotheses make the same prediction about the robot’s behavior and the
robot’s behavior is consistent with both.
For the second experiment (E2), a longer program with more steps is used: move
forward one unit, fire laser cannon twice, backup 1 unit. Then the RPT 2 command
is encountered. Hypothesis X predicts that the entire three-step program preced-
ing RPT will be executed two times. Hypothesis Y predicts that the full program
will be executed, followed by one repeat of the 2nd step (the FIRE2 command).
The robot’s behavior does not match either of these two predictions. Instead, it
repeats the last two steps in the program once, thus refuting both hypotheses. In
reality, RPT repeats the last N steps, preceding the RPT command, one time.
provides additional support for the idea that young children are able to understand
the “logic” of experiments long before they are able to produce them.
When task demands are reduced – such as in simple story problems or when one
can select (rather than produce) an experimental test from a set of a few alternatives –
even young children show competence with rudimentary science process skills.
Children, like adults, are sensitive to the context and the content of what is being
reasoned about. Such precursors are important for understanding the challenges of
17. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 83 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 83
teaching students how to conduct scientific investigations and the types of factors
that can be used to facilitate or scaffold developing skills.
Planning and carrying out investigations. Much of the research on the develop-
ment of investigation skills in older children and adults involves presenting partici-
pants with a multivariable causal system, such as physical apparatus or a computer
simulation. The participants’ goal is to investigate the system so as to identify the
causal and noncausal variables in the system; they propose hypotheses, make pre-
dictions, plan and conduct experiments, collect and evaluate evidence, make infer-
ences, and draw conclusions in the form of either new or updated knowledge (i.e.,
although the focus is on cells B or E in Table 4.1, other scientific practices come into
play). For example, Schauble’s (1996) participants conducted experiments in hydro-
dynamics, where the goal was to determine which variables have an effect on boat
One foundational, and domain-general, science process skill is the control-of-
variables strategy (CVS). The fundamental goal of an experiment is to unambigu-
ously identify causal factors and their effects, and the essential procedure for doing
this is to contrast conditions that differ only with the respect to the variable whose
causal status is under investigation. Procedurally, CVS includes the ability to create
experiments in which conditions differ with respect to only a single contrasting
variable as well as the ability to recognize confounded and unconfounded experi-
ments. Conceptually, CVS involves the ability to make appropriate inferences from
the results of unconfounded experiments (e.g., that only inferences about the causal
status of the variable being tested are warranted) as well as an awareness of “the
inherent indeterminacy of confounded experiments” (Chen & Klahr, 1999, p. 1098).
The conceptual aspects of CVS are relevant for argumentation and reasoning about
causality in science and everyday life, as CVS includes an understanding of the
invalidity of evidence from confounded experiments (or observations) and the
importance of comparing controlled conditions (Kuhn, 2005). Thus, CVS is relevant
to broader educational and societal goals, such as inquiry, reasoning skills, and
critical thinking.
Mastery of CVS is required for successful inquiry learning as it enables students to
conduct their own informative investigations. However, without instruction, stu-
dents – and even adults – have poor inquiry skills (e.g., Kuhn, 2007; for review, see
Zimmerman & Croker, 2014). Siler and Klahr (2012) identified the various “mis-
conceptions” that students have about controlling variables. Typical mistakes
include (1) designing experiments that vary the wrong (or “nontarget”) variable,
(2) varying more than one variable, and (3) not varying anything between the
contrasted experimental conditions (i.e., overextending the “fairness” idea so both
conditions are identical). Methods for determining children’s mastery of CVS have
varied from the kind of typical high-stakes test item shown in Figure 4.3 to computer-
based interactive assessments of the kind shown in Figure 4.4.
A recent meta-analysis of CVS instructional interventions (Schwichow, Croker, et
al., 2016) summarized the results of seventy-two studies. Possible moderators of the
overall effect size included design features (e.g., quasi-experimental vs.
18. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 84 [67–99] 5.11.2018 11:50AM
84 david klahr, corinne zimmerman, and bryan j. matlen
Bean Growth.
(Source: Texas Assessment of Knowledge and Skills (TAKS), 2003; 5th-Grade item)
Which of these questions can
be answered from the results of
this experiment?
Tap water Seawater a. Do beans need light in order
to grow?
b. Can beans grow faster in
Paper towel groups of eight?
c. Does seawater affect bean
d. How much water is needed
for beans to grow?
Figure 4.3 A typical item from a “high-stakes” state assessment of
domain-general experimentation skills and knowledge
experimental studies), instructional features (e.g., use of demonstrations), training
features (e.g., use of hands-on experiences), and assessment features (e.g., test
format). Of the various instructional features coded for, only two were found to be
effective: (1) interventions that induced a cognitive conflict and (2) teacher demon-
strations of good experimental design. In this context, a teacher draws attention to a
particular (confounded) comparison and asks what conclusions can be drawn about
the effect of a particular variable. For example, to return to the cake baking example
described earlier in the section, a teacher might note that although the cake made
with butter, wholewheat flour, and sugar tasted much better than the cake made with
margarine, white flour, and sugar, one could not tell for sure if the effect was due to
the type of flour or the type of sweetener. Because the comparison was confounded,
with two possible causal factors, either one of these potential causes might have
determined the outcome.
Cognitive conflict is induced in students by drawing attention to a current experi-
mental procedure or interpretation of data; the teacher attempts to get the student to
notice that the comparison is confounded or that the conclusion is invalid or inde-
terminate (Adey & Shayer, 1990). Interestingly, the cognitive conflict technique is
often presented via a demonstration by the teacher and so additional research is
necessary to disentangle the unique effects of these two instructional techniques
(Schwichow, Croker et al., 2016). Other instructional techniques that are often
presumed to be important, such as the need for “hands-on” engagement with experi-
mental materials, did not have an impact on student learning of CVS. And when
hands-on procedures are used, at least one study demonstrated that it does not matter
whether students’ hands are on physical or virtual materials (Triona & Klahr, 2008).
In a follow-up to the meta-analysis, Schwichow, Zimmerman, and colleagues (2016)
determined that it is important for there to be a match between the way students learn
CVS and the test format used to assess the extent to which they have learned it.
19. C:/ITOOLS/WMS/CUP-NEW/15359521/WORKINGFOLDER/DUNLOSKY/9781108416016C04.3D 85 [67–99] 5.11.2018 11:50AM
Improving Students’ Scientific Thinking 85
Q5. Suppose you want to find out what affects how much people can remember.
You can do this by setting up two ways to memorize and then comparing how
much people can remember.
There are three things that might make a difference in how much people can remember:
• The lighting of the room (Bright or Dim)
• A type of flashcard to memorize (10 Words or 10 Pictures)
• How much time they have to memorize (15 Seconds or 45 Seconds)
a. Figure out a way to find out whether the lighting of the room makes a difference in
how much people can remember.
For each person, choose the lighting of the room (Bright or Dim), a type of
flashcard to memorize (10 Words or 10 Pictures), and how much time they’re given
(15 seconds or 45 seconds).
Lighting Lighting
Bright Dim
Flashcards Flashcards
10 Pictures 10 Pictures
Time Time
45 Seconds 45 Seconds
Figure 4.4 Typical computer-interface item for assessing children’s ability to
design unconfounded experiments as part of their CVS training
Evaluating Evidence
The goal of most experiments is to produce evidence that bears on a hypothesis, and
once that evidence is generated it must be interpreted. (We say “most” here because,
in some cases, scientists may perform experiments in the absence of any clearly
articulated hypothesis, just to get a “feel” for the nature of the phenomenon.) The
final cognitive process and scientific practices we will discuss are those that enable
people to evaluate, analyze, and explain how evidence relates to the hypothesis that
inspired it (i.e., cells C and F in Table 4.1). Evidence evaluation is the part of the
cycle of inquiry aimed at determining whether the result of an experiment (or set of
experiments) is sufficient to reject or accept a hypothesis under consideration (or
whether the evidence is inconclusive), and to construct possible explanations for
how the hypothesis and evidence are related.
Evaluating patterns of evidence. One method of examining the developmental
precursors of skilled evidence evaluation with children involves presenting them
with pictorial representations of potential causes and effects. These are often simple