There is no unitary activity called “scientific discovery”; there are activities of designing experiments, gathering data, inventing and

developing observational instruments, formulating and modifying theories, deducing consequences from theories, making
predictions from theories, testing theories, inducing regularities and invariants from data, discovering theoretical constructs, and
others.
—Simon, Langley, & Bradshaw, 1981, p. 2

What Is Scientific Thinking and Reasoning?

There are two kinds of thinking we call “scientific.” The first, and most obvious, is thinking about the content of
science. People are engaged in scientific thinking when they are reasoning about such entities and processes as force,
mass, energy, equilibrium, magnetism, atoms, photosynthesis, radiation, geology, or astrophysics (and, of course,
cognitive psychology!). The second kind of scientific thinking includes the set of reasoning processes that permeate the
field of science: induction, deduction, experimental design, causal reasoning, concept formation, hypothesis testing,
and so on. However, these reasoning processes are not unique to scientific thinking: They are the very same
processes involved in everyday thinking. As Einstein put it:
The scientific way of forming concepts differs from that which we use in our daily life, not basically, but merely in the more
precise definition of concepts and conclusions; more painstaking and systematic choice of experimental material, and greater
logical economy. (The Common Language of Science, 1941, reprinted in Einstein, 1950, p. 98)

Nearly 40 years after Einstein’s remarkably insightful statement, Francis Crick offered a similar perspective: that
great discoveries in science result not from extraordinary mental processes, but rather from rather common ones. The
greatness of the discovery lies in the thing discovered.

I think what needs to be emphasized about the discovery of the double helix is that the path to it was, scientifically speaking,
fairly commonplace. What was important was not the way it was discovered, but the object discovered—the structure of DNA
itself. (Crick , 1988, p. 67; emphasis added)

Under this view, scientific thinking involves the same general-purpose cognitive processes—such as induction,
deduction, analogy, problem solving, and causal reasoning—that humans apply in non- scientific domains. These
processes are covered in several different chapters of this handbook: Rips, Smith, & Medin, Chapter 11 on
induction; Evans, Chapter 8 on deduction; Holyoak, Chapter 13 on analogy; Bassok & Novick, Chapter 21 on
problem solving; and Cheng & Buehner, Chapter 12 on causality. One might question the claim that the highly
specialized procedures associated with doing science in the “real world” can be understood by investigating the
thinking processes used in laboratory studies of the sort described in this volume. However, when the focus is on major
scientific breakthroughs, rather than on the more routine, incremental progress in a field, the psychology of problem
solving provides a rich source of ideas about how such discoveries might occur. As Simon and his colleagues put it:

It is understandable, if ironic, that ‘normal’ science fits . . . the description of expert problem solving, while ‘revolutionary’
science fits the description of problem solving by novices. It is understandable because scientific activity, particularly at the
revolutionary end of the continuum, is concerned with the discovery of new truths, not with the application of truths that are
already well-known . . . it is basically a journey into unmapped terrain. Consequently, it is mainly characterized, as is novice
problem solving, by trial-and-error search. The search may be highly selective—but it reaches its goal only after many halts,
turnings, and back- trackings. (Simon, Langley, & Bradshaw,
1981, p. 5)

The research literature on scientific thinking can be roughly categorized according to the two types of scientific
thinking listed in the opening paragraph of this chapter: (1) One category focuses on thinking that directly involves
scientific content. Such research ranges from studies of young children reasoning about the sun-moon-earth system
(Vosniadou & Brewer, 1992) to college students reasoning about chemical equilibrium (Davenport, Yaron, Klahr, &
Koedinger, 2008), to research that investigates collaborative problem solving by world-class researchers in real-world
molecular biology labs (Dunbar,
1995). (2) The other category focuses on “general” cognitive processes, but it tends to do so by analyzing people’s
problem-solving behavior when they are presented with relatively complex situations that involve the integration and
coordination of several different types of processes, and that are designed to capture some essential features of
“real-world” science in the psychology laboratory (Bruner, Goodnow, & Austin, 1956; Klahr & Dunbar, 1988;
Mynatt, Doherty, & Tweney, 1977).
There are a number of overlapping research traditions that have been used to investigate scientific thinking. We
will cover both the history of research on scientific thinking and the different approaches that have been used,
highlighting common themes that have emerged over the past 50 years of research.

A Brief History of Research on Scientific Thinking

Science is often considered one of the hallmarks of the human species, along with art and literature. Illuminating
the thought processes used in science thus reveal key aspects of the human mind. The thought processes
underlying scientific thinking have fascinated both scientists and nonscientists because the products of science
have transformed our world and because the process of discovery is shrouded in mystery. Scientists talk of the
chance discovery, the flash of insight, the years of perspiration, and the voyage of discovery. These images of
science have helped make the mental processes underlying the discovery process intriguing to cognitive scientists as
they attempt to uncover what really goes on inside the scientific mind and how scientists really think.
Furthermore, the possibilities that scientists can be taught to think better by avoiding mistakes that have been clearly
identified in research on scientific thinking, and that their scientific process could be partially automated, makes
scientific thinking a topic of enduring interest.
The cognitive processes underlying scientific discovery and day-to-day scientific thinking have been a topic of
intense scrutiny and speculation for almost 400 years (e.g., Bacon, 1620; Galilei, 1638; Klahr, 2000; Tweney,
Doherty, & Mynatt, 1981). Understanding the nature of scientific thinking has been a central issue not only for our
understanding of science but also for our understating of what it is to be human. Bacon’s Novumm Organum in 1620
sketched out some of the key features of the ways that experiments are designed and data interpreted. Over the
ensuing 400 years philosophers and scientists vigorously debated about the appropriate methods that scientists
should use (see Giere, 1993). These debates over the appropriate methods for science typically resulted in the espousal
of a particular type of reasoning method, such as induction or deduction. It was not until the Gestalt psychologists
began working on the nature of human problem solving, during the 1940s, that experimental psychologists began
to investigate the cognitive processes underlying scientific thinking and reasoning.
The Gestalt psychologist Max Wertheimer pioneered the investigation of scientific thinking (of the first type
described earlier: thinking about scientific content) in his landmark book Productive Thinking (Wertheimer, 1945).
Wertheimer spent a considerable amount of time corresponding with Albert Einstein, attempting to discover how
Einstein generated the concept of relativity. Wertheimer argued that Einstein had to overcome the structure of
Newtonian physics at each step in his theorizing, and the ways that Einstein actually achieved this restructuring
were articulated in terms of Gestalt theories. (For a recent and different account of how Einstein made his discovery,
see Galison, 2003.) We will see later how this process of overcoming alternative theories is an obstacle that both
scientists and nonscientists need to deal with when evaluating and theorizing about the world.
One of the first investigations of scientific thinking of the second type (i.e., collections of general- purpose
processes operating on complex, abstract, components of scientific thought) was carried out by Jerome Bruner and
his colleagues at Harvard (Bruner et al., 1956). They argued that a key activity engaged in by scientists is to
determine whether a particular instance is a member of a category. For example, a scientist might want to discover
which substances undergo fission when bombarded by neutrons and which substances do not. Here, scientists
have to discover the attributes that make a substance undergo fission. Bruner et al. saw scientific thinking as the
testing of hypotheses and the collecting of data with the end goal of determining whether something is a member of a
category. They invented a paradigm where people were required to formulate hypotheses and collect data that test their
hypotheses. In one type of experiment, the participants were shown a card such as one with two borders and
three green triangles. The participants were asked to determine the concept that this card represented by choosing
other cards and getting feedback from the experimenter as to whether the chosen card was an example of the
concept. In this case the participant may have thought that the concept was green and chosen a card with two green
squares and one border. If the underlying concept was green, then the experimenter would say that the card was an
example of the concept. In terms of scientific thinking, choosing a new card is akin to conducting an experiment,
and the feedback from the experimenter is similar to knowing whether a hypothesis is confirmed or disconfirmed.
Using this approach, Bruner et al. identified a number of strategies that people use to formulate and test hypotheses.
They found that a key factor determining which hypothesis-testing strategy that people use is the amount of
memory capacity that the strategy takes up (see also Morrison & Knowlton, Chapter 6; Medin et al., Chapter 11).
Another key factor that they discovered was that it was much more difficult for people to discover negative concepts
(e.g., not blue) than positive concepts (e.g., blue). Although Bruner et al.’s research is most commonly viewed as work
 on concepts, they saw their work as uncovering a key component of scientific thinking.
A second early line of research on scientific thinking was developed by Peter Wason and his colleagues (Wason, 1968).
Like Bruner et al., Wason saw a key component of scientific thinking as being the testing of hypotheses. Whereas
Bruner et al. focused on the different types of strategies that people use to formulate hypotheses, Wason focused on
whether people adopt a strategy of trying to confirm or disconfirm their hypotheses. Using Popper’s (1959) theory
that scientists should try and falsify rather than confirm their hypotheses, Wason devised a deceptively simple task
in which participants were given three numbers, such as 2-4-6, and were asked to discover the rule underlying the
three numbers. Participants were asked to generate other triads of numbers and the experimenter would tell the
participant whether the triad was consistent or inconsistent with the rule. They were told that when they were sure
they knew what the rule was they should state it. Most participants began the experiment by thinking that the rule
was even numbers increasing by 2. They then attempted to confirm their hypothesis by generating a triad like 8-
10-12, then 14-16-18. These triads are consistent with the rule and the participants were told yes, that the triads
were indeed consistent with the rule. However, when they proposed the rule—even numbers increasing by 2—they
were told that the rule was incorrect. The correct rule was numbers of increasing magnitude! From this research,
Wason concluded that people try to confirm their hypotheses, whereas normatively speaking, they should try to
disconfirm their hypotheses. One implication of this research is that confirmation bias is not just restricted to scientists
but is a general human tendency.
It was not until the 1970s that a general account of scientific reasoning was proposed. Herbert Simon, often in
collaboration with Allan Newell, proposed that scientific thinking is a form of problem solving. He proposed that
problem solving is a search in a problem space. Newell and Simon’s theory of problem solving is discussed in many
places in this handbook, usually in the context of specific problems (see especially Bassok & Novick, Chapter 21).
Herbert Simon, however, devoted considerable time to understanding many different scientific discoveries and scientific
reasoning processes. The common thread in his research was that scientific thinking and discovery is not a
mysterious magical process but a process of problem solving in which clear heuristics are used. Simon’s goal was to
articulate the heuristics that scientists use in their research at a fine-grained level. By constructing computer pro-
grams that simulated the process of several major scientific discoveries, Simon and colleagues were able to
articulate the specific computations that scientists could have used in making those discoveries (Langley, Simon,
Bradshaw, & Zytkow, 1987; see section on “Computational Approaches to Scientific Thinking”). Particularly
influential was Simon and Lea’s (1974) work demonstrating that concept formation and induction consist of a
search in two problem spaces: a space of instances and a space of rules. This idea has influenced problem-solving
accounts of scientific thinking that will be discussed in the next section.
Overall, the work of Bruner, Wason, and Simon laid the foundations for contemporary research on scientific
thinking. Early research on scientific thinking is summarized in Tweney, Doherty and Mynatt’s 1981 book On Scientific
Thinking, where they sketched out many of the themes that have dominated research on scientific thinking over the
past few decades. Other more recent books such as Cognitive Models of Science (Giere, 1993), Exploring Science
(Klahr, 2000), Cognitive Basis of Science (Carruthers, Stich, & Siegal, 2002), and New Directions in Scientific and
Technical Thinking (Gorman, Kincannon, Gooding, & Tweney, 2004) provide detailed analyses of different aspects of
scientific discovery. Another important collection is Vosnadiau’s handbook on conceptual change research
(Vosniadou, 2008). In this chapter, we discuss the main approaches that have been used to investigate scientific
thinking.
How does one go about investigating the many different aspects of scientific thinking? One common approach to
the study of the scientific mind has been to investigate several key aspects of scientific thinking using abstract tasks
designed to mimic some essential characteristics of “real-world” science. There have been numerous methodologies that
have been used to analyze the genesis of scientific concepts, theories, hypotheses, and experiments. Researchers have used
experiments, verbal protocols, computer programs, and analyzed particular scientific discoveries. A more recent
development has been to increase the ecological validity of such research by investigating scientists as they reason
“live” (in vivo studies of scientific thinking) in their own laboratories (Dunbar, 1995, 2002). From a “Thinking and
Reasoning” stand- point the major aspects of scientific thinking that have been most actively investigated are
problem solving, analogical reasoning, hypothesis testing, conceptual change, collaborative reasoning, inductive
reasoning, and deductive reasoning.

Scientific Thinking as Problem Solving

 One of the primary goals of accounts of scientific thinking has been to provide an overarching framework to
understand the scientific mind. One frame- work that has had a great influence in cognitive science is that scientific
thinking and scientific discovery can be conceived as a form of problem solving. As noted in the opening section of
this chapter, Simon (1977; Simon, Langley, & Bradshaw, 1981) argued that both scientific thinking in general and
problem solving in particular could be thought of as a search in a problem space. A problem space consists of all the
possible states of a problem and all the operations that a problem solver can use to get from one state to the next.
According to this view, by characterizing the types of representations and procedures that people use to get from
one state to another it is possible to understand scientific thinking. Thus, scientific thinking can be characterized as a
search in various problem spaces (Simon, 1977). Simon investigated a number of scientific discoveries by bringing
participants into the laboratory, pro- viding the participants with the data that a scientist had access to, and getting the
participants to reason about the data and rediscover a scientific concept. He then analyzed the verbal protocols that
participants generated and mapped out the types of problem spaces that the participants search in (e.g., Qin & Simon,
1990). Kulkarni and Simon (1988) used a more historical approach to uncover the problem- solving heuristics that
Krebs used in his discovery of the urea cycle. Kulkarni and Simon analyzed Krebs’s diaries and proposed a set of
problem-solving heuristics that he used in his research. They then built a computer program incorporating the
heuristics and biological knowledge that Krebs had before he made his discoveries. Of particular importance are the
search heuristics that the program uses, which include experimental proposal heuristics and data interpretation
heuristics. A key heuristic was an unusualness heuristic that focused on unusual findings, which guided search
through a space of theories and a space of experiments.
Klahr and Dunbar (1988) extended the search in a problem space approach and proposed that scientific
thinking can be thought of as a search through two related spaces: an hypothesis space and an experiment space.
Each problem space that a scientist uses will have its own types of representations and operators used to change the
representations. Search in the hypothesis space constrains search in the experiment space. Klahr and Dunbar found that
some participants move from the hypothesis space to the experiment space, whereas others move from the experiment
space to the hypothesis space. These different types of searches lead to the proposal of different types of
hypotheses and experiments. More recent work has extended the dual-space approach to include alternative
problem-solving spaces, including those for data, instrumentation, and domain-specific knowledge (Klahr & Simon,
1999; Schunn & Klahr, 1995, 1996).

Scientific Thinking as Hypothesis Testing

Many researchers have regarded testing specific hypotheses predicted by theories as one of the key attributes of
scientific thinking. Hypothesis testing is the process of evaluating a proposition by collecting evidence regarding its
truth. Experimental cognitive research on scientific thinking that specifically examines this issue has tended to fall into
two broad classes of investigations. The first class is concerned with the types of reasoning that lead scientists astray,
thus blocking scientific ingenuity. A large amount of research has been conducted on the potentially faulty
reasoning strategies that both participants in experiments and scientists use, such as considering only one favored
hypothesis at a time and how this prevents the scientists from making discoveries. The second class is concerned
with uncovering the mental processes underlying the generation of new scientific hypotheses and concepts. This
research has tended to focus on the use of analogy and imagery in science, as well as the use of specific types of
problem-solving heuristics.
Turning first to investigations of what diminishes scientific creativity, philosophers, historians, and
experimental psychologists have devoted a considerable amount of research to “confirmation bias.” This occurs
when scientists only consider one hypothesis (typically the favored hypothesis) and ignore other alternative
hypotheses or potentially relevant hypotheses. This important phenomenon can distort the design of experiments,
formulation of theories, and interpretation of data. Beginning with the work of Wason (1968) and as discussed
earlier, researchers have repeatedly shown that when participants are asked to design an experiment to test a
hypothesis they will predominantly design experiments that they think will yield results consistent with the
hypothesis. Using the 2-4-6 task mentioned earlier, Klayman and Ha (1987) showed that in situations where one’s
hypothesis is likely to be confirmed, seeking confirmation is a normatively incorrect strategy, whereas when the
probability of confirming one’s hypothesis is low, then attempting to confirm one’s hypothesis can be an
appropriate strategy. Historical analyses by Tweney (1989), concerning the way that Faraday made his discoveries,
and experiments investigating people testing hypotheses, have revealed that people use a confirm early, disconfirm
late strategy: When people initially generate or are given hypotheses, they try and gather evidence that is consistent
with the hypothesis. Once enough evidence has been gathered, then people attempt to find the boundaries of their
hypothesis and often try to disconfirm their hypotheses.
In an interesting variant on the confirmation bias paradigm, Gorman (1989) showed that when participants are told
that there is the possibility of error in the data that they receive, participants assume that any data that are
inconsistent with their favored hypothesis are due to error. Thus, the possibility of error “insulates” hypotheses
against disconfirmation. This intriguing hypothesis has not been con- firmed by other researchers (Penner & Klahr,
1996), but it is an intriguing hypothesis that warrants further investigation.
Confirmation bias is very difficult to overcome. Even when participants are asked to consider alternate
hypotheses, they will often fail to conduct experiments that could potentially disconfirm their hypothesis. Tweney
and his colleagues provide an excellent overview of this phenomenon in their classic monograph On Scientific
Thinking (1981). The precise reasons for this type of block are still widely debated. Researchers such as Michael
Doherty have argued that working memory limitations make it difficult for people to consider more than one
hypothesis. Consistent with this view, Dunbar and Sussman (1995) have shown that when participants are asked to
hold irrelevant items in working memory while testing hypotheses, the participants will be unable to switch
hypotheses in the face of inconsistent evidence. While working memory limitations are involved in the
phenomenon of confirmation bias, even groups of scientists can also display confirmation bias. For example, the
controversy over cold fusion is an example of confirmation bias. Here, large groups of scientists had other hypotheses
available to explain their data yet maintained their hypotheses in the face of other more standard alternative hypotheses.
Mitroff (1974) provides some interesting examples of NASA scientists demonstrating confirmation bias, which
highlight the roles of commitment and motivation in this process. See also MacPherson and Stanovich (2007) for
specific strategies that can be used to overcome confirmation bias.

Causal Thinking in Science

Much of scientific thinking and scientific theory building pertains to the development of causal models between
variables of interest. For example, do vaccines cause illnesses? Do carbon dioxide emissions cause global warming?
Does water on a planet indicate that there is life on the planet? Scientists and nonscientists alike are constantly
bombarded with statements regarding the causal relationship between such variables. How does one evaluate the
status of such claims? What kinds of data are informative? How do scientists and nonscientists deal with data that
are inconsistent with their theory?
A central issue in the causal reasoning literature, one that is directly relevant to scientific thinking, is the extent to
which scientists and nonscientists alike are governed by the search for causal mechanisms (i.e., how a variable
works) versus the search for statistical data (i.e., how often variables co-occur). This dichotomy can be boiled down
to the search for qualitative versus quantitative information about the paradigm the scientist is investigating.
Researchers from a number of cognitive psychology laboratories have found that people prefer to gather more
information about an underlying mechanism than covariation between a cause and an effect (e.g., Ahn, Kalish, Medin,
& Gelman, 1995). That is, the predominant strategy that students in simulations of scientific thinking use is to gather as
much information as possible about how the objects under investigation work, rather than collecting large amounts of
quantitative data to determine whether the observations hold across multiple samples. These findings suggest that a
central component of scientific thinking may be to formulate explicit mechanistic causal models of scientific events.
One type of situation in which causal reasoning has been observed extensively is when scientists obtain
unexpected findings. Both historical and naturalistic research has revealed that reasoning causally about
unexpected findings plays a central role in science. Indeed, scientists themselves frequently state that a finding
was due to chance or was unexpected. Given that claims of unexpected findings are such a frequent component
of scientists’ autobiographies and interviews in the media, Dunbar (1995, 1997, 1999; Dunbar & Fugelsang, 2005;
Fugelsang, Stein, Green, & Dunbar, 2004) decided to investigate the ways that scientists deal with unexpected
findings. In 1991–1992 Dunbar spent 1 year in three molecular biology laboratories and one immunology
laboratory at a prestigious U.S. university. He used the weekly laboratory meeting as a source of data on scientific
discovery and scientific reasoning. (He termed this type of study “in vivo” cognition.) When he looked at the types
of findings that the scientists made, he found that over 50% of the findings were unexpected and that these scientists
had evolved a number of effective strategies for dealing with such findings. One clear strategy was to reason causally
about the findings: Scientists attempted to build causal models of their unexpected findings. This causal model
building results in the extensive use of collaborative reasoning, analogical reasoning, and problem-solving heuristics
(Dunbar, 1997, 2001).
Many of the key unexpected findings that scientists reasoned about in the in vivo studies of scientific thinking were
 inconsistent with the scientists’ preexisting causal models. A laboratory equivalent of the biology labs involved
creating a situation in which students obtained unexpected findings that were inconsistent with their preexisting
theories. Dunbar and Fugelsang (2005) examined this issue by creating a scientific causal thinking simulation where
experimental outcomes were either expected or unex- pected. Dunbar (1995) has called the study of people reasoning in a
cognitive laboratory “in vitro” cogni- tion. These investigators found that students spent considerably more time
reasoning about unexpected findings than expected findings. In addition, when assessing the overall degree to which
their hypothesis was supported or refuted, participants spent the major- ity of their time considering unexpected findings.
An analysis of participants’ verbal protocols indicates that much of this extra time was spent formulating causal models
for the unexpected findings. Similarly, scientists spend more time considering unexpected than expected findings, and
this time is devoted to building causal models (Dunbar & Fugelsang, 2004).
Scientists know that unexpected findings occur often, and they have developed many strategies to take advantage
of their unexpected findings. One of the most important places that they anticipate the unexpected is in designing
experiments (Baker & Dunbar, 2000). They build different causal models of their experiments incorporating many
conditions and controls. These multiple conditions and controls allow unknown mechanisms to manifest
themselves. Thus, rather than being the victims of the unexpected, they create opportunities for unexpected events to
occur, and once these events do occur, they have causal models that allow them to determine exactly where in the
causal chain their unexpected finding arose. The results of these in vivo and in vitro studies all point to a more complex
and nuanced account of how scientists and nonscientists alike test and evaluate hypotheses about theories.

The Roles of Inductive, Abductive, and Deductive Thinking in Science

One of the most basic characteristics of science is that scientists assume that the universe that we live in follows
predictable rules. Scientists reason using a variety of different strategies to make new scientific discoveries. Three
frequently used types of reasoning strategies that scientists use are inductive, abductive, and deductive reasoning. In the
case of inductive reasoning, a scientist may observe a series of events and try to discover a rule that governs the event.
Once a rule is discovered, scientists can extrapolate from the rule to formulate theories of observed and yet-to-be-
observed phenomena. One example is the discover using inductive reasoning that a certain type of bacterium is a cause
of many ulcers (Thagard, 1999). In a fascinating series of articles, Thagard documented the reasoning processes that
Marshall and Warren went through in proposing this novel hypothesis. One key reasoning process was the use of
induction by generalization. Marshall and Warren noted that almost all patients with gastric entritis had a spiral
bacterium in their stomachs, and he formed the generalization that this bacterium is the cause of stomach ulcers.
There are numerous other examples of induction by generalization in science, such as Tycho De Brea’s induction
about the motion of planets from his observations, Dalton’s use of induction in chemistry, and the discovery of prions
as the source of mad cow disease. Many theories of induction have used scientific discovery and reasoning as examples
of this important reasoning process.
Another common type of inductive reasoning is to map a feature of one member of a category to another
member of a category. This is called categorical induction. This type of induction is a way of projecting a known
property of one item onto another item that is from the same category. Thus, knowing that the Rous Sarcoma virus is
a retrovirus that uses RNA rather than DNA, a biologist might assume that another virus that is thought to be a
retrovirus also uses RNA rather than DNA. While research on this type of induction typically has not been
discussed in accounts of scientific thinking, this type of induction is common in science. For an influential
contribution to this literature, see Smith, Shafir, and Osherson (1993), and for reviews of this literature see Heit
(2000) and Medin et al. (Chapter 11).
While less commonly mentioned than inductive reasoning, abductive reasoning is an important form of reasoning
that scientists use when they are seeking to propose explanations for events such as unexpected findings (see Lombrozo,
Chapter 14; Magnani, et al., 2010). In Figure 35.1, taken from King (2011), the differences between inductive,
abductive, and deductive thinking are highlighted. In the case of abduction, the reasoner attempts to generate
explanations of the form “if situation X had occurred, could it have produced the current evidence I am attempting
to interpret?” (For an interesting of analysis of abductive reasoning see the brief paper by Klahr & Masnick, 2001). Of
course, as in classical induction, such reasoning may produce a plausible account that is still not the correct one.
However, abduction does involve the generation of new knowledge, and is thus also related to research on creativity.
Turning now to deductive thinking, many thinking processes that scientists adhere to follow traditional rules of
deductive logic. These processes correspond to those conditions in which a hypothesis may lead to, or is deducible
to, a conclusion. Though they are not always phrased in syllogistic form, deductive arguments can be phrased as
“syllogisms,” or as brief, mathematical statements in which the premises lead to the conclusion. Deductive reasoning is
an extremely important aspect of scientific thinking because it underlies a large component of how scientists conduct
their research. By looking at many scientific discoveries, we can often see that deductive reasoning is at work.
Deductive reasoning statements all contain information or rules that state an assumption about how the world works,
as well as a conclusion that would necessarily follow from the rule. Numerous discoveries in physics such as the
discovery of dark matter by Vera Rubin are based on deductions. In the dark matter case, Rubin measured galactic
rotation curves and based on the differences between the predicted and observed angular motions of galaxies she
deduced that the structure of the universe was uneven. This led her to propose that dark matter existed. In
contemporary physics the CERN Large Hadron Collider is being used to search for the Higgs Boson. The Higgs
Boson is a deductive prediction from contemporary physics. If the Higgs Boson is not found, it may lead to a radical
revision of the nature of physics and a new understanding of mass (Hecht, 2011).

The Roles of Analog y in Scientific Thinking

One of the most widely mentioned reasoning processes used in science is analogy. Scientists use analogies to form a bridge
between what they already know and what they are trying to explain, understand, or discover. In fact, many scientists
have claimed that the making of certain analogies was instrumental in their making a scientific discovery, and almost
all scientific autobiographies and biographies feature one particular analogy that is discussed in depth. Coupled with
the fact that there has been an enormous research program on analogical thinking and reasoning (see Holyoak, Chapter
13), we now have a number of models and theories of analogical reasoning that suggest how analogy can play a role in
scientific discovery (see Gentner, Holyoak, & Kokinov, 2001). By analyzing several major discoveries in the history
 of science, Thagard and Croft (1999), Nersessian (1999, 2008), and Gentner and Jeziorski (1993) have all shown that
analogical reasoning is a key aspect of scientific discovery.
Traditional accounts of analogy distinguish between two components of analogical reasoning: the target and
the source (Holyoak, Chapter 13; Gentner, 2010). The target is the concept or problem that a scientist is attempting
to explain or solve. The source is another piece of knowledge that the scientist uses to understand the target or to
explain the target to others. What the scientist does when he or she makes an analogy is to map features of the source
onto features of the target. By mapping the features of the source onto the target, new features of the target may be
discovered, or the features of the target may be rearranged so that a new concept is invented and a scientific
discovery is made. For example, a common analogy that is used with computers is to describe a harmful piece of
software as a computer virus. Once a piece of software is called a virus, people can map features of biological viruses,
such as that it is small, spreads easily, self-replicates using a host, and causes damage. People not only map individual
features of the source onto the target but also the systems of relations. For example, if a computer virus is similar to
a biological virus, then an immune system can be created on computers that can protect computers from future
variants of a virus. One of the reasons that scientific analogy is so powerful is that it can generate new knowledge, such
as the creation of a computational immune system having many of the features of a real biological immune system.
This analogy also leads to predictions that there will be newer computer viruses that are the computational
equivalent of retroviruses, lacking DNA, or standard instructions, that will elude the computational immune
system.
The process of making an analogy involves a number of key steps: retrieval of a source from memory, aligning the
features of the source with those of the target, mapping features of the source onto those of the target, and possibly
making new inferences about the target. Scientific discoveries are made when the source highlights a hitherto unknown
feature of the target or restructures the target into a new set of relations. Interestingly, research on analogy has shown
that participants do not easily use remote analogies (see Gentner et al., 1997; Holyoak & Thagard 1995). Participants in
experiments tend to focus on the sharing of a superficial feature between the source and the target, rather than the
relations among features. In his in vivo studies of science, Dunbar (1995, 2001, 2002) investigated the ways that
scientists use analogies while they are conducting their research and found that scientists use both relational and
superficial features when they make analogies. Whether they use superficial or relational features depends on their goals. If
their goal is to fix a problem in an experiment, their analogies are based upon superficial features. However, if their goal
is to formulate hypotheses, they focus on analogies based upon sets of relations. One important difference between
scientists and participants in experiments is that the scientists have deep relational knowledge of the processes that
they are investigating and can hence use this relational knowledge to make analogies (see Holyoak, Chapter 13 for a
thorough review of analogical reasoning).
Are scientific analogies always useful? Sometimes analogies can lead scientists and students astray. For example,
Evelyn Fox-Keller (1985) shows how an analogy between the pulsing of a lighthouse and the activity of the
slime mold dictyostelium led researchers astray for a number of years. Likewise, the analogy between the solar
system (the source) and the structure of the atom (the target) has been shown to be potentially misleading to
students taking more advanced courses in physics or chemistry. The solar system analogy has a number of
misalignments to the structure of the atom, such as electrons being repelled from each other rather than attracted;
moreover, electrons do not have individual orbits like planets but have orbit clouds of electron den- sity.
Furthermore, students have serious misconceptions about the nature of the solar system, which can compound their
misunderstanding of the nature of the atom (Fischler & Lichtfeld, 1992). While analogy is a powerful tool in
science, like all forms of induction, incorrect conclusions can be reached.

Conceptual Change in Science

Scientific knowledge continually accumulates as scientists gather evidence about the natural world Over extended
time, this knowledge accumulation leads to major revisions, extensions, and new organizational forms for expressing
what is known about nature. Indeed, these changes are so substantial that philosophers of science speak of
“revolutions” in a variety of scientific domains (Kuhn, 1962). The psychological literature that explores the idea of
revolutionary conceptual change can be roughly divided into (a) investigations of how scientists actually make
discoveries and integrate those discoveries into existing scientific contexts, and (b) investigations of nonscientists ranging
from infants, to children, to students in science classes. In this section we summarize the adult studies of conceptual
change, and in the next section we look at its developmental aspects.
Scientific concepts, like all concepts, can be characterized as containing a variety of “knowledge elements”:
 representations of words, thoughts, actions, objects, and processes. At certain points in the history of science, the
accumulated evidence has demanded major shifts in the way these collections of knowledge elements are organized.
This “radical conceptual change” process (see Keil, 1999; Nersessian, 1998, 2002; Thagard, 1992; Vosniadou,
1998, for reviews) requires the formation of a new conceptual system that organizes knowledge in new ways, adds
new knowledge, and results in a very different conceptual structure. For more recent research on conceptual change,
The International Handbook of Research on Conceptual Change (Vosniadou, 2008) provides a detailed compendium
of theories and controversies within the field.
While conceptual change in science is usually characterized by large-scale changes in concepts that occur over
extensive periods of time, it has been possible to observe conceptual change using in vivo methodologies. Dunbar
(1995) reported a major conceptual shift that occurred in immunologists, where they obtained a series of
unexpected findings that forced the scientists to propose a new concept in immunology that in turn forced the change in
other concepts. The drive behind this conceptual change was the discovery of a series of different unexpected findings
or anomalies that required the scientists to both revise and reorganize their conceptual knowledge. Interestingly, this
conceptual change was achieved by a group of scientists reasoning collaboratively, rather than by a scientist
working alone. Different scientists tend to work on different aspects of concepts, and also different concepts, that
when put together lead to a rapid change in entire conceptual structures.
Overall, accounts of conceptual change in individuals indicate that it is indeed similar to that of conceptual change in
entire scientific fields. Individuals need to be confronted with anomalies that their preexisting theories cannot
explain before entire conceptual structures are overthrown. However, replacement conceptual structures have to be
generated before the old conceptual structure can be discarded. Sometimes, people do not overthrow their original
conceptual theories and through their lives maintain their original views of many fundamental scientific concepts.
Whether people actively possess naive theories, or whether they appear to have a naive theory because of the
demand characteristics of the testing context, is a lively source of debate within the science education community (see
Gupta, Hammer, & Redish, 2010).

Scientific Thinking in Children

Well before their first birthday, children appear to know several fundamental facts about the physical world. For
example, studies with infants show that they behave as if they understand that solid objects endure over time (e.g.,
they don’t just disappear and reappear, they cannot move through each other, and they move as a result of
collisions with other solid objects or the force of gravity (Baillargeon, 2004; Carey, 1985; Cohen & Cashon, 2006;
Duschl, Schweingruber, & Shouse, 2007; Gelman & Baillargeon, 1983; Gelman & Kalish, 2006; Mandler,
2004; Metz, 1995; Munakata, Casey, & Diamond, 2004). And even 6-month-olds are able to predict the future
location of a moving object that they are attempting to grasp (Von Hofsten, 1980; Von Hofsten, Feng, & Spelke,
2000). In addition, they appear to be able to make nontrivial inferences about causes and their effects (Gopnik et al.,
2004).
The similarities between children’s thinking and scientists’ thinking have an inherent allure and an internal
contradiction. The allure resides in the enthusiastic wonder and openness with which both children and scientists
approach the world around them. The paradox comes from the fact that different investigators of children’s thinking
have reached diametrically opposing conclusions about just how “scientific” children’s thinking really is. Some
claim support for the “child as a scientist” position (Brewer & Samarapungavan, 1991; Gelman & Wellman,
1991; Gopnik, Meltzoff, & Kuhl, 1999; Karmiloff-Smith, 1988; Sodian, Zaitchik, & Carey, 1991; Samarapungavan,
1992), while others offer serious challenges to the view (Fay & Klahr, 1996; Kern, Mirels, & Hinshaw, 1983;
Kuhn, Amsel, & O’Laughlin, 1988; Schauble & Glaser, 1990; Siegler & Liebert, 1975.) Such fundamentally
incommensurate conclusions suggest that this very field— children’s scientific thinking—is ripe for a conceptual
revolution!
A recent comprehensive review (Duschl, Schweingruber, & Shouse, 2007) of what children bring to their
science classes offers the following concise summary of the extensive developmental and educational research
literature on children’s scientific thinking:
• Children entering school already have substantial knowledge of the natural world, much of which is implicit.
• What children are capable of at a particular age is the result of a complex interplay among maturation,
experience, and instruction. What is developmentally appropriate is not a simple function of age or grade, but
rather is largely contingent on children’s prior opportunities to learn.
• Students’ knowledge and experience play a critical role in their science learning, influencing four aspects of
 science understanding, including (a) knowing, using, and interpreting scientific explanations of the natural world;
(b) generating and evaluating scientific evidence and explanations, (c) understanding how scientific knowledge is
developed in the scientific community, and (d) participating in scientific practices and discourse.
• Students learn science by actively engaging in the practices of science.
In the previous section of this article we discussed conceptual change with respect to scientific fields and
undergraduate science students. However, the idea that children undergo radical conceptual change in which old
“theories” need to be over- thrown and reorganized has been a central topic in understanding changes in scientific
thinking in both children and across the life span. This radical conceptual change is thought to be necessary for
acquiring many new concepts in physics and is regarded as the major source of difficulty for students. The factors that
are at the root of this conceptual shift view have been difficult to determine, although there have been a number of
studies in cognitive development (Carey, 1985; Chi, 1992; Chi & Roscoe, 2002), in the history of science
(Thagard, 1992), and in physics education (Clement, 1982; Mestre, 1991) that give detailed accounts of the
changes in knowledge representation that occur while people switch from one way of representing scientific
knowledge to another.
One area where students show great difficulty in understanding scientific concepts is physics. Analyses of
students’ changing conceptions, using interviews, verbal protocols, and behavioral out-come measures, indicate
that large-scale changes in students’ concepts occur in physics education (see McDermott & Redish, 1999, for a
review of this literature). Following Kuhn (1962), many researchers, but not all, have noted that students’ changing
conceptions resemble the sequences of conceptual changes in physics that have occurred in the history of science.
These notions of radical paradigm shifts and ensuing incompatibility with past knowledge- states have called
attention to interesting parallels between the development of particular scientific concepts in children and in the
history of physics. Investigations of nonphysicists’ understanding of motion indicate that students have extensive
misunderstandings of motion. Some researchers have interpreted these findings as an indication that many people
hold erroneous beliefs about motion similar to a medieval “impetus” theory (McCloskey, Caramazza, & Green,
1980). Furthermore, stu- dents appear to maintain “impetus” notions even after one or two courses in physics. In
fact, some authors have noted that students who have taken one or two courses in physics can perform worse on
physics problems than naive students (Mestre, 1991). Thus, it is only after extensive learning that we see a
conceptual shift from impetus theories of motion to Newtonian scientific theories.
How one’s conceptual representation shifts from “naive” to Newtonian is a matter of contention, as some have
argued that the shift involves a radical conceptual change, whereas others have argued that the conceptual change is
not really complete. For example, Kozhevnikov and Hegarty (2001) argue that much of the naive impetus notions of
motion are maintained at the expense of Newtonian principles even with extensive training in physics. However, they
argue that such impetus principles are maintained at an implicit level. Thus, although students can give the correct
Newtonian answer to problems, their reaction times to respond indicate that they are also using impetus theories
when they respond. An alternative view of conceptual change focuses on whether there are real conceptual changes at
all. Gupta, Hammer and Redish (2010) and Disessa (2004) have conducted detailed investigations of changes in
physics students’ accounts of phenomena covered in elementary physics courses. They have found that rather than
students possessing a naive theory that is replaced by the standard theory, many introductory physics students have no
stable physical theory but rather construct their explanations from elementary pieces of knowledge of the physical
world.

Computational Approaches to Scientific Thinking

Computational approaches have provided a more complete account of the scientific mind. Computational
models provide specific detailed accounts of the cognitive processes underlying scientific thinking. Early
computational work consisted of taking a scientific discovery and building computational models of the reasoning
processes involved in the discovery. Langley, Simon, Bradshaw, and Zytkow (1987) built a series of programs that
simulated discoveries such as those of Copernicus, Bacon, and Stahl. These programs had various inductive
reasoning algorithms built into them, and when given the data that the scientists used, they were able to propose the
same rules. Computational models make it possible to propose detailed models of the cognitive subcomponents of
scientific thinking that specify exactly how scientific theories are generated, tested, and amended (see Darden, 1997,
and Shrager & Langley, 1990, for accounts of this branch of research). More recently, the incorporation of
scientific knowledge into computer pro- grams has resulted in a shift in emphasis from using programs to simulate
discoveries to building programs that are used to help scientists make discoveries. A number of these computer
programs have made novel discoveries. For example, Valdes-Perez (1994) has built systems for discoveries in
chemistry, and Fajtlowicz has done this in mathematics (Erdos, Fajtlowicz, & Staton, 1991).
These advances in the fields of computer discovery have led to new fields, conferences, journals, and even
departments that specialize in the development of programs devised to search large databases in the hope of making
new scientific discoveries (Langley, 2000, 2002). This process is commonly known as “data mining.” This
approach has only proved viable relatively recently, due to advances in computer technology. Biswal et al. (2010),
Mitchell (2009), and Yang (2009) provide recent reviews of data mining in different scientific fields. Data mining is
at the core of drug discovery, our understanding of the human genome, and our understanding from kindergarten
through adulthood was a major shift in focus. Many of the particular scientific thinking skills that have been
emphasized are skills covered in previous sections of this chapter, such as teaching deductive and inductive thinking
strategies. However, rather than focusing on one particular skill, such as induction, researchers in education have
focused on how the different components of scientific thinking are put together in science. Furthermore, science
educators have focused upon situations where science is conducted collaboratively, rather than being the product
of one person thinking alone. These changes in science education parallel changes in methodologies used to
investigate science, such as analyzing the ways that scientists think and reason in their laboratories.
By looking at science as a complex multilayered and group activity, many researchers in science education have
adopted a constructivist approach. This approach sees learning as an active rather than a passive process, and it
suggests that students learn through constructing their scientific knowledge. We will first describe a few examples
of the constructivist approach to science education. Following that, we will address several lines of work that challenge
some of the assumptions of the constructivist approach to science education.
Often the goal of constructivist science education is to produce conceptual change through guided instruction
where the teacher or professor acts as a guide to discovery, rather than the keeper of all the facts. One recent and
influential approach to science education is the inquiry-based learning approach. Inquiry-based learning focuses on
posing a problem or a puzzling event to students and asking them to propose a hypothesis that could explain the
event. Next, the student is asked to collect data that test the hypothesis, make conclusions, and then reflect upon
both the original problem and the thought processes that they used to solve the problem. Often students use
computers that aid in their construction of new knowledge. The computers allow students to learn many of the
different components of scientific thinking. For example, Reiser and his colleagues have developed a learning
environment for biology, where students are encouraged to develop hypotheses in groups, codify the hypotheses,
and search databases to test these hypotheses (Reiser et al., 2001).
One of the myths of science is the lone scientist suddenly shouting “Eureka, I have made a discovery!” Instead,
in vivo studies of scientists (e.g.,Dunbar, 1995, 2002), historical analyses of scientific discoveries (Nersessian, 1999),
and studies of children learning science at museums have all pointed to collaborative scientific discovery mechanisms
as being one of the driving forces of science (Atkins et al., 2009; Azmitia & Crowley, 2001). What happens during
collaborative scientific thinking is that there is usually a triggering event, such as an unexpected result or situation that
a student does not understand. This results in other members of the group adding new information to the person’s
representation of knowledge, often adding new inductions and deductions that both challenge and transform the
reasoner’s old representations of knowledge (Chi & Roscoe, 2002; Dunbar, 1998). Social mechanisms play a key
component in fostering changes in concepts that have been ignored in traditional cognitive research but are crucial for
both science and science education. In science education there has been a shift to collaborative learning, particularly at the
elementary level; however, in university education, the emphasis is still on the individual scientist. As many domains of
science now involve collaborations across scientific disciplines, we expect the explicit teaching of heuristics for
collaborative science to increase.
What is the best way to teach and learn science? Surprisingly, the answer to this question has been difficult to
uncover. For example, toward the end of the last century, influenced by several thinkers who advocated a
constructivist approach to learning, ranging from Piaget (Beilin, 1994) to Papert (1980), many schools answered this
question by adopting a philosophy dubbed “discovery learning.” Although a clear operational definition of this
approach has yet to be articulated, the general idea is that children are expected to learn science by reconstructing
the processes of scientific discovery—in a range of areas from computer programming to chemistry to mathematics.
The premise is that letting students discover principles on their own, set their own goals, and collaboratively
explore the natural world produces deeper knowledge that transfers widely.
The research literature on science education is far from consistent in its use of terminology. However, our reading
suggests that “discovery learning” differs from “inquiry-based learning” in that few, if any, guidelines are given to
students in discovery learning contexts, whereas in inquiry learning, students are given hypotheses and specific goals
 to achieve (see the second paragraph of this section for a definition of inquiry-based learning). Even though
thousands of schools have adopted discovery learning as an of the universe for a number of reasons. First, vast
databases concerning drug actions, biological processes, the genome, the proteome, and the universe itself now exist.
Second, the development of high throughput data-mining algorithms makes it possible to search for new drug
targets, novel biological mechanisms, and new astronomical phenomena in relatively short periods of time. Research
programs that took decades, such as the development of penicillin, can now be done in days (Yang, 2009).
Another recent shift in the use of computers in scientific discovery has been to have both computers and people
make discoveries together, rather than expecting that computers make an entire scientific discovery. Now instead of
using computers to mimic the entire scientific discovery process as used by humans, computers can use powerful
algorithms that search for patterns on large databases and provide the patterns to humans who can then use the
output of these computers to make discoveries, ranging from the human genome to the structure of the universe.
However, there are some robots such as ADAM, developed by King (2011), that can actually perform the entire
scientific process, from the generation of hypotheses, to the conduct of experiments and the interpretation of
results, with little human intervention. The ongoing development of scientific robots by some scientists (King et
al., 2009) thus continues the tradition started by Herbert Simon in the 1960s. However, many of the controversies
as to whether the robot is a “real scientist” or not continue to the present (Evans & Rzhetsky, 2010, Gianfelici,
2010; Haufe, Elliott, Burian, & O’Malley, 2010; O’ Malley, 2011).

Scientific Thinking and Science Education

Accounts of the nature of science and research on scientific thinking have had profound effects on science
education along many levels, particularly in recent years. Science education from the 1900s until the 1970s was
primarily concerned with teaching students both the content of science (such as Newton’s laws of motion) or the
methods that scientists need to use in their research (such as using experimental and control groups). Beginning in
the 1980s, a number of reports (e.g., American Association for the Advancement of Science, 1993; National
Commission on Excellence in Education, 1983; Rutherford & Ahlgren, 1991) stressed the need for teaching
scientific thinking skills rather than just methods and content. The addition of scientific thinking skills to the
science curriculum alternative to more didactic approaches to teaching and learning, the evidence showing that it is
more effective than traditional, direct, teacher-controlled instructional approaches is mixed, at best (Lorch et al.,
2010; Minner, Levy, & Century, 2010). In several cases where the distinctions between direct instruction and more
open-ended constructivist instruction have been clearly articulated, implemented, and assessed, direct instruction
has proven to be superior to the alternatives (Chen & Klahr, 1999; Toth, Klahr, & Chen, 2000). For example, in
a study of third- and fourth-grade children learning about experimental design, Klahr and Nigam (2004) found that
many more children learned from direct instruction than from discovery learning. Furthermore, they found that
among the few children who did manage to learn from a discovery method, there was no better performance on
a far transfer test of scientific reasoning than that observed for the many children who learned from direct
instruction.
The idea of children learning most of their science through a process of self-directed discovery has some
romantic appeal, and it may accurately describe the personal experience of a handful of world-class scientists.
However, the claim has generated some contentious disagreements (Kirschner, Sweller, & Clark, 2006; Klahr,
2010; Taber, 2009; Tobias & Duffy, 2009), and the jury remains out on the extent to which most children can learn
science that way

Conclusions and Future Directions

The field of scientific thinking is now a thriving area of research with strong underpinnings in cognitive
psychology and cognitive science. In recent years, a new professional society has been formed that aims to facilitate
this integrative and interdisciplinary approach to the psychology of science, with its own journal and regular
professional meetings.1
Clearly the relations between these different aspects of scientific thinking need to be combined in order to produce
a truly comprehensive picture of the scientific mind.
While much is known about certain aspects of scientific thinking, much more remains to be discovered. In particular,
there has been little contact between cognitive, neuroscience, social, personality, and motivational accounts of scientific
thinking. Research in thinking and reasoning has been expanded to use the methods and theories of cognitive
neuroscience (see Morrison & Knowlton, Chapter 6). A similar approach can be taken in exploring scientific thinking
(see Dunbar et al., 2007). There are two main reasons for taking a neuroscience approach to scientific thinking. First,
functional neuroimaging allows the researcher to look at the entire human brain, making it possible to see the many
different sites that are involved in scientific thinking and gain a more complete understanding of the entire range of
mechanisms involved in this type of thought. Second, these brain- imaging approaches allow researchers to address
fundamental questions in research on scientific thinking, such as the extent to which ordinary thinking in non- scientific
contexts and scientific thinking recruit similar versus disparate neural structures of the brain.
Dunbar (2009) has used some novel methods to explore Simon’s assertion, cited at the beginning of this chapter,
that scientific thinking uses the same cognitive mechanisms that all human beings possess (rather than being an entirely
different type of thinking) but combines them in ways that are specific to a particular aspect of science or a specific
discipline of science. For example, Fugelsang and Dunbar (2009) compared causal reasoning when two colliding
circular objects were labeled balls or labeled subatomic particles. They obtained different brain activation patterns
depending on whether the stimuli were labeled balls or subatomic particles. In another series of experiments, Dunbar
and colleagues used functional magnetic resonance imaging (fMRI) to study patterns of activation in the brains of
students who have and who have not undergone conceptual change in physics. For example, Fugelsang and Dunbar
(2005) and Dunbar et al. (2007) have found differences in the activation of specific brain sites (such as the anterior
cingulate) for students when they encounter evidence that is inconsistent with their current conceptual understandings.
These initial cognitive neuroscience investigations have the potential to reveal the ways that knowledge is organized in
the scientific brain and provide detailed accounts of the nature of the representation of scientific knowledge. Petitto
and Dunbar (2004) proposed the term “educational neuroscience” for the integration of research on education,
including science education, with research on neuroscience. However, see Fitzpatrick (in press) for a very different
perspective on whether neuroscience approaches are relevant to education. Clearly, research on the scientific brain is
just beginning. We as scientists are beginning to get a reasonable grasp of the inner workings of the subcomponents of
the scientific mind (i.e., problem solving, analogy, induction). However, great advances remain to be made
concerning how these processes interact so that scientific discoveries can be made. Future research will focus on both
the collaborative aspects of scientific thinking and the neural underpinnings of the scientific mind