*3.1.2. Strategy development and use*

70 Current Topics in Children's Learning and Cognition

role for a teacher to scaffold children's scientific reasoning.

about science.

Specifically, this debate is about the ideal role for adults in helping children to encode information. In direct instruction, there is a clear role for a teacher, often actively pointing out effective examples as compared to ineffective ones, or directly teaching a strategy to apply to new examples. And, indeed, there is evidence that more direct guidance to test variables systematically can help students in learning, particularly in the ability to apply their knowledge to new contexts (e.g., Klahr & Nigam, 2004; Lorch et al., 2010; Strand-Cary & Klahr, 2008). There is also evidence that scaffolded discovery learning can be effective (e.g., Alfieri, Brooks, Adrich, & Tenenbaum, 2011). Those who argue for discovery learning often do so because they note that pedagogical approaches commonly labeled as "discovery learning," such as problem-based learning and inquiry learning, are in fact highly scaffolded, providing students with a structure in which to explore (Alfieri et al., 2011; Hmelo-Silver et al., 2007; Schmidt et al., 2007). Even in microgenetic studies in which children are described as engaged in "self-directed learning," researchers ask participants questions along that way that serve as prompts, hints, dialogue, and scaffolds that facilitate learning (Klahr & Carver, 1995). What there appears to be little evidence for is "pure discovery learning" in which students are given little or no guidance and expected to discover rules of problem solving or other skills on their own (Alfieri et al., 2011; Mayer, 2004). Thus, it is clear that formal education includes a critical

A common goal in science education is to correct the many misconceptions students bring to the classroom. Chinn and Malhotra (2002) examined the role of encoding evidence, interpreting evidence, generalization, and retention as possible impediments to correcting misconceptions. Over four experiments, they concluded that the key difficulty faced by children is in making accurate observations or properly encoding evidence that does not match prior beliefs. However, interventions involving an explanation of what scientists expected to happen (and why) were very effective in mediating conceptual change when encountering counterintuitive evidence. That is, with scaffolds, children made observations independent of theory, and changed their beliefs based on observed evidence. For example, the initial belief that a thermometer placed inside a sweater would display a higher temperature than a thermometer outside a sweater was revised after seeing evidence that disconfirmed this belief and hearing a scientist's explanation that the temperature would be the same unless there was something warm inside the sweater. Instructional supports can play a crucial role in improving the encoding and observational skills required for reasoning

In laboratory studies of reasoning, there is direct evidence of the role of adult scaffolding. Butler and Markman (2012a) demonstrate that in complex tasks in which children need to find and use evidence, causal verbal framing (i.e., asking whether one event caused another) led young children to more effectively extract patterns from scenes they observed, which in turn led to more effective reasoning. In further work demonstrating the value of adult scaffolding in children's encoding, Butler and Markman (2012b) found that by age 4, children are much more likely to explore and make inductive inferences when adults

intentionally try to teach something than when they are shown an "accidental" effect.

As discussed earlier in this chapter, learning which strategies are available and useful is a fundamental part of developing scientific thinking skills. Much research has looked at the role of adults in teaching strategies to children in both formal (i.e., school) and informal settings (e.g., museums, home; Fender & Crowley, 2007; Tenenbaum, Rappolt-Schlichtmann, & Zanger, 2004).

A central task in scientific reasoning involves the ability to design controlled experiments. Chen and Klahr (1999) found that directly instructing 7- to 10-year-old children in the strategies for designing unconfounded experiments led to learning in a short time frame. More impressively, the effectiveness of the training was shown seven months later, when older students given the strategy training were much better at correctly distinguishing confounded and unconfounded designs than those not explicitly trained in the strategy. In another study exploring the role of scaffolded strategy instruction, Kuhn and Dean (2005) worked with sixth graders on a task to evaluate the contribution of different factors to earthquake risk. All students given the suggestion to focus attention on just one variable were able to design unconfounded experiments, compared to only 11% in the control group given their typical science instruction. This ability to design unconfounded experiments increased the number of valid inferences in the intervention group, both immediately and three months later. Extended engagement alone resulted in minimal progress, confirming that even minor prompts and suggestions represent potentially powerful scaffolds. In yet another example, when taught to control variables either with or without metacognitive supports, 11-year-old children learned more when guided in thinking about how to approach each problem and evaluate the outcome (Dejonckheere, Van de Keere, & Tallir, 2011). Slightly younger children did not benefit from the same manipulation, but 4- to 6 year-olds given an adapted version of the metacognitive instruction were able to reason more effectively about simpler physical science tasks than those who had no metacognitive supports (Dejonckheere, Van de Keere, & Mestdagh, 2010).

#### **3.2. Cultural tools that support scientific reasoning**

Clearly, even with the number of studies that have focused on individual cognition, a picture is beginning to emerge to illustrate the importance of social and cultural factors in the development of scientific reasoning. Many of the studies we describe highlight that even "controlled laboratory studies" are actually scientific reasoning in context. To illustrate, early work by Siegler and Liebert (1975) includes both an instructional context (a control condition plus two types of instruction: *conceptual framework*, and *conceptual framework plus analogs*) and the role of cultural supports. In addition to traditional instruction about variables (factors, levels, tree diagrams), one type of instruction included practice with analogous problems. Moreover, 10- and 13-year-olds were provided with paper and pencil to keep track of their results. A key finding was that *record keeping* was an important mediating factor in success. Children who had the metacognitive awareness of memory limitations and therefore used the provided paper for record keeping were more successful

at producing all possible combinations necessary to manipulate and isolate variables to test hypotheses.

Emergence of Scientific Reasoning 73

Various elements of *number and number systems* are extremely important in science. Sophisticated scientific reasoning requires an understanding of data and the evaluation of numerical data. Early work on evidence evaluation (e.g., Shaklee, Holt, Elek, & Hall, 1988) included 2 x 2 contingency tables to examine the types of strategies children and adults used (e.g., comparing numbers in particular cells, the "sums of diagonals" strategy). Masnick and Morris (2008) used data tables to present evidence to be evaluated, and varied features of the presentation (e.g., sample size, variability of data). When asked to make decisions without the use of statistical tools, even third- and sixth-graders had rudimentary skills in detecting trends, overlapping data points, and the magnitude of differences. By sixth grade, participants had developing ideas about the importance of variability and the presence of

Although language, symbols, and number systems are used as canonical examples of cultural tools and resources within the socio-cultural tradition (Lemke, 2001), recent advances in *computing and computer simulation* are having a huge impact on the development and teaching of scientific reasoning. Although many studies have incorporated the use of physical systems (Figure 1, panel B) such as the canal task (Gleason & Schauble, 2000), the ramps task (e.g., Masnick & Klahr, 2003), mixing chemicals (Kuhn & Ho, 1980), and globes (Vosniadou, Skopeliti, & Ikospentaki, 2005), there is an increase in the use of interactive computer simulations (see Figure 1, panel C). Simulations have been developed for electric circuits (Schauble, Glaser, Raghavan, & Reiner, 1992), genetics (Echevarria, 2003), earthquakes (Azmitia & Crowley, 2001), flooding risk (Keselman, 2003), human memory (Schunn & Anderson, 1999), and visual search (Métrailler, Reijnen, Kneser, & Opwis, 2008). Non-traditional science domains have also been used to develop inquiry skills. Examples include factors that affect TV enjoyment (Kuhn et al., 1995), CD catalog sales (Dean & Kuhn, 2007), athletic performance (Lazonder, Wilhelm, & Van Lieburg, 2009), and shoe store sales

Computer simulations allow visualization of phenomena that are not directly observable in the classroom (e.g., atomic structure, planetary motion). Other advantages include that they are less prone to measurement error in apparatus set up, and that they can be programmed to record all actions taken (and their latencies). Moreover, many systems include a scaffolded method for participants to keep and consult records and notes. Importantly, there is evidence that simulated environments provide the same advantages as isomorphic

New lines of research are taking advantage of advances in computing and intelligent computer systems. Kuhn (2011b) recently examined how to facilitate reasoning about multivariable causality, and the problems associated with the visualization of outcomes resulting from multiple causes (e.g., the causes for different cancer rates by geographical area). Participants had access to software that produces a visual display of data points that represent main effects and their interactions. Similarly, Klahr and colleagues (Siler, Mowery, Magaro, Willows, & Klahr, 2010) have developed an intelligent tutor to teach experimentation

"hands on" apparatus (Klahr, Triona, & Williams, 2007; Triona & Klahr, 2007).

outliers for drawing conclusions from numerical data.

(Lazonder, Hagemans, & de Jong, 2010).
