A Psychometric Approach 1 Running head: R322-B A PSYCHOMETRIC APPROACH TO INTUITIVE PHYSICS A Psychometric Approach to Intuitive Physics

The literature on intuitive physics shows that many people exhibit systematic errors when predicting the behavior of simple physical events. Most previous research has attributed these errors to factors specific to a certain class of tasks. The present study investigated the possibility that intuitive physics performance may be related to general measures of cognitive ability. 204 adults (ages 20 to 91) were presented with 5 pairs of intuitive physics questions. It was found that performance on the intuitive physics items was moderately intercorrelated, suggesting that they are tapping into a unitary construct. Despite their correlation with factors which declined with advancing age, intuitive physics performance was not correlated with age (r = 0.00). The findings are discussed in the context of research on intuitive physics, as well as research in cognitive aging. A Psychometric Approach 3 Research into intuitive physics demonstrates that a large percentage of the adult population holds strikingly erroneous views about basic physical principles that govern the motions of objects in the world; a world in which people act and behave quite successfully. For example, when asked to draw the water level on a picture of a tilted glass, about 40% of young adults draw lines that are not horizontal (McAfee & Proffitt, 1991). When asked to predict the path that a pendulum bob takes when the string is cut at various points, a large portion of participants make systematically incorrect judgments (Caramazza, McCloskey, & Green, 1981). The prevalence of these systematic errors among otherwise sophisticated adults led some investigators to propose that incorrect performance on these tasks was due to specific “naïve” beliefs rather than one’s general ability to reason about mechanical systems (McCloskey and Kohl, 1983). In the present study, we investigated the possibility that performance on intuitive physics tasks also reflects individual differences in cognitive ability. Research seeking to explain performance on intuitive physics tasks can be grouped into three categories, based on the type of hypothesized explanation. First, some have posited explanations for performance that are specific to a certain task or class of tasks. Second, others have focused on factors that are related to general problem complexity or difficulty. A third approach asks whether intuitive physics performance can be explained in part by individual differences in cognitive abilities. While these three approaches are not mutually exclusive, most previous research has focused on taskspecific and problem complexity explanations. The current study focused on the third approach by relating intuitive physics performance to other cognitive abilities and age. A Psychometric Approach 4 By attempting to relate intuitive physics performance to other cognitive abilities across the adult lifespan, this approach offers the potential to both supplement and constrain the conclusions of previous research. Our first step was to determine whether a significant amount of variation in performance on intuitive physics tasks could be explained by a single individual difference construct. Is performance on a representative sample of intuitive physics tasks intercorrelated, thereby suggesting that these tasks tap a common disposition? If so, this would suggest that while the specific form that errors take on any given intuitive physics task may seem unrelated, the aggregate level of performance on several tasks can be partly explained by individual differences common to these tasks. Answering these questions serves to supplement previous explanations of intuitive physics performance by verifying that they are indeed a coherent group of tasks. Such results would also constrain previous conclusions by reminding researchers that despite the task-specific nature of systematic errors, individual differences in cognitive ability do account for some of the variation in performance across all of these tasks. Next, we looked at the relationship between intuitive physics performance and other cognitive abilities across the adult lifespan. In other words, if performance on these intuitive physics tasks is indeed intercorrelated, how does this intuitive physics performance relate to other measures of cognitive ability? These results will help place intuitive physics in the broader contexts of cognitive abilities and how cognitive abilities may change across the lifespan. A Psychometric Approach 5 Task Specific Approaches Initial approaches to explaining adult performance on intuitive physics tasks focused on the nature of the errors that people make. Guided by pioneering developmental research (e.g. Piaget & Inhelder, 1963), early investigators of intuitive physics tasks in adults saw the systematicity of errors as symptomatic of incorrect intuitive theories or beliefs held by those participants who exhibited the errors (Kaiser, McCloskey, & Proffitt, 1986; McCloskey & Kohl, 1983). This approach to intuitive physics invoked factors specific to a particular task to explain both the fact that errors existed as well as the particular form that the errors took. To illustrate, Shanon (1976) suggested that errors on questions regarding falling objects could be attributed to the fact that some people hold an Aristotelian theory (that objects fall with constant velocity) rather than the correct Newtonian theory (that objects accelerate as they fall). McCloskey and Kohl (1983) asked participants to predict the behavior of a ball exiting a c-shaped tube lying flat on a table. Forty percent of their participants incorrectly reported that the ball would continue on a curved path after it exited the tube. This incorrect prediction implies that a putative curvilinear force had been applied to the ball and that it would gradually dissipate after the ball left the tube. McCloskey suggested that these participants held an intuitive theory similar to medieval impetus theory, which did not discriminate between straight and curvilinear forces and asserted that forces are properties of objects themselves, not the interactions between objects. These “intuitive theory” approaches offered quite compelling explanations for the systematic nature of errors on several intuitive physics tasks. However, further A Psychometric Approach 6 examination revealed that there were limits to this approach. First, each intuitive theory explained performance on a limited class of tasks (e.g. impetus theory applies to the cshaped tube task, but not to the water level task). By invoking factors specific both to the task and to the people who failed the task, these approaches were quite limited in their scope. Second, if holding an incorrect theory causes errors, adherence to these theories should be a stable and consistent characteristic of an individual. Cooke and Breedin (1994) found evidence that suggested that individuals create implicit theories “on the fly”, in order to generate responses to intuitive physics questions. Both the form of the question and the form of the response (multiple choice or freehand drawing) affected the number and type of errors on intuitive physics tasks. This inconsistency suggests that performance on these tasks is neither a simple measure of knowledge of physics principles, nor a reflection of whether one holds an incorrect intuitive theory. For example, whereas participants may respond that a ball exiting a c-shaped tube follows a curved path, they are far less likely to respond that water exiting a curved hose (laying flat on the ground) will continue to curve (Kaiser, Jonides, & Alexander, 1986) even though the principle assessed is the same. More recent experiments in intuitive physics have attempted to address the shortcomings of the intuitive theories approach, while continuing to focus on why individuals fail particular tasks as well as the manner in which they fail. Some have suggested that rather than applying comprehensive (albeit incorrect) theories, erroneous participants are misapplying simple conceptual heuristics. For example, in an investigation of the development of beliefs about falling objects, Kaiser and colleagues found that the youngest group of participants (preschool and kindergarten) seemed to be A Psychometric Approach 7 using a situation-specific rule (“all objects fall straight down when dropped”) before they developed a more comprehensive account of the behavior of falling bodies (Kaiser, Proffitt, & McCloskey, 1985). Kozhevnikov and Hegarty (2001) combined an intuitive theory approach with heuristics and argued that participants use impetus beliefs as a default heuristic when other information is unavailable or insufficient. These approaches are similar to an intuitive theory approach in that they postulate differences in belief as a root cause of differences in performance on intuitive physics tasks; however, the explanations they provide are entirely specific to the tasks that they address. A similar approach is seen in an account of performance on the water level task that invokes perceptual biases to explain incorrect performance (McAfee & Proffitt, 1991). The Problem Complexity Approach Despite the limited success of these various task-specific approaches, they leave many aspects of the phenomena unexplained. By asking why certain people got specific problems wrong, and seeking to explain the form of these errors, these approaches do not provide a general account of what sorts of problems tend to induce errors and what sorts do not. Furthermore, correct performance is virtually ignored by these accounts. The problem complexity account described by Proffitt and Gilden (1989) complements these task-specific approaches by concentrating on general characteristics that make different problems more or less difficult. This account groups problems into two categories according to the number of physical dimensions of motion that must be combined to solve a problem. Problems involving particle motions require that participants pay attention to only one dimension of motion. Complex problems are defined as those that require the problem solver to combine multiple dimensions of motion in order to arrive at A Psychometric Approach 8 the correct solution. Proffitt and colleagues applied this approach to several cases of dynamical motion. Although this approach does not generalize to all of the intuitive physics tasks, its scope is larger than task-specific approaches. On the other hand, because this account essentially splits problems into easy and hard problems, it lacks the precision that characterizes the task-specific approaches discussed above. The problemcomplexity account provides a rough prediction of the proportion of errors expected given a certain problem, but does not explain the particular form that the errors take, nor does it account for which participants answer the problem correctly. The Psychometric Approach In addition to task-specific factors, performance on intuitive physics tasks may reflect individual differences in cognitive abilities. Psychologists working with other types of cognitive tasks have sought to supplement empirical investigations by administering many tasks to large samples of participants, and analyzing the data with correlational procedures often used by psychometricians. For the purposes of this paper, we will refer to this sort of investigation (large sample, many tasks and correlational analyses) as a psychometric approach. For example, some researchers in the field of cognitive aging often adopt this psychometric approach when looking at how cognitive abilities change over time. Our goal in using this approach is to identify the presence of common variance shared among performance measures of intuitive physics and measures of general cognitive abilities. In doing so, we hope to reach some tentative conclusions on the unity of intuitive physics as a set of tasks, and their relation (as a group) to individual differences such as age and cognitive ability. A Psychometric Approach 9 Research Goals Previous accounts of intuitive physics performance indicate that errors may be attributable to different causes, either dependent on relevant theories, beliefs, or heuristics, or dependent on the complexity of the task. Although many of these accounts are quite compelling, a brief survey of the field shows that there are nearly as many explanations as there are tasks. This plethora of explanations challenges the basic assumption that all of these tasks belong to the same grouping, as implied by the common label “intuitive physics”. Our study sought to assess whether intuitive physics tasks were, assessments of a common underlying ability. In other words, is there an individual difference construct that accounts for a large portion of the variation in the performance across all of these tasks? Additional questions follow from the first. If intuitive physics performance does reflect a unitary disposition or ability, how does this ability relate to other cognitive abilities? Finally, how does intuitive physics performance relate to age and those abilities that have different age trajectories?