Forming Spatial Groups
Over the last few years, I've turned my attention to the problem of how children and adults form spatial groups. Obviously, the ability to organize locations into groupings plays an important role in spatial memory. One consequence of grouping locations, however, is that estimates of location tend to be distorted. In particular, children and adults often underestimate distances between locations in the same region and overestimate distances between locations in different regions. The central issue these findings raise is what processes underlie these distortions? Recently, Huttenlocher, Hedges, and Duncan (1991) outlined a theory of spatial coding to account for these types of distortions in estimates of location. They propose that retrieval of locations from memory involves the use of both metric and categorical (i.e., spatial region) information. When trying to reproduce the location of a previously seen object, people initially rely on their memory of fine-grained, metric information and then make adjustments based on region membership. These adjustments distort memory for locations within a region in the direction of the most prototypical location (i.e., the region center). According to this theory, the pull toward the region center results in locations within the same region being remembered as closer together than they actually are and locations from different regions being remembered as farther apart than they actually are. Thus far, however, this theory has only been tested in the context of how children and adults remember single locations. Little is known about how the presence of other objects influences estimates of location. For example, objects within a region may be remembered as closer together than they really are not because individual objects gravitate toward the center of the region, but because individual objects gravitate toward the center of the set of objects.
|
 |
|
dots on a blue floor in an open square box. During the test phase, the floor with the yellow dots marking the locations is removed and replaced with a plain blue floor. Participants then attempt to place all of the objects in their correct locations. After participants place the objects, we measure the x and y coordinates of each object. Thus, we can determine whether children and adults place objects closer to the center of the regions than they actually are, and whether they underestimate distances among objects within the same region and overestimate distances between objects in different regions. |
||
What Role Do Spatial Prototypes Play in Memory for Multiple Locations?
In our first two experiments, Alycia Hund and I tested whether 7-, 9-, 11-year-olds and adults exhibit a bias toward the center of spatial regions in a task involving memory for multiple locations (Plumert & Hund, 2001). We found that although all age groups overestimated distances across regions, only 11-year-olds and adults in the most salient boundary condition displaced objects toward the center of spatial regions. One key issue these results raise is how to explain subdivision effects in the absence of prototype effects. That is, if children and adults did not displace objects toward the region centers, what then accounts for overestimation of distance across regions? It appears that participants displaced the objects toward the corners of the house rather than the centers of the regions. Displacing the objects toward the corners of the house necessarily resulted in greater distance between objects in different regions. Hence, overestimation of distance across regions was largely the result of bias toward the corners of the house rather than bias toward the centers of the regions. Clearly, these findings show that the idea of a prototype at the center of a geometrically-defined region does not account for all subdivision effects (i.e., overestimation of distance across regions).
What Role Does Spatial Temporal Contiguity Play in Forming Spatial Groups?
Alycia Hund and I recently conducted two more experiments to look more closely at the processes underlying the formation of spatial groups (Hund, Plumert, & Benney, under review). One factor that might play an important role in forming spatial groups is specific experience with locations. In particular, experiencing several locations together in time may highlight the associations among the locations and create a salient spatial grouping. For example, suppose a shopper spends Saturday morning shopping at several downtown businesses and stops for lunch at a nearby restaurant. This temporal experience (and similar experiences on other days) may lead the shopper to create a spatial grouping that includes downtown businesses and restaurants. Thus, everyday experience may influence spatial organization by highlighting associations among locations.
We investigated the role of temporal contiguity in forming spatial groups by examining whether experiencing nearby locations together in time during learning leads children and adults to underestimate distances between locations in the same region. In these two experiments, 7-, 9-, 11-year-olds and adults learned the locations of the objects region by region. (This contrasts with our previous experiments in which they learned the locations in a random order). After learning, participants attempted to place the objects in their correct locations without the aid of the dots marking the locations. Unlike our previous experiments, we found that all age groups underestimated distances among locations in the same region. This suggests that temporal contiguity may play a particularly important role in younger children's ability to form spatial groups.
Communicating about Nested Spatial Relations
Another aspect of children's spatial organizational skills that undergoes developmental change is their coding of nested spatial relations. Locations have a nested structure based on the spatial relations that hold between progressively larger landmarks and spatial regions. Thus, we might think of a coffee cup as near the toaster on the counter in the kitchen. Coding objects in relation to a series of nested landmarks and regions is essential for remembering and communicating about object locations. Imagine, for example, that you put an important document in a folder in a file cabinet in the departmental office. Later, one of your colleagues asks you for the document. If there are many file folders in the office, saying only that "it's in a folder" will not be sufficient for the other person to find it. Instead, you have to communicate about the folder in relation to one or more disambiguating landmarks (e.g., "the file folder is in the bottom drawer of the file cabinet next to the secretary's desk in the office"). As this example illustrates, remembering and communicating about object locations often requires coding hierarchical spatial relations among several landmarks and regions. In a series of papers, I have documented how children and adults represent hierarchical spatial relations in communication (Craton, Elicker, Plumert, & Pick, 1990; Plumert, 1996; Plumert, Carswell, DeVet, & Ihrig, 1995; Plumert, Ewert, & Spear, 1995; Plumert & Nichols-Whitehead, 1996; Plumert, Spalding, & Nichols-Whitehead, 2001; Plumert & Hawkins, 2001).
Throughout all of this work, I've been particularly interested in understanding the factors that influence the content and organization of spatial descriptions. One such factor that plays an important role in young children's ability to describe a location in relation to a small and a large landmark is the nature of the spatial relation between the two landmarks (Plumert et al., 1995; Plumert et al., 1996). In particular, both 3- and 4-year-olds are more likely to refer to the large landmark when the small landmark is on the large landmark (e.g., "The mouse is in the shoe on the bed") than when it is next to the large landmark (e.g., "The mouse is in the shoe next to the bed"). Interestingly, this phenomenon is not restricted to young children's descriptions of location. When adults are free to choose which pieces of spatial information to include in their descriptions, they are much more likely to include a reference to a small landmark when a target object is on rather than next to a small landmark (Plumert et al., 1996). The fact that both adults and young children show a preference for support relations over proximity relations suggests that the nature of the spatial relation exerts an important influence over the selection of spatial information in descriptions of location.
One question these findings raise is why does this preference for support over proximity relations exist? One possibility is that support relations are very salient because they have important functional consequences for how objects interact with one other. In particular, objects fall when surfaces of support are removed. The idea that functionality plays an important role in how young children communicate about location suggests that other functional spatial relations beside support should also have an advantage over proximity. One other spatial relation that has important consequences for how objects interact is containment. Like support, containment has implications for how objects move in the environment. For example, when toys are placed inside a box, the toys move when the box moves.
Aimee Hawkins and I recently examined this hypothesis by contrasting young children's ability to produce and comprehend spatial descriptions involving either a containment or a proximity relation between a large and a small landmark (Plumert & Hawkins, 2001) In this investigation, children either gave or followed directions for finding a miniature mouse hidden in a small dollhouse. As in my previous work, there were several pairs of identical small landmarks that served as hiding locations (e.g., bags, boxes, plants). The target member of the pair was either in or next to a large furniture landmark (e.g., under the towel in the playpen versus under the towel next to the playpen.). The results of these experiments clearly show that young children find it easier to communicate about containment relations than about proximity relations. When describing the location of a hidden object, 3- and 4-year-olds weremore likely to disambiguate the target small landmark by including a reference to the large landmark when it contained the small landmark than when it was nearby the small landmark. Likewise, 3-year-olds searched more quickly for a hidden object when the directions included a reference to a small landmark that was in rather than next to a large landmark. Together, these findings underscore the idea that both support and containment are more salient than proximity to young children.
Accidental injuries are the leading cause of death in children under age 18. Despite growing national concern over promoting children's safety, however, very little is known about why children have so many accidents. Increasingly, overviews of strategies for reducing unintentional childhood injuries have called for a better understanding of the developmental factors that contribute to the occurrence of these injuries. Over the last several years, I have been developing a program of research aimed at understanding the causes of unintentional childhood injuries. The primary question underlying this work is how do immature cognitive skills put children at risk for injury? In pursuing this question, I have attempted to integrate the study of basic and applied issues in a single program of research (for a more extended discussion of this approach, see Schwebel, Plumert, & Pick, 2000). The idea is to understand both how particular cognitive skills develop and how these skills (or lack thereof) are related to injury risk (Plumert, 1995; Plumert & Schwebel, 1997; Schwebel & Plumert, 1999).
What Role Does Ability Overestimation Play in Childhood Injuries?
According to Gibson (1979), adaptive behavior within the environment depends upon perceiving the fit between one's own physical characteristics and the properties of the environment in which actions take place. Errors in judging the relation between one's physical abilities and the demands of the situation may be one important factor contributing to accident risk. For example, although some pedestrian accidents may result when children fail to follow simple rules like looking both ways before crossing a street, others may result when children overestimate their ability to walk or run through traffic gaps.
How good are children at judging the fit between their physical capabilities and the demands of the situation? Like other researchers, we've found that 6- and 8-year-olds overestimate their physical abilities. In our studies, we ask children to judge whether they can perform particular physical activities such as reaching for an object on a shelf. Children make judgments both when the object is within reach and when the object is beyond reach. We've found that children in this age range often judge that they can perform actions that are beyond their capabilities. Moreover, we have found that 6-year-olds who overestimate their physical abilities tend to experience more unintentional injuries (Plumert, 1995; Plumert & Schwebel, 1997). Thus, it appears that overestimation of ability may be a risk factor for unintentional injuries.
Links between Temperamental Characteristics and Ability Overestimation
Another factor frequently implicated as a major contributor to childhood injury risk is temperament. Over the past several years, David Schwebel and I have examined the relations between temperament and injuries. In one study, we found that 8-year-olds whose parents described them as highly active, impulsive, and undercontrolled had more severe day-to-day injuries (Plumert & Schwebel, 1997). We have also found that children who were impulsive and undercontrolled at age 4 had a history of injuries requiring medical attention at age 6 (Schwebel & Plumert, 1999). Thus, early manifestations of impulsivity, noncompliance, and high activity level are predictive of later injury risk.
Thus, both temperament and ability overestimation appear to contribute to injury risk. However, one question these studies raise is what are the links between temperamental characteristics, ability overestimation, and injury risk? One possibility is that temperamental characteristics and ability overestimation make separate contributions to injury risk. A more interesting possibility, however, is that ability overestimation mediates the relation between temperamental characteristics and injury risk. That is, temperamental characteristics such as high impulsivity and low inhibitory control put children at risk for injuries because these characteristics lead to errors in judgment. When judging whether it is safe to cross a street, for example, children need to stop and think before going across. Children who make decisions impulsively may be more prone to errors in judgment than children who make decisions more carefully. Indeed, we have found that children who are impulsive, active, and approach-oriented are more likely to overestimate their physical abilities than are children who are less impulsive, active, and approach-oriented (Plumert & Schwebel, 1997; Schwebel & Plumert, 1999). This suggests that temperament and ability overestimation may work together to put certain children at greater risk for unintentional injuries.
The Bicycling Simulator Project
More recently, I have begun to study how ability overestimation may put children at risk for injuries in the context of children's bicycling safety. Although bicycle crashes are among the most common causes of severe injuries in childhood, the underlying causes of bicycle crashes remain poorly understood. One reason why we know so little about the behaviors that put children at risk for bicycling injuries is that it is difficult to study bicycling behavior without putting research participants at risk for injury. For example, one cannot study whether impulsive children are more likely to make errors in judging the size of traffic gaps by asking them to bicycle across busy roads in the real environment. Another reason we know so little about the causes of bicycling crashes is that it is difficult to study bicycling behavior in a controlled environment. Without control over the timing and nature of events in an experiment, it is difficult to draw valid conclusions about how these events influence children's bicycling behavior. Advances in virtual environment technology, however, offer a way of studying the problem of bicycling safety in a controlled manner without putting children at risk for injury. Using a high-fidelity, interactive bicycling simulator, we can safely present children with the same kinds of bicycling challenges as they confront in the real environment. We can then use their performance in the simulated environment to make inferences about their performance in the real environment.
I and two faculty members in the Computer Science Department, Joe Kearney and Jim Cremer, have recently launched an interdisciplinary project aimed at developing a bicycling simulator to study the factors that put children at risk for car-bicycle collisions. Like a driving or flight simulator, research participants ride a stationary bicycle through a computer-generated, virtual environment. The bicycling simulator is completely interactive, meaning that changes in what research participants do on the bicycle directly correspond to changes in what they see in the virtual environment. For example, a slowing down of pedaling speed on the bicycle is directly reflected in a slowing down of apparent motion in the virtual environment. Our current configuration of the bicycling simulator is also almost completely immersive, meaning that the display system provides research participants with a panoramic view of the virtual environment. To learn more about the simulator itself, click here.
Developing a Simulated Road-Crossing Task for Children
The aim of our initial pilot study funded by the University of Iowa Injury Prevention Research Center was to provide preliminary information about the feasibility of using a bicycling simulator to study children's bicycling behavior. We were particularly interested in the size of traffic gaps children accept when bicycling across busy roads. Eight-, 10-, 12-, and 14-year-old children rode a bicycle mounted on a stationary trainer through a simulated environment consisting of a straight, residential street with 12 intersections. Their task was to cross all 12 intersections without colliding with a car. Children faced cross traffic from their left-hand side and waited for gaps that they judged were adequate to allow safe crossing. The cross traffic traveled at continuous rates of either 25 MPH or 35 MPH. We found that older children left a large gap between themselves and the approaching cars than did the younger children. Likewise, children were more cautious when the cross traffic was moving faster. Specifically, children who faced cross traffic moving at 35 MPH allowed more cars to pass before crossing the road than did their counterparts in the 25 MPH group. These results suggest that using a bicycling simulator to study children's bicycling behavior is a promising approach for understanding car-bicycle collisions.
Validating Distance Estimates in the Simulated Environment
One issue that arises when using behavior in simulated environments to make inferences about behavior in real environments is ecological validity. That is, are people's responses in virtual environments representative of their responses in real environments? Although virtual environments are an exciting new medium for investigating children's bicycling behavior under safe and controlled conditions, the results of such experiments are of questionable value if virtual environments lack ecological validity. The most elegant approach to validation is to present the same subjects with exactly the same situations in both the simulated and the real environment. Using the same subjects in both the simulated and real environments makes it possible to factor out extraneous within-subject variation in behavior. Using exactly the same situations makes a direct comparison between simulated and real environments possible. Obviously, it is often very difficult to present subjects with the same situation in the real environment. For this reason, validation experiments work best when they test components of subjects' behavior. For example, an important component of making accurate judgments about whether a traffic gap affords crossing is the ability to make time-to-contact judgments about oncoming cars. That is, if a car is 300 ft away and traveling at 25 MPH, when will it pass in front of the bicyclist? Comparisons of judgments such as these in simulated and real environments can provide valuable information about the ecological validity of the simulated environment.
We have begun to address the validation issue by comparing people's distance judgments in real and simulated environments. In our study, 24 adults made distance judgments in the real environment and in the simulated environment. Our task involved estimating distance in terms of the time it would take to walk to a marked location. For each location, subjects pushed the start button on a stopwatch when they imagined themselves starting to walk toward the location, and pushed the stop button when they imagined themselves reaching the location. We then recorded the amount of time subjects estimated it would take to walk to the location. Across both the simulated and the real environment, subjects made distance judgments for six randomly ordered distances ranging between 20 and 120 ft. Half of the subjects made judgments in the real environment first, and half made judgments in the simulated environment first. The real environment consisted of a large, open grassy area in front of a university building. The simulated environment consisted of the same scene back-projected onto three 8 x 10 ft. screens (1280 x 1024 pixels on each screen) using high-resolution, textured graphics. The side screens were placed at right angles to the center screen, forming a three-walled, square room. Subjects stood in the middle of the room, providing them with an immersive view of the simulated environment.
The major question was did subjects' distance judgments differ in the real and simulated environments? As shown in the figure below, subjects' estimates of how long it would take to walk each distance were virtually identical in the two environments. Moreover, subjects significantly undershot actual times in both the real and the simulated environment. (We used a measure of each individual subject's walking speed to calculate estimates of how long it actually should have taken each subject to walk each distance). These results suggest that people's ability to make judgments about distance is similar in real and simulated environments.