499021 2013
HFSXXX10.1177/0018720813499021Month XXXX - Human FactorsTransfer of Distance Estimation Training
The Effects of Prism Adaptation on Egocentric Metric Distance Estimation Allyson R. Colombo, BCI, Inc., Dahlgren, Virginia, and Keith S. Jones, Texas Tech University, Lubbock, Texas, USA Objective: The present experiment evaluated whether training involving throwing transferred to metric distance estimation (i.e., describing in feet and inches the distance between oneself and targets). Background: In prior work, we found that metric estimation training negatively transferred to throwing. We explained our results in terms of cognitive intrusion. The present study tested that possibility by swapping our training and transfer tasks. Method: During pretesting, participants verbally estimated the metric distances between themselves and targets, or they threw a beanbag to targets. During training, participants donned goggles that distorted their vision. While wearing the goggles, they threw a beanbag to targets. Half received feedback. During posttesting, participants removed the distorting goggles and completed the same task that they performed during pretesting. Results: The results indicated that the distorting goggles degraded throwing at the beginning of training, visual feedback improved throwing during training, the effects of training with feedback persisted into the throwing posttest, and the effects of training with feedback did not transfer to the verbal metric estimation posttest. Conclusion: Training involving throwing was effective, but did not transfer to verbal metric distance estimation. This supports our argument that the negative transfer observed in our previous study stemmed from cognitive intrusion. Application: The present experiment suggests that the creation of distance estimation training should begin with a careful analysis of the transfer task, and that distance estimation training programs should explicitly teach trainees that their training will not generalize to all distance estimation tasks. Keywords: transfer, transfer of training, visual distortion, distorting goggles, throwing, verbal estimation, feedback, prism adaptation
Address correspondence to Keith S. Jones, Department of Psychology, Texas Tech University, Lubbock, TX 794092051, USA; e-mail: [email protected]. HUMAN FACTORS Vol. 56, No. 3, May 2014, pp. 605–615 DOI: 10.1177/0018720813499021 Copyright © 2013, Human Factors and Ergonomics Society.
People estimate distances between themselves and objects in their environments. This ability has been conceptualized in different ways (Foley, 1980; Gibson, 1950, 1979/1986; Helmholtz, 1910/1962; Hochberg, 1978; Milner & Goodale, 1995). For example, Helmholtz (1910/1962) conceptualized distance estimation as the elaboration of inherently ambiguous distance cues via unconscious inference. In contrast, Gibson (1950) conceptualized distance estimation as attention to environmental features that specify distance. Despite their differences, however, the various conceptualizations agree that distance estimation can be inaccurate. That expectation has empirical support. People have misestimated distances while flying aircraft (Gibson, 1947), firing weapons (Rogers, Sprol, Vitereles, Voss, & Wickens, 1945, as cited in Gibson & Bergman, 1954), working underwater (Ferris, 1972, 1973a, 1973b), wearing night-vision goggles (Dyer & Young, 1998), driving cars (Taieb-Maimon, 2007), and exploring virtual environments (Loomis & Knapp, 2003; Witmer & Kline, 1998). Such misestimations can have serious consequences. Accordingly, training programs were developed. They typically consisted of two steps. First, trainees estimated a distance in a given metric, such as yards (Gibson & Bergman, 1954) or feet (Ferris, 1972, 1973a, 1973b; Jones, DeLucia, Hall, & Johnson, 2009; Niall, Reising, & Martin, 1999; Reising & Martin, 1994, 1995; Waller, 1999). Second, trainees were told the actual metric distance as feedback about their performance (Ferris, 1972, 1973a, 1973b; Gibson & Bergman, 1954; Gibson, Bergman, & Purdy, 1955; Jones et al., 2009; Niall et al., 1999; Reising & Martin, 1994, 1995; Waller, 1999). This training will hereafter be referred to as metric estimation training. Such training improved metric distance estimation (Ferris, 1972, 1973a, 1973b; Gibson & Bergman, 1954; Gibson et al., 1955; Jones et al., 2009; Niall et al., 1999; Reising & Martin, 1994, 1995;
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
606 May 2014 - Human Factors
Waller, 1999). Furthermore, it positively transferred to a different task (Jones et al., 2009) and to certain settings (Ferris, 1972, 1973a). It also, however, negatively transferred to another task and other settings. For example, in prior work, we demonstrated that metric estimation training degraded participants’ accuracy when they threw an object to targets (Jones et al., 2009). Similarly, Ferris (1972) demonstrated that metric estimation training conducted in air negatively transferred to metric distance estimates made in turbid water. These results suggest that metric estimation training has promise and also has risk. Accordingly, it is important to determine why negative transfer occurred. With that knowledge, trainees could be taught when they should and should not apply what they learned to other distance estimation tasks and settings. Doing so could maximize the promise and minimize the risk associated with metric estimation training. In prior work, we explored why metric estimation training negatively transferred to throwing a beanbag to targets (Jones et al., 2009, Experiment 2). We explained our results in terms of what we called cognitive intrusion (Jones et al., 2009). Our participants threw a beanbag to targets during pretesting, completed metric estimation training with or without feedback (i.e., being or not being told the actual metric distance between themselves and targets, respectively), and then threw a beanbag to targets during posttesting. We argued that participants threw the beanbag in a primarily perceptual-motor way during pretesting. Furthermore, we argued that participants who did not receive feedback during training learned nothing, whereas those who received feedback during training learned cognitive strategies about metric estimation. For example, trainees who underestimated during training may have learned to compensate by overestimating subsequent distances. Finally, we argued that participants who learned nothing during training threw the beanbag in a primarily perceptual-motor way during posttesting, whereas those who learned cognitive strategies during training threw the beanbag in a way that exploited those strategies during posttesting. For example, participants who learned to overestimate may have generated an explicit mental estimate of the metric distance between themselves and the targets, and then threw farther than they would have otherwise. In doing so, participants who
learned cognitive strategies during training allowed cognition to intrude on what would have otherwise been a primarily perceptual-motor task. Such cognitive intrusion can degrade the performance of primarily perceptual-motor tasks. For example, putting accuracy degraded when participants consciously attended to their putting movements (Beilock & Carr, 2001; Beilock & Gonso, 2008), the accuracy of grasping movements degraded when they were based on participants’ memories of the to-be-grasped objects’ location (Hu & Goodale, 2000; Milner, Paulignan, Dijkerman, & Jeannerod, 1999), and the accuracy of reach-ability judgments degraded when participants analytically reflected on their judgments (Heft, 1993). According to our explanation, metric estimation training negatively transferred to our throwing task because participants injected cognition into what would have otherwise been a primarily perceptual-motor task. If so, then making that cognition a necessary part of our throwing task should prevent the negative transfer. To test that possibility, we changed the task from “throw to a target” to “throw to a distance.” Specifically, we had participants throw a beanbag to specific metric distances (e.g., throw to 30 feet, then 20 feet, etc.) during pretesting, complete metric estimation training with or without feedback, and then throw a beanbag to specific metric distances during posttesting (Jones et al., 2009, Experiment 4). Half received feedback, but only during training. Metric estimation training with feedback improved throwing to specific metric distances. In other words, making cognition (i.e., metric distance estimation) a necessary part of our throwing task (instead of the instruction “throw to a target”) prevented the negative transfer. This supported our argument that cognitive intrusion caused the negative transfer that we observed in our previous studies. To further test our explanation, we attacked the problem from a different angle. Specifically, we noted that our explanation did not discuss intrusion in general, for example, one task intruding on another. Rather, it discussed cognition intruding on what would have otherwise been a primarily perceptual-motor task. Accordingly, our explanation suggests that cognition can intrude on primarily perceptual-motor tasks, but primarily perceptual-motor tasks cannot intrude on primarily
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
607
Table 1: Overview of the Experimental Design
Pre/Posttest Task Throwing
Training Feedback Feedback No feedback
Verbal estimation
Feedback No feedback
Pretest Throwing without feedback Throwing without feedback Verbal estimation without feedback Verbal estimation without feedback
Phase
Training
Posttest
Throwing with feedback Throwing without feedback Throwing with feedback Throwing without feedback
Throwing without feedback Throwing without feedback Verbal estimation without feedback Verbal estimation without feedback
Note. Please note that training always involved throwing and that feedback, when provided, occurred only during training.
cognitive tasks. If correct, then swapping our original training and transfer tasks (i.e., throwing becomes training and estimating becomes transfer) should also prevent the negative transfer. To test that possibility, during pretesting, participants in the present study verbally estimated the metric distances between themselves and targets, or they threw a beanbag to targets. During training, participants donned goggles that made targets appear closer than they were to create an opportunity for learning. While wearing the goggles, participants threw a beanbag to targets. Half received feedback (i.e., they saw where the beanbag landed), but only during training. During posttesting, participants removed the goggles, and then completed the same task that they performed during pretesting. We predicted that this new type of training would affect subsequent throws, but would not transfer to metric estimation. In other words, we predicted that what was essentially our original transfer task (i.e., throwing) would affect subsequent throwing, but would not affect what was essentially our original training task (i.e., metric distance estimation). Method Participants
A total of 48 undergraduates (24 male, 24 female) participated for course credit. Their
ages ranged from 18 to 38 years (M = 22.15, SD = 3.85). Participants reported that they did not have motor impairments, and either that they did not need vision correction or that correction was accomplished via contact lenses. This was necessary because glasses prevented the goggles from fitting properly. Participants reported playing 0 to 49 (M = 10.98, SD = 10.68) seasons of 0 to 6 (M = 2.27, SD = 1.50) sports throughout their lifetime. All participants were raised measuring in feet and inches. Experimental Design
A mixed design was employed. The withinsubjects variable was phase (pretesting, throwing training, posttesting). The between-subjects variables were pre/posttesting task (throwing, verbal estimation), which concerned whether participants threw a beanbag to targets or verbally estimated the metric distance between themselves and targets during pre- and posttesting, as well as throwing training feedback (visual feedback, no visual feedback), which concerned whether participants saw where the beanbag came to rest during throwing training. Table 1 provides an overview of the experimental design. To be consistent with our previous work, the dependent variable was absolute percentage error.
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
608 May 2014 - Human Factors
Figure 2. The testing area, which was divided into two 22-foot areas. A viewing location was located on one end of each area. Two viewing locations were employed to ensure that the present results were not specific to a given viewpoint. Figure 1. One pair of the goggles worn by participants. Both pairs were identical except for the lenses, which either did or did not distort vision. Absolute percentage error (throwing) = distance beanbag was thrown – actual target distance actual target distance Absolute percentage error (verbal estimation) = verbal estimation of distance – actual target distance
× 100
× 100
actual target distance
Please note the following. First, signs were removed to prevent positive and negative differences from cancelling, which would suggest an incorrect level of accuracy when the data were averaged. Second, numerators were divided by the target’s actual distance to control for error increasing as distance increased (Gilinsky, 1951). This was important because our randomization process allowed distance ranges to vary slightly across trials, as well as across participants. Apparatus
The materials included a beanbag (7.5" × 5", 15.1 oz.), a standard barbell weight (4.75" diameter) that served as the target, and two pairs of goggles, each with 140° of horizontal and 115° of vertical field of view. Black opaque cardstock was attached to the outer and underneath portions of the goggles’ frames (Figure 1). It ensured that participants looked through the goggles’ lenses, and prevented participants from seeing their arm movements (Beckett, 1980; Martin, Keating, Goodkin, Bastian, & Thach, 1996; Redding, Rossetti, & Wallace, 2005). This was important because visual information about arm movements could have confounded the intended visual feedback. One pair of goggles housed
20-diopter base-up prism lenses (Bernell Optics). These lenses shifted the environment downward by approximately 11.4° (Ooi, Wu, & He, 2001; Redding & Wallace, 2006; Williams, Rasor, & Narasimham, 2009), which made environmental objects appear closer than they were (Ooi et al., 2001). The other pair of goggles, which was used during pre- and posttesting, housed clear plastic lenses that did not distort vision. The testing area consisted of two viewing locations that were 50 feet apart (Figure 2). Two viewing locations were employed to ensure that the results were not unique to a given viewpoint. Procedure
Participants were tested individually. They were assigned to one of the two viewing locations, and to one of the four combinations of pre/ posttesting task and throwing training feedback: (a) throwing–visual feedback, (b) throwing–no visual feedback, (c) verbal estimation–visual feedback, and (d) verbal estimation–no visual feedback (Table 1). Assignment was random with the constraint that viewing location and gender were balanced across conditions. Gender was balanced because sometimes men and women perform differently on tasks related to distance (for a review, see Kimura, 1999).
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
Practice. Participants practiced throwing so that actual testing reflected their throwing ability rather than learning the required task. Participants did not practice verbal metric estimation because participants intuitively understood the task requirements. All participants stood at their assigned viewing location. A target (i.e., a barbell weight) was 16 feet in front of them. This distance was not used during actual testing. Participants threw the beanbag underhanded to the target five times with their preferred hand. Their goal was to get the beanbag to rest as close as possible to the target’s center. After each throw, participants viewed the beanbag as it came to rest. Then, participants completed a second set of five throws during which an experimenter instructed participants to close their eyes and face away from the target after the beanbag was released. Participants quickly learned to do so. During later testing, this procedure eliminated visual feedback about throwing accuracy and allowed experimenters to measure the distance of the throw without biasing the participants. Throughout practice, experimenters stressed that participants should throw the beanbag underhanded so that it came to rest as close as possible to the target’s center. Pretesting. Participants put on the goggles that did not distort vision. Then, they either threw a beanbag to targets or verbally estimated distances between themselves and targets. Participants who threw stood at their assigned viewing location. They then threw the beanbag to a target that was positioned at four distances, which were chosen randomly with the following constraints: (a) one of the four distances was between 10 and 12.5 feet, 13 and 15.5 feet, 16.5 and 19 feet, and 19.5 and 22 feet and (b) distances were not reused during a participant’s testing. When the beanbag left their hands, participants were told to close their eyes and turn away from the testing area. Note that this differs from blind throwing wherein participants close their eyes before throwing (Eby & Loomis, 1987; Sahm, Creem-Regehr, Thompson, & Willemsen, 2005). While participants faced away from the testing area, experimenters measured the participant-to-beanbag distance and then repositioned the target.
609
Participants who made verbal estimations completed an identical pretest, except they provided a verbal estimation, in feet and inches, of the distance between themselves and the target’s center. They were not given any information about the size of a foot or an inch. Training. Participants put on the goggles that distorted vision. They then threw the beanbag to a target that was positioned at 16 distances. Distances were selected randomly with the following constraints: (a) four target distances were between 10 and 12.5 feet, 13 and 15.5 feet, 16.5 and 19 feet, and 19.5 and 22 feet and (b) distances were not reused during a participant’s testing. Participants assigned to the visual feedback condition saw where the beanbag landed relative to the target, and then were asked to face away from the testing area. Participants assigned to the no visual feedback condition were instructed to close their eyes and face away from the testing area after the beanbag left their hands, and to keep their eyes closed until they were told to face the testing area again. This prevented learning that might have otherwise taken place between trials. An experimenter checked that participants complied with instructions. While participants faced away from the testing area, experimenters measured the participant-to-beanbag distance and then repositioned the target. Posttesting. Participants put on the goggles that did not distort vision. They then completed a posttest that was identical to their pretest with the exception that the target distances were different. Results
Absolute percentage error was calculated for each trial. Those scores were screened for outliers by converting them into z scores, which were calculated separately for each trial in each condition. If a z score was more extreme than ±3, then its associated data point was an outlier (Tabachnick & Fidell, 1989). Four outliers were identified and replaced with the most extreme score in that portion of the data that was not an outlier. Doing so avoided missing data, reduced the outliers’ impact, and ensured that variance
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
610 May 2014 - Human Factors Table 2: Correlations Among Demographic Variables and Performance in the Throwing Condition Gender Pretest Training Trial 1 Training Trials 1–4 Training Trial 16 Posttest
rpb .174 .159 .408 –.012 .274
Sports
Seasons
p
r
p
r
p
.416 .457 .048 .957 .194
–.254 –.07 –.301 –.394 –.164
.230 .746 .153 .057 .443
–.295 –.062 –.07 .066 –.222
.161 .772 .744 .758 .297
Note. rpb denotes analyses using the point-biserial correlation.
Table 3: Correlations Among Demographic Variables and Performance in the Verbal Metric Estimation Condition Gender Pretest Training Trial 1 Training Trial 16 Posttest
rpb .120 –.242 .052 .250
Sports
Seasons
p
r
p
r
p
.576 .255 .811 .238
.348 –.187 –.109 .413
.095 .381 .611 .045
.267 –.242 –.165 .411
.206 .255 .440 .046
Note. rpb denotes analyses using the point-biserial correlation.
was not unnecessarily reduced (Tabachnick & Fidell, 1989). Did Performance Correlate With Gender or Sports Experience?
It was prudent to determine whether our dependent variable significantly correlated with gender, sports experience, or both. Relations between absolute percentage error and gender were measured via point-biserial correlations because the former was continuous and the latter was discrete dichotomous (Senter, 1969). Relations between absolute percentage error and the number of sports played, and the number of seasons played, were measured via Pearson’s product–moment correlations because all variables were continuous (Senter, 1969). Separate correlations were computed for each phase/ task combination (e.g., pretesting throwing) to preserve relations that might be lost if the data were aggregated. Correlations were grouped into families, each included the three correlations associated with a given phase/task combination. The significance
of each correlation was evaluated at the .017 level (i.e., .05/3). Thus, the Bonferroni correction maintained familywise error at .05 (Tabachnick & Fidell, 2007). Tables 2 and 3 provide the correlations and their p values. The correlations were not significant. Accordingly, gender and sports experience were not included in subsequent analyses. Analytic Approach
Our hypotheses called for specific comparisons, so omnibus analyses of variance were not employed. Rather, a minimally sufficient analysis, composed of paired samples t tests, was utilized (Wilkinson & APA Task Force on Statistical Inference, 1999). Tests associated with separate research questions were grouped into families. One-tailed t tests were employed when predictions were directional (Kirk, 1995). The Bonferroni correction maintained alpha at .05 per family (Tabachnick & Fidell, 2007). Effect sizes accounted for the correlation between means (Dunlap, Cortina, Vaslow, & Burke, 1996).
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
611
Did the Distorting Goggles Degrade Throwing?
Participants wore distorting goggles during training. If the goggles served their purpose, then throwing during the pretest should have been less errorful than throwing during training. To test that possibility, a paired samples t test compared the average of the pretest trials to the average of the first four training trials. Only the data from the throwing–no visual feedback condition were examined because it was the only condition in which pretesting throws could be compared to training throws without bias from feedback. That test was a family of one, so alpha was .05. The test indicated that error during pretesting (M = 7.6, SE = 0.6) was significantly less than error at the beginning of training (M = 12.2, SE = 1.1), t(11) = −4.10, p = .001 (one-tailed), d = 1.36. This probably does not reflect fatigue because these throws only took a few minutes, and none of the participants complained of fatigue. Accordingly, this outcome suggests that the distorting goggles degraded throwing at the beginning of training. Did Feedback Improve Throwing During Training?
Half of our participants watched the beanbag come to rest during training. This feedback should have improved throwing during training. If so, then throws at the beginning of training should have been more errorful than throws at the end of training. In contrast, throws at the beginning and end of training should have been equally errorful when feedback was not provided. Figure 3 presents throwing error as a function of training feedback (visual feedback, no visual feedback) and training trial (first, last). All participants threw during training, so data were collapsed across the throwing and verbal conditions. Data from Figure 3 were subjected to two paired samples t tests, which compared throwing error during the first training trial to throwing error during the last training trial in the visual feedback and no visual feedback conditions. These two tests were a family, so alpha was .025 per test. In the visual feedback condition, throwing error during the first training trial (M = 9.3, SE = 1.5) was
Figure 3. Mean absolute percentage error for throws as a function of training trial (first, last) and training feedback (visual feedback, no visual feedback). Error bars represent ± 1 standard error of the mean.
significantly different from throwing error during the last training trial (M = 4.6, SE = 0.9), t(23) = 2.76, p < .006 (one-tailed), d = 0.76. However, in the no visual feedback condition, throwing error during the first training trial (M = 9.7, SE = 1.5) was not significantly different from throwing error during the last training trial (M = 10.1, SE = 1.4), t(23) = −0.214, p = .833 (two-tailed), d = 0.06. Thus, feedback improved throwing during training. Did Training With Feedback Persist Into the Throwing Posttest?
Participants removed the distorting goggles after training. After adaptation, participants continue to account for the distortion, even when it is no longer present, until further adaptation takes place (Helmholtz, 1910/1962; Ooi et al., 2001; Redding & Wallace, 1997, 2002, 2006; Welch, 1978, 1986). Consequently, it was expected that pretest throws would be less errorful than posttest throws when feedback was provided during training because participants who received feedback would continue to account for the distortion that was no longer present. In contrast, it was expected that pre- and posttest throws would be equally errorful when feedback was not provided during training. To test that possibility, the four pretest throws were averaged to get each participant’s overall pretest score, and the four posttest throws were averaged to get each participant’s overall posttest score. The left half of Figure 4 presents throwing error as a function of training feedback
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
612 May 2014 - Human Factors
Figure 4. Mean absolute percentage error as a function of phase (pre, post), task (throwing, verbal estimation), and training feedback (visual feedback, no visual feedback). Error bars represent ± 1 standard error of the mean.
(visual feedback, no visual feedback) and phase (pretesting, posttesting). The data from the left half of Figure 4 were subjected to two paired samples t tests, which compared average pre- and posttest throwing error in the visual feedback and no visual feedback conditions. These two t tests were a family, so alpha was .025 per test. In the visual feedback condition, average pretest throwing error (M = 5.7, SE = 0.9) was significantly different from average posttest throwing error (M = 9.4, SE = 0.9), t(11) = −2.55, p = .014 (one-tailed), d = 0.94. However, in the no visual feedback condition, average pretest throwing error (M = 7.6. SE = 0.6) was not significantly different from average posttest throwing error (M = 7.0, SE = 1.2), t(11) = 0.574, p = .577 (two-tailed), d = 0.19. Thus, throwing degraded when participants who received feedback during training removed the distorting goggles. In contrast, throwing did not degrade when participants who did not receive feedback during training removed the distorting goggles. Accordingly, throwing training with feedback persisted into the throwing posttest. Did Training With Feedback Transfer to the Verbal Metric Estimation Posttest?
The final set of analyses concerned whether throwing training with feedback transferred to the verbal metric estimation posttest. If so, then pretest verbal estimations should have been less errorful than posttest verbal estimations when
participants received feedback during training. If not, then pre- and posttest verbal estimations should have been equally errorful irrespective of feedback during training. To examine that possibility, the four pretest verbal estimations were averaged to get each participant’s overall pretest score, and the four posttest verbal estimations were averaged to get each participant’s overall posttest score. The right half of Figure 4 presents verbal estimation error as a function of training feedback (visual feedback, no visual feedback) and phase (pretesting, posttesting). Data from the right half of Figure 4 were subjected to two paired samples t tests, which compared average pre- and posttest verbal estimation error in the visual feedback and no visual feedback conditions. These two tests were a family, so alpha was .025 per test. In the visual feedback condition, average pretest verbal estimation error (M = 31.5, SE = 3.0) was not significantly different from average posttest verbal estimation error (M = 29.8, SE = 2.8), t(11) = 0.805, p = .438 (two-tailed), d = 0.18. Likewise, in the no visual feedback condition, average pretest verbal estimation error (M = 28.2, SE = 4.4) was not significantly different from average posttest verbal estimation error (M = 26.5, SE = 4.5), t(11) = 0.734, p = .478 (two-tailed), d = 0.12. Thus, throwing training with feedback did not transfer to the verbal metric estimation task. Discussion
In prior work, we argued that metric estimation training negatively transferred to a throwing task because participants allowed cognitive strategies that they learned during training to intrude on what would have otherwise been a primarily perceptual-motor task (Jones et al., 2009). That explanation suggested that cognition can intrude on primarily perceptual-motor tasks, but primarily perceptual-motor tasks cannot intrude on primarily cognitive tasks. Accordingly, we predicted that swapping our original training and transfer tasks would prevent the negative transfer. More specifically, we predicted that what was essentially our original transfer task (i.e., throwing) would affect subsequent throwing, but would not affect what was essentially our original training task (i.e., metric
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
distance estimation). The results of the present study confirmed our predictions, which supported our argument that the negative transfer that we reported in Jones et al. (2009) stemmed from cognitive intrusion. To be clear, we do not claim that the negative transfer across settings that was reported in Ferris (1972, 1973a) also stemmed from cognitive intrusion. Future research should investigate why that negative transfer occurred. Practically speaking, our results suggested that trainees should not generalize what is learned during metric estimation training to distance estimation tasks that are primarily perceptual-motor in nature. The evidence to date suggests that doing so would degrade the performance of those perceptual-motor tasks. More generally, the possibility that cognitive intrusion led to negative transfer suggests that those who develop or implement metric estimation training programs can no longer assume that distance estimation is distance estimation. This may seem self-evident. To our knowledge, though, existing metric estimation training programs have sometimes been founded on the assumption that distance estimation is a singular process. For example, Ferris (1972) suggested that his variant of metric estimation training could improve a diver’s ability to perform tasks underwater. Similarly, Niall et al. (1999) argued that their variant of metric estimation training could improve a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle. In both cases, these claims were made without empirical validation or logical arguments that were based on a detailed understanding of the relevant tasks. As such, it appears that these claims were based on the assumption that improving one type of distance estimation would necessarily improve a potentially different type of distance estimation. Our data suggest that such an assumption is not tenable. Instead, metric estimation training programs need to account for the fact that different distance estimation tasks require different processes, and those differences can affect how metric estimation training transfers to other distance estimation tasks (Jones et al., 2009). Similarly, metric estimation training programs need to account for the fact that distance estimation in
613
one setting may be different than distance estimation in a different setting (Ferris, 1972, 1973a). For that to happen, we need a better understanding of the different types of distance estimation tasks and settings. Specifically, we need a better understanding of metric distance estimation, including what is learned during metric estimation training. Furthermore, we need a better understanding of operationally relevant distance estimation tasks. For example, if we want metric estimation training to improve a diver’s ability to perform various distance-related tasks underwater, then we must understand the processes that those distance-related tasks require. Likewise, if we want metric estimation training to improve a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle, then we must understand how such judgments are made. Finally, we need a better understanding of the settings in which metric distance estimation normally takes place. Without that knowledge, there is no a priori reason to expect that metric estimation training would improve the performance of those operationally relevant tasks in their natural settings. Furthermore, we need a better understanding of how metric estimation training transfers to other distance estimation tasks, as well as how such training transfers to other settings. For example, we need to know how metric estimation training affects a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle, as well as whether those results depend on when and where the training and transfer scenarios take place. To date, only a few studies have examined the transfer of metric estimation training (Ferris, 1972, 1973a; Jones et al., 2009). Table 4 depicts the essence of those studies, including their training and transfer scenarios as well as the direction of the observed transfer. Inspection of the table suggests that the transfer of metric estimation training to other distance estimation tasks and settings is not straightforward. In summary, the results of the present study supported our argument that the negative transfer across tasks that we reported in Jones et al. (2009) stemmed from cognitive intrusion. Our collective results suggest that trainees should not generalize what they learned during metric
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
614 May 2014 - Human Factors Table 4: Experiments That Investigated Whether Metric Estimation Training Transfers to Other Settings or Tasks Publication
Training
Ferris (1972)
Metric estimation in air Metric estimation in air Metric estimation in moderately turbid water Metric estimation in moderately turbid water Ferris (1973a) Metric estimation in clear water Metric estimation in turbid water Metric estimation Jones, DeLucia, Hall, & Johnson Metric estimation (2009) Metric estimation
Transfer
Direction
Metric estimation in clear water Metric estimation in turbid water Metric estimation in clear water
+ – –
Metric estimation in very turbid water Metric estimation in turbid water Metric estimation in clear water Throwing an object to targets Throwing an object to targets Throwing an object to a metric distance
+ – – – – +
Note. The second entry for Jones et al. (2009) was a replication.
estimation training to distance estimation tasks that are primarily perceptual-motor in nature. Furthermore, our results suggest that researchers must develop a better understanding of distance estimation tasks and settings, as well as how metric estimation training transfers to those tasks and settings. With that knowledge, one could supplement metric estimation training with information about when it would be safe and unsafe to generalize what was learned during metric estimation training to other distance estimation tasks and settings. In addition, one could develop a model of transfer for metric estimation training, which could support educated guesses about transfer to tasks, settings, or combinations thereof that have yet to be studied, as well as guide future research in this area. Key Points •• In prior work, we found that metric estimation training negatively transferred to throwing (Jones, DeLucia, Hall, & Johnson, 2009). •• We explained our results in terms of cognitive intrusion. •• The present study tested that possibility by swapping our original training and transfer tasks. •• Training involving throwing was effective, but did not transfer to verbal metric distance estimation. •• This supports our argument that the negative transfer observed in our earlier studies stemmed
from cognitive intrusion, and has important implications for the design and implementation of distance estimation training.
References Beckett, P. A. (1980). Development of the third component in prism adaptation: Effects of active and passive movement. Journal of Experimental Psychology, 6, 433–444. Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled performance: What governs choking under pressure? Journal of Experimental Psychology: General, 130, 701–725. Beilock, S. L., & Gonso, S. (2008). Putting in the mind versus putting on the green: Expertise, performance time, and the linking of imagery and action. Quarterly Journal of Experimental Psychology, 61, 920–932. Dunlap, W. P., Cortina, J. M., Vaslow, J. B., & Burke, M. J. (1996). Meta-analysis of experiments with matched groups or repeated measures designs. Psychological Methods, 1, 170–177. Dyer, J. L., & Young, K. M. (1998). Night vision goggle research and training issues for ground forces: A literature review (1082). Alexandria, VA: U.S. Army Research Institute. Eby, D. W., & Loomis, J. M. (1987). A study of visually directed throwing in the presence of multiple distance cues. Perception and Psychophysics, 41, 308–312. Ferris, S. H. (1972). Improvement of absolute distance estimation underwater. Perceptual and Motor Skills, 35, 299–305. Ferris, S. H. (1973a). Improving absolute distance estimation in clear and in turbid water. Perceptual and Motor Skills, 36, 771–776. Ferris, S. H. (1973b). Improving distance estimation underwater: Long-term effectiveness of training. Perceptual and Motor Skills, 36, 1089–1090. Foley, J. M. (1980). Binocular distance perception. Psychological Review, 87, 411–434. Gibson, J. J. (1947). Motion picture testing and research (Aviation Psychology Program Research Report No. 7). Washington, DC: Army Air Forces.
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training Gibson, J. J. (1950). The perception of the visual world. Boston, MA: Houghton Mifflin. Gibson, J. J. (1986). The ecological approach to visual perception. Hillsdale, NJ: Lawrence Erlbaum. (Original work published 1979). Gibson, E. J., & Bergman, R. (1954). The effect of training on absolute estimation of distance over the ground. Journal of Experimental Psychology, 48, 473–482. Gibson, E. J., Bergman, R., & Purdy, J. (1955). The effect of prior training with a scale of distance on absolute and relative judgments of distance over the ground. Journal of Experimental Psychology, 50, 97–105. Gilinsky, A. S. (1951). Perceived size and distance in visual space. Psychological Review, 58, 460–482. Heft, H. (1993). A methodological note on overestimates of reaching distance: Distinguishing between perceptual and analytical judgments. Ecological Psychology, 5, 255–271. Helmholtz, H., von. (1962). Treatise on physiological optics (Vol. 3, 3rd ed.; J. P. C. Southall, Ed. & Trans.). New York: Dover. (Original work published 1910) Hochberg, J. E. (1978). Perception. Englewood Cliffs, NJ: Prentice Hall. Hu, Y., & Goodale, M. A. (2000). Grasping after a delay shifts sizescaling from absolute to relative metrics. Journal of Cognitive Neuroscience, 12, 856–868. Jones, K. S., DeLucia, P. R., Hall, A. R., & Johnson, B. R. (2009). Can metric estimation training hinder actions involving distance? Human Factors, 51, 419–432. Kimura, D. (1999). Sex and cognition. Cambridge, MA: MIT Press. Kirk, R. E. (1995). Experimental design: Procedures for the behavioral sciences (3rd ed.). Pacific Grove, CA: Brooks/Cole. Loomis, J. M., & Knapp, J. M. (2003). Visual perception of egocentric distance in real and virtual environments. In L. J. Hettinger & M. W. Haas (Eds.), Virtual and adaptive environments (pp. 21–46). Mahwah, NJ: Lawrence Erlbaum. Martin, T. A., Keating, J. G., Goodkin, H. P., Bastian, A. J., & Thach, W. T. (1996). Throwing while looking through prisms: II. Specificity and storage of multiple gaze-throw calibrations. Brain, 119, 119–121. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford, UK: Oxford University Press. Milner, A. D., Paulignan, H. C., Dijkerman, F., & Jeannerod, M. M. (1999). A paradoxical improvement of misreaching in optic ataxia: New evidence for two separate neural systems for visual localization. Proceedings of the Royal Society: Biological Sciences, 266, 2225–2229. Niall, K. K., Reising, J. D., & Martin, E. L. (1999). Distance estimation with night vision goggles: A little feedback goes a long way. Human Factors, 41, 495–506. Ooi, T. L., Wu, B., & He, Z. J. (2001). Distance determined by the angular declination below the horizon. Nature, 414, 197–200. Redding, G. M., Rossetti, Y., & Wallace, B. (2005). Applications of prism adaptation: A tutorial in theory and method. Neuroscience and Biobehavioral Reviews, 29, 431–444. Redding, G. M., & Wallace, B. (1997). Adaptive spatial alignment. Mahwah, NJ: Lawrence Erlbaum. Redding, G. M., & Wallace, B. (2002). Strategic calibration and spatial alignment: A model from prism adaptation. Journal of Motor Behavior, 34, 126–138.
615 Redding, G. M., & Wallace, B. (2006). Generalization of prism adaptation. Journal of Experimental Psychology: Human Perception and Performance, 32, 1006–1022. Reising, M., & Martin, E. L. (1994). Distance estimation training with night vision goggles: A preliminary study (AL/HR-TR1994-0090). Dayton, OH: University of Dayton Research Institute. Reising, M., & Martin, E. L. (1995). Distance estimation training with night vision goggles under low illumination (AL/HR-TR1994-0138). Dayton, OH: University of Dayton Research Institute. Sahm, C. S., Creem-Regehr, S. H., Thompson, W. B., & Willemsen, P. (2005). Throwing vs. walking as indicators of distance perception in real and virtual environment. ACM Transactions on Applied Perception, 2, 1–14. Senter, R. J. (1969). Analysis of data: Introductory statistics for the behavioral sciences. Glenview, IL: Scott Foresman. Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics (4th ed.). Boston, MA: Allyn & Bacon. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Allyn & Bacon. Taieb-Maimon, M. (2007). Learning headway estimation in driving. Human Factors, 49, 734–744. Waller, D. (1999). Factors affecting interobject distance perception in virtual environments. Presence: Teleoperators and Virtual Environments, 8, 657–670. Welch, R. B. (1978). Perceptual modification: Adapting to altered sensory environment. London, UK: Academic Press. Welch, R. B. (1986). Adaptation of space perception. Handbook of Perception and Human Performance, 1, 1–45. Wilkinson, L., & APA Task Force on Statistical Inference. (1999). Statistical methods in psychology journals: Guidelines and explanations. American Psychologist, 54, 594–604. Williams, B., Rasor, T., & Narasimham, G. (2009). Distance perception in virtual environments: A closer look at the horizon and error. In Proceedings of the 6th Symposium on Applied Perception in Graphics and Visualization (pp. 7–10). New York, NY: Association for Computing Machinery. Witmer, B. G., & Kline, P. B. (1998). Judging perceived and traversed distance in virtual environments. Presence: Teleoperators and Virtual Environments, 7, 144–167.
Allyson R. Colombo is a human factors engineer at Basic Commerce and Industries (BCI), Inc. in Dahlgren, Virginia. She earned her PhD in human factors psychology from Texas Tech University in 2011. Keith S. Jones is an associate professor of psychology at Texas Tech University. He earned his PhD in human factors psychology from the University of Cincinnati in 2000. Date received: April 21, 2011 Date accepted: June 17, 2013
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
HFSXXX10.1177/0018720813499021Month XXXX - Human FactorsTransfer of Distance Estimation Training
The Effects of Prism Adaptation on Egocentric Metric Distance Estimation Allyson R. Colombo, BCI, Inc., Dahlgren, Virginia, and Keith S. Jones, Texas Tech University, Lubbock, Texas, USA Objective: The present experiment evaluated whether training involving throwing transferred to metric distance estimation (i.e., describing in feet and inches the distance between oneself and targets). Background: In prior work, we found that metric estimation training negatively transferred to throwing. We explained our results in terms of cognitive intrusion. The present study tested that possibility by swapping our training and transfer tasks. Method: During pretesting, participants verbally estimated the metric distances between themselves and targets, or they threw a beanbag to targets. During training, participants donned goggles that distorted their vision. While wearing the goggles, they threw a beanbag to targets. Half received feedback. During posttesting, participants removed the distorting goggles and completed the same task that they performed during pretesting. Results: The results indicated that the distorting goggles degraded throwing at the beginning of training, visual feedback improved throwing during training, the effects of training with feedback persisted into the throwing posttest, and the effects of training with feedback did not transfer to the verbal metric estimation posttest. Conclusion: Training involving throwing was effective, but did not transfer to verbal metric distance estimation. This supports our argument that the negative transfer observed in our previous study stemmed from cognitive intrusion. Application: The present experiment suggests that the creation of distance estimation training should begin with a careful analysis of the transfer task, and that distance estimation training programs should explicitly teach trainees that their training will not generalize to all distance estimation tasks. Keywords: transfer, transfer of training, visual distortion, distorting goggles, throwing, verbal estimation, feedback, prism adaptation
Address correspondence to Keith S. Jones, Department of Psychology, Texas Tech University, Lubbock, TX 794092051, USA; e-mail: [email protected]. HUMAN FACTORS Vol. 56, No. 3, May 2014, pp. 605–615 DOI: 10.1177/0018720813499021 Copyright © 2013, Human Factors and Ergonomics Society.
People estimate distances between themselves and objects in their environments. This ability has been conceptualized in different ways (Foley, 1980; Gibson, 1950, 1979/1986; Helmholtz, 1910/1962; Hochberg, 1978; Milner & Goodale, 1995). For example, Helmholtz (1910/1962) conceptualized distance estimation as the elaboration of inherently ambiguous distance cues via unconscious inference. In contrast, Gibson (1950) conceptualized distance estimation as attention to environmental features that specify distance. Despite their differences, however, the various conceptualizations agree that distance estimation can be inaccurate. That expectation has empirical support. People have misestimated distances while flying aircraft (Gibson, 1947), firing weapons (Rogers, Sprol, Vitereles, Voss, & Wickens, 1945, as cited in Gibson & Bergman, 1954), working underwater (Ferris, 1972, 1973a, 1973b), wearing night-vision goggles (Dyer & Young, 1998), driving cars (Taieb-Maimon, 2007), and exploring virtual environments (Loomis & Knapp, 2003; Witmer & Kline, 1998). Such misestimations can have serious consequences. Accordingly, training programs were developed. They typically consisted of two steps. First, trainees estimated a distance in a given metric, such as yards (Gibson & Bergman, 1954) or feet (Ferris, 1972, 1973a, 1973b; Jones, DeLucia, Hall, & Johnson, 2009; Niall, Reising, & Martin, 1999; Reising & Martin, 1994, 1995; Waller, 1999). Second, trainees were told the actual metric distance as feedback about their performance (Ferris, 1972, 1973a, 1973b; Gibson & Bergman, 1954; Gibson, Bergman, & Purdy, 1955; Jones et al., 2009; Niall et al., 1999; Reising & Martin, 1994, 1995; Waller, 1999). This training will hereafter be referred to as metric estimation training. Such training improved metric distance estimation (Ferris, 1972, 1973a, 1973b; Gibson & Bergman, 1954; Gibson et al., 1955; Jones et al., 2009; Niall et al., 1999; Reising & Martin, 1994, 1995;
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
606 May 2014 - Human Factors
Waller, 1999). Furthermore, it positively transferred to a different task (Jones et al., 2009) and to certain settings (Ferris, 1972, 1973a). It also, however, negatively transferred to another task and other settings. For example, in prior work, we demonstrated that metric estimation training degraded participants’ accuracy when they threw an object to targets (Jones et al., 2009). Similarly, Ferris (1972) demonstrated that metric estimation training conducted in air negatively transferred to metric distance estimates made in turbid water. These results suggest that metric estimation training has promise and also has risk. Accordingly, it is important to determine why negative transfer occurred. With that knowledge, trainees could be taught when they should and should not apply what they learned to other distance estimation tasks and settings. Doing so could maximize the promise and minimize the risk associated with metric estimation training. In prior work, we explored why metric estimation training negatively transferred to throwing a beanbag to targets (Jones et al., 2009, Experiment 2). We explained our results in terms of what we called cognitive intrusion (Jones et al., 2009). Our participants threw a beanbag to targets during pretesting, completed metric estimation training with or without feedback (i.e., being or not being told the actual metric distance between themselves and targets, respectively), and then threw a beanbag to targets during posttesting. We argued that participants threw the beanbag in a primarily perceptual-motor way during pretesting. Furthermore, we argued that participants who did not receive feedback during training learned nothing, whereas those who received feedback during training learned cognitive strategies about metric estimation. For example, trainees who underestimated during training may have learned to compensate by overestimating subsequent distances. Finally, we argued that participants who learned nothing during training threw the beanbag in a primarily perceptual-motor way during posttesting, whereas those who learned cognitive strategies during training threw the beanbag in a way that exploited those strategies during posttesting. For example, participants who learned to overestimate may have generated an explicit mental estimate of the metric distance between themselves and the targets, and then threw farther than they would have otherwise. In doing so, participants who
learned cognitive strategies during training allowed cognition to intrude on what would have otherwise been a primarily perceptual-motor task. Such cognitive intrusion can degrade the performance of primarily perceptual-motor tasks. For example, putting accuracy degraded when participants consciously attended to their putting movements (Beilock & Carr, 2001; Beilock & Gonso, 2008), the accuracy of grasping movements degraded when they were based on participants’ memories of the to-be-grasped objects’ location (Hu & Goodale, 2000; Milner, Paulignan, Dijkerman, & Jeannerod, 1999), and the accuracy of reach-ability judgments degraded when participants analytically reflected on their judgments (Heft, 1993). According to our explanation, metric estimation training negatively transferred to our throwing task because participants injected cognition into what would have otherwise been a primarily perceptual-motor task. If so, then making that cognition a necessary part of our throwing task should prevent the negative transfer. To test that possibility, we changed the task from “throw to a target” to “throw to a distance.” Specifically, we had participants throw a beanbag to specific metric distances (e.g., throw to 30 feet, then 20 feet, etc.) during pretesting, complete metric estimation training with or without feedback, and then throw a beanbag to specific metric distances during posttesting (Jones et al., 2009, Experiment 4). Half received feedback, but only during training. Metric estimation training with feedback improved throwing to specific metric distances. In other words, making cognition (i.e., metric distance estimation) a necessary part of our throwing task (instead of the instruction “throw to a target”) prevented the negative transfer. This supported our argument that cognitive intrusion caused the negative transfer that we observed in our previous studies. To further test our explanation, we attacked the problem from a different angle. Specifically, we noted that our explanation did not discuss intrusion in general, for example, one task intruding on another. Rather, it discussed cognition intruding on what would have otherwise been a primarily perceptual-motor task. Accordingly, our explanation suggests that cognition can intrude on primarily perceptual-motor tasks, but primarily perceptual-motor tasks cannot intrude on primarily
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
607
Table 1: Overview of the Experimental Design
Pre/Posttest Task Throwing
Training Feedback Feedback No feedback
Verbal estimation
Feedback No feedback
Pretest Throwing without feedback Throwing without feedback Verbal estimation without feedback Verbal estimation without feedback
Phase
Training
Posttest
Throwing with feedback Throwing without feedback Throwing with feedback Throwing without feedback
Throwing without feedback Throwing without feedback Verbal estimation without feedback Verbal estimation without feedback
Note. Please note that training always involved throwing and that feedback, when provided, occurred only during training.
cognitive tasks. If correct, then swapping our original training and transfer tasks (i.e., throwing becomes training and estimating becomes transfer) should also prevent the negative transfer. To test that possibility, during pretesting, participants in the present study verbally estimated the metric distances between themselves and targets, or they threw a beanbag to targets. During training, participants donned goggles that made targets appear closer than they were to create an opportunity for learning. While wearing the goggles, participants threw a beanbag to targets. Half received feedback (i.e., they saw where the beanbag landed), but only during training. During posttesting, participants removed the goggles, and then completed the same task that they performed during pretesting. We predicted that this new type of training would affect subsequent throws, but would not transfer to metric estimation. In other words, we predicted that what was essentially our original transfer task (i.e., throwing) would affect subsequent throwing, but would not affect what was essentially our original training task (i.e., metric distance estimation). Method Participants
A total of 48 undergraduates (24 male, 24 female) participated for course credit. Their
ages ranged from 18 to 38 years (M = 22.15, SD = 3.85). Participants reported that they did not have motor impairments, and either that they did not need vision correction or that correction was accomplished via contact lenses. This was necessary because glasses prevented the goggles from fitting properly. Participants reported playing 0 to 49 (M = 10.98, SD = 10.68) seasons of 0 to 6 (M = 2.27, SD = 1.50) sports throughout their lifetime. All participants were raised measuring in feet and inches. Experimental Design
A mixed design was employed. The withinsubjects variable was phase (pretesting, throwing training, posttesting). The between-subjects variables were pre/posttesting task (throwing, verbal estimation), which concerned whether participants threw a beanbag to targets or verbally estimated the metric distance between themselves and targets during pre- and posttesting, as well as throwing training feedback (visual feedback, no visual feedback), which concerned whether participants saw where the beanbag came to rest during throwing training. Table 1 provides an overview of the experimental design. To be consistent with our previous work, the dependent variable was absolute percentage error.
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
608 May 2014 - Human Factors
Figure 2. The testing area, which was divided into two 22-foot areas. A viewing location was located on one end of each area. Two viewing locations were employed to ensure that the present results were not specific to a given viewpoint. Figure 1. One pair of the goggles worn by participants. Both pairs were identical except for the lenses, which either did or did not distort vision. Absolute percentage error (throwing) = distance beanbag was thrown – actual target distance actual target distance Absolute percentage error (verbal estimation) = verbal estimation of distance – actual target distance
× 100
× 100
actual target distance
Please note the following. First, signs were removed to prevent positive and negative differences from cancelling, which would suggest an incorrect level of accuracy when the data were averaged. Second, numerators were divided by the target’s actual distance to control for error increasing as distance increased (Gilinsky, 1951). This was important because our randomization process allowed distance ranges to vary slightly across trials, as well as across participants. Apparatus
The materials included a beanbag (7.5" × 5", 15.1 oz.), a standard barbell weight (4.75" diameter) that served as the target, and two pairs of goggles, each with 140° of horizontal and 115° of vertical field of view. Black opaque cardstock was attached to the outer and underneath portions of the goggles’ frames (Figure 1). It ensured that participants looked through the goggles’ lenses, and prevented participants from seeing their arm movements (Beckett, 1980; Martin, Keating, Goodkin, Bastian, & Thach, 1996; Redding, Rossetti, & Wallace, 2005). This was important because visual information about arm movements could have confounded the intended visual feedback. One pair of goggles housed
20-diopter base-up prism lenses (Bernell Optics). These lenses shifted the environment downward by approximately 11.4° (Ooi, Wu, & He, 2001; Redding & Wallace, 2006; Williams, Rasor, & Narasimham, 2009), which made environmental objects appear closer than they were (Ooi et al., 2001). The other pair of goggles, which was used during pre- and posttesting, housed clear plastic lenses that did not distort vision. The testing area consisted of two viewing locations that were 50 feet apart (Figure 2). Two viewing locations were employed to ensure that the results were not unique to a given viewpoint. Procedure
Participants were tested individually. They were assigned to one of the two viewing locations, and to one of the four combinations of pre/ posttesting task and throwing training feedback: (a) throwing–visual feedback, (b) throwing–no visual feedback, (c) verbal estimation–visual feedback, and (d) verbal estimation–no visual feedback (Table 1). Assignment was random with the constraint that viewing location and gender were balanced across conditions. Gender was balanced because sometimes men and women perform differently on tasks related to distance (for a review, see Kimura, 1999).
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
Practice. Participants practiced throwing so that actual testing reflected their throwing ability rather than learning the required task. Participants did not practice verbal metric estimation because participants intuitively understood the task requirements. All participants stood at their assigned viewing location. A target (i.e., a barbell weight) was 16 feet in front of them. This distance was not used during actual testing. Participants threw the beanbag underhanded to the target five times with their preferred hand. Their goal was to get the beanbag to rest as close as possible to the target’s center. After each throw, participants viewed the beanbag as it came to rest. Then, participants completed a second set of five throws during which an experimenter instructed participants to close their eyes and face away from the target after the beanbag was released. Participants quickly learned to do so. During later testing, this procedure eliminated visual feedback about throwing accuracy and allowed experimenters to measure the distance of the throw without biasing the participants. Throughout practice, experimenters stressed that participants should throw the beanbag underhanded so that it came to rest as close as possible to the target’s center. Pretesting. Participants put on the goggles that did not distort vision. Then, they either threw a beanbag to targets or verbally estimated distances between themselves and targets. Participants who threw stood at their assigned viewing location. They then threw the beanbag to a target that was positioned at four distances, which were chosen randomly with the following constraints: (a) one of the four distances was between 10 and 12.5 feet, 13 and 15.5 feet, 16.5 and 19 feet, and 19.5 and 22 feet and (b) distances were not reused during a participant’s testing. When the beanbag left their hands, participants were told to close their eyes and turn away from the testing area. Note that this differs from blind throwing wherein participants close their eyes before throwing (Eby & Loomis, 1987; Sahm, Creem-Regehr, Thompson, & Willemsen, 2005). While participants faced away from the testing area, experimenters measured the participant-to-beanbag distance and then repositioned the target.
609
Participants who made verbal estimations completed an identical pretest, except they provided a verbal estimation, in feet and inches, of the distance between themselves and the target’s center. They were not given any information about the size of a foot or an inch. Training. Participants put on the goggles that distorted vision. They then threw the beanbag to a target that was positioned at 16 distances. Distances were selected randomly with the following constraints: (a) four target distances were between 10 and 12.5 feet, 13 and 15.5 feet, 16.5 and 19 feet, and 19.5 and 22 feet and (b) distances were not reused during a participant’s testing. Participants assigned to the visual feedback condition saw where the beanbag landed relative to the target, and then were asked to face away from the testing area. Participants assigned to the no visual feedback condition were instructed to close their eyes and face away from the testing area after the beanbag left their hands, and to keep their eyes closed until they were told to face the testing area again. This prevented learning that might have otherwise taken place between trials. An experimenter checked that participants complied with instructions. While participants faced away from the testing area, experimenters measured the participant-to-beanbag distance and then repositioned the target. Posttesting. Participants put on the goggles that did not distort vision. They then completed a posttest that was identical to their pretest with the exception that the target distances were different. Results
Absolute percentage error was calculated for each trial. Those scores were screened for outliers by converting them into z scores, which were calculated separately for each trial in each condition. If a z score was more extreme than ±3, then its associated data point was an outlier (Tabachnick & Fidell, 1989). Four outliers were identified and replaced with the most extreme score in that portion of the data that was not an outlier. Doing so avoided missing data, reduced the outliers’ impact, and ensured that variance
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
610 May 2014 - Human Factors Table 2: Correlations Among Demographic Variables and Performance in the Throwing Condition Gender Pretest Training Trial 1 Training Trials 1–4 Training Trial 16 Posttest
rpb .174 .159 .408 –.012 .274
Sports
Seasons
p
r
p
r
p
.416 .457 .048 .957 .194
–.254 –.07 –.301 –.394 –.164
.230 .746 .153 .057 .443
–.295 –.062 –.07 .066 –.222
.161 .772 .744 .758 .297
Note. rpb denotes analyses using the point-biserial correlation.
Table 3: Correlations Among Demographic Variables and Performance in the Verbal Metric Estimation Condition Gender Pretest Training Trial 1 Training Trial 16 Posttest
rpb .120 –.242 .052 .250
Sports
Seasons
p
r
p
r
p
.576 .255 .811 .238
.348 –.187 –.109 .413
.095 .381 .611 .045
.267 –.242 –.165 .411
.206 .255 .440 .046
Note. rpb denotes analyses using the point-biserial correlation.
was not unnecessarily reduced (Tabachnick & Fidell, 1989). Did Performance Correlate With Gender or Sports Experience?
It was prudent to determine whether our dependent variable significantly correlated with gender, sports experience, or both. Relations between absolute percentage error and gender were measured via point-biserial correlations because the former was continuous and the latter was discrete dichotomous (Senter, 1969). Relations between absolute percentage error and the number of sports played, and the number of seasons played, were measured via Pearson’s product–moment correlations because all variables were continuous (Senter, 1969). Separate correlations were computed for each phase/ task combination (e.g., pretesting throwing) to preserve relations that might be lost if the data were aggregated. Correlations were grouped into families, each included the three correlations associated with a given phase/task combination. The significance
of each correlation was evaluated at the .017 level (i.e., .05/3). Thus, the Bonferroni correction maintained familywise error at .05 (Tabachnick & Fidell, 2007). Tables 2 and 3 provide the correlations and their p values. The correlations were not significant. Accordingly, gender and sports experience were not included in subsequent analyses. Analytic Approach
Our hypotheses called for specific comparisons, so omnibus analyses of variance were not employed. Rather, a minimally sufficient analysis, composed of paired samples t tests, was utilized (Wilkinson & APA Task Force on Statistical Inference, 1999). Tests associated with separate research questions were grouped into families. One-tailed t tests were employed when predictions were directional (Kirk, 1995). The Bonferroni correction maintained alpha at .05 per family (Tabachnick & Fidell, 2007). Effect sizes accounted for the correlation between means (Dunlap, Cortina, Vaslow, & Burke, 1996).
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
611
Did the Distorting Goggles Degrade Throwing?
Participants wore distorting goggles during training. If the goggles served their purpose, then throwing during the pretest should have been less errorful than throwing during training. To test that possibility, a paired samples t test compared the average of the pretest trials to the average of the first four training trials. Only the data from the throwing–no visual feedback condition were examined because it was the only condition in which pretesting throws could be compared to training throws without bias from feedback. That test was a family of one, so alpha was .05. The test indicated that error during pretesting (M = 7.6, SE = 0.6) was significantly less than error at the beginning of training (M = 12.2, SE = 1.1), t(11) = −4.10, p = .001 (one-tailed), d = 1.36. This probably does not reflect fatigue because these throws only took a few minutes, and none of the participants complained of fatigue. Accordingly, this outcome suggests that the distorting goggles degraded throwing at the beginning of training. Did Feedback Improve Throwing During Training?
Half of our participants watched the beanbag come to rest during training. This feedback should have improved throwing during training. If so, then throws at the beginning of training should have been more errorful than throws at the end of training. In contrast, throws at the beginning and end of training should have been equally errorful when feedback was not provided. Figure 3 presents throwing error as a function of training feedback (visual feedback, no visual feedback) and training trial (first, last). All participants threw during training, so data were collapsed across the throwing and verbal conditions. Data from Figure 3 were subjected to two paired samples t tests, which compared throwing error during the first training trial to throwing error during the last training trial in the visual feedback and no visual feedback conditions. These two tests were a family, so alpha was .025 per test. In the visual feedback condition, throwing error during the first training trial (M = 9.3, SE = 1.5) was
Figure 3. Mean absolute percentage error for throws as a function of training trial (first, last) and training feedback (visual feedback, no visual feedback). Error bars represent ± 1 standard error of the mean.
significantly different from throwing error during the last training trial (M = 4.6, SE = 0.9), t(23) = 2.76, p < .006 (one-tailed), d = 0.76. However, in the no visual feedback condition, throwing error during the first training trial (M = 9.7, SE = 1.5) was not significantly different from throwing error during the last training trial (M = 10.1, SE = 1.4), t(23) = −0.214, p = .833 (two-tailed), d = 0.06. Thus, feedback improved throwing during training. Did Training With Feedback Persist Into the Throwing Posttest?
Participants removed the distorting goggles after training. After adaptation, participants continue to account for the distortion, even when it is no longer present, until further adaptation takes place (Helmholtz, 1910/1962; Ooi et al., 2001; Redding & Wallace, 1997, 2002, 2006; Welch, 1978, 1986). Consequently, it was expected that pretest throws would be less errorful than posttest throws when feedback was provided during training because participants who received feedback would continue to account for the distortion that was no longer present. In contrast, it was expected that pre- and posttest throws would be equally errorful when feedback was not provided during training. To test that possibility, the four pretest throws were averaged to get each participant’s overall pretest score, and the four posttest throws were averaged to get each participant’s overall posttest score. The left half of Figure 4 presents throwing error as a function of training feedback
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
612 May 2014 - Human Factors
Figure 4. Mean absolute percentage error as a function of phase (pre, post), task (throwing, verbal estimation), and training feedback (visual feedback, no visual feedback). Error bars represent ± 1 standard error of the mean.
(visual feedback, no visual feedback) and phase (pretesting, posttesting). The data from the left half of Figure 4 were subjected to two paired samples t tests, which compared average pre- and posttest throwing error in the visual feedback and no visual feedback conditions. These two t tests were a family, so alpha was .025 per test. In the visual feedback condition, average pretest throwing error (M = 5.7, SE = 0.9) was significantly different from average posttest throwing error (M = 9.4, SE = 0.9), t(11) = −2.55, p = .014 (one-tailed), d = 0.94. However, in the no visual feedback condition, average pretest throwing error (M = 7.6. SE = 0.6) was not significantly different from average posttest throwing error (M = 7.0, SE = 1.2), t(11) = 0.574, p = .577 (two-tailed), d = 0.19. Thus, throwing degraded when participants who received feedback during training removed the distorting goggles. In contrast, throwing did not degrade when participants who did not receive feedback during training removed the distorting goggles. Accordingly, throwing training with feedback persisted into the throwing posttest. Did Training With Feedback Transfer to the Verbal Metric Estimation Posttest?
The final set of analyses concerned whether throwing training with feedback transferred to the verbal metric estimation posttest. If so, then pretest verbal estimations should have been less errorful than posttest verbal estimations when
participants received feedback during training. If not, then pre- and posttest verbal estimations should have been equally errorful irrespective of feedback during training. To examine that possibility, the four pretest verbal estimations were averaged to get each participant’s overall pretest score, and the four posttest verbal estimations were averaged to get each participant’s overall posttest score. The right half of Figure 4 presents verbal estimation error as a function of training feedback (visual feedback, no visual feedback) and phase (pretesting, posttesting). Data from the right half of Figure 4 were subjected to two paired samples t tests, which compared average pre- and posttest verbal estimation error in the visual feedback and no visual feedback conditions. These two tests were a family, so alpha was .025 per test. In the visual feedback condition, average pretest verbal estimation error (M = 31.5, SE = 3.0) was not significantly different from average posttest verbal estimation error (M = 29.8, SE = 2.8), t(11) = 0.805, p = .438 (two-tailed), d = 0.18. Likewise, in the no visual feedback condition, average pretest verbal estimation error (M = 28.2, SE = 4.4) was not significantly different from average posttest verbal estimation error (M = 26.5, SE = 4.5), t(11) = 0.734, p = .478 (two-tailed), d = 0.12. Thus, throwing training with feedback did not transfer to the verbal metric estimation task. Discussion
In prior work, we argued that metric estimation training negatively transferred to a throwing task because participants allowed cognitive strategies that they learned during training to intrude on what would have otherwise been a primarily perceptual-motor task (Jones et al., 2009). That explanation suggested that cognition can intrude on primarily perceptual-motor tasks, but primarily perceptual-motor tasks cannot intrude on primarily cognitive tasks. Accordingly, we predicted that swapping our original training and transfer tasks would prevent the negative transfer. More specifically, we predicted that what was essentially our original transfer task (i.e., throwing) would affect subsequent throwing, but would not affect what was essentially our original training task (i.e., metric
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training
distance estimation). The results of the present study confirmed our predictions, which supported our argument that the negative transfer that we reported in Jones et al. (2009) stemmed from cognitive intrusion. To be clear, we do not claim that the negative transfer across settings that was reported in Ferris (1972, 1973a) also stemmed from cognitive intrusion. Future research should investigate why that negative transfer occurred. Practically speaking, our results suggested that trainees should not generalize what is learned during metric estimation training to distance estimation tasks that are primarily perceptual-motor in nature. The evidence to date suggests that doing so would degrade the performance of those perceptual-motor tasks. More generally, the possibility that cognitive intrusion led to negative transfer suggests that those who develop or implement metric estimation training programs can no longer assume that distance estimation is distance estimation. This may seem self-evident. To our knowledge, though, existing metric estimation training programs have sometimes been founded on the assumption that distance estimation is a singular process. For example, Ferris (1972) suggested that his variant of metric estimation training could improve a diver’s ability to perform tasks underwater. Similarly, Niall et al. (1999) argued that their variant of metric estimation training could improve a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle. In both cases, these claims were made without empirical validation or logical arguments that were based on a detailed understanding of the relevant tasks. As such, it appears that these claims were based on the assumption that improving one type of distance estimation would necessarily improve a potentially different type of distance estimation. Our data suggest that such an assumption is not tenable. Instead, metric estimation training programs need to account for the fact that different distance estimation tasks require different processes, and those differences can affect how metric estimation training transfers to other distance estimation tasks (Jones et al., 2009). Similarly, metric estimation training programs need to account for the fact that distance estimation in
613
one setting may be different than distance estimation in a different setting (Ferris, 1972, 1973a). For that to happen, we need a better understanding of the different types of distance estimation tasks and settings. Specifically, we need a better understanding of metric distance estimation, including what is learned during metric estimation training. Furthermore, we need a better understanding of operationally relevant distance estimation tasks. For example, if we want metric estimation training to improve a diver’s ability to perform various distance-related tasks underwater, then we must understand the processes that those distance-related tasks require. Likewise, if we want metric estimation training to improve a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle, then we must understand how such judgments are made. Finally, we need a better understanding of the settings in which metric distance estimation normally takes place. Without that knowledge, there is no a priori reason to expect that metric estimation training would improve the performance of those operationally relevant tasks in their natural settings. Furthermore, we need a better understanding of how metric estimation training transfers to other distance estimation tasks, as well as how such training transfers to other settings. For example, we need to know how metric estimation training affects a pilot’s ability to judge whether a helicopter’s blades will clear an obstacle, as well as whether those results depend on when and where the training and transfer scenarios take place. To date, only a few studies have examined the transfer of metric estimation training (Ferris, 1972, 1973a; Jones et al., 2009). Table 4 depicts the essence of those studies, including their training and transfer scenarios as well as the direction of the observed transfer. Inspection of the table suggests that the transfer of metric estimation training to other distance estimation tasks and settings is not straightforward. In summary, the results of the present study supported our argument that the negative transfer across tasks that we reported in Jones et al. (2009) stemmed from cognitive intrusion. Our collective results suggest that trainees should not generalize what they learned during metric
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
614 May 2014 - Human Factors Table 4: Experiments That Investigated Whether Metric Estimation Training Transfers to Other Settings or Tasks Publication
Training
Ferris (1972)
Metric estimation in air Metric estimation in air Metric estimation in moderately turbid water Metric estimation in moderately turbid water Ferris (1973a) Metric estimation in clear water Metric estimation in turbid water Metric estimation Jones, DeLucia, Hall, & Johnson Metric estimation (2009) Metric estimation
Transfer
Direction
Metric estimation in clear water Metric estimation in turbid water Metric estimation in clear water
+ – –
Metric estimation in very turbid water Metric estimation in turbid water Metric estimation in clear water Throwing an object to targets Throwing an object to targets Throwing an object to a metric distance
+ – – – – +
Note. The second entry for Jones et al. (2009) was a replication.
estimation training to distance estimation tasks that are primarily perceptual-motor in nature. Furthermore, our results suggest that researchers must develop a better understanding of distance estimation tasks and settings, as well as how metric estimation training transfers to those tasks and settings. With that knowledge, one could supplement metric estimation training with information about when it would be safe and unsafe to generalize what was learned during metric estimation training to other distance estimation tasks and settings. In addition, one could develop a model of transfer for metric estimation training, which could support educated guesses about transfer to tasks, settings, or combinations thereof that have yet to be studied, as well as guide future research in this area. Key Points •• In prior work, we found that metric estimation training negatively transferred to throwing (Jones, DeLucia, Hall, & Johnson, 2009). •• We explained our results in terms of cognitive intrusion. •• The present study tested that possibility by swapping our original training and transfer tasks. •• Training involving throwing was effective, but did not transfer to verbal metric distance estimation. •• This supports our argument that the negative transfer observed in our earlier studies stemmed
from cognitive intrusion, and has important implications for the design and implementation of distance estimation training.
References Beckett, P. A. (1980). Development of the third component in prism adaptation: Effects of active and passive movement. Journal of Experimental Psychology, 6, 433–444. Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled performance: What governs choking under pressure? Journal of Experimental Psychology: General, 130, 701–725. Beilock, S. L., & Gonso, S. (2008). Putting in the mind versus putting on the green: Expertise, performance time, and the linking of imagery and action. Quarterly Journal of Experimental Psychology, 61, 920–932. Dunlap, W. P., Cortina, J. M., Vaslow, J. B., & Burke, M. J. (1996). Meta-analysis of experiments with matched groups or repeated measures designs. Psychological Methods, 1, 170–177. Dyer, J. L., & Young, K. M. (1998). Night vision goggle research and training issues for ground forces: A literature review (1082). Alexandria, VA: U.S. Army Research Institute. Eby, D. W., & Loomis, J. M. (1987). A study of visually directed throwing in the presence of multiple distance cues. Perception and Psychophysics, 41, 308–312. Ferris, S. H. (1972). Improvement of absolute distance estimation underwater. Perceptual and Motor Skills, 35, 299–305. Ferris, S. H. (1973a). Improving absolute distance estimation in clear and in turbid water. Perceptual and Motor Skills, 36, 771–776. Ferris, S. H. (1973b). Improving distance estimation underwater: Long-term effectiveness of training. Perceptual and Motor Skills, 36, 1089–1090. Foley, J. M. (1980). Binocular distance perception. Psychological Review, 87, 411–434. Gibson, J. J. (1947). Motion picture testing and research (Aviation Psychology Program Research Report No. 7). Washington, DC: Army Air Forces.
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
Transfer of Distance Estimation Training Gibson, J. J. (1950). The perception of the visual world. Boston, MA: Houghton Mifflin. Gibson, J. J. (1986). The ecological approach to visual perception. Hillsdale, NJ: Lawrence Erlbaum. (Original work published 1979). Gibson, E. J., & Bergman, R. (1954). The effect of training on absolute estimation of distance over the ground. Journal of Experimental Psychology, 48, 473–482. Gibson, E. J., Bergman, R., & Purdy, J. (1955). The effect of prior training with a scale of distance on absolute and relative judgments of distance over the ground. Journal of Experimental Psychology, 50, 97–105. Gilinsky, A. S. (1951). Perceived size and distance in visual space. Psychological Review, 58, 460–482. Heft, H. (1993). A methodological note on overestimates of reaching distance: Distinguishing between perceptual and analytical judgments. Ecological Psychology, 5, 255–271. Helmholtz, H., von. (1962). Treatise on physiological optics (Vol. 3, 3rd ed.; J. P. C. Southall, Ed. & Trans.). New York: Dover. (Original work published 1910) Hochberg, J. E. (1978). Perception. Englewood Cliffs, NJ: Prentice Hall. Hu, Y., & Goodale, M. A. (2000). Grasping after a delay shifts sizescaling from absolute to relative metrics. Journal of Cognitive Neuroscience, 12, 856–868. Jones, K. S., DeLucia, P. R., Hall, A. R., & Johnson, B. R. (2009). Can metric estimation training hinder actions involving distance? Human Factors, 51, 419–432. Kimura, D. (1999). Sex and cognition. Cambridge, MA: MIT Press. Kirk, R. E. (1995). Experimental design: Procedures for the behavioral sciences (3rd ed.). Pacific Grove, CA: Brooks/Cole. Loomis, J. M., & Knapp, J. M. (2003). Visual perception of egocentric distance in real and virtual environments. In L. J. Hettinger & M. W. Haas (Eds.), Virtual and adaptive environments (pp. 21–46). Mahwah, NJ: Lawrence Erlbaum. Martin, T. A., Keating, J. G., Goodkin, H. P., Bastian, A. J., & Thach, W. T. (1996). Throwing while looking through prisms: II. Specificity and storage of multiple gaze-throw calibrations. Brain, 119, 119–121. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford, UK: Oxford University Press. Milner, A. D., Paulignan, H. C., Dijkerman, F., & Jeannerod, M. M. (1999). A paradoxical improvement of misreaching in optic ataxia: New evidence for two separate neural systems for visual localization. Proceedings of the Royal Society: Biological Sciences, 266, 2225–2229. Niall, K. K., Reising, J. D., & Martin, E. L. (1999). Distance estimation with night vision goggles: A little feedback goes a long way. Human Factors, 41, 495–506. Ooi, T. L., Wu, B., & He, Z. J. (2001). Distance determined by the angular declination below the horizon. Nature, 414, 197–200. Redding, G. M., Rossetti, Y., & Wallace, B. (2005). Applications of prism adaptation: A tutorial in theory and method. Neuroscience and Biobehavioral Reviews, 29, 431–444. Redding, G. M., & Wallace, B. (1997). Adaptive spatial alignment. Mahwah, NJ: Lawrence Erlbaum. Redding, G. M., & Wallace, B. (2002). Strategic calibration and spatial alignment: A model from prism adaptation. Journal of Motor Behavior, 34, 126–138.
615 Redding, G. M., & Wallace, B. (2006). Generalization of prism adaptation. Journal of Experimental Psychology: Human Perception and Performance, 32, 1006–1022. Reising, M., & Martin, E. L. (1994). Distance estimation training with night vision goggles: A preliminary study (AL/HR-TR1994-0090). Dayton, OH: University of Dayton Research Institute. Reising, M., & Martin, E. L. (1995). Distance estimation training with night vision goggles under low illumination (AL/HR-TR1994-0138). Dayton, OH: University of Dayton Research Institute. Sahm, C. S., Creem-Regehr, S. H., Thompson, W. B., & Willemsen, P. (2005). Throwing vs. walking as indicators of distance perception in real and virtual environment. ACM Transactions on Applied Perception, 2, 1–14. Senter, R. J. (1969). Analysis of data: Introductory statistics for the behavioral sciences. Glenview, IL: Scott Foresman. Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics (4th ed.). Boston, MA: Allyn & Bacon. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Allyn & Bacon. Taieb-Maimon, M. (2007). Learning headway estimation in driving. Human Factors, 49, 734–744. Waller, D. (1999). Factors affecting interobject distance perception in virtual environments. Presence: Teleoperators and Virtual Environments, 8, 657–670. Welch, R. B. (1978). Perceptual modification: Adapting to altered sensory environment. London, UK: Academic Press. Welch, R. B. (1986). Adaptation of space perception. Handbook of Perception and Human Performance, 1, 1–45. Wilkinson, L., & APA Task Force on Statistical Inference. (1999). Statistical methods in psychology journals: Guidelines and explanations. American Psychologist, 54, 594–604. Williams, B., Rasor, T., & Narasimham, G. (2009). Distance perception in virtual environments: A closer look at the horizon and error. In Proceedings of the 6th Symposium on Applied Perception in Graphics and Visualization (pp. 7–10). New York, NY: Association for Computing Machinery. Witmer, B. G., & Kline, P. B. (1998). Judging perceived and traversed distance in virtual environments. Presence: Teleoperators and Virtual Environments, 7, 144–167.
Allyson R. Colombo is a human factors engineer at Basic Commerce and Industries (BCI), Inc. in Dahlgren, Virginia. She earned her PhD in human factors psychology from Texas Tech University in 2011. Keith S. Jones is an associate professor of psychology at Texas Tech University. He earned his PhD in human factors psychology from the University of Cincinnati in 2000. Date received: April 21, 2011 Date accepted: June 17, 2013
Downloaded from hfs.sagepub.com at NORTHWESTERN UNIV LIBRARY on June 4, 2016
