Above-Real-Time Training (ARTT) Improves Transfer to a Simulated Flight Control Task

Abstract

Objective: The aim of this study was to measure the effects of above-real-time-training (ARTT) speed and screen resolution on a simulated flight control task.

Background: ARTT has been shown to improve transfer to the criterion task in some military simulation experiments. We tested training speed and screen resolution in a project, sponsored by Defence Research and Development Canada, to develop components for prototype air mission simulators.

Method: For this study, 54 participants used a single-screen PC-based flight simulation program to learn to chase and catch an F-18A fighter jet with another F-18A while controlling the chase aircraft with a throttle and side-stick controller. Screen resolution was varied between participants, and training speed was varied factorially across two sessions within participants. Pretest and posttest trials were at high resolution and criterion (900 knots) speed.

Results: Posttest performance was best with high screen resolution training and when one ARTT training session was followed by a session of criterion speed training.

Conclusion: ARTT followed by criterion training improves performance on a visual-motor coordination task. We think that ARTT influences known facilitators of transfer, including similarity to the criterion task and contextual interference.

Application: Use high-screen resolution, start with ARTT, and finish with criterion speed training when preparing a mission simulation.

Keywords

above-real-time training transfer of training attention cognitive load display resolution

Introduction

Above-real-time training (ARTT) is a training technique in which visual-motor tasks are trained at speeds faster than the criterion task speed (Miller, Stanney, Guckenberger, & Guckenberger, 1997). ARTT has reportedly been used by the National Aeronautics and Space Administration to train pilots and astronauts on advanced aircraft simulators as early as the 1960s (see Guckenberger, Uliano, & Lane, 1993). Although not extensively studied in other domains, ARTT has also been used in military training research, sometimes with military participants. Because it is relevant to understanding how ARTT has been implemented, we provide some detail about these experiments.

First, a representative experiment examined how ARTT influenced learning a tank gunnery task (Bliss, Lampton, & Boldovici, 1992). In this study, college undergraduates were trained and tested on a tank gunnery training simulator that presented a visual simulation of either night (infrared) or daytime gunnery targets. The researchers assigned 5 participants to each of five training conditions. Of these groups, three had all training trials at either 1.0, 1.6, or 2.0 times realistic (criterion) target speed, a fourth group had a random mixture of the three target speeds, and the fifth group had sequential increases of target speeds across the training session. Then each participant completed six criterion trials at the 1.0 criterion-speed condition. Criterion-task hit percentage increased steadily in order at 1.0, 1.6, and 2.0 times realistic target speed; in the sequential condition; and then in the random training condition. Training was completed fastest by the group trained at 2.0 times realistic speed, followed by the group trained at 1.6 times realistic speed, then the random and mixed training groups, and finally, the realistic-speed group (for a full discussion, see Bliss et al., 1992).

In short, sequentially increasing and randomly mixed training—both including ARTT—were better than criterion speed training. Not only did ARTT produce better transfer to criterion than did criterion speed training, but the fixed number of ARTT training trials was completed in less time.

Guckenberger and colleagues (1993) report on a study examining the use of ARTT in simulation and training with Air Force (USAF) pilots. In their study, four groups of six USAF F-16 pilots each completed three sets of 10 training trials in an F-16A avionics simulator. Each set was a separate maneuver problem. All the maneuver problems were included in four criterion speed trials. The groups differed on training trial speed: criterion speed, 1.5 times criterion speed, 2.0 times criterion speed, and mixed speed. A criterion measure was average number of hostile aircraft kills. Criterion performance was worst following criterion speed training and better following any one of the 1.5-times-speed, 2.0-times-speed, or mixed-speed training conditions. For all performance measures in most tasks, and most measures in all tasks, the faster training times led to better criterion trial performance (for a full discussion, see Guckenberger et al., 1993). In short, the data again show that ARTT was both more effective and more efficient than criterion speed training (see also Moroney & Moroney, 1999).

Although these studies provide evidence that ARTT is effective, training that is technically harder than the criterion task is not a new idea. For example, Guckenberger et al. (1993) cite research by Ecklund (1975) showing that basketball shooting transfer improved when training hoops had a smaller diameter than regulation hoops. Indeed, Roman legionnaires are reported to have practiced fighting with swords heavier than their combat swords (Gibbon, 1777, p. 731). Thus, one could argue that ARTT is a modern variant of this centuries-old training method. As such, ARTT can be construed of as one of a repertoire of methods designed to improve transfer to a criterion task, including phased adaptation to task stress (e.g., Keinan, Friedland, & Sarig-Naor, 1990), contextual interference (e.g., Memmert, 2006; Merbah & Meulemans, 2011), part-task training (e.g., Mané, Adams, & Donchin, 1989), control of attention (e.g., Gopher, 1982; Gopher, Weil, & Siegel, 1989; Kramer & Larish, 1995), and overlearning (e.g., Puttemans, Wenderoth, & Swinner, 2005).

The goal of the present study was twofold. First, we set out to replicate the general ARTT methodology to more firmly investigate the place of ARTT among visual-motor training methods. The experimental setting was the development of component technologies for military aviation fast-simulation systems. Second, we wanted to additionally examine the influence of screen resolution on learning in the context of ARTT. Transfer of training experiments, simulated aircraft identification experiments (Covas, Lindholm, & Eldman, 2006), and PC-based flight simulation studies all show that high resolution improves flight simulator training. But screen resolution influences data-processing load, which in turn influences hardware and software efficiency. So we included screen resolution as a factor in our design.

Because we tested people who were not pilots, we started by familiarizing them with our flight-based visual-motor task using part-task training in advance of ARTT training (for a similar approach, see Wightman & Sistrunk, 1987). We then gave participants a criterion task pretest and used the pretest data to control for individual differences. Last, to learn how ARTT interacts with other training speeds, we followed the familiarization training and the pretest with two training sessions with factorially balanced combinations of low, medium, and high (criterion) screen resolution and slow, criterion, and fast (ARTT) simulated aircraft speed. Each training session was followed by a posttest at high resolution and criterion aircraft speed.

Method

Participants

Participants were 54 McGill University undergraduates, graduate students, and technicians, who took part in the experiment in exchange for course credit and cash. The average age of the 27 men and 27 women completing the experiment was 21 (range = 18 to 46, SD = 5). Because of an equipment calibration error, 1 participant was replaced. In addition, 2 quit the task because of motion sickness, 2 dropped out because of unrelated illnesses, and 5 were replaced because they could not learn the task during familiarization.

Display

The display was an IBM ThinkVision L220 20-in. flat panel LED screen, 31 cm high by 41 cm wide, powered by an IBM IntelliStation Z Pro Type 6223 computer. At eye height, 60 cm from the participant, the display was 28° high by 36° wide. Low-resolution (738 × 1,024) pixels were 0.40 × 0.40 mm (2’14” × 2’76” visual angle), medium-resolution (1,024 × 1,200) pixels were 0.34 × 0.29 mm (1’30” × 1’43”), and high-resolution (1,200 x 1,600) pixels were 0.25 × 0.25 mm (1’26” × 1’26”). Human visual acuity at medium visual contrast is about 1’ of arc (Michaels, 1985), so all participants could discriminate one pixel from another at medium contrast. Ambient fluorescent room illumination was 312 lux. The full-color out-of-cockpit view was of a sky containing clouds and a sun hidden behind the clouds, a textured earth, and the target aircraft (Figure 1).

Figure 1.

Screen display used on most familiarization trials and on the pretest, training, and posttest sessions (original in color).

Flight Task

A speeded F-18A model was the target aircraft and another speeded F-18A model was the chase aircraft in a game of “aerial tag.” We used a “quasi-transfer” design (Taylor, Lintern, & Koonce, 2001): The criterion task, also a simulation task, was at the highest display resolution possible with our equipment and a 900-knot simulated airspeed.

The trajectories of target and chase aircraft were counterclockwise orbits (as participants were looking down) around an imaginary fixed vertical. The starting orbit radii, inclinations (relative to an orbit parallel to the ground), and aircraft speeds were software controlled. The orbit radius, inclination, and speed of the chase aircraft were changed in flight by the participant with use of a CH Products ProThrottle (left hand) to control acceleration and a CH Products 568 CombatStick (right hand) to control orbit radius and inclination. The throttle was a second-order controller. Its neutral position maintained the current orbit velocity, pushing the throttle forward increased linear acceleration, and pulling it back decreased linear acceleration. The x- and y-axes of the zero-order side stick controller controlled orbit radius and inclination. Pushing the side stick to the left decreased the radius; pushing it to the right increased it. Pushing the side stick forward tilted the orbit down relative to its current inclination; pulling it back tilted the orbit up.

Resolution was either low (738 × 1,024 pixels), medium (1,024 × 1,280 pixels), or high (1,200 × 1,600 pixels, criterion). Training speed was either slow (450 knots), criterion (900 knots), or fast (ARTT, 1350 knots). Pretest and posttest tasks were at high resolution and criterion speed.

Procedure

Participants completed three sessions lasting 45 to 90 min on successive days. During the first session, participants provided relevant personal information, gave informed consent, and completed 36 familiarization trials and 40 pretest trials. The second session included 50 training trials followed by 20 posttest trials. The third session included 50 more training trials followed by 20 posttest trials; then participants were debriefed, thanked, and paid. Minimum intertrial interval was a few seconds. Maximum intervals were participant controlled because trials began when a “Start next trial” button was clicked on the display. Participants received $30.00 or $10.00 per session if they did not complete the experiment.

Familiarization trial order was determined by four successive nine-trial Latin square combinations of low, medium, and high resolution with low, medium, and high speed. Training screen resolution (low, medium, or high) was a between-subjects factor. Training speed across two training sessions varied factorially within participants. Labeling the speeds S (slow), C (criterion), and F (fast), we presented the nine speed combinations in Sessions 2 and 3 as follows: S-S, S-C, S-F, C-S, C-C, C-F, F-S, F-C, F-F.

In Session 2, the 18 participants to be trained at each screen resolution were divided into three training-speed groups (slow, criterion, and fast), producing nine Training Resolution × Training Speed groups with 3 men and 3 women in each group. In Session 3, the Resolution × Speed groups were further subdivided, producing a total of 27 groups: Training Resolution (three levels) × First Training Speed (three levels) × Second Training Speed (three levels) with 1 man and 1 woman in each group.

Session 1: Familiarization and Pretest

Participants were told that they would play a game of aerial tag in which the goal was to intercept the target aircraft in their chase aircraft and that a successful intercept would freeze the display and end the trial. On Trials 1 through 9, participants adjusted only the throttle to approach and pace the target aircraft. A debriefing display showed a dynamic, schematic three-dimensional recreation of the chase and target aircraft flight paths for the two trials that gave each participant the greatest trouble (Figure 2).

Figure 2.

Screen display including a 3-D representation of the target and chase aircraft trajectory, used on some familiarization trials (original in color).

On Trials 10 through 18, the chase aircraft speed was set slightly faster than the target, and the side stick controller was used to change the chase aircraft orbit so as to intercept the target aircraft. The debriefing display again presented dynamic three-dimensional recreations of two difficult trials.

On Trials 19 to 27, the chase aircraft speed was set slightly faster than the target, and the target and chase aircraft start orbits had the same radius and differed only in inclination from the vertical. The side stick controller was used to vary the chase aircraft orbit inclination. Participants did not touch the throttle and did not move the side stick left or right. Debriefing again included three-dimensional recreations of two trials.

On Trials 28 to 36, the throttle and side stick were used simultaneously to control all three degrees of freedom (linear acceleration, orbit radius, and orbit inclination) of the chase aircraft. Debriefing again included three-dimensional recreations of two trials.

Participants then completed 40 pretest trials at high resolution (1,200 × 1,600 pixels) and aircraft speed of 900 knots (criterion). Each trial started with the target aircraft in sight. The chase aircraft speed, orbit radius, and inclination differed randomly from the target speed, radius, and inclination. Each trial ended with a successful intercept or after 2 min. Participants were then thanked and the next day’s appointment confirmed.

Session 2: First Training and Posttest

The training condition (the resolution– aircraft speed combination for Training Sessions 2 and 3) was chosen on the basis of the next unfilled spot in the experimental protocol for a man or woman. Participants were told that the training trials were like the previous day’s last set of 9 familiarization trials. Each trial began with the target aircraft in sight of the chase aircraft but with a different, randomly chosen speed, orbit radius, and inclination. Each trial ended with a successful intercept or after 2 min. After the 50 training trials, participants completed 20 posttest trials at high resolution and criterion aircraft speed (900 knots).

Session 3: Second Training and Posttest

Session 3 was like Session 2. The training trial speed and resolution depended on the predetermined condition that was completed by the participants. After the posttest trials, participants were debriefed, thanked, and paid.

Results

Performance Measures

The x, y, and z coordinate positions of the target and the chase aircraft were recorded every 0.1 s. The data were converted offline into the cumulated (every 0.1 s) performance measures of distance in kilometers between the target and chase aircraft across the entire trial and the total time in seconds taken to complete each trial. Data from the familiarization trials were not analyzed.

The performance measure distributions were positively skewed. The raw measures were transformed to common logarithms, and analyses were carried out on the log-transformed data. Results are reported as geometric mean values (the antilogs of the means of the log-transformed data). The mean performance score from the pretest trials on Day 1 was used as the covariate for the training and posttest trial data collected on Days 2 and 3.

Log time and log distance were correlated within conditions across observers (pretest, r = .92; first training, r = .79; first posttest, r = .75; second training, r = .83; second posttest, r = .90). Measure (time or distance) was a significant main effect in all sessions but interacted significantly only with first training session speed to influence first training session performance. Otherwise, distance and time were just differently scaled measures of the same performance. The log time and distance measures were averaged to reduce residual variance attributed to measurement error.

Two-min (timed-out) trials

The median number of timed-out trials (without an intercept) per participant was 7 on the 40 pretest trials (mode = 3, range = 0 to 34). The Pearson r between the number of timed-out trials per participant and the total time per pretest session was 0.80 (p < .04). On both the first and the second posttest, the median number of timed-out trials was 0 (range = 0 to 2), and they were not significantly correlated with total time to completion on either posttest (r = .39 on both posttests).

First Training and Posttest

Multivariate analysis of covariance of the common logs of the first training and first posttest scores was used to evaluate the significance of the covariate (average pretest performance), training trial resolution, first training session speed, time or distance performance measure, and their interactions. The factors and interactions accounted for more than 99% of the variance for both performance measures.

First training

The individual difference covariate was significant, F(1, 45) = 7.07, p < .01. Display resolution was significant, F(2, 45) = 7.96, p < .01. The medium-resolution display produced a significantly larger (worse) average training score (208.9) than did either the low- (166.0) or high-resolution (169.8) display, which did not differ significantly. The training speed factor was significant, F(2, 45) = 12.25, p < .0001. Slow training was best (157.8), followed by criterion (177.4) and then fast training (213.3). The faster the training speed, the harder the training task. The Training Speed × Performance measure interaction was significant, F(2, 45) = 3.54, p < .05, because the significant increase in distance traveled to interception from slow to criterion to fast training was greater than the nonsignificant but consistent increase in time to completion. There were no other significant main effects or interactions.

First posttest

The individual difference covariate was significant, F(1, 45) = 8.6, p < .01. Mean performance measures were significantly different, F(1, 45) = 10.5, p < .01: Geometric mean distance = 451.8 km; geometric mean time = 34.9 s. The display resolution factor was significant, F(2, 45) = 6.5, p < .01. Medium training resolution performance on the first posttest (131.6) was slightly worse than low- (122.7) or high-resolution (123.02) performance. The first training session speed factor was also significant. Criterion training speed produced worse performance (132.83) than either fast (121.81) or slow (122.91) first-session training speed. There were no significant interactions among the main effect factors. The worst overall first-posttest performance followed criterion speed training at medium resolution. The averaged geometric mean performance values for the significant main effects of screen resolution and training speed on the first training and the first posttest sessions are presented in Table 1.

Table 1:

Significant Main Effects of Training Resolution and Training Speed on First Training Session and First Posttest Session Average Geometric Mean Performance

Training Resolution (pixels)			Training Speed (knots)
Resolution	Training	Posttest^a	Speed	Training	Posttest^a
738 × 1,024	166.0	122.7	450 (slow)	157.8	122.9
1,024 × 1,280	208.9	131.6	900 (criterion)	177.4	132.8
1,200 × 1,600	169.8	123.0	1,350 (ARTT)	213.3	121.8

Note. ARTT = above-real-time training.

Posttest at 1,200 × 1,600 pixels and 900 knots.

Second Training and Posttest

Multivariate analysis of covariance was again used to evaluate the significance of the score differences related to the individual difference covariate, the performance measures, resolution, first and second training speed, and their interactions. The factors and interactions accounted for more than 99% of the variance for both measures.

Second training

The individual difference covariate was not significant, F(1, 27) = 1.50. Resolution (p < .0001), performance measure (p < .0001), and second training session speed main effects (p < .0003) were all significant. High resolution (101.20) produced significantly better performance than low resolution (119.81), which produced significantly better performance than medium resolution (144.7). Criterion speed training (110.4) was better than slow (120.8) or fast (131.8) training. Training resolution significantly interacted with first and second training speed, F(8, 27) = 9.16, p < .0001, in a complex pattern that defies meaningful interpretation. First-session training speed had no significant main effect on second-session training performance, and no other interactions were significant.

Second posttest

The individual difference covariate was marginally significant, F(1, 26) = 3.61, p < .07. All main effects were significant at least at p < .05, and no interactions with performance measure were even close to significant. On the second posttest, time and distance were alternate measures of the same performance. All of the interactions not including performance measure were significant at p < .0001 except Resolution × Second Training Speed, F(4, 26) = 2.99, p < .04. The three-way interaction of Resolution × First Training Speed × Second Training Speed was significant, F(8, 26) = 6.09, p < .0002. It was independent (with Type III sums of squares) of main effects and the other interactions.

High-resolution training produced significantly better second-posttest performance (90.08) than did either medium- (95.41) or low-resolution training (109.56), F(2, 26) = 7.08, p < .005. Fast (ARTT) first-training-session speed produced significantly better performance (90.79) than either slow (97.88) or criterion-speed (109.3) training, F(2, 26) = 7.09, p < .005. Criterion second-training-session speed produced better performance (84.50) than did either fast (ARTT, 98.00) or slow (117.31) second-training speed, F(2, 26) = 28.25, p < .0001.

The significant interaction between first and second training session speed on performance on the second posttest is shown in Table 2. The pairing of fast (ARTT) first training speed with criterion second training speed produced the best second-posttest performance. Next best was criterion speed on both training sessions, followed by fast (ARTT) speed on both training sessions. Worst was criterion speed on first training followed by slow speed on second training (143.72), F(4, 26) = 11.84, p < .0001. The significant three-way interaction among resolution, first training speed, and second training speed meant that at high and medium resolution, ARTT first-training speed and criterion second training speed produced best second-posttest performance (62.07 with high resolution, 60.82 with medium resolution), whereas at low resolution, ARTT during both first and second training sessions was best (92.95). At high and medium resolution, criterion first training followed by slow second training was worst (190.71 with high resolution, 154.74 with medium resolution), whereas at low resolution, fast (ARTT) first training followed by slow second training was worst (161.78), F(8, 26) = 6.5, p < 0001. The second posttest trials, like the first posttests, were carried out at high resolution and 900 knots (criterion) speed.

Table 2:

Effects of First- and Second-Training-Session Speed on Geometric Mean Second-Posttest Performance

	Second Training Speed (knots)
First Training Speed (knots)	450 (Slow)	900 (Criterion)	1,350 (ARTT)	Mean
450 (slow)	95.39	105.86	92.87	97.88
900 (criterion)	143.72	79.09	114.97	109.33
1,350 (ARTT)	117.76	72.08	88.17	90.79
Mean	117.31	84.50	98.00

Note. ARTT = above-real-time training.

Although we did not plan to study gender differences, we report that the covariance-corrected women’s performance score on the first posttest was 134.90; men’s was 158.40. Women’s second posttest score was 112.20; men’s was 128.82. Both women’s means were significantly lower (better) than the men’s (t tests, p ≤ .0001). We accepted anyone who volunteered, there were dropouts, and the principal experimenter was a woman, so there were uncontrolled selection and motivation factors, but we note the difference in favor of women as a matter of interest (e.g., Spence, Yu, Feng, & Marshman, 2009).

Summary

Training resolution, a between-subjects variable, had a small but significant effect on first-posttest performance (at high resolution and criterion speed), but second-posttest performance was worse following low- or medium-resolution training. Reducing the pixel count on the training display by 22% from 1,200 × 1,600 (criterion) to 1,024 × 1,280 (medium resolution) reduced second-posttest performance by 6%. Reducing the pixel count by 60% to 780 × 1,024 (low resolution) reduced performance by 22%.

First training and posttest

The faster the training speed, the longer the average distance and time taken per first training session trial. On the first posttest, medium-resolution training was worse than low- or high-resolution training, and criterion-speed training was worse than slow- or fast-speed training.

Second training and posttest

Following high- and medium-resolution training, second posttest performance was best when the first training session was at ARTT and the second training session was at criterion speed. Following low-resolution training, second-posttest performance was best when ARTT was presented on both first and second training sessions. Low-resolution training also produced the worst overall second-posttest results.

Discussion

In this study, we examined how simulation and training involving an ARTT-based methodology influences learning and performance. We additionally assessed the potential impact of screen resolution on training effectiveness. Our goal was to determine how ARTT is related to visual-motor training methods. In this final section, we summarize our findings and discuss them in the context of related research.

First, either slow training or ARTT on the first training session delivered better first- posttest performance than did criterion-speed training, independent of training resolution. Second-posttest results showed that a first training session at ARTT, followed by a second training session at criterion speed, delivered the best performance averaged across all training conditions. An addendum demanded by the significant interaction among first and second training session speed and training resolution is that following low-resolution training (a between-subjects factor), second test session performance (test sessions were always at high resolution) was best when both first and second low-resolution training sessions were at ARTT. In our simulation environment, some ARTT training improved criterion performance at all training resolutions.

Kolf (1973), cited by Ali, Guckenberger, Rossi, and Williams (2000), wrote that “regardless of type or amount of pre-flight simulator training accomplished by the pilot, the actual flight appears to take place at a much faster time frame than real time.” If real-time simulation seems slower to experienced pilots than does the actual flight experience, Kolf reasoned that speeding up the simulation should increase the apparent similarity between the simulator and the flight experience, possibly leading to better transfer. Guckenberger et al. (1993) also based themselves on Kolf when they wrote that programming a flight simulator in “fast time” would give test pilots a more accurate appreciation of the real-world stress that they would feel when flying the test aircraft—and by implication, that training in more stressful conditions would produce better transfer to the criterion flight task.

We assume that ARTT increases the effort required to complete each training trial and therefore focuses attention more effectively on the faster trials. Varying the trial speed across two sessions, as happened in many of our conditions, may also have increased the attention paid to the second training session by varying the training context (contextual interference; Merbah & Meulemans, 2011; Seidler, 2004). Increasing the effort required on some trials and increasing attention paid to the varied conditions of a second training task may both work to increase the amount learned during training and therefore contribute to greater transfer.

Our second-posttest results as well as those of Bliss et al. (1992) and Guckenberger et al. (1993) show that mixing ARTT and criterion speed trials, either randomly or in blocks, produced transfer as good as did all-ARTT training, although the all-ARTT training necessarily took less time. We leave unexplained the fact that second-posttest performance following low-resolution training was better when both training sessions were carried out at ARTT.

Our inexperienced university-based participants had undergone pretraining in the form of 36 familiarization trials and 40 pretest trials on Session 1 before being exposed to the 100 factorially balanced training trials and the 40 posttest trials on Sessions 2 and 3. Thus the context of our study of ARTT and resolution was with participants who began with an identical part-task learning experience and who had also been pretested to provide a benchmark against which further progress, influenced by later variations in training speed and screen resolution, could be evaluated.

Conclusion

Visual-motor skills simulation training with a quasi-transfer design should employ high-resolution displays, and high-resolution ARTT training trials should be followed by high-resolution criterion speed training trials to maximize transfer to the criterion task. We think that our results are based on two previously recognized causes: (a) the increase in the subjective speed (and difficulty) of the criterion transfer task, which is approximated by an increase in the actual speed (through ARTT) of the training task, thus improving transfer to the criterion task, and (b) a contextual interference transfer advantage (Merbah & Meulemans, 2011; Seidler, 2004) gained by varying the training tasks between ARTT and criterion speed.

Key Points

Practicing a computer-based, flight-related visual-motor skill at faster than criterion speed (above-real-time training [ARTT]) improves transfer to the computer-based criterion task.

With high- and medium-resolution displays, the benefits of ARTT increase when ARTT trials are followed by training trials at criterion speed. With low-resolution displays, continuous ARTT training produces better transfer to the criterion task.

We propose that ARTT increases training speed to match the subjective speed of the criterion task, thus improving transfer, and that varying ARTT with criterion-speed training produces contextual interference, which also improves transfer.

Footnotes

Acknowledgements

Human Factors North completed this research for Defence Research and Development Canada (Contract No. W7711-057961A). We thank Ray Obidowski of Array Systems Computing, who developed and installed the flight simulation software. The article was written while Donderi was on sabbatical leave at the San Francisco State University (SFSU) Psychology Department, and he thanks Kathy Mosier, who was chair at the time, for SFSU’s hospitality.

D.C. Donderi is a principal consultant with Human Factors North in Toronto, Ontario, Canada. He received a PhD in experimental psychology from Cornell University in 1963.

Keith K. Niall is first secretary, CDLS(W), at the Canadian Embassy in Washington, D.C. He received a PhD in experimental psychology from McGill University in 1987.

Karyn Fish is a student in the School of Communication Sciences and Disorders at McGill University in Montreal, Quebec, Canada. She received a BSc from McGill University in 2007.

Benjamin Goldstein is a student in the Faculty of Medicine at McGill University in Montreal, Quebec, Canada. He received a BSc from McGill University in 2008.

References

Ali

S. F.

Guckenberger

Rossi

Williams

(2000). Evaluation of above-real-time training and self-instructional strategies for airmanship tasks on a flight simulator (Report No AFRL-HE-AZ-TR-2000-0112). Mesa, AZ: U.S. Air Force Research Laboratory, Human Effectiveness Directorate.

Bliss

J. P.

Lampton

D. R.

Boldovici

J. A.

(1992). The effects of easy-to-difficult, difficult-only, and mixed-gunnery practice on performance of simulated gunnery tasks (Tech. Rep. 948). Alexandria, VA: Army Research Institute for the Behavioral Sciences. (Report ADA251866, National Technical Information Service)

Covas

C. M.

Lindholm

J. M.

Eldman

C. A.

(2006). P-28: Effects of image generation variables on aircraft-roll detection range. In Society for Information Display Digest (pp. 299–302). Campbell, CA: Society for Information Display.

Gibbon

(1777). The history of the decline and fall of the Roman Empire (Vol. 1). Retrieved from http://www.gutenberg.org/etext/731

Gopher

(1982). A selective attention test as a predictor of success in flight training. Human Factors, 24, 173–183.

Gopher

Weil

Siegel

(1989). Practice under changing priorities: An approach to the training of complex skills. Acta Psychologica, 71, 147–177.

Guckenberger

Uliano

K. C.

Lane

N. E.

(1993). Teaching high-performance skills using above-real-time training (NASA Contractor Report 4528). Washington, DC: National Aeronautics and Space Administration.

Keinan

Friedland

Sarig-Naor

(1990). Training for task performance under stress: The effectiveness of phased training methods. Journal of Applied Social Psychology, 20, 1514–1529.

Kramer

A. F.

Larish

J. F.

Strayer

D. L.

(1995). Training for attentional control in dual task settings: A comparison of young and old adults. Journal of Experimental Psychology: Applied, 1(1), 50–76.

10.

Mané

A. M.

Adams

J. A.

Donchin

(1989). Adptive and part-whole training in the acquisition of a complex perceptual-motor skill. Acta Psychologica, 71, 179–196.

11.

Memmert

(2006). Long-term effects of type of practice on the learning and transfer of a complex motor skill. Perceptual and Motor Skills, 103, 912–916.

12.

Merbah

Meulemans

(2011). Learning a motor skill: Effects of blocked versus random practice. A review. Psychologica Belgica, 51, 15–48.

13.

Michaels

D. D.

(1985). Visual optics and refraction: A clinical approach (pp. 258–276). St. Louis, MO: Mosby.

14.

Miller

Stanney

Guckenberger

(1997). Above real-time training. Ergonomics in Design, 5(3), 21–24.

15.

Moroney

W. F.

Moroney

B. W.

(1999). Flight simulation. In Garland

D. J.

Wise

J. A.

Hopkin

V. D.

(Eds.), Handbook of aviation human factors (pp. 355–388). Mahwah, NJ: Lawrence Erlbaum.

16.

Puttemans

Wenderoth

Swinner

(2005). Changes in brain activation during the acquisition of a multi-frequency bimanual coordination task: From the cognitive stage to advanced levels of automaticity. Journal of Neuroscience, 25, 4270–4278.

17.

Seidler

R. D.

(2004). Multiple motor learning experiences enhance motor adaptability. Journal of Cognitive Neuroscience, 16, 65–73.

18.

Spence

Feng

Marshman

(2009). Women match men when learning a spatial skill. Journal of Experimental Psychology: Learning, Memory and Cognition, 35, 1097–1103.

19.

Taylor

H. L.

Lintern

Koonce

(2001). Quasi-transfer as a predictor of transfer from simulator to airplane. Journal of General Psychology, 120, 257–276.

20.

Wightman

D. C.

Sistrunk

(1987). Part-task training strategies in simulated carrier landing final-approach training. Human Factors, 29, 245–254.