Factors predicting navigational accuracy and decision to land within crosswind limits
Causse, Dehais, and Pastor (20101) looked at relationships between cognitive capabilities, navigational accuracy, and decision-making in a simulator.
They used ordinary least squares regression (Illustration 1) and discriminant analysis (Illustration 2) to predict two outcomes for pilots using a flight simulator: Navigational accuracy (Flight Path Deviation) and decision to land based on crosswind limits.
Illustration 1 shows the mean value of Flight Path Deviation (over the flight from takeoff to Final Approach Fix), and the three significant predictors of that deviation, in decreasing order of their explanatory power within the model (ordered according to size of their standardized effects as shown in Figure 1 of the article).
|Illustration 1: Mean value of Flight Path Deviation and significant factors predicting it|
|Mean value of Flight Path Deviation (angular deviation from nominal flight path)||27.69|
|Significant (p < 0.05) predictors of Flight Path Deviation||Reasoning, total flight experience, updating in working memory|
Illustration 2 shows the percentage of pilots who continued to land despite exceeding crosswind limits, and the three significant predictors of correct crosswind go-around decision, in decreasing order of their explanatory power within the model.
|Illustration 2: Percentage of incorrect crosswind landing decisions and significant factors predicting correct go-around|
|Percentage of incorrect crosswind landing decisions (landing in spite of crosswind over limit)||41.6%|
|Significant (p < 0.05) predictors of correct crosswind go-around decision||Updating in working memory, total flight experience, motor impulsivity|
Other relationships of interest were that increased age correlated with both slower processing speed and reduced updates in working memory, and increased inhibition. Flight Path Deviation was negatively associated with reasoning ability.
One problematic issue in the article is that the multiple correlation coefficient given as .078 is labelled by authors as ‘good’ (p. 226), but is actually extremely low. Presumably this is a typographic error and the authors actually meant to write 0.78. If .078 were in fact correct, then the coefficient of determination shows that this model would explain only some 0.61% of the total variance (basically no better than chance), rendering it useless.
The impulsiveness variable (as measured by the Barratt Impulsiveness Scale) was left out of the analysis of Flight Path Deviation, as the authors considered it unrelated to piloting accuracy. This is debatable, as impulsiveness could conceivably affect the navigation strategy through abandonment of mental calculations or use of inappropriate heuristics.
Exploratory, quantitative research on relationships between pilots’ cognitive abilities and their performance.
24 French-speaking GA pilots, with mean total experience at 1676 hours, all with some PC simulator experience and with flight time in the last two years. All were right-handed, male, and university educated. Airline pilots, experts in logic, and those with sensory deficits, neurological, psychological, or emotional disorders, or under the influence of drugs affecting their central nervous systems were excluded.
No information is provided on how participants were sampled, which is something of an oversight.
Research design is correlational.
Independent (predictor) variables: Psychological test battery
Consisted of various tests to measure cognitive performance. Tests included:
• Target Hitting test (a reaction time and accuracy test)
• 2-Back Test (assesses working memory, matching current shape in sequence to that which appeared 2 cycles previously)
• deductive reasoning tests (a subset of classical logical reasoning tests)
• Computerized Wisconsin Card Sorting Test (sorting cards according to shape, colour and number)
• Spatial Stroop test (measures dissonance between meaning and location of word)
• Barratt Impulsiveness Scale (measures cognitive, motor, and non-planning impulsivity)
Independent (predictor) variables: Other
• Total flight experience
• 45 minute navigation exercise while mentally calculating ground speed. Measured by angular deviation from nominal flight path (Flight Path Deviation – FPD).
• Pilot crosswind landing performance, involving mental calculation of crosswind component. Binary variable, correct or incorrect decision to land vs. divert based on this calculation and the maximum crosswind limit of aircraft.
Pilots were given a navigation exercise on a PC flight simulator, set by flight instructors. Each participant did a training session prior to the study.
Navigation: Each participant was then given instructions, flight plan, and technical info such as aircraft crosswind limit. Participants were given the task of mentally calculating groundspeed, and required to navigate using only magnetic compass for part of the flight. Flights lasted three quarters of an hour, from takeoff to the Final Approach Fix (at which participants decided whether to land or not).
Landing decision: Pilots needed to decide whether crosswind (as given by destination ATIS) was within limits. They used the formula Crosswind = Wind velocity x sin (Angle between runway and wind).
Authors used Statistica 7.1 to perform multivariate regression and discriminant analysis.
Reasonable. However, 24 cases is possibly a rather small number for ordinary least squares regression, although the authors conducted a power calculation and used a sample size sufficient to give a power level of .80 at p = 0.05.
A very large range of pilot experience is also seen, from 57 to 13000 hours, potentially improving generalizability.