Ill I "L I B Rr A Ri-Yr- OF THE U N'lVLRSITY Of ILLINOIS 629-13 no.\-\0 Return this book on or before the Latest Date stamped below. University of Illinois Library J I i -7 !9S9 OCT -4 I3B8 ^^ lift AUG 2 o 19» 21 ■ 2 OCT 2 1«3 DEC I 4 n r 964 1364 AUG 2* - . H - * won ft/?/? 1 9 AP« MAR 08 1982 2 8 1985 L161— H41 UNIVERSITY no Q OF I LLINOIS *V»* tlOH iV*0» tX II.U** lH 5t»t AERONAUTICS BULLETIN NUMBER 9 UNIVERSITY OF ILLINOIS INSTITUTE OF AVIATION Leslie A. Bryan, Ph.D., LL.B., Director Bernice Schrader, A.M., Editor Return this book on or before the Latest Date stamped below. University of Illinois Library RPR -21955. IULLETII each month by tl ! Li6i^iui i 1912 at t he po University ' ' . ; |4- Z 9° 00' J ;fr- Y 17° 55' X 35° 00' X8' Y4' Z 2' ^30°- • Observer Magnification 0.86 Figure 4. Schematic diagrams of the nine experimental conditions showing the angular extent of horizontal and vertical outside visibility presented in each. windshield through which the periscope was installed caused a consider- able restriction on visibility to the left. However, a pilot flying from the righl seal could see ahead and to the right. Also, since pilots are accustomed to flying from the left scat, this control condition was placed at an additional disadvantage. It was anti< ipated thai the performance observed in this condition would be an underestimate of the level of proficiency that could be expected of the ame pilots flying from the left scat and with completely unrestricted . i ibility. 16 THE EXPERIMENTAL TASK It is generally agreed that the diffieulty of the task to be performed is an important factor that must be considered when evaluating aircraft instruments. This fact was clearly demonstrated by Williams (9) in 1947 and has since been confirmed in a series of experiments by the writer and others (8). In all of these experiments it was observed that when the flight task was too easy, differences in pilot performance were not signifi- cant even though the various instrument displays used were quite dis- similar. However, when the complexity or the difficulty of the flight task was increased, significant differences in pilot performance did appear. A distinction must be made between the complexity and the difficulty of a task even though the two are generally associated. Some relatively complex flight tasks are not particularly difficult to execute, although they may be difficult for the pilot to remember. Conversely, there are many uncomplicated and straightforward flight tasks that are not diffi- cult to remember but are difficult to execute. The former type is better for evaluating instruments that the pilot uses to make decisions as to what to do, such as navigation instruments. The latter type is better for evalu- ating instruments that the pilot uses to execute his decisions. These are known as flight instruments. The periscope could be used as a navigation instrument to show the pilot where he is going and to enable him to make heading decisions. However, its primary purpose is to serve as a flight instrument by which the pilot may control the attitude of his aircraft with respect to the horizon and thereby execute his decisions. The periscope as a pursuit meter A flight instrument is essentially a pursuit meter, and as such it must meet two requirements: It must present some index of desired perform- ance and some index of actual performance. The pursuit task is to align the index of actual performance with the index of desired performance. In most pursuit or tracking tasks the index of desired performance is presented to the operator automatically. In the flight task the index of desired performance is selected by the decision of the pilot. In either case, the index of actual performance is presented automatically. The periscope presents both of these indices. The index of desired performance is the horizon in straight and level flight, or some selected position in relation to the horizon in other cases. The index of actual performance is the position of the crosshairs on the face of the periscope in relation to the horizon. 17 Selection of the task Since the periscope is primarily a flight instrument, it should be tested as such. For this reason, the flight task selected should be easy to memorize but relatively difficult to execute. Such a task has the additional advantage of simplifying the problem of scoring. The task selected was the "Oboe" instrument flight pattern. This pattern includes climbing and gliding flight and turns in both directions. Specifically, the pattern is as follows: Starting from straight and level flight, the pilot makes a 360-degree standard rate (three degrees per second) climbing turn to the right at a vertical speed of 500 feet per minute, thus completing the turn and gaining 1,000 feet in two minutes; this climbing turn is followed immediately by a 360-degree gliding turn of the same rates but in the opposite direction that is completed with a loss of 1,000 feet of altitude in the next two minutes. The complete pattern requires four minutes to perform correctly. For a perfectly executed pattern, the aircraft will be at a particular altitude and have a particular heading at any given moment during the four-minute pattern. Inaccuracies in the performance of the task can thus be measured directly by comparing the momentary readings of the altimeter and directional gyro with these correct values. While the flight attitude of the airplane is not measured directly by this procedure, the attitude of an airplane and its power setting are the only variables under the immediate control of the pilot that determine the movement of the plane through the air, or, technically speaking, the relative wind. In these experimental flights the power settings of the air- plane were controlled in a specified manner by the experimenter, so that the only flight variable under the control of the pilot was the attitude of the airplane. By controlling attitude properly, the pilot could control the altitude and heading of the airplane and thereby make good his desired schedule. Thus it follows that any variations made by the pilot from the correct flight attitudes necessary to accomplish the specified pattern would result in deviations from the correct momentary altitude and heading values. Such deviations could also occur as a result of variables not under the control of the pilot, such as the turbulence of the air, but they should 0< ( u r in a random manner and should not bias the results of the experi- ment. By observing the altimeter and the directional gyro, the pilot could see whether or not he was on schedule and could judge what attitude < hanges should be made. The 'Oboe" instrument flight pattern had been used in a previous ' |" rimenl by the writer (7) and in a series of unpublished experiments 18 by the staff of the University of Illinois' aviation psychology laboratory. In these experiments it proved to be a relatively difficult task to perform accurately. However, pilots could easily memorize the correct altitude and heading values for each 15-second interval during the four-minute trial and did not become confused concerning these indices of desired performance. EXPERIMENTAL DESIGN AND PROCEDURE Balancing of intraserial effects The experiment was designed so that each subject performed the test pattern once under each of the experimental and control conditions. Because there were 1 1 conditions in all, 1 1 pilots were used so that each could perform his 11 trials in a different serial sequence. This was done in an effort to balance any intraserial effects of practice and fatigue. Any multiple of 11 subjects could have been used, but the expense of flying the aircraft prohibited the use of a larger pilot sample. The serial se- quence of the various conditions in which each subject performed the experimental task is shown in Table 2. Table 2. EXPERIMENTAL DESIGN. The serial sequence of the conditions in | which each of eleven subjects performed the experimental task is shown. Each condition appears once and only once for each subject and for each serial position. Serial Position Subject 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 1 0th 1 1 < 1 U H A Z B Y C X P R G 2 A U B H C Z P Y Q X R 3 B A C U P H Q Z R Y X 4 C B P A Q U R H X Z Y 5 P C Q B R A X U Y H 1 6 Q P R C X B Y A Z U f- 7 R Q X P Y C Z B H A l 8 X R Y Q Z P H C U B / 9 Y X Z R H Q U P A C E 10 Z Y H X U R A Q B P C 11 H Z U Y A X B R C Q F 19 Subjects The subjects were experienced instrument pilots, most of them flight instructors from the Institute of Aviation and the aviation psychology laboratory, University of Illinois. The subjects were given an indefinite amount of practice on the task in the various experimental and control conditions before performing their test trials. The purpose of the practice sessions was simply to familiarize the pilots with the particular airplane, the use of the periscope, the experimental procedure, and especially with the test pattern on which they were to be scored. When, in the judgment of the experimenter, a pilot had demonstrated that he had memorized the correct altitudes and headings for each 15-second interval of the four-minute pattern and was familiar with the experimental procedure, practice was discontinued. Most of the pilots received between one and three hours of practice. Each subject performed his 1 1 experimental trials in a single session. Weather Test flights were made only on days when the weather was such that a complete 360-degree horizon was visible. In most cases the test flights had to be made above the cloud or haze level in order to satisfy this requirement. Procedure Each pattern was started with the airplane headed north at an indi- cated altitude of 5,000, 6,000, or 7,000 feet and with an indicated airspeed of 1 20 m.p.h. A pattern was started as the sweep second hand of the clock passed through the 12 o'clock position. The power settings of the airplane were controlled by the experi- menter from the co-pilot's seat. As the pilot started the climbing turn, the experimenter advanced the throttles from the cruising position to a position appropriate for the desired 500 feet per minute rate of climb. I he throttles were left in this position during the first two minutes of the pattern. This position varied somewhat from flight to flight depending upon the altitude at which the pattern was started and the atmospheric < onditions at the time. At the end ol the first two minutes of the pattern, the experimenter reduced the throttle settings to the position appropriate for the desired >00 feel per minute rate of descent. The throttles were left in this position until thr end of the four-minute pattern. This position also varied from ighl to flight. On the average, manifold pressures of 23 in. Hg. and 14 20 in. Hg. were required for the climb and glide respectively. The propellers were set at 2,000 r.p.m. at all times. Before the first trial and at the completion of each subsequent trial, the periscope was set up for the next experimental condition. This was done by the subject under the direction of the experimenter. It included changing the magnification of the periscopic image and/or attaching the appropriate Manila cardboard mask to the face of the periscope as re- quired for the particular condition. The magnification of the image was controlled by adjusting the lens positioning rods by use of the aluminum gages described previously. The position of the mask was controlled by aligning indices drawn on the face of the mask with the horizontal and vertical crosshairs on the face of the screen. These indices bisected the aperture in the mask both horizontally and vertically so that the center of the aperture corresponded to the longitudinal axis of the airplane. For the unrestricted visibility trial (condition U), the subject and the experimenter exchanged seats. Method of scoring The performances of the pilots were scored by the experimenter in the following manner. The altitude and heading of the aircraft at speci- fied times during the pattern were recorded to the nearest ten feet and five degrees. Scores were recorded at ten-second intervals, starting 15 seconds after the beginning and ending 15 seconds before the completion of each half of the pattern. This resulted in distributions of ten scores for each half of the pattern for both altitude and direction. No scores could be taken during the first few seconds or during the time between the climbing and gliding portions of the pattern because the experimenter was occupied adjusting the power settings of the airplane. A copy of the protocol sheet used for recording the scores is shown in Figure 5. In addition to the times at which scores were taken during the four-minute pattern, the protocol sheet shows the correct altitude and heading for each of these intervals. It also provides space on the right side of the sheet for recording the discrepancies between the observed and correct values. Since the pilot's cockpit was surrounded by a hood, the experimenter could not see all of the instruments on the left side of the instrument panel that were used by the subject. By holding the hood slightly away from the panel, the experimenter could see the clock and the directional gyro but not the altimeter. Thus the altitude readings had to be taken from a second altimeter on the co-pilot's side of the panel. 21 360° Standard Climbing and Gliding Turns, 500'/min. Right Climb LeftG fide Right Climb Left Glide Direc- Alti- Direc- Alti- Direc- Alti- Direc- Alti- me tion tude Time tion tude tion tude Errors Errors tion tude Errors Errors :00 North X000' 2:00 North X000' X Y X Y :15 •95 45 75 125 210 :15 315 • 25 285 875 790 •35 105 290 •35 755 710 •45 135 375 •45 ??5 6?5 55 165 460 •55 195 540 •05 195 540 3 05 165 4*0 •1 5 225 625 :15 135 375 •95 255 710 :25 105 290 •35 985 790 •35 75 910 :45 ubje 315 875 •45 45 125 rt Serial Position Sum X Sum Y Sum X Sum Y Condition Sum X 2 Sum Y L> SumX 2 SumY 2 Figure 5. A copy of the protocol sheet used for scoring performance on the experimental task while in flight. The protocol sheet shows the times at which the scores were taken and the correct heading and altitude for each of these times. Space is provided at the right for recording the error scores. A number of possible sources of scoring error were thereby introduced. Iu the first place the experimenter had to view the clock and gyro obliquely through the small space between the hood and the panel. This resulted in a certain amount of parallax. The experimenter had to make allowances for this. Also, the altimeter from which the experimenter cored and the one by which the subject flew had to be carefully syn- < hronized al frequent intervals between trials in order to reduce the dis< repan< ies between them to a minimum. To determine the extent of th< scoring errors from all sources, motion pictures of the pilot's instru- 22 merit panel were taken during a number of pre-test trials. The scores that the experimenter had recorded for these trials were then compared with the photographs and were judged to be acceptably accurate. It is doubtful that more accurate scores could have been obtained by any other practical method. It would not have been possible to photograph the pilot's instrument panel during the test trials with the periscope without increasing the illumination inside the cockpit hood and thereby reducing the relative brightness of the periscopic image. Method of grading the records The altitude and direction readings that were recorded on the pro- tocol sheet during flight were later compared with the corresponding correct values and the discrepancies recorded as error scores in the spaces provided on the right side of the protocol sheet. If an observed altitude was higher than the corresponding correct altitude, the sign of the error was positive; if lower, the sign of the error was negative. If an observed direction value was ahead of the scheduled rate of turn, the error was positive, if behind, negative. ANALYSIS OF EXPERIMENTAL RESULTS Testing for constant errors The results of the experiment were evaluated in the manner described in the Appendix. The first step was to determine whether there were significant constant errors in the group performances in the various experimental and control conditions. This was done by determining fiducial limits for the population means for the various conditions. The sample mean for each condition was based upon the means of the 1 1 subjects for that condition. The subject means, in turn, were based upon the 20 error scores obtained from the per- formance of a single trial. The fiducial limits for the population means were based upon the value of t at the .01 level for N-l or ten degrees of freedom. These tests showed the sample constant errors for the various condi- tions to be of no practical significance and justified the assumption that both direction and altitude errors made on the experimental task were normally distributed about the standard or schedule. The deviations indi- vidual pilots made from the flight schedule usually balanced so that the group tendency coincided with the schedule rather than being consistently fast or consistently slow. 23 Standard deviations as performance scores The second step in the statistical treatment of the results was to com- pute the standard deviations of the error scores for each subject's trial in each of the conditions. Because the constant errors for the climbing and gliding portions of the pattern on individual trials were frequently of the opposite sign, the standard deviations were based upon the average of the variances for the two halves of the pattern. This was done to avoid over- estimating the true variability in performance for any given trial. The formula used to compute these average standard deviations for the two halves of the experimental task was V [SXVN C - (2X,/N,) 2 ] - [ZXVN, - (ZX./N,)'] 2 where N c and N g both equal ten. Since each of the 1 1 subjects performed one trial in each of the 1 1 conditions, distributions of 1 1 standard deviations were obtained for each of the conditions for both altitude and direction errors. These statistics, which describe within trial variability, were then con- sidered as measures of the levels of performance accomplished on the various trials, the lower variabilities representing the better performances. Throughout the subsequent statistical analyses they were treated simply as performance scores. The distributions of altitude performance scores for the 1 1 conditions are presented in Table 3 and those for direction in Table 4. An inspection of Tables 3 and 4 shows that there are differences in the central tendencies of the distributions for the various conditions, both in terms of altitude and direction performance. However, the char- acteristic distribution of standard deviations with ten degrees of freedom is positively skewed, and, in the present case, the variances of the distri- butions are highly correlated with their central tendencies. Thus the central tendencies of the distributions could not be compared by the use "I normal probability statistics. Transformation of the performance scores Tin third step in the analysis of the results was to find an appropriate transformation that would render the performance scores suitable for comparison by normal probability statistics. I he distributions resulting from a number of exponential and recipro- cal transformations were plotted on probability paper. Several of the tran formations resulted in distributions that approached normality, 24 but only one of these also resulted in apparently homogeneous varianees. This was the common logarithm of the standard deviation. For convenience in computation the transformed scores were coded in the following manner: The logarithms of the standard deviations were EXPERIMENTAL CONDITION Subject U A B C P Q R X Y Z H 1 20.73 24.73 27.00 41.18 28.02 27.45 32.40 24.28 25.77 33.45 49.12 2 24.44 30.44 22.58 29.36 39.18 36.33 37.58 23.45 23.07 23.66 55.86- 3 23.12 31.93 28.12 44.52 26.93 29.16 47.11 15.25 39.04 51.27 89.81 4 28.42 19.30 20.06 53.05 31.94 13.10 41.07 22.77 31.03 29.07 44.44 5 24.45 17.06 26.73 46.97 32.30 32.63 27.16 14.76 25.25 25.50 51.30 6 20.65 18.84 25.38 33.42 16.22 24.03 20.58 15.83 22.75 34.15 52.49 7 19.62 38.57 42.71 41.02 32.92 28.29 34.21 27.25 28.37 30.94 49.93 8 17.69 33.18 42.15 62.57 20.15 36.30 38.60 41.56 27.61 37.76 44.55 9 20.04 21.45 29.44 42.49 21.60 36.78 17.92 19.87 26.12 26.20 71.23 10 27.53 22.64 30.90 27.20 16.70 31.06 33.13 19.31 30.17 16.51 54.24 11 29.33 23.63 19.60 27.78 24.96 26.53 31.53 33.99 21.08 30.54 34.17 Table 3. THE DISTRIBUTIONS OF PERFORMANCE SCORES. Standard deviations for alti- tude control by eleven subjects in the nine experimental and two control conditions. EXPERIMENTAL CONDITION Subject U A B C P Q R X Y Z H 1 9.15 6.10 7.23 8.25 4.26 5.21 4.14 4.77 5.41 6.48 21.65 2 5.93 7.79 7.60 12.01 6.49 6.97 6.80 6.21 6.94 3.71 23.48 3 7.75 6.64 11.08 6.40 6.10 9.00 9.94 6.49 7.57 8.22 18.83 4 6.32 4.39 7.82 9.44 6.38 4.26 8.98 3.54 3.76 7.98 12.19 5 4.14 3.84 6.22 6.28 6.67 5.29 3.81 3.55 5.76 6.13 12.19 6 3.35 3.91 3.34 6.37 4.82 5.15 5.10 3.87 2.85 5.93 7.89 7 3.71 3.39 6.72 7.36 4.54 5.49 7.15 5.50 4.03 7.12 13.97" 8 5.53 5.80 6.17 16.07 9.22 8.64 10.69 7.07 5.53 10.28 16.14 9 3.54 5.72 4.56 7.86 7.28 5.83 4.87 6.01 5.57 5.27 19.30 10 4.76 6.20 5.49 3.95 5.35 7.16 6.33 4.42 5.07 5.81 8.72 11 4.49 5.53 4.90 8.62 3.84 4.14 5.70 5.06 3.91 6.64 8.31 Table 4. THE DISTRIBUTIONS OF PERFORMANCE SCORES. Standard deviations for direc- tion control by eleven subjects in the nine experimental and two control conditions. 25 first rounded off to three decimal places and then multiplied by 1,000 to remove the decimal point. The altitude scores were further coded by subtraction of 1,000 to reduce them to a more convenient size. Hereafter these transformed and coded scores will be referred to as "V-scores" since they originated from measures of variability. The distributions of V-scores for altitude and direction performances in each of the nine experimental conditions are presented in Tables 5 and 6, respectively. Testing the transformed distributions The fourth step in the analysis was to test the hypotheses that the transformed scores were normally distributed and that their variances were equal. The latter hypothesis was tested first, since the assumption of homogeneity of variance was required for the technique that was to be used in testing the hypothesis of normality. The distributions for the nine experimental conditions were tested for homogeneity of variance by use of the likelihood ratio criterion as described by Mood (4). The likelihood ratio criterion for testing the null hypothesis Ho *• C 2 i = C 2 2 = • • • ~ T 2 Jfc IS X = ri(-) ! W where Si = 2 ( X H ~ x 2 Si and 1 hua — = c 2 i. For cadi set of data (altitude and direction V-scores) in the present experiment, n 4 = nx = n 2 = ■ ■ • = n fc = 11, and k = 9. It is easily demonstrated that, lor these values of n* and k, log A can be expressed in the following convenient form: log X = 5.5 (log i variance. Before the distributions of transformed scores are used for comparative purposes, they should be tested to determine whether or not they satisfy these requirements. If they do not, a better transforma- tion must be found. Assuming the transformation has been found to be acceptable, the distributions of transformed scores may be compared by the use of sta- tistics that require the assumptions of normality and homogeneity of variance. Such common techniques as the analysis of variance and the /-test for correlated means will usually be adequate. 44 The level of significance that is found for any comparison between eonditions using the transformed seores is applicable to the differences between the original distributions of scores for those conditions. The logical justification for this statement is this: The probabilities that were determined are applicable only to the transformed distributions. However, the following crucial inference may be made: Since the transformed scores bear a definite, though nonlinear, relationship to the original scores, if the difference between the means of the distributions of the former is found to be significant, it may be inferred that the difference between the central tendencies of the corresponding distributions of the latter is equally significant. 45 REFERENCES 1. Chapanis, A., Garner, W. R., and Morgan, C. T. Applied experi- mental psychology: human factors in engineering design. New York: John Wiley & Sons, 1949, 55-57. 2. Johnson, B. E., and Williams, A. C, Jr. Obedience to rotation-indi- cating visual displays as a function of confidence in the displays. Port Washington, New York: U. S. Navy, Special Devices Center, Technical Report SDC 71-16-2, June, 1949. 3. McNemar, Q. Psychological statistics. New York : John Wiley & Sons, 1949, 289-316. 4. Mood, A. M. Theory of statistics. New York: McGraw-Hill, 1950, 269-270. 5. Mueller, C. G. Numerical transformations in the analysis of experi- mental data. Psychol. Bull., 1949, 46, 198-223. 6. Roscoe, S. N. The effects of eliminating binocular and peripheral monocular visual cues upon airplane pilot performance in landing. /. appl. Psychol., 1948, 32, 649-662. 7. Roscoe, S. N., and Williams, A. C., Jr. Pilot performance in instru- ment flight as a function of the extent and distribution of visible horizon. Port Washington, New York: U. S. Navy, Special Devices Center, Technical Report SDC 71-16-3, June, 1949. 8. Roscoe, S. N., Smith, J. F., Johnson, B. E., Dittman, P. E., and Wil- liams, A. C, Jr. Comparative evaluation of pictorial and symbolic VOR navigation displays in the 1-CA-l Link trainer. Washington, D. C: C.A.A., Division of Research, Report No. 92, October, 1950. 9. Williams, A. C, Jr. Evaluation and development of aircraft instru- ment designs. Port Washington, New York: U. S. Navy, Special Devices Center, University of Illinois' Progress Report No. 5, Sep- tember. 1947, 24-26. A 6 THE INSTITUTE OF AVIATION, established in 1945 as the Institute of Aeronautics, is operated as the administrative agency responsible for the fostering and correlation of the educational and research activities related to aviation in all parts of the University. Other functions include aca- demic instruction, flight training, management of the University of Illinois Airport, and aeronautical research. In connection with the latter function, the Institute issues two types of publications . . . first, a group of reports on research results, and second, a series of bulletins on aviation subjects of an extension service nature to the citizens of the State. The following publications have been issued: Bulletin One: Municipal Airport Management, Leslie A. Bryan, 1947. (Out of print) Bulletin Two: Landscape Planting for Airports, Florence B. Robinson, 1948. Bulletin Three : Labor Relations in the Air Transport Industry Under the Amended Railway Labor Act, E. B. McNatt, 1948. Bulletin Four: Airport Zoning, J. Nelson Young, 1948. (Out of print) Bulletin Five : Evaluation of the School Link as an Aid in Primary Flight Instruction, A. C. Williams, Jr., and Ralph E. Flexman, 1949. Bulletin Six : Lightplane Tires on Turf and Concrete, Leslie A. Bryan, 1949. Bulletin Seven: Light Aircraft Operating Costs, Leslie A. Bryan, 1949. Bulletin Eight: Evaluation of the School Link and Special Methods of Instruction in a Ten-Hour Private Pilot Flight-Training Program, Ralph E. Flexman, William G. Matheny, and Edward L. Brown, 1950. Bulletin Nine : Flight by Periscope : I. Performing an Instrument Flight Pattern; the Influence of Screen Size and Image Magni- fication, Stanley N. Roscoe, 1951. Publications of the Institute of Aviation will be sent free of charge upon request. 1 UNIVERSITY OF ILLINOIS-URBANA 3 0112 005630444