Suchergebnisse

Three fallacies about generalization are that so-called basic research is more generalizable than applied research, that general findings are immediately useful for design purposes, and that the use of taxonomies increases the generalizability of human factors studies. Some factors that limit generalizability are the use of unrepresentative subjects, insufficient training subjects receive before measurements are begun, inadequate sampling of tasks and situations, inappropriate selection of dependent variables, long-term changes in the world of work, and artifacts attributable to the measurement process itself In designing a study to predict behavior in a specific application, the guiding principle is similarity. The study should be as similar as possible to the real situation. Two principles should be followed to design studies whose findings can be extrapolated to a wide range of situations: (1) design heterogeneity into the studies and (2) replicate earlier studies with variations in subjects, variables, or procedures.

The aim of this paper is to call to attention a very large and important area of human factors engineering that is almost entirely neglected. This area consists of the language and the words that are attached to the tools, machines, systems, and operations with which human factors engineers are concerned. Examples, illustrations, and data are cited to show that changes in the words used in man-machine systems may produce greater improvements in performance than human engineering changes in the machine itself. Arguments are made that this province—the language and words of machines—is properly the concern of the human factors engineer, and not of the grammarian, linguist, or the communication theorist. The paper concludes with an outline of some of the kinds of work that needs to be done to fill these important gaps in our knowledge and technology.

Recommendations about preferred directions of movement for controls and displays are based almost exclusively on right-handed persons. This experiment has studied some common control-display movement stereotypes to find out if these stereotypes are equally valid for left- and right-handed subjects. Our apparatus presented a scale that could be oriented horizontally or vertically. The numbers on the scale increased either right-to-left or left-to-right for the horizontal scales, or up-to-down or down-to-up for the vertical scales. The linkage between the movement of a control knob and the movement of the hairline on the scale was reversible so that a clockwise rotation of the knob could be made to move the hairline toward either end of the scale. We recorded (1) the direction of the first movement made by the subject on each trial, (2) the total number of reversal movements made on each setting, (3) the time taken to make a setting, and (4) the actual setting made (and so, of course, setting errors). We tested 64 subjects, 32 left-handed and 32 right-handed, each using his preferred and non-preferred hands. Significant differences were found in the time to make settings, the number of reversal movements, and errors in initial direction of movement as functions of handedness and the various apparatus relationships. In addition to overall differences between the performance of the left- and right-handed subjects, the two groups behaved differently on some measures when using their preferred and non-preferred hands.

This experiment tested ten different linkages between four displays and four control buttons on a verticallymounted surface. The displays were arranged in a square and the controls were in a vertical column to the right of the displays. A single panel was used for all tests; different linkages were made by changing electrical connections. Eighty male subjects (eight for each linkage) were tested. The subject's task was to push the appropriate control button as soon as a light appeared in one of the displays. Each subject was given 96 consecutive trials on the linkage to which he was assigned. Dependent measures were: times to first response, times to correct response, and errors. The results show that one of the ten linkages is best in terms of both time and errors. There are also some significant differences between certain other linkages. These findings hold only for those situations in which, for some reason, it is necessary to have the displays arranged in the form of a square and the controls in a vertical column to the right of the displays.

Each of 52 (25 female and 27 male) high school students reproduced from memory 1000 eight-digit numbers after viewing each number for 5 s. Subjects were given unlimited time to reproduce the numbers and were allowed to change their reproductions. The range of errors was very large: from 71 to 2231 out of 8000 digits reproduced by each subject. Every subject showed a serial position effect and almost the same effect—about 70% of subjects made the greatest number of errors at the seventh digit. Female subjects made significantly more errors than did the males. Every subject improved his or her score by making changes. Data are given on the relative difficulty of recalling each of the 10 digits, the 100 doublets (pairs of digits), and the 1000 triplets (sets of three digits).

The purpose of this experiment was to study the readability of dials at different distances when the visual angle subtended by the dials is held constant. Five dials, the sizes and markings of which were proportional to viewing distance, were tested at distances ranging from 14 to 224 in. Each of 20 subjects made 20 readings on each dial. Care was taken to select subjects with excellent uncorrected near and far visual acuity. A focus sign above each dial was used to hold accommodation time constant. Directions to the subject stressed accuracy. Response times, errors of estimation, and questionnaire data were recorded. The results show a significant effect of distance on readability: Dials located at distances greater than 28 inches were read faster than two smaller, closer dials. Although there are no significant differences among the errors made on the five dials, the error data are consistent with the time data.

This experiment tested the effectiveness of sensor lines (heavy lines drawn on a control panel to show the linkages between displays and controls) on simple panels made up of lights (the displays) and keys (the controls). The three independent variables were (a) size of panel (two sizes were tested), (b) compatibility of the display-control linkages (a maximally compatible and a highly incompatible arrangement were used), and (c) the presence or absence of sensor lines. Eight panels were constructed to test all possible combinations of these three independent variables. Eighty male subjects in all (ten for each panel) were used. The subject's task was to push the appropriate key as soon as a light had been turned on. Each subject was given 240 consecutive trials on the panel to which he was assigned. Dependent measures were: time to first response, time to correct response, and errors. The results show that it is more imporant to make the linkages between displays and controls compatible than it is to use sensor lines which show schematically the linkages between displays and controls. Sensor lines appear to have a limited kind of usefulness for panels of the type tested here: The lines improved performance only when the linkages between displays and controls were not compatible.

Fifteen subjects were tested on each of four different wooden models of stoves containing various control-burner arrangements. The experiment measured reaction times and errors made by subjects who matched a control to a given burner as quickly as possible. Subjects were tested for 80 consecutive trials. The data show significant improvement in performance throughout the 80 trials. No significant improvement is evident in the second half of the trials, i.e., for trials 41 through 80. Analysis of the data for trials 41 through 80 shows one control-burner arrangement to be superior to all the others. Another arrangement is significantly worse than the best, but better than the remaining two.

Two-man teams solved credible, "real world" problems for which computer assistance has been or could be useful. Conversations were carried on in one of four modes of communication: (1) typewriting, (2) handwriting, (3) voice, and (4) natural, unrestricted communication. Both experienced and inexperienced typists were tested in the typewriting mode. Performance was assessed on three classes of dependent measures: time to solution, behavioral measures of activity, and linguistic measures. Significant differences among the communication modes were found in each of the three classes of dependent variable. This paper is concerned mainly with the results of the linguistic analyses. Linguistic performance was assessed with 182 measures, most of which turned out to be redundant and some of which were useless or meaningless. Those that remain show that although problems can be solved faster in the oral modes than in the hard-copy modes, the oral modes are characterized by many more messages, sentences, words, and unique words; much higher communication rates; but lower type-token ratios. Although a number of significant problem and job role effects were found, there were relatively few significant interactions of modes with these variables. It appears, therefore, that the mode effects hold for both problems and for both job roles assigned to the subjects.

Two-man teams solved credible, "real-world" problems for which computer assistance has been or could be useful. Conversations were carried on in one of four modes of communication: (1) typewriting, (2) handwriting, (3) voice, and (4) natural, unrestricted communication. Two groups of subjects (experienced and inexperienced typists) were tested in the typewriting mode. Performance was assessed on three classes of dependent measures: time to solution, behavioral measures of activity, and linguistic measures. Significant and meaningful differences among the communication modes were found in each of the three classes of dependent variable. This paper is concerned mainly with the results of the activity analyses. Behavior was recorded in 15 different categories. The analyses of variance yielded 34 statistically significant terms of which 27 were judged to be practically significant as well. When the data were transformed to eliminate heterogeneity, the analyses of variance yielded 35 statistically significant terms of which 26 were judged to be practically significant.