- Email: [email protected]
Legibility of symbols on CRT displays
Applied Ergonomics 1971,2.3,130-132
Legibility of symbols on CRT displays Allen G. Vartabedian American Telephone & Telegraph Company (This work was undertaken while the author was at Bell. Telephone Laboratories, Holmdel, New Jersey,
The effects of several parameters - such as letter orientation, slanting versus upright, dot matrix size, etc - of symbol formation on the legibility of CRT displays were studied. Symbols were tested by subjects attempting to identify them when presented briefly on the display screen. Speed and accuracy of identification were used as the measures of legibility. The 7 x 9 dot matrix symbols drawn with circular dots were superior to all others for both reaction time and errors. Slanting had a detrimental effect on dot and stroke symbols and circular dot symbols were superior to elongated dot symbols.
A major question in considering cathode ray tube (CRT) displays is how the CRT, with its wide range of characteristics, will affect human performance. Of basic importance to this question are the characteristics of displayed symbols, both from the standpoint of performance and the cost of the resulting device. This paper considers how various methods for generating symbols are related to legibility. Legibility, in this case, is determined by the subject's ability to identify a symbol, shown briefly, to the exclusion of all other symbols of a defined set. Seyeral commercial methods are available for generating symbols, among which are the dot matrix and stroke techniques. The dot matrix method is what its name implies: each character is formed of a series of dots in a matrix of possible dot positions. Symbols are drawn by blanking and unblanvdng the electron beam according to a pattern of light points within the fixed matrix. In the stroke method, by comparison, the unblanked electron beam is moved like a pen to trace out the shape of a given symbol. The end points of segments of the symbol fall upon fixed points within a matrix. While it was felt by many that the stroke-generated method might result in more effective displays, the dot matrix method is both more economical and more versatile, since it can be applied to other display devices in addition to CRTs - for example, displays using gas-discharge devices, light-emitting diodes, electroluminescent materials and fibre optics. The parameters under study were: 1. Symbol generation method, ie symbols drawn with dots versus symbols drawn with continuous strokes. 2. Dot matrix size: the generally encountered 5 x 7 matrix size versus the 7 x 9. 3. Symbol orientation: symbols that are upright versus symbols slanted 20 ° from the vertical. 4. Dot geometry: circular dots versus'vertically elongated dots (these tended to "bleed" together into vertical lines).
130
AppliedErgonomics September1971
These parameters were studied in six experimental conditions on two display devices (Fig 1). The dot matrix symbols used in this study were designed by the author for optimal legibility. Both.5 x 7 and 7 x 9 symbol fonts were stylistically similar within the constraints imposed by the two different matrix sizes. The stroke font was based on the Leroy font which is of known good legibility. The letter 'o' of the stroke font was modified to conform to the dot matrix font by the addition of a loop at the top. In this experiment there is inevitably some confounding of the effects of symbol generation method and symbol design. The various generation methods impose restrictions on character shaping which ultimately constrain symbol design to some extent. The design of both 5 x 7 and 7 x 9 dot symbols followed the same design criteria but resulted in somewhat differently styled symbols in some cases. This is true especially for the symbol '8'. The stroke symbols were designed according to an entirely different criterion by a different designer. Again certain striking differences resulted in symbol designs especially with the letter 'Q'.
Method The experimental task required each of the 23 subjects to depress a button which caused a single symbol to appear briefly on a CRT. The subject then verbally identified the symbol shown and continued in the same manner for all the trials of the experiment. Each subject experienced all six conditions of the experiment. In each of the conditions the subject saw 5 blocks of the 36 capital letters and numbers. Each block was a random permutation of the 36 symbols. There were 180 trials per condition for each subject. Two display consoles employing P31 and P4 phosphors were used as the experimental apparatus. Four conditions, namely the stroke, stroke slant, 7 × 9 (circle dots) and 5 x 7 (circle dots), were conducted using the P31 phosphor. The remaining two conditions, the 7 x 9 (elongated dots) and
7 x 9 slant (elongated dots), were conducted using the P4 phosphor. Both of these phosphors emit energy in a range near the centre of the visible spectrum. The duration of the image was controlled by the decay of phosphorescence of the CRT. This decay is exponential and, for these phosphors, drops to ten percent of its initial brightness in 1 to 2 milliseconds. Thus, with this method the duration of the CRT image is governed by the type of CRT phosphor and has a very small variability. Symbols were displayed in the centre of the CRT with a symbol spot brightness of 37"69 cd/m 2 (11 ft-L) and a background brightness of 6"8 cd/m 2 (2"0 ft-L). The distance between the subject's eyes and the display was fixed at 711 mm (28 in) by means of a head rest. Symbols were 3"55 mm (0.14 in) in height and subtended a visual angle of 17-2 minutes at the observer's eye. The width/ height ratio of a nominal width symbol was set at 0-75. The time from the display of the symbol to the subject's response was recorded by a PDP-9 computer control program by means of a voice switch. Errors in the subject's response were recorded by the experimenter. Results Table 1 presents the overall error rates and reaction times for each of the six experimental conditions. Error data are available for 23 subjects and response time data for 12 subjects. This table shows that the 7 x 9 circle dot font yielded the lowest average reaction time and error rate.
A three factor analysis of variance (generation method x symbols x subjects) was conducted on the reaction time scores. It was found that the generation method had a statistically significant effect (p < 0-005). Symbols had a significant effect (p < 0.001). Also the method - with symbols interaction was significant (p < 0"001). With respect to errors, a two factor (methods x subjects) analysis of variance was conducted after pooling the errors across all symbols. It was found that generation method had a significant effect (p < 0"001). Several comparisons among the six experimental conditions were available to ascertain the effects of the four parameters outlined in the introduction. These comparisons are presented in Table 2.To judge the statistical significance of the differences, the 99 "9 percent confidence interval was calculated from the analysis of variance to be 47"7 ms. Table 2 shows that all but two of the observed
Table 1 Average reaction time and error rates for the six conditions. Condition
Reaction time (ms)
% Errors
5 x 7 circle dot 7 x 9 circle dot
697 602
2"9 2"6
stroke upright
659
4" 5
stroke slant 7 x 9 elongated dot 7 x 9 slant elongated dot
677 678 814
6"7 3"2 7"7
Fig 1 Samples of the six conditions of the experiment. From top to bottom, 5 x 7 circle dot, 7 x 9 circle dot, stroke upright, stroke slant, 7 x 9 elongated dot and 7 x 9 slant elongated dot. In the experiment, symbols were of equal height and equal brightness. Also the last two conditions were not blurred as in the photograph.
Applied Ergonomics September 1971
131
differences are greater than 47'7 ms. A similar confidence interval for error rates requires a difference greater than 0"2% to assert that the error rates are significantly different. All differences in error rates are greater than O'2%. To ascertain the effect of the first parameter, generation method, we examine comparisons '1' and '2' which test the stroke upright font against both the 7 x 9 and 5 x 7 circle dot fonts. In both instances, differences in error rates are in favour of the dot fonts. Differences in reaction times are in favour of the 7 x 9 dot matrix size. For the third parameter, symbol orientation, we examine comparisons '4' and '5'. With respect to errors both comparisons are in favour of the upright orientation with the difference for the stroke font being half that of the dot font. For stroke symbols the observed difference of 18 ms is well below 47"7 ms in reaction time. For 7 x 9 elongated dot symbols, however, the difference is 136 ms the largest difference detected in any comparison. The interaction of the slanting and stroke versus 7 x 9 elongated dot parameters can be noted. The final parameter under question is dot geometry, examined in comparison '6'. Here we note the differences in error rates and reaction times are both in favour of the circular dots for the 7 x 9 matrix symbols.
Table 2 Selected comparisons among the six methods of the experiment (differences are in favour of the top method for each comparison). Difference Comparison (1) Stroke upright vs 5 x 7 circle dot (2) 7 x 9 circle dot vs stroke upright (3) 7 x 9 circle dot vs 5 x 7 circle dot (4) Stroke upright vs stroke slant (5) 7 x 9 elongated dot upright vs 7 x 9 elongated dot slant (6) 7 x 9 circle dot vs 7 x 9 elongated dot
1. The 7 x 9 circle dot font was superior in reaction time and error measures to all other fonts. 2. Dot elongation adversely affected legibility. 3. The 7 x 9 dot matrix font was superior to the 5 x 7 dot matrix font. 4. Dot matrix symbol generation was supeTior in legibility to stroke symbol generation.
132
Applied Ergonomics September 1971
Error rate
38
- 1 "6%
57
1 "9%
95
0-3%
18
2-2%
136
4"5%
76
0"6%
5. The slanting of symbols adversely affected legibility with 7 x 9 elongated dot symbols suffering greater degradation than stroke symbols.
Summary and conclusions The results of the experiment are summarised as follows:
Reaction time (ms)
Based on the results of this experiment it is concluded that, of the three methods of symbol generation tested, the 7 x 9 dot font is superior in legibility to either the 5 x 7 dot font or stroke font. The use of slanted symbols in a display would result in a loss in legibility. In the implementation of the dot matrix, legibility would be enhanced by using clearly circular dots rather than elongated ones. ©
AllenG. Vartabedian 1971
Legibility of symbols on CRT displays Allen G. Vartabedian American Telephone & Telegraph Company (This work was undertaken while the author was at Bell. Telephone Laboratories, Holmdel, New Jersey,
The effects of several parameters - such as letter orientation, slanting versus upright, dot matrix size, etc - of symbol formation on the legibility of CRT displays were studied. Symbols were tested by subjects attempting to identify them when presented briefly on the display screen. Speed and accuracy of identification were used as the measures of legibility. The 7 x 9 dot matrix symbols drawn with circular dots were superior to all others for both reaction time and errors. Slanting had a detrimental effect on dot and stroke symbols and circular dot symbols were superior to elongated dot symbols.
A major question in considering cathode ray tube (CRT) displays is how the CRT, with its wide range of characteristics, will affect human performance. Of basic importance to this question are the characteristics of displayed symbols, both from the standpoint of performance and the cost of the resulting device. This paper considers how various methods for generating symbols are related to legibility. Legibility, in this case, is determined by the subject's ability to identify a symbol, shown briefly, to the exclusion of all other symbols of a defined set. Seyeral commercial methods are available for generating symbols, among which are the dot matrix and stroke techniques. The dot matrix method is what its name implies: each character is formed of a series of dots in a matrix of possible dot positions. Symbols are drawn by blanking and unblanvdng the electron beam according to a pattern of light points within the fixed matrix. In the stroke method, by comparison, the unblanked electron beam is moved like a pen to trace out the shape of a given symbol. The end points of segments of the symbol fall upon fixed points within a matrix. While it was felt by many that the stroke-generated method might result in more effective displays, the dot matrix method is both more economical and more versatile, since it can be applied to other display devices in addition to CRTs - for example, displays using gas-discharge devices, light-emitting diodes, electroluminescent materials and fibre optics. The parameters under study were: 1. Symbol generation method, ie symbols drawn with dots versus symbols drawn with continuous strokes. 2. Dot matrix size: the generally encountered 5 x 7 matrix size versus the 7 x 9. 3. Symbol orientation: symbols that are upright versus symbols slanted 20 ° from the vertical. 4. Dot geometry: circular dots versus'vertically elongated dots (these tended to "bleed" together into vertical lines).
130
AppliedErgonomics September1971
These parameters were studied in six experimental conditions on two display devices (Fig 1). The dot matrix symbols used in this study were designed by the author for optimal legibility. Both.5 x 7 and 7 x 9 symbol fonts were stylistically similar within the constraints imposed by the two different matrix sizes. The stroke font was based on the Leroy font which is of known good legibility. The letter 'o' of the stroke font was modified to conform to the dot matrix font by the addition of a loop at the top. In this experiment there is inevitably some confounding of the effects of symbol generation method and symbol design. The various generation methods impose restrictions on character shaping which ultimately constrain symbol design to some extent. The design of both 5 x 7 and 7 x 9 dot symbols followed the same design criteria but resulted in somewhat differently styled symbols in some cases. This is true especially for the symbol '8'. The stroke symbols were designed according to an entirely different criterion by a different designer. Again certain striking differences resulted in symbol designs especially with the letter 'Q'.
Method The experimental task required each of the 23 subjects to depress a button which caused a single symbol to appear briefly on a CRT. The subject then verbally identified the symbol shown and continued in the same manner for all the trials of the experiment. Each subject experienced all six conditions of the experiment. In each of the conditions the subject saw 5 blocks of the 36 capital letters and numbers. Each block was a random permutation of the 36 symbols. There were 180 trials per condition for each subject. Two display consoles employing P31 and P4 phosphors were used as the experimental apparatus. Four conditions, namely the stroke, stroke slant, 7 × 9 (circle dots) and 5 x 7 (circle dots), were conducted using the P31 phosphor. The remaining two conditions, the 7 x 9 (elongated dots) and
7 x 9 slant (elongated dots), were conducted using the P4 phosphor. Both of these phosphors emit energy in a range near the centre of the visible spectrum. The duration of the image was controlled by the decay of phosphorescence of the CRT. This decay is exponential and, for these phosphors, drops to ten percent of its initial brightness in 1 to 2 milliseconds. Thus, with this method the duration of the CRT image is governed by the type of CRT phosphor and has a very small variability. Symbols were displayed in the centre of the CRT with a symbol spot brightness of 37"69 cd/m 2 (11 ft-L) and a background brightness of 6"8 cd/m 2 (2"0 ft-L). The distance between the subject's eyes and the display was fixed at 711 mm (28 in) by means of a head rest. Symbols were 3"55 mm (0.14 in) in height and subtended a visual angle of 17-2 minutes at the observer's eye. The width/ height ratio of a nominal width symbol was set at 0-75. The time from the display of the symbol to the subject's response was recorded by a PDP-9 computer control program by means of a voice switch. Errors in the subject's response were recorded by the experimenter. Results Table 1 presents the overall error rates and reaction times for each of the six experimental conditions. Error data are available for 23 subjects and response time data for 12 subjects. This table shows that the 7 x 9 circle dot font yielded the lowest average reaction time and error rate.
A three factor analysis of variance (generation method x symbols x subjects) was conducted on the reaction time scores. It was found that the generation method had a statistically significant effect (p < 0-005). Symbols had a significant effect (p < 0.001). Also the method - with symbols interaction was significant (p < 0"001). With respect to errors, a two factor (methods x subjects) analysis of variance was conducted after pooling the errors across all symbols. It was found that generation method had a significant effect (p < 0"001). Several comparisons among the six experimental conditions were available to ascertain the effects of the four parameters outlined in the introduction. These comparisons are presented in Table 2.To judge the statistical significance of the differences, the 99 "9 percent confidence interval was calculated from the analysis of variance to be 47"7 ms. Table 2 shows that all but two of the observed
Table 1 Average reaction time and error rates for the six conditions. Condition
Reaction time (ms)
% Errors
5 x 7 circle dot 7 x 9 circle dot
697 602
2"9 2"6
stroke upright
659
4" 5
stroke slant 7 x 9 elongated dot 7 x 9 slant elongated dot
677 678 814
6"7 3"2 7"7
Fig 1 Samples of the six conditions of the experiment. From top to bottom, 5 x 7 circle dot, 7 x 9 circle dot, stroke upright, stroke slant, 7 x 9 elongated dot and 7 x 9 slant elongated dot. In the experiment, symbols were of equal height and equal brightness. Also the last two conditions were not blurred as in the photograph.
Applied Ergonomics September 1971
131
differences are greater than 47'7 ms. A similar confidence interval for error rates requires a difference greater than 0"2% to assert that the error rates are significantly different. All differences in error rates are greater than O'2%. To ascertain the effect of the first parameter, generation method, we examine comparisons '1' and '2' which test the stroke upright font against both the 7 x 9 and 5 x 7 circle dot fonts. In both instances, differences in error rates are in favour of the dot fonts. Differences in reaction times are in favour of the 7 x 9 dot matrix size. For the third parameter, symbol orientation, we examine comparisons '4' and '5'. With respect to errors both comparisons are in favour of the upright orientation with the difference for the stroke font being half that of the dot font. For stroke symbols the observed difference of 18 ms is well below 47"7 ms in reaction time. For 7 x 9 elongated dot symbols, however, the difference is 136 ms the largest difference detected in any comparison. The interaction of the slanting and stroke versus 7 x 9 elongated dot parameters can be noted. The final parameter under question is dot geometry, examined in comparison '6'. Here we note the differences in error rates and reaction times are both in favour of the circular dots for the 7 x 9 matrix symbols.
Table 2 Selected comparisons among the six methods of the experiment (differences are in favour of the top method for each comparison). Difference Comparison (1) Stroke upright vs 5 x 7 circle dot (2) 7 x 9 circle dot vs stroke upright (3) 7 x 9 circle dot vs 5 x 7 circle dot (4) Stroke upright vs stroke slant (5) 7 x 9 elongated dot upright vs 7 x 9 elongated dot slant (6) 7 x 9 circle dot vs 7 x 9 elongated dot
1. The 7 x 9 circle dot font was superior in reaction time and error measures to all other fonts. 2. Dot elongation adversely affected legibility. 3. The 7 x 9 dot matrix font was superior to the 5 x 7 dot matrix font. 4. Dot matrix symbol generation was supeTior in legibility to stroke symbol generation.
132
Applied Ergonomics September 1971
Error rate
38
- 1 "6%
57
1 "9%
95
0-3%
18
2-2%
136
4"5%
76
0"6%
5. The slanting of symbols adversely affected legibility with 7 x 9 elongated dot symbols suffering greater degradation than stroke symbols.
Summary and conclusions The results of the experiment are summarised as follows:
Reaction time (ms)
Based on the results of this experiment it is concluded that, of the three methods of symbol generation tested, the 7 x 9 dot font is superior in legibility to either the 5 x 7 dot font or stroke font. The use of slanted symbols in a display would result in a loss in legibility. In the implementation of the dot matrix, legibility would be enhanced by using clearly circular dots rather than elongated ones. ©
AllenG. Vartabedian 1971