Experiment Design

Independent Variables

The experiment was structured in a 2 x 2 design. The independent variables were the type of display (CRT or LCD), and the size of the viewable area, which was varied for each display type, yielding two conditions for each display—large, with a static display; or small, requiring scrolling, and hence a dynamic display.

Controlled Variables

In order to obtain a meaningful comparison between the two display technologies, we used displays that could be configured to provide the same resolution and the same pixel density. To represent the AMLCD technology, we used a Taxan Crystalvision 650, a full color active matrix TFT device with 1024 x 768 resolution and a 368 mm (14.5") diagonal. The CRT chosen was an Innovision MAG Computronic Colorview 15, also with 1024 x 768 resolution. The CRT allowed the adjustment of both the horizontal and vertical image sizes, and for this experiment, the CRT was adjusted to feature the same pixel density as the AMLCD.

The illumination ambient on each of the screens was measured with a Minolta LS-100 Luminance meter, directed at a 98% reflectance surface provided by NIST. The room lighting was adjusted so that the ambient illumination was 24.5 cd/m², and the appropriate adjustment to the room lighting was made before each experimental run.

The font displayed subtended a visual angle of 90 minutes of visual arc at the experimental viewing distance of 40 cm. Neal (1968) found that character legibility approaches 100% as the visual angle exceeds 10 minutes of arc. Our task utilizes a character size will in excess of this load. Mansfield (1996) , in a study of the effect of fonts on reading acuity and reading speed, detected a small but significant advantage of Courier over Times for subjects with low visual acuity. For subjects with normal visual acuity, the difference in reading speed was smaller and favored Times over Courier. The font chosen for our experiment was Times Roman, since this is commonly used in the office environment and our subjects were required to have normal visual acuity. We used black characters on white a background.

For each display, the level of the white background was adjusted to 125 cd/m². The black level produced by the two types of display differed by a factor of two, being 4.41 cd/m² for the CRT and 1.91 cd/m² for the AMLCD. For reference, the contrast ratios these measurements yield are

CR_LCD = 125/1.91 = 65:1

and

CR_CRT = 125/4.41 = 28:1

To prevent an unintentional bias, the name of the manufacturer of the display and the type of display was concealed.

Dr. Shneiderman reviewing the test procedure.

Dependent Variable

The dependent variable measured was the total time required to locate single letter targets in a block of pseudo-text. Search Time began immediately after the pseudo-text is presented on the display, initiated by the subject through pressing a button on the keyboard (Return key). Search Time ended when the subject indicated completion of the page of pseudo-text. Subjects used a button on the keyboard (Space Bar) to indicate each time a target was detected. Another button (Return Key) was used to indicate completion of the search. For the dynamic case, the up and down arrows were used to scroll the pseudo-text, one half line at a time. (A pilot study showed that scrolling integer number of lines at a time caused the subjects to lose orientation of their place in scanning the text.)

Pseudo-text was chosen to minimize the cognitive workload, which can cause large variability among subjects in performance achieved. Each subject should have performed the task in all of the experimental conditions, performing the static test first, followed by the dynamic test. The order for the type of display was assigned at random. On the completion of the search testing, the subjects were asked to evaluate the visual quality of the test displays, rating each on a 9 point scale, with 1 being "poor" and 9 being "excellent". The subjects were also asked to rate, using the same 9 point scale, their perception of each display with respect to visual comfort.

Hypothesis

We hypothesize that there is no difference in search time between the display types for each condition of testing (static and dynamic), and also that there is no difference in the subjective ratings between the display types. This hypothesis implies that the distribution of the difference in search times for each subject (e.g., Search Time_CRT - Search Time_LCD) in each experimental condition is normally distributed with a mean of zero.

Subjects

The experiment participants were volunteers from the undergraduate and graduate student body at the University of Maryland, restricted to those with corrected or uncorrected visual acuity of 20/40 or better. The visual acuity was measured with a modified Snellen chart generated on a high resolution printer at the 40 cm viewing distance, and was administrated on a pass/fail basis.

Test Material

The test material was pseudo-text generated with the 8-bit single-byte coded graphic character set. The work of McConkie (1976) and Fisher (1976) indicate that

1. The effects of degraded signal quality on the display has a similar effect on both searching for single letter targets and searching for whole word targets.
2. If the number of spaces is consistent with normal text, eye movements will be approximately the same in the search task as with reading.

In order to maintain consistency between the pseudo-text and normal text, the pseudo-text was generated subject to the following constraints:

1. Pseudo-text shall consist of random strings of letters and digits separated by single spaces.
2. Each line of pseudo-text shall consist of between 30 and 35 characters, including embedded spaces. The pseudo-text shall contain 10 to 15 targets of a single kind. The position of the targets shall be randomly chosen, with the restriction that a line may not start or end with the target character.
3. The pseudo-text shall contain, on the average, 5 spaces per line, position randomly chosen, with restrictions that a line shall not start or end with a space character; a space character shall not be adjacent to another space; and the minimum string length shall be 2 characters.
4. The block of pseudo-text shall be placed in the center of the screen. The target character should have intermediate discriminability with respect to the other characters (e.g. do not use O, 0 or Q as target characters)
5. The number of pseudo-texts per subject should be large enough to prevent memorizing effects.

Figure 1: Sample of Pseudo-Text

Subject Training

The subjects were trained before the main experiment by performing the task for at least 2 blocks of pseudo-text in the static condition. The number of counted targets will be registered as a check for the subject's concentration. In the course of the experiment, the performance measurement will be neglected (discarded) if the number of missed or misidentified targets is too large ( ±10% of the actual count of targets—i.e. 9 to 11 indicated targets is acceptable in a text with 10 targets). In the training session, practice trials continued until the subject's performance on the last 2 training trials has an error rate consistent with the error rate required for valid data in the experiment.

The subjects were allowed a rest break of up to one minute between the trials, with a minimum break of 10 seconds. The minimum break was often ignored. Subjects were told to minimize errors and work as quickly as possible, and they were given feedback on the accuracy of their response after each trial. Subjects continued performance of the task for each experimental condition until they had accumulated 5 valid data points in each experimental condition.

Continue
Return to the Title Page for Comparison of AMLCD and CRT Displays