59.gif)
The consistent lack of this inverse relationship suggests that speed and pre-action duration are not capturing different parts of a single selection procedure but instead group selection is a different process from item selection. This is not an indication, however, that speed and pre-action duration are capturing unique processes. The most likely explanation is that pre-action duration is capturing both group search and some item search, or the speed measure captures most of the item search as well as some of the group search. Figure 20 illustrates the possible combinations of group and item search. The relationship between the two measures is most likely a reflection on some third variable that effects both measures. Eye/hand coordination, cognitive speed, and spatial ability have all been suggested as factors that may effect both measures.
The Effects of Input Device
As previous research has shown (Cohen, Meyer & Nilsen; 1993) both pointing
input devices (touch screen or graphics tablet) and mouse tend to have very
similar performance in selection tasks. The present study confirms this
result. There were no significant effects of input device on speed of
selection. There was, however, a main effect of input device on pre-action
duration. This effect can best be explained as an artifact of the SumaGraphics
Tablet. When a subject lifts the stylus off of the trigger area on the tablet,
the tablet tends to lose magnetic contact with the stylus. Not knowing the
current location of the stylus, the tablet freezes the on-screen pointer in
it's last known position until the stylus regains magnetic contact with the
tablet. This has the effect of leaving the cursor inside of the trigger area
even though the subjects' movement had began. This made the pre-action
duration of selections made with the tablet artificially high. Fortunately,
this artifact did not affect the speed measure because it is a measure of rate
and not duration. Also supporting the findings of no device effect is the non
significant effect of device on fatigue.
Learning Effects
Past research into menu selection has found diminishing differences between
different treatments as practice increased (Mehlenbacher, Duffy & Palmer;
1989). These treatments are usually different organizations of items within
group, or different numbers of groups. It's not surprising that people will
learn where things are. Mehlenbacher, et al. found that in only 75 trials
their manipulations of group labels and item ordering had diminished effect.
Treisman & Gelade (1980), on the other hand, found no learning effects in
1664 trials of a visual search task. Figures 13 - 18 show no evidence of
learning effects in this study. Even though subjects in this study performed
only 384 selections, it's fair to speculate that the effects of grouping format
may diminish over repeated exposure. On the other hand, it is unlikely that
the effect of grouping cue (color or text) will diminish over time.
Methodological
Considerations
The color printing on the templates for the tablet were not perfectly matched.
There were brown/orange confusions that lead subjects to double check the
target on these trials. This is probably not a significant problem because the
text templates matched well and still didn't show any effect of input device.
Some subjects reported that they did not use the template on the tablet in making selections. The main reason given was that the tablet's angle led to a glare on the template that made seeing the template difficult. This is a more difficult problem because it occurs in all conditions and may have masked a main effect of device.
Practical
Implications
When formatting items within groups it is important to choose the appropriate
form for the information presented. Column organizations are useful for single
word searches found in menus and indexes. This will facilitate scanning and
speed search. On the other hand, rows are most useful for alphabetically
organized single characters that are becoming popular in today's software. A
frequent sight on the World Wide Web is a row of alphabetically ordered links
that serve as an index to large databases in the same way that Roladexes use
single letter tabs for locating groups of similar items. In cases where space
is at a premium, block organizations of alphabetically organized information
may be used, providing that groups are delineated with color or spacing.
Color is a powerful cue for group separation and may be used in place of spacing or borders in situations where items are grouped. For the task of group separation, there is no limit of the number of colors used in the display as long as neighboring groups have moderate to high contrast. It should be noted that particularly dark colors should be avoided due to their poor contrast with black text. Borders of red and blue groupings should also be avoided as this color combination tend to induce the perception of depth. Although this will enhance the group selection, the illusion of depth tends to draw undue attention to itself and away from the items in the group. The use of color for grouping allows the elimination of space wasting separators which increase the allowable density of the screen without decreasing the usability of the display.
Color can also be used as a search aid in menu systems, improving searches in many ways. Using up to ten maximally discriminability colors (equally spaced selections from a psychometrically weighted color wheel) as group identifiers allows accelerated searches among a large number of items. This approach to grouping also allows for grouping of non-consecutive items, that is, items that are not located in a spatial group. This is a useful feature when groupings change dynamically, as in dynamic queries (Williamson & Shneiderman; 1992).
Of particular use in menuing systems is the ability to map semantic factors onto the grouping colors so that the colors relay information about the items without requiring close examination of items. For instance, relief maps use color to represent real features on the earth so that these features can be located, grouped, and examined quickly. A quick glance at such a map relays enough information for you to estimate the proportion of water to land, the locations of at least 6 major land masses, and the locations of all the highest mountain ranges (represented in white). Similarly, colors can be mapped to levels of some factor such as annual rainfall, or specific states of a factor like operational, failing, or failed equipment. As in all semantic mapping conditions, cultural factors that effect the users understanding of certain colors must be taken into account. In all these cases, the colors used have some semantic link to the category items, and a search for any category of items is merly a color feature search.