Results

Table 2:

Effect of grouping cue type on selection speed.

Table 3:

Effect of grouping format on selection speed.

Table 4:

Effect of input device and grouping cue on selection speed.

Table 5:

Effect of grouping cue and grouping format on selection speed.

A 2X3 within by 2 between factor ANOVA was performed on the factors of cue (color or text), form (row, column, block), and input device (mouse or tablet). The three-way interaction of cue, form, and device was not significant ( F(2,120) = 0.94, p>0.01). However, the interaction of cue and form was significant ( F(2,120) = 82.15, p<0.01). Neither of two-way interactions involving input device were significant (form by device: F(2,120) = 0.66, p>0.01) and ( cue by device: F(1,76) = 0.20, p>0.01), and the main effect of input device was not significant ( F(1,60) = 5.26, p>0.01). On the other hand, the main effects of both cue ( F(1,60) = 456.54, p<0.01) and form ( F(2,120) = 44.13, p<0.01) were both significant (see Figure 9).

Figure 9:

The significant interaction of grouping cue and grouping form showing a strong divergence between color blocks and text blocks.

Post-hoc comparisons show a significant difference between the color row and color column groups (t(61) = 9.27, p< 0.01) while the difference between text row and text column groups was not significant (t(61) = 2.37, p> 0.01). This suggests that performance in row and block grouping formats are not significantly different when combined with color grouping cues, but that they both yield better performance than the column grouping form. This contrasts sharply to the text grouping cue conditions in which row and column grouping forms perform equally but the block format leads to significantly poorer performance.

Even though there was a significant interaction of grouping cue and grouping format, the factor of grouping cue alone accounted for the most variance in performance ( = 0.51). The interaction ( = 0.13) still accounted for more variance then did the main effect of grouping format ( = 0.08).

Pre-Action Duration (Response Latency):

The pre-action duration is the time between presentation of the target and the movement of the cursor out of the trigger area measured in 60th of a second.

Use of the tablet input device led to a longer pre-action duration (M = 0.92, min = 0.45, max = 1.92) than did use of the mouse (M = 0.68, min = 0.19, max = 1.63) when other factors are not considered (see Table 6). When comparing all color grouped conditions in both input conditions, the color grouping conditions yielded shorter pre-action durations (M = 0.70, min = 0.19, max = 1.92) than the conditions employing text grouping cues (M = 0.81, min = 0.23, max = 1.80; see Table 7) . The spatial format of the grouping (row, column, or block) indicates that the row format yields the shortest pre-action durations (M = 0.72, min = 0.19, max = 1.92) with column and blocks formats trailing in performance (M = 0.73, min = 0.28, max = 1.42 and M = 0.82, min = 0.23, max = 1.82 respectively; see Table 8). In contrast to the speed of selection, trials involving the mouse input device regardless of grouping cue lead to the shortest pre-action durations. The combination of mouse input device and color grouping cue lead to the best performance (M =0.62), with the trials combining text grouping cues and the mouse input device leading to better performance (M = 0.73) than both the tablet conditions (M = 0.86 for the color grouping condition and M = 0.99 for the text grouping condition). This suggests that if this interaction is significant, the effect of input device is more important to performance than grouping cue for pre-action duration (see Table 9).

Putting input device aside for a moment, the combination of color grouping cues with all grouping forms (row, column, and block) leads to shorter pre-action durations (M = 0.68, 0.70, and 0.71 respectively) than does the combination of text grouping cue with all grouping forms (M = 0.75, 0.77, and 0.92 respectively). In Table 10 you will note that the combination of a text grouping cue and a block grouping form yields remarkably long pre-action durations. This contrasts strongly against the color grouping block form condition that lead to only marginally longer pre-action durations than both color row and color column conditions.

Table 6:
Effect of input device on pre-action duration.

Table 7:
Effect of grouping cue on pre-action duration.

Table 8:
Effect of grouping format on pre-action duration.

Table 9:
Effect of input device and grouping cue on pre-action duration.

Table 10:

Effect of grouping cue and format on pre-action duration.

A 2X3 within by 2 between factor ANOVA was performed on the factors of cue (color or text), form (row, column or block), and input device (mouse or tablet). The three-way interaction of cue, form, and device was not significant ( F(2,120) = 0.09 p>0.01), however, the interaction of cue and form was significant ( F(2,120) = 8.24, p<0.01). Neither of two-way interactions involving input device were significant (Form by device: F(2,120) = 0.61, p>0.01 and cue by device: F(1,76) = 0.19, p>0.01), but the main effect of input device was significant ( F(1,60) = 16.66, p<0.01). The main effects of both cue ( F(1,60) = 43.45, p<0.01) and form ( F(2,120) = 19.87, p<0.01) were also significant.

Figure 10:
The significant interaction of grouping cue and format shows the different effect grouping format has for the color and text grouping cues.

Post-hoc comparisons between the color row and color column groups and text row and text column groups were not significant (t(61) = 0.62, p> 0.01 and t(61) = 0.87, p> 0.01 respectively). This bears out the remarkable differences found among the two block format conditions and between the block format and other group format conditions.

Again, the main effect of grouping cue accounted for the most variance ( = 0.14) with grouping format ( = 0.08) and the interaction of these two factors ( = 0.03) following. It is interesting to note the difference between a high level of accounted variance in the speed measure and a relatively low level in the pre-action duration.

Errors in Selection

Selections of the wrong goal item or selections that missed the target and fell outside of the goal item are counted as errors.

Surprisingly, 41 of the 62 participants performed with two or fewer errors. The fact that the vast majority of errors were committed by a very small number of participants, mainly under the first condition that they were tested under, makes it difficult to draw any conclusions from the data. Nevertheless, some familiar trends are present. As in speed and pre-action duration, color grouping cues lead to fewer errors (M = 0.60) than the text grouping cue conditions (M = 0.72, see Table 12). Also familiar is the effect of grouping format on errors. Conditions that used the row group form (M = 0.60) performed better than column forms (M = 0.62) and block forms (M = 0.76, see Table 13).

Contradicting previous trends is the effect of input device on errors. Participants that used the tablet input device produced fewer errors (M = 0.36) than did users of the mouse input device, (M = 0.80, see Table 11).

Table 11:
Effect of input device on errors.

Table 12:
Effect of grouping cue on errors.

Table 13:
Effect of grouping format on errors.

A 2X3 within by 2 between factor ANOVA was performed on the factors of cue (color or text), form (row, column or block), and input device (mouse or tablet). The three-way interaction of cue, form, and device was not significant ( F(2,120) = 2.15 p>0.01), nor was the interaction of cue and form ( F(2,120) = 0.84, p>0.01). Neither of two-way interactions involving input device were significant (Form by device: F(2,120) = 0.04, p>0.01 and cue by device: F(1,76) = 0.17, p>0.01). Similarly, the main effect of input device was not significant ( F(1,60) = 1.89, p>0.01). The main effects of both cue and form were also not significant ( F(1,60) = 1.07, p>0.01 and F(2,120) = 1.45, p>0.01 respectively). The large variability and sparse results make finding any differences difficult.

Fatigue Measure

At the end of each condition, subjects were asked to rate their level of fatigue on a 9 point scale. "Very fatigued" was coded as a 1 and "Not fatigued" was coded as a 9, with seven points in between. All measures yielded a minimum score of 1 and a maximum of 7.

Continuing the trend found in speed of selection and pre-action duration, the fatigue measure shows a preference for the mouse input device (M = 4.64) over the tablet input device (M = 3.82, see Table 14). Also supporting the results of previous analysis, Table 15 shows conditions that used color grouping cues lead to less fatigue (M = 4.59) than did those incorporating text grouping cues (M = 3.91). Table 16 also supports the trends of speed and pre-action duration with block grouping forms yielding more fatigue (M = 3.88) than both row and column formats which differed minimally (M = 4.45 and 4.43 respectively). Table 17 suggests that the block form of the text grouping cue is the most fatiguing condition (M = 3.40).

Table 14:
Effect of input device on fatigue.

Table 15:
Effect of grouping cue on fatigue.

Table 16:
Effect of grouping format on fatigue.

Table 17:
Effect of grouping cue and format on fatigue.

A 2X3 within by 2 between factor ANOVA was performed on the factors of cue (color or text), form (row, column or block), and input device (mouse or tablet). The three-way interaction of cue, form, and device was not significant ( F(2,76) = 0.20 p>0.01). Similarly, the interaction of cue and form was not significant ( F(2,76) = 1.34, p>0.01). Neither of two-way interactions involving input device were significant (Form by device: F(2,76) = 0.54, p>0.01) and ( cue by device: F(1,38) = 0.62, p>0.01, see Figure 10), nor was the main effect of input device significant ( F(1,38) = 3.52, p>0.01). Figure 11 shows the main effects of both cue ( F(1,38) = 15.27, p<0.01) and form ( F(2,76) = 6.33, p<0.01) which were significant. As in previous measures, grouping cue accounts for the most variance ( = 0.28) with format of grouping following ( = 0.12).

Figure 11:
The significant main effects of grouping cue and grouping format. Text cues and block forms lead to greater fatigue.

Figure 12:
The effect of input device was not significant, nor was the interaction of device and grouping cue.

Post-hoc comparisons between the color row and color column groups (t(39) = 0.69, p> 0.01) and text row and text column groups (t(39) = 0.46, p> 0.01) were not significant.

Learning effects:

The effects of task learning within conditions and over the entire experiment were controlled for in the previous analysis, but the possibility of a learning effect is interesting to many applications. Three subjects at random were chosen for further analysis. Figures 13 - 16 show performance over each block of 64 selections in speed and pre-action duration as a function of time. Subjects in these graphs are using the computer mouse. No learning effects are apparent. Speed ratings are uniformly variable with no significant signs of improvement from mid to late trials. Pre-action durations are remarkably stable which suggests a process that either is not affected by learning or must continue for many hundreds of trials before its affects are seen. Figures 17 and 18 show similar results for a single user of the tablet input device.

Figure 13:
Sample trials using the color grouping cue and the mouse input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 14:
Sample trials using the text grouping cue and the mouse input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 15:
Sample trials from a second subject using the color grouping cue and mouse input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 15:
Sample trials from a second subject using the text grouping cue and mouse input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 16:
Sample trials from a third subject using the color grouping cue and tablet input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 17:
Sample trials from a third subject using the text grouping cue and tablet input device show no systematic improvements in speed or pre-action duration over 64 trials.

Figure 18 shows how speed and pre-action duration are constant across the 6 blocks of 64 trials. The trials are in order of presentation and are compared against the samples mean for that trial type that is controlled for learning effects. Subject performance does not diverge from the sample mean as trials progress. Any learning effect here would present itself with subjects performing below the sample mean in early trials, but above the sample mean in later trials. Learning effects are not evident in this subset of the sample.

Figure 18:

Aggregated performance for 3 subjects as they progress across 6 display conditions.

Relationship between Speed and Pre-action Duration
Table 18 shows the correlation between selection speed and pre-action duration. The moderate negative correlation indicates that subjects who perform well in selection speed also have shorter pre-action durations. The fact that higher speeds and lower pre-action durations are considered better performance leads to the negative relationship. This finding indicates that subjects are not trading off speed for pre-action duration, but are performing similarly in both measures.

Table 18:
The correlation of speed and pre-action duration for 3 randomly selected subjects. The inverse nature of pre-action duration leads negative correlations to indicate positive relationships.

Condition Subject 1 Subject 2 Subject 3

Color row -0.317 -0.077 -0.106 Color column -0.247 -0.222 -0.292 Color block 0.082 -0.289 -0.337 Text row -0.203 -0.053 -0.362 Text column -0.230 -0.258 -0.523 Text block -0.231 -0.463 -0.057

<< Back to Methods Section , On to Discussion Section >>