Figure
2 shows the average trial duration, saccade count and saccade latencies as a function of prior information about the target locations. From left to right on each graph in the figure, results show performance as the amount of prior information about the target locations is restricted from all seven targets displayed simultaneously, to only the next five targets displayed at a time, the next three targets are shown, finally to the next target only being revealed when a saccade was executed to the preceding target. This shows that, as prior information about the target locations was reduced trial duration increased,
F(3, 51) = 52.5, MSE = 312,918,
p < 0.001,
η2 = 0.755 (with contrasts between sequential information levels, 7 vs 5, 5 vs 3 and 3 vs 1, showing significance at
p’s < 0.011). This was a function of an increase in saccade count,
F(3, 51) = 85.3, MSE = 0.818,
p < 0.001,
η2 = 0.834, and average saccade latency,
F(3, 51) = 19.4, MSE = 977,
p < 0.001,
η2 = 0.532. Contrasts between each sequential level for saccade count show no significant difference between levels 3 and 1 (
p = 0.30) but do show significance (
p’s < 0.001) for 7 vs 5 and 5 vs 3. Contrasts for saccade latency also show a significant decrease in saccade latency as prior information increases (1 vs all other levels of prior information; 3 vs 7,
p’s < 0.05; but not 5 vs 7). Having greater information about subsequent target locations improved task completion times through a combination of there being an overall reduction in the number of saccades being executed and reduced response latencies.
1 It is worth noting that there was no difference in the first saccade latencies for each prior information level: prior information level 7,
M = 318 (SD = 11); prior information level 5,
M = 330 (SD = 27); prior information level 3,
M = 317 (SD = 17); prior information level 1,
M = 304 (SD = 17).
Overall saccade amplitudes generally decreased as prior information about target locations decreases (Fig.
3a),
F(3, 51) = 12.1, MSE = 0.083,
p < 0.001,
η2 = 0.416. Contrasts between each sequential level for saccade amplitude shows no significant different between levels 3 and 1 (
p = 0.813) but start to show a trend toward significance between levels 5 and 3 (
p = 0.079) and a significant difference between levels 7 and 5 (
p < 0.001). Examination of the underlying distribution of amplitudes across participants shows a bimodal one with peaks at about 0.5° and 3° (not shown). These reflect target-driven saccades (the larger ones) and the corrective saccades (smaller or shorter ones). Separating amplitudes at the 1.5° trough between these peaks results in the graphs shown in Fig.
3b, c. Saccade counts can be seen to increase as prior information decreases for both the larger saccades and the smaller amplitude saccades (Fig.
3b; amplitude magnitude:
F(3, 51) = 805, MSE = 1.41,
p < 0.001,
η2 = 0.979; prior information:
F(3, 51) = 85.3, MSE = 0.409,
p < 0.001,
η2 = 0.834; interaction:
F(3, 51) = 3.17, MSE = 0.267,
p < 0.001,
η2 = 0.411). While the interaction is significant, separate one-way ANOVAs for prior information depending on saccade amplitude continue to reveal significant effects for both large and small amplitude saccades: large amplitude saccades:
F(3, 51) = 29.0, MSE = 0.390,
p < 0.001,
η2 = 0.814 and small amplitude saccades:
F(3, 51) = 31.5, MSE = 0.286,
p < 0.001,
η2 = 0.650. Contrasts also show the same pattern across each amplitude magnitude with each level being significantly different from the next (all
p’s < 0.01) for both the number of smaller and longer amplitude saccades. Average saccade amplitudes for large and smaller saccades (Fig.
3c) also show significant effects (amplitude magnitude:
F(3, 51) =
F(3, 51) = 2893.703, MSE = 0.188,
p < 0.001,
η2 = 0.994; prior information:
F(3, 51) = 4.81, MSE = 0.128,
p = 0.005,
η2 = 0.220; interaction:
F(3, 51) = 5.74, MSE = 0.128,
p = 0.002,
η2 = 0.252). Separate one-way ANOVAs depending on prior information show significant effects for the larger saccades,
F(3, 51) = 5.34, MSE = 0.252,
p = 0.003,
η2 = 0.239, but no effect of prior information on the smaller ones,
F(3, 51) = 1.046, MSE = 0.004,
p = 0.380,
η2 = 0.058. Contrasts showed no differences across information level for the smaller saccades (
p’s > 0.148) but did show significant differences across information level for the longer amplitude saccades (7 vs 5 and 5 vs 3:
p’s < 0.001; with levels 3 vs 1 also showing a trend,
p = 0.058).
The average landing position error from the nearest target location is shown in Fig.
3d as a function of prior information. It can been seen that error reduces as the prior information is reduced reflecting the reduced influence of future target locations on saccade landing position control,
F(3, 51) = 9.55, MSE = 0.874,
p < 0.001, η2 = 0.360. Contrasts between each sequential level for saccade count shows no significant different between levels 7 and 5 (
p = 0.268) but do show a significant difference (
p’s < 0.001) for 5 vs 3 and a trend for 7 vs 5 (
p = 0.058).