Effort drives saccade selection
- Experimental Psychology, Helmholtz Institute, Utrecht University, Netherlands
- Department of Psychology, University of Groningen, Netherlands
Figures
Figure 1
Pupil size differs during saccade planning across directions.
(a) Twenty participants planned saccades in a cued direction. Saccades were executed as fast as possible upon cue offset. (b) All 36 possible saccade targets around the visual field. Only eight equally spaced locations were shown per trial. (c) Pupil size over time, split and averaged during saccade planning in oblique/cardinal, upward/downward, and left/rightward directions, locked to cue offset. Shaded areas indicate ± 1 s.e.m. (d) Averaged z-transformed pupil size during planning (–150 ms–170 ms post cue, gray area in e) across directions. (e) Linear mixed-effects model using obliqueness, verticalness, horizontalness of directions, and saccade properties to predict pupil size during saccade planning. (f) Standardized partial coefficients per predictor with 95% confidence intervals. *p < 0.05, **p < 0.01, ***p < 0.001.
Figure 2
Saccade preferences differ across directions and are predicted by saccade costs.
(a) The same twenty participants freely selected one of two saccade targets. (b) The average saccade preferences across directions (sum selected/sum offered). Shaded bands indicate ± 1 s.e.m. (c) Saccade costs correlated negatively with saccade preferences across directions: costly directions were avoided and affordable directions preferred. Black datapoints represent directions (averaged across participants). (d) Pupil size was larger for avoided compared with preferred directions. (e) Saccade costs predicted saccade selection on a trial-by-trial basis (56.64%). Together, the saccade costs in the first task predicted saccade preferences in the subsequent task. (c-e) Error bars reflect bootstrapped 95% confidence intervals. (d-e) Black datapoints represent participants. **p < 0.01, ***p < 0.001.
Figure 3
Saccade curvature and latency reveal active weighing of cost during saccade selection.
(a) Schematic layout of saccade trajectories curving away (magenta) or toward (cyan) non-selected options. Curvature was calculated as the peak deviation from a straight line between gaze positions at saccade onset and offset. The top-right histogram shows that more saccades curved away than toward the non-selected option. (b) Difference in pupil size during the saccade planning task is linked to the peak curvature deviation in the saccade preference task. (c) Same as b, but now linked to saccade latency in the saccade preference task. Larger differences in pupil size are related to more oculomotor conflict between the two options, as reflected in more curvature away from the non-selected option and longer saccade latencies. (b, c) Black line depicts the relationship across all trials, gray lines denote regression fits per participant. (d) Saccade-cost based prediction of saccade selection split for toward and away curving trials. On a trial-by-trial basis, saccade costs predicted saccade selection above chance (59.72%) when saccades curved away from the non-selected option. In contrast, saccade costs did not predict saccade selection for ’toward’ saccades. Black datapoints represent participants. All error bars reflect bootstrapped 95% confidence intervals. ***p < 0.001.
Figure 4
Saccade costs underlie saccade preferences in natural viewing.
(a) Forty-one participants searched for small letters (‘Z’ or ‘H’) in natural scenes (Exp. 1; n = 16), and either ignored (single task) or additionally attended (dual task) an auditory number stream (Exp. 2; n = 25). (b) Saccade preferences during search without auditory stimulation. (c) Preferred directions were associated with a smaller pupil size prior to the saccade (Exp. 1). (d, e) Same as b, c but now for Exp. 2 without attending the auditory number stream (single task). Preferred directions were again associated with a smaller pupil size preceding the saccade (Exp. 2). (f) Same as d but now under the increased cognitive demand of the (primary) auditory digit counting (dual) task. (g) Adjustment in saccade preferences between single- and dual-task conditions in percentage points. (h) Less saccades were executed in the more demanding dual-task condition. Black datapoints represent participants. (i) Pupil size during the single task predicted direction adjustments under additional cognitive demand. Costly saccades as assessed during the single-task condition were especially cut in the dual-task condition. (b, d, f, g) Shaded bands represent ± 1 s.e.m. Other error bars reflect bootstrapped 95% confidence intervals. (c, e, i) Black lines depict the relationship across all trials, gray lines denote regression fits per participant. ***p < 0.001.
Appendix 1—figure 1
Saccade-locked pupil traces as index of saccade costs in different directions.
Using the average pupil size in the 350 ms before saccade onset (saccade-locked), results remained qualitatively identical. Planning oblique saccades was associated with a larger pupil size than cardinal ones (β = 9.897, SE = 2.223, t = 4.451, p < 0.001), and downward saccades were associated with a larger pupil size than upward saccades (β = 0.471, SE = 0.112, t = 4.189, p < 0.001). A slightly increased pupil size for leftward compared with rightward saccades was observed as well (β = 0.260, SE = 0.107, t = 2.436, p = 0.015).
Author response image 1
The relationship between saccade latency (from the saccade planning task) and saccade preferences averaged across participants.
Individual points reflect directions and shading represents bootstrapped 95% confidence intervals.
Tables
Appendix 1—table 1
Full outcomes of the linear mixed-effects model analyzing pupil size assessed saccade costs across directions.
| Predictor | β | SE | t | p |
|---|---|---|---|---|
| Intercept | –26.510 | 9.059 | –2.926 | 0.003 |
| Obliqueness | 7.662 | 1.957 | 3.916 | <0.001 |
| Verticalness | –0.556 | 0.171 | –3.261 | 0.001 |
| Horizontalness | –0.226 | 0.095 | –2.388 | 0.017 |
| Obliqueness × Duration | –0.109 | 0.031 | –3.495 | <0.001 |
| Duration | 0.169 | 0.085 | 1.981 | 0.048 |
| Amplitude | –2.423 | 0.656 | –3.691 | <0.001 |
| Landing error | 7.344 | 2.168 | 3.387 | 0.001 |
| Peak velocity | 0.054 | 0.015 | 3.692 | <0.001 |
| Participant Var | 602.851 | 2.786 | ||
| Participant × Obliqueness Cov | –55.890 | 0.395 | ||
| Participant × Verticalness Cov | –1.625 | 0.058 | ||
| Obliqueness × Verticalness Cov | 0.934 | 0.011 | ||
| Obliqueness Var | 13.790 | 0.083 | ||
| Verticalness Var | 0.380 | 0.003 |
Appendix 1—table 2
Full outcomes of the linear mixed-effects model predicting pupil size using saccade preferences and control variables in Experiment 1.
| Predictor | β | SE | t | p |
|---|---|---|---|---|
| Intercept | 0.825 | 0.061 | 13.504 | <0.001 |
| Direction preferences | –1.784 | 0.324 | –5.412 | <0.001 |
| X coordinate | –0.0002 | 0.00001 | –21.12 | <0.001 |
| Y coordinate | –0.0002 | 0.00001 | –11.316 | <0.001 |
| Luminance | –0.635 | 0.017 | –36.251 | <0.001 |
| Saliency | –0.001 | 0.0001 | –10.536 | <0.001 |
| Trial number | –0.006 | 0.00006 | –91.139 | <0.001 |
| Fixation number (in trial) | 0.006 | 0.0002 | 55.513 | <0.001 |
| Saccade duration | –0.004 | 0.0009 | –42.150 | <0.001 |
| Fixation duration | 0.0003 | 0.00004 | 7.105 | <0.001 |
| Amplitude | 0.009 | 0.0007 | 13.979 | <0.001 |
| Participant Var | 0.053 | |||
| Direction preferences Var | 0.833 | 0.646 | ||
| Participant × Direction preferences Cov | 0.003 | 0.250 |
Appendix 1—table 3
Full outcomes of the linear mixed-effects model predicting pupil size using saccade preferences and control variables in Experiment 2.
| Predictor | β | SE | t | p |
|---|---|---|---|---|
| Intercept | 1.093 | 0.060 | 18.301 | <0.001 |
| Direction preferences | –0.644 | 0.170 | –3.780 | <0.001 |
| X coordinate | –0.0001 | 0.000005 | –24.442 | <0.001 |
| Y coordinate | –0.00004 | 0.000007 | –5.072 | <0.001 |
| Luminance | –0.341 | 0.009 | –38.269 | <0.001 |
| Saliency | –0.001 | 0.00006 | –10.405 | <0.001 |
| Trial number | –0.007 | 0.00006 | –126.687 | <0.001 |
| Fixation number (in trial) | 0.006 | 0.00009 | 64.174 | <0.001 |
| Saccade duration | –0.007 | 0.00005 | –152.775 | <0.001 |
| Fixation duration | 0.000005 | 0.00002 | 0.732 | 0.464 |
| Amplitude | 0.013 | 0.0003 | 37.737 | <0.001 |
| Participant Var | 0.086 | 0.041 | ||
| Direction preferences Var | 0.322 | 0.339 | ||
| Participant × Direction preferences Cov | 0.034 | 0.086 |
Appendix 1—table 4
Full outcomes of the linear mixed-effects model predicting pupil size using saccade direction adjustment and control variables.
| Predictor | β | SE | t | p |
|---|---|---|---|---|
| Intercept | 1.070 | 0.060 | 17.880 | <0.001 |
| Direction adjustment | –9.333 | 0.966 | –9.659 | <0.001 |
| X coordinate | –0.0001 | 0.000005 | –23.388 | <0.001 |
| Y coordinate | –0.00004 | 0.000007 | –5.261 | <0.001 |
| Luminance | –0.340 | 0.009 | –38.269 | <0.001 |
| Saliency | –0.001 | 0.00006 | –10.109 | <0.001 |
| Trial number | –0.007 | 0.00006 | –126.380 | <0.001 |
| Fixation number (in trial) | 0.006 | 0.00009 | 62.648 | <0.001 |
| Saccade duration | –0.007 | 0.00005 | –153.597 | <0.001 |
| Fixation duration | 0.00002 | 0.00002 | 1.537 | 0.124 |
| Amplitude | 0.012 | 0.0003 | 37.127 | <0.001 |
| Participant Var | 0.087 | 0.041 | ||
| Direction adjustment Var | 16.878 | 11.338 | ||
| Participant × Direction adjustment Cov | 0.142 | 0.499 |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Effort drives saccade selection
eLife 13:RP97760.
https://doi.org/10.7554/eLife.97760.3
