Contextual saccade adaptation induced by sequential saccades

Abstract

Saccade adaptation is the gradual adjustment of saccade end point to maintain spatial accuracy. Contextual adaptation refers to a situation in which the adaptation-related change in saccade end point is contingent on the behavioral context in which the saccade is made. For example, in some situations, the same saccade to the same retinotopic location can be simultaneously adapted in opposite directions depending on the context in which it is made. Saccade adaptation has traditionally been studied in isolated movements, but in everyday life, saccades are often planned and executed in sequences. The oculomotor system may therefore have adaptive mechanisms specific to sequential saccades. Here, in five experiments, we investigated contextual saccade adaptation in sequences of saccades. In the first experiment, we demonstrate that saccades to a given retinotopic location can be simultaneously adapted in opposite directions depending on whether they occur in isolation or in a sequence. In the other experiments, we measured the extent to which properties of the previous and following saccades in a sequence can induce contextual saccade adaptation. Overall, we find that the existence, direction, and amplitude of previous and subsequent saccades, as well as the order of the current saccade within a movement sequence, can all induce contextual adaptation. These novel findings demonstrate the surprising flexibility of the system in maintaining end point accuracy, and support the idea that saccades made in a movement sequence are planned concurrently rather than independently.

NEW & NOTEWORTHY This study reveals a new type of contextual saccade adaptation: sequential saccades are able to induce contextual saccade adaptation when direction, amplitude, or the existence of preceding and following saccades are used as contexts. These novel findings are also consistent with the idea that saccades made in a sequence are planned concurrently rather than independently.

INTRODUCTION

Saccade adaptation is the process by which the oculomotor system modifies saccades to minimize the amount of saccadic error, defined as the mismatch between the saccade end point and the target location. Saccade adaptation has been traditionally considered a recalibration mechanism which, when given a constant visual error signal, changes the saccade vector gradually over time (1–3). In experimental situations, saccade adaptation is typically induced by consistently jumping the saccade target either inward (toward fixation) or outward (away from fixation) during the execution of the saccadic eye movement. Due to saccadic suppression (4), these target jumps are not perceived (5), but they gradually modify the gain of successive saccades to the same target position by either decreasing or increasing saccade gain, depending on the direction of the target jump. Although they are gradual, the gain changes can begin to become evident after only a few (1, 6) or even only one trial (7). However, it can take a few hundred trials to obtain the maximum adaptation effect (8).

If saccade adaptation were only a simple recalibration procedure, any systematic error for a specific target position would apply to all targets, independent of their visual properties. Indeed, saccade adaptation is typically independent of the visual features of the target, and transfers to other stimuli at the same retinotopic location (9). However, it has recently been shown that in some circumstances saccade adaptation does not completely transfer, and instead depends on the context of the movement (10–13). This contextual saccade adaptation can make it possible to simultaneously adapt in opposite directions (e.g., both increasing gain and decreasing gain), resulting in different saccadic gains for the same retinotopic target location depending on the context. In this scenario, the direction of the gain change that will be expressed for a particular movement depends on the context of the saccade toward the target location (13). Consistent with the motor recalibration role of saccade adaptation, spatial- and motor-related contexts such as orbital eye position (14, 15), head orientation (16), depth component in three-dimensional space (17), and gravity magnitude (18) can induce contextual saccade adaptation.

In contrast, most visual properties, such as the shape (19), color (20), or shape and color (9, 10) of saccadic targets cannot induce contextual saccade adaptation. Interestingly, however, this is not true for all visual contexts: previous studies have identified some visual cues that are able to induce contextual adaptation in opposite directions, such as flickering versus nonflickering targets (12) and different directions and speeds of moving targets (10). This indicates that saccade adaptation is not just a simple motor recalibration process, but instead represents a complex plasticity of the oculomotor system that can incorporate information about visual context.

Here, we investigate whether the properties of previous or future saccades in a sequence of saccadic eye movements can serve as a context for saccade adaptation. The possibility that a previous or future saccade could affect adaptation of the current saccade arises because we know from previous studies that sequential saccades can be planned in parallel (21–34). For example, when subjects perform a sequence of two saccades, the intersaccadic interval is typically shorter than the latency of a single saccade (23, 28, 34), and may be inversely correlated with the latency of the first saccade (23, 28, 35). Moreover, in a sequence of two saccades, displacement of the first saccadic target does not update the second saccade in within-object cases (36), consistent with overlapped programming and execution of two sequential saccades.

In this study, we will use contextual saccade adaptation to probe the extent to which planning and execution of a previous or future saccade can influence adaptation of the current saccade. This will not only illuminate the interactions between saccades planned in a sequence; it will also reveal the complex plasticity of the oculomotor system.

GENERAL METHODS

Subjects

Thirteen naive subjects and one of the authors were voluntarily recruited for this study (6 females and 8 males, aged 26–34 yr). Eight subjects were recorded for each of the five experiments. All protocols were approved by the State University of New York, College of Optometry Institutional Review Board, and subjects provided written, informed consent.

Apparatus

Subjects sat in a dark room with their heads stabilized via chin and forehead rests, 114 cm from a 32 in. 1,920 × 1,080 IPS LCD with 120 Hz refresh rate (Display++ LCD Monitor, Cambridge Research Systems Ltd., Rochester, UK) while the eye level was aligned with the center of the LCD. The stimuli were presented by PsychoPy, a presentation software package written in Python (37). For tracking and recording of eye movements, an SR-Research EyeLink 1000+ Desktop Mount system was used with a sampling rate of 1,000 Hz to record the right eye position. A standard 13-point calibration followed by a validation procedure was used preceding all experimental sessions.

Procedure

In experiment 1 (single saccade vs. saccade sequence), we asked whether the programming of saccades in a sequence can affect saccade adaptation. Specifically, we tested whether the presence or absence of a previous saccade in a sequence could serve as a context for saccade adaptation, thereby differentially affecting the gain of the current saccade.

In single-saccade trials, subjects began each trial by maintaining fixation on a fixation point, presented 4° above the center of the display, for a randomly chosen period of 1,000 ± 250 ms. Following this, the fixation point was extinguished and a saccade target simultaneously appeared 10° below the fixation point (Fig. 1). The fixation point and visual target were black rings with a diameter of 0.5° and line thickness of 0.2°, and subjects were instructed to perform a reflexive saccade to the target.

Figure 1.

Schematic diagram showing the trial sequence in experiment 1 during the adapting phase, for the backward and forward adaptation contexts. Backward adaptation: i) A fixation point appears and subjects start fixation. ii) After 1,000 ± 250 ms, the fixation point disappears and simultaneously a target appears 10° below the fixation point. Subjects are instructed to perform a single saccade to the target as soon as possible. iii) In the adapting phase, during the saccade, a backward intrasaccadic step (ISS) is applied to the target to induce saccade adaptation. iv) A corrective saccade may be performed by subjects to fixate the target, and subjects are required to maintain fixation on the target for 400 ms. Forward adaptation: i) The fixation point is presented on the upper-right side of the screen and subjects start fixation. ii) After 1,000 ± 250 ms, the fixation point disappears and simultaneously two targets appear on the midline. iii) Subjects are instructed to perform the first saccade horizontally to the upper target and iv) a second saccade to the lower target. Upon initiation of the second saccade, the upper target disappears. Simultaneously, during the adapting phase, a forward ISS is applied to the second target. v) Subjects are required to maintain fixation on the final target for 400 ms, after any corrective saccade has been made.

In sequential saccade trials, the fixation stimulus and initial fixation period were the same as in the single-saccade trials, but the fixation point was presented 4° above the center of the display and was also horizontally offset 10° to the right of midline. Upon fixation point offset, two target stimuli were presented simultaneously: one (the T1 target) was located 10° to the left of the fixation point (i.e., on the horizontal midline, where the fixation point was located in the single saccade trials), and the other (the T2 target) was located on the horizontal midline 10° below the first (i.e., at the same location as the target in the single saccade trials). Both target stimuli were identical to the target stimulus used in single saccade trials. Participants were instructed to execute a saccadic sequence consisting of a movement to T1 followed by a movement to T2 as quickly as possible. T1 disappeared as soon as the second saccade was detected. In both conditions, participants were required to maintain fixation on the first target for at least 10 ms and on the final saccade target for 400 ms, at which time a blank screen appeared.

After a baseline phase consisting of 25 trials of each context (single target vs. sequence) to measure the starting gain of the saccades, there was an adapting phase consisting of 200 trials of each context. In these trials, in half of the subjects we applied −2° intrasaccadic steps (ISS) in single saccade trials (backward adaptation) and +2° ISS in sequential saccade trials (forward adaptation) upon initiation of the saccade toward the final saccadic target. In the other half of the subjects, the direction of the ISSs for the contexts was switched. This adapting phase was followed by a postadapting phase (25 trials of each context) to measure the final saccade gain. This phase used the same procedure as the baseline phase, with no ISS applied to the saccadic targets. Trials from the two contexts were presented randomly interleaved across the experiment.

The procedures for experiments 2–5 were similar to experiment 1 insofar as the visual characteristics of the fixation point and saccadic targets, the target eccentricities, the number of trials, and the durations of the fixation periods were the same. The only differences were that the metrics of the saccades preceding or following the adapted saccade were modified to determine the factors that influence the presence and magnitude of contextual saccade adaptation. The specific differences and their motivations are described separately in the sections related to each experiment. When the same subject participated in more than one experiment, we ensured that at least 2 days intervened between testing in the different experiments.

Saccade Detection

Saccades were detected online during experiments when eye position crossed a boundary of 2° from the fixation point. However, for data analysis, saccades were detected offline using an eye velocity criterion of 50°/s. Saccade start and end points were marked when velocity fell under two standard deviations above the median saccade velocity during fixation. Saccade velocity was computed as the difference between two data samples (in degrees) divided by the sampling duration. Saccades having a gain of less than 0.5 and more than 1.5 were excluded from this study. Using these criteria, on average 2.73% of trials were rejected across all the experiments and subjects.

Statistical Analysis

To test for significant differences between saccade contexts, we compared mean saccade amplitudes using paired t tests applied to the data from each subject. These comparisons were done using the last 50 trials of each context in the adapting phase and using all 50 trials in the postadapting phase. Then we adjusted the P values to correct for multiple comparisons using the Benjamini–Hochberg procedure (BH) in each experiment for the adapting and postadapting phases to control the false discovery rate at the level of 0.05 (38, 39).

RESULTS

Experiment 1: Single Saccade versus Saccade Sequence

In experiment 1, saccade amplitude gradually changed during the adapting phase. In Fig. 2A, saccade amplitude is plotted as a function of trial number for a typical subject. As can be seen, from the beginning of the adapting phase, where the intersaccadic steps were applied, saccade amplitude gradually began to diverge for the two different contexts. In the postadapting phase (during which intersaccadic target steps were absent), the difference in saccade amplitude between the contexts gradually decayed somewhat but did not disappear. Across all subjects, we plotted the mean saccade amplitude as a function of trial number by binning the data points: first we averaged saccade amplitude in 25 trial bins in each context for each subject separately, and then we calculated the mean and 95% confidence intervals for each binned data point across all subjects (Fig. 2B). Here, we see the same pattern of context-dependent adaptation of saccade amplitude during the adapting and postadapting phases.

Figure 2.

Results for adaption of a single saccade vs. a saccade in a sequence (experiment 1). A: saccade amplitude as a function of trial number. Data from a typical subject; black dashed lines separate baseline, adapting, and postadapting phases. Each circle indicates saccade amplitude. Solid lines indicate the moving average over trials. Blue and red colors, respectively, represent forward and backward adaptation. The saccade amplitudes gradually differ for the two contexts during the adapting phase, and this difference begins to decay during the postadapting phase. B: similarly, saccade amplitude is plotted as a function of trial number, averaged across all subjects in experiment 1. Thick lines indicate the moving average and the shaded regions indicate 95% CIs. C: this panel shows the difference in saccade amplitude in the two contexts during the adapting and postadapting phases. Each circle indicates the difference for each individual subject. The black horizontal lines indicate the average of these differences across all the subjects, and the green and yellow shaded bars the 95% CIs, respectively for the adapting (green) and postadapting (yellow) phases. CI, confidence interval.

The mean difference in saccade amplitude between the two different contexts was statistically significant during both the adapting and postadapting phases: when we averaged the saccade amplitudes in the last 50 trials of each context during the adapting phase across all subjects, the difference in amplitude between the two contexts was 2.15° (Fig. 2C). When we applied the same procedure to the trials in the postadapting phase, the resulting difference between the saccade amplitudes was 1.57°. A paired t test followed by the BH procedure showed that these differences are statistically significant [adapting phase: t(7) = 5.25, P = 0.001; postadapting phase: t(7) = 5.29, P = 0.001].

Experiment 2: Direction of the Preceding Saccade

The results from the previous experiment show that a given saccade vector can be simultaneously adapted in opposite directions (i.e., contextual saccade adaptation) depending on the presence or absence of a preceding saccade. In this experiment, we tested whether a given saccade vector can be simultaneously adapted in opposite directions depending on the direction of the preceding saccade. This experiment was similar to experiment 1 (Fig. 3A); in one context the initial fixation point was presented 10° to the right side of the midline, whereas in the other context the fixation point was presented 10° to the left. After initial fixation, two targets appeared on the midline and subjects were instructed to perform a horizontal saccade to the upper target followed by a vertical saccade to the lower target. Upon initiation of the second saccade, the upper target disappeared, and in adapting trials, the second target was stepped according to context (backward when the initial saccade was to the right and forward when the initial saccade was to the left).

Figure 3.

Effect of the direction of the preceding saccade on adaptation (experiment 2). Conventions are as in Fig. 2. except where noted. A: schematic diagram showing the trial sequence during the adapting phase. i) The fixation point is presented either on the upper-left side or upper-right side of the screen, respectively, in backward and forward adaptation contexts; then subjects start fixation. ii) After 1,000 ± 250 ms, the fixation point disappears and simultaneously two targets appear on the midline. iii) Subjects are instructed to perform the first saccade horizontally to the upper target; iv) and a second saccade to the lower target. Upon initiation of the second saccade, the upper target disappears. Simultaneously, during the adapting phase, an intrasaccadic step (ISS) is applied to the second target. v) Subjects are required to maintain fixation on the final target for 400 ms, after any corrective saccade has been made. B: saccade amplitude as a function of trial number, across all subjects. Thick and thin lines indicate moving averages with 95% CIs for forward (blue) and backward (red) adaptation. C: the difference in saccade amplitudes for the two contexts during the adapting and postadapting phases with 95% CIs. CI, confidence interval.

Similar to experiment 1, in this experiment, saccade amplitudes in the two contexts gradually diverged during the adapting phase (Fig. 3B), and they continued to differ during the postadapting phase. The average difference in saccade amplitude between the two contexts (i.e., previous saccade to the left vs. previous saccade to the right) across the last 50 trials of the adapting phase was 1.51°, and the average difference during the postadapting phase was 1.33° (Fig. 3C). Paired t tests followed by the BH procedure showed that both of these differences in saccade amplitude between the two contexts were significantly different from zero [adapting phase: t(7) = 5.79, P < 0.001; postadapting phase: t(7) = 7.02, P < 0.001].

Experiment 3: Direction of the following Saccade

Results from experiments 1 and 2 showed that the presence/absence (experiment 1), or a difference in direction (experiment 2), of a preceding saccade can provide a context that allows differential (opposite) changes in gain in the current saccade (contextual saccade adaptation). In experiment 3, we investigated if differences in the direction of a subsequent saccade can also induce contextual saccade adaptation. Trials began with a red fixation circle presented on the horizontal midline, 4° above the center of the display (Fig. 4A). After the fixation point disappeared, three horizontally aligned visual stimuli were presented 10° below the fixation point: one on the vertical midline and the other two at 10° horizontal eccentricity to the right and left. Each subject was instructed to make a sequence of two saccades as quickly as possible; first a vertical saccade to the middle target and then a horizontal saccade to either the left- or right-side target. The direction of the second saccade was cued by the direction of a 45°-angled gap inside the fixation circle, presented during the initial fixation period. During the adapting phase, a backward ISS was applied to the target of the first saccade in trials in which the second saccade was directed to the left, and a forward ISS was applied to the target of the first saccade in trials in which the second saccade was directed to the right.

Figure 4.

Effect of the direction of the following saccade on adaptation (experiment 3). Conventions are as in Fig. 2. A: schematic of the trial sequence during the adapting phase. The fixation point appears and subjects start fixation; the orientation of a small gap in the fixation circle cues the direction of the second saccade. ii) The fixation point disappears after 1,000 ± 250 ms, and simultaneously three peripheral visual stimuli appear. iii) Subjects first make a saccade to the middle target as quickly as possible; in the adapting phase, upon saccade onset, a forward or backward intrasaccadic step (ISS) (depending on the context) is applied to the target. iv) Subjects may perform a corrective saccade. v) A second saccade toward the final target is made, and subjects are required to hold fixation for 400 ms. B: saccade amplitude as a function of trial number, across all subjects with 95% CIs. C: the difference between the saccade amplitudes in the two contexts during the adapting and postadapting phases with 95% CIs. CI, confidence interval.

Similar to the previous experiments, the gains of the initial saccades in the two contexts gradually diverged during the adapting phase (Fig. 4B): the gain of the first saccade decreased when the second saccade was directed to the left and increased when the second saccade was directed to the right. The average difference in saccade amplitude between the two contexts was 0.73° for the final 50 trials of the adapting phase, and 0.50° during the postadapting phase (Fig. 4C). Both of these differences were significantly different from zero [paired t tests followed by the BH procedure: adapting phase: t(7) = 5.05, P = 0.002; postadapting phase: t(7) = 2.67, P = 0.031].

Experiment 4: Amplitude of the following Saccade

In experiment 3, we found that the direction of a future saccade can induce contextual saccade adaptation. Here, we asked whether the amplitude of a future saccade can also provide a context allowing the gain of a given saccade vector to be simultaneously adapted in opposite directions depending on the amplitude of the subsequent saccade. The procedure was similar to experiment 3, except that the visual stimuli were presented at three locations (Fig. 5A): on the horizontal midline, as well as 5° and 10° to the right. Subjects were instructed to first perform a vertical saccade toward the left target and then a second saccade toward either the closer or farther target based on the angle of a cue that was presented on the initial fixation circle: a gap oriented at 20° or 45° indicated that the second saccade should be directed to the closer or farther target, respectively.

Figure 5.

Effect of the amplitude of the following saccade on adaptation (experiment 4). Conventions are as in Fig. 2. A: schematic of the trial sequence during the adapting phase. i) The fixation point appears and subjects start fixation; a small gap in the fixation circle cues the eccentricity of the target for the second saccade. ii) The fixation point disappears after 1,000 ± 250 ms, and simultaneously three peripheral visual stimuli appear. iii) Subjects first make a saccade to the middle target as quickly as possible; in the adapting phase, upon saccade onset, a forward or backward intrasaccadic step (ISS) (depending on the context) is applied to the target. iv) Subjects may perform a corrective saccade. v) A second saccade toward the final target is made, and subjects are required to hold fixation for 400 ms. B: saccade amplitudes as a function of trial number, across all subjects with 95% CIs. C: the difference between the saccade amplitudes in adapting and postadapting phases with 95% CIs. CI, confidence interval.

The saccade amplitudes for the two conditions gradually diverged during the adapting phase in this experiment too, but this difference was smaller and decayed rapidly during the postadapting phase (Fig. 5B). The average difference in saccade amplitude between the two contexts was 0.50° during the adapting phase (Fig. 5C), which was significantly different form zero [paired t tests followed by the BH procedure: t(7) = 3.08, P = 0.035]. However, during the postadapting phase, the average difference in saccade amplitude for the two contexts was 0.06°, which was not significantly different from zero [paired t tests followed by the BH procedure: t(7) = 0.61, P = 0.560]. Thus, it appears that the amplitude of the following saccade in a sequence provides a weaker context for saccade adaptation than does the presence/absence of a previous saccade and the direction of a previous or following saccade in the sequence.

Experiment 5: Order of the Saccade in a Sequence

The previous experiments showed that different preceding or following saccades are able to induce contextual saccade adaptation. In this experiment, we tested whether the order of a saccade in a sequence is also able to induce contextual saccade adaptation (i.e., the first saccade vs. the second saccade, Fig. 6A). In the first context, in which the first saccade in the sequence was adapted (similar to experiment 2), the initial fixation point appeared 10° to the right side of the midline. After initial fixation, two targets appeared on the midline and subjects performed a horizontal saccade to the upper target, followed by a vertical saccade to the lower target. In trials of the other context, in which the second saccade was adapted (similar to experiment 3), the initial fixation point appeared on the midline. After initial fixation, two horizontally aligned visual stimuli were presented 10° below the fixation point: one on the vertical midline and the other one 10° to the left. Subjects were instructed to perform a vertical saccade on the midline, followed by a horizontal saccade to the left as quickly as possible.

Figure 6.

Effect of the order of the saccade within a sequence on adaptation (experiment 5). Conventions are as in Fig. 2 except where noted. A: schematic diagram showing the trial sequence during the adapting phase. Backward adaptation: i) The fixation point appears and subjects start fixation. ii) After 1,000 ± 250 ms, fixation point disappears and simultaneously two targets appear 10° below the fixation point. Subjects are instructed to first perform a saccade to the middle target as quickly as possible. iii) In the adapting phase, during the saccade, a backward intrasaccadic step (ISS) is applied to the target to induce saccade adaptation. iv) A corrective saccade may be performed by subjects to fixate the target, and subjects are required to maintain fixation on the target for 400 ms. Forward adaptation: i) The fixation point is presented on the upper-right side of the screen and subjects start fixation. ii) After 1,000 ± 250 ms, the fixation point disappears and simultaneously two targets appear on the midline. iii) Subjects are instructed to perform the first saccade horizontally to the upper target and iv) a second saccade to the lower target. Upon initiation of the second saccade, the upper target disappears. Simultaneously, during the adapting phase, a forward intrasaccadic step applied to the second target. v) Subjects are required to maintain fixation on the final target for 400 ms, after any corrective saccade has been made. B: saccade amplitudes as a function of trial number, across all subjects with 95% CIs. C: the difference between the saccade amplitudes in adapting and postadapting phases with 95% CIs. CI, confidence interval.

Similar to the previous experiments, saccade amplitudes for the two conditions gradually diverged during the adapting phase and continued to differ during the postadapting phase (Fig. 6B). The average difference in saccade amplitude between the two contexts across the last 50 trials of the adapting phase was 2.57°, and the average difference during the postadapting phase was 1.61° (Fig. 6C). Both of these differences in saccade amplitude between the two contexts were significantly different from zero [paired t tests followed by the BH procedure: adapting phase: t(7) = 9.95, P < 0.001; postadapting phase: t(7) = 4.91, P = 0.001]. Their results show that the order of a saccade in a saccade sequence can induce contextual saccade adaptation (i.e., different adaptation for the first saccade vs. the second saccade in the sequence).

Summary and Comparison of Experimental Results

Figure 7 summarizes the results across all five experiments. We found that adaptation of a given saccade is influenced by whether the saccade is part of a sequence of movements, including the presence/absence of a previous saccade, the direction and amplitude of previous and/or future saccades, and the ordinal position of the saccade within a sequence. These sequence-dependent contextual modulations of saccade adaptation led to significant differences in saccade amplitude between the contexts for all the experiments, as measured in the final 50 trials of the adapting phase. Significant differences were also observed in the postadapting phase, except in experiment 4 in which we tested the effect of differences in the amplitude of the following saccade on saccade adaptation. Experiment 4 also resulted in the smallest overall gain change (Fig. 7), perhaps because the difference between the two tested contexts was relatively small (5° difference in the amplitude of the following saccade) compared with the other experiments (10° or 20° difference in the end points of the following or preceding saccades). Therefore, the nonsignificant result in the postadapting phase of experiment 4 might be due to the weakness of the adaptation in this experiment.

Figure 7.

Summary across all five experiments. Summary bar graph representing the amount of contextual saccade adaptation in each experiment. The bars indicate the amplitude difference between the two contexts in each experiment, during the adapting (green) and postadapting (yellow) phases. The error bars represent 95% CIs. Schematic diagrams of saccade sequences and intrasaccadic steps (ISSs) in each experiment are also presented in this figure. The blue arrows indicate the saccade direction and amplitudes, and red arrows indicate the direction of ISSs. Note that the backward and forward intersaccadic steps were switched between the contexts for half of the subjects. CI, confidence interval.

To make valid comparisons of contextual saccade adaptation in sequences of saccades, it is important to verify that the adapted saccades in the two contexts have comparable directions. We found that direction differences in our study were quite small. Specifically, when we calculated the difference between the saccadic angles across the comparison context pairs for each subject, and averaged the angle between the corresponding contexts (e.g., left vs. right following saccades) across the subjects, we found only minor differences between the saccadic angles, ranging between −0.4° and 0.6° for all the experiments in the adapting and postadapting phases. As might be expected, paired t tests showed that none of these differences were statistically significant, with t-test statistic values (with 7 degrees of freedom) across all comparisons ranging from −1.53 to 1.66, and P values from 0.14 to 0.84 (uncorrected for multiple comparisons; corrected values would be even higher). Therefore, it would be difficult to account for our results on the basis of direction differences alone, considering that saccade adaptation has been shown to generalize over a much larger range of saccade vectors [the “adaptation field” (40, 41)].

DISCUSSION

In this study, we made the novel finding that it is possible to induce contextual saccade adaptation in saccades executed as part of a sequence based on the properties of the preceding and following saccades in the sequence. The properties capable of eliciting contextual saccade adaptation include the existence and direction of a preceding saccade (experiments 1 and 2), the direction and amplitude of a following saccade (experiments 3 and 4), and the order of a given saccade within a saccade sequence (experiment 5). These findings demonstrate the surprising flexibility of the saccadic system in maintaining end point accuracy, and support the idea that saccades made in a movement sequence are planned concurrently rather than independently.

Specifically, the results show that sequential saccades must interact during their planning and/or execution, thereby influencing the adaptation state of the oculomotor system. Here, our demonstration that the same saccade can be adapted in opposite directions depending on the sequence of movements in which it is embedded indicates that saccades with different preceding and following saccades must be associated, at least in part, with differing patterns of neuronal activity in the oculomotor system, even if the vector of that saccade and the properties of the visual target stimulus are unchanged. This is consistent with the idea that sequential saccades are not strictly independent; rather, the fact that they are in a sequence influences the generation of each of the individual component movements (42) and supports the idea that planning of sequential saccades can overlap in time (21–35, 43).

Could the changes in saccade amplitude that we observe arise from a volitional strategy on the part of subjects rather than from plasticity of the oculomotor system? We argue that this cannot be the case in our experiments, for the following reasons: first, most naive subjects cannot recognize the ISSs during the adapting phase, and therefore could not apply a voluntary strategy (4, 44). Second, saccade adaptation is usually gradual (1, 7, 45–47) whereas a conscious strategy would be expected to cause a more abrupt change in saccade amplitude near the beginning of adapting phase, when subjects notice the direction and amplitude of the ISSs in each context. In contrast to this, the results from all the experiments in this study showed a gradual modification of saccade amplitude during the adapting phase. Third, after adaptation, the difference in saccade amplitude between the contexts remained significant even in the absence of ISSs, during the postadapting phase. The only exception to this is experiment 4, and this can be explained by the relatively weak contextual adaptation that was obtained during the adapting phase in this experiment. Fourth, our results showed different amounts of saccade adaptation for each experiment, a result which is difficult to explain with a conscious modification of saccade amplitude. Finally, previous experiments showed that some contexts cannot induce contextual saccade adaptation (9, 10, 20), a finding which is, again, difficult to explain if one assumes a conscious strategy.

Earlier studies showed that the transfer of adaptation between a single reflexive saccade and the same saccade made in a sequence of scanning saccades may be limited or absent (11, 48). In these experiments, the arrangements of the sequential movements were complex, consisting of saccades traversing the corners of a square. However, subsequent studies using a simpler sequence of two saccades showed a more complex picture: single-saccade adaption can transfer to the second saccade, but only when that saccade targets a new object (49). Moreover, adaptation of the second saccade of two transfers to a single saccade aiming at a new object (50), but not to single saccades that explore the same object. Our results demonstrate that the existence of just a single preceding or following saccade in a planned sequence can be sufficient for contextual adaptation of the same saccade in opposite directions (experiments 1 and 5). Our results also demonstrate that adaptation state can be linked to the particular characteristics (e.g., direction and amplitude) of previous and future saccades in a sequence (experiments 2–4).

We previously reported that different directions and speeds of moving saccade targets can induce contextual saccade adaptation (10). Our current results suggest that performing a subsequent saccade in response to the target motion might have played a role in the contextual saccade adaptation that we observed in that experiment. However, the study also showed that varying the speed of a moving target induces stronger adaptation than varying its direction. This finding seems counterintuitive considering our results here which indicate that the direction of a following saccade provides a stronger context for adaptation than its amplitude. This apparent conflict could perhaps be resolved by having a greater difference between the amplitudes of the following saccades in the current experiments. Nevertheless, further experiments are needed to reject the role of target motion in contextual saccade adaptation.

Saccade adaptation is a complex motor learning process that can also modify visual perception (51). Following saccade adaptation, localization of a visual stimulus that flashes before an adapted saccade (41, 48, 52, 53) and even during fixation (54) tends to shift in the same direction as the saccade adaptation. The amount of this mislocalization is correlated with the amount of induced saccade adaptation. Distortion of object perception has also been reported after whole field saccade adaptation (55). Moreover, saccade adaptation has been successfully induced by using a perceptual task in the absence of any bottom-up visual position error signal (56), indicating that a top-down error signal can also modify the oculomotor system.

It is possible that differences in the locus of visual attention around the time of saccade execution might also be able to induce contextual saccade adaptation. In particular, previous studies have shown that the spatial allocation of attention during sequences of saccades depends not just on the goal of the current saccade (57–59), but also on the goals of the other saccades in the sequence (60–62). Here, we propose that it might be possible to induce contextual saccade adaptation by using the different sites of visual attention for the different saccades in the sequence as the context, but this conjecture will also have to await future experimentation.

In conclusion, our results introduce a new type of contextual saccade adaptation that is contingent of the properties of previous and future saccades in a sequence, and also support the idea that the individual movements in a saccade sequence are processed concurrently rather than independently.

GRANTS

This work was supported by National Institutes of Health (NIH) Grant R01-EY030669 (to R.M.M.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.A. and R.M.M. conceived and designed research; R.A. performed experiments; R.A. analyzed data; R.A. and R.M.M. interpreted results of experiments; R.A. prepared figures; R.A. drafted manuscript; R.A. and R.M.M. edited and revised manuscript; R.A. and R.M.M. approved final version of manuscript.