Automaticity of online control processes in manual aiming

Abstract

Experiments that manipulated the visual feedback of the moving limb have suggested the existence of efficient and automatic online correction processes. We wanted to determine whether the latency/gain of the correction for a cursor jump are only influenced by the size of the cursor jump or whether they are also influenced by the need of a correction for the target to be reached. In Experiment 1, we used two target sizes (5 and 30 mm) and three cursor-jump amplitudes (5, 15, and 25 mm), so that for some target size/cursor-jump combinations, no correction would be needed to reach the target. Participants were not aware of the cursor jump, but we observed a 65% correction regardless of target size. In Experiment 2, participants pointed at a large target for which a 15-mm cursor jump never impeded target attainment. Participants modified the trajectory of their movement in the direction opposite to the cursor jump (42% of the cursor jump). Our results indicate that the latency of the correction for a cursor jump was not influenced by the size of the cursor jump or that of the target. However, the correction tailored the movement's initial impulse according to the target's characteristics.

Introduction

Reaching movements toward a visual target put into play a series of processes for identifying the target and its location and transforming this information into appropriate motor commands (for a review, see Desmurget, Pélisson, Rossetti, & Prablanc, 1998; Elliott & Khan, 2010). The processes leading to movement planning and then to movement execution have intrinsic variability. Because of this variability inherent in all biological systems, and because of the high level of accuracy required in many of our daily activities, the CNS must quickly update movement planning and amend movement execution. 

Error detection and correction processes have been at the center of many research efforts since the seminal work of Woodworth (1899). Although many research strategies have been used in the last century, many authors opted to use a perturbation paradigm. In this paradigm, some aspects of the task are changed just prior to, at, or soon after movement initiation. Because these perturbations occur unexpectedly and often for only a small proportion of the trials, it is expected that participants would plan their movements as if no perturbation would occur. Thus, to reach the target, participants need to correct the movement they have planned and initiated to counteract the perturbation, which opens a window on error detection and correction processes. 

In many experiments, the perturbation changed the target location, its visually perceived location, or the velocity of the participant's hand. In target-jump experiments (Bridgeman, Lewis, Heit, & Martha, 1979; Desmurget et al., 1999; Goodale, Pélisson, & Prablanc, 1986; Gritsenko, Yakovenko, & Kalaska, 2009; Prablanc & Martin, 1992), participants first gazed at a fixation point and, following a variable foreperiod, a target was lit in his or her peripheral visual field. Participants were asked to look and to aim at the target quickly and accurately. Once the eyes' movement reached peak velocity (i.e., during the saccadic suppression period), the location of the target was switched to a different one. Results revealed that, although participants were unaware of the target jump, they quickly and accurately modified the trajectory of their hand (or cursor) so that their movement ended close to the target. Because participants could not refrain from initiating a correction toward the new target location (Day & Lyon, 2000; Johnson, Van Beers, & Haggard, 2002; Pisella et al., 2000), it was concluded that an “automatic pilot” drives fast corrective arm movements (see as well, however, Cameron, Franks, Enns, & Chua, 2007). Similar observations were reported in cursor-jump experiments (Brenner & Smeets, 2003; Brière & Proteau, 2010; Franklin & Wolpert, 2008; Nijhof, 2003; Proteau, Roujoula, & Messier, 2009; Sarlegna & Blouin, 2010; Sarlegna et al., 2003, 2004; Saunders & Knill, 2003, 2004, 2005). 

Typically, in cursor-jump experiments, participants move a cursor shown on a visual display to a target illustrated on the same display. The location of the cursor representing the participant's hand could be translated, for example, by 2 cm soon after movement initiation. Again, participants were unaware of the cursor jump (no participant reported having consciously perceived the jump) but nonetheless corrected their movement in the direction opposite to that of the cursor jump. Because efficient corrections were apparent even for the first—and not consciously perceived—perturbed trial to which participants were exposed (Proteau et al., 2009), it was concluded that these error detection and correction processes did not require learning or adaptation. In addition, because participants could not refrain from initiating a correction in the direction opposite to that of the cursor jump (Franklin & Wolpert, 2008), even when asked to move their hand in the same direction as the cursor jump, it was proposed that the correction was reflexive. 

In the present study, we were interested in the apparent automatic (and even reflex-like) nature of the correction process revealed by target-jump and cursor-jump studies. Recent research by Gritsenko et al. (2009) has revealed that latency of the correction for a target jump was not influenced by the size of the target jump, whereas the correction scaled linearly with the amplitude of the target jump throughout movement execution. Moreover, the gain of the correction process was scaled to the overall movement speed and movement duration of each participant. This indicated that the correction process was coordinated with the anticipated delay remaining until movement completion. What is still unknown is whether the latency and the gain of the correction are also influenced by the size of the target. To our knowledge, this is the first attempt at determining whether a correction will be initiated if the perturbation (here, a cursor jump) does not jeopardize target attainment, and if so, whether the gain of the correction would be affected by target size. 

Experiment 1

Methods

Participants

Thirty-six young adults (18–25 years old) took part in this experiment. They were all naive regarding the goal of the study. All participants were self-declared right-handed and reported normal or corrected-to-normal vision. They took part in a single 40-min experimental session and were paid CAN $10 for their time. The Health Sciences Ethics Committee of the Université de Montréal approved this study. 

Task and apparatus

The task was to move a computer-mouse-like device from a fixed starting position located close to the body toward one of two possible targets located further away from the body. The apparatus is illustrated in Figure 1. It consisted of a table, a computer screen, a headrest, a mirror, and a two-degrees-of-freedom manipulandum. 

Figure 1

 

(Top) View of the apparatus. (Bottom) Illustration of the 25-mm cursor jump for movements aimed at (left) 30-mm and (right) 5-mm targets.

Participants sat in front of the table. The computer screen (cathode ray tube, Mitsubishi, Color Pro Diamond 37 inches, refresh rate 60 Hz) was mounted face down on a ceiling support positioned directly over the table; the computer screen was oriented parallel to the surface of the table. The image of the computer screen was reflected on a mirror placed directly beneath it and parallel to the tabletop and was thus visible to the participant. The distance between the computer screen and the mirror was 20 cm, while the distance between the mirror and the tabletop was 20 cm, which permitted free displacement of the manipulandum on the tabletop. A headrest was affixed on the side of the computer screen. It was aligned with the lateral center of the computer screen, thus standardizing the information reflected in the mirror for all participants. 

The tabletop was covered by a piece of Plexiglas over which a starting base and the manipulandum were affixed. The starting base consisted of a thin strip of Plexiglas glued to the tabletop. It was parallel to the leading edge of the table and had a small indentation on its distal face. This indentation was aligned with the headrest and the participant's midline and served as the starting base for the stylus (see below). The indentation made it easy for the participant to position the stylus at the beginning of each trial. 

The manipulandum consisted of two pieces of rigid Plexiglas (43 cm) joined at one end by an axle. One free end of the manipulandum was fitted with a second axle encased in a stationary base. The other free end of the manipulandum, hereafter called the stylus, was fitted with a small vertical shaft (length: 3 cm, radius: 1 cm), which could be easily gripped by the participant. From the participant's perspective, the far end of the manipulandum was located 40 cm to the left of the starting base and 70 cm in the sagittal plane. Each axle of the manipulandum was fitted with a 13-bit optical shaft encoder (U.S. Digital, model S2-2048, sampled at 500 Hz, angular accuracy of 0.0439°), which enabled us to track the displacement of the stylus online and to illustrate it with a 1:1 ratio on the computer screen. Displacement of the stylus resulted in an identical and superimposed displacement of the cursor on the computer screen. The bottom of the stylus and the bottom of the optical encoder located at the junction of the two arms of the manipulandum were covered with a thin piece of Plexiglas. By lubricating the working surface at the beginning of each experimental session, displacement of the stylus was near frictionless. 

Procedures

Participants used their right dominant hand to move a cursor shown on the computer screen toward targets illustrated on the same screen. Participants performed 10 familiarizing trials, followed by 160 experimental trials. For all trials, the yellow cursor (5 mm in diameter) and the white targets (see below) were illustrated on a black background. The cursor and target remained visible throughout movement execution. 

For the familiarization trials, the targets (5 mm in diameter) were located at 320 mm from the starting base and located at every 10° to the right and left of the starting base (from −50° to −10° and from +10° to +50°). In this phase, one trial was performed toward each one of the 10 targets in random order. The experimental trials were equally divided among two new target locations (7.5° and 15° to the right of the participant's midline) and two target sizes (5 mm or 30 mm in diameter). In all cases, the distance between the starting base and the center of the target was 320 mm. For each target location/target size combination, a cursor jump occurred for 20% of the trials (8 out of 40 trials aimed at each target). Thus, there were 32 cursor-jump trials and 128 no-jump trials overall. Target (size and location) and cursor-jump trials were presented randomly (same order across participants), with one restriction being that cursor-jump trials be separated by at least two no-jump trials. 

Participants were assigned randomly to one of three groups ( n = 12). For the first group, the cursor jump was 5 mm, while for the two remaining groups, it was either 15 mm or 25 mm. The cursor jump occurred 100 ms after movement initiation (plus an additional random delay fluctuating between 14 and 21 ms due to equipment) and translated the position of the cursor perpendicularly to a straight line connecting the starting base and the center of the target. The cursor always jumped to the right. Thus, a correction for the cursor jump would be observed if the position of the stylus migrated closer to the participant's midline for the cursor-jump trials than it did for the no-jump trials. 

At the beginning of each trial, all participants could see the cursor they had to move resting on the starting base. A target was presented once the stylus was stabilized on the starting base for 500 ms. Participants were first asked to gaze at the target and then to initiate a single straight and smooth movement (i.e., not a stop-and-go movement) as they pleased (i.e., not a reaction-time task). They were also required to complete their movement in a movement time ranging between 680 ms and 920 ms (800 ms ± 15%). During data acquisition, movement initiation was detected when the cursor had been moved by 1 mm, whereas movement completion was detected when the cursor did not move by more than 2 mm in a time frame of 100 ms. A preliminary study revealed that the procedure we used to detect movement completion made it difficult for participants to use a stop-and-go strategy. When movements were completed outside the prescribed movement time bandwidth, the participant was reminded of the target movement time. A movement time bandwidth (Proteau et al., 2009; Saunders & Knill, 2003, 2004, 2005) reduces the possibility of different speed–accuracy trade-offs between the different experimental conditions (Fitts, 1954). 

Data reduction

The tangential displacement data of the stylus over time were first smoothed using a second-order recursive Butterworth filter with a cut-off frequency of 10 Hz. The filtered data were then numerically differentiated once using a central finite technique to obtain the velocity profile of the aiming movement, a second time to obtain the acceleration profile, and a third time to obtain a jerk profile. From the kinematic profiles, we determined the end of the movement's primary impulse (Meyer, Abrams, Kornblum, Wright, & Smith, 1988). This occurred when one of the following events was first detected on the kinematic profiles: (a) movement reversal (velocity going from positive to negative), (b) movement lengthening (presence of a secondary movement impulse as indexed by the acceleration profile crossing the zero value for a second time), or (c) a significant disruption in the deceleration profile as indexed by zero-crossing on the jerk profile. For a secondary movement impulse to be considered a discrete correction, its duration had to be of at least 80 ms and its extent had to be of at least 2 mm. Note that less than 3.5% of the trials in all conditions showed a secondary corrective impulse. These trials were withdrawn from all analyses. 

To provide a quick feedback to the participant during data acquisition, movement initiation was detected once the stylus had been moved by 1 mm. However, for the main analyses, movement initiation was defined as the moment at which the tangential velocity of the cursor reached 10 mm/s and was maintained above this value for at least 20 ms. Visual inspection of the data revealed that once the 10 mm/s had been reached, movements were clearly underway. 

To determine the efficacy of the correction for the cursor jump, we determined the end of the movement's primary impulse for all trials (hereafter called movement endpoint). Note that the data reported concern the position of the stylus moved by the participants, not that of the cursor shown on the computer screen. Endpoint frontal and sagittal errors were computed in Cartesian coordinates. These refer to the position of the stylus in relation to the center of the target. The frontal error was the signed difference on the frontal axis (in mm) between the movement endpoint and the target. A positive value indicates a movement ending to the right of the target, and a negative value indicates a movement ending to the left of the target. The sagittal error of a trial was defined as the signed difference between movement endpoint and the target on the sagittal axis (in mm). A positive value indicates that the target had been overshot, and a negative value indicates that it had been undershot. For all groups and conditions, data that differed by more than two standard deviations from the cell mean were excluded from all analyses; less than 3% of the trials were excluded. From these data, we computed the constant and variable aiming errors on the frontal and sagittal dimensions of the task. The constant error is the mean signed difference between the target and endpoint location. It indicates whether participants showed a bias in their movements (too long, too short, to the right, or to the left of the target). The variable error is the within-participant variability in endpoint location. 

To avoid repetition and to facilitate reading of this article, details concerning the dependent variables of interest and the statistical analyses that were computed are defined at the beginning of each subsection of the results presentation. All significant main effects and interactions involving more than two means were broken down using Dunn's procedures. All effects are reported at p < 0.05 (adjusted for the number of comparisons using Bonferonni's technique). Data of one participant in the 25-mm cursor-jump group were excluded from all analyses because they differed by more than two standard deviations from the mean results of that group. 

Results

Cursor jump was not consciously detected

Even after having been debriefed, participants reported that they were not aware that the cursor jumped on some trials. This aspect of the results replicates previous observations (Bédard & Proteau, 2003; Brière & Proteau, 2010; Proteau et al., 2009), even when a cursor jump occurred for a high proportion of the trials (though during saccadic visual suppression, Sarlegna et al., 2003, 2004; see also Saunders & Knill, 2003, 2004, 2005). It should be noted that, on average, the cursor jump occurred at 38.7° (SD = 2.5°) of visual angle, which is quite far in the periphery of the retina and may explain why participants did not notice it. 

Although participants were not aware of the cursor jump, the large corrections observed (see below) indicate that the participants reacted to compensate for it. Because the cursor jump was always to the right, one could argue that the participants might have adapted for the cursor jump by biasing movement planning/execution to the left of the target (direction of the needed correction for a cursor jump). If so, orientation planning for the no-jump trials should progressively migrate to the left of the target, whereas frontal velocity of the stylus for the no-jump trials should progressively decline. Moreover, the planning bias and declining frontal velocity for the no-jump trials should be a function of the cursor jump size. 

To test for this possibility, we contrasted movement orientation and frontal velocity 200 ms after movement initiation had been detected for the first and last four no-jump trials. The data were submitted to a 3 Groups (5-, 15-, or 25-mm cursor jump) × 2 Phases (first vs. last four no-jump trials) × 2 Targets (7.5° and 15°) × 4 Trials ANOVA using repeated measurements on the last three factors. The ANOVAs revealed a significant main effect of Target, F(1, 32) = 739.9 and 731.3, ps < 0.01, for movement orientation and frontal velocity, respectively. These main effects revealed that movements were initiated more to the right, resulting in larger frontal velocity for the 15° target than for the 7.5° target. The ANOVAs also revealed a significant Phase × Trial interaction, F(3, 96) = 4.2 and 7.1, ps < 0.01, for movement orientation and frontal velocity, respectively. The breakdown of these interactions revealed that for the first four no-jump trials, movement orientation and frontal velocity gradually decreased from Trial 1 to Trial 4 (from 10.3° to 9.2°, and from 148.3 mm/s to 120.1 mm/s, respectively), whereas movement orientation and frontal velocity did not significantly differ between the last four no-jump trials (means of 8.4° and 115.4 mm/s, respectively). Finally, for both movement orientation and frontal velocity, neither the Group main effect nor the Group × Phase interaction were significant, all ps > 0.34. The results of these analyses revealed that movement planning/execution was not biased (consciously or unconsciously) by the occurrence of cursor-jump trials. 

Correction for a cursor jump

In the present section, we wanted to determine whether participants corrected their movements for different sizes of cursor jumps and targets, and if so, how efficiently. Examples of stylus trajectories of control and cursor-jump trials aimed toward the different targets are illustrated in Figure 2. This figure clearly illustrates that these participants corrected their movements for the cursor jump. In addition, they show larger corrections for larger cursor jumps (Brière & Proteau, 2010; Sarlegna et al., 2004), regardless of target size. It should also be noted that movement endpoint of no-jump trials are located near the center of the target, regardless of the latter's size. 

Figure 2

 

Stylus trajectories for no-jump (dashed line) and cursor-jump (full line) trials aimed at 5-mm and 30-mm targets located at 7.5° to the right of the starting base. The results are from one participant in each cursor-jump condition.

Mean movement endpoint of all participants are reported in Table 1. Because target location had no significant effect on endpoint accuracy or variability, data of the two target locations were collapsed in the following statistical analyses. Endpoint constant and variable frontal and sagittal errors and movement time data were submitted individually to an ANOVA contrasting 3 Groups (5-, 15-, or 25-mm cursor jump) × 2 Types of trials (no jump vs. cursor jump) and 2 Target sizes (5 mm and 30 mm in diameter) using repeated measurements on the last two factors. 

Table 1

 

Mean (standard deviation) frontal and sagittal distance between the stylus and the center of the target at movement endpoint and movement time as a function of the type of trials, target size, and cursor-jump amplitude. Frontal position: a negative value indicates that the stylus ended to the left of the target center. Sagittal position: a negative value indicates that the stylus ended short of the target center.

Type of trial	Target size	Cursor-jump amplitude
		5 mm	15 mm	25 mm
		Frontal position (mm)
No jump	5 mm	−1.0 (2.3)	−1.9 (2.1)	−3.8 (2.7)
	30 mm	−1.1 (2.7)	−2.0 (3.2)	−2.9 (2.2)
Cursor jump	5 mm	−4.7 (2.5)	−11.5 (1.6)	−20.2 (2.2)
	30 mm	−4.0 (1.6)	−12.1 (1.6)	−18.4 (2.9)
Type of trial	Target size	Cursor-jump amplitude
		5 mm	15 mm	25 mm
		Sagittal position (mm)
No jump	5 mm	0.8 (2.2)	0.9 (4.1)	2.2 (2.1)
	30 mm	−1.8 (2.2)	−0.2 (3.4)	3.0 (4.7)
Cursor jump	5 mm	−0.1 (2.2)	−0.8 (6.6)	−1.2 (2.0)
	30 mm	−1.3 (1.9)	−2.1 (6.3)	0.8 (2.8)
Type of trial	Target size	Cursor-jump amplitude
		5 mm	15 mm	25 mm
		Movement time (ms)
No jump	5 mm	674 (50)	698 (51)	670 (42)
	30 mm	686 (49)	686 (43)	669 (49)
Cursor jump	5 mm	679 (44)	716 (52)	681 (52)
	30 mm	692 (50)	700 (49)	673 (42)

Endpoint constant error. The ANOVA computed on the frontal constant error revealed a significant Group × Type interaction, F(2, 32) = 66.4, p < 0.001. The breakdown of this interaction revealed that endpoint of cursor-jump trials was significantly to the left of endpoint of no-jump trials for the 5-mm (3.3 mm, p < 0.001), 15-mm (9.9 mm, p < 0.001), and 25-mm groups (16.0 mm, p < 0.001). This indicates that participants corrected their movements even for the smallest cursor jump. The interaction revealed that the correction grew larger as cursor-jump size increased. However, in relative terms, participants corrected their movements for 66%, 66%, and 64% of the perturbation for the 5-, 15-, and 25-mm cursor jump, respectively. Neither the Target size main effect, F(1, 32) = 1.60, p = 0.21, nor any interaction involving that factor ( p > 0.13 for all interactions) was significant. In addition, to determine whether the correction became larger with practice, we contrasted the frontal endpoint location of the cursor jump trials in a 3 Groups (5-mm, 15-mm, and 25-mm cursor jump) × 32 Trials (from the first 1st to the 32nd cursor jump) ANOVA using repeated measurements on the second factor. The ANOVA revealed a significant main effect of trial, F(31, 992) = 2.36, p < 0.001. Post hoc comparisons revealed a smaller correction for Trial 1 (8.03 mm) than it did for Trials 6, 7, 11, 14, and 26 (14.1 mm, SD = 0.19 mm) but not for the remaining 26 trials (11.5 mm, SD = 1.14 mm). As illustrated in Figure 3, we found no strong evidence that the size of the correction increased with practice. Therefore, it appears that the cursor jump elicited a strong correction from the very first cursor-jump trial. The ANOVA computed on the sagittal constant error did not reveal any significant main effect or interaction ( p > 0.16 for all main effects and interactions). 

Figure 3

 

Mean trial-by-trial correction for the 5-, 15-, and 25-mm cursor jumps. Note that the correction was present immediately from the first cursor-jump trial and remained largely unchanged across trials.

Endpoint variable error. The ANOVAs computed on frontal and sagittal endpoint variable errors did not reveal any significant main effect or interaction ( p > 0.30 for all main effects and interactions). On average, endpoint variability for the cursor-jump and no-jump trials was 5.71 mm and 5.96 mm on the frontal component of the task and 7.37 mm and 7.76 mm on the sagittal component, respectively. 

Movement time. The ANOVA computed on movement time revealed two significant interactions (see Table 1). First, the breakdown of the Type × Target size interaction, F(2, 32) = 6.741, p = 0.014, revealed significantly longer movement times for the cursor-jump trials than it did for the no-jump trials. This difference was significantly larger when participants aimed at the 5-mm (692 ms vs. 681 ms, p < 0.001) rather than at the 30-mm (688 ms vs. 680 ms, p = 0.051) target. Second, the breakdown of the Group × Type interaction, F(2, 32) = 7.868, p = 0.002, revealed significantly longer movement times for the cursor-jump trials than it did for the no-jump trials for the 25-mm (677 ms vs. 670 ms, p = 0.001) and 15-mm (708 ms vs. 692 ms, p < 0.001) cursor jumps but not for the 5-mm cursor jump (686 ms vs. 680 ms, p = 0.845). 

Characteristics of the correction

To determine when participants initiated a correction for a cursor jump, we used the unfiltered displacement data of the stylus on the frontal axis. As in previous work from our laboratory (Brière & Proteau, 2010; Proteau et al., 2009), we chose to analyze the frontal displacement of the stylus because the cursor jump, and thus the expected correction, largely occurred on this axis. For each participant, we computed a mean trajectory for jump and no-jump trials aimed at each target. Then, at every 20 ms, we computed the difference in location between these mean trajectories. A correction for the cursor jump was detected when (a) cursor-jump trials deviated from the no-jump trials by more than 1 mm in the direction opposite to the cursor jump compared with the position of the cursor at the occurrence of the cursor jump and (b) the deviation continued as movements progressed toward the targets. The 1-mm criterion was chosen arbitrarily. This technique was used by Proteau et al. (2009), who reported latency in the same range as reported by others (Brenner & Smeets, 2003; Saunders & Knill, 2003, 2005). Finally, and perhaps more importantly, it should be remembered that the absolute value of the latency is not our point of interest. What is important is to determine, using the same method, whether the latency differed across experimental conditions. Figure 4 illustrates the difference in the frontal location and velocity of the stylus between the no-jump and cursor-jump trials as a function of time. 

Figure 4

 

Difference in the (upper left panel) frontal displacement and (lower left panel) velocity of the stylus between no-jump and cursor-jump trials as a function of time. Note the similar correction latency for cursor-jump trials regardless of the size of the cursor jump and that of the target. Note that correction latency is reduced by approximately 40 ms when determined from the velocity rather than the displacement data. Note that the gain of the correction (slope) was significantly larger for larger cursor jumps but did not differ across target size. Standard error of the mean is illustrated on the upper and lower right panels.

The latency data were submitted to a 3 Groups (5-, 15-, or 25-mm cursor jump) × 2 Target sizes (5 vs. 30 mm) ANOVA using repeated measurements on the second factor. Note that two participants from the 5-mm group were excluded from this analysis because they showed no evidence of a correction, and correction latency could therefore not be computed. The ANOVA did not reveal any significant main effect ( Fs < 1, p > 0.64) or interaction, F(2, 30) = 1.374, p = 0.27. Mean latency data fluctuated between 137 ms and 172 ms (see Figure 4, upper right panel); these estimations of correction latency are reduced by approximately 40 ms when estimated from the velocity profiles (see Figure 4, lower left panel). The correction was initiated when the cursor reached 11.1° ( SD = 6.2°) of visual angle. 

Finally, to determine whether the initial portion of the correction was related to the size of the cursor jump, we computed the slope of this correction. Specifically, for each participant, we determined by how much the mean trajectory of cursor jump trials deviated from that of no-jump trials 200 ms after a correction had been detected. The ANOVA computed on this dependent variable revealed a significant main effect of cursor-jump size, F(2, 30) = 43.09, p < 0.001. Post hoc comparisons revealed that participants initiated a more abrupt correction as the size of the cursor jump increased (17.15, 35.25, and 63.3 mm/s for the 5-, 15-, and 25-mm cursor jump, respectively; see also Figure 4). However, the slope of this correction did not differ significantly as a function of target size, F(1, 30) < 1. 

Supplementary analyses. The results presented so far revealed slightly longer movement times for the cursor-jump trials than it did for the no-jump trials. A supplementary analysis revealed that this difference in movement time did not reflect a longer movement path for the former than for the latter trials. On the contrary, movement path was slightly longer for the no-jump (325.9 mm) trials than it was for the cursor jump trials (323.1 mm, 321.7 mm, and 321.4 mm for the 5-, 15-, and 25-mm cursor-jump trials, respectively). We also wanted to determine whether larger corrections were associated with a longer correction latency, a longer movement time, or only with a steeper slope of correction, as suggested by the results illustrated in Figure 4. For this purpose, we computed a series of coefficients of correlation across participants ( n = 33); data for each participant were averaged across target location and target size. The results revealed that a longer latency was not predictive of a larger correction for a cursor jump, a steeper slope of correction, or a shorter movement time ( r[31] = −0.03, 0.22, and −0.29, respectively, p > 0.05). However, the size of the correction was closely correlated with the initial slope of the correction ( r[31] = 0.88, p < 0.001), and more modestly, albeit significantly, with movement time ( r[31] = 0.36, p < 0.05). 

Discussion

The goal of this experiment was to determine whether the processes responsible for the detection and correction of unintentional deviations of one's movement aimed at a fixed target are solely based on the characteristics of the deviation or whether they are also adapted to the target's characteristics. We used a perturbation paradigm in which the cursor moved by the participants suddenly jumped between 5 mm and 25 mm to the right of a vector joining the starting base and the target, while the cursor was seen relatively far in the periphery of the retina (∼39° of visual angle). Although participants were not aware of this cursor jump, they soon corrected their movement (130–170 ms) in the opposite direction of the cursor jump. Neither the size, the latency, nor the slope of the correction was modified by target size. 

Our observation that participants modified the trajectory of their movements to counteract the cursor jump concurs with previous observations (Brenner & Smeets, 2003; Brière & Proteau, 2010; Franklin & Wolpert, 2008; Proteau et al., 2009; Sarlegna et al., 2003, 2004; Saunders & Knill, 2003, 2004, 2005). Our results also concur with previous observations in that the corrections were initiated while the cursor was beyond 10° of visual angle (Proteau et al., 2009), its latency was not related to the size of the cursor jump (Gritsenko et al., 2009; Sarlegna et al., 2004, although for target jumps), and the correction did not fully compensate for the size of the cursor jump (Proteau et al., 2009; Sarlegna et al., 2003, 2004; Saunders & Knill, 2003, 2004, 2005). An important new finding of the present experiment is that only the slope of the correction for a cursor jump was a function of the size of the cursor jump. A similar observation was reported recently by Gritsenko et al. (2009) for target jumps of different amplitudes. This suggests that the error detection process put into play in the present study was sensitive enough not only to detect a deviation in the cursor trajectory but also to grade it. 

Our results also concur in one important way with those reported by Brenner and Smeets (2003). These authors tested a condition in which both the cursor moved by the participants and the target jumped simultaneously and in the same direction. In this condition, the cursor and target jumps did not modify the relative position of the cursor and the target. Nonetheless, the authors noted that participants initiated a correction that apparently summed up the correction usually observed when only a target jump took place, as compared to that usually observed in reaction to a cursor jump. This suggests independent correction processes for the correction of the cursor jump and that of the target jump. In the present experiment, we observed the same correction for a cursor jump regardless of target size. Specifically, participants similarly corrected their movements for a 5-mm cursor jump that would have resulted in missing a 5-mm target or that was irrelevant for successfully reaching a 30-mm target. This observation strongly suggests that the correction for a cursor jump is solely related to the size of the induced error, without consideration for target size. This suggests that the correction is based on a comparison between (a) expected sensory consequences that can be derived from a forward model of movement control and (b) actual feedback ensuring that the movement is initiated/continued as planned. However, the correction process did not allow participants to complete their movement on the target. We will return to this point in the General discussion section. 

Before concluding that the error detection and correction processes triggered by the cursor jump have, as sole input, the size of the detected error regardless of movement goal, we must consider that, in the present experiment, participants apparently always aimed at the center of the target (see Figure 2). Endpoint location and variability of no-jump trials did not significantly differ when participants aimed at the 5-mm or 30-mm targets. Thus, it could be argued that the effective target size did not differ between these two experimental conditions, which would explain why participants behave similarly. We completed a second experiment to determine whether a cursor jump would still elicit a correction when participants were asked to plan and execute a movement that would end anywhere within the boundaries of a very large target and for which not correcting for a cursor jump would not jeopardize target attainment. 

Experiment 2

Methods

Participants

Eleven new participants took part in this experiment. They were aged between 18 and 22 years and naive regarding the goal of the study. All participants were self-declared right-handed and reported normal or corrected-to-normal vision. They took part in a single 40-min experimental session and were paid CAN $10 for their time. The Health Sciences Ethics Committee of the Université de Montréal approved this study. 

Task, apparatus, and procedures

We used the same task and apparatus as in Experiment 1. We also used the same procedures as for Group 15 mm in Experiment 1 but with a few exceptions. Participants completed 10 familiarization trials followed by 160 experimental trials. There were 128 no-jump trials and 32 cursor-jump trials (20% perturbation, as in Experiment 1). For all cursor-jump trials, the cursor jumped 100 ms after movement initiation. In all cases, the cursor jumped 15 mm to the right of its actual position. We used a single target. Its center was located at 320 mm from the starting base. The target had an arc of 30° starting at 7.5° and ending at 37.5° to the right of the participants' midline. The target was 30 mm deep. Participants were asked to aim in the general direction of the target and to stop the cursor within its boundaries in approximately 800 ms. We opted not to use a larger circular target than that in Experiment 1 (for example, a 50-mm target) because pilot data indicated that participants were still aiming at the center of the target, which was not the case when we used a wedge-like target. 

Data reduction

As expected, the size of the target and our instructions to aim in its general direction introduced large inter-trial variability (see Figure 5A). This made it difficult to determine appropriate no-jump control trials. To circumvent this difficulty, we used a bootstrapping technique to estimate the properties of movement trajectories for no-jump trials as well as for cursor-jump trials (see Efron & Tibshirani, 1993, for details on the bootstrapping technique; see Georgopoulos, Pellizzer, Poliakov, & Schieber, 1999; Merchant, Naselaris, & Georgopoulos, 2008 for an application on neurophysiological data). Bootstrapping consists of building a series of independent samples from the data. Thus, for each participant, we built 100 samples of no-jump and cursor-jump trials. For each sample, we drew 32 control trials with replacement from our sample of 160 no-jump trials. Similarly, we drew 32 cursor-jump trials with replacement from our sample of 32 cursor-jump trials. For each participant, the mean results obtained for all 100 samples of no-jump and cursor-jump trials were averaged. 

Figure 5

 

(Left) Stylus trajectories for the first four no-jump (red) and cursor-jump (black) trials for one typical participant (A). Distribution of trials (and standard error for the mean) within the target. Difference in the (B) frontal displacement and (C) velocity of the stylus between no-jump and cursor-jump trials as a function of time. Note the similar correction latency for 15-mm cursor jump aimed at a 5-mm target in Experiments 1 and 2. Note that correction latency is reduced by approximately 40 ms when determined from the velocity rather than the displacement data. Note the smaller gain of the correction in Experiment 2 compared to Experiment 1.

Results

Figure 5 (left panel) illustrates the trajectory of the first four no-jump and cursor-jump trials for one typical participant. These results and those illustrated in Figure 5A indicate that participants distributed their movements over a large portion of the target. The results of the frontal coordinates of endpoint location relative to the starting position were submitted to t-tests (two-tailed) contrasting 2 Types of trials (no jump vs. cursor jump). This analysis revealed that cursor-jump trials ended significantly to the left of no-jump trials (95.7 mm vs. 102.2 mm, respectively), t(10) = 4.036, p = 0.002, that is, in a direction opposite to the cursor jump. Thus, participants corrected their movements for the cursor jump (see also Figure 5, left panel). This correction did not result in a significant increase in movement time. On the contrary, cursor-jump trials had slightly shorter movement times than did no-jump trials (813 ms and 824 ms for cursor-jump and no-jump trials, respectively), t(10) = −1.961, p = 0.078. 

To determine whether the characteristics of the correction for the cursor jump differed from those reported in Experiment 1, we contrasted the results of Group 15 mm when aiming at the 5-mm target in Experiment 1 (same cursor-jump size as in the present experiment) with those of the present experiment (bootstrapped data). The results (see Figures 5B and 5C) revealed that the latency of the correction did not differ significantly across experiments (147 ms and 140 ms for Experiments 1 and 2, respectively), t(21) = 0.27, p = 0.79. However, the slope (35.9 mm/s vs. 24.4 mm/s), t(21) = 2.58, p = 0.017, and ultimately the size (9.6 mm vs. 6.5 mm), t(21) = 2.07, p = 0.051, of this correction were significantly larger in Experiment 1 than they were in Experiment 2. 

Discussion

The results of this second experiment are straightforward. Participants corrected their movements for the cursor jump, even though this correction was not needed to perform the task successfully. The latency of this correction did not differ from that reported in Experiment 1. This suggests that the error detection process engaged a corrective process, even if the correction was irrelevant considering the ultimate goal. However, the more gradual correction observed for the same cursor-jump size in the present experiment, as compared to that of Experiment 1, suggests that the correction process takes into consideration the “urgency” of the correction. In Experiment 1, this urgency was reflected by steeper corrections for the larger cursor jumps. In the present experiment, no correction was needed to reach the target; therefore, it was more gradual than and not as large as it was in Experiment 1. 

General discussion

In the present study, we wanted to determine whether error detection and correction processes based on visual information and relative to the displacement of one's hand toward a fixed target were solely driven by the stimulus that elicited them or also by the target's characteristics. The results of the two experiments reported in the present paper revealed that the latency of the correction triggered by the cursor jump was not influenced by the target's characteristics. However, the execution of the correction was affected significantly by both the size of the cursor jump and the target's characteristics. 

Correction latency

Our observation that correction latency was not influenced by the size of the cursor jump concurs with results previously reported by Sarlegna et al. (2004) for a cursor jump and by Gritsenko et al. (2009) for a target jump. In addition, as in Proteau et al. (2009), it indicates that even small deviations of the cursor from its intended path were detected by a very efficient error detection process that acts outside the central visual field (>10° of visual angle; see also Abahnini, Proteau, & Temprado, 1997; Abhanini & Proteau, 1999; Bard, Hay, & Fleury, 1985, 1990; Bédard & Proteau, 2001, 2003; Blouin, Bard, Teasdale, & Fleury, 1993; Blouin, Gauthier, Vercher, & Cole, 1996; Blouin, Teasdale, Bard, & Fleury, 1993; Paillard, 1980; Paillard & Amblard, 1985; Proteau, Boivin, Linossier, & Abahnini, 2000). The initiation of this correction does not appear to require attention from the participants because a correction was initiated while participants were not aware of the cursor jump. The slope of the correction was a function of the size of the cursor jump (see also Gritsenko et al., 2009; Sarlegna et al., 2004). This indicates that the error detection process put into play in the present study was sensitive enough not only to detect a difference between the expected and actual cursor position (Brenner & Smeets, 2003; Shabbott & Sainburg, 2009) and trajectory (Saunders & Knill, 2004) but also to grade it. 

Correction execution

Although the correction that was initiated to counteract the cursor jump was graded for its size, the results of Experiment 2 indicated that the slope of this correction was also influenced by target size. Specifically, the slope of the correction for a 15-mm cursor jump was significantly larger in Experiment 1 than it was in Experiment 2. It should be remembered that increasing the diameter of the target from 5 mm to 30 mm ( Experiment 1) was not sufficient for observing this difference. This indicates that a difference in target size that has been shown to influence movement time (Fitts, 1954), movement variability, or the proportion of trials requiring a secondary corrective impulse (Tinjust & Proteau, 2009) when the cursor is located close to the target was not sufficient to influence the slope of the correction in Experiment 1. This suggests that the error correction process revealed in Experiment 2 has a different origin than that usually observed when the cursor and target are seen close to one another in central vision. 

Because the latency of the correction was not influenced by target size, whereas target size modified the slope of the correction, this suggests that detection of an error and execution of a correction are based on distinct processes. This position concurs with recent observations reported by Glover, Miall, and Rushworth (2005). In that study, participants reached to grasp a small or large illuminated cylinder. During the reach, the cylinder could change from large to small or small to large. When repetitive transcranial magnetic stimulation (rTMS) was applied over the left intraparietal sulcus of the parietal cortex, Glover et al. observed a disruption in the online correction for a change in target size, but only when the rTMS was applied concurrent with the initiation of the adjustment and not when it was applied after the adjustment had already begun. This proposition fits well with the minimal intervention principle advocated by Todorov and Jordan (2002; see also Sarlegna & Blouin, 2010) and suggests that initiation of a correction is automatic but that its execution is functional. 

An alternative interpretation of our finding could be that the gain of the error correction process triggered by the cursor jump could be modulated as a function of the task's constraints. When these constraints are relatively loose, as in Experiment 2, it could be that the gain of the error correction process is set relatively low. Again, this proposition is in agreement with the minimal intervention principle advocated by Todorov and Jordan (2002). Specifically, effective target size was relatively small in Experiment 1, whereas it was quite large in Experiment 2, which explains why steeper corrections were observed in the former experiment than they were in the latter experiment. In turn, this suggests that one has some control over which component(s) of the movement to more loosely or strictly control. 

Partial corrections for cursor jumps

The results of this study concur with all previous research using a cursor-jump paradigm (Brière & Proteau, 2010; Proteau et al., 2009; Sarlegna et al., 2004; Saunders & Knill, 2003, 2004, 2005) in indicating that there is a limit to the size of the correction that can be performed by this apparently attention-free online corrective process. For instance, the correction for the cursor jump compensated for only between 45% (Sarlegna et al., 2004) and 80% (Saunders & Knill, 2003) of the imposed bias. In the present study, this correction was approximately 65%. Proteau et al. (2009) have suggested that the limits of the correction process could be exceeded when the error falls outside the normal variability of one's movement. Although this could be the case with relatively large cursor jumps, this explanation does not fit well with the partial correction observed in Experiment 1 for a 5-mm cursor jump. Rather, it could be that asking participants to complete their movement in a single motion (present study; Brière & Proteau, 2010; Proteau et al., 2009), considering the movement completed when their hand was still traveling at a relatively high velocity (100 mm/s; Saunders & Knill, 2003), or asking participants to produce very fast movements (Bédard & Proteau, 2003; Sarlegna & Blouin, 2010; Sarlegna et al., 2003, 2004) prevented them from completing their corrections. These results all contrast with those reported by Gritsenko et al. (2009), who showed that the gain of the correction for a target jump was set for movements to end up on the target. This difference might indicate that the correction processes revealed in cursor-jump and target-jump experiments differ in some important ways. However, a cursor jump—but not a target jump—creates a conflict between the seen position of the cursor and the felt position of the hand, which may have limited the participant's ability to produce a complete correction. 

Conclusion

When a reaching movement deviates from its intended path, quick and graded corrections are initiated while one's hand (or a cursor) is seen in peripheral vision. This correction occurs even when it is not required to complete the task successfully. This suggests that the initiation of the correction is based on a comparison between the intended and actual movement trajectories, without much consideration for the ultimate movement goal. On the contrary, the execution of the correction takes into consideration not only the size of the error that has been detected but also the task's constraints. These constraints likely influence the gain of the correction processes put into play early after movement initiation, when one's movement deviates from its intended trajectory. 

Acknowledgments

This work was supported by a Discovery grant (L.P.) provided by the Natural Sciences and Engineering Research Council of Canada. 

Commercial relationships: none. 

Corresponding author: Luc Proteau. 

Email: [e-mail]. 

Address: Département de Kinésiologie, Université de Montréal, C.P. 6128, Succursale Centre Ville, Montréal H3C 3J7, Canada.