Abstract

Cognitive control is engaged to facilitate stimulus–response mappings for novel, complex tasks and supervise performance in unfamiliar, challenging contexts—processes supported by pFC, ACC, and posterior parietal cortex. With repeated task practice, however, the appropriate task set can be selected in a more automatic fashion with less need for top–down cognitive control and weaker activation in these brain regions. One model system for investigating cognitive control is the ocular motor circuitry underlying saccade production, with basic prosaccade trials (look toward a stimulus) and complex antisaccade trials (look to the mirror image location) representing low and high levels of cognitive control, respectively. Previous studies have shown behavioral improvements on saccade tasks after practice with contradictory results regarding the direction of functional MRI BOLD signal change. The current study presented healthy young adults with prosaccade and antisaccade trials in five mixed blocks with varying probability of each trial type (0%, 25%, 50%, 75%, or 100% anti vs. pro) at baseline and posttest MRI sessions. Between the scans, participants practiced either the specific probability blocks used during testing or only a general 100% antisaccade block. Results indicated an overall reduction in BOLD activation within pFC, ACC, and posterior parietal cortex and across saccade circuitry for antisaccade trials. The specific practice group showed additional regions including ACC, insula, and thalamus with an activation decrease after practice, whereas the general practice group showed a little change from baseline in those clusters. These findings demonstrate that cognitive control regions recruited to support novel task behaviors were engaged less after practice, especially with exposure to mixed task contexts rather than a novel task in isolation.

INTRODUCTION

Cognitive control is a term encompassing multiple supervisory processes that coordinate sensory and motor functions to flexibly adapt behavior to current goals (Diamond, 2013; Braver, Paxton, Locke, & Barch, 2009; Fernandez-Duque & Knight, 2008; Cole & Schneider, 2007; Miller & Cohen, 2001). This often requires facilitation of new, unfamiliar task rules over habitual, familiar responses. To learn the appropriate stimulus–response associations for a new task or context, a cognitive control “scaffolding” system involving pFC, ACC, and posterior parietal cortex (PPC) is recruited across cognitive paradigms (Brass, Ullsperger, Knoesche, von Cramon, & Phillips, 2005; Chein & Schneider, 2005; Kelly & Garavan, 2005). With repeated exposure to or practice of a task, however, activation measured with the BOLD fMRI signal typically is reduced in cognitive control and attentional networks over time (Chein & Schneider, 2005). Presumably, novel tasks recruit large neural populations to establish an unfamiliar task set, while over time, less activity in these neurons or activity only in a more focal set of neurons is sufficient to produce the correct response (Kelly & Garavan, 2005; Poldrack, 2000). This occurs once task rules are learned and relevant connections are strengthened, so that neural networks can perform the requisite processing more efficiently (Bassett, Yang, Wymbs, & Grafton, 2015).

One model system for studying cognitive control and changes after task practice is the ocular motor circuitry associated with saccade production. Saccades are rapid eye movements made to foveate a location of interest in the visual field, and saccade tasks include visually guided prosaccades (PSs; look toward a newly appearing stimulus) and volitionally driven antisaccades (ASs; look to the mirror image location of a stimulus). The need to suppress a saccade toward the stimulus, to transform the spatial location of the stimulus into the opposite visual field, and to facilitate an endogenous saccade to a blank location during an AS trial require higher levels of cognitive control than a PS trial (Hutton, 2008; Munoz & Everling, 2004). AS trials typically have slower RTs and higher error rates (uninhibited saccades towards the stimulus) than PS trials (e.g., Pierce & McDowell, 2016b; Pierce, McCardel, & McDowell, 2015; Herweg et al., 2014; Weiler & Heath, 2012; Ethridge, Brahmbhatt, Gao, McDowell, & Clementz, 2009; Brown, Goltz, Vilis, Ford, & Everling, 2006). Basic saccade circuitry showing activation during fMRI tasks has been thoroughly characterized in the literature and includes visual cortex, PPC, FEF and supplementary eye field (SEF), thalamus, BG, and cerebellum; greater strength or extent of activation in these regions, and recruitment of additional cognitive control regions such as pFC and ACC, is observed during AS tasks (Jamadar, Fielding, & Egan, 2013; Reuter, Kaufmann, Bender, Pinkpank, & Kathmann, 2010; Curtis & Connolly, 2008; McDowell, Dyckman, Austin, & Clementz, 2008; Brown, Vilis, & Everling, 2007; Dyckman, Camchong, Clementz, & McDowell, 2007; Ford, Goltz, Brown, & Everling, 2005; Curtis & D'Esposito, 2003).

Cognitive control of saccade tasks is dependent on the context in which they are performed (Ethridge et al., 2009; Dyckman et al., 2007); one means of manipulating task context is by varying the proportion or probability of PSs and ASs within a mixed saccade block. As the probability of a given trial type decreases, its task set is less active, and participants make more errors and have slower RTs (Pierce & McDowell, 2016a; Chiau et al., 2011; Massen, 2004). Alternately, a high probability of an AS trial can slow down RTs of both trial types, presumably because of lingering motor inhibition (Pierce & McDowell, 2016b; Pierce et al., 2015; Barton, Greenzang, Hefter, Edelman, & Manoach, 2006; Dorris, Pare, & Munoz, 2000). The current study explores the interaction between trial type probability and saccade task practice to investigate how increased experience with a challenging context can improve behavior.

Previous studies on the practice of saccade tasks include a study (Dyckman & McDowell, 2005) that examined the behavioral effects of practicing ASs, PSs, or fixation for 2 weeks. All groups performed the tasks more quickly (faster RTs) after practice, whereas only individuals who practiced the AS task had fewer AS errors. Critically, those who practiced the PS task produced more AS errors at the end of the study, likely because the PS task reinforced the visually driven response that constitutes an error in the AS task. In a similar study examining BOLD activation changes in a blocked-design AS task after a week of AS, PS, or fixation practice (Lee et al., 2013), only the AS practice group demonstrated consistent decreases in saccade circuitry activation at posttest, despite a lack of significant changes in behavior. Furthermore, a recent study (Jamadar, Johnson, Clough, Egan, & Fielding, 2015) investigated the practice of interleaved saccade trials using event-related fMRI. Participants performed mixed blocks of PS and AS trials in the scanner and then practiced a shortened version of the task daily for 2 weeks before a second fMRI session. Their results indicated a reduction in AS RT with no change in error rate and markedly increased BOLD activation across saccade circuitry for both trial types at the second session. These studies offer conflicting evidence as to how saccade behavior and BOLD activation change with practice, perhaps because of the nature of the task practice performed in each study. The current study thus utilized two distinct practice groups with either mixed PS and AS practice or all-AS practice to help clarify the previous results while using the same test paradigm for both groups.

This study combined practice of saccade tasks with a trial type probability manipulation to examine how AS and PS trials in different task contexts were affected by either specific or general saccade practice. Participants completed an initial fMRI session with five blocks of varying probability of AS to PS trials (0%, 25%, 50%, 75%, or 100%) and then were divided into two practice groups for 4 days of saccade practice in the laboratory before an identical posttest fMRI session. Half of the participants were in the “specific” practice group—they practiced the same mixed probability blocks each day that they performed during the fMRI testing sessions. The other half of the participants were in the “general” group and practiced blocks of 100% AS trials. It was predicted that, as the contexts of the saccade trials became more familiar over the course of the practice days, less cognitive control would be required to activate the appropriate task set. Thus, even in blocks with a low probability of a given trial type, the correct task set would be selected in a more automatic manner with less interference from the more probable task set.

Those who practiced the cognitive control components required by the specific probability blocks were expected to improve saccade behavior the most at posttest as a result of repeated exposure to all mixed contexts. In contrast, those who practiced general cognitive control with only AS trials were expected to show less improvement in blocks with low AS probability because they were not trained on conditions with frequent (or any) PS trials. PS behavior was not expected to change significantly for either group.

As participants in the current study became more familiar with the probability contexts and/or AS task set, it was predicted that the need for effortful supervision by control processes would be reduced. Therefore, cognitive control regions associated with facilitating task rules and adapting to context, such as pFC and ACC, were expected to show a decrease in BOLD activation at posttest. Furthermore, participants who practiced the specific probability contexts were predicted to reduce activation in these cognitive control regions and saccade circuitry to a greater degree overall than the general AS practice group. Their familiarity with the mixed contexts and expectation of the probability manipulation at posttest should have better prepared them for both trial types even in the low probability blocks. Overall, the increased exposure and extended training with saccade tasks should have allowed participants to better utilize contextual information and activate the appropriate task set more efficiently.

METHODS

Participants

Sixty-five undergraduate students were recruited from the University of Gerogia Psychology Department online research pool and given course credit for their participation (as described in Pierce & McDowell, 2016b). Thirty-three individuals fulfilled the exclusion criteria or voluntarily opted out before completing the study, leaving 32 right-handed participants who experienced no current major psychiatric disorders or substance abuse, had no metal implants, and had normal or corrected-to-normal vision (via self-report). Sixteen of the participants (mean age = 19 years, SD = 1 year; five men) were assigned to the general AS practice group; and 16 (mean age = 20 years, SD = 5 years; five men), to the specific probability practice group (described below). All participants provided written informed consent, and activities were approved by the institutional review board of the University of Georgia.

Task Design

Participants were presented with five rapid event-related saccade blocks with varying probability of occurrence of an AS trial (relative to a PS trial): 0%, 25%, 50%, 75%, and 100% AS. The blocks consisted of 60 saccade trials presented according to the overall probability (e.g., 25% AS block had 15 AS and 45 PS trials), of which participants were not informed, interspersed with jittered fixation periods. All stimuli consisted of a 1° gray shape presented on a black background, and central fixation appeared for 2000–8000 msec (average = 3500 msec) between trials. For saccade trials, the trial type cue was illuminated around the cross for 500 msec (for PSs, a square; for ASs, a diamond). This was followed by a blank screen for 200 msec (“gap” presentation) and, finally, the peripheral stimulus at 5° or 10° right or left of the center for 800 msec. Two peripheral stimulus eccentricities were included to reduce the likelihood of participants anticipating the response location and preparing a motor response in advance (data collapsed across amplitudes for analyses). The practice tasks followed the same timing scheme as the fMRI scans but were generated separately so that the exact trial timing and order differed between practice and MRI sessions; five unique 100% AS blocks were created for the general practice group.

Procedure

Participants attended an initial session where they completed demographic surveys and were screened for exclusion criteria. During this session, participants were introduced to the saccade paradigm by performing 20 mixed PS (“look as quickly and accurately as possible toward the peripheral stimulus”) and AS (“look to the mirror image location of the stimulus, same distance from the center”) trials. During the subsequent MRI session, participants were positioned on the scanner table with the head secured. A high-resolution (T1-weighted) structural scan was obtained first for each participant, followed by the functional (T2*-weighted) scans. Stimuli were displayed using the Presentation software (Neurobehavioral Systems, Albany, CA) and a dual mirror system attached to the head coil that allowed the participant to view a projection screen at his or her feet and researchers to monitor the participant's eye. Right eye pupil position was sampled at 60 Hz (IView X MRI-LR system; SensoMotoric Instruments, Teltow, Germany) and recorded for offline analysis. Before beginning the saccade tasks, eye position was calibrated using IView's 5-point calibration and an in-house horizontal calibration.

After completing the baseline MRI, participants were assigned to one of two practice groups. Each group practiced five saccade blocks a day for 4 weekdays in the laboratory. The first group practiced the five “specific” probability blocks (0%, 25%, 50%, 75%, and 100% AS) in counterbalanced order across days. The second group practiced only “general” AS blocks (100% AS). On the practice days, participants were seated in the laboratory with their head in a chin rest in front of the display monitor (Samsung 40-in. LCD), and the eye-tracking apparatus (EyeLink II; SR Research, Kanata, Ontario, Canada) was placed on their head and adjusted. Eye position relative to the monitor was calibrated with both EyeLink's built-in 9-point calibration and an in-house horizontal 7-point calibration. Stimuli were displayed in a darkened room while the relative pupil positions of both eyes were sampled and digitized at 500 Hz. After the 4 practice days, both groups completed a posttest fMRI session with the same scan order as at baseline.

Imaging Parameters

MR images were collected on a 3-T GE Signa Excite HDx system (General Electric Medical Systems, Milwaukee, WI) at the University of Georgia Bio-Imaging Research Center. A high-resolution anatomical image was collected using a T1-weighted 3-D fast spoiled gradient echo sequence (echo time = 3 msec, flip angle = 20°, field of view = 240 mm × 240 mm, matrix size = 256 × 256, 150 axial slices, in-slice resolution = 0.94 × 0.94 mm, slice thickness = 1.2 mm, scan time = 6 min 32 sec). The functional scans were collected using a T2*-weighted gradient-echo EPI sequence (echo time = 30 msec, repetition time = 2000 msec, flip angle = 90°, field of view = 220 mm × 220 mm, matrix size = 64 × 64, 33 interleaved oblique slices aligned with the AC–PC plane, in-slice resolution = 3.4 × 3.4 mm, slice thickness = 4 mm, slice gap = 0 mm, 4 dummy volumes for magnet stabilization, 158 volumes, scan time = 5 min 24 sec).

Analyses

Eye position data were analyzed using custom scripts written in MATLAB (The MathWorks, Natick, MA). Trials were manually scored for the initial direction of response (eye movements with velocities surpassing 20°/sec were classified as saccades) and correct response RT. Error rate was defined as the number of trials with an initial saccade in the incorrect direction out of the total number of analyzable trials multiplied by 100; RT was defined as the time from the appearance of the peripheral circle to the initiation of the first saccade. Trials with no response, blinks at stimulus onset or anticipatory saccades (faster than 90-msec RT or during the gap window), or with insufficient data quality due to loss of pupil tracking were excluded from further analyses. Of 150 trials per condition, an average of 137 (baseline)/130 (posttest) AS trials and 138 (baseline)/132 (posttest) PS trials were included in the analysis. Statistical analyses on eye movement metrics were performed using SAS Version 5.1 (SAS Institute Inc., Cary, NC) and SPSS Version 22 (IBM Corp., Armonk, NY) software packages. To quantify the effects of practice on saccade behavior, a 2 × 2 × 4 (Practice group [specific/general] × Session [baseline/posttest] × Probability block) ANOVA was performed on error rate and correct trial RT. For AS trials, the levels of probability were 25%, 50%, 75%, and 100% AS, and for PS trials, they were 0%, 25%, 50%, and 75% AS.

Functional MRI data were analyzed using the AFNI software package (Cox, 1996, 2012) with initial processing steps including slice-timing correction, volume alignment to account for participant motion, resampling to a 4-mm3 voxel grid, spatial standardization to a Talairach template, spatial smoothing (4-mm FWHM Gaussian kernel), and voxel-wise scaling to a mean of 100 (making task activation equivalent to percent BOLD signal change from fixation baseline). On the basis of each participant's behavioral responses, a general linear model was fit with stimulus regressors for correct ASs, error ASs, correct PSs, and error PSs. Regressors of no interest were also included for baseline drift (linear, quadratic, cubic) and rotational movement in the x, y, and z planes. Coefficients for correct trials then were entered into a 2 × 2 × 4 (Practice group [specific/general] × Session [baseline/posttest] × Probability block) ANOVA for AS and PS trials separately. For AS trials, the levels of probability were 25%, 50%, 75%, and 100% AS, and for PS trials, they were 0%, 25%, 50%, and 75% AS. A t test was performed for the effect of practice session (posttest minus baseline) to illustrate the direction of changes for AS and PS trials, and then correlations were calculated between clusters with BOLD activation reductions after practice and average individual changes in RT to assess whether shortened processing time was driving the observed results. Finally, a 2 × 2 × 2 × 3 (Practice group [specific/general] × Trial type [anti/pro] × Practice session [baseline/posttest] × Probability block [25%, 50%, 75%]) ANOVA was performed on the three mixed saccade blocks with AS and PS trials together, specifically to determine whether trial type effects typically observed (i.e., greater AS activation) changed with practice.

All group analyses were confined to regions within a custom brain mask created from the average gray matter segmentation from all participants' anatomical images using FSL's FAST (FMRIB Software Libraries Automated Segmentation Tool; Zhang, Brady, & Smith, 2001) in conjunction with putamen, caudate, and thalamus regions as defined by AFNI's Talairach–Tournoux atlas (Talairach & Tournoux, 1988). To protect against false positives resulting from multiple comparisons across voxels, a clustering method derived from Monte Carlo simulations was applied to the group maps (AFNI's 3dclustsim). With a voxel-wise p < .01, a family-wise α < .05 was preserved by clusters with a minimum of 23 voxels.

RESULTS

Behavioral Responses

The ANOVA on AS behavior indicated a significant effect of Practice session on both AS error rate (F(1, 30) = 5.7, p < .05, η2 = .16) and correct RT (F(1, 30) = 14.7, p < .01, η2 = .33). There were fewer errors and faster RTs at posttest than baseline (Table 1). There was also a significant main effect of Probability for both measures (error rate: F(3, 90) = 4.6, p < .01, η2 = .13; RT: F(3, 90) = 6.7, p < .001, η2 = .19) and a significant Practice session × Probability interaction for RT (F(3, 90) = 5.4, p < .01, η2 = .15). Blocks with a higher AS probability had fewer errors and slower RTs, with this pattern being most dominant at baseline. There were no significant main or interaction effects involving practice group.

Table 1. 

Error Rate and Correct RT at Baseline and Posttest MRI Sessions for All Participants

Error Rate (%)RT (msec)
BaselinePosttestBaselinePosttest
Antisaccades 
25% AS 26.4 (18.6) 18.0 (18.1) 266 (42) 257 (41) 
50% AS 27.2 (14.7) 20.2 (14.9) 273 (42) 256 (44) 
75% AS 23.2 (13.5) 17.7 (12.4) 283 (42) 258 (48) 
100% AS 18.4 (13.7) 17.0 (11.7) 290 (40) 257 (49) 
 
Prosaccades 
0% AS 2.0 (2.2) 2.1 (2.7) 191 (23) 188 (20) 
25% AS 3.1 (4.4) 3.9 (4.6) 192 (23) 191 (23) 
50% AS 3.5 (4.0) 3.7 (5.3) 201 (27) 194 (23) 
75% AS 2.2 (5.1) 4.1 (5.4) 199 (30) 197 (26) 
Error Rate (%)RT (msec)
BaselinePosttestBaselinePosttest
Antisaccades 
25% AS 26.4 (18.6) 18.0 (18.1) 266 (42) 257 (41) 
50% AS 27.2 (14.7) 20.2 (14.9) 273 (42) 256 (44) 
75% AS 23.2 (13.5) 17.7 (12.4) 283 (42) 258 (48) 
100% AS 18.4 (13.7) 17.0 (11.7) 290 (40) 257 (49) 
 
Prosaccades 
0% AS 2.0 (2.2) 2.1 (2.7) 191 (23) 188 (20) 
25% AS 3.1 (4.4) 3.9 (4.6) 192 (23) 191 (23) 
50% AS 3.5 (4.0) 3.7 (5.3) 201 (27) 194 (23) 
75% AS 2.2 (5.1) 4.1 (5.4) 199 (30) 197 (26) 

Values are given as mean (SD).

The ANOVA on PS behavior showed a significant effect of Probability on correct RT (F(3, 90) = 7.0, p < .001, η2 = .19). Blocks with a higher PS probability had faster RTs. There was an interaction between Practice session and Group on error rate (F(1, 30) = 6.6, p < .05, η2 = .18), with the general practice group committing more PS errors at posttest, although average PS error rates were always less than 5% of trials. No other effects reached statistical significance.

BOLD Signal Changes

Main Effect of Practice Session

For AS trials, the Practice group × Session × Probability ANOVA revealed a main effect of Session in seven clusters that included the canonical saccade circuitry (Figure 1; Table 2). For PS trials, there was a main effect of Session in six clusters, many of which were located in similar regions as the AS trial clusters (Table 3). t tests demonstrated that all clusters showed decreased BOLD activation from baseline to posttest. Notably, bilateral pFC showed positive task activation for AS trials at baseline but no significant activation at posttest; the bilateral parietal/temporal clusters showed the same pattern for PS trials. The average change in activation across blocks from baseline to posttest for AS trials did not correlate with the average change in AS RT in any of these regions. For PS trials, several clusters (bilateral cuneus, r = −.38; right insula, r = −.37; left middle occipital gyrus [MOG], r = −.39; p < .05) showed a negative correlation between average percent signal change and RT change from baseline to posttest (the remaining clusters showed a similar relationship that did not reach significance). Figure 2 shows the correlation between change in average PS activation in bilateral cuneus and RT, where individuals who had slower PS RTs at posttest than baseline showed the greatest decrease in percent signal change at posttest relative to baseline, implying that the activation decreases after practice were not due simply to shortened neural processing time.

Figure 1. 

(Left) Maps of the practice session t test for AS and PS trials: Blue colors represent a decrease in percent signal change from the baseline to posttest session; no regions showed a significant increase after practice. AS trials resulted in significant clusters in (A1) bilateral FEF/SEF/ACC/bilateral insula, (A2) left inferior parietal lobule/precuneus, (A3) right inferior parietal lobule/precuneus, (A4) right MFG/superior frontal gyrus, (A5) left MFG/superior frontal gyrus, (A6) left cuneus, and (A7) left lingual gyrus. PS trials resulted in significant clusters in (P1) left inferior parietal lobule/superior temporal gyrus/postcentral gyrus/cingulate, (P2) right inferior parietal lobule/superior temporal gyrus, (P3) left precentral gyrus, (P4) right insula, (P5) bilateral cuneus, and (P6) left MOG. Brain images are displayed in radiological convention (right is left) with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Right) Plots of the percent signal change from all clusters for each trial type showing decreased activation from baseline to posttest sessions (cluster details provided in Tables 2 and 3).

Figure 1. 

(Left) Maps of the practice session t test for AS and PS trials: Blue colors represent a decrease in percent signal change from the baseline to posttest session; no regions showed a significant increase after practice. AS trials resulted in significant clusters in (A1) bilateral FEF/SEF/ACC/bilateral insula, (A2) left inferior parietal lobule/precuneus, (A3) right inferior parietal lobule/precuneus, (A4) right MFG/superior frontal gyrus, (A5) left MFG/superior frontal gyrus, (A6) left cuneus, and (A7) left lingual gyrus. PS trials resulted in significant clusters in (P1) left inferior parietal lobule/superior temporal gyrus/postcentral gyrus/cingulate, (P2) right inferior parietal lobule/superior temporal gyrus, (P3) left precentral gyrus, (P4) right insula, (P5) bilateral cuneus, and (P6) left MOG. Brain images are displayed in radiological convention (right is left) with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Right) Plots of the percent signal change from all clusters for each trial type showing decreased activation from baseline to posttest sessions (cluster details provided in Tables 2 and 3).

Table 2. 

Description of the Significant Clusters for AS Trials in the Practice Session × Probability × Practice Group ANOVA

AS Anatomical RegionPeak StatisticxyzSize (Voxels)
Practice Session t Test t     
A1. Bilateral FEF, SEF, ACC, bilateral insula −6.7 46 792 
A2. Left inferior parietal, precuneus −5.4 −14 −61 52 216 
A3. Right inferior parietal, precuneus −5.5 54 −41 20 365 
A4. Right MFG/superior frontal gyrus −6.1 22 47 32 127 
A5. Left MFG/superior frontal gyrus −5.0 −34 27 32 75 
A6. Left cuneus, precuneus −3.7 −14 −81 20 30 
A7. Left cuneus, lingual gyrus −4.4 −6 −69 79 
 
Probability Effect F     
A8. ACC/medial frontal gyrus 5.7 11 40 23 
A9. Right MFG 6.6 22 −1 44 30 
A10. Left MFG 5.8 −26 11 44 25 
A11. Left precentral/postcentral gyrus 6.4 −30 −25 64 23 
A12. Left precuneus, cuneus 7.6 −14 −69 40 67 
A13. Left middle/superior temporal, angular gyrus 7.3 −42 −57 20 41 
A14. Right cuneus, MOG 8.4 30 −85 20 32 
A15. Left MOG 6.7 −30 −77 26 
 
Group × Session Interaction      
A16. Right precentral gyrus, superior temporal gyrus, insula 25.1 38 24 733 
A17. Left precentral gyrus, superior temporal gyrus, insula 27.3 −46 −13 439 
A18. Right ACC, medial frontal gyrus 19.8 10 44 40 
A19. Left cuneus 17.4 −2 −93 24 
A20. Bilateral thalamus 19.5 18 −13 16 79 
A21. Left lingual gyrus, declive 14.4 −6 −77 −4 31 
 
Group × Probability Interaction      
A22. Left posterior cingulate, culmen 9.9 −6 −41 23 
 
Session × Probability Interaction      
A23. Bilateral cingulate, precuneus 7.8 −41 40 44 
A24. Right precuneus 11.0 14 −69 36 26 
AS Anatomical RegionPeak StatisticxyzSize (Voxels)
Practice Session t Test t     
A1. Bilateral FEF, SEF, ACC, bilateral insula −6.7 46 792 
A2. Left inferior parietal, precuneus −5.4 −14 −61 52 216 
A3. Right inferior parietal, precuneus −5.5 54 −41 20 365 
A4. Right MFG/superior frontal gyrus −6.1 22 47 32 127 
A5. Left MFG/superior frontal gyrus −5.0 −34 27 32 75 
A6. Left cuneus, precuneus −3.7 −14 −81 20 30 
A7. Left cuneus, lingual gyrus −4.4 −6 −69 79 
 
Probability Effect F     
A8. ACC/medial frontal gyrus 5.7 11 40 23 
A9. Right MFG 6.6 22 −1 44 30 
A10. Left MFG 5.8 −26 11 44 25 
A11. Left precentral/postcentral gyrus 6.4 −30 −25 64 23 
A12. Left precuneus, cuneus 7.6 −14 −69 40 67 
A13. Left middle/superior temporal, angular gyrus 7.3 −42 −57 20 41 
A14. Right cuneus, MOG 8.4 30 −85 20 32 
A15. Left MOG 6.7 −30 −77 26 
 
Group × Session Interaction      
A16. Right precentral gyrus, superior temporal gyrus, insula 25.1 38 24 733 
A17. Left precentral gyrus, superior temporal gyrus, insula 27.3 −46 −13 439 
A18. Right ACC, medial frontal gyrus 19.8 10 44 40 
A19. Left cuneus 17.4 −2 −93 24 
A20. Bilateral thalamus 19.5 18 −13 16 79 
A21. Left lingual gyrus, declive 14.4 −6 −77 −4 31 
 
Group × Probability Interaction      
A22. Left posterior cingulate, culmen 9.9 −6 −41 23 
 
Session × Probability Interaction      
A23. Bilateral cingulate, precuneus 7.8 −41 40 44 
A24. Right precuneus 11.0 14 −69 36 26 

Coordinates refer to the Talairach–Tournoux atlas and voxel size is based on 4 mm3 voxels.

Table 3. 

Description of the Significant Clusters for PS Trials in the Practice Session × Probability × Practice Group ANOVA

Prosaccade Anatomical RegionPeak StatisticxyzSize (Voxels)
Practice Session t Test t     
P1. Left inferior parietal lobule, postcentral gyrus, insula, superior temporal gyrus, cingulate −6.0 −38 −33 40 749 
P2. Right inferior parietal lobule, postcentral gyrus, superior temporal gyrus −4.4 42 −25 16 323 
P3. Left precentral gyrus −3.9 −58 −1 24 30 
P4. Right insula −4.2 50 −5 40 
P5. Bilateral cuneus, lingual gyrus, right fusiform gyrus −4.5 14 −65 12 214 
P6. Left fusiform, MOG −4.1 −38 −65 28 
 
Probability Effect F     
P7. Bilateral ACC, medial frontal gyrus 6.4 −6 48 24 
P8. Right MFG/medial frontal gyrus, precentral gyrus 6.2 26 −13 47 38 
P9. Left MFG/medial frontal gyrus, precentral gyrus 9.2 −18 −5 48 70 
P10. Left precuneus, superior parietal lobule 10.5 −22 −57 28 58 
P11. Left inferior parietal lobule, postcentral gyrus 7.5 −30 −41 32 41 
P12. Right precuneus, cuneus 8.3 14 −65 28 49 
P13. Right inferior frontal gyrus/MFG, precentral gyrus 8.1 46 32 55 
P14. Left fusiform gyrus, MOG, declive 12.1 −38 −53 −4 249 
P15. Right fusiform gyrus, MOG, declive 12.3 26 −77 −12 428 
 
Group Effect      
P16. Right MFG 16.1 38 23 44 27 
P17. Left cerebellum 11.2 −34 −69 −32 24 
 
Group × Session Interaction      
P18. Right postcentral/precentral gyrus 14.1 42 −21 52 34 
P19. Right superior/middle temporal gyrus 15.5 50 −9 58 
Prosaccade Anatomical RegionPeak StatisticxyzSize (Voxels)
Practice Session t Test t     
P1. Left inferior parietal lobule, postcentral gyrus, insula, superior temporal gyrus, cingulate −6.0 −38 −33 40 749 
P2. Right inferior parietal lobule, postcentral gyrus, superior temporal gyrus −4.4 42 −25 16 323 
P3. Left precentral gyrus −3.9 −58 −1 24 30 
P4. Right insula −4.2 50 −5 40 
P5. Bilateral cuneus, lingual gyrus, right fusiform gyrus −4.5 14 −65 12 214 
P6. Left fusiform, MOG −4.1 −38 −65 28 
 
Probability Effect F     
P7. Bilateral ACC, medial frontal gyrus 6.4 −6 48 24 
P8. Right MFG/medial frontal gyrus, precentral gyrus 6.2 26 −13 47 38 
P9. Left MFG/medial frontal gyrus, precentral gyrus 9.2 −18 −5 48 70 
P10. Left precuneus, superior parietal lobule 10.5 −22 −57 28 58 
P11. Left inferior parietal lobule, postcentral gyrus 7.5 −30 −41 32 41 
P12. Right precuneus, cuneus 8.3 14 −65 28 49 
P13. Right inferior frontal gyrus/MFG, precentral gyrus 8.1 46 32 55 
P14. Left fusiform gyrus, MOG, declive 12.1 −38 −53 −4 249 
P15. Right fusiform gyrus, MOG, declive 12.3 26 −77 −12 428 
 
Group Effect      
P16. Right MFG 16.1 38 23 44 27 
P17. Left cerebellum 11.2 −34 −69 −32 24 
 
Group × Session Interaction      
P18. Right postcentral/precentral gyrus 14.1 42 −21 52 34 
P19. Right superior/middle temporal gyrus 15.5 50 −9 58 

Coordinates refer to the Talairach–Tournoux atlas, and voxel size is based on 4-mm3 voxels.

Figure 2. 

Correlation between PS RT and percent signal change difference (posttest minus baseline). The percent signal change was extracted from the cluster in the bilateral cuneus (labeled P5 in the PS trails in Figure 1). Negative values indicate a reduction in percent signal change or faster RTs after practice, and data points represent individual participant averages across the four probability blocks. Participants who had slower RTs at posttest than baseline had the greatest reduction in BOLD percent signal change. A similar relationship was observed in PS clusters in the right insula and left MOG (labeled P4 and P6); AS trials showed no significant correlations.

Figure 2. 

Correlation between PS RT and percent signal change difference (posttest minus baseline). The percent signal change was extracted from the cluster in the bilateral cuneus (labeled P5 in the PS trails in Figure 1). Negative values indicate a reduction in percent signal change or faster RTs after practice, and data points represent individual participant averages across the four probability blocks. Participants who had slower RTs at posttest than baseline had the greatest reduction in BOLD percent signal change. A similar relationship was observed in PS clusters in the right insula and left MOG (labeled P4 and P6); AS trials showed no significant correlations.

Main Effect of Trial Type Probability

For AS trials, the main effect of Probability resulted in eight clusters (Figure 3; Table 2) that showed two general patterns. The first pattern observed in the left precuneus, ACC, and right middle frontal gyrus (MFG) showed the greatest percent signal change in the block with the fewest ASs (25% AS) and the least percent signal change in the block with the most ASs (100% AS). The second pattern observed in the right and left MOG, left angular gyrus, left MFG, and left precentral gyrus showed weak or negative activation for the block with the fewest ASs (25% AS) and positive or no signal change for the other blocks. For PS trials, an effect of Probability was observed in nine clusters (Figure 3; Table 3) with the greatest percent signal change in the block with the fewest PSs (75% AS) and the least percent signal change in the blocks with the most PSs (0% and 25% AS). This effect is similar to the first pattern described for ASs with the direction of response strength reversed because of the reversed PS trial type probability (i.e., high AS probability means low PS probability).

Figure 3. 

(Top) Maps of the Probability main effect for AS and PS trials; brighter colors represent higher F values. AS trials resulted in significant clusters in (A8) ACC, (A9) right MFG, (A10) left MFG, (A11) left precentral/postcentral gyrus, (A12) left precuneus, (A13) left middle/superior temporal gyrus, (A14) right MOG, and (A15) left MOG. PS trials resulted in significant clusters in (P7) ACC, (P8) right MFG/medial frontal gyrus, (P9) left MFG/medial frontal gyrus, (P10) left precuneus, (P11) left inferior parietal lobule, (P12) right precuneus, (P13) right inferior frontal gyrus/MFG, (P14) left MOG, and (P15) right MOG. Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Bottom left) Plots of the BOLD signal from AS trials in two cognitive control clusters; in ACC and left precuneus, activation was the highest in the low AS probability blocks (25% and 50% AS) and lowest in the all-AS block (100% AS). (Bottom right) Plots of BOLD signal from PS trials in two PS clusters similar to those locations shown for AS trials; PS activation showed a comparable response to probability with the strongest BOLD signal in the low PS probability block (75% AS). Details of all clusters are provided in Tables 2 and 3.

Figure 3. 

(Top) Maps of the Probability main effect for AS and PS trials; brighter colors represent higher F values. AS trials resulted in significant clusters in (A8) ACC, (A9) right MFG, (A10) left MFG, (A11) left precentral/postcentral gyrus, (A12) left precuneus, (A13) left middle/superior temporal gyrus, (A14) right MOG, and (A15) left MOG. PS trials resulted in significant clusters in (P7) ACC, (P8) right MFG/medial frontal gyrus, (P9) left MFG/medial frontal gyrus, (P10) left precuneus, (P11) left inferior parietal lobule, (P12) right precuneus, (P13) right inferior frontal gyrus/MFG, (P14) left MOG, and (P15) right MOG. Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Bottom left) Plots of the BOLD signal from AS trials in two cognitive control clusters; in ACC and left precuneus, activation was the highest in the low AS probability blocks (25% and 50% AS) and lowest in the all-AS block (100% AS). (Bottom right) Plots of BOLD signal from PS trials in two PS clusters similar to those locations shown for AS trials; PS activation showed a comparable response to probability with the strongest BOLD signal in the low PS probability block (75% AS). Details of all clusters are provided in Tables 2 and 3.

Main Effect of Practice Group

The main effect of Practice group did not yield any significant clusters for AS trials. For PS trials, the effect of Practice group showed two significant clusters in the right MFG and left cerebellum (Table 3). These regions had greater activation for those in the general practice group than those in the specific practice group.

Practice Group × Session Interaction

The interaction between Practice group and Session resulted in six significant clusters for AS trials and two clusters for PS trials (Figure 4; Tables 2 and 3). Across regions and trial types, the specific Practice group showed a strong decrease in activation from baseline to posttest. In contrast, the general practice group showed either no change or a slight increase in task activation after practice.

Figure 4. 

(Left) Maps of the Practice group × Session interaction for AS and PS trials; brighter colors represent higher F values. AS trials resulted in significant clusters in (A16) right precentral/superior temporal gyrus, (A17) left precentral/superior temporal gyrus, (A18) right ACC, (A19) left cuneus, (A20) thalamus, and (A21) left lingual gyrus. PS trials resulted in significant clusters in (P18) right postcentral/precentral gyrus and (P19) right superior/middle temporal gyrus. Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Top right) Plot of the percent signal change from AS trials adjusted for baseline values in the right ACC cluster (all clusters showed a similar pattern) indicating that the specific practice group decreased activation from baseline to posttest, whereas the general practice group did not change. (Bottom right) Plot of percent signal change from PS trials in the right precentral gyrus showing a similar effect with a slight increase at posttest for the general practice group. Details of all clusters are provided in Tables 2 and 3.

Figure 4. 

(Left) Maps of the Practice group × Session interaction for AS and PS trials; brighter colors represent higher F values. AS trials resulted in significant clusters in (A16) right precentral/superior temporal gyrus, (A17) left precentral/superior temporal gyrus, (A18) right ACC, (A19) left cuneus, (A20) thalamus, and (A21) left lingual gyrus. PS trials resulted in significant clusters in (P18) right postcentral/precentral gyrus and (P19) right superior/middle temporal gyrus. Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. (Top right) Plot of the percent signal change from AS trials adjusted for baseline values in the right ACC cluster (all clusters showed a similar pattern) indicating that the specific practice group decreased activation from baseline to posttest, whereas the general practice group did not change. (Bottom right) Plot of percent signal change from PS trials in the right precentral gyrus showing a similar effect with a slight increase at posttest for the general practice group. Details of all clusters are provided in Tables 2 and 3.

Other Interactions

For AS trials, the interaction between practice group and AS probability resulted in one small cluster in the posterior cingulate (Table 2). This region showed greater signal change for the general practice group in the block with the fewest ASs (25% AS), greater signal change for the specific practice group in the block with all ASs (100% AS), and little difference between groups in the other blocks (50% and 75% AS). AS trials also resulted in a Session × Probability interaction in two clusters in the cingulate and right precuneus. These regions had strong percent signal change in the all-AS block (100% AS) and weak signal change in the low-probability AS block (25% AS) at baseline, with the opposite pattern at posttest. Prosaccade trials did not show any other two-way interactions. Neither AS nor PS trials showed a significant three-way interaction between practice group, session, and probability.

Antisaccades versus Prosaccades

The ANOVA on AS and PS trials in the mixed blocks (25%, 50%, and 75% AS) showed a significant interaction between Trial type and Practice session in one cluster in the right supramarginal gyrus. In this region, positive AS activation was evident at baseline, but not at posttest, whereas PSs showed no significant activation at either MRI session (Figure 5; Table 4). A t test of AS versus PS BOLD activation at posttest revealed that the typically greater activation for AS trials (as seen in these participants at baseline; Pierce & McDowell, 2016b) remained significant in most of the canonical saccade circuitry including FEF, SEF, PPC/precuneus, BG, and cerebellum. One small cluster in the occipital cortex showed the opposite pattern with a greater signal change for PSs than ASs.

Figure 5. 

AS versus PS trials within the three mixed blocks (25%, 50%, and 75% AS); brighter colors represent higher statistic values. (Left) Interaction effect of trial type and practice session showing one cluster in the (1) right supramarginal gyrus. The extracted BOLD percent signal change illustrates that AS trials had positive activation at baseline and no task-related activation at posttest, whereas PS trials had no task activation at either session in this region. (Right) t test of AS minus PS BOLD activation during the posttest session showing five clusters with greater activation for AS trials (orange) in (2) bilateral FEF, SEF, ACC, bilateral thalamus, and BG; (3) bilateral precuneus; (4) right MFG; (5) right inferior temporal gyrus; and (6) bilateral culmen/declive (cerebellum). One cluster in the (7) right inferior occipital gyrus/MOG had greater activation for PS trials (blue). Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. Details of all clusters are provided in Table 4.

Figure 5. 

AS versus PS trials within the three mixed blocks (25%, 50%, and 75% AS); brighter colors represent higher statistic values. (Left) Interaction effect of trial type and practice session showing one cluster in the (1) right supramarginal gyrus. The extracted BOLD percent signal change illustrates that AS trials had positive activation at baseline and no task-related activation at posttest, whereas PS trials had no task activation at either session in this region. (Right) t test of AS minus PS BOLD activation during the posttest session showing five clusters with greater activation for AS trials (orange) in (2) bilateral FEF, SEF, ACC, bilateral thalamus, and BG; (3) bilateral precuneus; (4) right MFG; (5) right inferior temporal gyrus; and (6) bilateral culmen/declive (cerebellum). One cluster in the (7) right inferior occipital gyrus/MOG had greater activation for PS trials (blue). Brain images are displayed in radiological convention with functional results (voxel-wise p < .01, family-wise α < .05) overlaid on the average of the standardized anatomical images from all participants. Details of all clusters are provided in Table 4.

Table 4. 

Description of the Significant Clusters for the Comparison of AS and PS Trials within the Three Mixed Blocks (25%, 50%, and 75% AS)

Anatomical RegionPeak StatisticxyzSize (Voxels)
Trial Type × Practice Session Interaction F     
1. Right supramarginal gyrus 21.3 58 −45 32 25 
 
Anti > Pro Posttest t     
2. Bilateral FEF, SEF, ACC, thalamus, BG, insula 10.2 −26 −9 56 1393 
3. Bilateral precuneus/superior parietal lobule/inferior parietal lobule 9.0 22 −65 36 927 
4. Right MFG 4.7 38 19 28 31 
5. Right middle/inferior temporal gyrus 4.0 46 −53 −4 38 
6. Bilateral culmen/declive 5.4 34 −37 −32 504 
 
Pro > Anti Posttest      
7. Right inferior occipital gyrus/MOG −4.5 30 −85 −4 27 
Anatomical RegionPeak StatisticxyzSize (Voxels)
Trial Type × Practice Session Interaction F     
1. Right supramarginal gyrus 21.3 58 −45 32 25 
 
Anti > Pro Posttest t     
2. Bilateral FEF, SEF, ACC, thalamus, BG, insula 10.2 −26 −9 56 1393 
3. Bilateral precuneus/superior parietal lobule/inferior parietal lobule 9.0 22 −65 36 927 
4. Right MFG 4.7 38 19 28 31 
5. Right middle/inferior temporal gyrus 4.0 46 −53 −4 38 
6. Bilateral culmen/declive 5.4 34 −37 −32 504 
 
Pro > Anti Posttest      
7. Right inferior occipital gyrus/MOG −4.5 30 −85 −4 27 

Coordinates refer to the Talairach–Tournoux atlas, and voxel size is based on 4-mm3 voxels.

DISCUSSION

Cognitive control is recruited according to current goals to facilitate performance of a novel task set in an unfamiliar context, yet with practice, the task set can be strengthened and executed with fewer demands for top–down supervision. In this study, the effects of specific and general saccade task practice on the performance of blocks with varying trial type probabilities were investigated in healthy young adults using BOLD fMRI. The specific group practiced the same mixed probability blocks as assessed during testing, and the general group practiced only all-AS blocks. The analysis of these fMRI data indicated that the specific practice group decreased BOLD activation strongly from baseline to posttest in several clusters, whereas the general practice group showed a little change in those clusters. In addition, widespread decreases in activation were observed for all participants across saccade circuitry for both trial types after practice, although ASs still recruited greater saccade network activation than prosaccades in most regions.

Behavioral results demonstrated that, regardless of whether a participant practiced the specific mixed probability task or the general AS task, both AS error rate and RT decreased at posttest, whereas prosaccade behavior changed minimally. Consistent with previous reports (Pierce & McDowell, 2016a, 2016b; Massen, 2004), an effect of probability was observed in AS error rate and RT as well as prosaccade RT: Blocks with a high probability of AS trials had lower AS error rate and slower RTs for both trial types. Taken together, these results suggest that, although both types of saccade practice were effective at improving behavior and generally reducing activation in saccade circuitry, the specific practice group's more extensive exposure to the probability contexts allowed them to create more efficient task representations that led to reduced activation of additional brain regions at posttest than the general practice group. Therefore, training on a difficult task in isolation is not as effective at reducing the demand for cognitive control as practicing the task within a mixed context.

Reduced BOLD Activation after Saccade Practice

In the comparison between baseline and posttest fMRI scans for both AS and prosaccade trials, widespread saccade circuitry showed a reduction in BOLD activation. AS trials showed more extensive significant decreases than prosaccade trials, with decreases in bilateral PPC (precuneus, inferior parietal lobule), cuneus, FEF/SEF, insula, ACC, and bilateral pFC. pFC clusters, in particular, showed positive activation at baseline and no significant task activation at posttest for AS trials. This supports the notion of a “scaffolding” cognitive control system for learning novel stimulus–response mappings (Chein & Schneider, 2005; Kelly & Garavan, 2005). The AS response (look to the mirror image location) requires an unfamiliar transformation of visual stimulus information and engages pFC (and other regions) to facilitate a volitional saccade over a visually driven response. With practice, this mapping is strengthened, and less top–down control is required to correctly execute the necessary inversion.

This is consistent with a previous report of reduced AS activation after AS practice (Lee et al., 2013), although it conflicts with the increases observed by Jamadar et al. (2015). Differences in the extent of practice and corresponding behavior (see below) may have contributed to these contrasting results if saccade-related activation undergoes both increases and decreases during practice over time. Interestingly, the secondary analysis in the current study comparing both trial types within the mixed blocks at posttest indicated that the greater cognitive demands of AS trials still engaged saccade circuitry more than for PS trials after practice. The inherent asymmetry between the two trial types, therefore, may be maintained despite practice-related changes in activation, to support unique AS task functions. Only the right supramarginal gyrus showed an interaction between trial type and practice session with a large decrease in activation for AS trials and no change for PS trials. This region is associated with orienting of attention (Corbetta & Shulman, 2002) and may have been recruited especially to support the novel AS task set at baseline but disengaged at posttest after practice strengthened the AS response.

For PS trials, a practice-related decrease was observed in parietal cortex activation extending into the left postcentral/precentral gyrus (FEF) and cingulate/SEF as well as in the cuneus. In these core saccade visual–motor regions, decreased BOLD activation at posttest may be due to increased network efficiency, strengthened task set representations, or reduced attentional engagement. The basic neurophysiological coding of the saccades themselves is unlikely to have changed with practice because eye movements to certain locations in visual space are generated many times every day and are already well established. The association of these eye movements with the current task setup, stimuli, and overall context, however, could have improved with practice, resulting in more effective recruitment of the saccade network with minimal top–down control. Because the current study included only groups with active task practice, another consideration is that some changes in activation may have been caused by test–retest effects (Ettinger et al., 2003) or because participants were more comfortable with the MRI/eye tracking environment in general and not because of saccade task practice per se.

The pattern of behavioral changes demonstrates that saccade practice was effective at improving AS error rates and reducing AS RTs for both groups at posttest, while diminishing the difference among probability blocks as measured at baseline. This is broadly consistent with previous saccade practice studies: Dyckman and McDowell (2005) reported fewer AS errors for those who practiced ASs and faster AS RTs overall, and Jamadar et al. (2015) reported faster RTs for both ASs and PSs after mixed saccade practice, although there was no change in AS error rates. The error rates in that study, however, were lower (∼10%) than those reported here (∼20%), and participants may have approached peak performance during the baseline session. Interestingly, in the current results, the average individual change in AS RT after practice did not correlate with changes in AS BOLD activation, but PS RT change negatively correlated with PS activation change. Participants who reduced their RT with practice thus were not driving BOLD signal changes simply because of shorter neural processing time (Poldrack, 2000; D'Esposito et al., 1997)—those who had the largest increase in PS RT showed the greatest reduction in BOLD activation, perhaps indicating that weakened visual stimulus processing slowed motor output. Nonetheless, PSs still showed stronger activation than ASs at posttest in a cluster in the inferior occipital gyrus/MOG. This suggests that participants inhibited visual processing to a greater degree during AS trials to prevent stimulus-driven directional errors.

Effects of Trial Type Probability

The analysis of trial type probability revealed effects in posterior temporal/occipital, parietal, and frontal regions for both trial types. In most of these clusters, both trial types showed the greatest BOLD percent signal change in blocks with a low probability of that trial type and the least signal change in blocks with a high trial type probability. This pattern suggests that, when a trial type was unexpected, its task set was poorly activated and greater effort or attention (and therefore stronger BOLD activation) was required to correctly execute the saccade. In the study of Pierce and McDowell (2016b), we reported trial type probability effects from these participants at baseline, focusing only on the three mixed blocks (25%, 50%, and 75% AS). In that analysis, strong probability effects were observed for PS trials, with activation in the low PS probability block (75% AS) reaching AS levels in the precuneus, occipital/temporal cortex, medial frontal gyrus, and right MFG (Pierce & McDowell, 2016b). With the inclusion of the single trial type blocks and posttest data in the current analysis, probability effects for AS trials were detected in similar regions. Thus, with a broader range of probabilities and more trials, complex AS trials showed sensitivity to context as well as basic PS trials.

There also were interactions of trial type probability with practice session and with practice group for AS trials. The interaction with session indicated that, at posttest, clusters in the parietal cortex showed an increase in BOLD activation for the low AS probability block and a decrease for the all-AS block. Participants' knowledge at posttest of the probability manipulation (due to their initial exposure or daily task practice) may have reduced demands for visual attention and spatial transformation processes during blocks with greater likelihood of an AS trial. The interaction between probability and practice group resulted in a small cluster in the posterior cingulate with greater activation for the general practice group in the low AS probability block and weaker activation in the all-AS block. This may reflect differences in visual processing based on the practice groups' differing exposure to each probability block.

Specific versus General Saccade Task Practice

Although the main effect of Practice session highlighted many regions that showed decreased BOLD activation after practice for all participants, the Practice group × Session interaction identified clusters in which the groups showed divergent responses. The specific practice group showed a clear decrease in all of these clusters (including insula/superior temporal gyrus, right precentral gyrus, and ACC) over time, whereas the general practice group showed no change or a small increase in activation. This pattern implies that the specific practice group's greater experience with the different probability blocks allowed them to reduce demand for the cognitive control and attention processes required to select the appropriate task set for both trial types (Bassett et al., 2015). As with the main effect of practice session, AS trials showed more extensive regions of significance, whereas PS trials showed a similar direction of effect in more circumscribed clusters. The insula, in particular, showed a marked effect for AS trials and has been shown previously to be activated in response to the greater cognitive and motor demands of a novel AS task (Jamadar et al., 2013). ACC also showed this same pattern and has been related to conflict monitoring during task performance, predicting outcome likelihood, and signaling that greater cognitive control should be exerted by pFC (Brown, 2013; Carter & van Veen, 2007; Botvinick, Braver, Barch, Carter, & Cohen, 2001; Braver, Barch, Gray, Molfese, & Snyder, 2001; MacDonald, Cohen, Stenger, & Carter, 2000). Thus, for the specific practice group, more familiarity with the probability contexts at posttest may have diminished the effective conflict or prediction error on low probability and AS trials and the corresponding need for ACC recruitment.

The lack of reduction in BOLD activation in ACC for the general practice group, however, suggests that trial conflict was not reduced to the same degree by practice of the challenging AS task by itself. This kind of single trial type practice presumably did not account for the additional task switching or working memory demands in the mixed saccade blocks (cf. contextual interference in motor training; Lage et al., 2015; Magill & Hall, 1990). The unexpected occurrence of a low probability trial type and the need to switch task sets in these blocks evidently engaged some task selection processes at least as much at posttest as during the baseline scan. This limited familiarity, and the stronger attentional demands associated with it may account for the fact that the general practice group showed greater activation overall for PS trials in the right MFG and left cerebellum. Together, these effects suggest that the specific practice group was able to more thoroughly strengthen task set and context representations, increase saccade circuitry efficiency, and reduce the demand for cognitive control, despite behavioral improvements being observed for both groups.

Conclusions

This study investigated the impact of specific versus general saccade practice on behavior and BOLD signal activation in a mixed saccade task. The task included blocks of randomly interleaved AS and PS trials with varying trial type probabilities. Both AS error rate and RT decreased after practice, as did the BOLD activation across saccade circuitry for both trial types. Cognitive control regions such as pFC, ACC, and PPC showed positive task activation at baseline for AS trials that then diminished or disappeared at posttest as the novel task set was strengthened and could be executed in a more automated manner. The trial type probability manipulation led to increased activation for low probability trials in visual and motor pathways, with similar effects for both simple and complex trial types. Finally, the practice groups showed opposing changes after practice in several regions, with the specific practice group decreasing BOLD activation at posttest and the general practice group changing little. This likely resulted from the increased familiarity with the different probability contexts that the specific practice group gained. Greater exposure to a mixed context afforded additional training with switching and maintaining both task sets within a block, whereas general practice reinforced a single mode of responding that was not as beneficial in the mixed contexts. These findings generally demonstrate that, with practice of a complex task in varying contexts, participants can learn and strengthen new task sets and reduce demand for the cognitive control supervision of task performance.

Acknowledgments

The authors thank the UGA Bio-Imaging Research Center, K. Mason for technical assistance, and B. McCardel, J. Coppiano, A. Rodrigue, R. Hart, C. Burton, D. Schaeffer, and S. Arkin for help with data collection.

Reprint requests should be sent to Jennifer E. McDowell, Department of Psychology, University of Georgia, 125 Baldwin Street, Athens, GA 30602, or via e-mail: jemcd@uga.edu.

REFERENCES

Barton
,
J. J.
,
Greenzang
,
C.
,
Hefter
,
R.
,
Edelman
,
J.
, &
Manoach
,
D. S.
(
2006
).
Switching, plasticity, and prediction in a saccadic task-switch paradigm
.
Experimental Brain Research
,
168
,
76
87
.
Bassett
,
D. S.
,
Yang
,
M.
,
Wymbs
,
N. F.
, &
Grafton
,
S. T.
(
2015
).
Learning-induced autonomy of sensorimotor systems
.
Nature Neuroscience
,
18
,
744
751
.
Botvinick
,
M. M.
,
Braver
,
T. S.
,
Barch
,
D. M.
,
Carter
,
C. S.
, &
Cohen
,
J. D.
(
2001
).
Conflict monitoring and cognitive control
.
Psychological Review
,
108
,
624
652
.
Brass
,
M.
,
Ullsperger
,
M.
,
Knoesche
,
T. R.
,
von Cramon
,
D. Y.
, &
Phillips
,
N. A.
(
2005
).
Who comes first? The role of the prefrontal and parietal cortex in cognitive control
.
Journal of Cognitive Neuroscience
,
17
,
1367
1375
.
Braver
,
T. S.
,
Barch
,
D. M.
,
Gray
,
J. R.
,
Molfese
,
D. L.
, &
Snyder
,
A.
(
2001
).
Anterior cingulate cortex and response conflict: Effects of frequency, inhibition and errors
.
Cerebral Cortex
,
11
,
825
836
.
Braver
,
T. S.
,
Paxton
,
J. L.
,
Locke
,
H. S.
, &
Barch
,
D. M.
(
2009
).
Flexible neural mechanisms of cognitive control within human prefrontal cortex
.
Proceedings of the National Academy of Sciences, U.S.A.
,
106
,
7351
7356
.
Brown
,
J. W.
(
2013
).
Beyond conflict monitoring cognitive control and the neural basis of thinking before you act
.
Current Directions in Psychological Science
,
22
,
179
185
.
Brown
,
M. R. G.
,
Goltz
,
H. C.
,
Vilis
,
T.
,
Ford
,
K. A.
, &
Everling
,
S.
(
2006
).
Inhibition and generation of saccades: Rapid event-related fMRI of prosaccades, antisaccades, and no-go trials
.
Neuroimage
,
33
,
644
659
.
Brown
,
M. R. G.
,
Vilis
,
T.
, &
Everling
,
S.
(
2007
).
Frontoparietal activation with preparation for antisaccades
.
Journal of Neurophysiology
,
98
,
1751
1762
.
Carter
,
C. S.
, &
van Veen
,
V.
(
2007
).
Anterior cingulate cortex and conflict detection: An update of theory and data
.
Cognitive, Affective & Behavioral Neuroscience
,
7
,
367
379
.
Chein
,
J. M.
, &
Schneider
,
W.
(
2005
).
Neuroimaging studies of practice-related change: fMRI and meta-analytic evidence of a domain-general control network for learning
.
Brain Research. Cognitive Brain Research
,
25
,
607
623
.
Chiau
,
H. Y.
,
Tseng
,
P.
,
Su
,
J. H.
,
Tzeng
,
O. J.
,
Hung
,
D. L.
,
Muggleton
,
N. G.
, et al
(
2011
).
Trial type probability modulates the cost of antisaccades
.
Journal of Neurophysiology
,
106
,
515
526
.
Cole
,
M. W.
, &
Schneider
,
W.
(
2007
).
The cognitive control network: Integrated cortical regions with dissociable functions
.
Neuroimage
,
37
,
343
360
.
Corbetta
,
M.
, &
Shulman
,
G. L.
(
2002
).
Control of goal-directed and stimulus-driven attention in the brain
.
Nature Reviews Neuroscience
,
3
,
201
215
.
Cox
,
R. W.
(
1996
).
AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages
.
Computers in Biomedical Research
,
29
,
162
173
.
Cox
,
R. W.
(
2012
).
AFNI: What a long strange trip it's been
.
Neuroimage
,
62
,
743
747
.
Curtis
,
C. E.
, &
Connolly
,
J. D.
(
2008
).
Saccade preparation signals in the human frontal and parietal cortices
.
Journal of Neurophysiology
,
99
,
133
145
.
Curtis
,
C. E.
, &
D'Esposito
,
M.
(
2003
).
Success and failure suppressing reflexive behavior
.
Journal of Cognitive Neuroscience
,
15
,
409
418
.
D'Esposito
,
M.
,
Zarahn
,
E.
,
Aguirre
,
G. K.
,
Shin
,
R. K.
,
Auerbach
,
P.
, &
Detre
,
J. A.
(
1997
).
The effect of pacing of experimental stimuli on observed functional MRI activity
.
Neuroimage
,
6
,
113
121
.
Diamond
,
A.
(
2013
).
Executive functions
.
Annual Review of Psychology
,
64
,
135
168
.
Dorris
,
M. C.
,
Pare
,
M.
, &
Munoz
,
D. P.
(
2000
).
Immediate neural plasticity shapes motor performance
.
Journal of Neuroscience
,
20
,
RC52
.
Dyckman
,
K. A.
,
Camchong
,
J.
,
Clementz
,
B. A.
, &
McDowell
,
J. E.
(
2007
).
An effect of context on saccade-related behavior and brain activity
.
Neuroimage
,
36
,
774
784
.
Dyckman
,
K. A.
, &
McDowell
,
J. E.
(
2005
).
Behavioral plasticity of antisaccade performance following daily practice
.
Experimental Brain Research
,
162
,
63
69
.
Ethridge
,
L. E.
,
Brahmbhatt
,
S.
,
Gao
,
Y.
,
McDowell
,
J. E.
, &
Clementz
,
B. A.
(
2009
).
Consider the context: Blocked versus interleaved presentation of antisaccade trials
.
Psychophysiology
,
46
,
1100
1107
.
Ettinger
,
U.
,
Kumari
,
V.
,
Crawford
,
T. J.
,
Davis
,
R. E.
,
Sharma
,
T.
, &
Corr
,
P. J.
(
2003
).
Reliability of smooth pursuit, fixation, and saccadic eye movements
.
Psychophysiology
,
40
,
620
628
.
Fernandez-Duque
,
D.
, &
Knight
,
M.
(
2008
).
Cognitive control: Dynamic, sustained, and voluntary influences
.
Journal of Experimental Psychology: Human Perception and Performance
,
34
,
340
355
.
Ford
,
K. A.
,
Goltz
,
H. C.
,
Brown
,
M. R.
, &
Everling
,
S.
(
2005
).
Neural processes associated with antisaccade task performance investigated with event-related fMRI
.
Journal of Neurophysiology
,
94
,
429
440
.
Herweg
,
N. A.
,
Weber
,
B.
,
Kasparbauer
,
A.
,
Meyhofer
,
I.
,
Steffens
,
M.
,
Smyrnis
,
N.
, et al
(
2014
).
Functional magnetic resonance imaging of sensorimotor transformations in saccades and antisaccades
.
Neuroimage
,
102
,
848
860
.
Hutton
,
S. B.
(
2008
).
Cognitive control of saccadic eye movements [review]
.
Brain and Cognition
,
68
,
327
340
.
Jamadar
,
S. D.
,
Fielding
,
J.
, &
Egan
,
G. F.
(
2013
).
Quantitative meta-analysis of fMRI and PET studies reveals consistent activation in fronto-striatal-parietal regions and cerebellum during antisaccades and prosaccades
.
Frontiers in Psychology
,
4
,
749
.
Jamadar
,
S. D.
,
Johnson
,
B. P.
,
Clough
,
M.
,
Egan
,
G. F.
, &
Fielding
,
J.
(
2015
).
Behavioral and neural plasticity of ocular motor control: Changes in performance and fMRI activity following antisaccade training
.
Frontiers in Human Neuroscience
,
9
,
653
.
Kelly
,
A. M.
, &
Garavan
,
H.
(
2005
).
Human functional neuroimaging of brain changes associated with practice
.
Cerebral Cortex
,
15
,
1089
1102
.
Lage
,
G. M.
,
Ugrinowitsch
,
H.
,
Apolinario-Souza
,
T.
,
Vieira
,
M. M.
,
Albuquerque
,
M. R.
, &
Benda
,
R. N.
(
2015
).
Repetition and variation in motor practice: A review of neural correlates
.
Neuroscience & Biobehavioral Reviews
,
57
,
132
141
.
Lee
,
J.
,
Park
,
C.
,
Dyckman
,
K. A.
,
Lazar
,
N. A.
,
Austin
,
B. P.
,
Li
,
Q.
, et al
(
2013
).
Practice-related changes in neural activation patterns investigated via wavelet-based clustering analysis
.
Human Brain Mapping
,
34
,
2276
2291
.
MacDonald
,
A. W.
,
Cohen
,
J. D.
,
Stenger
,
V. A.
, &
Carter
,
C. S.
(
2000
).
Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control
.
Science
,
288
,
1835
1838
.
Magill
,
R. A.
, &
Hall
,
K. G.
(
1990
).
A review of the contextual interference effect in motor skill acquisition
.
Human Movement Science
,
9
,
241
289
.
Massen
,
C.
(
2004
).
Parallel programming of exogenous and endogenous components in the antisaccade task
.
Quarterly Journal of Experimental Psychology A
,
57
,
475
498
.
McDowell
,
J. E.
,
Dyckman
,
K. A.
,
Austin
,
B. P.
, &
Clementz
,
B. A.
(
2008
).
Neurophysiology and neuroanatomy of reflexive and volitional saccades: Evidence from studies of humans [review]
.
Brain and Cognition
,
68
,
255
270
.
Miller
,
E. K.
, &
Cohen
,
J. D.
(
2001
).
An integrative theory of prefrontal cortex function
.
Annual Review of Neuroscience
,
24
,
167
202
.
Munoz
,
D. P.
, &
Everling
,
S.
(
2004
).
Look away: The anti-saccade task and the voluntary control of eye movement
.
Nature Reviews Neuroscience
,
5
,
218
228
.
Pierce
,
J. E.
,
McCardel
,
J. B.
, &
McDowell
,
J. E.
(
2015
).
Trial-type probability and task-switching effects on behavioral response characteristics in a mixed saccade task
.
Experimental Brain Research
,
233
,
959
969
.
Pierce
,
J. E.
, &
McDowell
,
J. E.
(
2016a
).
Effects of preparation time and trial type probability on performance on anti- and pro-saccades
.
Acta Psychologica
,
164
,
188
194
.
Pierce
,
J. E.
, &
McDowell
,
J. E.
(
2016b
).
Modulation of cognitive control levels via manipulation of saccade trial-type probability assessed with event-related BOLD fMRI
.
Journal of Neurophysiology
,
115
,
763
772
.
Poldrack
,
R. A.
(
2000
).
Imaging brain plasticity: Conceptual and methodological issues—A theoretical review
.
Neuroimage
,
12
,
1
13
.
Reuter
,
B.
,
Kaufmann
,
C.
,
Bender
,
J.
,
Pinkpank
,
T.
, &
Kathmann
,
N.
(
2010
).
Distinct neural correlates for volitional generation and inhibition of saccades
.
Journal of Cognitive Neuroscience
,
22
,
728
738
.
Talairach
,
J.
, &
Tournoux
,
P.
(
1988
).
Co-planar stereotaxic atlas of the human brain. 3-Dimensional proportional system: An approach to cerebral imaging
.
New York
:
Thieme
.
Weiler
,
J.
, &
Heath
,
M.
(
2012
).
Task-switching in oculomotor control: Unidirectional switch-cost when alternating between pro- and antisaccades
.
Neuroscience Letters
,
530
,
150
154
.
Zhang
,
Y.
,
Brady
,
M.
, &
Smith
,
S.
(
2001
).
Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm
.
IEEE Transactions on Medical Imaging
,
20
,
45
57
.