Abstract
With every saccadic eye movement, humans bring new information into their fovea to be processed with high visual acuity. Notably, perception is enhanced already before a relevant item is foveated: During saccade preparation, presaccadic attention shifts to the upcoming fixation location, which can be measured via behavioral correlates such as enhanced visual performance or modulations of sensory feature tuning. The coupling between saccadic eye movements and attention is assumed to be robust and mandatory and considered a mechanism facilitating the integration of pre- and postsaccadic information. However, until recently it had not been investigated as a function of saccade direction. Here, we measured contrast response functions during fixation and saccade preparation in male and female observers and found that the pronounced response gain benefit typically elicited by presaccadic attention is selectively lacking before upward saccades at the group level—some observers even showed a cost. Individual observer’s sensitivity before upward saccades was negatively related to their amount of surface area in primary visual cortex representing the saccade target, suggesting a potential compensatory mechanism that optimizes the use of the limited neural resources processing the upper vertical meridian. Our results raise the question of how perceptual continuity is achieved and how upward saccades can be accurately targeted despite the lack of—theoretically required—presaccadic attention.
- contrast sensitivity
- cortical magnification
- polar angle asymmetries
- presaccadic attention
- primary visual cortex
- saccadic eye movements
Significance Statement
When we make a saccadic eye movement to a target location in the visual field, perception improves at the saccade target, already before the eyes start moving. This benefit afforded by presaccadic attention is thought to be mandatory and independent of eye movement direction. We show that this is not the case; moving our eyes horizontally or downwards, but not upwards, enhances contrast sensitivity. At the neural level, however, humans with less V1 cortical tissue representing the target location for upward saccades have some presaccadic enhancement. The finding that presaccadic attention is dependent upon eye movement direction challenges the view that the presaccadic benefit is automatic and mandatory in nature.
Introduction
As the high acuity of the fovea markedly decreases with eccentricity, we constantly make saccadic eye movements to actively explore the visual world and gather information. To maintain a continuous percept across saccades, the human visual system must seamlessly integrate blurry peripheral information with its high-acuity equivalent, once brought into the fovea by a saccade. To achieve a smooth transition, we preferentially process visual information at the future eye fixation already before the eyes start moving: During saccade preparation, “presaccadic attention” is deployed to the saccade target, where it, for example, enhances visual sensitivity (Kowler et al., 1995; Deubel and Schneider, 1996; Montagnini and Castet, 2007; Rolfs and Carrasco, 2012) and sharpens orientation and spatial frequency tuning (Li et al., 2016, 2019; Ohl et al., 2017; Kroell and Rolfs, 2021). These presaccadic perceptual modulations render the peripheral information at the future eye fixation more fovea-like, easing the integration of pre- and postsaccadic visual input in support of perceptual continuity across saccades. Presaccadic attention is considered robust and mandatory (Deubel and Schneider, 1996; Li et al., 2019; Kreyenmeier et al., 2020; Hanning et al., 2022b); sensitivity at nontarget locations—including the currently still foveated location (Hanning and Deubel, 2022)—inevitably decreases just before saccade onset (Montagnini and Castet, 2007; Deubel, 2008; Collins et al., 2010).
The assumed neural basis of presaccadic perceptual modulations is feedback signals from oculomotor structures [e.g., frontal eye fields (FEF) and superior colliculus (SC)] to early visual cortices (Ekstrom et al., 2008; Bisley and Mirpour, 2019). Subthreshold microstimulation of FEF or SC (which elicits a saccade when stimulated above the threshold) changes activity in the visual cortex (Moore and Armstrong, 2003) and enhances sensitivity at the movement field of the stimulated neurons (Moore and Fallah, 2004; Müller et al., 2005), resembling the behavioral correlates of presaccadic attention.
The strong functional coupling of oculomotor programming and attention is undisputed and neural and behavioral effects of presaccadic attention are thought to be ubiquitous throughout the visual field—independent of where the saccade is executed. Visual performance, however, is heterogeneous around the visual field (Himmelberg et al., 2023b): At a fixed eccentricity, it is better along the horizontal than vertical meridian and along the lower than upper vertical meridian. These polar angle asymmetries emerge for various dimensions, for example, contrast sensitivity (CS) and spatial resolution (Abrams et al., 2012; Barbot et al., 2021), and are robust enough to not be attenuated by covert attention (Carrasco et al., 2001; Roberts et al., 2016; Purokayastha et al., 2021), deployed without concurrent eye movements. Presaccadic attention even exacerbates them: Measured at the perceptual threshold, saccade preparation enhances CS before horizontal and downward, but not upward saccades (Hanning et al., 2022a; Fig. 1a).
Background and experimental design. a, Schematic representation of polar angle performance asymmetries. Points further from the center represent higher performance. Baseline visual performance (during fixation) is better along the horizontal than that in the vertical meridian and along the lower than that in the upper vertical meridian. Covert spatial (Carrasco et al., 2001; Cameron et al., 2002; Talgar and Carrasco, 2002; Roberts et al., 2016; Purokayastha et al., 2021) and temporal attention (Fernández et al., 2019), tested under similar stimulus conditions, benefit performance to the same degree around the visual field. Presaccadic attention, however, exacerbates the asymmetries by enhancing CS at all locations except the upper vertical meridian (Hanning et al., 2022a). b, Experimental task. Vertical saccade condition. After a fixation period, a central direction cue (black line) indicated the saccade target 8° above or below central fixation, marked by black dots (note that in the horizontal conditions, stimuli were placed 8° left and right). ∼100 ms after direction cue onset (delay adjusted for each observer), a test Gabor was presented at the saccade target; observers reported its orientation after the saccade offset. In the fixation condition, two lines cued both placeholder locations; observers were instructed to maintain fixation. See Movie 1 for a demonstration of the trial sequence. c, Visualization of the different contrast levels (5–95%, log-spaced) tested to measure contrast response functions.
Video demonstration of an exemplary vertical saccade trial (grating contrast 70%).
This finding is surprising: (1) Presaccadic attention is assumed to be mandatorily deployed before any saccade is made (Shepherd et al., 1986; Deubel and Schneider, 1996; Hanning et al., 2022b); (2) by rendering peripheral representations at the saccade target more fovea-like, presaccadic attention enables smooth integration of pre- and postsaccadic information and thus aids perceptual continuity (Herwig, 2015; Rolfs, 2015; Kwon et al., 2019)—a mechanism that should be indispensable also for upward saccades; (3) sensitivity at the upper vertical meridian (during fixation) is typically lowest (for a review, see Himmelberg et al., 2023b); thus, perceptual performance could improve there the most.
Our former study measured presaccadic attention at the perceptual threshold (Hanning et al., 2022a), but presaccadic attention primarily modulates performance at high stimulus contrasts, where performance asymptotes (Li et al., 2021b; Hanning et al., 2023). To optimally investigate the extent of presaccadic attention as a function of saccade direction, here we measured full contrast response functions and evaluated presaccadic response gain around the visual field. Additionally, we obtained fMRI-derived retinotopic maps for our observers from the NYU Retinotopy Dataset (Himmelberg et al., 2021). Across individuals, greater local V1 surface area encoding a visual field location benefits perception at that location (Duncan and Boynton, 2003; Himmelberg et al., 2022). We explored whether individual differences in the amount of local V1 surface area encoding the saccade target region are related to the inconsistent magnitude of individual observer’s presaccadic enhancement, which at the group level was lacking before upward saccades.
Materials and Methods
In this study, we compared 2AFC orientation discrimination performance during fixation—a neutral cue instructed observers to maintain fixation throughout the trial—and saccade preparation, where observers made an immediate saccade to a centrally cued target location 8° left, right, above, or below fixation (Fig. 1b and Movie 1; Experimental design). We varied the contrast of the test grating on a trial-by-trial basis across the full contrast range (Fig. 1c), to measure presaccadic modulations of the contrast response function (CRF) at the saccade target. Observers’ gaze positions were monitored continuously. Importantly, test presentation time was adjusted separately to each observer's horizontal and vertical saccade latencies (see Stimulus timing), to ensure a measurement right before saccade onset, where presaccadic attention has its maximum effect (Deubel, 2008; Rolfs and Carrasco, 2012; Li et al., 2016; Hanning et al., 2018, 2019a). Only trials in which the test was presented during the last 100 ms before saccade onset were included in the analysis. We adjusted the tilt angle (clockwise or counterclockwise relative to vertical) in a pretest (see Tilt titration) to compensate for polar angle asymmetries and ensure equal task difficulty around the visual field during fixation—and thus equal room for a presaccadic benefit.
Observers
Eight observers (four female, four male; aged 21–33 years, including two authors, N.M.H. and M.M.H.) with normal or corrected-to-normal vision participated in the study. We chose a sample size in the typical range of studies investigating presaccadic attention (Montagnini and Castet, 2007; Ohl et al., 2017; Li et al., 2019, 2021b; Hanning et al., 2019b; Kreyenmeier et al., 2020; Hanning and Deubel, 2022). All observers provided written informed consent and (except for the two authors) were naive to the purpose of the experiment. The protocols for the study were approved by the University Committee on Activities involving Human Subjects at New York University, and all experimental procedures were in accordance with the Declaration of Helsinki.
Setup
Observers sat in a dimly illuminated room with their heads stabilized by a chin and forehead rest and viewed the stimuli at 57 cm distance on a gamma-linearized 20 in ViewSonic G220fb CRT screen with a spatial resolution of 1,280 by 960 pixels and a vertical refresh rate of 85 Hz. The gaze position of the dominant eye was recorded using an EyeLink 1000 Desktop Mount eye tracker (SR Research) at a sampling rate of 1 kHz. Manual responses were recorded via a standard keyboard. An Apple iMac Intel Core 2 Duo computer, running MATLAB (MathWorks) with Psychophysics (Brainard, 1997; Pelli, 1997) and EyeLink (Cornelissen et al., 2002) toolboxes, controlled stimulus presentation and response collection.
Experimental design
The experiment comprised two experimental conditions—saccade and fixation—tested at the two meridians, horizontal and vertical, in separate experimental blocks. Observers fixated a central fixation target (black ring; ∼0 cd/m2, diameter 0.35°, width 0.1° of visual angle) on a gray (∼26 cd/m2) background (Fig. 1b). Two placeholders indicated the isoeccentric locations of the upcoming stimuli (and potential saccade targets), either 8° left and right or above and below fixation (depending on the tested meridian). Each placeholder comprised four dark gray dots (diameter 0.1°), forming the corners of a square (diameter 4.5°). The trial started once we detected stable fixation within a 1.75° radius virtual circle centered on the fixation target.
In saccade blocks, after a 700 ms fixation period, a central direction cue (black line; ∼0 cd/m2, length 0.175°, width 0.1°) pointed to one of the opposing placeholders (randomly selected), cueing the saccade target. Observers were instructed to look as fast and precisely as possible to the center of the indicated placeholder. Approximately 100 ms after cue onset (interstimulus interval adjusted based on each observer's direction-specific movement latency, see Stimulus timing), a Gabor grating (4-cpd, random phase, Gaussian envelope σ = 0.5°), which was slightly tilted relative to vertical (see Tilt titration), appeared for ∼35 ms at the cued saccade target. Gabor contrast was varied on a trial-by-trial basis following the method of constant stimuli (Michelson contrast, nine log-spaced steps from 5 to 95%, Fig. 1c). Note that the grating was presented within the movement latency, that is, when the gaze still rested at fixation. Then, 400 ms after stimuli offset (well after the eye had landed at the saccade target), we played a high- or low-pitch response sound to specify the location that had contained the Gabor patch. Observers indicated their orientation judgment via button press (clockwise or counterclockwise, two-alternative forced choice) and were informed that the orientation report was nonspeeded. They received auditory feedback for incorrect responses. Stimulus parameters and timing for the fixation blocks were identical to the saccade blocks, with one difference: Two (rather than one) black direction cue lines appeared, pointing to both placeholder locations, and observers were instructed to keep eye fixation while their gaze was monitored.
Observers performed multiple sessions of 12 experimental blocks each: 4 horizontal and 8 vertical blocks (counterbalanced within sessions), of which half were fixation and half were saccade blocks, randomly interleaved. Note that fewer experimental blocks were dedicated to testing the horizontal than vertical meridian because we collapsed data across the left and right test locations (but separately analyzed upward and downward locations). Each block comprised 105 trials. We monitored gaze position online and controlled for correct eye fixation, that is, gaze remaining within 1.75° from the central fixation target until (1) response cue onset (fixation blocks) or (2) direction cue onset (saccade blocks). Trials in which the gaze deviated from fixation were aborted and repeated at the end of each block. In saccade blocks, we also repeated trials with too short (<100 ms) or long (>400 ms) saccade latency or incorrect eye movements (initial saccade landing beyond 2.5° from the indicated target). Overall data collection for an observer was stopped once at least 50 trials per stimulus contrast level and experimental condition were acquired after offline trial exclusion (see Eye data preprocessing).
Tilt titration
To match overall task difficulty to each observer’s visual sensitivity and to account for sensitivity differences around the visual field during fixation, we titrated the tilt angle (±0.125–5° relative to vertical) separately for each observer and test position (horizontal, upper, lower) with best PEST [Pentland, 2010; using custom code (https://github.com/michaeljigo/palamedes_wrapper) that ran subroutines implemented in the Palamedes toolbox (Prins and Kingdom, 2018)]. Using the procedure of the fixation condition, we concurrently ran three independent adaptive procedures (each comprising 36 trials) for each test location, to determine the tilt angle at which observers’ orientation discrimination performance at the highest contrast level (95%) was ∼d′ = 2. The location-specific tilt angle used for the first main experimental session (the same for saccade and fixation blocks) was computed by averaging medians of the last five trials of each staircase across the three staircases per location. Following each experimental session, the tilt angle for each location was adjusted as needed to ensure comparable accuracy across locations in the fixation condition. Note that the tilt angle for a given test location was the same for fixation and saccade blocks within an experimental session. Tilt angles did not correlate with the observed presaccadic benefit for any tested saccade direction (horizontal, r(6) = −0.48, p = 0.224; downward, r(6) = 0.65, p = 0.080; upward, r(6) = −0.24, p = 0.573).
Eye data preprocessing
We scanned the recorded eye position data offline and detected saccades based on their velocity distribution (Engbert and Mergenthaler, 2006) using a moving average over 20 subsequent eye position samples. Saccade onset and offset were detected when the velocity exceeded or fell below the median of the moving average by three standard deviations for at least 20 ms. We included trials in which no blink occurred during the trial and correct eye fixation was maintained within a 1.75° radius centered on central fixation throughout the trial (fixation trials) or until cue onset (saccade trials). Moreover, we only included those eye movement trials in which the initial saccade landed within 2.5 from the required target location and in which the test signal was presented within 100 ms before saccade onset (i.e., the saccade started only after test signal presentation, but not later than 100 ms after signal offset). In total, we included 30,168 trials in the analysis of the behavioral results [on average 3,771 ± 114 (mean ± 1 SEM) trials per observer].
Stimulus timing
The effect of presaccadic attention increases throughout saccade preparation and peaks shortly before saccade onset (Deubel, 2008; Rolfs and Carrasco, 2012; Li et al., 2016; Hanning et al., 2018, 2019a). Saccade latencies, however, vary as a function of saccade direction (Honda and Findlay, 1992; Tzelepi et al., 2010; Hanning et al., 2022a; Kwak et al., 2023; Liu et al., 2023), for example, downward saccades are typically initiated ∼25 ms later than horizontal or upward saccades (Fig. 3d). To measure and compare the effects of presaccadic attention equally close to saccade onset for all saccade directions, we matched the test presentation time relative to the saccade by adjusting the delay between saccade cue and test onset to each observer's direction-specific saccade latency. For the initial experimental session, we randomly selected the stimulus onset asynchrony (SOA) between 141 ms and 235 ms on a trial-by-trial basis. After each session, we evaluated vertical and horizontal saccade latencies and adjusted the SOA range to only include those SOAs for which test presentation fell in the desired presaccadic window (test offset −100 to 0 ms relative to saccade onset) in at least 70% of trials. This adjusted SOA range was used for both fixation and saccade blocks in the upcoming session.
Quantification and statistical analysis
Orientation discrimination performance, indexed by visual sensitivity [d′ = z(hit rate) − z(false alarm rate)], was measured as a function of stimulus contrast using the method of constant stimuli (nine log-spaced Michelson contrast levels: 5, 10, 14, 20, 27, 37, 51, 70, and 95%). We arbitrarily defined counterclockwise responses to counterclockwise oriented gratings as hits and counterclockwise responses to clockwise oriented gratings as false alarms (Huihui et al., 2019; Jigo and Carrasco, 2020; Li et al., 2021b). To avoid infinite values when computing d′, we substituted hit and false alarm rates of 0 and 1 by 0.01 and 0.99, respectively (Wollenberg et al., 2018; Hanning et al., 2019a,b; Hanning and Deubel, 2020).
As only saccade trials with test offset occurring within the last 100 ms prior to saccade onset were included in the sensitivity analysis (see Eye data preprocessing), trials with certain SOAs (see Stimulus timing) were more likely to contribute to the averaged visual sensitivity per contrast level. To ensure identical timing parameters in the fixation condition (in which no trials were excluded based on temporal criteria), for each observer we first calculated d’ for each stimulus contrast level separately for each used SOA, before computing a weighted average per contrast level, that matched the respective contrast level's SOA distribution at the respective test location of the saccade condition.
To evaluate asymptotic CS, we obtained contrast response functions for each condition and test location by fitting each observer's data with Naka–Rushton functions (Naka and Rushton, 1966), parameterized as d′(C) = dmaxCn/(Cn + Cn50), where C is the contrast level, dmax is the asymptotic performance, C50 is the semisaturation constant (contrast level corresponding to half the asymptotic performance), and n determines the slope of the function. The error was minimized using a least-squares criterion; dmax and C50 were free parameters, and n was fixed. Contrast levels were log-transformed prior to fitting. A change in dmax indicates a response gain change, and a change of C50 indicates a contrast gain change. We used repeated-measures ANOVAs to assess statistical significance, followed by paired t tests for post hoc comparisons. All post hoc comparisons were Bonferroni-corrected for multiple comparisons. All p-values for repeated-measures ANOVAs in which the assumption of sphericity was not met were Greenhouse–Geisser corrected. Parameter estimates for psychometric functions and all statistical tests were computed in MATLAB.
fMRI retinotopy and V1 surface computation
Each observer's population receptive field (pRF) data (Dumoulin and Wandell, 2008) and anatomical data were taken from the publicly available NYU Retinotopy Dataset (Himmelberg et al., 2021). The pRF stimulus, MRI and fMRI acquisition parameters, MRI and fMRI preprocessing, the implementation of the pRF model, and the computation of V1 surface area representing wedge ROIs along the cardinal meridians of the visual field are identical to those described in our previous work (Himmelberg et al., 2021, 2022). In brief, to calculate the amount of V1 surface area representing each saccade target region, we defined wedge ROIs in the visual field that were centered along each of the cardinal meridians. Each wedge ROI was ±25° in width and extended 4–12° of eccentricity (i.e., surrounding the saccade target and stimulus locations at 8° eccentricity, as including more data within the wedge ROI leads to a more accurate estimate of V1 surface area). We calculated the amount of V1 surface area (in mm2) encapsulated by each of these wedge ROIs by first generating nine subwedge ROIs (which are later combined to form the full wedge ROI). Each subwedge ROI was constrained to a narrow eccentricity band (nine log-spaced bands from 4 to 12°). For each subwedge ROI, the 25° angular border on the cortex was calculated by identifying the average distance of a selected pool of vertices whose pRF polar angle coordinates lay around the 25° polar angle pRF centers from each respective meridian. All vertices between a meridian and the 25° boundary were included in the subwedge ROI. The subwedge ROI was then used as a mask and overlaid on the anatomical surface, for which each vertex has an assigned surface area value (in mm2). The surface areas of the vertices within the subwedge ROI mask were then summed to give a measurement of the surface area for the respective subwedge ROI. This was repeated for each subwedge ROI. Finally, the surface area of each subwedge ROI was summed to compute the amount of surface area representing the full wedge ROI representing a saccade target wedge in the visual field. This was repeated for each cardinal saccade target wedge. For the current study, the surface areas of the horizontal (left and right) saccade target wedge ROIs were averaged together, as there are no differences in surface or performance between the left and right horizontal meridian (Himmelberg et al., 2023b). We conducted (two-tailed) Pearson’s correlations to evaluate the relation between V1 surface area and presaccadic enhancement for horizontal, upward, and downward saccades at the individual observer level.
Results
For each test location (horizontal, lower, upper) and condition (fixation, saccade), we fit performance (d′) as a function of target contrast with a Naka–Rushton function (Naka and Rushton, 1966; see Quantification and statistical analysis in Materials and Methods). Contrast responses during fixation were highly comparable across test locations (Fig. 2a–c, gray), also indicated by the absence of a difference between their asymptotic performance level dmax (F(2,14) = 1.09, p = 0.355). We conducted repeated-measures ANOVAs with the factors test location and condition. Consistent with a recent study (Li et al., 2021b), presaccadic attention did not affect the semisaturation constant C50 (F(1,6) = 1.81; p = 0.221) but significantly modulated the asymptotic performance level dmax (F(1,6) = 29.63, p = 0.001), indicating a response gain change. Importantly, this modulation interacted with the saccade direction (F(2,14) = 15.63, p < 0.001). Bonferroni-corrected post hoc comparisons showed that saccade preparation enhanced dmax at the saccade target (relative to fixation) before horizontal (p < 0.001; Fig. 2a) and downward (p < 0.001; Fig. 2b) saccades. Crucially, there was no presaccadic response gain benefit before upward saccades (p = 0.217; Fig. 2c). Correspondingly, the presaccadic benefit Δdmax (dmax saccade − dmax fixation) significantly varied with saccade direction (F(2,14) = 15.63; p < 0.001). Post hoc comparisons demonstrated no statistical difference between the benefit caused by horizontal and downward saccades (p = 1.00; Fig. 2d), whereas the benefit before upward saccades (when present) was significantly reduced compared with both horizontal (p = 0.003; Fig. 2e) and downward (p = 0.004; Fig. 2f) saccades. Note that this differential effect was present for all observers.
Contrast responses depending on saccade direction. Group-averaged psychometric contrast response functions (CRF; d′ vs contrast) measured during fixation (gray) or saccade preparation (colored) at the horizontal (a), lower vertical (b), and upper vertical (c) meridian. Group-averaged dmax and C50 extracted from individual observer’s CRFs are plotted at the right and the bottom of the figure. Error bars depict ±1 SEM. ***p < 0.001. (d–f) Pairwise comparison of individual observer’s presaccadic benefit Δdmax (dmax saccade − dmax fixation) at the horizontal (green background), lower (purple), and upper (orange) test locations. The black crosses depict the group average ±1 SEM.
We evaluated eye movement parameters to ensure that the missing presaccadic benefit before upward saccades cannot be explained by differences in saccade latency or precision, for example, comparatively slower or less precise upward saccade execution. A two-way repeated-measures ANOVA with factors saccade direction and test contrast level showed a significant main effect of saccade direction (F(3,21) = 8.94, p = 0.006) on latency (Fig. 3a), but there was no effect of stimulus contrast (F(8,56) = 1.50, p = 0.248) or an interaction effect between the two (F(24,168) = 2.15, p = 0.116). Post hoc comparisons indicated significantly longer latencies for downward saccades compared with left (p = 0.006), right (p < 0.001), and upward (p = 0.037) saccades. As this pattern was expected (Honda and Findlay, 1992; Tzelepi et al., 2010; Grujic et al., 2018; Hanning et al., 2022a; Kwak et al., 2023; Liu et al., 2023), we adjusted the delay between cue and test presentation to each observer's horizontal and vertical saccade latencies (see Stimulus timing in Materials and Methods) so that test presentation time relative to saccade onset was matched across saccade directions (Fig. 3d); pairwise comparisons did not show any significant difference between upward saccades and those to other directions (all p ≥ 0.106). Note that presaccadic perceptual enhancement was not correlated with saccade latencies for any saccade direction (all p ≥ 0.671). Likewise, a saccade latency median split (within each participant) showed no significant difference in the presaccadic effect between relatively shorter and longer latencies for each saccade direction (all p ≥ 0.467).
Eye movement parameters. Group average saccade latency (a) and amplitude (b) as a function of saccade direction (top) and test contrast level (bottom). The shaded error areas indicate ±1 SEM. c, Normalized saccade endpoint frequency maps averaged across observers depicting saccade landing variance. d, Density plots of saccade onset times relative to test offset, for each observer (thin colored lines) and across observers (bold line). The dashed vertical lines indicate the average saccade onset across observers; the gray background indicates the time window of trials included in the analysis (test offset ≤100 ms before saccade onset).
Neither saccade amplitudes (Fig. 3b) nor saccade landing errors (i.e., the mean Euclidean distance between saccade endpoints and saccade target center; Fig. 3c) were affected by saccade direction (amplitude, F(3,21) = 1.585, p = 0.241; landing error, F(3,21) = 1.292, p = 0.306), stimulus contrast (amplitude, F(8,56) < 1; landing error, F(8,56) = 1.41, p = 0.259), or their interaction (amplitude, F(24,168) < 1; landing error, F(24,168) = 1.37, p = 0.265). In sum, latency and movement precision of upward saccades were comparable to those of horizontal and downward saccades, ruling out these parameters as sources for the absent presaccadic benefit before upward saccades. In any case, the presaccadic shift of attention is not linked to the saccade landing position but to the intended saccade target (Deubel and Schneider, 1996; Van der Stigchel and de Vries, 2015; Wollenberg et al., 2018; Hanning et al., 2019b; Li et al., 2021b).
Our data show that presaccadic attention not only varies around the visual field but also considerably among individual observers. Whereas observers consistently showed a presaccadic benefit of similar magnitude before horizontal and downward saccades (Fig. 4a), we observed a mixed pattern for upward saccades, ranging from moderate presaccadic benefits to even presaccadic costs (i.e., lower presaccadic sensitivity at the upward saccade target than during fixation; 37.5% of observers). Note that for those observers displaying a presaccadic benefit at all locations, the benefit before upward saccades was still lower than that before horizontal and downward saccades.
Linking presaccadic attention to individual differences in the surface area of primary visual cortex. a, Individual observer’s dmax (extracted from their fitted CRF) for each test location and condition. The dashed black lines depict the group average, the error bars depict ±1 SEM. b, Individual observer’s presaccadic benefit (dmax saccade − dmax fixation) before horizontal, downward, and upward saccades plotted as a function of their respective localized measurements of V1 surface area. The circular icon depicts the extracted area: 4–12° eccentricity, ±25° of radial angle extending either side of each cardinal meridian (note that the horizontal surface area was computed as the average of left and right). *p < 0.05.
Notably, the visual field is nonuniformly represented in primary visual cortex (V1)—with a greater surface area representing the horizontal than vertical meridian and lower vertical than upper vertical meridian (Silva et al., 2018; Benson et al., 2021; Himmelberg et al., 2021, 2022, 2023a). This directly reflects (Himmelberg et al., 2020) and likely (at least partially) underlies (Himmelberg et al., 2022) perceptual differences around the visual field (Fig. 1a). Individual differences in CS at perceptual threshold (measured during fixation) correlate with the amount of V1 surface dedicated to processing the respective region of visual space (Himmelberg et al., 2022). Thus, higher CS at the horizontal meridian is a perceptual consequence of more dedicated V1 surface area. We confirmed that this link also holds in our study by correlating CS measurements obtained from six of the eight observers at the perceptual threshold in our previous study (Hanning et al., 2022a) with their V1 surface area measurements [using the same parameters as Himmelberg et al. (2022); Fig. 5]. We replicated the positive correlation between V1 surface area and CS during fixation (Himmelberg et al., 2022; Jigo et al., 2023), which provides further support for the notion that targeted fMRI measurements are robust even with small sample sizes (Himmelberg et al., 2021). Could this relation between primary visual cortex and CS also hold for presaccadic perceptual modulations? To explore whether presaccadic attention benefits are linked to V1 surface area, we correlated each observer's presaccadic benefit before horizontal, downward, and upward saccades with their amount of V1 surface area representing the respective saccade target region. Individual observer’s surface estimates were obtained in a separate fMRI experiment (Himmelberg et al., 2021; see fMRI retinotopy and V1 surface computation in Materials and Methods).
Local V1 surface area predicts CS during fixation. Individual observer’s normalized CS measured during fixation in the context of our previous study (Hanning et al., 2022a) as a function of the respective localized V1 surface area measurement. The circular icon depicts the extracted area: 1–8° eccentricity, ±15° of radial angle extending either side of each cardinal meridian, matching the original study (Himmelberg et al., 2022). Note that the horizontal surface area was computed as the average of left and right. **p < 0.01.
For horizontal and downward saccades, which showed the typical presaccadic sensitivity enhancement, we observed no correlation between presaccadic benefits and the amount of V1 surface area representing the saccade target (horizontal target, r(6) = −0.03, p = 0.947; downward target, r(6) = −0.17, p = 0.681). In contrast, interindividual variations in the presaccadic effect preceding upward saccades were negatively correlated with the amount of V1 surface dedicated to processing the upward saccade target (Fig. 4b). Thus, for individual observers, the less cortical surface representing the saccade target, the larger the presaccadic attentional benefit (upward target, r(6) = −0.79, p = 0.019). This finding documents a link between cortical anatomy and the magnitude of presaccadic perceptual modulations (or lack thereof) preceding upward saccades at the individual observer level.
To underscore the significance of this finding, we have validated the consistency of the observed inverse relation between V1 surface area and perceptual presaccadic enhancement at the upper vertical meridian by evaluating data from our previous psychophysical study measuring presaccadic CS (Hanning et al., 2022a). Observers for whom fMRI retinotopy data were available from the NYU Retinotopy Dataset (n = 5 independent, new observers) show the same pattern of a negative relation between V1 surface area and presaccadic benefit at the upper vertical meridian: Figure 6 depicts the presaccadic enhancement, now computed as the ratio of presaccadic CS, indexed at dmax (present study, n = 8) or contrast threshold [previous study (Hanning et al., 2022a), n = 5] and the respective measurement during fixation. Combined, these data from 13 observers confirm the inverse relation between the V1 surface area representing the upward saccade target and the corresponding (diminished) presaccadic enhancement (r(11) = −0.71; p = 0.007).
Presaccadic benefit in asymptotic and threshold CS as a function of local V1 surface area representing the respective saccade target region. Benefit computed as the ratio of asymptotic CS dmax saccade/CS dmax fixation (colored dots, matching n = 8 observers in Figs. 2–4) or CS threshold saccade/CS threshold fixation (white dots, n = 5). The circular icons depict the extracted area: 4–12° eccentricity, ±25° of radial angle extending either side of each cardinal meridian (note that the horizontal surface area was computed as the average of left and right). **p < 0.01.
Discussion
To investigate how presaccadic attention for different saccade directions impacts the CRF, we have optimized the detection of presaccadic benefits as a function of saccade direction: We (1) measured presaccadic CS across the full contrast range to capture the response gain effect by which presaccadic attention modulates CS; (2) adjusted overall task difficulty during fixation—which served as a baseline condition to quantity the effect of presaccadic attention—to compensate for performance field asymmetries; and (3) adjusted the test presentation time separately to each observer's direction-specific saccade latency to ensure a measurement right before saccade onset, when presaccadic attention has its maximum effect.
We found that horizontal and downward saccades increased response gain, consistent with the well-established sensitivity benefit at the saccade target. Before upward saccades, however, presaccadic benefits were absent at the group level—and inconsistent across observers—resembling our findings at the perceptual threshold (Hanning et al., 2022a). We verified that differences in saccade latency, amplitude, or landing precision (between observers or saccade directions) could not explain the diminished to absent upward presaccadic enhancement; some observers even showed a cost. Interestingly, we identified a neural correlate for this differential effect: Observers with a comparatively greater upward presaccadic benefit have relatively less V1 surface area dedicated to processing that target region.
The absent presaccadic benefit before upward saccades contradicts the notion that presaccadic attention is automatically deployed before any saccade; however, presaccadic attention studies typically rely on measurements along the horizontal meridian or do not evaluate potential differences among tested saccade directions (for a review, see Li et al., 2021a). Likewise, neurophysiological measurements of presaccadic attention have not been evaluated as a function of saccade direction. Our present and recent (Hanning et al., 2022a; Liu et al., 2023) findings demonstrate that observations made for one saccade angle do not generalize to other directions and call for an investigation of presaccadic attention for other properties and tasks as a function of polar angle.
Although the strong coupling between saccade planning and visual attention has been undisputed, the direction of this relation is still debated. Proponents of the premotor theory of attention argue that visual attention is a product of the motor system; even covert shifts of attention, during fixation, result from (eye) movement planning (Rizzolatti et al., 1987; Craighero et al., 1999; for a review, see Smith and Schenk, 2012). Oculomotor brain structures (e.g., FEF and SC) are also selectively modulated during covert attention tasks in human- and nonhuman primates (Nobre et al., 2000; Bogadhi et al., 2018; Bollimunta et al., 2018). However, distinct neuronal populations within these regions underlie covert attention and saccade preparation activity (Ignashchenkova et al., 2003; Müller et al., 2005; Gregoriou et al., 2012; Messinger et al., 2021). In sum, recent research refutes the view that the control of spatial attention is dependent on oculomotor control circuits (Hanning et al., 2019b, 2023; Hanning and Deubel, 2020; Masson et al., 2020; Messinger et al., 2021; review: Li et al., 2021a) and suggests the reverse: goal-directed movements depend on preceding attentional selection to specify motor target coordinates—selection for action (Schneider, 1995; Deubel and Schneider, 1996; Baldauf and Deubel, 2010). Our results are inconsistent with both accounts: (1) Eye movement parameters for upward saccades are no different from horizontal and downward saccades; given that upward saccade motor programming is “typical” (has comparable saccade latency and precision), according to the premotor theory of attention, it should cause a regular shift of attention to the motor target; (2) according to the selection-for-action account, the presaccadic shift of attention functions as a motor target marker and is a prerequisite for saccade execution. Here, we show regularly executed upward saccades without a measurable attentional marker of motor target selection.
Assuming presaccadic attention is required for motor target selection but absent prior to upward saccades, how can we perform (accurate) upward saccades? Could presaccadic attention be deployed to upward saccade targets, without resulting in the typical corresponding perceptual advantages? This possibility could be tested by measuring neurophysiological correlates of presaccadic attention. ERPs, which track covert attentional (Luck et al., 1994; Luck and Yard, 1995) and motor target selection (Baldauf and Deubel, 2009) via sensory-evoked P1/N1 components, could reveal a dissociation between perceptual and neurophysiological correlates of presaccadic attention at the upper vertical meridian.
This is a plausible scenario given that sensitivity at the upper vertical meridian is lower than at other isoeccentric locations due to anatomical constraints in the retina and visual cortex (Himmelberg et al., 2023b). Perceptual polar angle asymmetries are likely explained by the distribution of V1 tissue (Benson et al., 2021; Himmelberg et al., 2022, 2023a; Kupers et al., 2022), which parallels perceptual sensitivity differences around the visual field (Benson et al., 2021), even at individual observer level (Himmelberg et al., 2022). Could the reduced neural resources devoted to processing the upper visual field (partially) explain why saccade preparation fails to improve sensitivity at the upper vertical meridian? The absent group-level benefit at the upper vertical meridian could be related to its smallest surface area representation, largest receptive fields, and lowest preferred spatial frequency (Aghajari et al., 2020; Broderick et al., 2022). Our 4-cpd test stimulus may be less suited for driving neural responses at the upper vertical than the horizontal meridian (Jigo et al., 2023) at the 8° eccentric saccade target.
At the individual level, however, observers with comparatively less V1 surface representing the upward saccade target had a relatively greater—but compared with other directions still severely reduced—presaccadic benefit. According to the above explanation, however, one would expect observers with relatively larger surface areas to have larger presaccadic benefits. Presaccadic attention causes the neural representation of a target stimulus to become more “fovea-like” and likely sharpens receptive field sizes at the target location (Li et al., 2016, 2019; Ohl et al., 2017; Kroell and Rolfs, 2021; Kwak et al., 2023). If observers with less surface area representing the upper vertical meridian have larger receptive fields, and presaccadic attention sharpens them, this could result in a (relatively) greater presaccadic benefit for observers with smaller cortical upper vertical visual field representation.
This effect was not observed at the other cardinal locations, indicating that computations of neurons encoding the upper vertical meridian may differ from those at the other meridians. Magnifying the stimulus size according to cortical representation eliminates CS differences across eccentricity but not polar angle (Jigo et al., 2023), suggesting that polar angle performance differences are likely mediated by neurons with differential image-processing capabilities (i.e., differently tuned spatial filters). Using reverse correlation, we have shown that higher sensitivity to task-relevant orientation and spatial frequency and lower internal noise at fovea than perifovea underlie eccentricity-dependent variations (Xue et al., 2023). Polar angle differences, however, seem to stem from different orientation and spatial frequency tuning functions; for example, the upper vertical meridian is less sensitive to orientation and tuned to low spatial frequencies (Xue and Carrasco, 2023).
Our results challenge the mandatory link between saccade programming and attention and raise the question of how a smooth transition from (presaccadic) blurry peripheral information to its high-acuity equivalent (once foveated after the saccade) is achieved also across upward saccades: Transsaccadic integration is assumed to rely on predictive presaccadic sensitivity modulations to render peripheral representations at the saccade target fovea-like (Herwig, 2015; Rolfs, 2015). Interestingly, CS is reduced at the postsaccadic center of gaze after upward saccades (Liu et al., 2023). This effect could result from a diminished preparatory shift of presaccadic attention to the upward saccade target. Likewise, perisaccadic perceptual mislocalization persists longer after upward saccades than saccades in other directions (Grujic et al., 2018)—which has been linked to asymmetries in SC neural properties representing the upper versus lower visual hemifield (Hafed and Chen, 2016). This study, however, does not record from neurons with receptive fields at the upper vertical meridian (±30°), where our analyses focus. Future work should document whether reported asymmetries also extend to the vertical meridian, where perceptual and cortical asymmetries are more pronounced (Himmelberg et al., 2023b).
We have recently investigated presaccadic acuity thresholds, which were equally increased (compared with fixation) for all cardinal saccade directions, including upward saccades (Kwak et al., 2023). This saccade direction–unspecific effect is seemingly inconsistent with the direction-specific effect of presaccadic attention on CS. However, acuity and CS measurements target different points on the contrast sensitivity function (CSF), which is characterized by a joint manipulation of contrast and spatial frequency. Whereas visual acuity, measured at full contrast, corresponds to the cutoff point of the CSF, CS is measured closer to the peak of the CSF. If presaccadic attention reshapes the CSF differently depending on the saccade direction, this could explain its differential effect on (enhanced) acuity and (unaffected) CS at the upper vertical meridian.
To conclude, even when accounting for known polar angle differences in peripheral vision during fixation and testing in a contrast range optimized for detecting presaccadic sensitivity modulations, saccades made to the upper vertical meridian lack the typical presaccadic benefit. This effect may be related to individual differences in the V1 cortical surface shown here and to different computations underlying perceptual modulations at the upper vertical meridian. Our findings call into question the generalizability of behavioral and neurophysiological measurements obtained from one saccade angle to another. Future studies should test whether oculomotor feedback signals (between eye movement and visual areas), which are thought to underlie presaccadic modulations of visual perception, are different for upward saccades than those in other directions, and whether further behavioral correlates of presaccadic attention—such as orientation tuning or the selective shift of sensitivity to higher spatial frequencies—are diminished or even absent prior to upward saccades.
Data Availability
Raw eye tracking and behavioral data are available from the OSF database at https://osf.io/xrtbf/; preprocessed fMRI data are available from the openNeuro platform at https://openneuro.org/datasets/ds003787/versions/1.0.0.
Footnotes
This research was supported by a Marie Skłodowska-Curie Individual Fellowship by the European Commission (898520) to N.M.H. and NIH NEI Grants R01-EY-019693 and R01-EY-027401 to M.C. We thank the members of the Carrasco Lab and Kate Salamatov and Monty Cox for their helpful discussions.
The authors declare no competing financial interests.
- Correspondence should be addressed to Nina M. Hanning at hanning.nina{at}gmail.com.
