Abstract
Each time we make an eye movement, attention moves before the eyes, resulting in a perceptual enhancement at the target. Recent psychophysical studies suggest that this pre-saccadic attention enhances the visual features at the saccade target, whereas covert attention causes only spatially selective enhancements. While previous nonhuman primate studies have found that pre-saccadic attention does enhance neural responses spatially, no studies have tested whether changes in neural tuning reflect an automatic feature enhancement. Here we examined pre-saccadic attention using a saccade foraging task developed for marmoset monkeys (one male and one female). We recorded from neurons in the middle temporal area with peripheral receptive fields that contained a motion stimulus, which would either be the target of a saccade or a distracter as a saccade was made to another location. We established that marmosets, like macaques, show enhanced pre-saccadic neural responses for saccades toward the receptive field, including increases in firing rate and motion information. We then examined if the specific changes in neural tuning might support feature enhancements for the target. Neurons exhibited diverse changes in tuning but predominantly showed additive and multiplicative increases that were uniformly applied across motion directions. These findings confirm that marmoset monkeys, like macaques, exhibit pre-saccadic neural enhancements during saccade foraging tasks with minimal training requirements. However, at the level of individual neurons, the lack of feature-tuned enhancements is similar to neural effects reported during covert spatial attention.
Significance Statement
Attention leads eye movements producing perceptual enhancements at saccade targets. Recent psychophysical studies indicate that increases in pre-saccadic sensitivity are concentrated around features of the target. We tested at the neural level how pre-saccadic attention modulates the tuning curves of visual neurons in area MT of marmoset monkeys. While neurons exhibited clear pre-saccadic enhancements that were consistent with previous studies in macaques, the changes were uniform across the tuning curve. These results show pre-saccadic enhancements are a general feature of visual processing, shared by New World monkeys, but at the level of individual neurons, enhancements are uniform across features much like what has been reported previously for covert attention.
Introduction
Visual attention is strongly linked to eye movement planning and saccadic control (Bisley, 2011; Squire et al., 2013). Every saccade is preceded by a shift of attention that enhances the perception of the saccade target, called pre-saccadic attention (Kowler et al., 1995; Deubel and Schneider, 1996; Rolfs et al., 2011; White et al., 2013). Pre-saccadic attention is automatic and occurs rapidly within 50–100 ms before saccades (Deubel, 2008; Rolfs et al., 2011; Rolfs and Carrasco, 2012; Li et al., 2016; Ohl et al., 2017). It also appears to be obligatory, occurring even when it is detrimental to task demands (Montagnini and Castet, 2007; Deubel, 2008; Steinmetz and Moore, 2010). Neural studies with nonhuman primates have shown that pre-saccadic attention involves enhancements in firing that increase neural sensitivity and, to a first approximation, are highly similar to that seen during covert attention where the eyes remain at central fixation (Steinmetz and Moore, 2010; Squire, et al., 2013). However, psychophysical studies highlight ways in which pre-saccadic attention differs from covert attention (Li et al., 2021a,b). The differences between these mechanisms at the neural level remain uncertain.
Recent human psychophysics suggests that pre-saccadic attention involves a concentration of enhancement around the saccade target's features (Li et al., 2016, 2021a; Ohl et al., 2017). The concentration of sensitivity around a target feature could be implemented at the neural level by feature gain, similar to a feature-based attention. Neurophysiology studies of feature-based attention have found that neurons sharing features with an attended target increase the gain of their responses while those selective to opposite features are suppressed (Treue and Martinez-Trujillo, 1999; Martinez-Trujillo and Treue, 2004). In contrast, a pure spatial selection of the target, as in covert spatial attention, is associated with increases in the gain that are uniform across features (McAdams and Maunsell, 1999). If pre-saccadic attention engages the automatic selection of target features, we would predict feature-specific gain rather than spatial gain. To test between these alternatives requires detailed measurement of neural tuning.
Previous studies of pre-saccadic attention in macaques have found improvements in neural sensitivity but without careful examination of feature tuning. It is known that visual neurons in several brain areas increase firing rates and stimulus selectivity during pre-saccadic attention when a saccade is planned to a stimulus within a neuron's receptive field (RF) (Moore et al., 1998; Li and Basso, 2008; Moore and Chang 2009; Steinmetz and Moore 2010; Merrikhi et al., 2017). It has also been shown at off-target locations that there are enhancements when the RF stimulus matches features of the saccade target (Burrows et al., 2014). However, no studies have examined if changes in tuning curves might reflect feature or spatial gain.
We examined how pre-saccadic attention modulates tuning curves in the middle temporal visual area and the middle temporal crescent (MT/MTC) of the marmoset monkey. The marmoset is a small-bodied New World primate that has gained interest in neural investigations due to the feasibility of genetic manipulation (Belmonte et al., 2015) and advantages for imaging and array recordings (Solomon and Rosa, 2014; Mitchell and Leopold, 2015). Area MT/MTC neurons exhibit similar tuning for motion direction as macaques (Elston and Rosa, 1999). However, it remains untested if marmosets share similar mechanisms of visual attention. In the macaque, it is established how MT neurons are modulated by feature-based attention (Treue and Martinez-Trujillo, 1999; Martinez-Trujillo and Treue, 2004) and also that they show pre-saccadic enhancements (Merrikhi et al., 2021). We developed a saccade foraging paradigm for marmoset monkeys to test for changes in neural tuning under pre-saccadic attention. We first established that, like macaques, marmoset neurons show pre-saccadic increases in neural firing and sensitivity for motion direction. Then by varying motion direction independently across trials, we sampled full tuning curves for motion direction and tested if they exhibited feature-specific enhancements for the saccade target.
Methods
Subjects and surgery
All experimental protocols were approved by the University of Rochester Institutional Animal Care and Use Committee and were conducted in compliance with the National Institutes of Health guidelines for animal research. Two adult common marmosets (Callithrix jacchus), Marmoset E (female) and Marmoset M (male), were used for neurophysiology recording experiments to measure changes in neuronal tuning during pre-saccadic attention. Subjects were single housed at the University of Rochester with a circadian cycle of 12 h light/dark. Subject M was briefly food scheduled with full access to water during early training but had no restrictions by the time neurophysiological data were collected. Subject E was never food scheduled and always had full access to food and water.
Both subjects were surgically implanted with head caps to stabilize them for head-fixed eye tracking and neural recordings. Two months prior to surgery, subjects were trained to sit in a small primate chair following methods previously described (Lu et al., 2001; Remington et al., 2012; Osmanski et al., 2013; Nummela et al., 2017). Then subjects underwent surgery under sterile conditions to implant an acrylic head cap with titanium posts to stabilize the head using methods described in detail previously (Nummela et al., 2017). During the implant surgery, recording chambers were placed over visual areas MT and V1 based on stereotaxic coordinates (Paxinos et al., 2012). Recording chambers consisted of custom 3D prints (Protolabs) and adhered to the skull using C&B Metabond (Parkell). The skull inside the recording chambers was also covered by a thin layer of C&B Metabond. After initial head implant and chamber placement, marmosets were trained to acclimate to head restraint while sitting comfortably in a custom-designed primate chair. Over several months, they were trained to perform several basic tasks including central fixation (Mitchell et al., 2014) and a saccade task toward a peripherally detected Gabor grating, which we used to measure their visual acuity (Nummela et al., 2017).
After preliminary training, a second surgery was performed to create a craniotomy (2–3 mm in diameter) in the recording chamber over area MT. Craniotomies were sealed with a thick layer of silastic gel (Kwik-Sil; World Precision Instruments) to protect the brain from infection and reduce granulation growth (Spitler and Gothard, 2008). If any bleeding occurred or the Silastic seal leaked clear fluids in the days following surgery, then the chamber was cleaned with sterile saline, and a new Silastic layer was applied. Typically, the chamber stabilized and remained dry with a tight seal after a few days to a week. At that time, we performed a dural scrape to remove any excess tissue over the dura and applied a thin layer (<1 mm) of Silastic, which was thin enough to enable passage of electrodes for recordings. The silastic remained in place for the duration of the study. When applied correctly, the silastic has been observed to limit the growth of granulation tissue on the dura and prevent infection (Spitler and Gothard, 2008) and can also be recorded through tungsten electrodes (Miller et al., 2015, MacDougall et al., 2016). Additionally, it was possible to record through Silastic using linear array silicon probes (NeuroNexus) when the dura was thin.
Electrophysiology
Over the duration of the study, our recordings improved from using single-channel tungsten electrodes to using multichannel linear silicon arrays. The first monkey was recorded entirely with tungsten single electrodes (47 sessions), with fewer tungsten electrode recordings in the second monkey (31 sessions). The bulk of data from the second monkey originated from linear arrays that provide much higher cell counts per session (21 sessions). Due to animal health issues, we were unable to employ linear arrays in the first monkey, which passed away during the pandemic. The current findings focus on behavior and the single-unit neural effects on tuning curves during pre-saccadic attention, which can be addressed well with either of the recording methods used and can include both animals.
In single tungsten electrode recordings, we sampled across recording sites in a 1 mm spaced grid with electrodes advanced through metal guide tubes placed in the grid. We inserted tungsten 2.5–5 MΩ electrodes (1–3 FHC) that were mounted onto a lightweight screw micro-drive (Crist Instrument, 3-NRMD drive) positioned over the spacing grid. Electrodes were passed through a metal guide tube that touched but did not penetrate the Silastic layer covering the brain. Tungsten electrodes reliably penetrated the thin silastic layer and intact dura to enter the brain. A stainless-steel reference wire was implanted under the skull in a 1 mm craniotomy.
Later recording sessions using multichannel linear silicon arrays (NeuroNexus) used a custom-built X–Y stage for submillimeter targeting of recording sites. The X–Y stage was mounted onto the recording chamber and carried a lightweight screw micro-drive (Crist Instrument, 3-NRMD) that could deliver linear arrays mounted to a steel tube (28 gauge) into the brain. Designs for the 3D printed parts used in the X–Y stage and recording chamber are online (https://marmolab.bcs.rochester.edu/resources.html).
All neurophysiology data were amplified and digitized at 30 kHz with Intan headstages (Intan) using the Open Ephys GUI (https://github.com/open-ephs/plugin-GUI). The wideband signal was highpass filtered by the headstage at 0.1 Hz. We corrected for the phase shifts from this filtering (Jun et al., 2017). For linear arrays, the resulting traces were also preprocessed by common-average referencing.
Tungsten spike sorting and cluster isolation
Single-unit and multiunit clusters from tungsten electrodes were identified using custom MATLAB software. First, the raw signal was bandpass filtered from 800 Hz to 6,000 Hz with a sixth-order Butterworth filter. To reduce movement artifacts in recordings (i.e., licking or other movements), we chose a narrower filter for initial spike detection to threshold spike events (1,500 Hz to 4,500 Hz) and then used the wider band pass (800–6,000 Hz) to classify single units based on clustering of their threshold triggered spike waveforms in a principal components analysis (PCA) space that included the first two principal components and time as variables. Clusters identified in PCA space were compared to a noise floor, based on random sampling of threshold events, and clusters that could not be fully separated from other clusters or the noise floor, or that exhibited more than 1% of inter-spike interval violations under 1 ms, were counted as multiunit activity.
Laminar electrophysiology and spike sorting
In later recordings, we were able to use multisite silicon electrode arrays that provided much higher cell counts. These arrays included 1–2 shanks, and each shank consisted of 32 channels with 35 microns spacing between contacts. All arrays were 50 microns thick and had sharpened tips. Arrays that included two shanks spaced 200 μm apart were from NeuroNexus (http://www.neuronexus.com). Although the arrays could penetrate the dura when it was thinned, we found that dimpling of the tissue could still suppress neural activity during insertion and that the best recording quality was achieved by applying a small 1–2 mm horizontal slit in the dura during a dural scape and sealing it under Silastic to prevent infection. For the best recording quality, it was further useful to electrode-plate the silicon electrode arrays with PEDOT, a method that has been shown to increase signal/noise ratios (Ludwig et al., 2006, 2011). Last, the yield of neurons recorded was generally improved by inserting the array electrodes into the cortex slowly. We first lowered the silicon shank quickly until we observed units at the array tip, and then retracted one turn of the micro-drive (250 μm) slowly. Then we lowered the arrays advancing approximately 4–6 turns (1–1.5 mm) over a 20–30 min duration until neurons were evenly distributed across the length of the shank. We aimed to span the entire depth of the cortex (∼1.5 mm) to obtain a more uniform sampling of units across layers. We then slowly retracted the array 1–2 turns (0.25–0.5 mm) to reduce pressure on the tissue during recordings. During this final retraction, neurons did not typically shift vertical locations on the array, suggesting that it primarily acted to reduce pressure on the tissue and that otherwise the arrays would have continued advancing slowly during the recording as the tissue relaxed. After retracting, we waited 20 min before starting the main behavioral task and recordings.
We spike sorted array data after initial filtering from the Intan system (as described earlier) using Kilosort2. Outputs from the spike sorting algorithms were manually labeled using the “phy” GUI (https://github/kwikteam/phy). Units with tiny or physiologically implausible waveforms were classified as noise and excluded. Kilosort can identify multiunit clusters that are not physiologically possible, spanning any channels with unrealistic waveforms. Therefore, to be conservative, we only included units from Kilosort that had clear clusters in PCA space, less than 1% inter-spike interval violations, and biphasic spike waveforms localized to adjacent channels on the linear array.
Stimulus presentation and timing
Stimuli were generated using the Psychophysics toolbox (Brainard, 1997; Pelli, 1997; Kleiner et al., 2007) in MATLAB 2015b (MathWorks) on a PC (Intel i7 CPU, Windows 7, 8 GB RAM, GeForce Ti graphics card). They were presented on a gamma-corrected display (BenQ X2411z LED monitor; resolution, 1920 × 1080 p; refresh rate, 100 Hz; gamma correction, 2.2) that had a dynamic luminance range from 0.5 to 230 cd/m2 at a distance of 57 cm in a dark room and viewed under head restraint in custom-designed primate chair as described previously (Nummela et al., 2017). Brightness on the display was set to 100 and contrast to 50, and additional visual features of the monitor, such as blur reduction and low blue light, were turned off. Gamma corrections were verified with measurement by a photometer. Task events and neural responses are recorded using a DATAPixx I/O box (VPixx Technologies) for temporal registration. The MATLAB code is available online (https://github.com/jcbyts/MarmoV5).
Random dot motion fields were used as targets for saccade foraging and also provided a stimulus to validate the motion selective responses on individual neurons as an inclusion criterion. Each aperture contained a field of black dots (each dot 0.15 dva diameter with a density of 2.54 dots per visual degree squared), which moved at 15°/s in one direction (100% coherent). Dot speed was fixed at 15°/s to roughly match the preferred median speed previously reported for neurons in marmoset MT (Lui et al., 2007; Solomon et al., 2011). The dots had limited lifetimes of 50 ms with asynchronous updating to new locations. The radius of the dot field was equal to half the eccentricity of where it was located from the center of the screen and appeared over a gray background (115 cd/m2). The contrast of dots was decreased from black at the center of field (0.5 cd/m2) to zero on the aperture edges according to a Gaussian envelope with a sigma equal to a one-sixth of the aperture's diameter.
Eye tracking
Eye position was acquired at 220 Hz using an Arrington Eye Tracker and ViewPoint software (Arrington Research) or at 1,000 Hz using an EyeLink 1000 Plus eye tracker (SR research). Eye position was collected from infrared light reflected off of a dichroic mirror (part #64-472, Edmund Optics). Each subject's vision was corrected using spherical concave lenses (Optimark Perimeter Lens Set) that were centered 4–5 mm in front of the face as described previously (Nummela et al., 2017). The lens of −2.5 diopters was used for Marmoset M and of −2.0 diopters for Marmoset E. Eye position was calibrated at the start of each behavioral session using a Gabor windowed face detection task described previously (Mitchell et al., 2014, Nummela et al., 2017).
Eye position data were collected during the entire recording session. Raw horizontal and vertical eye position signals were smoothed offline with a median filter (five samples, 5 ms) and convolved with a Gaussian kernel (5 ms half width, out to 3 SD, −15 to 15 ms) to minimize high-frequency noise. For offline detection of saccadic eye movements, we used an automatic procedure that detected deviations in 2D eye velocity space (Engbert and Mergenthaler, 2006; Kwon et al., 2019). We computed horizontal and vertical eye velocity by taking the difference of the smooth eye position and then marked saccades by where the 2D velocity exceeded the median velocity by 10 SD for at least 15 ms (Engbert and Mergenthaler, 2006; Kwon et al., 2019) and merged any two saccadic events into a single saccade if they were separated by less than 5 ms. Saccade onset and offset were determined by the first and last time the 2D velocity crossed the median velocity threshold. Epochs with eye blinks (based on pupil size) are removed from analysis.
RF mapping of MT/MTC during free viewing task
Spatial receptive fields were estimated from the responses to a wide field stimulus consisting of large moving white dots (Fig. 2A). In brief, marmosets freely viewed a full-field display that consisted of white dots (230 cd/m2, 1 dva diameter) that appeared against a gray background (115 cd/m2) spanning ±20 dva on the horizontal and 15 ±dva on the vertical of the display. Each dot moved at 15° per second for 50 ms before being replotted to a new location in the full-field display. Dot motion was selected at random from 1 of 16 motion directions sampled around the circle. In each task trial, the screen would contain a fixed number of dots (sampled from 4, 8, 16, or 32) that would be viewed for 10 s. To encourage foraging near the center of the screen, a Gabor target (20% Michelson contrast, 1 dva diameter, one cycle/degree, with random orientation) appeared within 5° of the center superimposed with the dots flashing dots. If the marmoset's eye position acquired the Gabor target within a 2° diameter, a juice reward was delivered, and the Gabor target was replotted to a new location.
Off-line we corrected for eye position to represent the flashed stimuli in a grid of retinal coordinate locations and correlated the stimulus history with spike counts to estimate the RF (Fig. 2B). The full methods for estimating receptive fields have been described previously (Yates et al., 2023) and analysis code is available online (https://github.com/VisNeuroLab/yates-beyond-fixation). In brief, firing rate was computed as a function of the x–y retinal-based grid location (2 × 2° bins) where each bin contained a flashed dot or did not. To reduce correlation in the stimulus history of the moving dots, we only represented dots on the first video frame from their 50 ms lifetime, and the spatial position was registered by their location at the middle of the lifetime. The firing rate was computed for the onset of dots across the grid at different lag times in 10 ms spike counting bins (Fig. 2B, left). Those locations exhibiting firing responses significantly above the pre-stimulus baseline firing (from −100 to 0 ms lag, p < 0.001) were labeled as significant to mark the RF. A smoothed 2D contour was computed to circumscribe the peak of the RF at its half-maximum height relative to the baseline for the peak temporal lag (Fig. 2B, top right). Then the direction selective evoked responses were computed from the flash events inside the defined RF contour as a function of their motion direction, with error bars indicating two standard errors of the mean at each direction (Fig. 2B, bottom).
The tuning for motion in full-field mapping and also later for dot field stimuli in foraging were fit using a modified von Mises function. The tuning curve for firing rate $$mathtex$$R$$mathtex$$ was defined as a piecewise continuous function based on a bandwidth parameter, $$mathtex$$K$$mathtex$$, as:$$mathtex$$R = b + A\;{\rm exp}\lpar {K\lpar \cos \lpar {\theta -\hat{\!\theta }} \rpar -1} \rpar \rpar$$mathtex$$when $$mathtex$$K \gt 0$$mathtex$$ and otherwise as:$$mathtex$$R = b + A\;\lpar {1-{\rm exp}\lpar {-K\lpar \cos \lpar {\theta -\;\hat{\!\theta }-180} \rpar -1} \rpar } \rpar \rpar$$mathtex$$where $$mathtex$$b$$mathtex$$ is the baseline firing rate, $$mathtex$$A$$mathtex$$ is the amplitude, $$mathtex$$K$$mathtex$$ is the bandwidth, and $$mathtex$$\hat{\!\theta }$$mathtex$$ is the preferred direction. Von Mises functions have been used previously to describe motion tuning in area V1 (Patterson et al., 2013). We adapted that function to allow for curves with wider-than-cosine tuning. Cosine tuning in the function occurs as $$mathtex$$K$$mathtex$$ approaches 0. We allowed the curve to continue to be defined for negative values of $$mathtex$$K$$mathtex$$ adopting an inverted von Mises with the opposite direction preference (180° opposite) such that the peak location remained at the same preference but was wider-than-cosine tuning. The modified curve was fit by maximizing the likelihood of the spike counts observed for each motion direction bin assuming Poisson firing statistics (Truccolo et al., 2005). The error bars were generated by a 10-fold Jackknife procedure, and tuning width was estimated from the half width of the curve in degrees.
Neural inclusion criteria
Areas MT and adjacent MTC were identified in targeted recordings based on their direction selective responses and retinotopy (Rosa and Elston, 1998). While we targeted neurons in area MT, a small number of motion selective neurons from adjacent MTC (also called V4t in lower hemifield and MST lateral in upper hemifield for the macaque) may have been included that lay near the vertical meridian. MTC has been reported to contain neurons with similar sized RFs and a significant portion of those cells also have motion selective responses (Rosa and Elston, 1998). We used the response to random dot motion patches placed inside the RF during the foraging task (irrespective of saccade condition) to evaluate if neurons had significant visual and motion selective responses. We only included neurons if they had a significant visual response defined by an increase in spike counts from 50 to 100 ms following the onset of a dot motion stimulus (averaged across sampled directions) as compared to a baseline from −100 to 0 ms before onset (Signrank test, p < 0.05). We also required that neurons had a significant direction selective index (DSI) indicative of motion tuning. The DSI was computed from spike counts from 50 to 150 ms after the stimulus onset as$$mathtex$${\rm DSI} = \displaystyle{{\left\vert {\mathop \sum \nolimits_{{\rm n} = 1}^{16} R_ne^{i\theta_n}} \right\vert } \over {\mathop \sum \nolimits_{{\rm n} = 1}^{16} R_n}}$$mathtex$$where Rn represents the mean spike count in response to a motion direction. To compute confidence intervals on the DSI, we used a 10-fold Jackknife procedure and units were included if the lower bound on the 95% confidence interval was above a DSI of 0.05. Additionally, we required that a minimum of 48 trials with straight single-step saccades be completed for the RF location and at least 48 trials combined for the other two foraging locations.
Saccade foraging task
To study pre-saccadic attentional modulations, we designed a saccade foraging task performed by marmoset monkeys in which a saccade was performed to one of three equally eccentric peripheral motion apertures (Fig. 1A). In each session, one of the apertures was positioned to fall near the center of the RF of a neuron or the set of neurons under study, while the other two were placed 120° apart from it at the same eccentricity. The task trial began with the fixation of a small spot (0.3° radius, 0.5 cd/m2 center, 230 cd/m2 surround) within a 1.5° radius window for a delay uniformly distributed between 0.1 and 0.3 s, presented on a gray background (115 cd/m2). After the fixation period, the fixation point was offset and three dot motion apertures (as described earlier) appeared in the periphery. The monkey was given up to 1.5 s to make a saccade out of the fixation window to one of the apertures with the final eye position remaining within the target window for 0.25 s to confirm the saccade endpoint. We rewarded saccades to any location as long as they differed from the previous trial to encourage foraging. A correct choice was rewarded with 10–20 μl of liquid reward and the appearance of a marmoset face at the aperture location for 1 s, providing positive feedback. The juice reward consisted of marshmallows blended with water that were prepared fresh for each daily session. An incorrect choice back to a location sampled in the previous trial resulted in a black Gaussian spot filling the chosen aperture, as feedback for choosing the wrong location. The next trial proceeded at an interval of 1–2 s depending on juice rewards.
Trial inclusion criteria
We limited analyses to trials in which saccades to foraging targets were executed in a single step without preceding movements and after a minimum latency from stimulus onset. First, we detected any micro-saccades of amplitude greater than 0.5 visual degrees that occurred during the fixation period and excluded those trials. To ensure that the animal had time to see the stimulus before initiating their saccade, we excluded trials where the reaction time (saccade onset from stimulus onset) was smaller than 0.12 s. We also excluded trials where the animal made two smaller saccades that stepped to the aperture and trials in which the saccade end point fell short of the aperture center (>50% of the target eccentricity). Last, we require that saccades fall within a window that has a radius of 50% of the eccentricity of the center of the aperture in order to be counted within that aperture. When all criteria were applied, this resulted in excluding 21.4% and 13.1% of trials across sessions for Marmoset E and Marmoset M, respectively. We included a small percentage (10%) of catch trials that required the monkey to maintain central fixation. In these trials, the fixation period was extended to 0.5 s without any target apertures ever appearing. For holding fixation, the monkey was rewarded with both juice reward and a marmoset face that appeared at fixation.
Temporal separation of the stimulus and saccade onset epochs
We examined the neural firing response time-locked relative to the moment of stimulus and saccade onset. Because there is no extended delay period between stimulus and saccade onset under natural foraging tasks, as in a delayed saccade task, these two intervals will be partly overlapping depending on the saccadic reaction time, and it is important to choose analysis intervals that minimize their overlap. To examine the temporal response of neurons, we smoothed the peri-stimulus time histograms (PSTHs) for the “toward” and “away” conditions using a Gaussian temporal kernel (σ = 5 ms). The stimulus locked onset response revealed a visual latency of approximately 40 ms with a transient peak that rose and fell by 70–80 ms into a sustained response (Fig. 4A,B, left). We defined a stimulus–response epoch to be between 40 and 90 ms after stimulus onset to capture the early peak. The response when instead time-locked to saccade onset reveals a rise in firing rate continuing up to 20 ms after the saccade followed by suppression (Fig. 4A,B, right). We defined the pre-saccadic window to be between −30 and +30 ms from saccade onset. While previous studies using delayed saccade paradigms have used windows that begin from −100 ms, in our paradigm that interval would include trials with shorter saccadic latencies that include the stimulus evoked peak response. It might be difficult to identify the stimulus evoked response, as it would be convolved with the variability in reaction times, but it would still influence the firing rate in those epochs preceding saccade onset. We thus restricted analyses to reaction times later than 120 ms and set our pre-saccadic window to be no earlier than −30 ms from saccade onset such that no part of the stimulus evoked transient, which resolved to a sustained level by 90 ms, would be mixed with the saccade onset response. For the end of the saccade onset epoch, we chose +30 ms because the visual latency of neurons was no earlier than 40 ms, and thus responses out to that period would yet reflect visual motion induced by the saccade itself.
While selection of the stimulus and saccade onset epochs as described above reduces their temporal overlap, it also restricts the extent to which we can time-lock responses after stimulus onset or before the saccade. To address this limitation and extend those intervals, we performed a second analysis on firing responses that preferentially sampled longer reaction time trials at extended latencies. Specifically, at each stimulus locked latency, only trials that were not within 30 ms of saccade onset were included in computing the trial averaged firing rate. Likewise, at each saccade locked latency, only trials that were not within 120 ms of stimulus onset were included. This approach was applied to extend stimulus locked analyses for 65–115 ms and 90–140 ms epochs and also when using 50 ms sliding windows to compute the number of units significantly modulated by attention condition across time (Fig. 5A–C).
Although temporal intervals were selected to minimize overlap between stimulus and saccade epochs, we must also control for any differences in saccade reaction times between “toward” and “away” conditions when comparing modulation in their firing rates or tuning curves. For example, shorter reaction times in one condition could lead to differences in the adaptation state of the neuron, which if not corrected for, could produce differences in firing at the time of saccade onset. To control for those differences, we created resampled trial distributions for the “toward” and “away” conditions that matched the saccadic reaction times between conditions. For each session, the distribution of reaction times was computed in 10 ms bins, and stepping through each bin, we randomly sampled an equal number of trials from the condition with more trials in order to match the number to the condition with fewer trials. There are no clear signs of the stimulus evoked response contaminating the saccade onset epoch, nor of differences in adaptation between conditions. All comparisons between “toward” and “away” conditions applied this correction to be conservative, although all effects reported remained consistent without it throughout the results.
Firing rate and variability analyses
The mean and variability of the firing rate were assessed in the stimulus and saccade onset intervals. The mean firing rate was computed from all trials for either the “toward” and “away” conditions, and thus each of those conditions was sampled randomly over the 16 motion directions used in the task. To measure variability, we computed the Fano factor (FF), which provides a measure of variance in spike counts across trials that is normalized by rate. For a Poisson process, the spike count variability scales in proportion to the mean count, giving unity FF. However, nonlinearities such as the spike refractory period, burst firing, or super-Poisson fluctuations in rate produce dependencies on how the FF scales with a rate that deviates from linearity. Thus, it is necessary to match firing rates between the “toward” and “away” conditions before comparing them in order to disassociate changes in FF from changes in mean rate (Mitchell et al., 2007; Churchland et al., 2010). To match firing rates, we first computed the mean spike counts for each of the 16 motion directions in each of the two saccade conditions (16 points of mean vs variance for each condition). We performed a search in random order for each of the 16 points in the “toward” condition to find the point with the mean rate in the “away” condition most closely matching its mean rate, without reference to whether motion direction matched, and accepted those as a pair if the two rates matched within 5%. Repeating this procedure without replacement across all points identified the region of overlap between the two distributions that were constrained to have nearly matching rates (<5%). The FF was then computed by dividing the spike count variance by the rate at each point and then averaging those ratios for each saccade condition. In one monkey where we performed recordings using linear arrays, it was also possible to isolate many simultaneously recorded pairs of neurons in order to estimate the noise correlations. Spike counts were computed during the pre-saccadic epoch from −30 to 30 ms at saccade onset, and correlations were computed within trial sets where the same motion direction was in the RF and then pooled across motion directions.
Noise correlations
In one marmoset where array recordings were made, it was further possible to analyze noise correlations between pairs of simultaneously recorded neurons. Spike counts for noise correlations were computed during the pre-saccadic epoch from −30 to 30 ms at saccade onset. For each neuron, the mean response to each of the 16 motion directions per attention condition (toward and away) was first subtracted from those pre-saccadic counts to eliminate any stimulus-driven response. Then the correlation coefficient between the residual counts of the pair was computed to give the noise correlation. To measure the signal correlation between pairs, we first averaged the 16 points of the tuning curve for the motion direction of each neuron across both attentional conditions during the stimulus onset period (40–90 ms from stimulus onset). The signal correlation was computed as the correlation coefficient between the 16 points on the tuning curve for each pair of neurons considered. Signal correlation defines how much overlap there is in the motion tuning between neural pairs. Across the population of pairs, the noise correlations were binned and plotted for different signal correlations from −1 to 1 with bins ±0.1 wide and stepping in bin size with increments of 0.05.
Mutual information analysis
In the saccade onset interval, we examined if there was an increase in neural sensitivity for motion direction. We measured sensitivity from the distributions of firing rates across the 16 motion directions by computing the mutual information (MI). The MI quantifies how much information one variable (such as firing rate) provides about another variable (such as motion direction) measured in bits. For our purposes, it provides an estimate of sensitivity, much as classic measures like the AUC in ROC do for the case of discriminating two sensory conditions, but it is readily generalized to more than two stimulus conditions and thus is highly suitable for use with tuning curves (Hatsopoulos et al., 1998). We calculated a MI for both the “toward” and “away” conditions using the equation:$$mathtex$$I\lpar {x\semicolon \;y} \rpar = \sum\nolimits_{x}\sum\nolimits_{y} p\lpar {x \;y} \rpar {\rm log}\lpar {p\lpar {x \;y} \rpar {\rm \sol }p\lpar x \rpar p\lpar y \rpar } \rpar {\rm \;\;\;}$$mathtex$$where x is the motion direction stimulus shown and y is the measured firing rate. Computing the MI requires that we estimate the probability distributions (p(x), p(y), and p(y|x), using histograms and binning of firing rate data per stimulus direction. For the stimulus variable x, we binned based on the motion direction, where we had sampled from 16 possible directions. However, to further enforce a smoothing constraint on our data and represent that adjacent motion directions are related, we further pooled the firing rate data from the two adjacent motion direction bins arranged on the circle, thus giving on average three times as much firing rate data per motion direction bin and forcing a smoothing constraint on the data. The binning of the firing rate, y, was determined based on a “goodbins” function described in Scott (1979), applied to the total firing rate distribution across all motion directions to estimate the marginal distribution p(y). Then the histogram of firing rates conditioned on each stimulus direction, p(y|x), was computed using those bins that were established to describe p(y).
Comparing motion tuning between toward and away conditions
To examine neural tuning changes across the “toward” and “away” conditions, we fit the motion direction tuning of individual MT units with a modified von Mises function as described earlier in methods. The von Mises curves provide parameters for amplitude, baseline, and width for each tuning curve in the two conditions. We constrained fits to share the same preferred direction between saccade conditions. Only units where the net quality of the curve fits for the two conditions (pooled over both curves) exceeding an R-squared of 0.5 were included. This significantly reduced our neural population included for von Mises-based analyses (Marmoset E had 33 units, and Marmoset M had 287 units).
To quantify the modulation of parameters between the “toward” and “away” conditions, we calculated an attention index (AI) for each of the parameters considered as:$$mathtex$${\rm AI} = \lpar {{\rm Towards}-{\rm Away}} \rpar {\rm \sol}\lpar {\rm Towards} + {\rm Away\rpar }$$mathtex$$where AI ranges from −1 to 1, with zero indicating no change. The AI metric is useful for emphasizing percentage changes in a variable. All statistical tests on AI indices were nonparametric (signed-rank or rank-sum tests, p < 0.05).
Linear prediction model
To evaluate which modulations in tuning curves contributed to increases in neural sensitivity, we fit a linear model. The input parameters to the linear regression included the AI for the von Mises fit parameters of baseline, gain, and curve half width, as well as the AI for the rate-matched FF. We also include a constant term, giving five input parameters. The output aimed to predict the AI for MI of each unit. Linear fit coefficients (and error bars) were fit using the “regress” function in MATLAB (version 2018). Only units that had reliable von Mises fits for both the “toward” and “away” conditions (R2 = 0.5) were included.
Nonparametric fit of tuning curves
Von Mises fits are biased to fit better for neurons with high firing rates and high motion selectivity. Therefore, we sought an alternative metric to quantify changes in baseline, amplitude, and tuning width in the total population. We smoothed the mean responses per motion direction (raw tuning curve) by pooling across the two adjacent motion directions for each point in the tuning curve. Then we computed the preferred motion direction by computing the resultant vector (circular statistic) on the mean smoothed tuning curve averaged over the “toward” and “away” conditions. Then smoothed responses in each attention condition were rank ordered by their angular distance from that preferred direction, with the least preferred direction being defined as 180° opposite of preferred. Baseline was computed from the rate at the least preferred index while amplitude was the difference between the rate at the most and the least preferred. Tuning width was computed by subtracting the baseline (zeroing any resulting negative values) and computing the width from the preferred direction at which half the area under the baseline subtracted curve was contained.
Results
We used a saccade foraging task to measure the modulation of neural firing and motion tuning during pre-saccadic attention (Fig. 1A). In each trial, the monkey was trained to maintain fixation on a central point for 100–300 ms, after which three random dot field motion stimuli appeared in peripheral apertures of equal eccentricity and separation from each other. The monkey responded by making a saccade to one of the three apertures immediately after stimulus onset. While monkeys performed this foraging task, we recorded from individual neurons in visual areas MT/MTC (Fig. 1B). The apertures were positioned such that one of them was centered inside the RF of the neurons under study. Thus, across trials the monkeys performed saccades either toward the RF (“toward” condition) or away from the RF (“away” condition). Because we examined responses in the pre-saccadic epoch while the eyes were still at fixation, the sensory stimuli were matched between these conditions so we could isolate the effects of saccade planning on neural responses. Our goal was to measure how neural tuning curves differ for saccades toward the RF as opposed to away from it.
Recent studies in human psychophysics suggest that pre-saccadic attention differs from covert attention in important ways, specifically involving an automatic narrowing of sensitivity around the feature of the saccade target, whereas covert spatial attention applies a uniform gain at the attended location independent of tuning. A narrowing in feature sensitivity could manifest at the level of individual neuronal tuning curves in a variety of ways. For example, in covert attention tasks, previous studies have identified individual MT neurons that change their tuning according to widely used gain models, including spatial gain (McAdams and Maunsell, 1999) and feature similarity gain as observed when feature-based attention is involved (Treue and Martinez-Trujillo, 1999; Martinez-Trujillo and Treue, 2004). If pre-saccadic attention only enhances spatial gain, there will be a uniform multiplicative increase across all motion directions (Fig. 1C, blue curve). Considering how that would impact the proportional gain across the tuning curve, such a spatial gain would give constant positive value (flat line) across directions (Fig. 1D). Alternatively, the feature similarity model would preferentially increase the gain for preferred motion directions while suppressing gain for nonpreferred directions (Fig. 1C, green curve). In that case, the proportional gain would not be uniform across the curve but rather show an enhancement at the peak of tuning and suppression away from the peak (Fig. 1D, green curve). We designed our saccade foraging task to sample different motion directions in the pre-saccadic epoch of response in order to reconstruct full neural tuning curves and test if changes in tuning favor feature-specific gain as opposed to spatial gain.
Measurement of MT/MTC receptive fields during free viewing
To properly position peripheral stimuli in the saccade foraging task, we must first determine the receptive fields of MT/MTC neurons. As marmosets are less able to maintain fixation on central locations for extended periods during RF mapping (Mitchell et al., 2014; Yates et al., 2023), this presented a unique challenge. We developed a novel free viewing approach to map the receptive fields of MT/MTC neurons (Yates et al., 2023; see Methods). In brief, marmosets were allowed to explore a full-field display of moving dots, with 4–16 dots displayed at any time and each dot being 1° visual angle (dva) in diameter. The dots flashed at random locations and then moved along a single direction of motion at 15 dva/s for a duration of 50 ms (Fig. 2A). Off-line we corrected for eye movements during viewing of the full-field stimulus and reconstructed the stimulus history in a retinotopic coordinate frame to assess visual receptive fields. Those visual locations that exhibited responses significantly above the pre-stimulus baseline time-locked to a dot appearing were labeled to identify the RF (Fig. 2B, top). Then the responses to dots within the RF were further broken down based on their direction of motion to estimate the neuron's motion tuning (Fig. 2B, bottom). The example cell illustrated had a visual latency of 40–50 ms with a RF in the lower left visual quadrant and strong motion tuning as reflected by a direction selective index (DSI) significantly above zero, consistent with a typical MT/MTC neuron.
We recorded single and multiunit activity from two marmoset monkeys across 38 and 52 experimental sessions, respectively. Neural recordings were targeted for MT area based on retinotopy and motion selectivity, but some neurons with receptive fields near the vertical meridian may have been included from adjacent MTC area, which like area MT has a significant portion of neurons with motion selective responses and comparably sized receptive fields (Rosa and Elston, 1998). We only included neurons for analyses if they exhibited a visual response with significant direction selective tuning (DSI > 0.05; see Methods) and a minimum evoked firing rate of 1 spike/s (see Methods). The first marmoset monkey (Marmoset E) was recorded using single tungsten electrodes during initial studies while advanced array recording methods were still in development. We obtained 116 units of which 87% showed significant visual responses, and of those 75% had significant motion tuning, giving 73 units in total (39 single units, 34 multiunits). The second marmoset monkey (Marmoset M) was recorded after we had refined our recording methods to include a 64-channel linear array yielding higher cell counts. We obtained 872 units of which 60% had significant visual responses, and of those 90% had significant motion tuning, giving 472 units in total (444 single units, 28 multiunits). In all subsequent analyses, the two marmosets are presented separately, not only because the second animal would dominate the sample based on neuron numbers but also because the different recording methods impose different sampling biases across neuron types and cortical layers.
The distribution of visual field locations sampled for individual recording sessions across the two marmosets covered the upper and lower left hemifield (Fig. 2C). For array recordings, we typically isolated several neurons in a single recording but because the marmoset cortex is smooth our linear arrays were oriented down a single cortical column and thus had largely overlapping receptive fields. We thus were able to test a single visual field location that encompassed different neurons on the same array. We sampled from the left hemifield in both animals. Biases toward lower or upper visual fields varied between animals because of the position of blood vessels in their tissue, which we avoided in placing electrodes. In one monkey (Marmoset E), the majority of recordings were obtained from the lower left quadrant, whereas in the other (Marmoset M), most recordings were obtained from the upper left quadrant. Both marmosets exhibited a wide range of motion direction preferences (Fig. 2D), which is roughly consistent with uniform direction preferences reported in macaque MT (Albright, 1989). The larger sample of neurons (N = 472) from Marmoset M did not differ significantly from a uniform distribution (Rayleigh test, p = 0.3220, z = 1.1337). Marmoset E (N = 73) showed a significant bias away from uniform (p = 0.0208, z = 3.8460) favoring leftward motion (180°). The width of motion tuning was clustered between 30° and 120° in both animals (Fig. 2E), with a median of 60° and 76° in Marmoset E and M, respectively, values that are comparable to that previously reported for dot motion in macaque MT (Albright, 1984). Marmoset E did show significantly tighter width tuning on average than that of Marmoset M (rank-sum test, p = 0.00227). Despite variation in electrode placement, recording methods, and tuning, we find similar qualitative patterns of neural modulation across the two animals.
Marmoset behavior in a saccade foraging task for study of pre-saccadic attention
While the smooth cortical surface of marmosets facilitated access of areas MT/MTC for neural investigation, a key disadvantage of working with marmosets is the number of trials they can perform in highly constrained behavioral tasks (Mitchell et al., 2014). In the macaque, studies of pre-saccadic attention have imposed constraints to maintain central fixation that are similar to covert attention task, with an extended period of fixation prior to making a saccade to a cued target (Moore and Chang, 2009; Steinmetz and Moore, 2014). Here we took a different approach with marmoset monkeys to instead optimize the number of saccade trials while minimizing the duration of fixation and thus the total duration of individual trials. Marmosets completed 413 and 578 trials on an average session that met criteria for obtaining accurate initial fixation for a brief fixation epoch (1.5 dva fixation window) and then making an accurate saccade targeting one peripheral aperture (see Methods). To encourage foraging between different locations across trials, we provide a juice reward to the animal if it selected an aperture that differed from that selected in the previous trial. Initial piloting of the task demonstrated that marmosets foraged for more trials when the task included three apertures instead of two. During each recording session, we positioned one of the three apertures over the neural receptive fields under study. By encouraging foraging between locations, we were able to sample neural responses both when saccades were made toward or away from the RF.
Marmoset monkeys acquired fixation accurately to initiate trials and sample across aperture locations during the foraging. The fixation and saccade end points from a typical behavioral session are illustrated in Figure 3A. The color indicates which aperture location was selected in each trial, with red indicating saccades toward the RF and dark or light blue locations away from the RF. Zooming in on the period of central fixation, the 2D eye position clustered within the fixation window during the 100 ms epoch preceding saccade onset and overlapped regardless of the target selected (Fig. 3B). The overlap in central fixation was consistent across sessions and the two monkeys when comparing between conditions where the saccade was toward (red) or away (blue) from the RF (Fig. 3C). Across trials, monkeys were sampled across all three aperture locations as illustrated by the distribution of saccade end points for the example session (Fig. 3A). Average trial counts favored sampling of the RF location (red) in both animals reflecting an unintended alignment in their spatial biases (Fig. 3D). Although marmosets foraged different locations across trials, they were not perfect at avoiding a return to the location selected in the previous trial (Fig. 3D, filled regions). Their return saccades to previous locations reflected the sampling bias toward the RF location. This bias was not intentionally rewarded by the task design but may have reflected a biased sampling across days for those recording locations that successfully yielded an adequate number of trial repeats (minimum, 48 trials) at the RF location. However, we controlled for the potential impact of location bias in subsequent analyses. For example, the biases toward the RF were reflected in the distribution of saccadic latencies. Both monkeys made faster saccades toward the RF location (shown in red) as compared to the away locations (shown in blue) with median latency shifts of 10 and 15 ms, respectively (Fig. 3E). To control for differences in the saccadic timing between toward and away conditions, we resampled the trials from each session so that all subsequent analyses comparing the firing rates between those conditions first match the saccadic latency distributions (see Methods). This enables us to match saccade timing in addition to sensory conditions between toward and away conditions for further analyses.
Marmoset monkeys exhibit single-unit neural signatures of pre-saccadic attention
As illustrated for a single example cell, the firing rate in pre-saccadic epoch increased for saccades made toward the RF location (Fig. 4). Spiking was time-locked across trials to the saccade onset for the toward RF condition (in red) and away RF conditions (in blue) (Fig. 4A). Averaging firing across the trials the mean rate showed no significant rate modulation when time-locked to stimulus onset but had grown toward a modest increase around the time of saccade onset (Fig. 4B). The response at the stimulus onset from 40 to 100 ms did not exhibit a significant difference for toward versus away conditions (58.5 vs 59.8 sp/s; rank-sum, p = 0.7625). The rate had a modest increase approaching significance for the toward condition from −30 to 30 ms around the saccade onset (49.9 vs 41.5 sp/s; rank-sum, p = 0.056). The mean firing rate, however, is averaged across trials that included different stimulus motion directions that sampled from 16 different directions from 0° to 360°. When instead breaking out changes in rate as a function of direction (Fig. 4C), we observed a highly significant increase in the gain of the tuning curve of the example neuron in the toward as compared to away condition (toward Amp, 100.4; away Amp, 76.26; Z-transform based on von Mises fit confidence intervals, p = 0.00024).
Previous studies have reported increases in neural sensitivity during pre-saccadic attention based on comparing responses to preferred and nonpreferred stimuli (Moore and Chang, 2009; Steinmetz and Moore, 2010). To assess the sensitivity of neurons to stimulus motion based on their tuning curve, we instead computed the MI. The MI provides a measure that is similar to the AUC in the receiver operating characteristic (ROC) function but generalized for a complete tuning curve rather than just the case of two stimuli (Hatsopoulos et al., 1998). Similar to AUC, the MI depends not only on the separation in firing rates across the different stimulus conditions but also on the variability in responses, which describes neural sensitivity based on how well the spike count distributions for different stimuli can be separated. For the example neuron, there was almost no difference in the AUC measure for the toward (0.99) versus away (0.97) condition (Fig. 4D). This simply reflects a saturation in that measure as the preferred and nonpreferred stimuli in the tuning curve were extremely well separated in response. However, the MI remains sensitive to the separation across the entire curve and reflects an increase from 0.570 to 0.756 bits of information (Fig. 4E), which approached significance for this neuron (Z-transform based on confidence intervals, p = 0.065).
Across the population, we found a diverse pattern of modulation in tuning curves. Similar to what has been observed in covert attention, we find that some cells show an increase in baseline firing rate with attention (Fig. 4F), while others show an increase in gain (Fig. 4C). However, we also find example cells with a narrowing in width (Fig. 4G) that would be consistent with a feature-based gain. However, there are also many neurons that exhibit the opposite pattern, showing a broadening in half width (Fig. 4H). Thus, we sought to determine how the population changed across a variety of measures and relate them. We examined the increase in mean firing rate, the increase in MI, and finally how those changes related to the variety of modulations observed for neural tuning curves.
Across the population of neurons, there was an increase in firing rate and MI for saccades toward the RF at saccade onset (Fig. 5). In each monkey, we plot the firing rates over time averaged across neurons after normalizing to the peak response. Marmoset M was recorded using linear arrays with higher neuron yields, so we plot each animal separately to avoid it from dominating a pooled sample (Fig. 5A,B). To quantify the distribution of effects on rate in the population for each monkey, we used an AI defined as (toward rate—away rate)/(toward rate + away rate). We first quantified if there was a significant increase in firing rate time-locked to the stimulus onset epoch (between 40 and 90 ms after stimulus onset). One of the two monkeys exhibited a modest 3% rate increase at stimulus onset for trials where saccades were made toward the RF (Marmoset M: signed-rank test, median +3.0%, AI = 0.015, p = 0.00001), while the other monkey had no significant change (Marmoset E: signed-rank test, median −0.3%, AI = −0.002, p = 0.065). A previous study of covert attention in macaque V4 found weak attentional modulation for high contrast stimuli in the transient response at stimulus onset that grew larger into the sustained response (Reynolds et al., 2000). Thus, we also examined if modulation increased in the sustained response in later 50 ms epochs at 65–115 ms and 90–140 ms. Again, we found similar modest increases that were significant for Marmoset M (65–115 ms: signed-rank test, median +3.3%, p = 0.0013; 90–140 ms: signed-rank test, median +3.8%, p = 0.0003) but no consistent change for Marmoset E (65–115 ms: signed-rank test, median +1.5%, p = 0.49; 90–140 ms: signed-rank test, median −1.1%, p = 0.94). Further, we examined the fraction of neurons with individually significant increases in rate using a sliding window around the stimulus onset (Fig. 5C, left). Marmoset M had a fraction significantly above chance peaking 100 ms after stimulus onset, while Marmoset E showed a similar trend that did not reach significance. Taken together, we found that modulation in firing rate was not strong when time-locked to the stimulus onset, even progressing into the sustained stimulus response.
In contrast to the stimulus onset, both monkeys showed reliable increases in firing rate immediately before and during saccade onset (Fig. 5A,B, right). Examining the fraction of neurons with individually significant increases in rate (Fig. 5C, right), we find both animals showed a positive effect. This was also reflected by the distribution of AI indices taken from the epoch centered ±30 ms around saccade onset, as reflected by a rightward shift in the distribution for both animals (Fig. 5D). At that time, Marmoset E showed a 14.2% median increase, and Marmoset M showed a 13.8% median increase (signed-rank test, Marmoset E: median AI = 0.0738, p = 1.134 × 10−15; Marmoset M: AI = 0.0615, p = 4.857 × 10−29), and these increases did not differ from each other significantly (rank-sum test, p = 0.7418). Of 79 units, Marmoset E had 8 with significant increases and none with decreases, while of the 472 units in Marmoset M, there were 59 units with individually significant increases and 9 with decreases. Thus, changes in firing were predominantly related to the onset of the saccade, rather than stimulus onset, consistent with previous studies of pre-saccadic attention. Subsequent analyses focus on changes in tuning and sensitivity in the pre-saccadic epoch.
The observed increases in firing rate, however, do not necessarily imply an increase in neural sensitivity. Changes in sensitivity depend on how the tuning curve and variability in firing is modulated. We therefore examined FF, a measure of firing variability that is normalized by the firing rate. We found only modest reductions in FF for the two marmoset monkeys (Fig. 5E) that were not significantly different for toward versus away conditions (signed-rank test, Marmoset E: −3.6%, AI = −0.0188, p = 0.595; Marmoset M: −1.2%, AI = −0.0065, p = 0.595). Furthermore, few units were individually significant for changes in the AI for FF (Marmoset E had no units with increases or decreases of 73; Marmoset M had 1 unit with an increase and 0 with decreasing of 472).
In the second monkey, where we had performed linear array recordings, we examined the shared variability between pairs of neurons and noise correlations and found a modest reduction consistent with the modest changes in FF during the pre-saccadic epoch (Marmoset M: N = 21,169 pairs, −4.1%, AI = −0.019, p = 0.012) (Fig. 5F). This result differs from covert attention, where previous studies have found much larger reductions both in FF and noise correlations (Cohen and Maunsell, 2009; Mitchell et al., 2009).
However, despite the modest reductions in variability, we nonetheless observed consistent increases in the MI across both monkeys for the toward RF as compared to away RF condition (Fig. 5G). Both animals had an individually significant increases, though they did differ in magnitude with Marmoset E having a median increase of 27.33% (signed-rank test, AI = 0.1202, p = 1.5709 × 10−4) and Marmoset M showing a modest increase of 3.9% (signed-rank test, AI = 0.0193, p = 0.0161), a difference between animals that was significant (rank-sum test, p = 0.0015). Marmoset E had 11 units with significant increases and 2 units with significant decreases in AI of 73 units. Marmoset M had 21 units with significant increases and 14 units with significant decreases in AI of 472 units. Thus, both marmoset monkeys show increases in rate and in sensitivity, which qualitatively are similar to sensitivity enhancements found in macaques. However, there remains a question about why the magnitude of their effects differ. We thus sought to relate these changes to underlying changes in neural tuning curves and, in doing so, also address to what extent either of these animals might show changes consistent with a feature-specific selection in gain.
Changes in neural tuning enhance sensitivity but not feature gain
To quantify changes in tuning, we fit an adjusted von Mises function to the motion direction tuning in the saccade toward and away conditions. The von Mises function fits a single peaked tuning curve to the neural responses as a function of motion direction or orientation (Patterson et al., 2013). It is defined by four parameters: a baseline, an amplitude (gain), a width of the tuning curve, and a preferred direction. To make meaningful comparisons of the tuning between the saccade conditions, we limited our analyses to those neurons for which the model fit was better than a minimum R-squared criterion (R2 > 0.5; see Methods). This included 33 cells from Marmoset E and 284 cells from Marmoset M. Although this sample was biased toward neurons with higher mean firing rates and directional tuning, it allowed more reliable comparison of changes in tuning.
Across the population, we found that tuning curves showed increases in either baseline firing or gain but no net change in tuning width (Fig. 6). To quantify the effects, we computed attention indices [AI = (toward-away)/(toward + away)] for each of the fit parameters. Baseline firing rates increased reflected by a rightward shift in the distribution of attention indices (Fig. 6A), with Marmoset M showing an 18.9% median increase that was significant (signed-rank test, AI = 0.0865; p = 1.5010 × 10−7) and Marmoset E approaching significance with an 8.4% median increase (signed-rank test, AI = 0.0404; p = 0.2278). The difference between animals in baseline, however, was not significant (rank-sum test, p = 0.425). Marmoset E had 6 units with significant increases and 3 with decreases in AI of 33 units. Marmoset M had 58 units with significant increases and 13 with decreases in AI. The modulation of gain also showed significant increases for both animals (Fig. 6B) but with a robust increase of 29.5% in Marmoset E (signed-rank test, AI = 0.123; p = 0.0011) and only a modest increase of 5.2% in Marmoset M (signed-rank test, AI = 0.0253; p = 0.0018), which did differ significantly between monkeys (rank-sum test, p = 4.9683 × 10−4). Marmoset E had 13 units with significant increases and 1 with decreases in AI. Marmoset M had 42 units with significant increases and 25 units with decreases in AI. In contrast to baseline and gain, the tuning width exhibited no net increase or decrease in either monkey (Marmoset M: +1.1%, AI = 0.0029; p = 0.9321; Marmoset E: −0.9%, AI = −0.0024; p = 0.5259), which was reflected by attention indices clustered around zero (Fig. 6C). Marmoset E had 2 units with significant increases and 0 with significant decreases in AI. Marmoset M had 35 units with significant increases and 26 units with decreases in AI. Thus, across the two animals, increases in baseline and gain contributed to average changes in tuning, while changes in tuning width were not consistent.
Although there was no net change in tuning width across the population, individual neurons did exhibit significant changes, with nearly equal numbers showing narrowing as much as broadening. The number of individually significant changes in tuning width was much higher than would be expected by chance with 19.9% showing significant changes (Jackknifed confidence intervals on von Mises fits, 63 of 317 total units, which differed from 5% based on a binomial test, p < 0.0001). We validated in a control analysis with trials shuffled between the saccade conditions that 5% or fewer units would be expected to show significant effects by chance. Thus, while there is no net trend for motion tuning to narrow, different subpopulations exhibit significant narrowing and could, in principle, preferentially contribute to read-out mechanisms downstream.
While increases in gain and baseline were significant in both animals, the two animals differed substantially in the magnitude of these relative increases with one monkey showing predominantly increases in gain and the other a stronger baseline increases with modest changes in gain (Fig. 6D). We considered to what extent those differences might also reflect changes consistent with feature-specific changes in gain. A spatial modulation would show a consistent increase in gain across all directions regardless of a neuron peak in tuning, while a feature-specific modulation for the saccade target would preferentially increase the gain at the peak while suppressing it in the flanks (Fig. 1D). Overall, the gain remained uniform across motion directions in both monkeys, more consistent with a spatial rather than feature gain modulation (Fig. 6E). Marmoset E shows a weak trend for a larger gain at the preferred direction than at nonpreferred direction, but comparing AI values between the preferred and nonpreferred directions, this difference was not significant (rank-sum, median AI difference = +0.055, p = 0.3669). Meanwhile Marmoset M showed the opposite trend with smaller gain at the preferred direction than the nonpreferred direction, a difference that was significant but in the opposite direction predicted by the feature gain model (rank-sum, median AI difference = −0.049, p = 0.0098). Thus, the changes in tuning largely support a uniform spatial gain, rather than feature gain, which would be more consistent with previous studies of covert attention.
Finally, we examined how well we could predict the observed changes in neural sensitivity based on the underlying changes in tuning. A linear predictor of the attention indices (AI) for MI (sensitivity) for each neuron was fit using the AI for FF, baseline, amplitude, and tuning width as input parameters (Fig. 6F, left). Although the MI is a nonlinear measure of sensitivity, it was fit well overall by these variables with an R = 0.764 (Spearman's rho, p = 1.663 × 10−54). The strength of the linear coefficients in the fit prediction also provides an estimate of their importance for increases in sensitivity (Fig. 6F, right). We find that changes in gain predominately drove the changes in sensitivity with the largest prediction coefficient, while baseline and FF had a weaker and opposite influence. Increases in tuning width also tended to reduce predicted sensitivity but did not reach significance. Overall, changes in sensitivity were primarily driven by increases in gain, and thus the larger increases in sensitivity for Marmoset E can be explained by that animal's larger increases in gain.
To probe for distinct functional groups within the population, we examined how modulations in baseline, amplitude, or width co-varied across neurons. The full population of neurons was included in this analysis using nonparametric measures of baseline, amplitude, and width based on a smoothed version of the raw tuning curve (see Methods, nonparametric fits). This provides an alternative measure to control for any biases present in fitting the von Mises function, which if nothing else, performs poorly in fitting units with either low firing rates or weak tuning. We found that the main results based on von Mises fits (Fig. 6) replicated using these nonparametric measures with the full population. Again we find pre-saccadic increases in baseline and gain, but no significant changes in width [Marmoset E (N = 73) and Marmoset M (N = 472); baseline, median modulation +3.4% (p = 0.1691) and +18.6% (p < 0.0001); amplitude, +34.6% (p < 0.0001) and %+5.5 (p = 0.0003); width, −4.1% (p = 0.7897) and −0.6% (p = 0.4254)]. We then examined if there was any distinct clustering of the attention indices (AI) using these metrics. Although we did not observe any significant clustering into distinct types, we did find that increase in gain during pre-saccadic attention was negatively correlated with increases in baseline rate (Fig. 7A) and with increases in tuning width (Fig. 7B), whereas there was only a marginal correlation between baseline changes and tuning width (Fig. 7C). These results were consistent in both monkeys for the negative correlation between gain and baseline increases (pooled: r = −0.368, Spearman's rho, p = 8.694 × 10−18; Marmoset E: r = −0.494, Spearman's rho, p = 3.14 × 10−5; Marmoset M: r = −0.335, Spearman's rho, p = 2.99 × 10−13) and for the negative correlation between gain and width increases (pooled: r = −0.370, Spearman's rho, p = 1.23 × 10−16; Marmoset E: r = −0.276, Spearman's rho, p = 3.14 × 10−5; Marmoset M: r = −0.391, Spearman's rho, p = 2.99 × 10−13). The negative correlation between width and baseline was marginal (pooled: r = −0.084, Spearman's rho, p = 0.057; Marmoset E: r = −0.067, Spearman's rho, p = 0.576; Marmoset M: r = −0.083, Spearman's rho, p = 0.079). If we examine the subset of neurons well fit by von Mises functions, they had similar correlations for gain versus baseline and gain versus width based on the AI from their von Mises fit parameters (pooled: baseline vs gain, r = −0.175, Spearman's rho, p = 0.0018; gain vs width, r = −0.255, Spearman's rho, p < 0.0001) but additionally showed a significant negative correlation between baseline and width increases (r = −0.347, Spearman's rho, p < 0.0001). The negative correlation between width and baseline for the von Mises fits could reflect a bias in how the model parameters trade off in the fit, particularly as that correlation was weak for the nonparametric fits. In summary, while we did not find distinct clustering of types by attention modulation, there were clear systematic relationships between the changes in tuning with pre-saccadic attention, and in particular the negative relation between increases in gain versus baseline and gain versus width was robust using different methods of estimating the parameters.
To further validate the relationship between gain increases and baseline reductions and width narrowing, we performed median splits of the population based on the modulation (AI) of each parameter and examined the averaged normalized tuning curves with each median split shown as half of a curve reflected around the preferred direction (Fig. 7D–F). Consistent with the negative correlation between AI baseline and gain, we find that neurons exhibiting above median increases in baseline show minimal changes in gain (Fig. 7D, right split) whereas those exhibiting weak or negative changes in baseline show gain increases (Fig. 7D, left split). Also consistent, we find that neurons exhibiting above median gain increases show minimal baseline increases (Fig. 7E, right split) whereas those with weak or negative gain increases show baseline increases (Fig. 7E, left split). Last, the negative correlation between AI gain and width was also reflected in that those units showing an increase in tuning width showed minimal changes or reductions in gain (Fig. 7F, right split) whereas those showing narrowing (decrease) in width had gain increases (Fig. 7F, left split). Thus, although we do not find distinct clustering of types of modulation, we do find that those neurons showing stronger increases in gain typically also show weaker to no increases in baseline rate and a narrowing of tuning width, both of which are consistent with enhanced tuning and motion sensitivity.
As a final consideration, we performed a control analysis to rule out that behavioral preferences to select the RF location over other away locations could have explained our results. We computed an AI to represent the location bias for each behavior session as the number of saccades toward the RF minus the number away, normalized by their sum (i.e., an AI for saccade location preference). We tested for correlation between AI for location preference and each of the tuning parameters (baseline, gain, and tuning width). For each monkey and across all measures, there was no significant correlation (Spearman's rho, p > 0.05). For Marmoset M, there was a correlation approaching significance for the baseline parameter (r = 0.083, Spearman's rho, p = 0.058). We further tested if this correlation might be present in the epoch from 40 to 100 ms after stimulus onset, where this animal had exhibited a modest increase in mean rate for the toward RF condition (3.0% increase in rate, p = 0.00001, Fig. 5B). Indeed, we did find a significant correlation in that early epoch with location biases in Marmoset M (r = 0.102, Spearman's rho, p = 0.029). This suggests that locations biases influenced early firing in the stimulus epoch for that animal; however, they did not account for the larger modulations found in the saccade onset period, which remained similar in magnitude after removing location biases as covariate (Marmoset M: baseline increase, +22.3%, AI = 0.100, p = 1.37 × 10−5). Thus, while location biases may have contributed to early increases in rate at stimulus onset for Marmoset M, they do not fully explain the changes reported at saccade onset.
Discussion
The current findings did not support that pre-saccadic attention automatically engages feature selection for the saccade target, at least not at the level of single units. Recent human psychophysics suggested that pre-saccadic attention differed in important ways from covert attention, one of which involved an automatic feature selection for the target reflected by narrowing in psychophysical sensitivity around its features (Ohl et al., 2017; Li et al., 2021a). At the neural level, studies of feature-based attention have suggested there should be an increase in gain at the preferred direction while nonpreferred directions are suppressed (Treue and Martinez-Trujillo, 1999; Martinez-Trujillo and Treue, 2004). In our study, we analyzed neural tuning curves for motion direction in areas MT/MTC during the pre-saccadic epoch, aiming to discern if there were any indications of modulation reflecting feature selection of the saccade target. However, we found no strong evidence for feature gain across the total population. While some neurons exhibited narrowing in their tuning, others showed the opposite pattern just as frequently. Overall, the magnitude of gain was largely uniform across motion directions, which is instead consistent with the spatial gain, which has been previously observed in studies of covert attention (McAdams and Maunsell, 1999). While our findings were limited in addressing the read-out of activity at the population level, which may be more closely linked to behavior, we could at least conclude that the patterns of modulation among single neurons were highly similar to that found in covert attention, rather than feature-based attention.
Previous human studies that found feature-specific enhancements employed tasks that additionally require a stimulus discrimination. The foraging task required only that marmosets plan a saccade toward or away from the receptive fields of neurons under study, without any demands to discriminate the target. Thus, our task isolated aspects of pre-saccadic attention that were only obligatory to saccade planning. While we found increases in neural firing and sensitivity to confirm the automaticity of pre-saccadic enhancement, we found no feature-specific enhancements for the saccade target as in some previous human studies (Li et al., 2016; Ohl et al., 2017). Those differences may simply reflect that the brain areas we were studying are not involved in those aspects of pre-saccadic attention, or differences between species. However, it was also possible that task differences underlie the lack of feature-specific enhancements. Human tasks involved a discrimination of a grating titled left or right from vertical that flashed briefly within a sequence of 1/f masking noise. Such a task is more likely to engage feature attention for the reference grating to isolate it from masking noise. While one of these previous studies showed that covert and pre-saccadic attention differed under identical discrimination demands (Li et al., 2016), it is possible that some degree of feature attention had to be engaged in the task in order to reveal the amplification by pre-saccadic attention. Thus, it could be possible that feature-specific enhancements were not an automatic feature of pre-saccadic attention.
This was the first study to examine neural mechanisms of attention in the marmoset monkey, a small New World primate. While it is already established that the marmoset shares similar feed-forward circuits as macaques from retina to cortex for visual processing (Troilo et al., 1993; Solomon and Rosa, 2014; Mitchell and Leopold, 2015), as well as frontoparietal networks involved in eye movement control (Solomon and Rosa, 2014; Ghahremani et al., 2017; Johnston et al., 2018), little is yet known about the mechanisms of selective attention. In the macaque, almost everything we know about the neural mechanisms of attention originates from tasks that involve covert attention tasks with long delay periods of sustained central fixation prior to making a judgment or saccade to a peripheral target (Moran and Desimone, 1985; Treue and Maunsell, 1996; Seidemann and Newsome, 1999). Studies of pre-saccadic attention in macaques have also used delayed saccade tasks with sustained fixation periods before the saccade (Moore and Chang, 2009; Steinmetz and Moore, 2010). Here we found that a saccade foraging paradigm with minimal fixation delays was sufficient to examine neural modulation of pre-saccadic attention in the marmoset monkey. Marmosets completed 400–600 accurate saccade trials in daily sessions enabling us to map neural tuning curves across a range of 16 motion directions. Neurons in marmoset area MT showed neural modulation consistent with that of macaques using more demanding delayed saccade tasks (Moore and Chang, 2009; Steinmetz and Moore, 2010, 2014; Merrikhi et al., 2021). Neurons increased their mean firing rate and sensitivity to motion direction in the period immediately prior to saccade onset. This supported that the neural mechanisms of pre-saccadic attention were conserved from Old World to New World primates and that they generalized to more natural task conditions.
One difference from findings in previous studies of attention in macaques is that we found no significant change in firing variability. There was a modest reduction in the FF in both animals (4.8% and 0.6% respectively), but it did not reach significance. In one monkey where we had made array recordings, we also found a modest (−4%) reduction in noise correlations between pairs of neurons, but while significant this reduction was modest relative to the nearly 50% reductions reported in covert attention (Cohen and Maunsell, 2009; Mitchell et al., 2009). This difference might have reflected the emphasis of our experimental design to sample from complete tuning curves, wherein we collected fewer trial repetitions per motion direction. Previous studies have focused on getting larger trial repetitions using fewer stimuli, often only a preferred and/or nonpreferred stimulus for each neuron (Mitchell et al., 2007; Li and Basso, 2008; Moore and Chang, 2009; Steinmetz and Moore, 2010). However, these differences could also reflect differences in the behavioral tasks, which would be interesting to explore in future studies. Specifically, while previous paradigms measured variability during extended periods of sustained central fixation (Mitchell et al., 2007; Li and Basso, 2008; Cohen and Maunsell, 2009; Moore and Chang, 2009; Steinmetz and Moore, 2010; Merrikhi et al., 2021), we allowed marmosets to initiate saccades with no delay at stimulus onset. It is known that stimulus onsets can quench noise correlations (Churchland et al., 2010), which could explain the modest reductions in our task where saccades were initiated within 150–250 ms of stimulus onset. However, it is also worth pointing out that during natural saccades, foraging saccades will typically occur in rapid succession every 200–300 ms (Mitchell et al., 2014), so in many ways, the timing in our task may better reflect what is relevant during natural vision.
Our study not only establishes that pre-saccadic attention modulates neural responses in the extra-striate cortex of the marmoset monkey but also reveals a diversity in how neural tuning curves are modulated. We found no support for a net narrowing in neural tuning curves. However, across the population, individual neurons do support a diversity of changes in tuning, even with some showing a clear narrowing in tuning, and this leaves open a question about what types of sensitivity changes might be supported in the read-out of population activity and ultimately in behavior. Recent evidence from covert attention studies demonstrates that depending on the read-out, information can be reshaped in the read-out to enhance specific types of information, even while there may be no net improvement in the total population (Ruff and Cohen, 2019). Distinctions in how different parts of the population encode information are supported by laminar distinctions found in other studies of covert attention (Buffalo et al., 2011; Nandy et al., 2017; Pettine et al., 2019). A recent study of pre-saccadic attention found that stimulus orientation was better encoded by superficial layers neurons whereas movement direction was better encoded by deep layer neurons (Pettine et al., 2019). We found that subsets of neurons in the population do show significant narrowing in their tuning and that narrowing in tuning is correlated with increases in gain, as are reductions in baseline (Fig. 7). Thus, depending on how information is read-out, it remains possible that changes in tuning with pre-saccadic attention could support a feature selection of the target. However, at the level of individual units, pre-saccadic modulation appears consistent with changes reported in covert attention.
Footnotes
We thank Dina Graf and members of the Mitchell lab for help with marmoset care and handling. We thank Martin Rolfs and Lisa Kroell for their comments on an earlier draft of this manuscript. This work was supported by NIH Grants R01 EY030998 (J.F.M. and S.C.), R00 EY032179 (J.Y.), and T32 EY007125 (S.C.).
The authors declare no competing financial interests.
- Correspondence should be addressed to should be addressed to Jude F. Mitchell at jude.francis.mitchell{at}gmail.com.