Ongoing Projects
Measuring the Speed of Perception
Accurate measurement is a cornerstone of science. In our case, our research relies critically on time-resolved measurements of cognitive processing (sometimes labeled as "mental chronometry") that are more precise than those obtained with alternative techniques. In our urgent paradigms, the amount of time available to view the relevant cue stimulus (rPT) is limited and is determined for each individual trial. This way we can track the evolution of a choice process millisecond by millisecond, essentially. We can then measure, for instance, how perceptual processing speed differs across individual subjects, their motivation (say, when a large versus a small reward is at stake), their experience with the task, or the visual features being discriminated (say, color versus shape).
In general, we have found that the dynamics of the oculomotor system are much faster and richer at such short timescales than previously appreciated. For example, saccades are considered highly stereotyped in that, in looking repeatedly from point A to point B, the observed movements are nearly identical. Nearly — but not exactly. We found (Seideman et al., 2018) that when the subject looks at B but is guessing, the peak saccade velocity is slightly lower than average, whereas when the subject looks at B knowing that B is the correct choice, the peak velocity is slightly higher than average (see figure below). The time-resolved data unlock many previously hidden clues about the underlying neuronal dynamics.
The peak velocity of a saccade carries information about the decision-making process leading to it. (a) Percentage of correct choices as a function of raw processing time (rPT), or cue viewing time, in an urgent color discrimination task. The yellow shade separates informed choices (long rPT trials) from guesses (short rPT trials). (b, c) Mean eye velocity profiles for all guesses (b) and all informed choices (c) indicated in panel a. Note a slight difference between correct (green) and incorrect (red) traces for cue-guided trials but not for guesses. Such deflections are highly reliable, and relate in lawful ways to the target selection process that precedes each eye movement. See Seideman et al., 2018.
Stanford TR, Salinas E (2021) Urgent decision making: resolving visuomotor interactions at high temporal resolution. Annu Rev Vis Sci 7:323–348. [pdf] [doi]
Salinas E, Steinberg BR, Sussman LA, Fry SM, Hauser CK, Anderson DD, Stanford TR (2019) Voluntary and involuntary contributions to perceptually guided saccadic choices resolved with millisecond precision. eLife 8:e46359. [pdf] [doi]
Seideman J, Stanford TR, Salinas E (2018) Saccade metrics reflect decision-making dynamics during urgent choices. Nat Commun 9:2907. [pdf] [doi]
Shankar S, Massoglia DP, Zhu D, Costello MG, Stanford TR, Salinas E (2011) Tracking the temporal evolution of a perceptual judgment using a compelled-response task. J Neurosci 31:8406–8421. [pdf] [doi]
Stanford TR, Shankar S, Massoglia DP, Costello MG, Salinas E (2010) Perceptual decision making in less than 30 milliseconds. Nat Neurosci 13:379–385. [pdf] [doi]
How Distinct Neuronal Types Contribute to Saccadic Choices
Oculomotor areas contain three classically-defined neuronal types: visual (V), visuomotor (VM), and motor (M). Although their properties are well understood for simple oculomotor tasks, such as looking at a single, bright spot of light on a dark background, their roles in saccadic choices are less clear. For instance, we discovered that the responses of visual neurons depend critically on the salience of visual stimuli (see figure below), such that when the choice is between potential targets that are equally salient, visual neurons do not contribute to the target selection process at all. These cells can only help select objects that stand out. One aim of our research is to further characterize the functional properties of different oculomotor cell types, as well as of different oculomotor areas in the brain.
Sensitivity to salience in the three classic neuronal types found in the Frontal Eye Field, or FEF. (Top row) Traces represent neuronal activity as a function of time for populations of FEF neurons classified as visual (V), visuomotor (VM), and motor (M). The recorded activity is aligned on the presentation of a single visual target (left panels) and on the onset of a saccade to that target (right panels). In different trials the target fell either inside (red traces) or outside (green traces) of each cell's response field. The V neurons respond most strongly shortly after target presentation, the M neurons are activated just before saccade onset, and the VM neurons combine both bursts of activity. (Bottom row) Differential responses of the same neuronal populations in two choice tasks, the compelled saccade task (CS, pink traces), in which target and distracter are equally salient, and the compelled oddball task (CO, blue traces), in which the target stands out against a backdrop of identical distracters. The V response depends fundamentally on salience. See Scerra et al., 2019.
Scerra VE, Costello MG, Salinas E, Stanford TR (2019) All-or-none context dependence delineates limits of FEF visual target selection. Curr Biol 29:294–305. [pdf] [doi]
Costello MG, Zhu D, May PJ, Salinas E, Stanford TR (2016) Task dependence of decision- and choice-related activity in monkey oculomotor thalamus. J Neurophysiol 115:581–601. [pdf] [doi]
Costello MG, Zhu D, Salinas E, Stanford TR (2013) Perceptual modulation of motor — but not visual — responses in the frontal eye field during an urgent-decision task. J Neurosci 33:16394–16408. [pdf] [doi]
Computational Characterization of the Saccadic Choice Process
It is well established that a saccade is triggered when the activity of motor neurons in oculomotor areas builds up and crosses a critical, or threshold level; once the threshold is exceeded, the system commits to producing an eye movement within a few tens of milliseconds. Elaborating on this triggering mechanism, we have developed a computational model whereby a saccadic choice is conceived as a race between two (or more) populations of neurons, such that the first population to exceed the threshold determines the direction of the next saccade. This saccadic competition model reproduces in quantitative detail not only the psychophysical behavior of the subjects in our urgent tasks (see figure below), but also many key features of the oculomotor activity recorded during task performance. Many of our studies include a variant of this model. An important goal of our research program is to develop this computational framework further, to generate testable hypotheses and design new experiments.
Reaction time (RT) distributions from two monkeys that performed an urgent color discrimination task. Color shades correspond to experimental data from correct (cyan) and incorrect (pink) trials. Black traces are simulated data from our race-to-threshold model fitted to each monkey. Each row corresponds to a different value of the gap parameter (in ms, indicated on the top left corners), which controlled the difficulty of the task. Vertical lines divide the trials in each histogram approximately into guesses (to the left of the line) and informed choices (to the right). Toward the bottom most trials are guesses, whereas toward the top most trials are informed choices. The model reproduces in great detail the varied shapes of the RT distributions obtained in this urgent task — while, at the same time, mimicking the oculomotor selection dynamics recorded during the same task. See Salinas et al., 2010.
Salinas E, Steinberg BR, Sussman LA, Fry SM, Hauser CK, Anderson DD, Stanford TR (2019) Voluntary and involuntary contributions to perceptually guided saccadic choices resolved with millisecond precision. eLife 8:e46359. [pdf] [doi]
Hauser CK, Zhu D, Stanford TR, Salinas E (2018) Motor selection dynamics in FEF explain the reaction time variance of saccades to single targets. eLife 7:e33456. [pdf] [doi]
Shankar S, Massoglia DP, Zhu D, Costello MG, Stanford TR, Salinas E (2011) Tracking the temporal evolution of a perceptual judgment using a compelled-response task. J Neurosci 31:8406–8421. [pdf] [doi]
Salinas E, Shankar S, Costello MG, Zhu D, Stanford TR (2010) Waiting is the hardest part: comparison of two computational strategies for performing a compelled-response task. Front Comput Neurosci 4:153. [pdf] [doi]
Stanford TR, Shankar S, Massoglia DP, Costello MG, Salinas E (2010) Perceptual decision making in less than 30 milliseconds. Nat Neurosci 13:379–385. [pdf] [doi]
Relationship between Attention and Saccade Planning
Spatial attention refers to the capacity to preferentially process information from a specific region of space, filtering out information from other regions. Attention can be deployed either overtly, by moving the eyes, or covertly, while maintaining the eyes fixed. Although it is well established that covert spatial attention is controlled by the same brain circuits that generate saccades, its implementation is still uncertain. For instance, attention can be directed either voluntarily (at will) or involuntarily (reflexively), but what is the neuronal basis of this distinction? We suspect that so-called visual neurons in oculomotor areas (see above) are responsible for the involuntary, reflexive effects, and are actively investigating this.
Again, timing is critical. We recently designed an urgent version of the classic antisaccade task, in which a single bright stimulus appears and the participant is instructed to look away from it (Salinas et al., 2019; Goldstein et al., 2022). Under time pressure this is not easy, because there is a natural tendency to look at salient things that appear abruptly; they grab our attention automatically. We were able to characterize this reflexive tendency very precisely, and found that it exhibits a distinct dynamic: it is very strong but only during a brief period of time of about 40 ms, and thereafter it can be overcome by the willful intent to look away (see figure below). The results demonstrate an internal struggle that unfolds extremely rapidly.
A unique behavioral curve. In the urgent antisaccade task, participants are instructed to look away from a single, salient cue stimulus. Plots show the percentage of correct responses as a function of raw processing time (or cue viewing time). Traces are for individual participants (P1–P6) and their pooled data (All participants). When the cue is seen for less than 100 ms or so, performance is at chance (50% correct), as expected. Then, in trials with processing times of approximately 100–140 ms, performance approaches 0% correct. In this "capture" range (gray shaded area), it is almost impossible not to look at the cue and make an error. When the cue is viewed for 170 ms or more, performance is nearly perfect. A race-to-threshold model (black lines) reproduced the results in detail, pointing to specific neural mechanisms likely responsible for the potent oculomotor capture observed here. See Salinas et al., 2019.
Oor EE, Stanford TR, Salinas E (2023) Stimulus salience conflicts and colludes with endogenous goals during urgent choices. iScience 26:106253. [pdf] [doi]
Seideman JA, Stanford TR, Salinas E (2022) A conflict between spatial selection and evidence accumulation in area LIP. Nat Commun 13:4463. [pdf] [doi]
Goldstein AT, Stanford TR, Salinas E (2022) Exogenous capture accounts for fundamental differences between pro- and antisaccade performance. eLife 11:e76964. [pdf] [doi]
Salinas E, Steinberg BR, Sussman LA, Fry SM, Hauser CK, Anderson DD, Stanford TR (2019) Voluntary and involuntary contributions to perceptually guided saccadic choices resolved with millisecond precision. eLife 8:e46359. [pdf] [doi]
Scerra VE, Costello MG, Salinas E, Stanford TR (2019) All-or-none context dependence delineates limits of FEF visual target selection. Curr Biol 29:294–305. [pdf] [doi]
Individual Differences
Our behavioral techniques allow us to accurately measure the time course of fundamental perceptual processes. In a dark background, detecting that a light was turned on takes about 100 ms, finding a ripe tomato among many lemons takes about 130 ms, and distinguishing a coin from a scrabble tile about 160 ms. This is very fast; no wonder that most eye movements are preceded by a deliberation period of just 200–250 ms. Importantly, these numbers depend on the participant. For example, the time required to look away from a bright light may vary by 30 ms or more from one individual to another (see figure above). Because such idiosyncratic differences are highly reliable, we are exploring whether they can serve as measures of mental capacity more generally. Such ‘cognitive fingerprints’ could work as diagnostic markers of mental dysfunction for conditions such as ADHD, in which attention and perceptual processing are significantly altered.