To understand how neural activity gives rise to behavior, a powerful approach is to manipulate the activity of groups of neurons defined by particular functional or anatomical properties in the context of a suitable behavioral task. This strategy has grown in popularity over recent years with the advent of sophisticated tools for manipulating neural circuit function. However, recent perspectives have cautioned that so-called causal evidence is not always as decisive as it may appear (Jazayeri and Afraz, 2017), and by itself does not generate the level of understanding we wish to attain (Krakauer et al., 2017). These and other arguments serve to renew a longstanding dictum in systems neuroscience, namely the primary importance of developing a rigorous theoretical or conceptual framework for understanding the behavior of interest. Only then can clear hypotheses be articulated about the causal role of neural populations or circuits in generating the behavior.

The study of perceptual decision making has achieved a degree of progress toward this goal, in part by leveraging detailed knowledge of the neural representation of sensory evidence supporting the decision (reviewed by Cohen and Newsome, 2004; Romo and Salinas, 2001; Shadlen and Kiani, 2013). One well-studied behavioral task requires judgment of the net direction of motion in a noisy visual display designed to promote the integration of motion information across the display and over time. Neurons in motion-sensitive cortical areas, especially the middle temporal (MT) and medial superior temporal (MST) areas, are well suited to provide the evidence for the task, and theoretical considerations (Gold and Shadlen, 2001; Shadlen et al., 1996) point toward a simple and plausible computation: the subtraction of spike counts (or rates) between pools of neurons favoring each of the direction alternatives. This difference furnishes a quantity proportional to the log likelihood ratio favoring a given alternative, a statistical measure of the weight of evidence that can be accumulated over time to support optimal statistical decision making (Gold and Shadlen, 2002; Wald, 1947)

Building on the work of Newsome and colleagues (Celebrini and Newsome, 1995; Salzman et al., 1992), Ditterich et al. (2003) took advantage of the columnar organization of MT (Albright et al., 1984) to demonstrate, beyond simply a causal role for these neurons in the task, the specific differencing-and-integration mechanism that was previously hypothesized on theoretical grounds. Using electrical microstimulation (µStim) to activate neurons that were largely within a single direction column, they confirmed earlier reports showing that monkeys’ choices were biased toward the preferred direction of the activated neurons. More importantly, µStim increased the speed of decisions in favor of the preferred direction while slowing decisions made in favor of the opposite direction. The results supported a model in which the decision is formed by temporal integration of momentary evidence, defined as the difference in activity of pools of MT neurons tuned for the preferred and anti-preferred motion direction (Ditterich et al., 2003). More recently, we showed that decision confidence is affected by µStim in a way that is quantitatively commensurate with the effect on choices and well explained by a bounded evidence accumulation model (Fetsch et al., 2014a). Taken together, these findings suggest that a common process of evidence accumulation underlies all three behavioral measures of decision making in this task: choice, reaction time, and confidence.

Causal activation of neurons informs our understanding of the sufficiency, but not the necessity, of neural activity for driving behavior. Thus, reversible inactivation offers an important complement to stimulation, but conventional methods in primate neuroscience (e.g., muscimol or cooling) are lacking in both spatial and temporal resolution. The temporal component is crucial and often neglected. Indeed, the insights gained from the µStim studies described above depended on targeting the perturbation not only to the appropriate neurons but also the appropriate time frame within a trial (Seidemann et al., 1998). µStim itself has other drawbacks, including the possibility of antidromic activation, the generation of unwanted temporal and spatial patterns of inactivation (Butovas and Schwarz, 2003; Logothetis et al., 2010; Seidemann et al., 2002), and the difficulty of quantifying concurrent changes in neural activity due to voltage artifacts (but see Gnadt et al., 2003).

To address these limitations, we used an optogenetic approach to suppress MT activity during a motion discrimination task with post-decision wagering (PDW; Figure 1A). While some primate optogenetics studies set out to illuminate the largest possible volume of tissue (Acker et al., 2016; Gerits et al., 2012), we used very low light power levels to target relatively small clusters of excitatory neurons with similar selectivity for direction of motion. The rationale for this approach is that, in MT, such clusters constitute a critical functional unit for the conversion of a sensory representation into evidence for the decision. We found that optogenetic suppression was capable of inducing a choice bias against the neurons’ preferred direction, and a qualitatively similar change in the pattern of post-decision wagering, broadly consistent with a common mechanism underlying choice and confidence. We also found intriguing evidence for compensatory changes in the readout of MT activity, on both the sub-second and tens-of-minutes time scales. The results highlight the importance of spatial and temporal specificity in causal interventions, and they suggest a possible explanation for weak behavioral effects of optogenetic manipulations in some previous nonhuman primate studies.

Figure 1 with 1 supplement see all

Download asset Open asset

Figure 1—source data 1

Data and Matlab code for reproducing all panels and figure supplements for Figure 1.

https://doi.org/10.7554/eLife.36523.004

Download elife-36523-fig1-data1-v2.zip

We examined the effects of optogenetic inactivation (hereafter, photosuppression) in extrastriate visual cortex (area MT) on perceptual choices and decision confidence. Two monkeys were trained to perform a random-dot motion (RDM) direction discrimination task with post-decision wagering (PDW; Figure 1A). The task was to decide whether the net direction of dot motion was to the left or to the right, and to indicate the choice with a saccade to a leftward or rightward target when prompted. In addition to the choice targets, a ‘sure-bet’ target was presented during the delay period on a random half of trials, allowing the monkey to receive a guaranteed but smaller reward and thereby indicate its lack of confidence in the binary left-right decision. As in previous studies (Fetsch et al., 2014a; Kiani and Shadlen, 2009; Zylberberg et al., 2016), monkeys chose the sure-bet most frequently when the motion was weak (Figure 1B) and of short duration (Figure 1—figure supplement 1B), and their accuracy was greater when the sure-bet was available but waived, versus when it was unavailable (Figure 1C; p<10⁻¹⁰, logistic regression). These behavioral observations, and quantitative analyses published previously (Fetsch et al., 2014a; Kiani and Shadlen, 2009), serve to validate the PDW assay as a measure of confidence—that is, a prediction of accuracy based on the state of a decision variable on a given trial, rather than a low-level estimate of trial difficulty or an index of lapses of attention.

Histological and physiological characterization of Jaws expression

At least 8 weeks before commencing experiments, area MT in one hemisphere of each animal was injected with an AAV vector to drive expression of the red light-sensitive chloride pump Jaws (cruxhalorhodopsin; Chuong et al., 2014) under the control of the CaMKIIα promoter. The two animals continue to participate in experiments and are unavailable for histology. Therefore, to test the overall efficacy of the AAV-CaMKIIα-Jaws vector (to our knowledge, not previously used in the macaque), we performed immunohistochemical analysis in a third animal following injections in the lateral intraparietal area (LIP). This analysis revealed robust Jaws-GFP expression in superficial and deep layers (Figure 2A), and a tropism for excitatory pyramidal cells (Figure 2B), as shown in previous studies using the CaMKIIα promoter (Han et al., 2009; Nassi et al., 2015). This tropism was not as strong as in previous work: 9% of Jaws-GFP-positive neurons (83 of 903) were double-labeled for the inhibitory marker parvalbumin (Figure 2B), giving an upper bound on selectivity of 91%, compared to >98% in macaque primary visual cortex (2 of 119 cells double-labeled for any of three different inhibitory markers; Nassi et al., 2015), and 100% in a study of the frontal eye field (0 of 78 cells double-labeled for GABA; Han et al., 2009). However, in addition to being conducted in different cortical areas, these studies used lentiviral vectors rather than AAV vectors. Thus, the effectiveness of promoter-based targeting of excitatory neurons likely depends on the viral vector and/or serotype (Nathanson et al., 2009; Gerits et al., 2015), and could also vary across brain areas. We emphasize that any conclusions from Figure 2 in relation to the results presented below should be rendered with caution, considering possible differences in cell-type distributions and laminar organization of MT versus LIP.

Figure 2

Download asset Open asset

In each experimental session, we advanced a custom optrode into area MT and began searching for candidate sites for testing the effects of photosuppression on behavior. If a suitable site was found (see Materials and methods), the monkey commenced the task described above, with low-power red illumination (633 nm, total power = 0.25–2.0 mW, irradiance = 2–16 mW/mm² at a distance of 300 µm) delivered concurrently with visual stimulation (laser delayed by a 20 ms) on a random half of trials. Note that unlike previous studies, a red-shifted opsin was chosen not for its suitability for large-volume tissue illumination (Acker et al., 2016) but for its superior photocurrent at very low irradiance (Chuong et al., 2014). Optrode design, laser power, and site selection criteria were all intended to limit the light-induced suppression of activity to a relatively small cluster of MT neurons with consistent receptive field and tuning properties (i.e., preferred speed and direction of motion). We estimated tissue irradiance as a function of laser power and distance from the fiber tip using available online tools (https://web.stanford.edu/group/dlab/cgi-bin/graph/chart.php) as well as Monte Carlo simulations (Stujenske et al., 2015) (Figure 3—figure supplement 1). These estimates suggest that physiologically effective irradiance levels extended at most 300–500 µm from the fiber tip, which we confirmed qualitatively in pilot experiments using a fiber affixed to a linear electrode array (V-Probe, Plexon, Inc., Dallas, TX). After identifying a candidate site, we titrated the laser power to achieve an amount of suppression just shy of maximum (saturating) levels (Figure 3—figure supplement 2).

During the mapping procedure, we occasionally isolated single neurons and used this opportunity to better quantify the physiological effects of the light. Single-unit visual responses to high-coherence RDM stimuli—presented at the preferred direction and speed—were strongly suppressed on average (N = 26; Figure 3A), although a handful of neurons showed no effect or an increase in firing rate. Presumably this handful of cells includes those that did not express Jaws, as well as a small minority that were indirectly activated through polysynaptic mechanisms. Among putative Jaws-expressing neurons, the visually evoked response, after subtracting baseline activity, was suppressed by 101% (±13% SEM; N = 20; Figure 3B). Without baseline subtraction, the average firing rate during stimulus presentation was reduced by 72% ± 5%. In contrast, MU responses during the discrimination task were reduced by an average of 33% (±0.3% SEM, N = 20058 trials; Figure 3C). The fractional degree of suppression (Equation 4; Materials and methods), calculated from MU activity, varied substantially both within and across sessions (Figure 3A, inset; Figure 4A). We take this variability into consideration in the analyses to follow.

Figure 3 with 2 supplements see all

Download asset Open asset

Figure 3—source data 1

Data and Matlab code for reproducing all panels and figure supplements for Figure 3.

https://doi.org/10.7554/eLife.36523.009

Download elife-36523-fig3-data1-v2.zip

Figure 4 with 3 supplements see all

Download asset Open asset

Figure 4—source data 1

Data and Matlab code for reproducing all panels and figure supplements for Figure 4.

https://doi.org/10.7554/eLife.36523.014

Download elife-36523-fig4-data1-v2.zip

Note that the rebound of spiking after abrupt laser offset (Figure 3B), a known byproduct of photosuppression, was mitigated by introducing a 140 ms ramp-down of laser power in the majority of behavioral sessions (Figure 3D; Chuong et al., 2014). We did not detect a difference in behavioral effects when the rebound was suppressed, compared to an earlier subset of sessions without the ramp-down (Figure 3D, inset), and therefore pooled all sessions for subsequent analyses.

Temporal factors influence behavioral effects: compensation on multiple time scales?

The result shown in Figure 4B implies that photosuppression was only effective at altering behavior for a subset of trials. Indeed, most individual sessions failed to show statistically reliable effects on behavior when analyzed in their entirety (Figure 4—figure supplement 3). However, while collecting the data we noticed a tendency for systematic biases on laser trials to appear early within a session, only to dissipate over the course of tens of minutes (100–500 trials; ~5 s per trial on average, including intertrial intervals). This can be seen in the example session shown in Figure 5. Over the first 500 trials, the monkey tended to choose the preferred direction less often on laser trials vs. no-laser trials, effectively shifting the psychometric curve to the right (Figure 5A; shift = −4.6%, p=0.14, Equation 1). In contrast, there was essentially no effect on choices in the remaining ~1000 trials of the session (Figure 5B; shift = −0.7% coh, p=0.75). The pattern was similar for the confidence assay: a rightward shift of the confidence function on early trials (Figure 5C; shift = −4.2%, p=0.34) but not later trials (Figure 5D, 1.2%, p=0.80). This trend was evident when pooling across all sessions with enough trials to measure it (minimum 800 trials per session, N = 10 sessions): rightward shifts early in the session (choice shift = −3.5% coh±1.1%, p=0.001, Figure 5A'; confidence shift = −3.8% ± 1.9% [Equation 3], p=0.05, Figure 5C') but not later (choice shift = −0.7%, p=0.26, Figure 5B'; confidence shift = 0.1%, p=0.93; Figure 5D'). Post-hoc tests using generalized linear models (Equation 5, see Materials and methods) showed that the difference between early and late trials was statistically significant for choice (p=0.01) and confidence (p=0.02, Bonferroni corrected).

Figure 5

Download asset Open asset

Figure 5—source data 1

Data and Matlab code for reproducing all panels for Figure 5.

https://doi.org/10.7554/eLife.36523.016

Download elife-36523-fig5-data1-v2.zip

Salzman et al. (1992) reported that the effects of µStim dissipated over the course of the behavioral sessions, which they interpreted as indicating neuronal fatigue, damage, or shifts in the position of the stimulating electrode. In our case, however, the decrease in effectiveness of photosuppression on behavior was not accompanied by an attenuation of its effect on neural responses. In the example session (Figure 5E vs. Figure 5F), the degree of suppression was slightly greater in early versus late trials (∆R = −0.80 ± 0.01 and −0.74 ± 0.01, respectively; p<0.05, bootstrap confidence interval, resampling trials with replacement), but this quantitative difference seems unlikely to explain the complete nullification of the behavioral effects (Figure 5B,D). Across all 10 sessions in the pooled dataset (Figure 5E' vs. Figure 5F') the trend was in the opposite direction, toward greater suppression on late trials (∆R early = −0.28 ± 0.008, late = −0.40 ± 0.004, p<0.05). Furthermore, we did not observe a qualitative change in the direction selectivity of the affected neurons between early and late trials (Figure 5G vs. Figure 5H; and Figure 5G' vs. Figure 5H'). At the conclusion of a subset of sessions, we re-measured direction tuning at multiple depths and found little evidence for optrode drift, tissue damage or response fatigue. Instead, the results suggest a compensatory change in the downstream readout of MT neural signals to reduce or nullify the effects of photosuppression on a time scale of 10–40 minutes.

To examine this putative compensation at a finer grain, we plotted the choice and confidence effects calculated in a sliding window of trials sorted by trial number after pooling across sessions (Figure 6A; N = 8057 trials from 23 sessions; see Materials and methods). Interestingly, the effect on confidence (blue trace) appears to dissipate more quickly than the effect on choice, and ‘overshoots’ zero for a period of time roughly between trial numbers 500–800 (~40–65 min), before settling back to zero later in the session (Figure 6A). Note that positive values on the ordinate indicate leftward shifts of the function, as observed previously with µStim, and are thus opposite of the expected effect from inactivation. Although the trend was not statistically significant (p=0.14; Equation 3 applied to trial numbers 550–750), it raises the possibility of a dissociation between choice and confidence on this particular subset of trials (see Discussion). Lastly, for comparison with previous work, we performed the same analysis on the µStim data from monkey D (Fetsch et al., 2014a). Unlike photosuppression, and contrary to the findings of Salzman et al. (1992), the biases in choice and confidence induced by µStim remained stable throughout the behavioral session (Figure 6—figure supplement 1A). This result implies that the capacity of cortical circuits to compensate for artificially induced activity patterns may be greater for photosuppression compared to µStim, although we cannot ascertain whether the key difference is the sign (suppression versus stimulation) or the modality (optogenetic versus electrical) of the perturbation.

Figure 6 with 2 supplements see all

Download asset Open asset

Figure 6—source data 1

Data and Matlab code for reproducing all panels and figure supplements for Figure 6.

https://doi.org/10.7554/eLife.36523.020

Download elife-36523-fig6-data1-v2.zip

The presence of compensatory changes in readout across trials led us to wonder whether such changes could occur on a faster time scale as well. We addressed this question by grouping trials according to the experimenter-controlled stimulus (and laser) duration, which varied randomly across trials (range = 95–925 ms, truncated exponential distribution; Figure 1—figure supplement 1A). Remarkably, we found that the behavioral effects of photosuppression were largely restricted to trials with short and intermediate durations (duration <~350 ms; Figure 6B). Figure 7 depicts this trend pooled across all sessions, in the same format as Figure 5 and limited to the first 500 trials of each session. Short-duration trials exhibited light-induced shifts of −4.7% (p=0.0006, Figure 7A) and −5.7% (p=0.004, Figure 7C) for choice and confidence respectively, whereas long-duration trials showed no systematic biases in either measure (choice: −0.6%, p=0.50; confidence: 0.38%, p=0.84). The effect of duration was statistically significant (Equation 5, Materials and methods) for confidence (p=0.002, Bonferroni corrected) and marginally so for choice (p=0.04).

Figure 7

Download asset Open asset

Figure 7—source data 1

Data and Matlab code for reproducing all panels for Figure 7.

https://doi.org/10.7554/eLife.36523.022

Download elife-36523-fig7-data1-v2.zip

As with the slower form of compensation, this sub-second attenuation of behavioral effects was not readily explained by differences in MT neural responses on short- versus long-duration trials (Figure 7E–H). We did observe weaker direction selectivity during the latter half of the stimulus presentation on long-duration trials (Figure 7H). It is conceivable that these late epochs contributed less to behavior, under the assumption that neural signals contribute in proportion to their sensitivity or informativeness (Britten et al., 1996; Gu et al., 2007; Purushothaman and Bradley, 2005). However, the relevant comparison here is between short- and long-duration trials, not early versus late within a trial. Unsurprisingly, neuronal sensitivity was greater on long-duration trials when including all spiking activity throughout the stimulus epoch (neuronal threshold = 27.2 ± 1.6% coh versus 37.5 ± 2.5% for short durations; see Materials and methods). Thus, without additional assumptions, the weaker direction selectivity at time points >300 ms (less separation of the traces on the right side of Figure 7H versus 7G) cannot explain the inability of photosuppression to affect behavior on long-duration trials. Indeed, any candidate explanation for this result must contend with the fact that activity early in the stimulus epoch was suppressed equally on short- and long-duration trials (Figure 7E vs. 7F), yet only the former showed behavioral effects. It seems improbable that downstream circuits could treat such early activity differently, because at any given time during a trial there would be no way to predict whether the stimulus (and laser) would terminate in the next moment or continue. Below we discuss possible explanations and implications of this unexpected finding.

Matlab code (m-files) and data (.mat files) have been provided for Figure 1 and Figures 3-7.

1. 1. Acker L
   2. Pino EN
   3. Boyden ES
   4. Desimone R

   (2016) FEF inactivation with improved optogenetic methods

   PNAS 113:E7297–E7306.

   https://doi.org/10.1073/pnas.1610784113

   • PubMed
   • Google Scholar
2. 1. Afraz A
   2. Boyden ES
   3. DiCarlo JJ

   (2015) Optogenetic and pharmacological suppression of spatial clusters of face neurons reveal their causal role in face gender discrimination

   PNAS 112:6730–6735.

   https://doi.org/10.1073/pnas.1423328112

   • PubMed
   • Google Scholar
3. 1. Afraz SR
   2. Kiani R
   3. Esteky H

   (2006) Microstimulation of inferotemporal cortex influences face categorization

   Nature 442:692–695.

   https://doi.org/10.1038/nature04982

   • PubMed
   • Google Scholar
4. 1. Albright TD
   2. Desimone R
   3. Gross CG

   (1984) Columnar organization of directionally selective cells in visual area MT of the macaque

   Journal of Neurophysiology 51:16–31.

   https://doi.org/10.1152/jn.1984.51.1.16

   • PubMed
   • Google Scholar
5. 1. Averbeck BB

   (2015) Theory of choice in bandit, information sampling and foraging tasks

   PLoS Computational Biology 11:e1004164.

   https://doi.org/10.1371/journal.pcbi.1004164

   • PubMed
   • Google Scholar
6. 1. Bang D
   2. Fleming SM

   (2018) Distinct encoding of decision confidence in human medial prefrontal cortex

   PNAS 115:6082–6087.

   https://doi.org/10.1073/pnas.1800795115

   • PubMed
   • Google Scholar
7. 1. Britten KH
   2. Newsome WT
   3. Shadlen MN
   4. Celebrini S
   5. Movshon JA

   (1996) A relationship between behavioral choice and the visual responses of neurons in macaque MT

   Visual Neuroscience 13:87–100.

   https://doi.org/10.1017/S095252380000715X

   • PubMed
   • Google Scholar
8. 1. Britten KH
   2. Shadlen MN
   3. Newsome WT
   4. Movshon JA

   (1992) The analysis of visual motion: a comparison of neuronal and psychophysical performance

   The Journal of Neuroscience 12:4745–4765.

   https://doi.org/10.1523/JNEUROSCI.12-12-04745.1992

   • PubMed
   • Google Scholar
9. 1. Butovas S
   2. Schwarz C

   (2003) Spatiotemporal effects of microstimulation in rat neocortex: a parametric study using multielectrode recordings

   Journal of Neurophysiology 90:3024–3039.

   https://doi.org/10.1152/jn.00245.2003

   • PubMed
   • Google Scholar
10. 1. Carrera E
    2. Tononi G

    (2014) Diaschisis: past, present, future

    Brain 137:2408–2422.

    https://doi.org/10.1093/brain/awu101

    • PubMed
    • Google Scholar
11. 1. Celebrini S
    2. Newsome WT

    (1995) Microstimulation of extrastriate area MST influences performance on a direction discrimination task

    Journal of Neurophysiology 73:437–448.

    https://doi.org/10.1152/jn.1995.73.2.437

    • PubMed
    • Google Scholar
12. 1. Chandrasekaran C
    2. Peixoto D
    3. Newsome WT
    4. Shenoy KV

    (2017) Laminar differences in decision-related neural activity in dorsal premotor cortex

    Nature Communications 8:614.

    https://doi.org/10.1038/s41467-017-00715-0

    • PubMed
    • Google Scholar
13. 1. Chen TW
    2. Li N
    3. Daie K
    4. Svoboda K

    (2017) A map of anticipatory activity in mouse motor cortex

    Neuron 94:866–879.

    https://doi.org/10.1016/j.neuron.2017.05.005

    • PubMed
    • Google Scholar
14. 1. Chuong AS
    2. Miri ML
    3. Busskamp V
    4. Matthews GA
    5. Acker LC
    6. Sørensen AT
    7. Young A
    8. Klapoetke NC
    9. Henninger MA
    10. Kodandaramaiah SB
    11. Ogawa M
    12. Ramanlal SB
    13. Bandler RC
    14. Allen BD
    15. Forest CR
    16. Chow BY
    17. Han X
    18. Lin Y
    19. Tye KM
    20. Roska B
    21. Cardin JA
    22. Boyden ES

    (2014) Noninvasive optical inhibition with a red-shifted microbial rhodopsin

    Nature Neuroscience 17:1123–1129.

    https://doi.org/10.1038/nn.3752

    • PubMed
    • Google Scholar
15. 1. Churchland AK
    2. Huang X
    3. Lisberger SG

    (2007) Responses of neurons in the medial superior temporal visual area to apparent motion stimuli in macaque monkeys

    Journal of Neurophysiology 97:272–282.

    https://doi.org/10.1152/jn.00941.2005

    • PubMed
    • Google Scholar
16. 1. Cohen MR
    2. Newsome WT

    (2004) What electrical microstimulation has revealed about the neural basis of cognition

    Current Opinion in Neurobiology 14:169–177.

    https://doi.org/10.1016/j.conb.2004.03.016

    • PubMed
    • Google Scholar
17. 1. Dai J
    2. Brooks DI
    3. Sheinberg DL

    (2014) Optogenetic and electrical microstimulation systematically bias visuospatial choice in primates

    Current Biology 24:63–69.

    https://doi.org/10.1016/j.cub.2013.11.011

    • PubMed
    • Google Scholar
18. 1. de Lafuente V
    2. Jazayeri M
    3. Shadlen MN

    (2015) Representation of accumulating evidence for a decision in two parietal areas

    Journal of Neuroscience 35:4306–4318.

    https://doi.org/10.1523/JNEUROSCI.2451-14.2015

    • PubMed
    • Google Scholar
19. 1. DeAngelis GC
    2. Cumming BG
    3. Newsome WT

    (1998) Cortical area MT and the perception of stereoscopic depth

    Nature 394:677–680.

    https://doi.org/10.1038/29299

    • PubMed
    • Google Scholar
20. 1. Ditterich J
    2. Mazurek ME
    3. Shadlen MN

    (2003) Microstimulation of visual cortex affects the speed of perceptual decisions

    Nature Neuroscience 6:891–898.

    https://doi.org/10.1038/nn1094

    • PubMed
    • Google Scholar
21. 1. El-Shamayleh Y
    2. Kojima Y
    3. Soetedjo R
    4. Horwitz GD

    (2017) Selective optogenetic control of Purkinje cells in monkey cerebellum

    Neuron 95:51–62.

    https://doi.org/10.1016/j.neuron.2017.06.002

    • PubMed
    • Google Scholar
22. 1. Fetsch CR
    2. Kiani R
    3. Newsome WT
    4. Shadlen MN

    (2014a) Effects of cortical microstimulation on confidence in a perceptual decision

    Neuron 83:797–804.

    https://doi.org/10.1016/j.neuron.2014.07.011

    • PubMed
    • Google Scholar
23. 1. Fetsch CR
    2. Kiani R
    3. Shadlen MN

    (2014b) Predicting the accuracy of a decision: a neural mechanism of confidence

    Cold Spring Harbor Symposia on Quantitative Biology 79:185–197.

    https://doi.org/10.1101/sqb.2014.79.024893

    • PubMed
    • Google Scholar
24. 1. Fleming SM
    2. Maniscalco B
    3. Ko Y
    4. Amendi N
    5. Ro T
    6. Lau H

    (2015) Action-specific disruption of perceptual confidence

    Psychological Science 26:89–98.

    https://doi.org/10.1177/0956797614557697

    • PubMed
    • Google Scholar
25. 1. Gerits A
    2. Farivar R
    3. Rosen BR
    4. Wald LL
    5. Boyden ES
    6. Vanduffel W

    (2012) Optogenetically induced behavioral and functional network changes in primates

    Current Biology 22:1722–1726.

    https://doi.org/10.1016/j.cub.2012.07.023

    • PubMed
    • Google Scholar
26. 1. Gerits A
    2. Vancraeyenest P
    3. Vreysen S
    4. Laramée ME
    5. Michiels A
    6. Gijsbers R
    7. Van den Haute C
    8. Moons L
    9. Debyser Z
    10. Baekelandt V
    11. Arckens L
    12. Vanduffel W

    (2015) Serotype-dependent transduction efficiencies of recombinant adeno-associated viral vectors in monkey neocortex

    Neurophotonics 2:031209–031218.

    https://doi.org/10.1117/1.NPh.2.3.031209

    • PubMed
    • Google Scholar
27. 1. Gnadt JW
    2. Echols SD
    3. Yildirim A
    4. Zhang H
    5. Paul K

    (2003) Spectral cancellation of microstimulation artifact for simultaneous neural recording in situ

    IEEE Transactions on Biomedical Engineering 50:1129–1135.

    https://doi.org/10.1109/TBME.2003.816077

    • PubMed
    • Google Scholar
28. 1. Gold JI
    2. Shadlen MN

    (2001) Neural computations that underlie decisions about sensory stimuli

    Trends in Cognitive Sciences 5:10–16.

    https://doi.org/10.1016/S1364-6613(00)01567-9

    • PubMed
    • Google Scholar
29. 1. Gold JI
    2. Shadlen MN

    (2002) Banburismus and the brain: decoding the relationship between sensory stimuli, decisions, and reward

    Neuron 36:299–308.

    https://doi.org/10.1016/S0896-6273(02)00971-6

    • PubMed
    • Google Scholar
30. 1. Gu Y
    2. DeAngelis GC
    3. Angelaki DE

    (2007) A functional link between area MSTd and heading perception based on vestibular signals

    Nature Neuroscience 10:1038–1047.

    https://doi.org/10.1038/nn1935

    • PubMed
    • Google Scholar
31. 1. Han X
    2. Qian X
    3. Bernstein JG
    4. Zhou HH
    5. Franzesi GT
    6. Stern P
    7. Bronson RT
    8. Graybiel AM
    9. Desimone R
    10. Boyden ES

    (2009) Millisecond-timescale optical control of neural dynamics in the nonhuman primate brain

    Neuron 62:191–198.

    https://doi.org/10.1016/j.neuron.2009.03.011

    • PubMed
    • Google Scholar
32. 1. Hanks TD
    2. Kopec CD
    3. Brunton BW
    4. Duan CA
    5. Erlich JC
    6. Brody CD

    (2015) Distinct relationships of parietal and prefrontal cortices to evidence accumulation

    Nature 520:220–223.

    https://doi.org/10.1038/nature14066

    • PubMed
    • Google Scholar
33. 1. Jazayeri M
    2. Afraz A

    (2017) Navigating the neural space in search of the neural code

    Neuron 93:1003–1014.

    https://doi.org/10.1016/j.neuron.2017.02.019

    • PubMed
    • Google Scholar
34. 1. Kepecs A
    2. Mainen ZF

    (2012) A computational framework for the study of confidence in humans and animals

    Philosophical Transactions of the Royal Society B: Biological Sciences 367:1322–1337.

    https://doi.org/10.1098/rstb.2012.0037

    • Google Scholar
35. 1. Kiani R
    2. Corthell L
    3. Shadlen MN

    (2014) Choice certainty is informed by both evidence and decision time

    Neuron 84:1329–1342.

    https://doi.org/10.1016/j.neuron.2014.12.015

    • PubMed
    • Google Scholar
36. 1. Kiani R
    2. Hanks TD
    3. Shadlen MN

    (2008) Bounded integration in parietal cortex underlies decisions even when viewing duration is dictated by the environment

    Journal of Neuroscience 28:3017–3029.

    https://doi.org/10.1523/JNEUROSCI.4761-07.2008

    • PubMed
    • Google Scholar
37. 1. Kiani R
    2. Shadlen MN

    (2009) Representation of confidence associated with a decision by neurons in the parietal cortex

    Science 324:759–764.

    https://doi.org/10.1126/science.1169405

    • PubMed
    • Google Scholar
38. 1. Klein C
    2. Evrard HC
    3. Shapcott KA
    4. Haverkamp S
    5. Logothetis NK
    6. Schmid MC

    (2016) Cell-targeted optogenetics and electrical microstimulation reveal the primate koniocellular projection to supra-granular visual cortex

    Neuron 90:143–151.

    https://doi.org/10.1016/j.neuron.2016.02.036

    • PubMed
    • Google Scholar
39. 1. Komatsu H
    2. Wurtz RH

    (1988) Relation of cortical areas MT and MST to pursuit eye movements. I. Localization and visual properties of neurons

    Journal of Neurophysiology 60:580–603.

    https://doi.org/10.1152/jn.1988.60.2.580

    • PubMed
    • Google Scholar
40. 1. Krakauer JW
    2. Ghazanfar AA
    3. Gomez-Marin A
    4. MacIver MA
    5. Poeppel D

    (2017) Neuroscience needs behavior: correcting a reductionist bias

    Neuron 93:480–490.

    https://doi.org/10.1016/j.neuron.2016.12.041

    • PubMed
    • Google Scholar
41. 1. Lambelet P
    2. Sayah A
    3. Pfeffer M
    4. Philipona C
    5. Marquis-Weible F

    (1998) Chemically etched fiber tips for near-field optical microscopy: a process for smoother tips

    Applied Optics 37:7289–7292.

    https://doi.org/10.1364/AO.37.007289

    • PubMed
    • Google Scholar
42. 1. Lewis JW
    2. Van Essen DC

    (2000) Corticocortical connections of visual, sensorimotor, and multimodal processing areas in the parietal lobe of the macaque monkey

    The Journal of Comparative Neurology 428:112–137.

    https://doi.org/10.1002/1096-9861(20001204)428:1<112::AID-CNE8>3.0.CO;2-9

    • PubMed
    • Google Scholar
43. 1. Liu LD
    2. Pack CC

    (2017) The contribution of area MT to visual motion perception depends on training

    Neuron 95:436–446.

    https://doi.org/10.1016/j.neuron.2017.06.024

    • PubMed
    • Google Scholar
44. 1. Logothetis NK
    2. Augath M
    3. Murayama Y
    4. Rauch A
    5. Sultan F
    6. Goense J
    7. Oeltermann A
    8. Merkle H

    (2010) The effects of electrical microstimulation on cortical signal propagation

    Nature Neuroscience 13:1283–1291.

    https://doi.org/10.1038/nn.2631

    • PubMed
    • Google Scholar
45. 1. Meyniel F
    2. Sigman M
    3. Mainen ZF

    (2015) Confidence as bayesian probability: from neural origins to behavior

    Neuron 88:78–92.

    https://doi.org/10.1016/j.neuron.2015.09.039

    • PubMed
    • Google Scholar
46. 1. Middlebrooks PG
    2. Sommer MA

    (2012) Neuronal correlates of metacognition in primate frontal cortex

    Neuron 75:517–530.

    https://doi.org/10.1016/j.neuron.2012.05.028

    • PubMed
    • Google Scholar
47. 1. Morales J
    2. Lau H
    3. Fleming SM

    (2018) Domain-general and domain-specific patterns of activity supporting metacognition in human prefrontal cortex

    The Journal of Neuroscience 38:3534–3546.

    https://doi.org/10.1523/JNEUROSCI.2360-17.2018

    • PubMed
    • Google Scholar
48. 1. Nassi JJ
    2. Avery MC
    3. Cetin AH
    4. Roe AW
    5. Reynolds JH

    (2015) Optogenetic activation of normalization in alert macaque visual cortex

    Neuron 86:1504–1517.

    https://doi.org/10.1016/j.neuron.2015.05.040

    • PubMed
    • Google Scholar
49. 1. Nathanson JL
    2. Yanagawa Y
    3. Obata K
    4. Callaway EM

    (2009) Preferential labeling of inhibitory and excitatory cortical neurons by endogenous tropism of adeno-associated virus and Lentivirus vectors

    Neuroscience 161:441–450.

    https://doi.org/10.1016/j.neuroscience.2009.03.032

    • PubMed
    • Google Scholar
50. Conference

    1. Ohayon S

    (2017) Towards an all optical interrogation of deep neural circuits with ultra-thin optical fibers

    Neuroscience 2017.

    http://www.abstractsonline.com/pp8/#!/4376/presentation/282

    • Google Scholar
51. 1. Otchy TM
    2. Wolff SB
    3. Rhee JY
    4. Pehlevan C
    5. Kawai R
    6. Kempf A
    7. Gobes SM
    8. Ölveczky BP

    (2015) Acute off-target effects of neural circuit manipulations

    Nature 528:358–363.

    https://doi.org/10.1038/nature16442

    • PubMed
    • Google Scholar
52. 1. Pouget A
    2. Drugowitsch J
    3. Kepecs A

    (2016) Confidence and certainty: distinct probabilistic quantities for different goals

    Nature Neuroscience 19:366–374.

    https://doi.org/10.1038/nn.4240

    • PubMed
    • Google Scholar
53. 1. Purcell BA
    2. Kiani R

    (2016) Hierarchical decision processes that operate over distinct timescales underlie choice and changes in strategy

    PNAS 113:E4531–E4540.

    https://doi.org/10.1073/pnas.1524685113

    • PubMed
    • Google Scholar
54. 1. Purushothaman G
    2. Bradley DC

    (2005) Neural population code for fine perceptual decisions in area MT

    Nature Neuroscience 8:99–106.

    https://doi.org/10.1038/nn1373

    • PubMed
    • Google Scholar
55. 1. Quick RF

    (1974) A vector-magnitude model of contrast detection

    Kybernetik 16:65–67.

    https://doi.org/10.1007/BF00271628

    • PubMed
    • Google Scholar
56. 1. Raposo D
    2. Sheppard JP
    3. Schrater PR
    4. Churchland AK

    (2012) Multisensory decision-making in rats and humans

    Journal of Neuroscience 32:3726–3735.

    https://doi.org/10.1523/JNEUROSCI.4998-11.2012

    • PubMed
    • Google Scholar
57. 1. Romo R
    2. Hernández A
    3. Zainos A
    4. Brody CD
    5. Lemus L

    (2000) Sensing without touching: psychophysical performance based on cortical microstimulation

    Neuron 26:273–278.

    https://doi.org/10.1016/S0896-6273(00)81156-3

    • PubMed
    • Google Scholar
58. 1. Romo R
    2. Salinas E

    (2001) Touch and go: decision-making mechanisms in somatosensation

    Annual Review of Neuroscience 24:107–137.

    https://doi.org/10.1146/annurev.neuro.24.1.107

    • PubMed
    • Google Scholar
59. 1. Salzman CD
    2. Murasugi CM
    3. Britten KH
    4. Newsome WT

    (1992) Microstimulation in visual area MT: effects on direction discrimination performance

    The Journal of Neuroscience 12:2331–2355.

    https://doi.org/10.1523/JNEUROSCI.12-06-02331.1992

    • PubMed
    • Google Scholar
60. 1. Seidemann E
    2. Arieli A
    3. Grinvald A
    4. Slovin H

    (2002) Dynamics of depolarization and hyperpolarization in the frontal cortex and saccade goal

    Science 295:862–865.

    https://doi.org/10.1126/science.1066641

    • PubMed
    • Google Scholar
61. 1. Seidemann E
    2. Zohary E
    3. Newsome WT

    (1998) Temporal gating of neural signals during performance of a visual discrimination task

    Nature 394:72–75.

    https://doi.org/10.1038/27906

    • PubMed
    • Google Scholar
62. 1. Self MW
    2. van Kerkoerle T
    3. Supèr H
    4. Roelfsema PR

    (2013) Distinct roles of the cortical layers of area V1 in figure-ground segregation

    Current Biology 23:2121–2129.

    https://doi.org/10.1016/j.cub.2013.09.013

    • PubMed
    • Google Scholar
63. 1. Shadlen MN
    2. Britten KH
    3. Newsome WT
    4. Movshon JA

    (1996) A computational analysis of the relationship between neuronal and behavioral responses to visual motion

    The Journal of Neuroscience 16:1486–1510.

    https://doi.org/10.1523/JNEUROSCI.16-04-01486.1996

    • PubMed
    • Google Scholar
64. 1. Shadlen MN
    2. Kiani R

    (2013) Decision making as a window on cognition

    Neuron 80:791–806.

    https://doi.org/10.1016/j.neuron.2013.10.047

    • PubMed
    • Google Scholar
65. 1. Shadlen MN
    2. Shohamy D

    (2016) Decision making and sequential sampling from memory

    Neuron 90:927–939.

    https://doi.org/10.1016/j.neuron.2016.04.036

    • PubMed
    • Google Scholar
66. 1. Shekhar M
    2. Rahnev D

    (2018) Distinguishing the roles of dorsolateral and anterior PFC in visual metacognition

    The Journal of Neuroscience 38:5078–5087.

    https://doi.org/10.1523/JNEUROSCI.3484-17.2018

    • PubMed
    • Google Scholar
67. 1. So N
    2. Stuphorn V

    (2016) Supplementary eye field encodes confidence in decisions under risk

    Cerebral Cortex 26:764–782.

    https://doi.org/10.1093/cercor/bhv025

    • PubMed
    • Google Scholar
68. 1. Stauffer WR
    2. Lak A
    3. Yang A
    4. Borel M
    5. Paulsen O
    6. Boyden ES
    7. Schultz W

    (2016) Dopamine neuron-specific optogenetic stimulation in rhesus macaques

    Cell 166:1564–1571.

    https://doi.org/10.1016/j.cell.2016.08.024

    • PubMed
    • Google Scholar
69. 1. Stujenske JM
    2. Spellman T
    3. Gordon JA

    (2015) Modeling the spatiotemporal dynamics of light and heat propagation for in vivo optogenetics

    Cell Reports 12:525–534.

    https://doi.org/10.1016/j.celrep.2015.06.036

    • PubMed
    • Google Scholar
70. 1. Stöckle R
    2. Fokas C
    3. Deckert V
    4. Zenobi R
    5. Sick B
    6. Hecht B
    7. Wild UP

    (1999) High-quality near-field optical probes by tube etching

    Applied Physics Letters 75:160–162.

    https://doi.org/10.1063/1.124305

    • Google Scholar
71. 1. Tanaka K
    2. Hikosaka K
    3. Saito H
    4. Yukie M
    5. Fukada Y
    6. Iwai E

    (1986) Analysis of local and wide-field movements in the superior temporal visual areas of the macaque monkey

    The Journal of Neuroscience 6:134–144.

    https://doi.org/10.1523/JNEUROSCI.06-01-00134.1986

    • PubMed
    • Google Scholar
72. 1. van den Berg R
    2. Anandalingam K
    3. Zylberberg A
    4. Kiani R
    5. Shadlen MN
    6. Wolpert DM

    (2016a) A common mechanism underlies changes of mind about decisions and confidence

    eLife 5:e12192.

    https://doi.org/10.7554/eLife.12192

    • PubMed
    • Google Scholar
73. 1. van den Berg R
    2. Zylberberg A
    3. Kiani R
    4. Shadlen MN
    5. Wolpert DM

    (2016b) Confidence is the bridge between multi-stage decisions

    Current Biology 26:3157–3168.

    https://doi.org/10.1016/j.cub.2016.10.021

    • PubMed
    • Google Scholar
74. Book

    1. Wald A

    (1947)

    Sequential Analysis

    Mineola, NY: Dover Publications.

    • Google Scholar
75. 1. Yeung N
    2. Summerfield C

    (2012) Metacognition in human decision-making: confidence and error monitoring

    Philosophical Transactions of the Royal Society B: Biological Sciences 367:1310–1321.

    https://doi.org/10.1098/rstb.2011.0416

    • Google Scholar
76. 1. Yu AJ
    2. Dayan P

    (2005) Uncertainty, neuromodulation, and attention

    Neuron 46:681–692.

    https://doi.org/10.1016/j.neuron.2005.04.026

    • PubMed
    • Google Scholar
77. 1. Zylberberg A
    2. Fetsch CR
    3. Shadlen MN

    (2016) The influence of evidence volatility on choice, reaction time and confidence in a perceptual decision

    eLife 5:e17688.

    https://doi.org/10.7554/eLife.17688

    • PubMed
    • Google Scholar

Thank you for submitting your article "Focal optogenetic suppression in macaque area MT biases direction discrimination and decision confidence, but only transiently" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by David Van Essen as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Richard Born (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The reviewers agree that this is an outstanding study that makes use of a powerful visual discrimination task in which choice and confidence decision mechanisms that can be linked to the directionality of the preferred response of MT neurons. The reviewers agree that your suppression optogenetic approach in the preferred direction related MT neurons and its effects on choice and confidence constitutes an important advance to linking neural activity and behavior. Having said this, there are some comments by reviewers #2 and 3 and to a lesser extent by reviewer #1 that must be considered to improve/strengthen your paper:

1) It is not clear whether choice and confidence relate to the same decision variable, at least for the photo suppression results provided here (reviewer #2).

2) It would be useful to know whether your weak effects obtained on both choice and confidence were due to the low light power used; the reviewers wonder whether you used higher light power in other sessions (reviewers #2 and 3). This could show that the effects are more comparable to those reported before with your electrical micro-stimulation method. This might also shed light on the spatial extent of suppression by light and its power dependence (reviewers #2 and 3).

3) Please comment on whether you tried up-down discriminations while inactivating right-left preferred columns as a control (reviewers #2 and 3).

4) The short and long temporal suppressive effects on decisions were pooled from sessions in the two monkeys; the reviewers wonder about the behavioral performances of each animal.

Reviewer #1:

This study is technically well executed. This research group previously proved that electrical micro-stimulation of MT increased the speed of decisions in favor of the preferred direction while slowing decisions in favor of the opposite directions. However and according to the authors, there are technical limitations using the micro-stimulation technique, and therefore the need of some other ways of perturbing clusters of neurons associated with behavior. Here, they used an optogenetic approach in which they suppressed small clusters of excitatory neurons with selectivity for visual motion. They found that optogenetic suppression of small clusters of MT neurons induced a weak choice bias against the neurons' preferred direction, consistent with a common mechanism underlying choice and confidence. These weak effects on choice and confidence were supposedly due to compensatory changes in the read out of MT activity. A more intriguing result was that these weak changes were present only during the first third of the perturbing session (first 500 trials). This was consistently observed across many sessions. Importantly, in all the recorded neuronal sessions, single and multi-unit activity decreased during the optogenetic test, showing the reliability of this suppression approach. As said above, the paper is technically well executed and the experiment is well designed. Of course, one major concern relies on the suppression effects of MT activity and therefore on the behavioral results. The experiment is fair since it shows the opposite of what a micro-stimulation effect was in this task. I understand the authors for using new experimental tools, and appreciate the finesse of the statistical analyses and the explanation they gave on the temporal suppression effects on behavior. I should say that this paper is one of the finest ones that I have seen using the optogenetic tool in behavior. But, honestly, I did not learn much of this study.

Reviewer #2:

The transfer of optogenetic (and other virally or genetically targeted) methods to the nonhuman primate is a big deal, and the authors have made several important contributions to this endeavor. First, and most importantly, they tested the new method in the context of a behavioral task that is extremely well characterized at the anatomical, physiological, behavioral and theoretical levels. Moreover, they used a slight variation of this task ("post-decision wagering") that buys them better interpretability of any effects produced. The effects they observed, while relatively modest in size by microstim criteria, are convincing and highly intriguing. They go well beyond demonstrating the efficacy of the method to reveal new and interesting results, particularly with respect to the apparent downstream compensation for inactivation at both short (within trial) and longer (across trials) time scales. A critical advance over the microstim experiments is that, here, the authors have a trial-by-trial measure of the efficacy of their inactivation. They also reveal what may be subtle differences (subject to limitations of the data, to which the authors make appropriate emphasis – subsection “Compensation on slow time scales”, last paragraph) in circuit mechanisms for choice vs. confidence. Overall, I think this is an outstanding paper that makes major methodological and conceptual contributions to our understanding of sensory decision making.

The biggest negative, and it is a relatively minor one, was the uncertainty in the spatial extent of the inactivation. In the main text, the authors supply us with only a rather vague description: "a relatively small cluster." They actually do much better in the second paragraph of the Materials and methods subsection “Optrode design, site selection and photosuppression protocol”. I suggest that some version of these lines – especially the point about confirming the lack of effect at 1 mm distances-be moved to (or repeated in) the main text. Obviously, the directional selectivity of the effects is also evidence for columnar specificity, but the authors only ever assessed the one (i.e. right-left) axis. It would be interesting to know if they ever tried the orthogonal axis (i.e. up-down when inactivating right-preferring neurons) as a negative control, or did some of the other specificity checks that have been previously performed for microstimulation: e.g. finding a lack of effect when moving the discriminand away from the recorded/inactivated receptive fields; increasing laser power to inactivate multiple direction columns to produce a flattening of the psychometric function. These are all obvious directions for future experiments and are absolutely not required for this study to have a large impact.

One other general concern: it appears that data from the two monkeys was pooled for all analyses. This is understandable. But I was not able to find a statement declaring a justification for this. Were the main results consistent between the two monkeys (except for the one-monkey peculiarity mentioned in the legend to Figure 5)?

Reviewer #3:

This paper uses optogenetic suppression of cortical activity to investigate how visual motion signals in area MT are read out during perceptual judgments. The experimental design is based on that used previously with electrical microstimulation – monkeys viewed a random dot visual motion display of variable coherence and reported their left/right judgment by making a saccade to the appropriate target and were also sometimes given a third opt-out option that gave a guaranteed smaller reward. The opt-out option provides a measure of confidence (higher opt-out rate implies lower decision confidence) and previous microstimulation results suggested that a common mechanism contributed to both confidence and choice. The major difference in this study is the use of focal optogenetic suppression that was orchestrated so that the causal effects were mostly limited to a population of MT neurons with similar directional tuning, which was confirmed with electrophysiological recordings. The ability to reversibly suppress neuronal activity with this spatial and temporal precision is a substantial and elegant advance over previous techniques.

The effects reported in the Results are subtle but clear – both choices and confidence are systematically altered by the optogenetic suppression. But unlike the case with electrical microstimulation, the causal effects are subject to some form of compensatory change, such that the effects on choice behavior are present only for the first 100 or so trials within each session and only for visual motion stimuli that are presented for shorter durations. The reasons for these compensatory changes are not known and lie beyond the scope of this paper, although the authors outline some reasonable explanations to be tested elsewhere. The main strength of the paper lies in its impressive demonstration of methodology for applying optogenetics to pose questions that cannot be addressed with older traditional techniques (e.g., stimulation, drug injection) that lack the necessary temporal and spatial precision. The results also point to some interesting distinctions between choice and confidence that are unexpected, and it seems that the authors themselves are somewhat unsure whether to fully embrace these effects, since they run counter to their previous interpretations.

Overall, this is a very well-written paper that presents novel results on a very interesting topic. My comments are mainly aimed at clarifying the results and interpretation.

The data in the paper provide a basis for possibly drawing novel distinctions between the mechanisms of choice versus confidence. The authors acknowledge that choice and confidence are not always coupled in their results, but nonetheless conclude that their data are consistent with a common mechanism for both (Abstract). As they point out, their analysis in Figure 6 shows that confidence and choice have different effects as a function of trial number. They also report that the shifts in the choice and confidence functions were not correlated, although this negative result may be hard to interpret. More to the point, when I look at the confidence curves in Figures 4D, 5A, etc., I see a very different pattern than the obvious shift that is observed during electrical stimulation (e.g., Figure 4—figure supplement 1B). Rather than a shift, it looks like optogenetic suppression results in a 1-sided (uni-directional) increase in confidence (i.e., a drop in opt-out choices for low signal strengths when the opposite motion direction is suppressed). This makes me wonder about the assumption that confidence is based on the same decision variable as the choice – that is, the difference between rightward and leftward motion signals. Is it possible that confidence is influenced by the total amount of rightward and leftward motion, and not just the magnitude of the difference? To be clear, I don't expect this issue can be settled with this data set. But there does seem to be an incongruity between the data presented and the conclusion about a 'common mechanism for choice and confidence' that should be resolved.

As part of the experimental design, very low light powers were used in order to target local clusters of MT neurons with similar selectivity. This explains why the effect sizes are modest and may not reach significance on individual sessions. For comparison with other studies, have you tried using higher light powers and illuminating larger volumes of tissue? I would expect to see stronger disruptions of performance, but for all directions of visual motion. Nonetheless, the disruptions would be specific for visual motion stimuli placed at the appropriate retinotopic locations. This might serve as a useful positive control and point of comparison with other studies.

The histology on the Jaws expression is an important aspect of the results. An obvious concern is that the histology was done in area LIP not area MT, and it is not obvious to me whether the expression patterns in area MT would be expected to closely follow those in LIP or whether they might end up being very different. Are there neuronal cell type similarities (or differences) between MT and LIP that should be pointed out when interpreting these results?

Unless I missed it, the paper presents the effects of photosuppression only for the trials that included the opt-out option. What effects do you find if the opt-out is not available? Does this substantially change the size of the effect?

The reviewers agree that this is an outstanding study that makes use of a powerful visual discrimination task in which choice and confidence decision mechanisms that can be linked to the directionality of the preferred response of MT neurons. The reviewers agree that your suppression optogenetic approach in the preferred direction related MT neurons and its effects on choice and confidence constitutes an important advance to linking neural activity and behavior. Having said this, there are some comments by reviewers #2 and 3 and to a lesser extent by reviewer #1 that must be considered to improve/strengthen your paper:

1) It is not clear whether choice and confidence relate to the same decision variable, at least for the photo suppression results provided here (reviewer #2).

We agree this question was left somewhat unresolved, and we have now attempted to address it more fully through modeling and changes to the text. These additions clarify what we can and cannot say about this issue with the current dataset. The upshot is that we have maintained, although softened, our conclusion that the results support a common decision variable for choice and confidence. For details please see response to reviewer 3 and Author response image 4.

2) It would be useful to know whether your weak effects obtained on both choice and confidence were due to the low light power used; the reviewers wonder whether you used higher light power in other sessions (reviewers #2 and 3). This could show that the effects are more comparable to those reported before with your electrical micro-stimulation method. This might also shed light on the spatial extent of suppression by light and its power dependence (reviewers #2 and 3).

We suspect there is a fundamental limit to the size of direction-specific biasing effects that can be caused by photosuppression, for three main reasons: (i) a nontrivial proportion of neurons in the injected region fail to express the opsin, a limitation of current viral vector technology; (ii) increasing the laser power much beyond the levels we used would suppress nearby columns with different direction preferences, thus diluting the direction specificity of the effect; and (iii) the magnitude of the change in firing rate with suppression is inherently limited by floor effects, more so than microstimulation is limited by ceiling effects. We did not have the opportunity to examine the question empirically, e.g., by systematically varying light power for a given behavioral session. We did modify the text (moving some from Materials and methods to Results) and added two figure supplements to emphasize pilot experiments addressing the spatial extent and power dependence of photosuppression, based on a suggestion from reviewer 2. See Author response image 2 and new figure supplements associated with main text Figure 3.

3) Please comment on whether you tried up-down discriminations while inactivating right-left preferred columns as a control (reviewers #2 and 3).

We thank the reviewers for this idea and we hope to pursue such experiments in the future. Here we were unable to test up-down discriminations, owing to a limitation in the behavioral capacities of the particular animals used in this study. This limitation is now mentioned in Materials and methods (subsection “Behavioral task”, first paragraph); see also response to reviewer 2 below.

4) The short and long temporal suppressive effects on decisions were pooled from sessions in the two monkeys; the reviewers wonder about the behavioral performances of each animal.

The effects of suppression on early/short trials (Figure 5A’+C’, Figure 7A+C) were highly significant in each animal individually, and the trends depicted in Figure 6 were qualitatively similar in both monkeys (Author response image 3), although with some minor differences. See response to reviewer 2 below for details.

Lastly, we corrected an error in the handling of data from one of the 23 inactivation sessions. This session showed clear behavioral effects on choice and confidence, but in the opposite direction than predicted by the recorded multiunit activity (see data point in the upper-right quadrant of Figure 4—figure supplement 3, now highlighted in red). Such ‘paradoxical’ effects occasionally happen with electrical µStim as well (e.g. Salzman et al., 1992; DeAngelis et al., 1998; Fetsch et al., 2014) and are believed to imply that a population of oppositely tuned neurons – nearby but not being picked up by the recording electrode – were affected by the manipulation and contributed more strongly to behavior than the neurons being recorded (Salzman et al., 1992).

Interestingly, the effects in this individual session were sensitive to trial number (Author response image 1A-D), duration (E-H), and degree of suppression (not shown), consistent with the patterns described for the full dataset. As the study’s emphasis shifted to the surprising compensatory effects, we decided to include this paradoxical site in these analyses. To do so required reversing the assignment of preferred and null directions, such that the sign of the behavioral shifts was negative (rightward); otherwise the compensation would have canceled out similar trends in other ‘non-paradoxical’ sessions. However, we mistakenly included the reversed session in the analyses of Figure 4, which has nothing to do with compensation. We have now remedied this by explaining the above reasoning in the Materials and methods (subsection “Data analysis”, last paragraph) and replacing Figure 4 with one that does not include the paradoxical session. We have also added a figure supplement (Figure 6—figure supplement 2) showing that the main results hold when the paradoxical session is omitted.

Author response image 1

Download asset Open asset

Reviewer #1:

This study is technically well executed. This research group previously proved that electrical micro-stimulation of MT increased the speed of decisions in favor of the preferred direction while slowing decisions in favor of the opposite directions. However and according to the authors, there are technical limitations using the micro-stimulation technique, and therefore the need of some other ways of perturbing clusters of neurons associated with behavior. Here, they used an optogenetic approach in which they suppressed small clusters of excitatory neurons with selectivity for visual motion. They found that optogenetic suppression of small clusters of MT neurons induced a weak choice bias against the neurons' preferred direction, consistent with a common mechanism underlying choice and confidence. These weak effects on choice and confidence were supposedly due to compensatory changes in the read out of MT activity. A more intriguing result was that these weak changes were present only during the first third of the perturbing session (first 500 trials). This was consistently observed across many sessions. Importantly, in all the recorded neuronal sessions, single and multi-unit activity decreased during the optogenetic test, showing the reliability of this suppression approach. As said above, the paper is technically well executed and the experiment is well designed. Of course, one major concern relies on the suppression effects of MT activity and therefore on the behavioral results. The experiment is fair since it shows the opposite of what a micro-stimulation effect was in this task. I understand the authors for using new experimental tools, and appreciate the finesse of the statistical analyses and the explanation they gave on the temporal suppression effects on behavior. I should say that this paper is one of the finest ones that I have seen using the optogenetic tool in behavior. But, honestly, I did not learn much of this study.

We thank the reviewer for this concise summary of our findings and the positive assessment overall. We do not understand the major concern mentioned in the sentence beginning “Of course, one major concern relies on…” We assume the final remark reflects the reviewer’s acceptance of the conclusions drawn from previous studies, especially our electrical µStim study (Fetsch et al., 2014). Nevertheless, we feel it is not a foregone conclusion that inactivating a population of neurons would have clear, interpretable effects in the opposite direction as activating them. Or perhaps the reviewer feels this study does not represent a major technical advance. We would agree, in the sense that optogenetics in nonhuman primates still has yet to reach its full potential, but hope the reviewer would concede that this is nevertheless an important step.

Reviewer #2:

[…] The biggest negative, and it is a relatively minor one, was the uncertainty in the spatial extent of the inactivation. In the main text, the authors supply us with only a rather vague description: "a relatively small cluster." They actually do much better in the second paragraph of the Materials and methods subsection “Optrode design, site selection and photosuppression protocol”. I suggest that some version of these lines – especially the point about confirming the lack of effect at 1 mm distances-be moved to (or repeated in) the main text.

We agree that the issue of spatial extent was not given sufficient emphasis or clarity in the main text, and have now addressed this in several ways. First, as suggested, we reexamined the pilot sessions with 1 mm separation, but found that we had not tested high enough power levels, leaving it unclear whether the lack of an effect was because the light did not reach the neurons near the electrode tip or whether those neurons had simply failed to express the opsin. We do have some relevant data from the handful of V-probe + fiber sessions alluded to in the Materials and methods, now shown in Author response image 2, but we do not consider this direct estimate of light spread to be reliable enough (N=3 sessions) to warrant inclusion in the main text. We also provide a figure showing multiunit activity during a subset of mapping (fixation) blocks in which light power was systematically varied prior to beginning the behavioral task. This has been added to manuscript as Figure 3—figure supplement 2. Lastly, we performed new Monte Carlo simulations of light absorption/scattering in brain tissue (Figure 3—figure supplement 1), based on code made available by Stujenske et al. 2015. Each of these analyses provides support for our claim that photosuppression was typically limited to a region 300-500 µm across.

Author response image 2

Download asset Open asset

Obviously, the directional selectivity of the effects is also evidence for columnar specificity, but the authors only ever assessed the one (i.e. right-left) axis. It would be interesting to know if they ever tried the orthogonal axis (i.e. up-down when inactivating right-preferring neurons) as a negative control, or did some of the other specificity checks that have been previously performed for microstimulation: e.g. finding a lack of effect when moving the discriminand away from the recorded/inactivated receptive fields; increasing laser power to inactivate multiple direction columns to produce a flattening of the psychometric function. These are all obvious directions for future experiments and are absolutely not required for this study to have a large impact.

These are all excellent suggestions and on the docket for future experiments. We do know that microstimulation of columns with preferred directions orthogonal to the discriminanda can affect choices in a systematic way (Salzman and Newsome, 1994; Groh et al., 1997; Nichols and Newsome 2002). These effects have been interpreted as favoring either a winner-take-all or vector-averaging readout mechanism, depending on the angular separation and the specifics of the task. Unfortunately we were unable to perform this type of experiment here because the particular monkeys we used were not trained to perform up-down discriminations, and in fact were unusually resistant to such training. We have clarified the text in Materials and methods that mentions this fact (subsection “Behavioral task”, first paragraph). We also did not try placing the dots stimulus outside of the receptive field. We know this eliminates the effect of microstimulation on choices (Salzman et al., 1992), but it would be interesting to look for effects on confidence. Regarding a high-power manipulation to flatten the psychometric function, please see response to reviewer 3 below.

One other general concern: it appears that data from the two monkeys was pooled for all analyses. This is understandable. But I was not able to find a statement declaring a justification for this. Were the main results consistent between the two monkeys (except for the one-monkey peculiarity mentioned in the legend to Figure 5)?

Yes, the main results were consistent between the two monkeys: the effects of suppression on early/short trials (Figure 5A’+C’, Figure 7A+C) were significant in each animal individually, and the time course of attenuation of these effects across trials and time within a trial were qualitatively similar, as shown in Author response image 3. All four trends were statistically significant in monkey D (blue and red traces in Author response image 3A-B), but not monkey N (Author response image 3C-D). In panel D the trends are there, but weaker (p = 0.14 for both). In panel C they are nonsignificant because the data exhibited non-monotonicity (e.g., blue trace for trial numbers ~950-1400; see inset), whereas the statistical approach we used assumes a monotonic relationship. We do not have an explanation for the transient ‘rebound’ of the confidence effect on late trials in monkey N (inset), but it is worth noting that only a minority of sessions (~7) lasted long enough to contribute to these late data points. Overall, we felt the results were similar enough to justify pooling the two monkeys, especially given the constraints on sample size owing to technical challenges involved in acquiring the data. But we agree this should have been mentioned in the text, and now do so (Materials and methods, subsection “Data analysis”, fourth paragraph).

Author response image 3

Download asset Open asset

Reviewer #3:

[…] The effects reported in the Results are subtle but clear – both choices and confidence are systematically altered by the optogenetic suppression. But unlike the case with electrical microstimulation, the causal effects are subject to some form of compensatory change, such that the effects on choice behavior are present only for the first 100 or so trials within each session and only for visual motion stimuli that are presented for shorter durations. The reasons for these compensatory changes are not known and lie beyond the scope of this paper, although the authors outline some reasonable explanations to be tested elsewhere. The main strength of the paper lies in its impressive demonstration of methodology for applying optogenetics to pose questions that cannot be addressed with older traditional techniques (e.g., stimulation, drug injection) that lack the necessary temporal and spatial precision. The results also point to some interesting distinctions between choice and confidence that are unexpected, and it seems that the authors themselves are somewhat unsure whether to fully embrace these effects, since they run counter to their previous interpretations.

Overall, this is a very well-written paper that presents novel results on a very interesting topic. My comments are mainly aimed at clarifying the results and interpretation.

We are grateful for these remarks, and agree there was some ambiguity in the interpretation relating to whether a common mechanism underlies choice and confidence. We have addressed this by fitting the data to a bounded evidence accumulation model (see below) and by improving the clarity with which we address this issue in the text.

The data in the paper provide a basis for possibly drawing novel distinctions between the mechanisms of choice versus confidence. The authors acknowledge that choice and confidence are not always coupled in their results, but nonetheless conclude that their data are consistent with a common mechanism for both (Abstract). As they point out, their analysis in Figure 6 shows that confidence and choice have different effects as a function of trial number. They also report that the shifts in the choice and confidence functions were not correlated, although this negative result may be hard to interpret. More to the point, when I look at the confidence curves in Figures 4D, 5A, etc., I see a very different pattern than the obvious shift that is observed during electrical stimulation (e.g., Figure 4—figure supplement 1B). Rather than a shift, it looks like optogenetic suppression results in a 1-sided (uni-directional) increase in confidence (i.e., a drop in opt-out choices for low signal strengths when the opposite motion direction is suppressed). This makes me wonder about the assumption that confidence is based on the same decision variable as the choice – that is, the difference between rightward and leftward motion signals. Is it possible that confidence is influenced by the total amount of rightward and leftward motion, and not just the magnitude of the difference? To be clear, I don't expect this issue can be settled with this data set. But there does seem to be an incongruity between the data presented and the conclusion about a 'common mechanism for choice and confidence' that should be resolved.

We thank the reviewer for these insightful comments. We hope the discussion below and the changes to the manuscript help resolve the incongruity between our conclusion of a common mechanism and the apparent discrepancies that the reviewer perceptively points out.

Regarding the asymmetry in the confidence function, we agree that the data seem to indicate a greater effect on confidence when the motion was in the null direction (negative coherence) compared to the preferred direction (positive coherence). We tried a couple different ways to quantify this observation (The first method was to calculate the odds ratio (OR) for the contingency table made up of the occurrence (or not) of a sure-bet choice and the presence/absence of the laser. We did this separately for each side of the confidence function (coherence > 3.2% versus < -3.2%) and computed 95% confidence intervals for the log-OR (method of Woolf). The 95% CI for positive coh was highly overlapping the abs(CI) for negative coh, meaning the asymmetry was not statistically significant. The second method was to use the width parameter of the fitted Gaussian, using a modification of Equation 3 with an indicator term applied to σ. A shift plus a decrease in width on laser trials would generate the asymmetry, but the width change was also not statistically significant.), and for both approaches it did not reach statistical significance. However, even if the asymmetry were significant, whether it entails a distinct decision variable (or mechanism) for confidence is unclear. First, we confess an inability to connect the dots between the observation and the interesting proposal the reviewer makes about confidence being influenced by the total amount of rightward and leftward motion. Trials with greater or lesser total motion could arise by chance, but were equally likely to occur for positive and negative coherence, and thus cannot straightforwardly explain the asymmetry (we did not systematically vary the total motion energy within a given coherence level). Still, we do not reject the possibility of such a mechanism, and it would be interesting to test other types of stimuli that independently vary the strength of evidence supporting each option (e.g., Zylberberg, Barttfeld, and Sigman, 2012; Odegaard et al., PNAS 2018).

It should be noted that the asymmetry could arise under a common mechanism, if photosuppression were to cause both a shift and a decrease in the width of the bell-shaped confidence function. In a bounded accumulation model (e.g., Fetsch et al., 2014), narrowing of the confidence curve is captured by an increase in the sensitivity parameter k, which converts the (experimenter-controlled) motion strength to the drift rate of the decision variable. Naturally, increasing k also entails an increase in the sensitivity, or slope, of the choice function, and indeed there is a hint of this in the data, especially in Figure 7A (although it too does not reach significance; p = 0.10, logistic regression).

To illustrate this point, we fit the model to the data shown in main text Figures 5A’+C’ and 7A+C. The fits are shown below in Author response image 4, showing that the model can account for the asymmetry in the confidence effect despite having a single decision variable that governs choice and confidence. Our sample size does not allow for rigorous model comparison, so we cannot make strong conclusions about whether and how photosuppression alters properties of sensory signal and noise. But the fitting exercise suggests that the observation of asymmetry is not impetus for rejecting the hypothesis of a common mechanism. We have now included this point in the Discussion (subsection “A common mechanism for choice and confidence”, third paragraph).

Author response image 4

Download asset Open asset

Regarding the discrepancies in time course (Figure 6), the current dataset cannot rule out that they occurred by chance. Using the standard errors of the shift parameters (β2 in Equation 1, and δ in Equation 3) – and conceding that the lateral shift is only one of several aspects of the data one could examine – none of the differences between the shifts in choice and confidence rise to the level of statistical significance (see reply to comment below about the ‘overshoot’).

Nevertheless, in light of the reviewer’s concern (which we share), we have walked back some of the emphasis in the manuscript on a common mechanism for choice and confidence. For instance, we removed a sentence relating to Figure 4B: "Note that the effects (or lack thereof) on choice and confidence were reasonably well matched across the full range of ∆R (Figure 4B), supporting the idea of a common decision mechanism linking the two measures," added phrases like “broadly consistent”, and rewrote a large part of the Discussion (subsection “A common mechanism for choice and confidence”) to include several of the considerations described above.

As part of the experimental design, very low light powers were used in order to target local clusters of MT neurons with similar selectivity. This explains why the effect sizes are modest and may not reach significance on individual sessions. For comparison with other studies, have you tried using higher light powers and illuminating larger volumes of tissue? I would expect to see stronger disruptions of performance, but for all directions of visual motion. Nonetheless, the disruptions would be specific for visual motion stimuli placed at the appropriate retinotopic locations. This might serve as a useful positive control and point of comparison with other studies.

We appreciate these suggestions, although in our view it is unclear whether low light power explains the small effect sizes. Increasing laser power might have suppressed the targeted column more strongly, but this would come at the cost of affecting nearby columns/clusters and thus diluting the directional specificity of the manipulation. It is conceivable that, if laser power were to be steadily increased, one might see larger bias effects kick in before being swamped by the overall disruption in performance (Murasugi et al., 1993). We suspect that this would not occur or be too subtle to detect, mainly due to floor effects; i.e., increasing the strength of suppression has diminishing returns as neurons approach zero spikes/s, whereas stimulation has a much higher ceiling. Also working against us here is the fact that current viral vector technology limits the percentage of neurons that express the transgene. Still, we hope to address these issues with future experiments that include more systematic variations in light power, and better reagents.

The histology on the Jaws expression is an important aspect of the results. An obvious concern is that the histology was done in area LIP not area MT, and it is not obvious to me whether the expression patterns in area MT would be expected to closely follow those in LIP or whether they might end up being very different. Are there neuronal cell type similarities (or differences) between MT and LIP that should be pointed out when interpreting these results?

We agree with this reviewer that presenting some histology will be useful to readers. We also agree with all three reviewers that there are limits to what one can infer from expression in a different brain area of a different monkey. The revised manuscript now clarifies these limitations. We are not aware of any studies that investigate the similarities or differences in cell types or laminar organization of MT and LIP, but if there is such a study we would gladly cite it.

Unless I missed it, the paper presents the effects of photosuppression only for the trials that included the opt-out option. What effects do you find if the opt-out is not available? Does this substantially change the size of the effect?

We apologize for leaving this ambiguous. In fact, to maximize statistical power, we pooled together trials with and without the sure-bet option for all graphs and analyses of choice data, with the exception of Figure 1C. We have now clarified this in the Materials and methods (subsection “Data analysis”, first paragraph). If we analyze these conditions separately, it turns out that the photosuppression effects were about twice as large on trials where the sure-bet option was unavailable compared to when it was available. We don’t know how seriously to take this result since the effect sizes were modest overall. One possible interpretation is that the sure-bet option absorbed some of the potential biasing effect – that is, some choices that would have been biased away from the preferred direction were instead rendered as opt-outs. But this would seem to predict an overall increase in opt-out choices on laser trials, which we did not observe (if anything, it was the opposite). We have added this point to the newly rewritten Discussion section regarding possible deviations from a straightforward model in which confidence and choice are yoked via a common decision variable (subsection “A common mechanism for choice and confidence”).

Author details

1. Christopher R Fetsch

   1. Zanvyl Krieger Mind/Brain Institute, Johns Hopkins University, Baltimore, United States
   2. Solomon H. Snyder Department of Neuroscience, Johns Hopkins University, Baltimore, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   For correspondence

   cfetsch@jhu.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7921-8306

2. Naomi N Odean

   1. Kavli Institute, Columbia University, New York, United States
   2. Howard Hughes Medical Institute, Columbia University, New York, United States
   3. Department of Neuroscience, Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States

   Contribution

   Conceptualization, Investigation, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

3. Danique Jeurissen

   1. Kavli Institute, Columbia University, New York, United States
   2. Howard Hughes Medical Institute, Columbia University, New York, United States
   3. Department of Neuroscience, Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States

   Contribution

   Investigation, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3835-5977

4. Yasmine El-Shamayleh

   Department of Physiology & Biophysics, Washington National Primate Research Center, University of Washington, Washington, United States

   Contribution

   Investigation, Methodology, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5396-2823

5. Gregory D Horwitz

   Department of Physiology & Biophysics, Washington National Primate Research Center, University of Washington, Washington, United States

   Contribution

   Resources, Supervision, Funding acquisition, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-5130-5259

6. Michael N Shadlen

   1. Kavli Institute, Columbia University, New York, United States
   2. Howard Hughes Medical Institute, Columbia University, New York, United States
   3. Department of Neuroscience, Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States

   Contribution

   Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Methodology, Writing—original draft, Writing—review and editing

   For correspondence

   shadlen@columbia.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2002-2210

• 3,332

  Page views

• 616

  Downloads

• 26

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.