Perceptual decision related activity in the lateral geniculate nucleus

Yaoguang Jiang,1 Dmitry Yampolsky,2 Gopathy Purushothaman,2 and Vivien A. Casagrande1,2,3

1Department of Psychology, Vanderbilt University, Nashville, Tennessee; 2Department of Cell and Developmental Biology,Vanderbilt University, Nashville, Tennessee; and 3Department of Ophthalmology and Visual Sciences, Vanderbilt University,Nashville, Tennessee

Submitted 23 January 2015; accepted in final form 26 May 2015

Jiang Y, Yampolsky D, Purushothaman G, Casagrande VA.Perceptual decision related activity in the lateral geniculate nucleus. JNeurophysiol 114: 717–735, 2015. First published May 27, 2015;doi:10.1152/jn.00068.2015.—Fundamental to neuroscience is the un-derstanding of how the language of neurons relates to behavior. In thelateral geniculate nucleus (LGN), cells show distinct properties suchas selectivity for particular wavelengths, increments or decrements incontrast, or preference for fine detail versus rapid motion. No studies,however, have measured how LGN cells respond when an animal ischallenged to make a perceptual decision using information within thereceptive fields of those LGN cells. In this study we measured neuralactivity in the macaque LGN during a two-alternative, forced-choice(2AFC) contrast detection task or during a passive fixation task andfound that a small proportion (13.5%) of single LGN parvocellular (P)and magnocellular (M) neurons matched the psychophysical perfor-mance of the monkey. The majority of LGN neurons measured in bothtasks were not as sensitive as the monkey. The covariation betweenneural response and behavior (quantified as choice probability) wassignificantly above chance during active detection, even when therewas no external stimulus. Interneuronal correlations and task-relatedgain modulations were negligible under the same condition. A bot-tom-up pooling model that used sensory neural responses to computeperceptual choices in the absence of interneuronal correlations couldfully explain these results at the level of the LGN, supporting thehypothesis that the perceptual decision pool consists of multiplesensory neurons and that response fluctuations in these neurons caninfluence perception.

thalamus; vision; perceptual decision; choice probability

WE HAVE KNOWN FOR DECADES that in mammals the retina sendsinput directly to the lateral geniculate nucleus (LGN) of thethalamus, which, in turn, relays this information to the primaryvisual cortex. In primates this pathway is critical for consciousvisual perception (Casagrande and Ichida 2011; Jones 2007;Logothetis 1998; Saalmann and Kastner 2011; Schmid et al.2010; Sherman and Guillery 2001). What has remained a hugepuzzle is what the LGN contributes to this process. Neurons inthe LGN have been found to behave physiologically very muchlike their retinal inputs (reviewed in Kaplan 2013; Shapley1990; Sherman and Guillery 2001), even though the LGNreceives most of its synapses from nonretinal sources (re-viewed in Casagrande and Norton 1991; Nassi and Callaway2009). The argument has been made that these nonretinalinputs gate, or modulate, the main driving input from the retinadepending on the behavioral context of the animal (reviewed inCasagrande et al. 2005). No studies to date, however, have

examined LGN neural responses when an animal is activelymaking a perceptual decision. Specifically, the few studies thathave measured LGN neural responses in awake behavinganimals all reported that the basic receptive field characteristicsare similar between awake and anesthetized preparations butthat these responses are often modulated by eye movements,attention, and the arousal/alertness states of the awake animal(Alitto et al. 2011; Martinez-Conde et al. 2002; McAlonan etal. 2008; Ramcharan et al. 2001; Reppas et al. 2002; Royal etal. 2006; Ruiz et al. 2006). None of these studies, however,directly measured neural responses in the LGN when theanimal was required to make a perceptual decision using theinformation available within the receptive fields of those LGNcells.

To explore the functional involvement of LGN during per-ceptual decision making, we aimed to address three mainquestions in this study. First, how sensitive are LGN parvocel-lular (P) cells and magnocellular (M) cells in detecting smallchanges in contrast compared to a monkey performing thesame task? Second, do LGN cells show neural signaturesdifferentiating a task that requires active detection from onethat only requires passive fixation? Finally, can the activity ofsingle LGN cells actually reflect a perceptual decision?

These questions are particularly interesting and relevant tostudy at the level of the LGN for several reasons. First, earlyphysiological recordings revealed that single sensory neuronsin the retina and LGN were exquisitely sensitive in detectingcontrast (Barlow et al. 1971; Derrington and Lennie 1984;Enroth-Cugell and Robson 1966; Hubel and Wiesel 1961;Kaplan and Shapley 1986; Shapley et al. 1981), but rarely wasthe sensitivity of these single neurons directly compared withthe simultaneously obtained psychophysical performanceof the animal (but see Kang and Malpeli 2009). Second, earlystudies in anesthetized preparations also established that dif-ferent LGN cell classes show distinct response profiles. Spe-cifically, M cells, on average, are more sensitive to contrastcompared with P cells (Kaplan 2013; Shapley 1990), implyingthat these M cells are largely responsible for sustaining ourperception of contrast. In reality, however, behavioral mea-surements in monkeys with localized LGN lesions revealed theopposite effect, with P layer lesions producing a pronounceddeficit in contrast sensitivity, whereas M layer lesions mostlyonly affected motion sensitivity (for example see Merigan et al.1991; Merigan and Maunsell 1990; Schiller et al. 1990).Finally, it is known that stimulus-independent, random fluctu-ations of sensory neural responses can covary with the percep-tual decisions of the animal. The strength of this covariation,quantified as “choice probability” (Britten et al. 1996), maycarry information about the causal contribution of a sensory

Address for reprint requests and other correspondence: V. A. Casagrande,Dept. of Cell and Developmental Biology, Vanderbilt Medical School, PMB407935, 465 21st Ave. S., Nashville, TN 37240 (e-mail: vivien.casagrande@vanderbilt.edu).

J Neurophysiol 114: 717–735, 2015.First published May 27, 2015; doi:10.1152/jn.00068.2015.

7170022-3077/15 Copyright © 2015 the American Physiological Societywww.jn.org

neuron to an emergent perceptual decision (Parker and New-some 1998). Many studies have confirmed the presence of aweak but significantly above-chance choice probability in anumber of visual cortical areas (Britten et al. 1996; Cook andMaunsell 2002; Dodd et al. 2001; Grunewald et al. 2002; Liuand Newsome 2005; Nienborg and Cumming 2006; Palmer etal. 2007; Purushothaman and Bradley 2005; Uka and DeAngelis2004; Uka et al. 2005), but similar measurements are seldommade in subcortical structures (however, see Liu et al. 2013)and never in the LGN.

Therefore, in this study we measured the sensitivity andchoice probability of LGN neurons during contrast detection,taking advantage of identified cell types in the LGN and therelative simplicity of its neural circuitry (Briggs and Usrey2011; Casagrande and Norton 1991; Nassi and Callaway2009). In addition, we also examined whether LGN P and Mcells differ in their sensitivity, choice probability, and responsedynamics. Finally we compared, in detail, LGN neural re-sponses during active detection and passive fixation tasks.Aspects of the data presented in this article were previouslypublished in abstract form (Jiang et al. 2012, 2013).

MATERIALS AND METHODS

Subjects

Two macaque monkeys (monkey 1: Macaca radiata, male, 7 kg, 10yr old; monkey 2: Macaca mulatta, male, 8 kg, 12 yr old) served assubjects. These monkeys were treated and cared for under an ap-proved protocol in accordance with the National Institutes of HealthGuide for the Care and Use of Laboratory Animals and the guidelinesof the Vanderbilt University Animal Care and Use Committee.

LGN localization and Surgeries

The surgical procedures were described in detail in our previouspublications (Royal et al. 2006; Ruiz et al. 2006). Briefly, theseprocedures were as follows. The LGNs first were localized viaanatomic images created with a GE Signa 1.5-Tesla MRI scanner.Under general anesthesia (15 mg/kg ketamine), each monkey wassecured in a titanium stereotaxic apparatus fitted with water-filled earbars, which were used to define the anterior-posterior (AP) zero andthe horizontal plane. A series of 1-mm-thick coronal images weretaken anterior to the ear bars in overlapping 0.5-mm increments. LGNcoordinates were calculated from images produced by GE 3.9 soft-ware and compared with the coordinates in a standard macaquestereotaxic atlas (Paxinos et al. 1999).

After the LGNs were located, the monkeys underwent surgeries forthe implantation of head posts and recording chambers. Under generalanesthesia (1–2% isoflurane in O2) and using sterile procedures, astainless steel head post (custom-made at Vanderbilt or courtesy ofDr. Andrew Rossi, National Institute of Mental Health) was secured tothe front of each monkey’s skull with titanium screws (2–4 mm;Synthes, West Chester, PA) and methyl methacrylate cement (Biomet,Warsaw, IN). A recording chamber (stainless steel, �20-mm diame-ter, custom-made at Vanderbilt or purchased from Crist, Hagerstown,MD) was centered over the right (monkey 1) or left (monkey 2) LGNusing coordinates calculated from the MRI images (monkey 1: AP �7, ML � 12.5; monkey 2: AP � 7, ML � 12) and was secured to theskull using titanium screws and methyl methacrylate cement. Thebone enclosed within the recording chamber was removed in asubsequent sterile surgery.

Visual Stimulus Presentations and Behavioral Tasks

A computer-based real-time experimental operating system (VisualTEMPO; Reflective Computing, St. Louis, MO) controlled stimuluspresentation time, monitored eye movements, and delivered reward orpunishment in every trial. Eye position data were collected via aninfrared camera (1 kHz, spatial resolution 0.1° � 0.1°, EyeLink 1000;SR Research, Kanata, ON, Canada) and fed to both the operatingsystem (TEMPO) and the recording system (Plexon, Dallas, TX) viaanalog channels. Visual stimuli were generated in MATLAB Psych-toolbox (The MathWorks, Natick, MA) running on a Macintoshcomputer (G5) and displayed on a SONY Multiscan G420 monitor(refresh rate 90 Hz, resolution 1,280 � 1,024 pixels, visible area 36° � 29°;SONY, San Jose, CA). The RGB-to-luminance table for the monitorwas calibrated periodically, and the nonlinearity was corrected ac-cordingly.

Fixation task. After recovering from the surgeries, the monkeyswere initially trained to enter and exit primate chairs on command andto tolerate head fixation for extended periods. They were then placedin a dark recording room (�2 cd/m2) and conditioned to fixate on acentral fixation spot (0.1° � 0.1°, Michelson contrast 80%, on auniform gray background with average luminance of �15 cd/m2) onthe computer monitor positioned 57 cm in front of the monkeys. Themonkeys were required to maintain fixation within a 1° � 1° (range:0.5° � 0.5° to 1.5° � 1.5°) invisible window centered on the fixationspot. Continuous fixation for an extended period of time (200 to 800ms) was rewarded with a drop of juice (0.1 to 1 ml).

Contrast detection task. After learning to maintain stable fixation,the monkeys were trained to perform a two-alternative, forced-choice(2-AFC) contrast detection task (see Fig. 1A). The detection task wasdistinguished from the simple fixation task by changing the centralfixation spot from a filled to a hollow square. This fixation squareappeared at the beginning of each detection trial, and the monkeyinitiated the trial by bringing its gaze into the fixation window within1,000 ms. After an initial fixation period of 200–500 ms (fixed at 350ms during recordings), a monochromatic circular contrast patch (2° indiameter) or a Gaussian contrast profile (larger in size but containingthe same overall energy as the circular patch) appeared on either theleft or the right of the fixation spot. The contrast stimuli were 5.5°away (horizontally) from the central fixation spot during initial train-ing and were subsequently moved around to establish thresholdperformance at different locations. During recordings, the diameterand position of the stimulus were adjusted daily so that in 50% of thetrials the stimulus completely covered the classical receptive field(center and surround) of the cell despite small fixational eye move-ments, and in the other 50% of the trials the same stimulus appearedat a symmetrical location in the opposite visual hemifield. Both thecontrast stimulus and the fixation point remained on for 200 ms (pulsefunction), after which two target dots (50–80% contrast) appeared,one in each hemifield. During initial training, the target dots were 8°away (horizontally) from the central fixation spot. During recordings,target dot positions were sometimes adjusted, but always nonoverlap-ping with the stimulus locations. Upon target onset, the monkey wasallowed to make a saccade to one of the two target locations toindicate the side on which it had detected the contrast stimulus. If themonkey made a saccade to the correct target location within 2,000 msafter target onset and maintained its gaze within the invisible targetwindow (1° � 1° to 10° � 10°, centered on the target dot) for 200 ms,it would receive a drop of juice reward (0.1 to 1 ml) at the end of thetrial.

Circular patches and Gaussian contrast profiles. To ensure that ourthreshold measurements were not biased by the type of contraststimulus we chose, we compared the psychometric and neurometricperformance using circular contrast patches as well as Gaussiancontrast profiles in both the fixation and the detection tasks. For thepatch, the stimulus diameter equaled the classical receptive field(center and surround) plus the fixation window diameter, ensuring that

718 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

the same contrast would always fall onto the cell’s receptive fielddespite the small fixational eye movements that continuously occur.To enable comparisons between the patch and the Gaussian profile,the Gaussian stimuli always had the same contrast levels (defined bythe peak of the Gaussian curve) and the same energy levels (definedby the area under the Gaussian curve) as the corresponding patchstimuli. This meant the matching Gaussian profiles were larger indiameter compared with their circular patch counterparts. Fourieranalysis on individual stimuli determined that during recordings, most(�90%) of the contrast stimuli used (both patch and Gaussian) had asignificant amount (�50%) of energy distributed in the low to me-dium spatial frequency range (�10 cycles/deg), which can effectivelydrive most LGN P and M neurons under dim light conditions such asours (Derrington and Lennie 1984). We did not, however, tailor ourstimuli to optimally drive the receptive field center of each LGN cell,as is commonly done in anesthetized and paralyzed preparations.Because of a combination of factors such as the constant fixational eyemovements in awake monkeys, the small receptive field sizes of LGN

neurons, and the large number of trials required to establish stablepsychometric and neurometric functions, we determined that it wouldnot be feasible to optimize the stimulus for every neuron on everytrial. We consider our choice of the visual stimulus to be appropriatefor the purpose of this study because 1) our main goal was to study thelink between LGN activity and perceptual choices, not to measure theabsolute sensitivity of LGN neurons; and 2) our contrast patchstimulus was able to elicit brisk responses from both P and M LGNneurons (see Fig. 1, B–E). Compared with the contrast patches, theGaussian profiles with the same overall energy yielded slightly poorerneurometric performance, presumably due to the fact that only alimited proportion of their total energy fell in the receptive field centerof the cell.

Measuring psychophysical thresholds. During each recording ses-sion, visual stimuli of 5 or 9 different contrast levels (ranging from�99% to 80% in contrast and encompassing the psychophysicalthreshold) were displayed in each hemifield. Contrast in this case wasdefined as the luminance difference between the stimulus and its

−100 −50 0 50 100

1

2

3

Time (ms)

Firin

g R

ate

(Nor

mal

ized

) 0% contrast 0.8% contrast 2.4% contrast 38% contrast 80% contrast

−100 −50 0 50 100

1

3

5

7

9

Time (ms)

Firin

g R

ate

(Nor

mal

ized

) 0% contrast 0.8% contrast 2.4% contrast 38% contrast 80% contrast

A

B C

Choose targetStimulus on/Maintaining fixationMaintaining fixation

Stimulus off/Choose target

Targets off/Reward or punishment

350 ms 200 ms 0-2000 ms 100-1500 ms

Time

P Neuron M Neuron

−100 −50 0 50 1000

10

20

30

Time (ms)

Tria

l Num

ber

D

−100 −50 0 50 1000

10

20

30

40

Time (ms)

Tria

l Num

ber

E

Fig. 1. Contrast detection task. A: the two-alternative, forced-choice (2-AFC) contrast detection task. A contrast patch profile was briefly presented either in thereceptive field of the cell (red dotted circle) or at a symmetric location in the other hemifield. The monkey made a saccade to 1 of 2 target locations to indicateits choice. B and C: peristimulus time histograms (PSTHs) of example lateral geniculate nucleus (LGN) P (B) and M neurons (C) at different contrasts. Stimulusonset, 0 ms. D and E: raster plots of the same LGN P (D) and M neurons (E) at high contrast (80%). Stimulus onset, 0 ms.

719LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

background, divided by their sum. The polarity of the stimulusmatched the sign of the cell being recorded (�0% contrasts forON-center cells and �0% contrasts for OFF-center cells). Differentcontrast levels and stimulus locations were randomly mixed, withequal probabilities of left or right appearance and higher proportionsof low- to medium-contrast trials to ensure accurate estimations of thethreshold. When no stimulus was presented (i.e., 0% contrast, catchtrials), the monkey was randomly rewarded in 50% of the trials forchoosing either side. These trials were used to monitor the behavioralbias of the animal as well as to establish the link between LGN neuralresponses and the monkeys’ choices in the absence of any physicalstimulus.

Correction of behavioral biases in the detection task. Precautionswere taken to ensure unbiased estimations of the true threshold(Purushothaman and Bradley 2005). For example, during initial train-ing, only high-contrast stimuli were used to teach the monkeys theprocedures and rules of the detection task. After the monkeys’performance reached �95% correct, contrasts were gradually droppeduntil a stable psychophysical threshold was reached and an overallperformance of 70%–85% correct was maintained around that thresh-old for �3 days. After threshold performance was established for theinitial set of stimulus parameters, the monkey was taught to generalizethe task over a number of dimensions including variations in stimulussize, stimulus eccentricity, and stimulus location (upper or lowervisual field). The psychophysical threshold often increased when oneor a few of these parameters were initially changed, in which caseadditional training was necessary to reestablish stable threshold per-formance under the new configuration. During training, behavioralbiases (i.e., preferring to choose one side over the other) weremonitored online via catch trials (i.e., 0% contrast, no stimuluspresent), and if found to be significant, biases were corrected byadjusting the relative frequencies of stimulus presentation on eachside. It is noteworthy, however, that the 50:50 presentation ratio wasalways maintained during LGN recordings. Random guesses werestrongly discouraged by introducing a between-trial “time-out” periodfor wrong choices (discounting trials in which no stimulus waspresent). Correct choices were encouraged by progressively increas-ing the reward size for consecutive correct trials. Additionally, beforemeasuring the threshold each day, the monkeys spent at least 45 minin the recording room to fully adapt to the dim light condition.

Single-Unit Recordings

Extracellular activities from single and pairs of LGN neurons wererecorded with Parylene-coated tungsten microelectrodes (1–3 M�;FHC, Bowdoinham, ME). The electrode was placed within a stainlesssteel guide tube (both sterilized), and then the electrode and the guidetube were both mounted on the recording chamber of the monkey viaan adapter (FHC). The guide tube was manually pushed down topenetrate the dura, after which the electrode was slowly advanced intothe brain via a motorized microdrive (FHC). Neural responses wereamplified, band-pass filtered, and recorded (30 kHz; Plexon). Single-unit activities were sorted offline using a combination of principlecomponent analysis (PCA), the Template Matching algorithm, and theValley Seeking algorithm in three-dimensional feature space (OfflineSorter; Plexon).

Cell mapping and classification. LGN cells were hand mapped firstwith a flashlight and then with an elongated bar (drawn in Psychtoolbox,MATLAB) sweeping across the visual field while the monkey wasrewarded for maintaining fixation (0.1–1 ml of juice/500 ms). The bar’sheight, width, luminance, orientation, and direction of motion wereadjusted by the experimenter to best define the receptive field borders.According to its location and response properties, each LGN cell wascategorized as either ON- or OFF-center and either P or M type.Specifically, cells that showed increases in firing rates to contrast incre-ments in their receptive field centers were classified as ON-center cells,whereas those that showed increases in firing rates to contrast decrements

in their receptive field centers were classified as OFF-center cells. P or MLGN cells were classified offline on the basis of their physiologicalproperties. According to our previous experience (Norton et al. 1988;Royal et al. 2006; Ruiz et al. 2006; Xu et al. 2001), two of the mostreliable classification criteria were visual response latency and visualresponse transiency: 1) visual response latency: compared with P cells, Mcells demonstrated shorter onset and peak latencies. To determine eachneuron’s latency, we plotted its peristimulus time histogram (PSTH) athigh contrast (80% contrast for ON-cells and �99% contrast for OFF-cells) in 5-ms bins. We defined the onset latency as the time fromstimulus onset to the first bin on the rising edge of the PSTH surpassing3 SDs above base firing rate, and the peak latency as the time fromstimulus onset to the bin with the highest amplitude in the PSTH. In ourdata set, compared with P neurons, M neurons had, on average, shorteronset latencies (P neurons: 25.2 2.0 ms; M neurons: 19.6 2.2 ms;P � 0.041, t-test) and shorter peak latencies (P neurons: 72.9 5.3 ms;M neurons: 43.5 3.3 ms; P � 0.000, t-test) (see Fig. 1, B–D, forexample PSTHs and raster plots for typical P and M neurons). 2) Visualresponse transiency: compared with P cells, M cells exhibited greatertransient bursts to the onset of visual stimuli. Using the same PSTH athigh contrast, we computed for each neuron a transiency index, whichwas similar to the phasic/tonic index used in our previous studies (Irvinet al. 1986; Norton et al. 1988; Norton and Casagrande 1982). In thisstudy the transiency index � 100 � (100–200 ms response �spontaneous response)/(0–100 ms response � spontaneous response)� 100. A larger transiency index indicated a greater level of tran-siency in the neural response. In our data set, M neurons respondedmore transiently than P neurons (P index: 22.95 5.97; M index:39.53 7.5; P � 0.044, t-test) (see Fig. 1, B–D, for example PSTHsand raster plots for typical P and M neurons).

Some or all of the following classification criteria also were used inthis study depending on the cell: 1) relative depth of the electrode andshifts in ocular dominance within the same penetration: the macaqueLGN has 6 layers in the central 0°–17° of visual field representation, withthe M neurons (layers 1–2) lying below the P neurons (layers 3–6). Theipsilateral eye projects to layers 2, 3 and 5, whereas the contralateral eyeterminates in layers 1, 4 and 6. Shifts in ocular dominance therefore wereuseful for cell classification if all the receptive fields encounteredin the same penetration remained at or below the horizontal meridianrepresentation, in which case the electrode passed from the P layers atthe top to the M layers at the bottom without reentering P layers(Erwin et al. 1999). 2) Receptive field size: at the same eccentricity,M neurons had larger receptive fields than P neurons.

Finally, because previous literature suggests significant overlap forthe two populations in most response features (for example, seeNorton et al. 1988; Irvin et al. 1993; Xu et al. 2001; White et al. 2001),we did not establish strict cutoff points for any of these classificationcriteria. Instead, we took into consideration all of these featurestogether to achieve a more holistic, albeit not completely quantitative,method of cell classification.

Single-unit data set. All the cells that could be clearly mapped andmaintained long enough to characterize both psychophysical and neuralresponses (�150 trials) in the detection task were included in our data set(overall: n � 89; monkey 1: n � 61; monkey 2: n � 28). We identifiedin this data set 41 ON-center P neurons, 27 OFF-center P neurons, 19ON-center M neurons, and 2 OFF-center M neurons. Among these 89neurons, 49 were also recorded in the fixation task. An additional 20 pairsof neurons were recorded only in the fixation task and incorporated onlyin our analysis of interneuronal correlation (see below).

Data Analysis

Behavioral, physiological, and eye movement data were recordedin Plexon; data analyses were performed using self-developed scriptsin NeuroExplorer (Nex Technologies, Madison, AL) and MATLAB.

Psychometric functions. Only recording sessions in which themonkey completed �150 trials and achieved an overall performance

720 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

of �65% correct were included in this analysis. To generate apsychometric function, the proportion of correct responses at eachcontrast was plotted, and a Weibull function was fitted to the data:

P�c� � 1 � �0.5 · e��c⁄���� (1)

where P(c) is the probability of correct responses at a certain contrastlevel c, � is the contrast that supports threshold performance (82%correct), and � is the slope of the function.

Neurometric functions. Basic procedures in computing neurometricfunctions were similar to those described in previous studies (Barlowet al. 1971; Britten et al. 1992; Purushothaman and Bradley 2005). Togenerate neurometric functions, we used spike counts in fixed timewindows (0–150 ms after stimulus onset). For each neuron at eachcontrast level, an ROC (receiver operating characteristic) curve wascomputed, plotting for all possible signal detection criteria (spikes/s)the proportion of stimulus-inside-receptive-field (test) trials againstthe proportion of stimulus-outside-receptive-field (reference) trials inwhich the spike count exceeded the criteria. The area under the ROCcurve indicated the predicted detection performance of an idealobserver using only the information available in this neuron’s re-sponses (Green and Swets 1966). To generate a neurometric function,each value for area under the ROC curve was plotted against itscorresponding contrast, a Weibull function (Eq. 1, as described above)was fitted, and the neurometric threshold and slope were obtainedfrom the fitted curve.

Eye movement analysis. We analyzed fixational eye movements bycomputing 1) the average eye position deviation from central fixation,and 2) the average eye speed (i.e., moment-to-moment change in eyeposition) during fixation. An eye movement was characterized as asaccade if it met the following criteria: at least 8 ms of monotonicchange in eye position and a peak velocity of at least 30°/s. The onsetof the saccade was defined as the moment it reached a speed of 10°/s.The reaction time of the monkey was defined as the time fromstimulus onset to saccade onset.

Choice probability. Basic procedures in computing choice proba-bility were similar to those described in previous studies (Britten et al.1996). To compute choice probabilities, we used spike counts in fixedtime windows (0–150 ms after stimulus onset). The choice probabilityat each contrast was measured by plotting as an ROC curve theproportion of choice-inside-receptive-inside trials (i.e., trials in whichthe monkey chose to saccade toward the receptive field location)against the proportion of choice-outside-receptive-field trials thatexceeded each spike count criterion, and then computing the areaunder the ROC curve. A one-way ANOVA was used to test whetherthe stimulus contrast level had a significant effect on choice proba-bility values. For all the contrast levels at which choice probabilitydistributions did not differ (�4% contrast), neural responses werez-score normalized (with reference to the mean and SD of the neuralresponse at each contrast level) and pooled together to generate asingle choice probability value for each neuron. A grand choiceprobability was derived for the entire population by z-score normal-izing each neuron’s response, pooling together all the choice-insideand choice-outside-receptive-field trials, and then performing an ROCanalysis on the population response distributions. Recent literaturesuggests that such grand choice probabilities derived from z-scorenormalized responses can sometimes underestimate the populationchoice probability (Kang and Maunsell 2012), so we also reported themean choice probability alongside the grand choice probability. Thesignificance of these choice probabilities was assessed using a per-mutation (i.e., bootstrapping) test (see below). To accurately estimatechoice probability, only neural recordings that met the followingcriteria were included in this analysis: 1) behavioral bias ratio (choice-inside/choice-outside-receptive-field trials) �0.25 and �4, and 2) atleast 10 choice-inside and 10 choice-outside-receptive-field trialswere recorded. Of the 89 neurons recorded in the detection task, 75 (Pneurons: 54, M neurons: 21) were included in the choice probabilityanalysis according to these criteria.

Permutation test. The significance of a choice probability valuewas assessed using a permutation (i.e., bootstrapping) test. Specifi-cally, for all the z-score normalized trials used to compute choiceprobability, each trial was randomly reassigned as either a choice-inside or choice-outside-receptive-field trial, with the same probabilitythat the monkey chose that location during the detection task. Achoice probability value was then recomputed for the new spike countdistributions using an ROC analysis. For each neuron this process wasrepeated 1,000 times to construct a “random” choice probabilitydistribution. A measured choice probability was considered to besignificant if it fell outside the 95% confidence interval of the mean ofthis “random” choice probability distribution.

Temporal dynamics of neurometric threshold, choice probability,and Fano factor. The initial assessment of neurometric threshold andchoice probability was done in a fixed time period (0–150 ms afterstimulus onset). To characterize the temporal changes of these vari-ables on a finer scale, we used two types of moving time windows: 1)a “growing” window started at 0 ms and grew in its width (i.e.,duration) in steps of 12.5 ms. For example, the neurometric thresholdat 100 ms was computed from the neural response during the 0–100ms period, and the threshold at 200 ms was computed from the neuralresponse during the 0–200 ms period. 2) A “sliding” window had afixed width (i.e., duration) of 50 ms and advanced in steps of 12.5 ms.In this case, the neurometric threshold at 100 ms was computed bycounting spikes in the 50–100 ms period, and the threshold at 200 mswas derived by counting spikes in the 150–200 ms period. Toaccurately describe the fine temporal changes in threshold and choiceprobability, only the most sensitive M and P neurons were included inthis analysis. The criteria for choosing such neurons were as follows:1) neurometric threshold �100% in at least 4 consecutive sliding timewindows, 2) behavioral bias ratio (choice-inside/choice-outside-re-ceptive-field trials) �0.25 and �4, and 3) at least 10 choice-inside and10 choice-outside-receptive-field trials were recorded. According tothese criteria, 57 neurons (P neurons: 37; M neurons: 20) from thedetection task were included in the temporal dynamics analysis. Mean SEvalues are reported for the population dynamics of threshold andchoice probability. Additionally, for the same data set, we alsoanalyzed the temporal dynamics of the Fano factor using fixed 50-mssliding windows that advanced in steps of 12.5 ms. For each neuronat each contrast level (stimulus-inside-receptive-field trials only), wecomputed the mean and variance of its firing rate. For each timewindow, individual variances were plotted against their correspondingmeans, with each data point representing one neuron at one contrastcondition. The Fano factor for that time window was defined as theslope of the regression relating the variance to the mean. The esti-mated slope and its 95% confidence interval values are reported forthe population dynamics of the Fano factor.

Interneuronal noise correlation. Interneuronal correlations be-tween pairs of LGN neurons (recorded on the same electrode) werecomputed by first z-score normalizing each neuron’s response in eachtrial, according to its response mean and SD at each contrast, and thenestimating the (nonparametric) correlation coefficient between the twoz-scored firing rates. Because the Pearson correlation coefficient (r) isa biased estimator of the true correlation, especially for small samplesizes (n), an adjusted, unbiased estimator r= (Lehmann 1957; Olkinand Pratt 1958) was reported alongside the Pearson r:

r=� r�1 � �1 � r2� ⁄ 2�n � 3��. (2)

Additionally, fixed 50-ms sliding windows were used to characterizethe temporal changes in interneuronal correlations during the detec-tion and the fixation tasks.

Pooling Model

To simulate a perceptual decision pool of n units, n single neuronswere randomly chosen, with replacement, from our entire data set. Toconstruct a single trial at a given contrast, each chosen neuron’s

721LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

response was simulated by randomly drawing a number from aGaussian distribution. The mean and variance of this Gaussian distri-bution were determined by that neuron’s measured response at thatcontrast. Because the spontaneous response rates of LGN neuronswere quite high (16.15 1.73 spikes/s in our sample), we consideredit appropriate to approximate the Poisson distributions of neuralresponses with Gaussian distributions. In each trial, a decision wasmade by comparing the summed activity of the neural pool at the testcontrast to the summed activity of the same neurons at the reference(i.e., 0%) contrast. Fifty such trials were simulated for each of the fivecontrast levels. The “psychophysical” performance was calculated asthe percentage of correct choices at each contrast and was then fittedwith the cumulative Weibull function (Eq. 1). The “psychophysical”threshold was extracted from the fitted function as described above.The choice probability for each neuron was quantified as the covari-ation between the simulated neural response and the simulated choiceat 0% contrast. At each parameter combination (see below), the sameset of simulations (50 trials � 5 contrasts) was repeated 200 times,each time with a new random sample of n neurons, thus giving reliableestimations of model performance. The fitness of the model wasevaluated by computing a goodness-of-fit (GoF) index:

GoF � 1 ⁄ 3 � �simulated threshold � measured threshold ⁄measured threshold � simulated P choice probability � measuredP choice probability ⁄ measured P choice probability � simulatedM choice probability � measured M choice probability/

measured M choice probability� � 100% (3)

Model parameters. The following parameters were varied in thepooling model: the number of P and M neurons in the pool (n �1–512), the integration time window (t � 25–200 ms), the Fano factor(f � 0.5–2.0), interneuronal noise correlation (r � 0–0.3), anddownstream pooling noise (p � 0–4.0). For most simulations, theFano factor, interneural correlation, and pooling noise were fixed attheir experimentally measured values. Specifically, an average Fanofactor was derived for each integration time window, ranging from 0.8to 1.4. The interneuronal correlation for LGN P-P and M-M neuronpairs was measured as 0.028. The interneuronal correlation betweenM and P neurons could not be estimated from our data, because it wasnearly impossible to record from an M-P neuron pair on the sameelectrode at the same eccentricity. Several M-P correlation valueswere tested in our model, ranging from 0 to 0.05. The simulationresults did not change significantly within this range of M-P correla-tions (data not shown). For the simulations reported in this article, afixed value of 0.01 was used. The pooling noise was assumed to be2.0. The choice of these values was considered neurobiologicallyrealistic and meaningful for the following reasons: 1) Fano factor:recordings in anesthetized as well as alert animals have reportedsignificant variability in the responses of single cortical neurons (Fanofactor � 1–3) (McAdams and Maunsell 1999; Oram et al. 1999;Tolhurst et al. 1983). The Fano factor of subcortical visual neurons,however, was relatively low (�1.0). This is true for retinal ganglioncells (RGCs) (Berry et al. 1997; Levine et al. 1992; Reich et al. 1997)as well as LGN cells (Kara et al. 2000). The Fano factors measured inour detection task and adopted by the model were thus in agreementwith those previously reported measurements. 2) Interneuronal corre-lation: in sensory cortex, interneuronal correlations between pairs ofnearby neurons are typically weak but positive (�0.1–0.2) (Averbecket al. 2006). In the LGN, convergent feedforward, divergent feedfor-ward, and lateral connections are sparser than those in the cortex(Casagrande and Xu 2004; Nassi and Callaway 2009). It is thereforewithin expectation that the average interneuronal correlation for LGNP-P and M-M neuron pairs would be close to 0 (0.028). The interneu-ronal correlation between M and P neurons is likely to be evensmaller, because the M and P pathways remain segregated in differentlayers of the LGN, have different RGC inputs, and retain separatefeedback loops with V1 (Briggs and Usrey 2011; Ichida and Casa-

grande 2002; Ichida et al. 2014; Nassi and Callaway 2009). Inaddition, our simulations could always approach �99% GoF at someparameter combinations, given any interneuronal correlation value be-tween 0–0.05 for P-P, M-M, and P-M pairs. 3) Pooling noise: thedownstream pooling noise in our model can be considered the averageFano factor of cortical neurons onto which LGN neurons converge(Shadlen et al. 1996). In most simulations this number was fixed at2.0, which is the average estimation of the Fano factor in cortex (seeabove). The success of our simulations (i.e., approaching �99% GoFat some parameter combinations), however, did not depend on thisassumption about the pooling noise. Similar model results could beobtained by simply assuming a true Poisson distribution for alldownstream neurons (i.e., pooling noise � 1.0). It is noteworthy thatwe only simulated independent downstream pooling noise values thatwere not associated with the local interneuronal correlations withinthe LGN, and that these noise values were fixed and did not changewith the duration of the integration window. Finally, the poolingscheme for the model was assumed to be uniform (i.e., responses ofall neurons were equally weighted).

RESULTS

We describe our main findings in several subsectionsbelow. First, we present data comparing the contrast sensi-tivity of the monkeys to the sensitivity of single LGN P andM neurons measured during active detection. We thenpresent neural sensitivity data for LGN neurons measuredduring passive fixation. The next section compares thepairwise interneuronal correlations measured during thedetection and fixation tasks. We follow with a section onchoice probability and then another on the temporal dynam-ics of neural sensitivity as well as choice probability.Finally, we model these measured relationships using dif-ferent numbers of LGN P and M neurons.

LGN Single Neurons Were Less Sensitive Than the MonkeysDuring Active Detection

Compared with the simultaneously obtained psychometricfunctions, typical neurometric functions had higher contrastthresholds and shallower slopes, as demonstrated by an exam-ple P cell in Fig. 2A and an example M cell in Fig. 2B. A smallpercentage of the cells in both classes, however, were assensitive as the monkeys, as shown by an example P cell in Fig.2C and an example M cell in Fig. 2D. For the entire data set(n � 89), the average neurometric threshold (54.4 4.78%contrast) was significantly different from the simultaneouslymeasured psychometric threshold (5.76 0.72% contrast; P �0.000, Wilcoxon signed rank test; Fig. 2E). The average ratioof neurometric to psychometric threshold was 40.74 10.75,indicating that the average LGN neuron was much less sensi-tive than the monkey during active contrast detection. None-theless, a subpopulation of LGN neurons (n � 12/89, 13.5%)demonstrated contrast sensitivities comparable to the monkey(permutation test, P � 0.05). Additionally, in our sample, Mneurons had slightly lower thresholds than P neurons (P thresh-old � 58.85 3.84%, M threshold � 49.22 7.37%; P �0.047, Kolmogorov-Smirnov test for small sample size). Torelate this result to previously reported LGN contrast responsesin anesthetized monkeys (Derrington and Lennie 1984; Pur-pura et al. 1988; Shapley et al. 1981), we also plotted z-scorenormalized contrast response functions (CRFs) for each cellgroup (Fig. 2F). Two-way ANOVA on the population CRFs ofM and P neurons showed that M neurons were indeed more

722 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

sensitive than P neurons (F � 3.87, P � 0.049), but there wasconsiderable overlap between the two populations.

The relatively high neurometric threshold observed in theLGN led us to run a series of necessary controls. First, finersampling of contrasts (measuring 9 contrast levels instead of 5)did not significantly alter the measured psychometric andneurometric threshold (psychometric: n � 12, P � 0.88,Wilcoxon signed-rank test; neurometric: n � 5, P � 0.56,Wilcoxon signed-rank test). Using Gaussian contrast profilesinstead of circular patches resulted in slightly higher psycho-metric (n � 6, P � 0.037, Wilcoxon signed-rank test) andneurometric thresholds (n � 6, P � 0.009, Wilcoxon signed-rank test) but did not alter the ratio of neurometric to psycho-metric threshold (neurometric/psychometric ratio � 46.48

12.32; P � 0.93, Wilcoxon signed-rank test). Additionally,within the range of receptive field eccentricities in our data set(average � 4.72 0.32°), neurometric threshold was notsignificantly influenced by eccentricity (r � �0.14, P �0.191). Similarly, the spontaneous firing rates of the neurons(average � 16.15 1.73 spikes/s) did not significantly influ-ence their thresholds either (r � �0.15, P � 0.161). To ruleout the possibility that the poor sensitivity of LGN neurons wasdue to unstable fixation (including fixational eye movementsand premature saccades), we performed several eye movementanalyses. First, we computed the threshold for each neuronusing only the 50% best-fixated trials and found that theneurometric threshold derived from these best-fixated trials(58.84 3.61%) was not significantly different from the

A B

E F

C

1 2 3 4 50

0.5

1

1.5

Contrast Level

Firin

g R

ate

(z−s

core

) P M

P Neuron M Neuron

P Neuron M Neuron

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ctPsychometric Function Neurometric Function

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ct

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ct

D

100 101 102

100

101

102

Neurometric Threshold

Psyc

hom

etric

Thr

esho

ld P (Monkey 1)

P (Monkey 2)M (Monkey 1)M (Monkey 2)

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ct

Fig. 2. Psychometric and neurometric functions. A–D: example neurometric functions (magenta) and simultaneously recorded psychometric functions (blue) inP neurons (A and C) and M neurons (B and D). Filled circles indicate psychometric performance, filled squares indicate neurometric performance, and the solidline is the fitted Weibull function. E: the neurometric thresholds of most LGN neurons were higher than the simultaneously recorded psychometric thresholds.Filled magenta symbols represent P neurons, and open green symbols represent M neurons: circles, monkey 1; inverted triangles, monkey 2; solid line, unity line;neurometric threshold � psychometric threshold. F: z-score normalized contrast response functions (CRFs) for LGN P and M populations, showing highercontrast sensitivity for M neurons. The magenta line with circles represents P neurons, and the green line with inverted triangles represents M neurons. Valuesare means SE (error bars); contrast levels: 1, 0%; 2, 0.8–1.3%; 3, 2.4–4.8%; 4, 19.5–38%; 5, 80–99%.

723LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

threshold derived from all trials (54.4 4.78%; P � 0.374,Wilcoxon signed-rank test for dependent variables; Fig. 3A).Second, to control for fixational tremors and drifts, we calcu-lated the average deviation of the eye (from central fixationpoint) for each neuron and ordered neurons by increasing valueof eye position deviation. By this criterion, the top 50% ofneurons (average eye position deviation � 0.18 0.01°, n �44) had slightly smaller neurometric thresholds (49.74 4.65%) than the rest of the neurons (eye position deviation �0.4 0.02°, n � 45, neurometric threshold � 59.28 4.9%;P � 0.044, Wilcoxon rank sum test; Fig. 3B). Thresholds forthese top 50% of neurons, however, were still significantlyhigher than the simultaneously measured psychometric thresh-olds (psychometric threshold � 5.22 0.87%; P � 0.000,Wilcoxon signed-rank test). Third, to control for microsac-cades, we calculated the average eye speed during fixation(i.e., first derivative of eye position) for each neuron andordered neurons by increasing value of eye speed. By thiscriterion, the top 50% of neurons (average eye speed � 1.12 0.03°/s, n � 44) had neurometric thresholds (54.69 5.01%) that were not significantly different from the rest ofthe neurons (eye speed � 1.84 0.05°/s, n � 45, threshold �54.47 4.72%; P � 0.928, Wilcoxon rank sum test; Fig.3C). Finally, to rule out the possibility of premature sac-cades, we characterized the relationship between reactiontime (average � 0.43 0.04 s) and neurometric thresholdand found no significant overall correlation between the two(r � 0.17, P � 0.111; Fig. 3D).

Neural Performance Improved During Simple Fixation

Because of the well-documented effects of attention andother types of modulations on the responses of single LGNneurons (for review see Casagrande et al. 2005; Saalmann andKastner 2009), we had predicted that contrast thresholds forsingle neurons would be elevated when the monkey wasrequired to passively fixate rather than to actively engage inmaking perceptual decisions. Instead, we found the opposite.Specifically, for a few LGN neurons, the neurometric functionsmeasured in fixation and detection were quite similar (see Fig.4A for an example). For most others, however, thresholdsmeasured during fixation were lower than those duringdetection (see Fig. 4B for an example). For the 49 neurons(P: n � 46; M: n � 3) that were tested in both tasks, theirneurometric thresholds were significantly correlated acrosstasks (r � 0.74, P � 0.000; Fig. 4C), and their thresholdsduring passive fixation were consistently lower comparedwith those obtained in active detection (fixation � 46.17 4.76%; detection � 54.38 4.78%; P � 0.001, Wilcoxonsigned-rank test; Fig. 4D).

To determine if the higher neural thresholds measured in theactive detection task could be explained by behavioral orphysiological changes across the two tasks, we made severaladditional comparisons. First, eye movement analyses revealedsignificant differences between the two tasks, in terms of bothaverage eye position deviation (detection � 0.26 0.02°,fixation � 0.33 0.02°; P � 0.000, pairwise t-test) and

3.125 12.5 50 >=1000

5

10

15

Neurometric Threshold (% contrast)

Num

ber o

f Neu

rons

50% Best Fixated (By Position) Other 50%

3.125 12.5 50 >=1000

5

10

15

Neurometric Threshold (% contrast)

Num

ber o

f Neu

rons

50% Best Fixated (By Speed) Other 50%

A B

C D

P (Monkey 1)P (Monkey 2)M (Monkey 1)M (Monkey 2)

0 50 1000

50

100

Threshold (All Trials)

Thre

shol

d (B

est F

ixat

ed)

400 450 5000

20

40

60

80

100

Reaction Time (ms)

Thre

shol

d (%

con

tras

t)

Fig. 3. Controls for eye movements during contrast detection. A: the neurometric threshold derived from the 50% best-fixated trials was similar to the threshold derivedfrom all trials. Filled magenta symbols represent P neurons, and open green symbols represent M neurons: circles, monkey 1; inverted triangles, monkey 2; solid line,unity line; 50% best-fixated threshold � all-trial threshold. B: the top 50% of neurons with smaller average eye position deviations had neurometric thresholds that wereslightly smaller than those for the rest of the neurons. C: the top 50% of neurons with smaller average eye speeds had neurometric thresholds that were similar to thosefor the rest of the neurons. D: the neurometric threshold was not significantly correlated with reaction time. Symbols are defined as in A.

724 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

average eye speed (detection � 1.54 0.08°/s, fixation � 1.65 0.07°/s; P � 0.000, pairwise t-test). Specifically, both eyeposition deviation and eye speed were smaller during activedetection (detection/fixation ratio, eye position � 0.83 0.05,eye speed � 0.93 0.02), indicating that the decrease inneural sensitivity observed in the detection task could not beattributed to unstable fixations of the monkeys. Furthermore,population PSTHs measured at the same contrast for bothmonkeys (Fig. 4E, monkey 1; Fig. 4F, monkey 2) confirmedthat recordings from the same neurons were stably maintainedacross the two tasks, and all neurons were sufficiently stimu-lated in both tasks.

Interneuronal Correlations Differed Between Detection andFixation

The next variable that we examined was the interneuronalcorrelation between pairs of simultaneously recorded LGNneurons during either active detection or passive fixation. Todetermine interneuronal noise correlation, each neuron’s re-sponse in each trial was first z-normalized to remove the“signal” (i.e., contrast response) and leave only the trial-to-trial“noise” fluctuations. Such scaled firing rates of one neuronwere then plotted against the correspondingly scaled firingrates of the other neuron, and the correlation coefficient be-

3.125 12.5 50 >=1000

5

10

15

Neurometric Threshold (% contrast)

Num

ber o

f Neu

rons

Fixation Detection

A B

C D

E

−100 −50 0 50 100

1

2

Time (ms)

Firin

g R

ate

(Nor

mal

ized

)

F

−100 −50 0 50 1001

2

3

Time (ms)

Firin

g R

ate

(Nor

mal

ized

) Detection Fixation

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ct Neurometric Function (Detection) Neurometric Function (Fixation)

0 1 10 100

0.5

0.6

0.7

0.8

0.9

1

Contrast (%)

Prop

ortio

n C

orre

ct

101 102

101

102

Threshold (Detection)

Thre

shol

d (F

ixat

ion)

P (Monkey 1)P (Monkey 2)M (Monkey 1)M (Monkey 2)

Fig. 4. Comparison of neurometric thresholds during detection and fixation. A and B: example neurometric functions of the same neurons during fixation (magentaline with squares) and detection (gray line with circles). C: neurometric thresholds during fixation and detection were significantly correlated with each other.Filled magenta symbols represent P neurons, and open green symbols represent M neurons: circles, monkey 1; inverted triangles, monkey 2; solid line, unity line;fixation threshold � detection threshold. D: neurometric thresholds during fixation (hatched bars) were on average lower than those during detection (solid bars).Black arrow indicates mean fixation threshold; gray arrow indicates mean detection threshold. E and F: population PSTHs for monkey 1 (E) and monkey 2 (F)measured at the same contrast (high contrast, 80–99%), showing that LGN neurons maintained their firing rate patterns across tasks. Solid gray line indicatesdetection and dashed magenta line indicates fixation; thick lines are mean response and thin lines are mean SE.

725LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

tween the two responses was taken as their interneuronalcorrelation. An unbiased estimator of the correlation coeffi-cient (r=) (Lehmann 1957; Olkin and Pratt 1958; see Eq. 2) wasreported for each neuron pair. In the detection task, noisecorrelations between pairs of LGN neurons were typicallyclose to 0 (see Fig. 5A for an example pair). During fixation,however, LGN neuron pairs often were significantly and pos-itively correlated (see Fig. 5B for an example pair). In our dataset the overall noise correlation during detection was notsignificantly different from 0 (average � 0.028 0.023, n �37 pairs, P � 0.423, t-test), whereas the noise correlationduring fixation was significantly different from 0 (average �0.175 0.025, n � 37 pairs, P � 0.001, t-test; Fig. 5C). Usingthe Pearson r instead of the unbiased estimator r= yieldedsimilar results (data not shown). Furthermore, the noise corre-lations in blank trials (i.e., 0% contrast, no stimulus present)still differed significantly between the detection (0.031 0.024) and the fixation (0.16 0.02) tasks (P � 0.000, t-test),indicating that the decrease in noise correlation observed in theactive perceptual detection task was most likely due to globalmodulations such as changes in the motivation or the alertnessstate of the animal (Fig. 5D). Finally, as shown above, thestability of fixation differed slightly between detection andfixation. To rule out the possibility that the noise correlationduring fixation was inflated by excessive fixational eye move-ments, we obtained eye position matched samples of interneu-

ronal correlations from the two tasks (detection: n � 36 of 37,average eye position deviation � 0.27 0.02°; fixation: n �26 of 37, average eye position deviation � 0.30 0.02°;P � 0.37, t-test). Within these samples, the interneuronalcorrelation distributions still differed significantly between thedetection (0.028 0.024) and the fixation (0.159 0.025)tasks (P � 0.000, t-test).

LGN Responses Were Correlated with the Monkeys’ ChoicesDuring Detection

We quantified the covariation between LGN responses andbehavioral choices as choice probability (Britten et al. 1996).When no stimulus was presented (i.e., 0% contrast trials only),overall choice probability was 0.54 for P neurons and 0.54 forM neurons, both above chance according to permutation (i.e.,bootstrapping) tests (P neurons: P � 0.015; M neurons: P �0.033). Next, we combined all low (�4%)-contrast trialswithin which choice probabilities did not significantly differfrom each other (F � 0.05, P � 0.995, 1-way ANOVA; Fig.6A) and derived a combined choice probability for each cell aswell as a grand choice probability for all cells. We found thatthe neural activities of LGN P and M neurons were bothsignificantly correlated with choice (P neurons: 0.53 0.01; Mneurons: 0.54 0.01; grand choice probability: 0.54; P � 0.05,permutation tests; Fig. 6B). Choice probability showed a small

−3 −2 −1 0 1 2 3−3

−2

−1

0

1

2

3

Firing Rate, Unit A (z−score)

Firin

g R

ate,

Uni

t B (z

−sco

re)

M072912AM072912Br = 0.004P > 0.1

−3 −2 −1 0 1 2 3

−3−2−10123

Firing Rate, Unit A (z−score)

Firin

g R

ate,

Uni

t B (z

−sco

re) Fixation Task

M021413AM021413Br = 0.309P < 0.001

−0.6 −0.4 −0.2 0 0.2 0.4 0.60

5

10

15

Interneuronal Correlation (unbiased r)

Num

ber o

f Neu

rons

Detection Fixation

A B

C D

−0.6 −0.4 −0.2 0 0.2 0.4 0.60

5

10

15

Interneuronal Correlation (unbiased r)

Num

ber o

f Neu

rons

Detection Fixation

Detection Task

Fig. 5. Interneuronal correlation. A and B: interneuronal correlations between example pairs of LGN neurons during detection (A) and fixation (B). Filled circlesindicate z-score normalized responses; solid line indicates linear regression fit. C: interneuronal correlation distributions during detection (solid bars) and fixation(hatched bars) significantly differed from each other. Solid line indicates correlation � 0; gray arrow indicates mean detection correlation; and black arrowindicates mean fixation correlation. D: the noise correlations computed using only blank trials (i.e., 0% contrast) still differed significantly between the detection(solid bars) and the fixation tasks (hatched bars). Line and arrows as in C.

726 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

negative correlation with neurometric threshold that was notstatistically significant (r � �0.12, P � 0.305).

The observed above-chance choice probabilities in the LGNwere not dependent on the monkeys’ behavioral bias in choos-ing the receptive field location over its mirror image location(r � 0.08, P � 0.495; Fig. 6C). Additionally, choice proba-bility was not correlated with the spontaneous response rate ofthe cell (r � �0.08, P � 0.495). We also characterized therelationship between eye movements and choice probabilityand found no significant correlation between the two (data notshown). Furthermore, choice probability was not correlatedwith reaction time (r � 0.03, P � 0.798), and using only the50% of trials with the fastest reaction times or the slowestreaction times did not yield significantly different choice prob-ability distributions (P � 0.854, permutation test; Fig. 6D).Finally, previous literature suggests that choice probability canbe significantly influenced by top-down, task-related gain mod-ulations (Nienborg and Cumming 2009). For our data wecharacterized task-related gain modulations of LGN neurons inseveral different ways by comparing the neural responsesduring the detection and fixation tasks, and we found suchmodulations could not explain the observed choice probabilityvalues either (data not shown). Similarly, choice probabilitywas not correlated with other factors that could have influenced

the monkeys’ motivation, including day-to-day changes inreward rate, overall reward value, or the monkeys’ psycho-physical performance (data not shown).

Neural Sensitivity, Choice Probability, and the Fano factor,But Not Interneuronal Correlation, Were Modulated By Time

We next asked how the above-characterized physiologicalproperties, including neural sensitivity, choice probability, andinterneuronal correlation, changed with time. We first charac-terized the changes in neurometric threshold using time inter-vals that started at stimulus onset (0 ms) and increased induration in steps of 12.5 ms (i.e., growing windows) and foundthat P (n � 37) and M (n � 20) neurons had significantlydifferent temporal dynamics (Fig. 7A). Specifically, comparedwith P neurons, M neurons exhibited more sensitive profiles(F � 7.51, P � 0.006, 2-way ANOVA, main effect for cellgroup) that developed faster [P: 62.5–200 ms thresholds devi-ated from 0 ms threshold; M: 50–200 ms thresholds deviatedfrom 0 ms threshold; P � 0.05, Tukey’s honestly significantdifference test (HSD)]. Next, computing neurometric thresh-olds in sliding time windows that were 50 ms wide andadvanced in steps of 12.5 ms, we confirmed that P and Mneurons had significantly different temporal dynamics (Fig.7B), with M neurons showing more transient developments in

0.3 0.4 0.5 0.6 0.70

5

10

15

20

25

Choice Probability

Num

ber o

f Neu

rons

P P (*) M M (*)

A B

0 0.8 1.2 2.4 3.6

0.45

0.5

0.55

0.6

Contrast (%)

Cho

ice

Prob

abili

ty

C D

0.3 0.4 0.5 0.6 0.70

5

10

15

20

25

Choice Probability

Num

ber o

f Neu

rons

Short RT Trials Short RT Trials (*) Long RT Trials Long RT Trials (*)

0.5 1 2 40.4

0.5

0.6

0.7

Choice Ratio (Inside RF/Outside RF)

Cho

ice

Prob

abili

ty

P (Monkey 1)P (Monkey 2)M (Monkey 1)M (Monkey 2)

Fig. 6. Choice probability. A: choice probability was not significantly influenced by contrast (�4% contrast). Solid line indicates choice probability � 0.5. Valuesare means SE (error bars). B: LGN neurons had significantly above-chance choice probabilities. Magenta bars represent P neurons, and green bars representM neurons; hatched bars represent P (magenta) and M neurons (green) with significant choice probabilities. Solid line indicates choice probability � 0.5; arrowsindicate mean choice probability. C: choice probability was not significantly correlated with the monkey’s behavioral bias in choosing the receptive field locationover its mirror image location. Filled magenta symbols represent P neurons, and open green symbols represent M neurons: circles, monkey 1; inverted triangles,monkey 2. D: recomputing choice probability using the 50% of trials with the fastest reaction time (short RT) or the slowest reaction time (long RT) did not yieldsignificantly different choice probability distributions. Blue bars represent choice probability computed with short RT trials, magenta bars represent choiceprobability computed with long RT trials; hatched bars represent choice probabilities computed with short RT (blue) and long RT trials (magenta) that weresignificantly different from 0.5. Solid line indicates choice probability � 0.5.

727LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

sensitivity that degraded faster in the later half of the stimuluspresentation (P: 62.5–200 ms thresholds deviated from 0 msthreshold; M: 62.5–75 ms thresholds deviated from 0 msthreshold; P � 0.05, Tukey’s HSD).

Using growing windows on choice probability, we deter-mined that compared with P neurons, M neurons also showedstronger covariations with choice (F � 11.09, P � 0.001,2-way ANOVA, main effect for cell group) that developed

faster in time (P: 100–200 ms choice probabilities deviatedfrom chance; M: 62.5–200 ms choice probabilities deviatedfrom chance; P � 0.05, permutation tests) (Fig. 7C). Similarly,in sliding windows, choice probabilities in the 75–112.5 msrange for P neurons and in the 62.5–125 ms range for Mneurons were significantly different from chance (P � 0.05,permutation tests; Fig. 7D). Z-normalized temporal changes inneurometric threshold were negatively correlated with the

0 50 100 150 200

0.48

0.5

0.52

0.54

Time (ms)

Cho

ice

Prob

abili

ty

0 50 100 150 200

40

50

60

70

80

90

100

Time (ms)

Neu

rom

etric

Thr

esho

ld

0 50 100 150 200

0.48

0.5

0.52

0.54

Time (ms)

Cho

ice

Prob

abili

ty

A B

0 50 100 150 200

40

50

60

70

80

90

100

Time (ms)

Neu

rom

etric

Thr

esho

ld

PM

C D

FE

0 50 100 150 2000.5

0.6

0.7

0.8

0.9

Time (ms)

Fano

Fac

tor

0 50 100 150 200

0

0.05

0.1

0.15

Time (ms)

Inte

rneu

rona

l Cor

rela

tion

Detection Fixation

Growing Window

Growing Window

Sliding Window

Sliding Window

Fig. 7. Temporal dynamics. A and B: the temporal dynamics of neurometric threshold computed using growing time windows (A) and sliding time windows (B).Solid magenta lines represent P neurons, dashed green lines represent M neurons; thick lines are mean threshold, and thin lines indicate mean SE. X-axisindicates when the time window ends (after stimulus onset). C and D: the temporal dynamics of choice probability computed using growing time windows (C)and sliding time windows (D). Solid magenta lines represent P neurons, dashed green lines represent M neurons; thick lines are mean choice probability, andthin lines indicate mean SE. Solid black line indicates choice probability � 0.5. X-axis indicates when the time window ends (after stimulus onset). E: thetemporal dynamics of Fano factor computed using sliding time windows. Solid magenta lines represent P neurons, dashed green lines represent M neurons; thicklines represent Fano factor estimation, and thin lines indicate the 95% confidence interval. F: the temporal dynamics of interneuronal correlation computed usingsliding time windows. Solid gray lines represent detection, and dashed magenta lines represent fixation; thick lines are mean correlation, and thin lines indicatemean SE. Solid black line indicates correlation � 0.

728 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

temporal changes in choice probability (in sliding windows, Pneurons: r � �0.84, P � 0.000; M neurons: r � �0.85, P �0.000), indicating that when the neuron was the most sensitive,it also correlated the most with the monkey’s choice. Sinceneural sensitivity developed earlier in time for M neurons,choice probability also evolved significantly earlier for Mneurons, capturing a subtle, very early difference in the tran-sient onset responses of the two cell types. Additionally, theFano factor (i.e., variance/mean firing rate) of LGN P and Mneurons also changed significantly with time in a similarfashion (sliding window, Fig. 7E), as previously reported in anumber of cortical areas (Churchland et al. 2010) as well as thethalamus (Victor et al. 2007). Specifically, for both M and Pcells, the Fano factor showed a sharp dip following stimulusonset (F � 14.10, P � 0.000, 2-way ANOVA, main effect fortime), and the temporal course of the Fano factor was alsosignificantly correlated with that of the neurometric threshold(P: r � 0.62, P � 0.008; M: r � 0.5, P � 0.041).

Finally, we characterized the temporal dynamics of interneu-ronal correlations in both the detection and the fixation tasks(sliding window, Fig. 7F). As described above, the averageinterneuronal correlation was found to be significantly differentbetween the detection and the fixation tasks (F � 205.1, P �0.000, 2-way ANOVA, main effect for task), and this differ-ence was persistent throughout time (F � 0.51, P � 0.943,2-way ANOVA, main effect for time).

Psychophysical Threshold and Choice Probability Could BeModeled with a Simple Pooling Scheme with No SignificantInterneuronal Correlations

The pooling model used in this study was similar to thosepreviously proposed to account for psychophysical perfor-mance and choice probability based on sensory neural re-sponses (Cohen and Newsome 2009; Haefner et al. 2013; Liuet al. 2013; Purushothaman and Bradley 2005; Shadlen et al.1996). Inputs to the model were single LGN P and M neuralresponses at different contrasts. Outputs were the simulatedpsychophysical threshold of the model and choice probabilitiesfor individual “neurons.” Variable parameters in the modelincluded 1) the number of P and M neurons (n), 2) the durationof the integration time window (t), 3) the Fano factor (f), 4) theinterneuronal noise correlation (r), 5) the downstream poolingnoise (p), and 6) the pooling strategy. In agreement withShadlen et al. (1996), our model behaved in a predictable waywhen one of these parameters was changed. For example,increasing the number of neurons in the pool decreased thepsychophysical threshold and choice probability (Fig. 8, A andB). Increasing interneuronal correlation increased threshold aswell as choice probability (Fig. 8, C and D). Increasing poolingnoise increased threshold and decreased choice probability(Fig. 8, E and F). For the simulation results reported in thisarticle, the Fano factor, interneuronal correlation, downstreampooling noise, and the pooling strategy were fixed according toeither measurements in our data set or previous literature (seeMATERIALS AND METHODS for a detailed description of the choiceof these parameters). Only the number of P and M neurons inthe pool (n) and the time window (t) were manipulated here toanswer a critical question: at different integration durations,how many P neurons and how many M neurons were likely tocontribute to the monkey’s perceptual decision?

To evaluate the fitness of the model, a GoF index (see Eq. 3)was reported for each parameter combination. The GoF index(ranging from 0 to 100%, with 100% indicating perfect fitting)reflected three factors equally: 1) how close the model thresh-old approached the measured psychophysical threshold, 2) howclose P choice probability in the model approached the mea-sured choice probability for P neurons, and 3) how close Mchoice probability in the model approached the measuredchoice probability for M neurons. Simulation results indicatedthat 1) at extremely short integration intervals (25 ms), incor-porating many neurons from both classes still failed to repro-duce the observed threshold and choice probabilities (Fig. 9A);2) at relatively brief intervals (50 ms), incorporating more Mneurons rather than P neurons in the pool could explain theobserved threshold as well as choice probabilities (Fig. 9B);and 3) at medium to long intervals (100–200 ms), a widerrange of M/P neuron combinations (1–128 neurons) yieldedgood model performance (Fig. 9, C–E). Finally, when infor-mation from only the second 100 ms of the stimulus presen-tation was integrated, it was beneficial to incorporate a largenumber of P neurons, presumably because M neural responseswere significantly degraded in the second half of the stimuluspresentation (Fig. 9F).

DISCUSSION

Our study is the first to examine the link between neuralactivity and perceptual decisions in the LGN of awake mon-keys. Using a 2AFC contrast detection task, we found that asmall percentage (13.5%) of LGN P and M cells were assensitive as the monkey, whereas the majority of LGN P and Mneurons were far less sensitive in detecting contrast (Fig. 2E).Second, having the monkeys perform a simple fixation taskinstead of active detection slightly improved the neural sensi-tivity (Fig. 4, C and D). Third, the average interneuronalcorrelation between LGN neuron pairs was not different from0 during active detection but was significantly above 0 duringsimple fixation (Fig. 5, C and D). Importantly, the covariationbetween neural responses and perceptual decisions, measuredas choice probability, was significant for both P and M neurons(Fig. 6B). The temporal dynamics of choice probability closelytracked those of neural sensitivity, reflecting subtle temporaldifferences between P and M neural responses (Fig. 7, A–D).Finally, modeling work suggested that both psychophysicalthreshold and LGN choice probability could be fully explainedwith a simple pooling model without assuming significantinterneuronal correlations or top-down gain modulations (Fig.9). Together, these results support the hypothesis that theperceptual decision pool for contrast detection consists ofmultiple LGN P and M neurons, and that response fluctuationsin these neurons can influence perception. In the discussionbelow we consider the implications of these results in light ofpreviously published findings.

The Neural Sensitivities of LGN P and M Cells

In our detection task, the psychophysical threshold of themonkey (5.76 0.72% contrast) was similar to previous mea-surements made under similar conditions (De Valois et al. 1974;Merigan 1980). Single-unit recordings in anesthetized, paralyzedanimals, in which visual stimuli are presented only to the receptivefield centers of single visual neurons (for example see Derrington

729LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

and Lennie 1984; Purpura et al. 1988; Shapley et al. 1981),showed that the contrast sensitivity of the most sensitive LGNneurons was comparable to the psychophysical sensitivity of theanimal. Intuitively, one might expect LGN neurons to be moresensitive in an active detection task compared with the anesthe-tized condition. In reality, we found that only a small percentage(13.5%) of LGN neurons had neurometric thresholds comparableto that of the monkey, whereas the majority (86.5%) of LGNneurons were much less sensitive (Fig. 2E).

There are several explanations for this observed differencein sensitivity between our study in awake, behaving monkeysand previous studies in anesthetized animals. First, to study thelink between LGN activity and perceptual choices, we con-structed visual stimuli (circular patches and Gaussian profiles)that were large enough to ensure stable, consistent stimulationof LGN neurons despite the constant fixational eye movementsof awake animals. Because of surround inhibition (Enroth-Cugell and Robson 1966; Rodieck 1965) and extraclassical

1 10 1001

10

100

Number of Neurons

Thre

shol

d (%

con

tras

t) P M

1 100.5

0.6

0.7

0.8

Threshold (%)

Cho

ice

Prob

abili

ty

P M

n = 200

n = 1

1 100.5

0.6

0.7

0.8

Threshold (%)

Cho

ice

Prob

abili

ty

0.00 0.05 0.10 0.20 0.30

1 100.5

0.6

0.7

0.8

Threshold (%)

Cho

ice

Prob

abili

ty

0.00 0.05 0.10 0.20 0.30

1 100.5

0.6

0.7

0.8

Threshold (%)

Cho

ice

Prob

abili

ty

0.0 0.5 1.0 2.0 4.0

1 100.5

0.6

0.7

0.8

Threshold (%)

Cho

ice

Prob

abili

ty

0.0 0.5 1.0 2.0 4.0

A B

C D

E F

Fig. 8. Parametric analysis of the pooling model. A and B: increasing the number of neurons in the pool decreased the simulated threshold and choice probability.Interneuronal correlation and pooling noise were assumed to be 0; integration time was 0–150 ms. Magenta lines represent simulated values for P neurons, andgreen lines represent simulated values for M neurons; thick gray lines indicate measured mean psychophysical threshold/choice probability, and thin grayline/rectangles indicate mean SE. C and D: increasing interneuronal correlation increased simulated threshold and choice probability for both P neurons (C)and M neurons (D). Pooling noise was assumed to be 0; integration time was 0–150 ms. As in B, each simulation line moved from the upper right corner tothe lower left corner as more neurons were added to the pool. Gray lines indicate measured mean psychophysical threshold/choice probability, and rectanglesindicate mean SE. E and F: increasing pooling noise increased simulated threshold and decreased choice probability for both P neurons (E) and M neurons(F). Interneuronal correlation was assumed to be 0.028 (measured value); integration time was 0–150 ms. As in B, each simulation line moved from the upperright corner to the lower left corner as more neurons were added to the pool. Lines are defined as in C and D.

730 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

surround inhibition (Alitto and Usrey 2008; Solomon et al.2006), large stimuli such as ours are not optimal for estimatingthe absolute sensitivity of LGN neurons (see also Kang andMalpeli 2009 for similar results in using low spatial frequencystimuli and measuring LGN sensitivity in passively fixatingcats). Second, it is important to note that the majority of thecommonly used anesthetics alter the normal LGN inhibitorycircuitry as well as other modulatory inputs to the LGN (Franks2008; Rudolph and Antkowiak 2004). These circuitries are, ofcourse, critical to understanding the function of the LGN(Casagrande et al. 2005; Saalmann and Kastner 2011; Sherman2007). For example, anesthetics such are propofol, isoflurane,halothane, or barbiturates interact strongly with GABAA re-ceptors (Franks 2006; Garcia et al. 2010; Krasowski andHarrison 1999), which may play an important role in regulatingthe transfer ratio of retinal input to LGN output (Casagrandeand Ichida 2011; Norton and Godwin 1992). Therefore, it isunclear to what extent the properties of LGN cells measuredunder anesthesia reflect the normal functional state of the LGNin awake, behaving animals.

Despite this difference in absolute sensitivity, our studyagreed with previous work in terms of the relative differencesin sensitivity and temporal dynamics (see below) between theLGN P and M populations (Derrington and Lennie 1984;Kaplan and Shapley 1982; Purpura et al. 1988; Sclar et al.1990; Shapley et al. 1981). Specifically, we found that Mneurons had, on average, slightly higher contrast sensitivityand contrast gain than P neurons (Fig. 2F), but the populationresponses of M and P neurons overlapped extensively (Kaplan2008, 2013; Norton et al. 1988; Xu et al. 2001).

Comparing Active Detection with Passive Fixation

We observed the nonintuitive result that LGN neurons weremore sensitive during the passive fixation task than the activedetection task (Fig. 4, C and D). Given the fact that detectingcontrast (especially at or near the psychophysical threshold)can be attentionally demanding compared with simple fixation,and that attention is known to boost the visually driven re-sponses of LGN neurons (Casagrande et al. 2005; McAlonan etal. 2008), this result seems counterintuitive. Eye movement

Fig. 9. Goodness-of-fit (GoF) index for different pool sizes(n) and integration time windows (t). A GoF of 100% (white)indicates that the model perfectly matches the observedpsychometric threshold as well as choice probabilities forboth P and M neurons. A: t � 0–25 ms. B: t � 0–50 ms. C:t � 0–100 ms. D: t � 0–150 ms. E: t � 0–200 ms. F: t �100–200 ms.

731LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

and other control analyses revealed that this difference insensitivity could not be simply explained by the instability ofthe monkeys’ fixation or our physiological recordings. Wethink one possible explanation is that pronounced, long-lastingpresaccadic suppressions (Reppas et al. 2002; Royal et al.2006) occurred only during the detection task but not duringthe passive fixation task, as the monkeys were only required tomake saccades in the former case.

Another significant difference we observed between thedetection and the fixation tasks was in terms of interneuronalcorrelations between LGN neuron pairs (Fig. 5). Specifically,during passive fixation, LGN interneuronal correlations weresignificantly above 0, much like what has been reported in theretina (DeVries 1999; Mastronarde 1983, 1989; Meister et al.1995) as well as in a number of cortical areas (see below).During active detection, however, we observed that LGNinterneuronal correlations were not significantly different from0, in agreement with recent data obtained from anesthetizedmarmoset LGN (Cheong et al. 2011) as well as from largepopulations of V1 neurons in awake monkeys (Ecker et al.2010, 2014). It remains controversial why the average in-terneuronal correlation values vary greatly from study to study(Cohen and Kohn 2011), and the small sample sizes limit thetypes of conclusions that we can draw based on our data. Webelieve, however, that the relative decrease in interneuronalcorrelation we observed when the monkeys were engaged inactive detection could be attributed to the fact that bothattention and perceptual learning de-correlate pairs of singleneurons (Cohen and Maunsell 2010; Gu et al. 2011; Mitchell etal. 2009). Another possibility is that during the fixation task, itmight be more difficult for the monkeys to maintain longfixations or suppress the impulse to saccade to a very salienttarget. Therefore, the frequency of microsaccades might in-crease slightly during the fixation task, which in turn couldenhance neural sensitivities and raise interneuronal correlationsat the same time (Collewijn and Kowler 2008; Martinez-Condeet al. 2002, 2004, 2013; Rolfs 2009).

Parallel Pathways in Fast Detection

We reported in our temporal analyses that the average LGNneuron achieved its peak sensitivity and choice probabilitywithin 100 ms after stimulus onset, with M neurons beingslightly faster than P neurons in almost all cases (Fig. 7, A–D).These results are well within expectation given the nature ofour task and the known differences in response dynamicsbetween LGN P and M neurons. In many visual tasks, thepsychophysical performance of human and monkey subjectsplateaus within 100–300 ms after stimulus onset (Angelaki andCullen 2008; Nachmias 1967; Purushothaman and Bradley2005; Snowden and Braddick 1991; Tulunay-Keesey and Jones1976). Physiological recordings also have revealed great sin-gle-neuron sensitivities based solely on the early, transientcomponents of visually driven responses (Buracas et al. 1998;Ghose and Harrison 2009; Müller et al. 2001; Oram and Perrett1992; Price and Born 2010; Tovee et al. 1993). This transientresponse has been reported to have higher contrast gain, higherresponse gain, and lower variability (Müller et al. 2001) and tocontain �80% of all the information available throughout thestimulus presentation (Osborne et al. 2004; Tovee et al. 1993).Under natural viewing conditions, humans and monkeys sac-

cade once every �300 ms (Wilson and Goldman-Rakic 1994).Thus, for a simple detection task such as ours, it may be morebeneficial for the visual system to take in all the informationavailable “at a glance” and then quickly move on (Ghose andHarrison 2009; Osborne et al. 2004; Uchida et al. 2006). In thisscenario, the distinctive temporal dynamics of LGN P and Mneurons (Schiller and Malpeli 1978; Schmolesky et al. 1998)determine how they are differentially pooled together to per-form the task (Merigan and Maunsell 1993; Stone 1983).

Choice Probability in the LGN and Its Implications

Moderately above-chance choice probability values (0.5–0.6) have been found in numerous cortical areas during avariety of behavioral tasks, regardless of the sensitivity ofsingle neurons (Bosking and Maunsell 2011; Britten et al.1996; Cook and Maunsell 2002; Dodd et al. 2001; Grunewaldet al. 2002; Liu and Newsome 2005; Nienborg and Cumming2006; Palmer et al. 2007; Parker et al. 2002; Price and Born2010; Purushothaman and Bradley 2005; Uka and DeAngelis2004; Uka et al. 2005). Similarly, we found choice probabilityin the LGN (0.54) to be comparable in its magnitude to choiceprobabilities in the cortex, despite the relative insensitivity ofmost single LGN neurons (Fig. 6B).

Compared with the cortex, the LGN is closer to the sensorysurface and is thus expected to more faithfully reflect thephysical reality of the outside world. It is therefore surprisingthat in the absence of a physical stimulus, fluctuations in thespontaneous responses of LGN neurons were still significantlycorrelated with the monkey’s choices. The theoretical implica-tion of this above-chance choice probability is twofold. First,previous modeling work indicated that choice probability val-ues are heavily influenced by interneuronal correlations (Nien-borg et al. 2012; Nienborg and Cumming 2010; Shadlen et al.1996). Thus far, almost all cortical areas where choice proba-bilities have been measured also exhibit ubiquitous interneu-ronal correlations (Averbeck et al. 2006; Cohen and Kohn2011). In contrast, we found that LGN P and M neuronsexhibited above-chance choice probabilities during contrastdetection, even though the average interneuronal correlationwas not significantly different from 0. In agreement with ourresults, recordings in the subcortical vestibular nucleus (VTN)during a heading direction discrimination task (Liu et al. 2013)also observed above-chance choice probabilities in the absenceof positive noise correlations. Importantly, recent modelingresults reveal that it is not the average interneuronal correlationlevel but the structure of a specific type of differential corre-lation that determines choice probability values in neural poolsof different sizes (Haefner et al. 2013; Moreno-Bote et al.2014). Our experimental and modeling results are consistentwith this hypothesis. Additionally, it is known that above-chance choice probabilities can arise from both feedforwardconnections (i.e., response fluctuations in sensory neurons) andfeedback connections, such as top-down attentional modula-tion (Cohen and Maunsell 2010; Nienborg and Cumming2009). Our results support a bottom-up rather than top-downorigin of choice probability because 1) we found no significantgain modulations in the attentionally demanding detection task,and 2) in our models we were able to successfully simulatethreshold and choice probability results using a very shortintegration window (i.e., 50 ms). Taken together, our data and

732 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

the success of our model demonstrate that, at least for periph-eral sensory structures such as the LGN, random responsefluctuations in single neurons can be linked to changes in theperceptual behavior of the subject.

ACKNOWLEDGMENTS

We are grateful to our veterinarian Troy Apple, veterinary technician MaryFeurtado, and the staff of Division of Animal Care at Vanderbilt University forproviding excellent support and care for our animals. We appreciate Dr. JeffreySchall’s advice on experimental design and his willingness to share labequipment and programs. We thank Drs. Frank Tong and Alexander Maier aswell as three anonymous reviewers for taking time to help us improve themanuscript. Finally, we thank Julia Mavity-Hudson and Brandon Moore forcontinuous help with many aspects of this project.

GRANTS

This work was supported by National Institutes of Health (NIH) GrantsEY001778 and R21 EY019132 (to V. A. Casagrande), NIH Core GrantsEY008126 and HD15052, and funds from the Department of Cell and Devel-opmental Biology at Vanderbilt University.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Y.J., G.P., and V.A.C. conception and design of research; Y.J. performedexperiments; Y.J., D.Y., and G.P. analyzed data; Y.J., G.P., and V.A.C.interpreted results of experiments; Y.J. prepared figures; Y.J. drafted manu-script; Y.J., D.Y., G.P., and V.A.C. edited and revised manuscript; Y.J., D.Y.,G.P., and V.A.C. approved final version of manuscript.

REFERENCES

Alitto HJ, Rathbun DL, Martin Usrey W. A comparison of visual responsesin the lateral geniculate nucleus of alert and anaesthetized macaque mon-keys. J Physiol 589: 87–99, 2011.

Alitto HJ, Usrey WM. Origin and dynamics of extraclassical suppression inthe lateral geniculate nucleus of the macaque monkey. Neuron 57: 135–146,2008.

Angelaki DE, Cullen KE. Vestibular system: the many facets of a multimodalsense. Annu Rev Neurosci 31: 125–150, 2008.

Averbeck BB, Latham PE, Pouget A. Neural correlations, population codingand computation. Nat Rev Neurosci 7: 358–366, 2006.

Barlow HB, Levick WR, Yoon M. Responses to single quanta of light inretinal ganglion cells of the cat. Vision Res Suppl 3: 87–101, 1971.

Berry MJ, Warland DK, Meister M. The structure and precision of retinalspike trains. Proc Natl Acad Sci USA 94: 5411–5416, 1997.

Bosking WH, Maunsell JH. Effects of stimulus direction on the correlationbetween behavior and single units in area MT during a motion detectiontask. J Neurosci 31: 8230–8238, 2011.

Briggs F, Usrey WM. Corticogeniculate feedback and visual processing in theprimate. J Physiol 589: 33–40, 2011.

Britten KH, Newsome WT, Shadlen MN, Celebrini S, Movshon JA. Arelationship between behavioral choice and the visual responses of neuronsin macaque MT. Vis Neurosci 13: 87–100, 1996.

Britten KH, Shadlen MN, Newsome WT, Movshon JA. The analysis ofvisual motion: a comparison of neuronal and psychophysical performance.J Neurosci 12: 4745–4765, 1992.

Buracas GT, Zador AM, DeWeese MR, Albright TD. Efficient discrimina-tion of temporal patterns by motion-sensitive neurons in primate visualcortex. Neuron 20: 959–969, 1998.

Casagrande VA, Ichida JM. Processing in the lateral geniculate nucleus(LGN). In: Adler’s Physiology of the Eye (11th ed.), edited by Levin LA,Nilsson SFE, Ver Hoeve J, Wu S, Kaufman PL, Alm A. Amsterdam, TheNetherlands: Elsevier, 2011, p. 574–585.

Casagrande VA, Norton TT. The lateral geniculate nucleus: a review of itsphysiology and function. In: The Neural Basis of Visual Function, edited byLeventhal AG. London: MacMillan, 1991, p. 41–84.

Casagrande VA, Sary G, Royal D, Ruiz O. On the impact of attention andmotor planning on the lateral geniculate nucleus. Prog Brain Res 149:11–29, 2005.

Casagrande VA, Xu X. Parallel visual pathways: a comparative perspective.In: The Visual Neurosciences, edited by Chalupa L, Werner JS. Cambridge,MA: MIT Press, 2004, p. 494–506.

Cheong SK, Tailby C, Martin PR, Levitt JB, Solomon SG. Slow intrinsicrhythm in the koniocellular visual pathway. Proc Natl Acad Sci USA 108:14659–14663, 2011.

Churchland MM, Byron MY, Cunningham JP, Sugrue LP, Cohen MR,Corrado GS, Newsome WT, Clark AM, Hosseini P, Scott BB, BradleyDC, Smith MA, Kohn A, Movshon JA, Armstrong KA, Moore T, ChangSW, Snyder LH, Lisberger SG, Priebe NJ, Finn IM, Ferster D, Ryu SI,Santhanam G, Sahani M, Shenoy KV. Stimulus onset quenches neuralvariability: a widespread cortical phenomenon. Nat Neurosci 13: 369–378,2010.

Cohen MR, Kohn A. Measuring and interpreting neuronal correlations. NatNeurosci 14: 811–819, 2011.

Cohen MR, Maunsell JH. A neuronal population measure of attentionpredicts behavioral performance on individual trials. J Neurosci 30: 15241–15253, 2010.

Cohen MR, Newsome WT. Estimates of the contribution of single neurons toperception depend on timescale and noise correlation. J Neurosci 29:6635–6648, 2009.

Collewijn H, Kowler E. The significance of microsaccades for vision andoculomotor control. J Vis 8: 20, 2008.

Cook EP, Maunsell JH. Attentional modulation of behavioral performanceand neuronal responses in middle temporal and ventral intraparietal areas ofmacaque monkey. J Neurosci 22: 1994–2004, 2002.

De Valois RL, Morgan H, Snodderly DM. Psychophysical studies of monkeyvision. 3. Spatial luminance contrast sensitivity tests of macaque and humanobservers. Vision Res 14: 75–81, 1974.

Derrington AM, Lennie P. Spatial and temporal contrast sensitivities ofneurones in lateral geniculate nucleus of macaque. J Physiol 357: 219–240,1984.

DeVries SH. Correlated firing in rabbit retinal ganglion cells. J Neurophysiol81: 908–920, 1999.

Dodd JV, Krug K, Cumming BG, Parker AJ. Perceptually bistable three-dimensional figures evoke high choice probabilities in cortical area MT. JNeurosci 21: 4809–4821, 2001.

Ecker AS, Berens P, Cotton RJ, Subramaniyan M, Denfield GH, CadwellCR, Smirnakis SM, Bethge M, Tolias AS. State dependence of noisecorrelations in macaque primary visual cortex. Neuron 82: 235–248, 2014.

Ecker AS, Berens P, Keliris GA, Bethge M, Logothetis NK, Tolias AS.Decorrelated neuronal firing in cortical microcircuits. Science 327: 584–587, 2010.

Enroth-Cugell C, Robson JG. The contrast sensitivity of retinal ganglioncells of the cat. J Physiol 187: 517–552, 1966.

Erwin E, Baker FH, Busen WF, Malpeli JG. Relationship between laminartopology and retinotopy in the rhesus lateral geniculate nucleus: results froma functional atlas. J Comp Neurol 407: 92–102, 1999.

Franks NP. General anaesthesia: from molecular targets to neuronal pathwaysof sleep and arousal. Nat Rev Neurosci 9: 370–386, 2008.

Franks NP. Molecular targets underlying general anaesthesia. Br J Pharmacol147: S72–S81, 2006.

Garcia PS, Kolesky SE, Jenkins A. General anesthetic actions on GABAA

receptors. Curr Neuropharmacol 8: 2, 2010.Ghose GM, Harrison IT. Temporal precision of neuronal information in a

rapid perceptual judgment. J Neurophysiol 101: 1480–1493, 2009.Green DM, Swets JA. Signal Detection Theory and Psychophysics. New

York: Wiley, 1966.Grunewald A, Bradley DC, Andersen RA. Neural correlates of structure-

from-motion perception in macaque V1 and MT. J Neurosci 22: 6195–6207,2002.

Gu Y, Liu S, Fetsch CR, Yang Y, Fok S, Sunkara A, DeAngelis GC,Angelaki DE. Perceptual learning reduces interneuronal correlations inmacaque visual cortex. Neuron 71: 750–761, 2011.

Haefner RM, Gerwinn S, Macke JH, Bethge M. Inferring decoding strate-gies from choice probabilities in the presence of correlated variability. NatNeurosci 16: 235–242, 2013.

Hubel DH, Wiesel TN. Integrative action in the cat’s lateral geniculate body.J Physiol 155: 385–398, 1961.

733LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

Ichida JM, Casagrande VA. Organization of the feedback pathway fromstriate cortex (V1) to the lateral geniculate nucleus (LGN) in the owlmonkey (Aotus trivirgatus). J Comp Neurol 454: 272–283, 2002.

Ichida JM, Mavity-Hudson JA, Casagrande VA. Distinct patterns of corti-cogeniculate feedback to different layers of the lateral geniculate nucleus(LGN). Eye Brain 6: 57–73, 2014.

Irvin GE, Casagrande VA, Norton TT. Center/surround relationships ofmagnocellular, parvocellular, and koniocellular relay cells in primate lateralgeniculate nucleus. Vis Neurosci 10: 363–373, 1993.

Irvin GE, Norton TT, Sesma MA, Casagrande VA. W-like responseproperties of interlaminar zone cells in the lateral geniculate nucleus of aprimate (Galago crassicaudatus). Brain Res 362: 254–270, 1986.

Jiang Y, Yampolsky D, Purushothaman G, Casagrande V. Neural sensi-tivity in the lateral geniculate nucleus (LGN) of awake, behaving monkeysduring a contrast detection task: comparison of neurometric and psychomet-ric functions (Abstract). J Vis 13: 1026, 2013.

Jiang Y, Yampolsky D, Purushothaman G, Casagrande VA. Neural re-sponse in the lateral geniculate nucleus (LGN) of awake, behaving monkeysduring a contrast detection task: comparison of psychometric and neuromet-ric functions. Program 175.115. 2012 Neuroscience Meeting Planner. NewOrleans, LA: Society for Neuroscience, 2012. Online.

Jones EG. The Thalamus. Cambridge, UK: Cambridge University Press, 2007.Kang I, Malpeli JG. Dim-light sensitivity of cells in the awake cat’s lateral

geniculate and medial interlaminar nuclei: a correlation with behavior. JNeurophysiol 102: 841–852, 2009.

Kang I, Maunsell JH. Potential confounds in estimating trial-to-trial corre-lations between neuronal response and behavior using choice probabilities.J Neurophysiol 108: 3403–3415, 2012.

Kaplan E. The M, P and K pathways of the primate visual system revisited.In: The New Visual Neurosciences. Cambridge, MA: The MIT Press, 2013,p. 215–226.

Kaplan E. The P, M and K streams of the primate visual system: what do theydo for vision. In: The Senses: A Comprehensive Reference Vision. Amster-dam, The Netherlands: Elsevier, 2008, p. 369–381.

Kaplan E, Shapley RM. The primate retina contains two types of ganglioncells, with high and low contrast sensitivity. Proc Natl Acad Sci USA 83:2755–2757, 1986.

Kaplan E, Shapley RM. X and y cells in the lateral geniculate nucleus ofmacaque monkeys. J Physiol 330: 125–143, 1982.

Kara P, Reinagel P, Reid RC. Low response variability in simultaneouslyrecorded retinal, thalamic, and cortical neurons. Neuron 27: 635–646, 2000.

Krasowski M, Harrison N. General anaesthetic actions on ligand-gated ionchannels. Cell Mol Life Sci 55: 1278–1303, 1999.

Lehmann E. Theory of Point Estimation. Pacific Grove, CA: Wadsworth andBrooks/Cole, 1957.

Levine MW, Cleland BG, Zimmerman RP. Variability of responses of catretinal ganglion cells. Vis Neurosci 8: 277–279, 1992.

Liu J, Newsome WT. Correlation between speed perception and neuralactivity in the middle temporal visual area. J Neurosci 25: 711–722, 2005.

Liu S, Gu Y, DeAngelis GC, Angelaki DE. Choice-related activity andcorrelated noise in subcortical vestibular neurons. Nat Neurosci 16: 89–97,2013.

Logothetis NK. Single units and conscious vision. Philos Trans R Soc Lond BBiol Sci 353: 1801–1818, 1998.

Martinez-Conde S, Macknik SL, Hubel DH. The function of bursts of spikesduring visual fixation in the awake primate lateral geniculate nucleus andprimary visual cortex. Proc Natl Acad Sci USA 99: 13920–13925, 2002.

Martinez-Conde S, Macknik SL, Hubel DH. The role of fixational eyemovements in visual perception. Nat Rev Neurosci 5: 229–240, 2004.

Martinez-Conde S, Otero-Millan J, Macknik SL. The impact of microsac-cades on vision: towards a unified theory of saccadic function. Nat RevNeurosci 14: 83–96, 2013.

Mastronarde DN. Correlated firing of cat retinal ganglion cells. I. Spontane-ously active inputs to X- and Y-cells. J Neurophysiol 49: 303–324, 1983.

Mastronarde DN. Correlated firing of retinal ganglion cells. Trends Neurosci12: 75–80, 1989.

McAdams CJ, Maunsell JH. Effects of attention on the reliability of indi-vidual neurons in monkey visual cortex. Neuron 23: 765–773, 1999.

McAlonan K, Cavanaugh J, Wurtz RH. Guarding the gateway to cortex withattention in visual thalamus. Nature 456: 391–394, 2008.

Meister M, Lagnado L, Baylor DA. Concerted signaling by retinal ganglioncells. Science 270: 1207–1210, 1995.

Merigan W, Katz LM, Maunsell J. The effects of parvocellular lateralgeniculate lesions on the acuity and contrast sensitivity of macaque mon-keys. J Neurosci 11: 994–1001, 1991.

Merigan WH. Temporal modulation sensitivity of macaque monkeys. VisionRes 20: 953–959, 1980.

Merigan WH, Maunsell JH. Macaque vision after magnocellular lateralgeniculate lesions. Vis Neurosci 5: 347–352, 1990.

Merigan WH, Maunsell JH. How parallel are the primate visual pathways?Annu Rev Neurosci 16: 369–402, 1993.

Mitchell JF, Sundberg KA, Reynolds JH. Spatial attention decorrelatesintrinsic activity fluctuations in macaque area V4. Neuron 63: 879–888,2009.

Moreno-Bote R, Beck J, Kanitscheider I, Pitkow X, Latham P, Pouget A.Information-limiting correlations. Nat Neurosci 17: 1410–1417, 2014.

Müller JR, Metha AB, Krauskopf J, Lennie P. Information conveyed byonset transients in responses of striate cortical neurons. J Neurosci 21:6978–6990, 2001.

Nachmias J. Effect of exposure duration on visual contrast sensitivity withsquare-wave gratings. J Opt Soc Am 57: 421–427, 1967.

Nassi JJ, Callaway EM. Parallel processing strategies of the primate visualsystem. Nat Rev Neurosci 10: 360–372, 2009.

Nienborg H, Cohen MR, Cumming BG. Decision-related activity in sensoryneurons: correlations among neurons and with behavior. Annu Rev Neurosci35: 463–483, 2012.

Nienborg H, Cumming B. Correlations between the activity of sensoryneurons and behavior: how much do they tell us about a neuron’s causality?Curr Opin Neurobiol 20: 376–381, 2010.

Nienborg H, Cumming BG. Decision-related activity in sensory neuronsreflects more than a neuron’s causal effect. Nature 459: 89–92, 2009.

Nienborg H, Cumming BG. Macaque V2 neurons, but not V1 neurons, showchoice-related activity. J Neurosci 26: 9567–9578, 2006.

Norton TT, Casagrande VA. Laminar organization of receptive-field prop-erties in lateral geniculate nucleus of bush baby (Galago crassicaudatus). JNeurophysiol 47: 715–741, 1982.

Norton TT, Casagrande VA, Irvin GE, Sesma MA, Petry HM. Contrast-sensitivity functions of W-, X-, and Y-like relay cells in the lateral genic-ulate nucleus of bush baby, Galago crassicaudatus. J Neurophysiol 59:1639–1656, 1988.

Norton TT, Godwin DW. Inhibitory GABAergic control of visual signals atthe lateral geniculate nucleus. Prog Brain Res 90: 193–217, 1992.

Olkin I, Pratt JW. Unbiased estimation of certain correlation coefficients.Ann Math Stat 201–211, 1958.

Oram M, Perrett D. Time course of neural responses discriminating differentviews of the face and head. J Neurophysiol 68: 70–84, 1992.

Oram M, Wiener M, Lestienne R, Richmond B. Stochastic nature ofprecisely timed spike patterns in visual system neuronal responses. JNeurophysiol 81: 3021–3033, 1999.

Osborne LC, Bialek W, Lisberger SG. Time course of information aboutmotion direction in visual area MT of macaque monkeys. J Neurosci 24:3210–3222, 2004.

Palmer C, Cheng SY, Seidemann E. Linking neuronal and behavioralperformance in a reaction-time visual detection task. J Neurosci 27: 8122–8137, 2007.

Parker AJ, Krug K, Cumming BG. Neuronal activity and its links with theperception of multi-stable figures. Philos Trans R Soc Lond B Biol Sci 357:1053–1062, 2002.

Parker AJ, Newsome WT. Sense and the single neuron: probing the physi-ology of perception. Annu Rev Neurosci 21: 227–277, 1998.

Paxinos G, Huang XF, Toga AW. The Rhesus Monkey Brain in StereotaxicCoordinates. Salt Lake City, UT: Academic, 1999.

Price NS, Born RT. Timescales of sensory-and decision-related activity in themiddle temporal and medial superior temporal areas. J Neurosci 30: 14036–14045, 2010.

Purpura K, Kaplan E, Shapley RM. Background light and the contrast gainof primate P and M retinal ganglion cells. Proc Natl Acad Sci USA 85:4534–4537, 1988.

Purushothaman G, Bradley DC. Neural population code for fine perceptualdecisions in area MT. Nat Neurosci 8: 99–106, 2005.

Ramcharan EJ, Gnadt JW, Sherman SM. The effects of saccadic eyemovements on the activity of geniculate relay neurons in the monkey. VisNeurosci 18: 253–258, 2001.

Reich DS, Victor JD, Knight BW, Ozaki T, Kaplan E. Response variabilityand timing precision of neuronal spike trains in vivo. J Neurophysiol 77:2836–2841, 1997.

734 LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org

Reppas JB, Usrey WM, Reid RC. Saccadic eye movements modulate visualresponses in the lateral geniculate nucleus. Neuron 35: 961–974, 2002.

Rodieck RW. Quantitative analysis of cat retinal ganglion cell response tovisual stimuli. Vision Res 5: 583–601, 1965.

Rolfs M. Microsaccades: small steps on a long way. Vision Res 49: 2415–2441, 2009.

Royal DW, Sary G, Schall JD, Casagrande VA. Correlates of motorplanning and postsaccadic fixation in the macaque monkey lateral geniculatenucleus. Exp Brain Res 168: 62–75, 2006.

Rudolph U, Antkowiak B. Molecular and neuronal substrates for generalanaesthetics. Nat Rev Neurosci 5: 709–720, 2004.

Ruiz O, Royal D, Sary G, Chen X, Schall JD, Casagrande VA. Low-threshold Ca2-associated bursts are rare events in the LGN of the awakebehaving monkey. J Neurophysiol 95: 3401–3413, 2006.

Saalmann YB, Kastner S. Cognitive and perceptual functions of the visualthalamus. Neuron 71: 209–223, 2011.

Saalmann YB, Kastner S. Gain control in the visual thalamus duringperception and cognition. Curr Opin Neurobiol 19: 408–414, 2009.

Schiller PH, Logothetis NK, Charles ER. Role of the color-opponent andbroad-band channels in vision. Vis Neurosci 5: 321–346, 1990.

Schiller PH, Malpeli JG. Functional specificity of lateral geniculate nucleuslaminae of the rhesus monkey. J Neurophysiol 41: 788–797, 1978.

Schmid MC, Mrowka SW, Turchi J, Saunders RC, Wilke M, Peters AJ,Ye FQ, Leopold DA. Blindsight depends on the lateral geniculate nucleus.Nature 466: 373–377, 2010.

Schmolesky MT, Wang Y, Hanes DP, Thompson KG, Leutgeb S, SchallJD, Leventhal AG. Signal timing across the macaque visual system. JNeurophysiol 79: 3272–3278, 1998.

Sclar G, Maunsell JH, Lennie P. Coding of image contrast in central visualpathways of the macaque monkey. Vision Res 30: 1–10, 1990.

Shadlen MN, Britten KH, Newsome WT, Movshon JA. A computationalanalysis of the relationship between neuronal and behavioral responses tovisual motion. J Neurosci 16: 1486–1510, 1996.

Shapley R. Visual sensitivity and parallel retinocortical channels. Annu RevPsychol 41: 635–658, 1990.

Shapley R, Kaplan E, Soodak R. Spatial summation and contrast sensitivityof X and Y cells in the lateral geniculate nucleus of the macaque. Nature292: 543–545, 1981.

Sherman SM. The thalamus is more than just a relay. Curr Opin Neurobiol17: 417–422, 2007.

Sherman SM, Guillery RW. Exploring the Thalamus. San Diego, CA:Academic, 2001.

Snowden RJ, Braddick OJ. The temporal integration and resolution ofvelocity signals. Vision Res 31: 907–914, 1991.

Solomon SG, Lee BB, Sun H. Suppressive surrounds and contrast gain inmagnocellular-pathway retinal ganglion cells of macaque. J Neurosci 26:8715–8726, 2006.

Stone J. Parallel Processing in the Visual System. New York, NY: Plenum,1983.

Tolhurst DJ, Movshon J, Dean A. The statistical reliability of signals insingle neurons in cat and monkey visual cortex. Vision Res 23: 775–785,1983.

Tovee M, Rolls E, Treves A, Bellis R. Information encoding and theresponses of single neurons in the primate temporal visual cortex. J Neu-rophysiol 70: 640, 1993.

Tulunay-Keesey U, Jones RM. The effect of micromovements of the eye andexposure duration on contrast sensitivity. Vision Res 16: 481–488, 1976.

Uchida N, Kepecs A, Mainen ZF. Seeing at a glance, smelling in a whiff:rapid forms of perceptual decision making. Nat Rev Neurosci 7: 485–491,2006.

Uka T, DeAngelis GC. Contribution of area MT to stereoscopic depthperception: choice-related response modulations reflect task strategy. Neu-ron 42: 297–310, 2004.

Uka T, Tanabe S, Watanabe M, Fujita I. Neural correlates of fine depthdiscrimination in monkey inferior temporal cortex. J Neurosci 25: 10796–10802, 2005.

Victor J, Blessing E, Forte J, Buzás P, Martin P. Response variability ofmarmoset parvocellular neurons. J Physiol 579: 29–51, 2007.

White AJ, Solomon SG, Martin PR. Spatial properties of koniocellular cellsin the lateral geniculate nucleus of the marmoset Callithrix jacchus. JPhysiol 533: 519–535, 2001.

Wilson FA, Goldman-Rakic PS. Viewing preferences of rhesus monkeysrelated to memory for complex pictures, colours and faces. Behav Brain Res60: 79–89, 1994.

Xu X, Ichida JM, Allison JD, Boyd JD, Bonds AB, Casagrande VA. Acomparison of koniocellular, magnocellular and parvocellular receptive fieldproperties in the lateral geniculate nucleus of the owl monkey (Aotustrivirgatus). J Physiol 531: 203–218, 2001.

735LGN IN CONTRAST DETECTION

J Neurophysiol • doi:10.1152/jn.00068.2015 • www.jn.org