A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions

A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions View ORCID Profile Alice Tsvinev , View ORCID Profile April Pilipenko , View ORCID Profile Jason Samaha doi: https://doi.org/10.1101/2025.11.13.688140 Alice Tsvinev 1 University of California, Santa Cruz. Department of Psychology Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Alice Tsvinev For correspondence: atsvinev{at}ucsc.edu jsamaha{at}ucsc.edu April Pilipenko 1 University of California, Santa Cruz. Department of Psychology Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for April Pilipenko Jason Samaha 1 University of California, Santa Cruz. Department of Psychology Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jason Samaha Abstract Full Text Info/History Metrics Preview PDF Abstract Humans frequently make decisions based on sensory input from the external environment or information retrieved from memory. The centro-parietal positivity (CPP), an event-related EEG potential, has recently been identified as a neural correlate of sensory evidence accumulation during perceptual decision-making tasks. However, it remains unclear whether this component also reflects evidence accumulation during decisions based on long-term memory. The present study investigated whether the CPP serves as a domain-general signal of evidence accumulation across both perception and semantic memory-based decisions. We designed a visual discrimination task in which participants discriminated between the luminance of two alphanumeric strings across three levels of difficulty defined by the luminance difference. In a semantic memory task, participants discriminated between the populations of two different U.S. states, binned into three difficulty levels using US census data. After each decision, participants rated their confidence on a scale from 1 (very low) to 4 (very high). We observed a CPP component in both tasks, the slope of which scaled with the sensory or mnemonic evidence in both stimulus- and response-locked analyses. Furthermore, these CPP slopes were sensitive to reaction times (RT) and confidence in both tasks. Finally, the peak height of the CPP just prior to the response, indicative of one’s decision boundary, was strongly correlated across tasks, suggesting that a common threshold was applied to the abstract evidence quantity being accumulated. Our results indicate that the CPP can be used to track the unfolding dynamics underlying decisions made both on the basis of external sensory evidence and internal mnemonic evidence. Introduction Humans frequently make decisions based on sampling sensory evidence from our external environment but also by internally sampling and retrieving information held in memory ( O’Connell et al., 2012 ; Shadlen & Shohamy, 2016 ; Van Ede & Nobre, 2024 ). In recent decades, cognitive neuroscience has made significant strides in understanding how the brain supports this process, revealing that many decisions are made dynamically as information accumulates. One influential framework proposes that binary decisions are made by continuously accumulating evidence until a threshold is reached and a choice is made. Computational models based on this principle have successfully explained variation across a range of decision behaviors including response times (RT), difficulty, and confidence ( Desender et al., 2021 ; Kiani et al., 2014 ; O’Connell et al., 2018 ; Ratcliff et al., 2016 ; Ratcliff & McKoon, 2008 ). Importantly, recent work has uncovered an electroencephalogram (EEG) potential that closely tracks the sensory evidence accumulation process underlying simple perceptual decisions ( O’Connell et al., 2012 ). The centro-parietal positivity (CPP) has been identified as an event-related potential (ERP) signature of evidence accumulation, closely aligning with the temporal dynamics of sensory evidence accumulation ( O’Connell et al., 2012 ). The CPP reflects evidence accumulation as a gradual buildup of positive voltage that scales with the strength of the absolute sensory evidence, peaks around the time of behavioral response, and can predict decision accuracy, RT, and confidence in simple perceptual decisions ( Charles et al., 2020 ; Dou et al., 2024 ; Herding et al., 2019 ; Kelly & O’Connell, 2013 ; O’Connell et al., 2012 ). Furthermore, it has been argued that the CPP reflects a supramodal decision signal, occurring with a similar topography and temporal dynamic regardless of the sensory modality (e.g. auditory, visual) or response format (e.g., button presses or covert responses such as mental counting), suggesting a domain-general decision-making process ( O’Connell et al., 2012 ; Schaffhauser et al., 2021 ). Despite the substantial literature linking the CPP to evidence accumulation, most studies have focused on the CPP in the context of perceptual decision-making, with limited work examining decisions formed from evidence culled from long-term memory. Several recent studies have examined working-memory related tasks ( Van Ede & Nobre, 2024 ; van Vugt et al., 2019 ), where the CPP also appears to be related to decision processes. However, decisions made on the basis of long term memory retrieval remain unexplored in relation to the CPP. This raises an important question about the domain-generality of the CPP– if the underlying neural dynamics reflect a central decision-making mechanism, it should track evidence accumulation for both sensory and mnemonic evidence and be predictive of decision behaviors in both cases. The current study sought to address this gap by directly comparing the neural dynamics of decision-making in a perceptual luminance discrimination task and a semantic memory discrimination task involving retrieval of semantic knowledge about US state populations. We examined whether a CPP-like component emerges in both tasks, scales similarly with evidence strength, and predicts decision behaviors like RT and confidence. Building on prior work linking the CPP to perceptual decisions ( Kelly & O’Connell, 2013 ; O’Connell et al., 2012 ), we predicted that CPP slopes would increase with stronger evidence, faster RTs, and higher subjective confidence in both the perception and memory tasks. Methods Participants 24 total participants completed 640 experimental trials across two tasks; 320 visual task trials, and 320 memory task trials. All participants were recruited from the University of California, Santa Cruz (UCSC) in exchange for course credit, reported normal or corrected-to-normal vision, and provided written formal consent. All procedures performed in this study were reviewed and approved by the institutional review board at the University of California, Santa Cruz. Four participants were removed from analysis due to accuracy levels near or below chance (50%) in the easy difficulty level of the mnemonic task, suggesting little familiarity with U.S. state population sizes. The present sample size was chosen to be on par with our recent work on perceptual evidence accumulation ( Dou et al., 2024 ; Morrow et al., 2024 ). A power analysis of the results from Experiment 3 of Dou et al. (2024) , who reported effect sizes for the effect of evidence strength on stimulus- and response-locked CPP slopes, suggests that 5 to 18 subjects, respectively, are needed to detect an effect with 80% power. Stimuli and Apparatus In all tasks, stimuli were presented on a grey background (∼30 cd/m 2 ) using an electrically shielded VIEWPixx/EEG monitor (120 Hz refresh rate, resolution 1920 × 1080 pixels) that was ∼53 cm wide and was viewed at a distance of ∼70 cm from a chinrest. Stimulus presentation and behavioral data were controlled by Psychophysics Toolbox (Version 3; Kleiner et al., 2007 ; Pelli, 1997 ) running in the MATLAB environment. In the visual task, two strings of eight capitalized ‘X’ characters (Helvetica font, size 16) were presented directly (0.3 degrees of visual angle) above and below the central fixation point. One string was designated the “comparison” stimulus as it varied in luminance intensity across three levels (34.3, 34.8, 37 cd/m 2 ), randomly determined on each trial. The luminance of the other “standard” stimulus was fixed at 33.9 cd/m 2 . The difference in luminance between the standard and comparison string determined the strength of evidence for the visual decision. In the memory task, the two strings were the names of states in the US and were presented at the same locations above and below the central fixation point, with each string having the same luminance value (33.9) for all trials. For both tasks, the stimuli linearly ramped up in luminance across the first 300ms (from the background luminance to the final luminance value) to reduce sensory onset responses in the EEG that might overlap with decision-related signals. Stimuli remained on the screen until a response was made or until the trial timed out (see Procedure). The strength of memory evidence was manipulated by creating state population pairs that varied in their degree of population differences. To this end, we took 2020-2024 US census data ( US Census Bureau, 2024 ) for each US state and calculated the absolute value of population differences for every possible state combination. Three difficulty levels were defined based on the absolute differences between state pairs, chosen based on pilot behavioral data: Hard (0-20th percentile), Medium (20-45th percentile), and Easy (45-100th percentile). On each trial, a state pair was randomly sampled from all pairs that fell within the assigned difficulty bin. This binning ensured a balanced number of trials per difficulty level, as well as provided a quantifiable form of evidence strength comparable to the difficulty manipulation in luminance values in the visual task. Procedure Participants were tested individually in a dim, sound-attenuated room. The experimenter explained the instructions to the participant and verified that the participant understood the instructions before proceeding to the practice trials and critical blocks. The participant completed a practice block of 20 trials, followed by two critical blocks of 80 trials each. This sequence was repeated twice per task with the response hand mapping changed after two blocks in a counterbalanced manner. On both practice and critical trials, participants heard a beep if they did not respond within a designated response window for that task (1.8 seconds for the visual task and 3 seconds for the memory task), prompting the participant to continue to their confidence response without a decision response. In the visual task, participants performed a visual luminance discrimination task followed by confidence ratings. Participants were informed they would be presented with two strings of X’s, located above and below the central fixation point, varying in luminance from each other. Participants were instructed to identify which of the two strings had the higher luminance and to respond as quickly as possible, followed by their confidence rating. For the memory task, participants performed a population discrimination task, identifying which of two presented states in the United States had a greater population based on prior semantic knowledge with no training regarding population information prior to beginning the task. Similarly to the visual task, participants were instructed to respond as quickly as possible, followed by a confidence rating for the previous response. For both tasks, stimuli remained on the screen until a response was made (or until the response window timed out). The stimuli then disappeared after the response, prompting the (untimed) confidence report, given on a 4-point scale. For half the blocks, participants were instructed to keep their fingers on the right keys “J” and “K” representing the location of the target stimulus being either above or below the central fixation, respectively. Participants were additionally instructed to keep their fingers on the left keys “1”, “2”, “3”, and “4” (from pinky to index finger) to indicate confidence from lowest to highest, respectively. In blocks where the response mapping was flipped, participants were instructed to keep their fingers on the left keys “F” and “D” representing the location of the target stimulus being either below or above the central fixation, respectively. Participants were additionally instructed to keep their fingers on the right keys “7”, “8”, “9”, and “0” (from pinky to index finger) to indicate confidence from lowest to highest, respectively. Each participant completed 640 trials in total, consisting of 320 trials in the visual task and 320 trials in the memory task with three levels of difficulty varying randomly across trials for both tasks. Choice and confidence hand responses were counterbalanced both within and between participants. EEG and Recording Analysis Continuous EEG was acquired from 63 active electrodes (BrainVision actiCHamp), with impedance at each central-parietal electrode kept below 20kΩ. Recordings were digitized at 1000 Hz, and FCz was used as the online reference. EEG was processed offline using custom scripts in MATLAB (version R2019b) and the EEGLAB toolbox ( Delorme & Makeig, 2004 ). Data were high-pass filtered at 0.1 Hz, then downsampled to 500 Hz. Data were re-referenced offline to the average of all electrodes. Continuous signals were then segmented into epochs centered on stimulus onset using a time window of -1,000ms to 2,500ms for the visual task, or a time window of -1,000ms to 3,200ms for the memory task. Individual trials were rejected if any scalp channel exceeded ±100 μV at any time during the interval extending from -300 to 1000 ms relative to stimulus onset. On average, 36 trials were rejected per subject across both tasks, and trials excluded from the EEG data were similarly excluded from the analyses of behavior. Noisy channels were interpolated, and an independent components analysis was performed to identify and remove components reflecting eye-blinks or movements. On average, 1.2 components were removed per subject (range = 1-2). Statistical Analyses For the current experiment, the CPP buildup rate was defined as the slope of a line fit to each participants’ average CPP waveform. The CPP slope was the main focus of our EEG analysis as it is most theoretically related to the construct of evidence-accumulation rate, with steeper slopes reflecting a faster build-up of evidence ( Kelly & O’Connell, 2013 ; O’Connell et al., 2012 ). For each condition, EEG data were averaged across trials and over a predefined set of centro-parietal electrodes (CP1, CP2, CP3, CP4, CPz, P1, P2, P3, P4, Pz), yielding a single ERP waveform per subject per condition. We computed the slope of the CPP component by fitting a line to the ERP waveform for each subject, task (visual, memory), and difficulty condition (easy, medium, hard). For the response-locked analyses in the visual task, the slope was calculated over a window between -800ms to -50ms relative to the response, while the memory task slope was calculated over a window between -1200ms to -50ms. For the stimulus-locked analyses, the slope was calculated over a window between 300ms to 700ms in the visual task, or between 300 to 1000ms in the memory task. The memory task involved longer overall response times so the windows were adjusted accordingly to ensure the accumulation process was accurately captured. These time windows were determined based on visual inspection of the grand-average ERP from both tasks and best capture the rising phase of the stimulus- and response-locked CPP waveform. We conducted additional analyses to investigate if the CPP slope varied as a function of both reaction time and confidence within each difficulty level. For each subject and task, trials were split into fast and slow RTs based on a median split, and into high and low confidence trials based on a mean split, consistent with prior work linking CPP dynamics to decision timing and subjective certainty ( Dou et al., 2024 ; Herding et al., 2019 ; Kurtz et al., 2017 ). A line was fit to the resulting CPP waveforms using the same time windows as the previous analysis. Finally, we conducted exploratory correlational analyses (as our sample size was not chosen to examine individual differences) to test if participants use a common threshold for terminating perceptual and mnemonic decisions. If the CPP reflects bounded evidence accumulation, then the mean amplitude of the CPP just prior to the response should reflect the amount of evidence a given participant requires before committing to a choice. To test for a common threshold, we computed the Spearman correlation across participants between the pre-response (−100 to -25ms) amplitudes across the memory and perceptual tasks. A similar correlation was computed between median RTs in each task (averaging over difficulty levels), which could provide further behavioral evidence of a common threshold. Results Behavior Twenty subjects completed 640 trials of perceptual and mnemonic decision tasks, discriminating either greater luminance or greater state populations ( Figure 1A ) with varying levels of evidence strength (i.e., difficulty). Three 2 (task: perceptual, memory) x 3 (difficulty: easy, medium, hard) repeated-measures ANOVAs were conducted to examine the effects of task type and difficulty on behavioral measures of accuracy, reaction time, and subjective confidence ( Figure 1B ). Download figure Open in new tab Figure 1. Example trial and behavioral results. A) Each trial began with a central fixation dot and 300ms gradual luminance ramp up of the visual or memory stimuli, which were displayed until response (up to a maximum of 1800 or 3000 ms, respectively). Stimuli were presented at one of three difficulty levels randomly for each trial. Participants were tasked with choosing the brighter alphanumeric string (visual) or more populated state (memory) and rating their confidence in that decision (on a scale of 1-4). B) The mean accuracy (left plot) decreased as difficulty increased, median reaction time (middle plot) increased as difficulty increased, and mean confidence rating (right plot) decreased as difficulty increased. Error bars denote ±1 SEM (across participants). Accuracy The data revealed a significant main effect of Task on accuracy, F (1, 19) = 23.77, p < .001, and Difficulty, F (2, 38) = 419.66, p < .001, indicating that accuracy was higher in the visual task compared to the memory task, and decreased as task difficulty increased. The Task x Difficulty interaction was not significant, F (2, 38) = 2.78, p = .075, suggesting that the effect of difficulty on accuracy was similar across tasks. Reaction Time The data revealed a significant main effect of Task, F (1, 19) = 250.75, p < .001, and Difficulty, F (2, 38) = 84.94, p < .001, indicating that reaction times were faster in visual task compared to the memory task, and increased as task difficulty increased. No significant Task x Difficulty interaction was observed, F (2,38) = 2.02, p = .146, indicating that the effect of difficulty on reaction times was similar across tasks. Confidence There was a significant main effect of Difficulty, F (2, 38) = 99.41, p < .001, indicating that participants reported higher confidence on easier trials. The Task x Difficulty interaction was also significant, F (2,38) = 3.52, p = .040, suggesting that the effect of difficulty on confidence differed between the tasks. There was no significant main effect of Task, F (1,19) = 0.144, p = .708, indicating that overall confidence did not vary between the two tasks. A Neural Signature of Evidence Accumulation Across Perception and Memory We manipulated the strength of evidence while participants made perceptual- and memory-based decisions with concurrent EEG recording. If the CPP reflects evidence accumulation, its buildup rate (quantified by slope) should 1) increase as the quality of evidence increases, 2) increase as RTs decrease, and 3) increase as subjective confidence increases. As expected, we observed a positive ERP component over central parietal electrodes in the perceptual task which peaked around 650 ms poststimulus ( Figure 2A , left) and ramped up prior to motor decision response ( Figure 2A , right), resembling the CPP ( O’Connell et al., 2012 ). A similar positive ERP component was observed in the memory task, peaking around 1000 ms poststimulus ( Figure 2B , left) and ramping up prior to the motor response ( Figure 2B , right). Notably, the topographies of the perception and memory CPP were highly similar stimulus- (r = 0.943, p < .001) and response-locked (r = 0.963, p < .001) analyses, and the later-peaking CPP in the memory task fits with the longer RTs observed overall in the memory task. Thus, to a first approximation, the CPP component in the perception and memory tasks seem highly similar, providing initial evidence for a domain-general neural signature of evidence accumulation. Download figure Open in new tab Figure 2. Effects of perceptual and menemonic task difficulty on the CPP components. Grand average CPP waveforms were aligned to both stimulus onset (left column) and response execution (right column). The vertical dashed line denotes the time of stimulus onset (left) and time of response execution (right). The inset topographies of the ERP measured after stimulus onset for the visual perception task (A) and for the memory task (B) show a positive central-parietal component (electrodes used for analysis are shown in white dots) across both stimulus and response-locked windows. Note that the topography time windows are different to account for the longer memory-based decision time. Stimulus-aligned and response-aligned CPP slopes both increased as evidence strength increased for both the perception and memory tasks. To test if the CPP was sensitive to the strength of memory and/or perceptual evidence (and task), we ran 2 (task: perceptual, memory) x 3 (difficulty: easy, medium, hard) repeated-measures ANOVAs predicting CPP slope separately for response- and stimulus locked waveforms. We found significant main effects of Task ( F (1,19) = 26.62, p < .001) and Difficulty ( F (2,38) = 35.98, p < .001) on the stimulus-locked CPP slope, as well as a significant interaction between the two ( F (2,38) = 15.46, p < .001). These results indicate that CPP slopes were steeper for the visual task compared to the memory task, with slopes critically varying as a function of evidence strength. Time locking these components to the motor response additionally revealed significant main effects of Task ( F (1,19) = 27.31, p < .001) and Difficulty ( F (2,38) = 44.91, p < .001), indicating that slopes varied reliably across the difficulty levels. The Task x Difficulty interaction was also significant, F (2,38) = 4.74, p = .015, suggesting that the influence of difficulty on CPP slopes differed between the two tasks. Because the Task x Difficulty interaction effects reported in the previous section could be driven by a lack of difficulty effect in one task, we conducted additional one-way repeated measures ANOVAs and repeated-measures t-tests predicting CPP slopes from difficulty level within each task separately. If the CPP is domain-general, we would expect significant effects of difficulty on slopes in both tasks. As predicted, the effects of difficulty on CPP slope across both tasks were significant, with steeper slopes in the easier task difficulties. In the perceptual task, both stimulus-locked ( F (2,38) = 32.77, p < .001) and response-locked ( F (2,38) = 35.37, p < .001) analyses revealed significant main effects of Difficulty, with pairwise comparisons revealing steeper slopes in the Easy condition compared to both Hard ( t (19) = 6.24, p < .001 SL; t (19) = 6.72, p < .001 RL) and Medium ( t (19) = 5.78, p < .001 SL; t (19) = 6.97, p < .001 RL). Hard and Medium difficulty in the perceptual task did not differ ( t (19) = 0.65, p = .52 SL, t (19) = , p = .156 RL). Critically, in the memory task, both stimulus-locked ( F (2,38) = 11.35, p < .001) and response-locked ( F (2,38) = 18.42, p < .001) analysis revealed a significant main effect of Difficulty, with pairwise comparisons revealing steeper CPP slopes for Easy compared to both Hard ( t (19) = 4.60, p < .001 SL; t (19) = 4.49, p < .001 RL) and Medium ( t (19) = 3.63, p = .002 SL; t (19) = 5.35, p = .001 RL), with no difference between Medium and Hard ( t (19) = -0.10, p = .924 SL; t (19) = 1.39, p = .54 RL). Reaction Time Analyses To test whether the CPP could also predict decision-related behaviors across both tasks, we next examined whether CPP slopes varied with RT in both the mnemonic and perceptual tasks. Trials were categorized as fast or slow within each task and difficulty level using a median split on RTs. A 2 (RT: Fast, Slow) x 3 (Difficulty: Easy, Medium, Hard) repeated-measures ANOVA predicting CPP slopes was conducted separately for each task ( Figure 3 ). Download figure Open in new tab Figure 3. Effects of RT on the CPP. A) Grand average CPP waveforms aligned to both stimulus onset (left column) and response execution (right column). The vertical dashed line denotes the time of stimulus onset (left) and time of response execution (right). The CPP was broken down by difficulty level for each task and then split into fast (solid line) and slow (dashed line) RT trials within each difficulty level based on a median split. Both stimulus-aligned and response-aligned CPP slopes were steeper on fast-RT trials as compared to slow-RT trials. Results from the perceptual task revealed main effects of RT on CPP slope for both the stimulus-locked ( F (1,19) = 24.90, p < .001) and response-locked ( F (1,19) = 38.63, p < 0.05) analyses. Additionally, significant main effects of difficulty were observed in both stimulus-locked ( F (2,38) = 32.85, p < .001) and response-locked ( F (2,38) = 35.24, p < 0.05) analyses, as reported in the previous section, though no significant interaction effect between RT and Difficulty was observed in either the response-locked ( F (2,38) = 2.43, p = 0.12) or stimulus-locked ( F (2,38) = 1.96, p = .155) analyses. These results indicate that CPP amplitude varied systematically with both RT and difficulty, but the influence of RT did not differ across difficulty levels within the perceptual task ( Figure 3A ). Results from the memory task also revealed significant main effects of RT on CPP slopes for both the stimulus-locked ( F (1,19) = 14.33, p = .0013) and response-locked ( F (1,19) = 29.34, p < 0.05) analyses., Additionally, significant main effects of difficulty were found for the stimulus-locked ( F (2,38) = 11.35, p < .001) and response-locked ( F (2,38) = 18.49, p < 0.05) analyses, along with a significant interaction effect between RT and difficulty for stimulus-locked ( F (2,38) = 3.54, p = .039) and response-locked ( F (2,38) = 16.86, p < 0.05) analyses. These results indicate that CPP amplitude in the memory task was influenced by both RT and difficulty, with the relationship between RT and CPP varying across difficulty levels ( Figure 3B ). Confidence Analyses Prior studies have correlated the CPP with subjective confidence judgments, such that higher rates of reported confidence are associated with steeper CPP slopes in perceptual decisions ( Dou et al., 2024 ; Grimaldi et al., 2015 ; Herding et al., 2019 ). Following from the hypothesis that the CPP reflects a domain-general signal of evidence accumulation, we predicted that higher subjective confidence would correspond to steeper CPP slopes during both perceptual and mnemonic decisions. Trials were categorized as High or Low Confidence within each task and difficulty level using a mean split on confidence reports. A 2 (Confidence: High, Low) x 3 (Difficulty: Easy, Medium, Hard) repeated-measures ANOVA was conducted separately for the perceptual and memory tasks ( Figure 4 ). Download figure Open in new tab Figure 4. Effects of confidence on the CPP. A) Grand Average CPP waveforms aligned to both stimulus onset (left column) and response execution (right column). The CPP was broken down by difficulty level for each task and then split into high (solid line) and low (dashed line) confidence trials within each difficulty level based on a mean split. Both stimulus-aligned and response-aligned CPP slopes were steeper on high-confidence trials compared to low-confidence trials. As expected, confidence ratings were positively correlated with CPP slope, such that higher confidence corresponded to steeper CPP slopes across tasks. In the perceptual task, significant main effects of Confidence on CPP slopes were present in both stimulus-locked ( F (1,19) = 5.88, p = .0255) and response-locked ( F (1,19) = 12.42, p < 0.01) analysis. A main effect of Difficulty was also observed in both stimulus-locked ( F (2,38) = 19.97, p < .001) and response-locked ( F (2,38) = 12.37, p < 0.05) analyses. No interaction effect was observed in either stimulus- ( F (2,38) = 1.61, p = .214) or response-locked ( F (2,38) = 0.311, p = 0.734) analyses, suggesting that the influence of confidence was consistent across difficulty level within the perceptual task ( Figure 4A ). Similarly within the memory task, significant main effects of Confidence were observed for both stimulus-locked ( F (1,19) = 4.40, p = .0496) and response-locked ( F (1,19)=7.14, p < 0.01) analyses. A main effect of difficulty was found for the stimulus-locked ( F (2,38) = 11.43, p < .001) and response-locked ( F (2,38) = 15.80, p < 0.01) analysis. Notably, an interaction effect between Confidence and Difficulty was observed in the response-locked ( F (2,38)=11.29, p <0.001) but not the stimulus-locked ( F (2,38) = 2.24, p = .120) analyses, indicating that within the memory task, the effect of confidence on CPP slope depended on trial difficulty ( Figure 4B ). Threshold Analyses To examine whether the peak amplitude of the CPP just prior to response commitment reflects a domain-general threshold applied to both perceptual and mnemonic decisions in our tasks we correlated each subject’s pre-response (−100 to - 25 ms) response-locked CPP amplitude between the perceptual and memory tasks. We predicted that individuals who exhibit larger pre-response peak CPP amplitude in one task would also have a larger amplitude in the other task. Consistent with this prediction, peak response-locked CPP amplitudes were strongly correlated across perceptual and memory tasks, (ρ = 0.87, p < 0.001), indicating that peak pre-response CPP amplitude could reflect a stable signature of the threshold of required evidence which generalizes across domains ( Figure 5A ). As a result, we further predicted that individuals who exhibited faster RTs in one task would exhibit faster RTs in the other task. Consistent with this prediction, a medium positive correlation was observed between median reaction times in the visual task and memory task (ρ = 0.43, p = 0.058), suggesting a trend towards a common underlying factor influencing decision speed across domains ( Figure 5B ). Download figure Open in new tab Figure 5. A) Spearman correlation between response-locked memory (x axis) and perceptual (y axis) pre-response peak CPP amplitudes (−100 to -25 ms) collapsed across difficulty levels. Participants with higher pre-response peaks in the memory task also showed higher pre-response peak amplitudes in the perceptual task, suggesting a domain-general evidence accumulation threshold. B) Spearman correlation between response-locked memory (x axis) and perceptual (y axis) median RTs collapsed across difficulty levels,suggesting a common influence on decision speed. Discussion It has been suggested that the decision-making framework of evidence accumulation extends beyond the sampling of externally-driven sensory evidence to include retrieval of internal samples of evidence from memory, though little neural evidence has been presented to support this notion in humans (Ratcliff, 1978; Shadlen & Shohamy, 2016 ; Van Ede & Nobre, 2024 ). The present study addresses the link between evidence accumulation and mnemonic decision-making by investigating whether the CPP ERP component tracks evidence accumulation during perceptual (visual) and mnemonic (semantic) decision-making tasks. Behaviorally, as expected, easier trials were associated with higher accuracy, faster responses, and greater subjective confidence across both perceptual and memory tasks. At the neural level, we found that the CPP tracks evidence accumulation across both perceptual and mnemonic decisions with steeper CPP slopes on trials with stronger evidence, faster reaction times, and higher confidence in both tasks. Furthermore, we found overall shallower slopes in the memory task. The fact that similar evidence-sensitive dynamics were observed in our memory task suggests that the accumulation of internal evidence over time might be a key computation underlying some judgments based on semantic memory. In perceptual decisions, evidence accumulation begins shortly after stimulus presentation and continuously builds (so long as sensory evidence is available) until a decision threshold is reached ( Herding et al., 2019 ; Kelly & O’Connell, 2013 ; O’Connell et al., 2012 ; Shadlen & Shohamy, 2016 ; Twomey et al., 2015 ). In contrast, the accumulation of mnemonic evidence likely depends on factors such as retrieval and evaluation of stored representations, which could result in greater variability in accumulation onset, slower buildup rates, and longer reaction times compared to perception ( Bornstein & Norman, 2017 ; Ratcliff, 1978; Ratcliff & Starns, 2013 ; Shadlen & Shohamy, 2016 ; Weidemann & Kahana, 2016 ). Such retrieval-dependent variability would likely contribute to the delayed CPP onset and reduced slope observed in the memory condition, not to mention the overall greater complexity involved in first reading two memory cues before internal evidence sampling can begin. Although our memory task involved reading and understanding the word stimuli that cued the retrieval of state population information, which could be seen as an additional form of perceptual decision-making occurring during the memory task, the process of reading each state was independent of the ground-truth evidence that we experimentally manipulated (i.e., the actual difference in population between any two states), which is what the CPP slope was sensitive to. In other words, while our memory task did involve perceptual processing, the differences in CPP slopes were specifically modulated by the strength of mnemonic evidence (in addition to participant’s decision behavior), which was manipulated experimentally and independently of perceptual processing demands. Additionally, the fact that both behavior and the CPP was sensitive to the pre-defined difficulty of the semantic decisions suggests that even though participants were not trained on any state population information, the internal representations of population sizes that our participants came into the lab with were, on average, sensitive to the true underlying population differences. Notably, with our manipulation of sensory and mnemonic evidence, the effect of difficulty on CPP slopes seemed largely driven by the easy difficulty condition, which clearly differentiated from the medium and hard conditions (but which were not clearly differentiated from one another). This likely reflects the fact that in medium and hard conditions, although accuracy levels were clearly different, RTs and confidence were actually quite similar (see Figure 1B ). This pattern, which was also reflected in the CPP, suggests that the CPP dynamics were most sensitive to RT and confidence changes across difficulty levels, rather than accuracy. Our findings contribute to the broader theoretical debate about whether the nature of decision processes are embodied or abstract. Embodied cognition theories propose that cognitive processes are grounded in the body’s sensory and motor systems, such that decision-making may depend on modality-specific representations and actions ( Wilson, 2002 ). In contrast, the observation that the CPP exhibits similar evidence-accumulation dynamics across perceptual and mnemonic domains in the current study suggests that decision formation can operate abstractly, independent of the evidence source. Prior work has already shown that the CPP can be dissociated from overt motor preparation ( Kelly & O’Connell, 2013 ; O’Connell et al., 2012 ), indicating that it indexes an amodal accumulation-to-bound process rather than a motor-specific signal. Extending this logic, our finding that the CPP also tracks evidence derived from internal memory retrieval provides further evidence against a strictly embodied account of decision making. Instead, it supports the view that decision-making relies on shared computational principles across domains, such as evidence accumulation to a common decision threshold, providing a unified decision framework in which a single latent accumulation process underlies diverse forms of choice behavior ( Shadlen & Shohamy, 2016 ). To further test the domain-generality of the CPP, future research could examine the extent to which the CPP may underlie other forms of internal evidence accumulation. For example, whether the CPP encoded evidence accumulation during different types of mnemonic decision tasks, such as recognition or free recall remains unknown. Such a paradigm could extend our findings beyond semantic decision making and into the domain of episodic memory retrieval. It may also be of interest to explore tasks that rely on more economic, emotional, and social decision making rather than strictly perceptual or semantic evidence alone. Lastly, it should be noted that our findings so far show a neural signal of evidence accumulation from memory on a trial- and subject-averaged basis. It is possible that on single trials, or for certain subjects, retrieval dynamics do not conform to a gradual build-up of evidence, but rather a more step-like transition in memory access, or myriad other possible temporal profiles of internal evidence access. While our present results do not distinguish between these different accounts of single-trial decision making, this remains an important area for future research, perhaps addressable through joint computational modeling of evidence accumulation dynamics and single-trial EEG activity. The clear pattern of effects observed across decision modality supports the view that the CPP reflects a shared evidence accumulation mechanism across perceptual and semantic decisions. Together, these results indicate that the CPP reflects an abstract neural mechanism underlying decision formation, linking perceptual and mnemonic decisions through shared accumulation dynamics. Acknowledgements We would like to thank Jacob Chaudhry, Jaxon Elsner, Alexandra Mcgowan, and Aishwaroopa Narayanan for their help on this project. References ↵ Bornstein , A. M. , & Norman , K. A . ( 2017 ). Reinstated episodic context guides sampling-based decisions for reward . Nature Neuroscience , 20 ( 7 ), 997 – 1003 . doi: 10.1038/nn.4573 OpenUrl CrossRef ↵ Charles , L. , Chardin , C. , & Haggard , P . ( 2020 ). Evidence for metacognitive bias in perception of voluntary action . Cognition , 194 , 104041 . doi: 10.1016/j.cognition.2019.104041 OpenUrl CrossRef PubMed ↵ Delorme , A. , & Makeig , S . ( 2004 ). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis . Journal of Neuroscience Methods , 134 ( 1 ), 9 – 21 . doi: 10.1016/j.jneumeth.2003.10.009 OpenUrl CrossRef PubMed Web of Science ↵ Desender , K. , Donner , T. H. , & Verguts , T . ( 2021 ). Dynamic expressions of confidence within an evidence accumulation framework . Cognition , 207 , 104522 . doi: 10.1016/j.cognition.2020.104522 OpenUrl CrossRef PubMed ↵ Dou , W. , Martinez Arango , L. J. , Castaneda , O. G. , Arellano , L. , Mcintyre , E. , Yballa , C. , & Samaha , J . ( 2024 ). Neural Signatures of Evidence Accumulation Encode Subjective Perceptual Confidence Independent of Performance . Psychological Science , 35 ( 7 ), 760 – 779 . doi: 10.1177/09567976241246561 OpenUrl CrossRef ↵ Grimaldi , P. , Lau , H. , & Basso , M. A . ( 2015 ). There are things that we know that we know, and there are things that we do not know we do not know: Confidence in decision-making . Neuroscience & Biobehavioral Reviews , 55 , 88 – 97 . doi: 10.1016/j.neubiorev.2015.04.006 OpenUrl CrossRef PubMed ↵ Herding , J. , Ludwig , S. , von Lautz , A. , Spitzer , B. , & Blankenburg , F. ( 2019 ). Centro-parietal EEG potentials index subjective evidence and confidence during perceptual decision making . NeuroImage , 201 , 116011 . doi: 10.1016/j.neuroimage.2019.116011 OpenUrl CrossRef PubMed ↵ Kelly , S. P. , & O’Connell , R. G . ( 2013 ). Internal and External Influences on the Rate of Sensory Evidence Accumulation in the Human Brain . Journal of Neuroscience , 33 ( 50 ), 19434 – 19441 . doi: 10.1523/JNEUROSCI.3355-13.2013 OpenUrl Abstract / FREE Full Text ↵ Kiani , R. , Corthell , L. , & Shadlen , M. N . ( 2014 ). Choice Certainty Is Informed by Both Evidence and Decision Time . Neuron , 84 ( 6 ), 1329 – 1342 . doi: 10.1016/j.neuron.2014.12.015 OpenUrl CrossRef PubMed ↵ Kleiner , M. , Brainard , D. , Pelli , D. , Ingling , A. , Murray , R. , & Broussard , C . ( 2007 ). What’s new in psychtoolbox-3 . Perception , 36 ( 14 ), 1 – 16 . OpenUrl CrossRef PubMed ↵ Kurtz , P. , Shapcott , K. A. , Kaiser , J. , Schmiedt , J. T. , & Schmid , M. C . ( 2017 ). The Influence of Endogenous and Exogenous Spatial Attention on Decision Confidence . Scientific Reports , 7 ( 1 ), 6431 . doi: 10.1038/s41598-017-06715-w OpenUrl CrossRef PubMed ↵ Morrow , A. , Pilipenko , A. , Turkovich , E. , Sankaran , S. , & Samaha , J . ( 2024 ). Endogenous Attention Affects Decision-related Neural Activity but Not Afferent Visual Responses . Journal of Cognitive Neuroscience , 36 ( 11 ), 2481 – 2494 . doi: 10.1162/jocn_a_02239 OpenUrl CrossRef PubMed ↵ O’Connell , R. G. , Dockree , P. M. , & Kelly , S. P . ( 2012 ). A supramodal accumulation-to-bound signal that determines perceptual decisions in humans . Nature Neuroscience , 15 ( 12 ), 1729 – 1735 . doi: 10.1038/nn.3248 OpenUrl CrossRef PubMed ↵ O’Connell , R. G. , Shadlen , M. N. , Wong-Lin , K. , & Kelly , S. P . ( 2018 ). Bridging Neural and Computational Viewpoints on Perceptual Decision-Making . Trends in Neurosciences , 41 ( 11 ), 838 – 852 . doi: 10.1016/j.tins.2018.06.005 OpenUrl CrossRef PubMed ↵ Pelli , D. G . ( 1997 ). The VideoToolbox software for visual psychophysics: Transforming numbers into movies . Spatial Vision , 10 ( 4 ), 437 – 442 . doi: 10.1163/156856897X00366 OpenUrl CrossRef PubMed Web of Science Ratcliff , R. (n.d.). A Theory of Memory Retrieval . ↵ Ratcliff , R. , & McKoon , G . ( 2008 ). The Diffusion Decision Model: Theory and Data for Two-Choice Decision Tasks . Neural Computation , 20 ( 4 ), 873 – 922 . doi: 10.1162/neco.2008.12-06-420 OpenUrl CrossRef PubMed Web of Science ↵ Ratcliff , R. , Smith , P. L. , Brown , S. D. , & McKoon , G . ( 2016 ). Diffusion Decision Model: Current Issues and History . Trends in Cognitive Sciences , 20 ( 4 ), 260 – 281 . doi: 10.1016/j.tics.2016.01.007 OpenUrl CrossRef PubMed ↵ Ratcliff , R. , & Starns , J. J . ( 2013 ). Modeling confidence judgments, response times, and multiple choices in decision making: Recognition memory and motion discrimination . Psychological Review , 120 ( 3 ), 697 – 719 . doi: 10.1037/a0033152 OpenUrl CrossRef PubMed ↵ Schaffhauser , T. , Boubenec , Y. , & Mamassian , P . ( 2021 ). EEG signature of evidence integration suggests distinct visual and auditory representation in multisensory context . Journal of Vision , 21 ( 9 ), 2627 . doi: 10.1167/jov.21.9.2627 OpenUrl CrossRef ↵ Shadlen , M. N. , & Shohamy , D . ( 2016 ). Decision making and sequential sampling from memory . Neuron , 90 ( 5 ), 927 – 939 . doi: 10.1016/j.neuron.2016.04.036 OpenUrl CrossRef PubMed ↵ Twomey , D. M. , Murphy , P. R. , Kelly , S. P. , & O’Connell , R. G . ( 2015 ). The classic P300 encodes a build-to-threshold decision variable . European Journal of Neuroscience , 42 ( 1 ), 1636 – 1643 . doi: 10.1111/ejn.12936 OpenUrl CrossRef PubMed ↵ US Census Bureau . ( 2024 , May 28). State Population Totals and Components of Change: 2020-2024 . Census.gov . https://www.census.gov/data/tables/time-series/demo/popest/2020s-state-total.html ↵ Van Ede , F. , & Nobre , A. C. ( 2024 ). A Neural Decision Signal during Internal Sampling from Working Memory in Humans . The Journal of Neuroscience , 44 ( 19 ), e1475232024 . doi: 10.1523/JNEUROSCI.1475-23.2024 OpenUrl Abstract / FREE Full Text ↵ van Vugt , M. K. , Beulen , M. A. , & Taatgen , N. A. ( 2019 ). Relation between centro-parietal positivity and diffusion model parameters in both perceptual and memory-based decision making . Brain Research , 1715 , 1 – 12 . doi: 10.1016/j.brainres.2019.03.008 OpenUrl CrossRef PubMed ↵ Weidemann , C. T. , & Kahana , M. J . ( 2016 ). Assessing recognition memory using confidence ratings and response times . Royal Society Open Science , 3 ( 4 ), 150670 . doi: 10.1098/rsos.150670 OpenUrl CrossRef PubMed ↵ Wilson , M . ( 2002 ). Six views of embodied cognition . Psychonomic Bulletin & Review , 9 ( 4 ), 625 – 636 . doi: 10.3758/bf03196322 OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted November 13, 2025. Download PDF Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions Alice Tsvinev , April Pilipenko , Jason Samaha bioRxiv 2025.11.13.688140; doi: https://doi.org/10.1101/2025.11.13.688140 Share This Article: Copy Citation Tools A Common Neural Signal of Evidence Accumulation for Perceptual and Mnemonic Decisions Alice Tsvinev , April Pilipenko , Jason Samaha bioRxiv 2025.11.13.688140; doi: https://doi.org/10.1101/2025.11.13.688140 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7618) Biochemistry (17636) Bioengineering (13860) Bioinformatics (41847) Biophysics (21401) Cancer Biology (18536) Cell Biology (25424) Clinical Trials (138) Developmental Biology (13353) Ecology (19860) Epidemiology (2067) Evolutionary Biology (24287) Genetics (15583) Genomics (22463) Immunology (17701) Microbiology (40300) Molecular Biology (17141) Neuroscience (88434) Paleontology (666) Pathology (2825) Pharmacology and Toxicology (4813) Physiology (7633) Plant Biology (15107) Scientific Communication and Education (2042) Synthetic Biology (4285) Systems Biology (9808) Zoology (2268)