Pupil Old-New Effect Reflects an Automatic Rapid Match Between Perceptual Input and Memory Representation

Pupil Old-New Effect Reflects an Automatic Rapid Match Between Perceptual Input and Memory Representation | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 4 September 2025 V1 Latest version Share on Pupil Old-New Effect Reflects an Automatic Rapid Match Between Perceptual Input and Memory Representation Authors : Jonathon Whitlock 0000-0001-9815-5147 [email protected] , Ryan Hubbard , and Lili Sahakyan Authors Info & Affiliations https://doi.org/10.22541/au.175698424.41735811/v1 172 views 99 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Within episodic memory recognition tasks, the diameter of the pupil increases to a greater degree on correct trials viewing old, studied items compared to novel unstudied items, a finding known as the objective pupil old-new effect. Theoretical accounts of this effect relate this greater increase for correctly recognized old items to the memory strength accrued from being recently exposed to these items. However, pupil sizes on incorrect trials have shown the opposite effect: namely, greater pupil diameter increases to falsely recognized new items compared to old items that were not successfully recognized, a finding known as the subjective pupil old-new effect. The current investigation involves two experiments whose aim was to dissociate the objective and subjective pupil old-new effects across the pupil dilation time-course. Using temporal principal component analyses and hierarchical linear regression, we observed that confidence judgments dissociated the objective and subjective pupil old-new effects, and pupil sizes were larger on correct trials for matching item-scene pairs (match trials) than recombined item-scene pairs (re-pair trials), despite both kinds of pairs involving familiar (studied) items. Critically, the memory strength accounts of the pupil old-new effect cannot account for the difference in pupil sizes between match and re-pair trials due to both involving similarly familiar information. Therefore, we propose an automatic rapid match hypothesis that differentiates objective and subjective pupil effects. Introduction Pupil size fluctuations under controlled lighting provide insight into how general arousal is related to cognitive processing and task performance (for reviews, see Aston-Jones & Cohen, 2005; Sirois & Brisson, 2014). For decades, psychologists have linked pupil diameter changes to ongoing mental operations, including working memory (Kahneman & Beatty, 1966; Robinson & Unsworth, 2019), attention (Unsworth & Robinson, 2016), and emotional processing (Bradley et al., 2008). In episodic memory tasks, pupil sizes increase more for successfully recognized old items than for correctly rejected new items, a phenomenon known as the pupil old-new effect (Heaver & Hutton, 2011; Otero et al., 2011; Vo et al., 2008). This effect reflects the true status of recognized old items rather than mere endorsement of target information, as it persists even when novelty detection is emphasized (Kafkas & Montaldi, 2015). The pupil old-new effect is robust across stimulus types, emerging for verbal materials (Gerotto et al., 2023; Heaver & Hutton, 2011; Mill et al., 2016; Otero et al., 2011), positive, negative, and neutral words (Brocher & Graf, 2016; Vo et al., 2008), pseudowords (Brocher & Graf, 2016; 2017), as well as images of man-made objects (Kafkas & Montaldi, 2015) and scenes (Naber et al., 2013), and is robust even during attempts to suppress memory between learning and retrieval (Gerotto et al., 2023). The size of the pupil old-new effect is larger for stronger than weaker memories (Otero, Weekes, & Hutton, 2011; Papesh et al., 2012), and larger when test words are spoken in their original voice from study compared to a familiar or novel voice (Papesh et al., 2012). However, pupil old-new effects are reduced when tasks emphasize pseudoword detection in lexical discrimination tasks (Brocher & Graf, 2017), required speeded responses (Brocher & Graf, 2016; 2017), or involve non-mnemonic judgments like group categorization (Albi & Pajkossy, 2024). Finally, in continuous recognition tasks where the number of intervening trials before a word is repeated (i.e., lag) is manipulated—specifically lag values of 1, 4, 8, and 32—words repeated after a short lag (lag 1) evoked greater pupil dilation than those repeated after longer lags (lags 4 to 32; Oliveira et al., 2021). Therefore, the size of the pupil old-new effect varies according to memory strength, how recently an item was previously encountered, and task instructions, including those that emphasize certain kinds of information or task demands. Among several explanations for this phenomenon, the memory strength account is the most widely accepted, which suggests that encountering information activates memory representations for that information, increasing their strength and resulting in greater pupil dilation during subsequent recognition testing (Oliveira et al., 2021; Otero et al., 2011). The present study proposes a novel account, which we term ARAM: that pupil dilation to correctly recognized old items reflects an automatic rapid match (ARAM) between perceptual input and stored memory representations. On this view, the magnitude of pupil dilation depends on the degree of match between the test stimulus and the memory trace formed during learning—the better the match, the larger the dilation. If supported, this rapid match account would generate predictions that differ from those of the traditional memory strength explanation. Memory strength account of the pupil old-new effect According to the memory strength account, the greater the memory strength for correctly recognized old words, the greater the pupil size increases to these items. These predictions have been verified in various studies that either experimentally manipulated memory strength or merely indexed differences in memory strength. Pupil size increases are greater for correctly recognized old words studied in deep rather than shallow encoding conditions (Otero et al., 2011; but also see: Gross & Dobbins, 2021), for words endorsed at test with “Remember” rather than “Know” responses (Otero et al., 2011; Taikh & Bodner, 2022), and for words endorsed with high compared to low confidence judgments (Papesh et al., 2012). Even within correct responses to old items, the pupil size increase is greater for more precise responses (Albi & Pajkossy, 2025) and when recalling a greater number of original study attributes such as a word’s location in a display or font color (Siefert, He, Festa, & Heindel, 2024). These findings suggest that pupil size variations reflect a dynamic interaction between the strength of memory representations and their retrieval. Memories encoded with richer contextual or semantic details during learning may form stronger, higher-quality representations, making them more likely to be retrieved successfully and eliciting greater changes in pupil dilation when retrieved. This also aligns with the hypothesis that the pupil old-new effect reflects a rapid and automatic match between memory representations formed during learning and stimuli presented in a later recognition test. By bridging the gap between encoding and retrieval, the pupil old-new effect highlights how subtle physiological changes can index the strength and accessibility of stored memories in real time. Memory strength is a slippery construct that has been conceptualized in diverse ways in the literature, ranging from subjective Remember/Know judgments or confidence judgments to objective manipulations like repetitions, extra study time, or levels-of-processing instructions, and the present paper proposes a more precise and mechanistic account: that memory strength, as indexed by pupil dilation, reflects the degree of match between the memory representation formed during learning and the perceptual input encountered during recognition. This match-based perspective offers a more mechanistic and testable account for understanding how memory strength drives recognition-related physiological responses. Veridical vs. Subjective Pupil Old-New Effects While prior work has emphasized pupil dilation as a marker of successful recognition of studied items, research also highlights a related effect tied to the subjective experience of familiarity, even when recognition is inaccurate. Pupil size increases more when novel items are mistakenly judged as old than when old items go unrecognized — a phenomenon known as the subjective pupil old-new effect (Kafkas & Montaldi, 2015; Montefinese et al., 2013, Otero et. al., 2011). The memory strength account explains this reversed pattern of pupil size by proposing that larger pupils during false alarms reflect the activation of underlying memory traces associated with studied words. Similar to correctly recognized old items, studying material activates memory representations of both the studied items and semantically related (critical lure) items (Otero et al., 2011). As such, this subjective effect is more pronounced during false recognition of new items that are semantically related to old, studied items, compared to non-semantically related new items (Otero et al., 2011), and thus varies with the extent to which new items activate previously encountered old items due to their shared high similarity. Thus, the memory strength account suggests that there may be shared mechanisms underlying both the veridical and subjective pupil effects. Although the veridical and subjective effects might share underlying mechanisms, they exhibit distinct timing patterns in their emergence after stimulus presentation. For example, Kafkas and Montaldi (2015) used recognition tasks focusing separately on familiarity detection and novelty detection, and found that the veridical effect appeared earlier, while the subjective effect emerged later, closer to the moment of response execution. Importantly, these timing differences were observed through a comparison across different tasks, raising the possibility that the distinct timing may be influenced by the varying test conditions as opposed to reflecting differences in the underlying mechanisms contributing to these effects. We address this possibility in Experiment 1. Pupil dilation during false alarms relating to increasing evidence of “oldness” can be explained through the lens of signal detection theories. According to signal detection theories, in recognition memory tasks, participants evaluate whether a current item was previously studied by accumulating evidence of its “oldness” (Wixted, 2007). When this evidence exceeds a certain threshold, the item is judged as “old”; otherwise, it is deemed “new.” Semantic relatedness can enhance this evidence of “oldness”, making unstudied words seem familiar due to their association with studied words. According to our ARAM hypothesis, this occurs because overlapping features between the test item and the original memory trace create a partial match, which leads to increased pupil dilation compared to items with less overlap, such as semantically unrelated lures. Therefore, pupil sizes should vary according to the degree of overlap between the memory representation of studied material and the familiar stimuli presented during recognition testing. Indeed, one study (Montefinese et al., 2018) more directly tested this notion by presenting new items that were semantically related or unrelated (based on feature-based semantic norm) to studied old items, and found that pupil dilation to semantically similar new items endorsed as “old” was similar in magnitude to actually studied items endorsed as “old”. Thus, one possibility is that the pupil response during recognition testing reflects the degree to which test items match memory representations formed during learning. Pupil Size as a Rapid Match Signal We propose that the veridical pupil old-new effect emerges early in the time course of viewing a test item and reflects a rapid match between what was learned during study and the perceptual input of the information presented during testing. Thus, this effect serves as a rapid match signal that is dissociated from the actual response provided to studied items. Some preliminary evidence for this comes from the finding that pupil sizes during earlier periods of the pupillary time-course are enhanced for old relative to new items even during incorrect recognition (Kafkas & Montaldi, 2015), whereas the opposite is true during later periods. Additionally, visual inspection of results from other studies that have not explicitly tested differences in time windows of the pupillary time-course do suggest that veridical old-new effects emerge fairly early (e.g., Albi & Pajkossy, 2025; Pajkossy et al., 2020). Thus, regardless of the veridicality of recognition judgments, old items elicit larger pupil responses than new items shortly after item presentation. In Experiment 1, we examined the timing of the veridical and subjective pupil effects within a single task. Specifically, we collected participants’ confidence judgments, which served as an index of how closely the perceptual input at test matched stored memory representations. For correctly recognized old items, higher confidence judgments are assumed to reflect a stronger match between the test stimulus and the memory trace. In contrast, high confidence responses to incorrectly recognized new items—those falsely identified as old—do not indicate a true match, as these items were never studied. Therefore, confidence judgments should distinguish old from new items, even when new items are misclassified as old. This leads to differing predictions for the relationship between confidence and pupil responses across item types. Specifically, among old items, high-confidence correct recognitions should elicit greater pupil dilation than low-confidence correct recognitions. However, for new items incorrectly recognized as old, pupil size should not vary with confidence. Thus, early pupil responses are expected to reflect the degree of correspondence between test input and stored memory representations, as indicated by recognition confidence, rather than subjective confidence alone. Note that it has been proposed that the veridical and subjective pupil old-new effects share the same underlying mechanism (Otero et al., 2011). Therefore, finding that pupil sizes become larger on high confidence trials, regardless of whether the item was old or new, would provide evidence that the two effects indeed are driven by the same mechanism. We further examine predictions made from the rapid match signal account in Experiment 2 by controlling for the effect of veridical judgment and the degree of familiarity of test items. To do so, we implemented the use of associative recognition in order to contrast separate conditions that both involve old, studied items but varied the extent to which they matched earlier encoding conditions. Specifically, the associative recognition task involved studying item-scene pairs, and crucially, the test included items that were either studied with the test background scene (a match trial), a different background scene (a re-pair trial) during encoding, or novel unstudied items presented with a familiar background scene (a novel trial). Thus, for both match and re-pair trials, the test item and the background scene were previously studied and therefore should be equally familiar. However, only the match trials present the studied pair in its original form, whereas the re-pair trials present studied but non-matching pairs, and thus comparing the changes in pupil size for these two item types provides a solid test for the rapid match account. Furthermore, contrasting match and re-pair trials involving successful recognition (i.e., correct trials) ensures that recollection was successful in both instances, since correct responses indicate that the relationship between item and scene were successfully retrieved. Thus, any differences in pupil size between match and re-pair trials when recollection is successful suggests that the pupil old-new effect reflects a match between test pairs and memory representations for these pairs, rather than a difference in the degree of recollection. A critical aspect of our approach involves assessing the entire time course of when stimuli are presented during recognition testing, prior to when responses are made. In order to do so, we adopted the use of temporal principal component analysis (tPCA), a data-driven approach that allowed us to identify separate components in the pupillary time-course related to cognitive processes of interest. This method has previously been used to delineate separable overlapping components in the pupillary time-course related to recognition memory processes (Clewett et al., 2020; Johannson et al., 2018; Książek et al., 2023; Montefinese et al., 2018). Preliminary findings suggest that the pupil old-new effects have different temporal characteristics, with the veridical effects emerging earlier during viewing, shortly after the presentation of test items, whereas the subjective effects emerge later, closer to when recognition judgments are made (Kafkas & Montaldi, 2015). These differences in when the effects emerge suggest separable underlying mechanisms that can be readily identified through the use of tPCA. Since we suspect that separate mechanisms drive the veridical and subjective pupil effects, applying tPCA offers a way to disentangle these processes. Identifying these differences will also enable more targeted analyses to distinguish the mechanisms underlying the veridical and subjective pupil old–new effects. To summarize, two experiments were conducted to examine the underlying mechanisms of the veridical and subjective pupil old-new effects, and to offer a novel account. The aim of Experiment 1 is to dissociate the veridical and subjective pupil old-new effects using both confidence judgments and analysis of the entire time course to dissociate the emergence of the two effects in a single task. The aim of Experiment 2 is to closely examine predictions made by the rapid match signal account using associative recognition that varies the extent to which test pairs are a full or partial match of those studied during the learning phase. Data and analysis code are available using the following link: https://osf.io/v72x3/ Experiment 1 In the current study, we employed a typical single-item recognition memory procedure in which participants studied individual faces and were later tested with a mix of studied (old) and unstudied (new) faces. Pupil sizes were recorded during the recognition test while faces were displayed, prior to participants’ behavioral response. Pupil data were analyzed based on the memory status of each item (old or new) and the accuracy of recognition. After each recognition decision, participants rated their confidence in having correctly identified the face as old or new. These confidence judgments were used to dissociate veridical and subjective pupil effects during testing. If the pupil old-new effect reflects memory strength, then pupil size should vary with confidence for both old and new items. However, if it reflects a match between perceptual input and stored memory representations, only pupil responses to old items should scale with confidence, while responses to new items should remain unaffected. To examine how these effects evolve over time, we applied tPCA, enabling us to capture dynamic changes in pupil size throughout the face presentation period and potentially dissociate the time course of objective and subjective effects during recognition. Participants Participants were 45 undergraduates from the University of Illinois Urbana-Champaign who participated in return for course credit. The study was approved by the Institutional Review Board of the University of Illinois at Urbana-Champaign and complied with APA ethical standards in the treatment of participants. All participants gave informed consent prior to inclusion in the study. They were tested individually in the lab. The study complied with university-mandated safety protocols. Apparatus Eye movements were recorded in both study and test phases of the experiment using an Eyelink II eye-tracking system (SR Research LTD, Ontario, Canada) at a rate of 500Hz. Calibration involved a 3 x 3 automated spatial array prior to the study and test phases, ending with a centrally located cross hair that signaled the next phase would begin when participants fixated on it. The computer screen resolution was set to 1280 x 1024. Stimuli Stimuli consisted of 76 faces selected from a database of non-famous people used in previous research (Althoff & Cohen, 1999). Care was taken to match features such as clothing, hairstyles, and professional lighting across all faces. The faces were cropped to only include above the chin in each image, using Adobe Photoshop software. Face images were sized to 300 x 300 pixels, presented in color. Procedure Participants performed a single item recognition task involving 36 studied faces. The study began with a study phase, after which participants were immediately tested on the entire set of studied faces as well as 36 new faces. In the study phase, each trial began with a black fixation point on a grey background screen for 4 s, followed by a face superimposed on the grey background screen for 3 s, and ended with a 1 s intertrial interval where only the grey background scene was visible. Participants were instructed to think of whether they would be friends with each face in order to emphasize encoding of faces and were told that they would subsequently be tested on those faces. After the study phase, participants immediately underwent testing on the 36 studied faces and 36 novel faces for a total of 72 test trials. In the test phase, each trial began with a black fixation point on a grey background screen for 4 s, followed by a face for 4 s, and then the face was removed and replaced with a recognition probe asking participants whether they remember studying the previously presented face. Thus, participants were only permitted to make their response after the face was presented but no longer visible to the participants. Recognition decisions were untimed, and participants made their response by selecting either ‘1’ for old or ‘0’ for new. After they made their decision regarding the face, they were asked to provide a confidence judgment referring to how confident they were in recognizing the face as either studied or not studied, for both correct and incorrect responses. Confidence judgments were also untimed, and participants made their response by selecting either ‘1’ for Low Confidence, ‘2’ for Medium Confidence, and ‘3’ for High Confidence. Pupil Dilation Pre-Processing Pupil diameter was measured during the experiment from the left eye using the same eye-tracking apparatus. Following data collection, the data were pre-processed in order to interpolate data containing eye blinks and remove trials containing artifacts. The onsets and offsets of blinks were identified using an automated algorithm based on noise in the pupillometry signal (Hershman et al., 2018). The data between these onsets and offsets were then interpolated with shape-preserving piecewise cubic interpolation. Trial epochs from 100 ms prior to the onset to 4 seconds after the onset of each face were then extracted from the pupillometry time-course, and the 100 ms pre-stimulus period was used to baseline correct each trial by subtracting the average of the pre-stimulus period from the post-stimulus data. Lastly, trials with large rapid deviations in dilation, as well as trials containing large amounts of data interpolation were removed, and additional visual inspection was performed to remove trials with slower artifacts. Overall, an average of 4.6% of trials were removed from each subject’s data. Trial-level pupil dilation measurements were obtained by extracting the mean of the signal from 2.5-4 s after the onset of the face stimulus, as this measurement window has been used in previous work (Whitlock et al., 2023). Analytic Plan Recognition accuracy ( d’ ) was calculated from each participant’s data after hits and false alarms were transformed using log-linear transformation to avoid estimation problems arising from hit rates of 1 and false alarm rates of 0 (Snodgrass & Corwin, 1988). Untransformed hits and false alarms are shown in Appendix, Table A1. We report both overall d’ as well as analyze differences in d’ as a function of participant provided confidence judgments. Average pupil sizes were extracted from the last 1.5 seconds (i.e., 2.5-4 s) during which participants were viewing the faces while waiting for the opportunity to make their recognition response. Reaction times were log transformed and outliers that were +/- 3 SD from the mean were removed. In the following analyses, we looked at the pupil-size during test as a function of both the status of the item (old vs. new) as well as the accuracy of recognition (correct vs. incorrect). That is, rather than only looking at accurate trials, as is typical in many pupil old-new effect studies (Heaver et al., 2011; Otero et al., 2011; Papesh et. al., 2012), our goal was to assess the full range of outcomes during recognition testing in order to explore both objective and subjective pupil effects. We also collected confidence judgments in order to dissociate the objective and subjective pupil old-new effects, namely for correctly recognized old items and falsely recognized new items. Dissociating the two effects using confidence would provide evidence that they are driven by separate mechanisms, consistent with the rapid match hypothesis proposed by the current investigation. On the one hand, if both old and new items that elicit responses of ‘old’ are moderated by confidence ratings, this would suggest that it’s the amount of evidence with which responses are made that are driving both effects, consistent with the memory strength account. On the other hand, if only old but not new items are moderated by confidence ratings, this would suggest that the veridical pupil old-new effect reflects the match between the memory representation formed during learning and the stimuli presented during testing, whereas the subjective effect is driven by some other process likely related to decision making. Exploring the entire time-course of retrieval allows us to assess how rapidly these effects emerge and whether there are different temporal components driving each effect. Pupil size varied across the time-course of test trials as a function of accuracy and the status of test items, visualized in Figure 1 (left panel) as well as varying according to participant-provided confidence judgments, visualized in Figure 1 (right panel) . We also ran a separate analysis involving Confidence judgments involving both old and new test items. Figure 1 Pupil size fluctuations during the face period – Experiment 1 Note . Pupil size fluctuations across the full 4 s in which the test stimuli are displayed, prior to response, separately for old and new items as a function of correct and incorrect recognition decisions (left panel), and for correctly recognized old items endorsed with High, Medium, and Low Confidence (right panel), in Experiment 1. Shaded bars reflect the standard error around the mean. To investigate the temporal dynamics of the pupillary response, we conducted a temporal Principal Component Analysis (tPCA) of the pupil time-course, following previously described methods (Clewett et al., 2020; Johansson et al., 2018). The tPCA was conducted on averaged baseline-corrected pupil time-courses for each participant, for each condition of interest (i.e., the analysis was not conducted on the trial-level data). Thus, the input data matrix to the tPCA consisted of a participant-condition x time-point matrix, with rows corresponding to the participant and condition of interest, and the columns corresponding to each time-point sample (2 ms) of the 4 seconds following the onset of the face of the pupillary time-course. Accordingly, each time sample was treated as a dependent variable in the tPCA, such that components extracted from the analysis reflected periods of time where the samples were highly correlated with each other, but not with other periods of time in the time-course. We followed the method described by Kayser and Tenke (2003), employing an unrestricted PCA using the unstandardized covariance matrix with and Varimax rotation applying Kaiser normalization. The tPCA produces component loadings, which reflect the relationship between the dependent variables (time-point samples) and components. Dividing the loadings by the standard deviation of the data normalizes the range of the values of the loadings, and they can then be plotted to depict the temporal dynamics of each component. Component loadings can also be used to compute component scores, which reflect the contributions of each component to each participant and condition’s data. Importantly, since the tPCA is data-driven and agnostic to experimental condition, the component scores can be analyzed separately for each condition. Thus, statistical analyses to determine differences in the temporal components across conditions of interest were conducted on the resultant component scores from the tPCA. Across our analyses and experiments, the first 3 components explained roughly 90% of the data, and so we focused our statistical analyses on the top 3 components from the PCA. Since the tPCA was conducted on averaged data, some participants were excluded from the analysis, due to their behavioral accuracy for a certain condition (e.g., if a participant did not give any low confidence responses, we would not be able to include their data in a tPCA examining pupil changes following low confidence responses). The resultant N included in each tPCA conducted is reported along with the results. Statistical analyses were conducted using R (R Development Core Team, 2008). Recognition accuracy was analyzed on a trial-by-trial basis using Mixed Logit Regression Models (Jaeger, 2008), implemented with the glmer function in the lme4 package (Bates, Maechler, Bolker, & Walker, 2015). Pupil sizes were analyzed using Mixed Effects Regression Models with the lmer function in lme4 . Significance testing of coefficients was conducted using the lmerTest package (Kuznetsova et al., 2017). Follow-up analyses for significant interactions were performed using the interactions package (Long, 2022). Zero-sum contrast coding was used for each variable using the ‘contrast’ function in R for variable with two levels, whereas for Confidence, contrasts were set to separately compare high and medium confidence, and in turn medium and low confidence. Models included random intercepts for both participants and item type, with a random slope allowing the fixed effect of item type to vary across participants. Incorporating random slopes for fixed effects ensures that significant effects are robust and consistent across levels of the random slope, capturing variability in the data (Gelman & Hill, 2006). Results Recognition Accuracy (d’) Recognition accuracy ( d’ ) was analyzed with a Mixed Effects Regression Model, using Confidence (High vs. Medium vs. Low) as a fixed effect and a random intercept for Participants . Recognition accuracy was greater for High confidence ( M =2.17, SD =0.83) than Medium confidence (M = 1.11, SD =0.75) responses, ( β = 1.06, SE = 0.14, t = 7.65, p < .001), which in turn was greater than Low confidence (M = 0.71, SD =0.70) responses, ( β = 0.41, SE = 0.15, t = 2.78, p = .007). Overall recognition accuracy ( d’ ) was 1.62. Pupil Measures Pupil Old-New Effect The pupil size at test was analyzed with a Mixed Effects Regression Model, using Item Type (Old vs. New), and Accuracy (Correct vs. Incorrect) as fixed effects, and Participants and Item Type as random intercepts. There was a significant Item Type x Accuracy interaction, ( β = 67.49, SE = 19.20, t = 3.52, p < .001), visualized in Figure 2 (left panel) . This interaction was due to larger pupil sizes on correct trials for old items than new items ( β = 32.40, SE = 8.84, t = 3.66, p < .001), confirming the pupil old-new effect, whereas on incorrect trials larger pupil sizes were observed for new items than old items ( β = -35.10, SE = 16.75, t = 2.10, p = .036), confirming the subjective pupil old-new effect. Therefore, pupil sizes discriminated hits from correct rejections as well as between misses and false alarms. Exploring this interaction by splitting the analysis by Item Type revealed that pupil sizes to old items that were correctly recognized were larger than those that were not recognized ( β = 48.30, SE = 13.10, t = 3.68, p < .001), whereas no differences in pupil sizes were found for correctly recognized compared to incorrectly recognized new items ( β = -19.10, SE = 14.00, t = 1.37, p = .171). Planned comparisons for ‘old’ responses to both old and new items revealed similar pupil size increases for old items endorsed as ‘old’ as those for new items endorsed as ‘old’ ( β = 10.88, SE = 13.85, t = 0.79, p = .432). The main effects of Accuracy and Item Type were not significant, ps > .126. Confidence Judgments In order to explore whether pupil sizes varied for old responses as a function of confidence for both old and new items, we analyzed the effect of Confidence on pupil sizes separately for correctly recognized old and falsely recognized new items, visualized in Figure 2 (right panel) . For old items, there were greater pupil sizes for High confidence than Medium confidence responses ( β = 45.23, SE = 20.04, t =2.261, p = .024), but not between Medium confidence and Low confidence responses ( β = -12.99, SE = 21.70, t = 0.60, p = .550). For new items, there were no differences in pupil sizes as a function of Confidence (all p’s > .155). Figure 2 Pupil sizes for Old and New items Note. Pupil sizes for Old and New items Left: Accuracy (Correct and Incorrect) and Item Type (Old vs. New). Right: Old and New Items endorsed as “Old” varied by Confidence (High, Medium, and Low), in Experiment 1. Error bars reflect standard error of the mean. Temporal PCA We next conducted tPCA on the pupil time-course data to uncover temporal dynamics of the pupillary response that differed by experimental conditions. The first analysis focused on the four conditions of interest: Old item Correct, Old item Incorrect, New item Correct, and New item Incorrect. For each participant (N = 38), the average pupil response for each of these four conditions was entered into the tPCA matrix. Following the tPCA, the resultant component loadings for the top 3 components were plotted to depict the component time-courses, and statistical analyses were conducted on the component scores. The component loadings, scores for the top 3 components, and variance explained by each component are shown in Figure 3 (top-left panel) . The top 3 components were largely separable into 3 time windows, and thus were labeled as “Early”, “Middle” and “Late” components. Component loadings separated by condition are shown in Figure 3 (top-right panel). For each of the 3 components, we conducted a Repeated-Measures ANOVA on the component scores, with factors of Item Type (Old vs. New) and Accuracy (Correct vs. Incorrect). For the Early Component, there were no significant main effects or interaction between factors. For the Middle Component, a significant main effect of Item Type was found (F 1,37 = 7.26, p = 0.01), and a follow-up t -test confirmed that Old Correct items had larger component scores than New Correct items (t 37 = 2.37, p = .02); additional follow-up tests between conditions revealed no other significant differences (all p values > 0.15). For the Late component, a significant interaction between Item Type and Accuracy (F 1,37 = 8.47, p = .006); follow-up t -tests confirmed that New Incorrect items had larger component scores than Old Incorrect items (t 37 = 2.76, p = .009), and also identified a significant difference in component scores between Old Correct items and Old Incorrect items (t 37 = 2.97, p = .005). No other comparisons were significant (all p values > 0.15). These results suggest the veridical pupil old-new effect occurred in an earlier time-window than the subjective pupil old-new effect, and that these effects were selective to the two different time windows. Our next analysis focused on differences in pupil responses based on confidence. We conducted two different tPCAs; one focused on Old Correct items, separated by confidence judgment, and another focused on New Correct items, separated by confidence judgment. The component loadings, scores from the top 3 components, and variance explained by each component, for the analysis focused on Old Correct items (N = 39) is presented in Figure 3 (bottom-left panel) . The time-courses of the component loadings are similar to what was found in the previous tPCA. Component loadings separated by condition are shown in Figure 3 (bottom-right panel). Figure 3 Temporal PCA Results – Experiment 1 Note . Temporal PCA results comparing Old items (Correct and Incorrect) to New items (Correct and Incorrect), and correct Old items endorsed with Confidence. A) Temporal PCA results for Old and New items, Correct and Incorrect responses. Top-left: component loadings for the top 3 components. Top-right: boxplot of component scores for the top 3 components, separated into the 4 conditions. B) Temporal PCA results for correctly endorsed Old items, separated by confidence judgments. Bottom-left: component loadings for the top 3 components. Bottom-right: boxplot of component scores for the top 3 components, separated into the 3 confidence judgments. Data points are individual subjects. For each of the 3 components, we conducted paired t -tests testing for differences in component scores for the different levels of confidence. For the Early Component and the Late Component, there were no significant differences between levels of confidence (all p-values > 0.07). For the Middle Component, High confidence judgments had significantly higher component scores than Medium (t 38 = 2.27, p = .03) and Low (t 38 = 2.93, p = .006) confidence judgments; Medium and Low confidence did not significantly differ (p = .17). There were also no significant differences between levels of confidence in the Late Component (all p-values > 0.2). Thus, Old Correct items had the largest pupil responses when confidence was high in the Middle time window, in concordance with the previous tPCA result of the veridical old-new effect. Lastly, we examined differences in pupil dynamics for New Correct items based on confidence judgments using tPCA (N = 38). The analysis strategy here was identical to the previous analysis for the tPCA of Old Correct items. While the time-courses of the loadings for the top 3 components were similar here as in the previous analyses, the t -tests revealed no differences between component scores for any of the 3 components (all p-values > 0.2). Thus, the difference in pupil response based on conditions in the Middle time window was selective for Old Correct items.11It would have been ideal to also conduct an analysis examining Incorrect New items, or false alarms, that varied by confidence, using tPCA. However, very few participants made high confidence judgments to New items - across the data, only roughly 4% of New trials ended in a high confidence false alarm, and only 6 participants had more than 2 trials with high confidence incorrect responses to New items. Since tPCA is more dependent on averaged data across trials than linear mixed-effects models, we decided to rely on the analysis using mixed-effects models to interpret these effects. Discussion In Experiment 1, we dissociated the relationship between participant-provided confidence judgments and pupil sizes to old and new items, as well as the emergence in time of the veridical and subjective pupil old-new effects. Importantly, both the veridical and subjective pupil effects were observed in the period of viewing test items immediately prior to the recognition judgment. Thus, prior to the recognition judgment, faces that matched representations formed during learning elicited larger pupil size changes than faces that did not. These effects were further modulated by recognition confidence for old items but not new items, consistent with a rapid match signal account of the pupil old-new effect. Therefore, larger pupil size increases during testing involves a successful matching process between perceptual input and earlier formed memory representations, wherein a stronger match elicited even larger pupil size increases. Analyses involving the entire time course were also conducted, revealing several components that further supported the rapid match signal account. Specifically, the middle and late components dissociated the veridical and subjective pupil effects, such that the middle component showed old items elicited larger pupil sizes than new items regardless of accuracy and thus was more about the veridical status of the test item. In contrast, the late component showed a crossover between correct and incorrect trials such that responses of “old” elicited larger pupil sizes than responses of “new” regardless of the veridical status of the test item (i.e., larger pupil sizes for both old and new items that were endorsed as old). Therefore, the middle component was sensitive to a matching process between perceptual input at test and previously formed memory representations for the studied faces, whereas the late component was sensitive to the actual experience of “oldness” that drove participants’ recognition decisions. Furthermore, the tPCA revealed that it was the middle component that showed differences between levels of confidence to correctly recognized old items, further supporting the rapid time course of the matching process reflected in pupil size changes during recognition testing. Thus, the tPCA was instrumental in identifying different time-courses of the veridical and subjective pupil effects. What is particularly striking about these results is that they suggest that pupil sizes on correct trials that were greater for old than new items begin earlier in time in which subjective effects begin to emerge. The results from the tPCA motivated us to expand the time frame in which we analyze pupil sizes in Experiment 2. Specifically, the tPCA results identified separate components that displayed different timing characteristics for when the veridical and subject effects emerged and persisted throughout the time course. Our regression analyses in Experiment 2 focused on these two separate time windows relating to the components identified in Experiment 1. Experiment 2 The purpose of Experiment 2 was to further investigate the automatic rapid match account that predicts larger pupil sizes in response to matching compared to non-matching information. To evaluate this, we used associative object-scene pairs that either matched the original studied pair ( match trials ) or were a recombination of previously studied objects and scenes ( Re-pair trials ). Importantly, all objects and scenes had all been studied, with the key difference being whether the object and scene were originally studied together. Therefore, comparisons between match and Re-pair trials control for familiarity and instead vary the extent to which the trials matched the original memory representation formed during learning. According to the rapid match account, correctly recognized object-scene pairs should elicit greater pupil dilation when they reflect a full match to the memory representation—i.e., when both elements were studied together—compared to pairs made up of familiar elements studied separately. In other words, full matches should produce a stronger pupil response than recombinations of familiar elements that were studied separately. Whereas Experiment 1 inferred memory matching based on confidence judgments, Experiment 2 directly manipulated whether test pairs constituted full or partial matches to the original memory representation using an associative recognition task. Method Participants Participants were 74 undergraduates from the University of Illinois Urbana-Champaign who participated in return for course credit. The study was approved by the Institutional Review Board of the University of Illinois at Urbana-Champaign and complied with APA ethical standards in the treatment of participants. All participants gave informed consent prior to inclusion in the study. They were tested individually in the lab two years into the Covid-19 pandemic. The study complied with University-mandated safety protocols. Apparatus The apparatus was identical to that used in Experiment 1. Stimuli Stimuli consisted of 120 colored images of real-world, identifiable objects selected from various online sources, including Google Images, that were sized to 300 x 300 pixels, as well as 96 colored images of real-world scenes selected from the Fine-Grained Image Memorability (FIGRIM) dataset (Bylinskii et al., 2015). Object stimuli consisted of images of everyday objects, such as kitchen utensils, office supplies, food items, toys, musical instruments, etc. Scenes consisted of outdoor images comprised of natural and man-made landscapes, including rolling hills, farms, cities, parks, etc. Object-scene pairings were randomly determined across trials and across participants.11We conducted behavioral piloting with this paradigm to determine the difficulty level of the associative memory task, and found that participants’ memory performance was near chance levels when the same Face stimuli as in Experiment 1 were used. We therefore used object stimuli to improve performance. While this creates some differences between the two Experiments, we also note that finding similar results in Experiment 2 as in Experiment 1 would suggest that the results of the first experiment were not specific to processing face stimuli, but are generalizable to episodic memory processing across different types of stimuli. Procedure Participants performed an associative recognition task involving unique object-scene pairs. The study began with a study phase, beginning with a centrally located fixation cross for 0.5s, then a scene mask was presented for 1.5s, followed by a scene for 3s, then a blank screen for 3s. The scene mask consisted of a box-scrambled version of the scene image, which was presented prior to the scene in order to allow the pupil to respond to the luminance change, without presenting any of the semantic content of the scene (Whitlock et al., 2023). Finally, an object was presented in the middle of the screen for 4s, followed by another blank screen for 3s. The purpose of the blank screens was to allow the pupil size to stabilize after responding on each trial. Immediately following the study phase, participants were tested on their memory for object-scene pairs. Test trials began with a centrally located fixation cross for 0.5s, then a scene mask was presented for 1.5s, followed by a scene for 3s, and then a blank screen for 3s. Finally, an object was presented in the middle of the screen for 4s, followed by a test display asking for participants to endorse the object-scene pair as either a Match, Re-pair, or Novel pair. Thus, participants were only permitted to make their response after the object was presented but no longer visible to the participants. Recognition decisions were untimed, and participants made their response by selecting either ‘1’ for Match, ‘2’ for Re-pair, or ‘3’ for Novel. Finally, they were asked to make a confidence judgment of either High, Medium, or Low confidence for their response on that trial. Pupil Dilation Pre-Processing The pupil data pre-processing steps were largely identical to those performed in Experiment 1. Following artifact rejection, an average of 7.9% of trials were removed across participants. Analytic Plan Pupils varied according to the response made on those trials, visualized in Figure 4 . In addition to the tPCA analyses in Experiment 2, we also ran two separate analyses involving regression models. The first set of analyses analyzed pupil sizes as a function of trial type (Match vs. Re-pair vs. Novel) and Accuracy (Correct vs. Incorrect). Since incorrect responses in the current design could involve two separate types of responses (e.g., for Match trials, incorrect responses could involve endorsing the trial with either a Re-pair or Novel response), the second set of analyses were aimed at breaking down the incorrect responses by the kind of response made on individual trials. Analyzing the error responses according to which response was made on each trial sheds light on the nature of those error responses. In Match trials, selecting Re-pair as a response would suggest recognizing both the item and the scene, yet failing to recognize that they were learned together. Conversely, opting for Novel as a response would indicate a failure to recognize the item altogether. Regarding Re-pair trials, choosing Match as a response would signify recognizing both the item and the scene but incorrectly recognizing them as studied together. On the other hand, selecting Novel would imply a failure to recognize the item. In Novel trials, choosing Match would indicate mistakenly identifying a non-studied item as having been learned with the scene, while opting for Re-pair would suggest falsely recognizing the item and correctly recognizing it was not studied with that scene. The tPCA results from Experiment 1 identified middle and late components that differed according to accuracy and item type. Therefore, in Experiment 2, we analyzed pupil sizes in both an early and late time window. The purpose of which was to explore whether differences in pupil sizes between Match and Re-pair trials differed as a function of the time-course of test trials. According to the rapid match hypothesis, pupil sizes should be larger for correct responses to Match than Re-pair trial. Both Match and Re-pair trials involve old (studied) information and only differed according to whether that old information was studied with the test background scene or studied with a different scene than presented with at test. Importantly, this comparison involves holding both accuracy and the oldness of information constant and varying the degree to which the item matched the background scene that it was originally studied with. Therefore, any differences between correct Match and Re-pair trials would be due to matching information driving larger pupil size increases, consistent with the rapid match hypothesis. Furthermore, this difference was more likely to occur early in the time-course, consistent with the rapid nature of this match response. Figure 4 Pupil size fluctuations during the object period – Experiment 2 Note . Pupil sizes across the time-course of test trials in Match, Re-pair, and Novel trials as a function of the type of response provided on a trial-by-trial basis, in Experiment 2. Error bands reflect the standard error around the mean. Results Recognition Accuracy Recognition accuracy was analyzed with a Logit Mixed Effects Regression Model, using Trial Type (Match vs. Re-pair vs. Novel) as a fixed effect, random intercept for Participants and Item , and by-Participant random slopes for the fixed effect of Item Type . Recognition accuracy was significantly greater for Re-pair ( M =0.59, SD =0.49) trials than Match ( M =0.52, SD =0.46) trials ( β = -0.29, SE = 0.14, t = 2.09, p = .036), as well as greater for Novel ( M =0.69, SD =0.49) than Re-pair trials ( β = 0.66, SE = 0.16, t = 4.02, p < .001). Pupil Measures Pupil Associative Effect – Early Time Window The pupil size during the early time window at test was analyzed with a Mixed Effect Regression Model, using Trial Type (Match vs. Re-pair vs. Novel) and Accuracy (Correct vs. Incorrect) as fixed effects, random intercepts for Participants and Items, and random slopes for the fixed effect of Accuracy . Pupil sizes were larger for Re-pair than Novel trials ( β = 27.19, SE = 8.67, t = 3.14, p = .002), but were similar between Match and Re-pair trials ( β = 7.36, SE = 7.35, t = 1.00, p = .317). These main effects were qualified by a Trial Type (Match vs. Re-pair) and Accuracy interaction ( β = 40.35, SE = 13.29, t = 3.04, p = .002), due to larger pupil sizes for correct responses to Match trials than Re-pair trials ( β = 27.53, SE = 9.40, t = 2.93, p = .010), but not for incorrect responses ( β = -12.82, SE = 10.43, t = 1.23, p = .436), visualized in Figure 5 (left panel) . Pupil Associative Effect – Early Time Window The pupil size during the late time window at test was analyzed by a Mixed Effect Regression Model, using Trial Type (Match vs. Re-pair vs. Novel) and Accuracy (Correct vs. Incorrect) as fixed effects, random intercepts for Participants and Items, and random slopes for the fixed effect of Accuracy . Pupil sizes were significantly larger on Re-pair trials than Novel trials ( β = 40.18, SE = 8.63, t = 4.65, p < .001), whereas pupil sizes were not significantly different between Match trials and Re-pair trials ( β = -3.93, SE = 7.67, t = 0.51, p = .608). These main effects were qualified by a Trial Type (Re-pair & Novel) and Accuracy interaction ( β = 39.37, SE = 15.09, t = 2.61, p = .009), due to larger pupil sizes for correct responses to Re-pair trials than Novel trials ( β = -59.87, SE = 9.97, t = 6.00, p < .001), but not on incorrect trials ( β = -20.50, SE = 12.83, t = 1.60, p = .247), visualized in Figure 5 (right panel) . Splitting the interaction by Accuracy revealed larger pupil sizes on incorrect Novel trials than correct Novel trials ( β = -0.14, SE = 0.05, t = 2.62, p = .009), but not differences between correct and incorrect Re-pair trials. Figure 5 Pupil size in the early and late time windows: Accuracy x Trial Type – Experiment 2 Note . Pupil sizes for Match, Re-pair, and Novel trials as a function of Accuracy, in Experiment 2. Left: In the early time window. Right: In the late time window. Error bars reflect the standard error of the mean. Response Analysis – Early Time Window In order to explore the kinds of incorrect responses made on individual trials, we analyzed the pattern of pupil sizes as a function of the type of response (Match vs. Re-pair vs. Novel) made as a function of trial type. The pupil size during the early time window at test was analyzed by a Mixed Effect Regression Model, using Trial Type (Match vs. Re-pair vs. Novel) and Response (Match vs. Re-pair vs. Novel) as fixed effects, random intercepts for Participants and Items , and by-participant random slopes for the fixed effect of Response . Pupil sizes were larger for Match responses than Re-pair responses ( β = 21.59, SE = 9.79, t = 2.21, p = .028), visualized in Figure 6 (left panel) . There was a significant Trial Type (Match vs. Re-pair) and Response (Match vs. Re-pair) interaction ( β = 31.17, SE = 15.35, t = 2.03, p = .042), due to larger pupil sizes for Match responses than Re-pair responses on Match trials ( β = 37.59, SE = 11.30, t = 3.38, p = .003) but not on Re-pair trials ( β = 11.85, SE = 11.80, t = 1.00, p = .575), visualized in Figure 6 (right panel) . Response Analysis – Late Time Window The pupil size during the late time window at test was analyzed by a Mixed Effect Regression Model, using Trial Type (Match vs. Re-pair vs. Novel) and Response (Match vs. Re-pair vs. Novel) as fixed effects, random intercepts for Participants and Items, and random slopes for the fixed effect of Response . Pupil sizes were larger for Re-pair responses than Novel responses ( β = 24.33, SE = 9.24, t = 2.63, p = .010), whereas pupil sizes were not significantly different between Match responses and Re-pair responses ( β = 17.91, SE = 10.50, t = 1.71, p = .090), visualized in Figure 6 (right panel) . None of the interactions were significant, all p’s > .235. Figure 6 Pupil size in the early and late time windows: Accuracy x Response Type – Experiment 2 Note . A) Pupil sizes for Match, Re-pair, and Novel responses in the early time window of Match and Re-pair trials, in Experiment 2. Top-left: Response type in the early time window. Top-right: Trial type (Match and Re-pair) and Response (Match, Re-pair, and Novel). B) Pupil sizes for Match, Re-pair, and Novel responses in the late time window of Match and Re-pair trials, in Experiment 2. Bottom-left: Response type in the late time window. Bottom-right: Trial type (Match and Re-pair) and Response (Match, Re-pair, and Novel). Error bars reflect the standard error of the mean. Temporal PCA We next conducted tPCA on the pupil time-course data as was done in Experiment 1. The first analysis focused on the three conditions of interest: Match item Correct (participant made Match response), Re-pair item Correct (participant made Re-pair response), and Novel item Correct (participant made Novel response). For each participant (N = 75), the average pupil response for each of these three conditions was entered into the tPCA matrix. Following the tPCA, the resultant component loadings for the top 3 components were plotted to depict the component time-courses, and statistical analyses were conducted on the component scores. The component loadings, variance explained by each component, and scores for the top 3 components are shown in Figure 7 (top-left) . Here, the top 3 components identified by the tPCA bore a striking resemblance to those identified in Experiment 1, despite different stimuli and experimental instructions. As before, these 3 components were largely separable into 3-time windows, and thus were labeled as “Early”, “Middle” and “Late” components. For each of the 3 components, we conducted paired t -tests testing for differences in component scores for the different conditions. Component scores for each condition are visualized in Figure 7 (top-right) . For the Early Component, there were no significant differences between conditions (all p-values > 0.32). For the Middle Component, Correct Match items had significantly higher component scores than Correct Re-pair (t 74 = 4.73, p < 0.001) and Correct Novel (t 74 = 4.88, p < 0.001) items; Correct Re-pair and Novel items did not significantly differ (p = 0.63). For the Late Component, both Correct Match items (t 74 = 4.85, p < 0.001) and Correct Re-pair items (t 74 = 5.08, p < 0.001) had significantly higher component scores than Correct Novel items, but Correct Match and Correct Re-pair items did not significantly differ in their component scores (p = 0.88). In sum, Correct Match items had higher scores than the other conditions in the Middle time window, whereas Re-pair items had higher scores than Novel items in the Late time window, once again in concordance with the previous results of the importance of item match in the middle time window. Given our interest in the primary nature of the veridical old-new effect reflecting familiarity vs. item match, we next conducted an additional tPCA focused specifically on Match items given a Match response, compared to Re-pair items given a Match response11The initial goal was to also include Novel items given a Match response in the tPCA in order to include all 3 conditions; however, many participants never gave Novel items a Match response, and thus had empty cells for this condition. The inclusion of this condition would have reduced the N of the tPCA to 36. Therefore, since our focus was on items where familiarity was controlled, we conducted the analysis by only including Match and Re-pair items given a Match response. This tPCA (N = 72) revealed components with similar time-courses as the previous analyses, shown in Figure 7 (bottom-left), and component scores for each response condition are visualized in Figure 7 (bottom-right) . Importantly, paired t- tests on the component scores revealed that a significant difference between these two conditions was found only in the Middle component (t 71 = 2.21, p = 0.03); the Early and Late components showed no significant differences, once again providing evidence for the Middle component as an important window of time in the pupil time-course for separating item matches from non-matches. Figure 7 Temporal PCA Results: Experiment 2 Note . Temporal PCA results comparing Correct Match, Re-pair, and Novel items. A) Temporal PCA results for correct Match, Re-Pair, and Novel items. Top-left: component loadings for the top 3 components. Top-right: boxplot of component scores for the top 3 components, separated into the 3 conditions. B) Temporal PCA results for Match and Re-Pair items, when Match responses were given. Bottom-left: component loadings for the top 3 components. Bottom-right: boxplot of component scores for the top 3 components, separated into the 2 conditions. Data points are individual subjects. Discussion Experiment 2 tested the rapid match hypothesis in an associative recognition task by comparing pupil responses across Match, Re-pair, and Novel trials. The key prediction was that pupil size would reflect the degree of match between the perceptual input at test and the original memory representation, emerging shortly after the onset of test stimuli. Match trials preserved the exact item-scene pairing from encoding, while Re-pair trials recombined old items and scenes, and Novel trials present unstudied items onto familiar scenes. By focusing on comparisons between correct Match and correct Re-pair trials, we controlled for item familiarity and response accuracy, allowing us to isolate the effects of associative match. Importantly, the item and scene were presented separately to isolate the pupil response to the item itself. On each trial, a scene appeared first, followed by a blank screen, and then the item. Viewing a studied scene triggers retrieval of the information previously associated with it (Hannula et al., 2007; 2009). When the scene is followed by an item, recognition requires comparing the test item to the memory of the item paired with that scene. Thus, the pupil size patterns observed in Experiment 2 reflect how strongly the test item matched the memory representation formed during learning. Consistent with the ARAM account, greater pupil dilation was observed for correct Match than Re-pair responses early in the viewing period—well before the recognition response—suggesting an early, automatic match signal during item viewing. This difference diminished over time due to increasing pupil dilation on Re-pair trials in the later window. Temporal principal component analysis (tPCA) supported this temporal distinction: the middle component, which captured early pupil dynamics, showed stronger loadings for correct Match trials compared to Re-pair and Novel trials, reflecting the veridical match between test input and memory. In contrast, the late component showed no difference between Match and Re-pair trials but lower loadings for Novel trials, suggesting it tracked general familiarity rather than associative match. Further analyses of response types provided additional insight, as it afforded the opportunity to separate incorrect responses based on the actual response to a given trial (e.g., “Match” and “Novel” responses to Re-pair trials). On Re-pair trials, false alarms (i.e., “Match” responses to Re-pair trials) showed a pupil trajectory that shifted over time—from initially resembling Re-pair responses to aligning with true Match responses in the late window. This pattern suggests that later pupil dilation reflects a subjective sense of match, even in the absence of a true associative link. The tPCA mirrored this interpretation: while the middle component differentiated between true and false Match responses (greater loadings for true Matches), the late component did not, highlighting its sensitivity to subjective familiarity regardless of veridical match. Together, the behavioral and tPCA findings converge to support a two-stage model of pupil responses during recognition. Early pupil dilation, indexed by the middle tPCA component, reflects a rapid veridical match between test input and stored memory representations. Later dilation, captured by the late component, appears to reflect subjective familiarity or the perceived plausibility of a match, regardless of its accuracy. These findings clarify the temporal dynamics of the pupil old-new effect and suggest it is not solely driven by general familiarity but also by the quality of the match between memory and perception. General Discussion Recognition memory involves both veridical accuracy and subjective experience, and pupil dilation reflects both processes. Across two experiments, we dissociated a veridical pupil effect reflecting a rapid match signal between the perceptual input of test information and memory representations from a subjective pupil effect that is potentially associated with decision-making processes, particularly false alarms to new information. A key finding is that these effects emerge at different time points during item viewing, with the veridical pupil old-new effect appearing earlier and persisting throughout the trial, while the subjective effect arises closer to the response period. These temporal dynamics were identified using tPCA in Experiment 1 and further supported in distinct time windows in Experiment 2. In particular, Experiment 1’s tPCA revealed “middle components” relating more to veridical matches and “late components” relating more to the subjective recognition experience. Identifying these components guided our investigation of specific time periods in Experiment 2 where we again found evidence of the veridical effect emerging earlier than subjective effects and effects of overall familiarity. Therefore, pupil size increases during recognition testing reflects multiple components, one relating to veridical memory and another relating to the subjective experience of having encountered familiar information. Montefinese et al. (2018) used a similar methodology, showing that the pupil signal can be decomposed into distinct components, each modulated by different experimental factors (e.g., condition and response). In addition to dissociating the veridical and subjective effects based on their temporal emergence, we also identified distinct characteristics unique to each. Notably, the veridical effect was modulated by recognition confidence, whereas the subjective effect was not. This likely reflects the fact that confidence judgments index the degree of match between the test stimulus and the memory representation formed during encoding—the stronger the memory for studied items, the greater the likelihood of a perceptual match at test. Given that new items were, by definition, unstudied, confidence judgments are unable to reflect any strength of memory representation to these items, and thus pupil sizes to new items were unaffected by the confidence by which they were endorsed as old items, despite false alarms to these items signaling a subjective experience of “oldness” for these items. We note that, while we have used the terminology of an “earlier” component in the pupil time-course reflecting the rapid match signal, tPCA results from both experiments also identified a very early component, generally between 0 and 1 second post-stimulus onset, that explained a fair amount of variance in the pupil dilation response. Importantly, across both experiments, this component appeared to be unrelated to and did not differ between the conditions of interest, similar to what was found in other studies using this technique (Johannson et al., 2018; Montefinese et al., 2018; Raisig et al., 2007). We interpret this component of the time-course as potentially reflecting early sensory processing of the stimulus, unrelated to more elaborate cognitive processing, or perhaps differences in participants’ anticipation of the upcoming stimulus, during the preceding fixation time prior to the onset of the stimulus. The pupillary response is fairly sluggish compared to the rapid perceptual and cognitive processes happening on the order of milliseconds, and some processing prior to the onset of the stimulus could have influenced this early component (Denison et al., 2020; Hoeks & Levelt, 1993). Thus, while we are confident that the change in pupil dilation occurring around 2 seconds after the onset of the stimulus indexes a rapid match signal, it is more difficult to state exactly when this rapid match process was engaged , and answering this question will likely require incorporation of other methodologies. The current dominant theoretical account of the pupil old-new effect holds that greater pupil dilation to old versus new items reflects the activation of studied information that is re-encountered during recognition testing. Supporting this view, studies have shown an increased pupil size for Remember compared to Know responses, for deeply encoded items, and for high-confidence decisions (Otero et al., 2011; Papesh et al., 2012). However, these findings primarily emphasize the strength of the memory trace itself. In contrast, evidence in favor of the rapid match account suggests that studying an item activates associated memory representations (Montefinese et al., 2018), which remain active and lead to increased pupil responses when perceptual input at test matches those representations, due to either having encountered that item recently (Oliveira et al., 2021) or having encountered a semantically-related new item (Montefinese et al., 2018; Otero et al., 2011). The current investigation revisited this foundational idea, demonstrating that early pupil dilation during recognition testing reflects a rapid match between memory representations activated during learning and the perceptual characteristics of test stimuli. When studied information is sufficiently encoded, it is better able to be matched to the same stimuli when re-encountered during single-item recognition testing, and thus greater memory strength for learned information means test stimuli are better able to be matched to that representation. Hence, we observed greatest pupil size increases for high confidence hit responses, which we interpret as the outcome of a sufficiently strong match between the test probe and the memory representation. Relatedly, the semantic similarity between new and old words has also been shown to drive pupil size increases, varying with the degree to which the concepts overlap (Montefinese et al., 2018). Specifically, new words that shared a high feature similarity (HFS) to old words elicited larger pupil size increases than new words that shared a low feature similarity (LFS). Thus, the greater the semantic match due to overlapping features, the greater the pupil size increase, similar to what we observed in the current findings. Whereas Montefinese et al. (2018) focused on the degree of match on semantic qualities, the current investigation focused on the degree of match on perceptual qualities, and importantly both sets of findings are consistent with the matching process between test probe items and memory for recently encountered items. The concept of a matching process involving memory representations and test probes has a precedent in models of recognition memory such as REM (Shiffrin & Steyvers, 1997), MINERVA (Hintzman, 1988), and TODAM (Murdock, 1982). While these models vary on the characteristics of this matching process, all of these models converge on the notion that studying individual items involves forming representations of these items, such as a vector of unique values, and recognition involves matching those vectors to the test probes. Critically, the degree of match between the two is taken as evidence for having encountered that information during learning. Thus, the current set of experiments aligns these foundational models with more recent investigations involving pupillometry as a means of investigating the physiological basis of recognition memory. Future research could focus on whether the pupil old-new effects observed here can help inform the nature of this proposed matching process, furthering our understanding of the interaction between the LC-NE system and hippocampal-based memory (i.e., recollection of item-context relationships). Pupil dilation reflecting a match between test and studied information is evident also in associative recognition that, rather than relying on feelings of familiarity, requires recollecting specific details about the original learning episode. Recollection-based recognition assesses the extent to which participants can use the stimuli presented during recognition testing to retrieve specific details concerning the original episode in which they were encountered (Yonelinas, 2002). Specifically, the test stimuli are being compared to the contents of memory for the original episode, in order to discriminate between equally familiar information. Typical findings in studies assessing recollection relating to pupil size have revealed that pupil dilation is also greater when specific contextual details from encoding—such as an item’s original location or font color—are successfully recollected (Albi & Pajkossy, 2025; Siefert et al., 2024). Successful recollection of contextual details such as item location or other characteristics such as font color indicates a richer, more complete, memory representation that is being compared against test stimuli. Therefore, the greater the likelihood that test stimuli will activate memory representations for the full episode in which items were studied, the greater the pupil size increase to these items. Future studies could attempt to more directly link this rapid match response to behavioral metrics of recollection, through the use of ROC curves (Yonelinas, 1997), or perhaps other physiological signals, such as the LPC ERP component (Rugg & Curran, 2007). In Experiment 2, distinguishing match trials from re-pair trials required recollection, as both the scene and item had been studied before and were equally familiar. That is, determining whether a studied scene and item had appeared together necessitated retrieving the original learning episode. Thus, correct responses in both trial types reflected successful recollection, which was presumably comparable across conditions. Despite this, correct responses on match trials elicited greater pupil dilation than re-pair trials early in the viewing period. As such, the rapid match signal account was more strongly supported by the pupil size patterns found in Experiment 2. This difference diminished as the response approached, with pupil sizes converging, possibly reflecting a general familiarity signal or successful recollection. However, in other studies incorporating other modalities such as EEG, familiarity signals precede signals of recollection, rather than appear later (Rugg & Curran, 2007). An alternative explanation is that the late component of the pupil dilation response may reflect engagement of higher-order executive control or retrieval monitoring processes involved in evaluating the contents of retrieval and making decisions or goal-related actions (Albi & Pajkossy, 2025; Brocher & Graf, 2016; Dobbins et al., 2002; Euston et al., 2012). Evaluating the output of memory retrieval thus may be reflected in the late component, which more closely aligns to when responses were made. Thus, the pupil signal varies dynamically throughout the time course of item viewing, reflecting various processes that are engaged while recognition is simultaneously unfolding and being evaluated prior to making a response. Analyses of incorrect trials were potentially consistent with this pattern: early in viewing, match responses to match trials again elicited greater pupil dilation than match responses to re-pair trials, but these differences faded later in the trial, suggesting that subjective “match” judgments elicited similar pupil responses as the trial progressed regardless of actual study history. Thus, pupil sizes relating to the likelihood of match resembled those relating to veridical recognition of an exact match immediately prior to response. Further research examining pupil dilation changes when false alarms are made to lure stimuli will be critical for testing the rapid match hypothesis. Essentially, if stimuli are presented during the test that are highly similar to stimuli that were studied, then we would expect a similar change in pupil dilation to these stimuli, since they closely match the representation of a studied item. Experiment 2 of this study tested this idea in a way, by recombining the elements of some studied items during the test, thus keeping familiarity constant - however, this is different than testing individuals on items that were never studied, but are perceptually similar to studied items. While other research has not directly tested this hypothesis, there have been studies that have provided compelling results supporting our intuition. As previously mentioned, initial evidence supports this idea: false alarms to new items that were semantically related to studied items produced greater pupil dilation than false alarms to unrelated new items (Otero et al., 2011). Furthermore, Pajkossy et al. (2020) monitored pupil dilation as participants performed the Mnemonic Similarity Task, which contains images of objects during the recognition test that are similar to studied objects (Stark et al., 2019). The authors found that studied items and similar lures elicited similar changes in pupil dilation in a similar time window to our results (around 2 seconds post-stimulus); however, responses were also made during this time, which could have influenced the results. Similarly, when studied words are presented auditorily in one of two voices (male vs. female) and later repeated at test, pupil dilation is greater for words spoken in the same voice as at study compared to those spoken in a familiar or novel voice (Papesh et al., 2012). Thus, perceptual match effects can occur in the auditory domain and are not limited to visual features. Additionally, Montefinese et al. (2018) presented participants with lure words that were semantically similar to study words, and found similar pupil dilation to these lures as study words as well as PCA components that were similar to what we obtained. These intriguing results suggest that the match between the test item and the memory representation may not necessarily be strictly perceptual in nature. Overall, our findings indicate that the pupil old-new effect reflects multiple dynamic processes related to perception, memory, and decision-making, which interact over time during memory retrieval. The novelty of the rapid match hypothesis proposed in the current investigation is that it is both consistent with the underlying mechanism proposed by the memory strength and recollection accounts of the pupil old-new effect and also extends these accounts by proposing a mechanism by which current perceptual input is matched to active memory representations at the time of retrieval. Our results also suggest that deeper analysis of the time course of pupil variation during retrieval provides additional information about the underlying mechanisms producing either the veridical or subjective pupil effect. Analyzing the entire time course of retrieval, rather than specific time windows identified in previous research, affords opportunities to isolate different components that emerge separately depending on the status of the test item and whether that test item is correctly recognized, and the confidence with which responses are made. References Albi, Á., & Pajkossy, P. (2024). Pupil Old/New Effect is Modulated by Task Instruction-Evidence for an Attentional Saliency Account. Journal of Experimental Psychology: Learning, Memory, and Cognition, Advance online publication. Albi, Á., & Pajkossy, P. (2025). Pupil dilation accompanying successful recognition is linearly related to memory precision. Journal of Experimental Psychology: Learning, Memory, and Cognition . Althoff, R. R., & Cohen, N. J. (1999). Eye-movement-based memory effect: a reprocessing effect in face perception. Journal of Experimental Psychology: Learning, Memory, and Cognition , 25 (4), 997. Aston-Jones, G., & Cohen, J. D. (2005). An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance. Annual Review Neuroscience , 28 (1), 403-450. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of statistical software , 67 , 1-48. Bradley, M. M., Miccoli, L., Escrig, M. A., & Lang, P. J. (2008). The pupil as a measure of emotional arousal and autonomic activation. Psychophysiology , 45 (4), 602-607. Brocher, A., & Graf, T. (2016). Pupil old/new effects reflect stimulus encoding and decoding in short‐term memory. Psychophysiology , 53 (12), 1823-1835. Brocher, A., & Graf, T. (2017). Decision-related factors in pupil old/new effects: Attention, response execution, and false memory. Neuropsychologia , 102 , 124-134. Bylinskii, Z., Isola, P., Bainbridge, C., Torralba, A., & Oliva, A. (2015). Intrinsic and extrinsic effects on image memorability. Vision research , 116 , 165-178. Clewett, D., Gasser, C., & Davachi, L. (2020). Pupil-linked arousal signals track the temporal organization of events in memory. Nature communications , 11 (1), 4007. Denison, R. N., Parker, J. A., & Carrasco, M. (2020). Modeling pupil responses to rapid sequential events. Behavior Research Methods , 52, 1991-2007. Dobbins, I. G. (2021). Pupil dilation signals recognition salience. Psychonomic Bulletin & Review , 28 (2), 565-573. Dobbins, I. G., Foley, H., Schacter, D. L., & Wagner, A. D. (2002). Executive control during episodic retrieval: multiple prefrontal processes subserve source memory. Neuron , 35(5), 989-996. Euston, D. R., Gruber, A. J., & McNaughton, B. L. (2012). The role of medial prefrontal cortex in memory and decision making. Neuron , 76(6), 1057-1070. Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models . Cambridge: Cambridge University Press. Gerotto, J., Rêgo, G., Laurence, P., & Macedo, E. (2023). Effect of suppressing or facilitating tasks during retention interval in the Deese-Roediger-McDermott Paradigm: A behavioral and pupillometry study. Psychology & Neuroscience , 16 (1), 19. Gross, M. P., & Dobbins, I. G. (2021). Pupil dilation during memory encoding reflects time pressure rather than depth of processing. Journal of Experimental Psychology: Learning, Memory, and Cognition , 47 (2), 264. Heaver, B., & Hutton, S. B. (2011). Keeping an eye on the truth? Pupil size changes associated with recognition memory. Memory , 19 (4), 398-405. Hershman, R., Henik, A., & Cohen, N. (2018). A novel blink detection method based on pupillometry noise. Behavior research methods , 50 , 107-114. Hintzman, D. L. (1988). Judgments of frequency and recognition memory in a mmultiple-trace memory model. Psychological Review, 95, 528-551. Hoeks, B., & Levelt, W. J. (1993). Pupillary dilation as a measure of attention: A quantitative system analysis. Behavior Research Methods, Instruments, & Computers , 25(1), 16-26. Jaeger, T. F. (2008). Categorical data analysis: Away from ANOVAs (transformation or not) and towards logit mixed models. Journal of memory and language , 59 (4), 434-446. Johansson, R., Pärnamets, P., Bjernestedt, A., & Johansson, M. (2018). Pupil dilation tracks the dynamics of mnemonic interference resolution. Scientific reports , 8 (1), 4826. Kafkas, A., & Montaldi, D. (2015). The pupillary response discriminates between subjective and objective familiarity and novelty. Psychophysiology , 52 (10), 1305-1316. Kahneman, D., & Beatty, J. (1966). Pupil diameter and load on memory. Science , 154 (3756), 1583-1585. Kayser, J., & Tenke, C. E. (2003). Optimizing PCA methodology for ERP component identification and measurement: theoretical rationale and empirical evaluation. Clinical neurophysiology , 114 (12), 2307-2325. Książek, P., Zekveld, A. A., Fiedler, L., Kramer, S. E., & Wendt, D. (2023). Time-specific components of pupil responses reveal alternations in effort allocation caused by memory task demands during speech identification in noise. Trends in Hearing , 27, 23312165231153280. Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. (2017). lmerTest package: tests in linear mixed effects models. Journal of statistical software , 82 , 1-26. Long, J.A. (2022). Interactions: Comprehensive, User-Friendly Toolkit for Probing Interactions . R package version 1.1.6, https://cran.r-project.org/package=interactions. Mill, R. D., O’Connor, A. R., & Dobbins, I. G. (2016). Pupil dilation during recognition memory: Isolating unexpected recognition from judgment uncertainty. Cognition , 154 , 81-94. Montefinese, M., Ambrosini, E., Fairfield, B., & Mammarella, N. (2013). The “subjective” pupil old/new effect: Is the truth plain to see? International Journal of Psychophysiology , 89 (1), 48-56. Montefinese, M., Vinson, D., & Ambrosini, E. (2018). Recognition memory and featural similarity between concepts: The pupil’s point of view. Biological Psychology , 135, 159-169. Murdock, B. B. (1982). A theory for the storage and retrieval of item and associative information. Psychological Review, 89 , 609-626. Naber, M., Frässle, S., Rutishauser, U., & Einhäuser, W. (2013). Pupil size signals novelty and predicts later retrieval success for declarative memories of natural scenes. Journal of Vision , 13(2), 11-11. Otero, S. C., Weekes, B. S., & Hutton, S. B. (2011). Pupil size changes during recognition memory. Psychophysiology , 48 (10), 1346-1353. Pajkossy, P., Szőllősi, Á., & Racsmány, M. (2020). Pupil size changes signal hippocampus-related memory functions. Scientific Reports , 10(1), 16393. Papesh, M. H., Goldinger, S. D., & Hout, M. C. (2012). Memory strength and specificity revealed by pupillometry. International Journal of Psychophysiology , 83 (1), 56-64. Raisig, S., Welke, T., Hagendorf, H., & Meer, E. V. D. (2007). Investigating dimensional organization in scripts using the pupillary response. Psychophysiology , 44(6), 864-873. Robison, M. K., & Unsworth, N. (2019). Pupillometry tracks fluctuations in working memory performance. Attention, Perception, & Psychophysics , 81 , 407-419. Rugg, M. D., & Curran, T. (2007). Event-related potentials and recognition memory. Trends in Cognitive Sciences , 11(6), 251-257. Shiffrin, R. M., & Steyvers, M. (1997). A model for recognition memory: REM—retrieving effectively from memory. Psychonomic Bulletin & Review , 4 (2), 145-166. Siefert, E. M., He, M., Festa, E. K., & Heindel, W. C. (2024). Pupil size tracks cue–trace interactions during episodic memory retrieval. Psychophysiology , 61 (1), e14409. Sirois, S., & Brisson, J. (2014). Pupillometry. Wiley Interdisciplinary Reviews: Cognitive Science , 5 (6), 679-692. Snodgrass, J. G., & Corwin, J. (1988). Pragmatics of measuring recognition memory: applications to dementia and amnesia. Journal of Experimental Psychology: General , 117 (1), 34. Stark, S. M., Kirwan, C. B., & Stark, C. E. (2019). Mnemonic similarity task: A tool for assessing hippocampal integrity. Trends in Cognitive Sciences , 23(11), 938-951. Taikh, A., & Bodner, G. E. (2022). Pupil dilation during recognition reflects the subjective recollection/familiarity experience at test rather than the level of processing at encoding. Canadian Journal of Experimental Psychology, 76 (3), 186.s Unsworth, N., & Robison, M. K. (2016). Pupillary correlates of lapses of sustained attention. Cognitive, Affective, & Behavioral Neuroscience , 16 , 601-615. Whitlock, J., Hubbard, R., Ding, H., & Sahakyan, L. (2023). Trial-level fluctuations in pupil dilation at encoding reflect strength of relational binding. Journal of Experimental Psychology: Learning, Memory, and Cognition , 50 (2), 212. Wixted, J. T. (2007). Dual-process theory and signal-detection theory of recognition memory. Psychological review , 114 (1), 152. Võ, M. L. H., Jacobs, A. M., Kuchinke, L., Hofmann, M., Conrad, M., Schacht, A., & Hutzler, F. (2008). The coupling of emotion and cognition in the eye: Introducing the pupil old/new effect. Psychophysiology , 45 (1), 130-140. Yonelinas, A. P. (1997). Recognition memory ROCs for item and associative information: The contribution of recollection and familiarity. Memory & Cognition , 25, 747-763. Yonelinas, A. P. (2002). The nature of recollection and familiarity: A review of 30 years of research. Journal of memory and language , 46 (3), 441-517. Information & Authors Information Version history V1 Version 1 04 September 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Authors Affiliations Jonathon Whitlock 0000-0001-9815-5147 [email protected] Mississippi State University Department of Psychology View all articles by this author Ryan Hubbard University at Albany Department of Psychology View all articles by this author Lili Sahakyan University of Illinois Urbana-Champaign Department of Psychology View all articles by this author Metrics & Citations Metrics Article Usage 172 views 99 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Jonathon Whitlock, Ryan Hubbard, Lili Sahakyan. Pupil Old-New Effect Reflects an Automatic Rapid Match Between Perceptual Input and Memory Representation. Authorea . 04 September 2025. DOI: https://doi.org/10.22541/au.175698424.41735811/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175698424.41735811/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fed278b19d3df94',t:'MTc3OTI5OTExOQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();