Long term potentiation-like neural plasticity and performance-based memory function

OBJECTIVE
Experience-dependent modulation of the visual evoked potential (VEP) has emerged as a promising non-invasive proxy for assaying long term potentiation (LTP)-like plasticity in the cerebral cortex. LTP is considered the principal candidate mechanism underlying learning and memory. There is, however, a paucity of evidence exploring associations between LTP-like plasticity and performance-based learning and memory. The present study aimed to explore the relationship between VEP-plasticity and higher-order learning and memory in healthy adults.


METHOD
Visual and verbal learning and memory was assessed using the Aggie Figures Learning Test (AFLT) and the Rey Auditory Verbal Learning Test (RAVLT). The study included 111 healthy adults (61.1% females; mean age 37.6 years, range 17-71) who underwent a VEP paradigm employing visual high-frequency stimulation to induce a change in visual evoked responses recorded by scalp EEG. In addition, a more comprehensive neuropsychological assessment was administered.


RESULTS
Several significant moderate age-corrected positive correlations were found between modulation of the later VEP components (N1 and P1-N1 peak-to-peak) and both visual and verbal learning and memory performance. Further, there were significant differences in learning and memory performance between participants showing a higher degree of modulation (> 1 SD above mean) compared to participants showing a lower degree of modulation. No significant associations were found between VEP-plasticity and other neurocognitive domains.


CONCLUSIONS
The current results suggest that LTP-like plasticity indexed by VEP modulation reflect processes specific to learning and memory. Future research is needed to further delineate the complex relationship between neural plasticity and learning and memory, specifically concerning possible clinical implications in populations with deficits in learning and memory function.


Introduction
Long-term potentiation (LTP), first demonstrated in the animal hippocampus by Lømo and Bliss (1974), is considered the leading candidate mechanism underlying memory and learning in humans. The LTPphenomenon has been extensively studied at the molecular level in animal studies (Abraham, Logan, Greenwood, & Dragunow, 2002) and excised human tissue (Beck, Goussakov, Lie, Helmstaedter, & Elger, 2000). Though this phenomenon is well characterized at the molecular level, there have been challenges associated with the translation from basic research to assaying the phenomenon in the intact human brain due to the invasive nature of inducing canonical LTP. However, in recent years, stimulus-specific response modulation (SRM) of the visual (e.g. Teyler et al., 2005) or auditory evoked potential (e.g. Clapp, Kirk, Hamm, Shepherd, & Teyler, 2005) has emerged as a promising method for inducing an LTP-like effect in vivo. Repetitive high-frequent or prolonged visual or auditory sensory stimulation have been shown to induce an LTP-like effect in the sensory cortices, evident as modulation of amplitudes in the visual or auditory evoked potential poststimulation. Several studies have demonstrated this phenomenon in both sensory modalities, though the visual system seems to be more suited for expressing this phenomenon as the effect has been less consistent and robust in the auditory modality (Rygvold, Hatlestad-Hall, Elvsåshagen, Moberget, & Andersson, 2021). The measured effect of SRM is referred to as LTP-like as it shares several important core physiological and functional characteristics with LTP assessed by single-cell recordings in animals. The SRM effect have been shown to display both input-specificity (McNair et al., 2006;Ross et al., 2008) and NMDAreceptor dependence (Clapp, Eckert, Teyler, & Abraham, 2006), considered to be defining features of LTP.
Considering that LTP is assumed to represent the molecular basis of learning and memory, an association between the magnitude of the SRM effect and performance on neurocognitive tests measuring learning and memory function might be expected. However, even though the SRM phenomenon is increasingly well-characterized, there is a paucity of evidence exploring its relation to learning, both regarding basic perceptual and higher-order learning and memory function. Perceptual learning refers to experience-dependent changes in perceptual ability related to basic features such as orientation, motion and contrast (Aberg & Herzog, 2012), and is usually expressed as improved performance after practicing a perceptual task. It has been proposed that perceptual learning and LTP rely on the same underlying mechanisms, as both types of plasticity show input specificity and are dependent on a given stimulation frequency to occur (Aberg & Herzog, 2012). Clapp et al. (2012) demonstrated that the visual detection threshold improved significantly in a group exposed to visual high-frequent stimulation (HFS) prior to a perceptual learning task compared to a control group that did not receive HFS stimulation (Clapp, Hamm, Kirk, & Teyler, 2012). A recent study from our group did not, however, demonstrate significant correlations between modulation of visual evoked potential (VEP) amplitudes and performance on a visual perceptual learning task using identical checkerboard stimuli in both tasks (Lengali et al., 2021). Though this discrepancy in results might be attributed to differences in experimental design, the lack of significant correlations suggests that the association between synaptic LTP-like plasticity and performance on perceptual learning tasks is not necessarily that easily replicable.
Studies probing the association between induction of LTP-like effects through sensory stimulation and higher-order learning in humans are sparse. Higher-order learning in this context refers to the capacity for memory and learning, as well as the intentional access to this knowledge (Lezak, Howieson, & Loring, 2004). Animal studies have typically focused on spatial memory tasks, with varying results (Lynch, 2012). Some evidence suggests that LTP and spatial learning rely on the same underlying cellular mechanisms by demonstrating that saturating LTP in rodents caused a deficit in spatial learning, though the saturation did not affect already established spatial memory (Castro, C. A.;Silbert, L. H.;McNaughton, B. L.;Barnes, 1989). Several studies have, however, failed to replicate this finding (Bliss & Richter-Levin, 1993). Spriggs et al. (2019) used a visual SRM paradigm to demonstrate that LTP-like neural plasticity significantly predicted performance on tasks of visual memory in humans. The effect was limited to modulation in a temporally later recording of post-stimulation VEP amplitudes and was not seen in the earlier VEP recordings. This discrepancy in results was attributed to the later post-stimulation block representing a maintenance phase of LTP, whereas the earlier block represents the induction phase of LTP. These processes likely rely on different underlying molecular mechanisms, where protein synthesis is necessary for the maintenance phase to occur (Spriggs et al., 2019). The synaptic plasticity and memory hypothesis posits that the memory of prior experience is mediated by the reactivation of 'traces' involving alterations in synaptic efficacy (Takeuchi, Duszkiewicz, & Morris, 2014). There are however alternative hypotheses that argue that LTP reflects other cognitive processes, such as attention, that support and facilitate memory formation. According to this view, LTP may serve as a neural equivalent to an arousal or attention device in the brain (Shors & Matzel, 1997). Though some studies have implicated attentional processes in the modulation of VEP components (Rauss, Pourtois, Vuilleumier, & Schwartz, 2012;Santesso et al., 2008), others have argued that the LTP-like effect seen after high-frequency sensory stimulation is independent of attention (W. C. Clapp et al., 2005). Definitive evidence distinguishing the contributions of these cognitive processes in LTP-expression remain elusive. As an exploratory analysis, assessment of attention as well as the assessment of the neurocognitive domains executive function, processing speed and verbal fluency was included in the current study. Exploration of the possible association between performance on these neurocognitive domains and LTP-like visual plasticity were done in an effort to increase the specificity of the potential relationship with learning and memory performance.
We have previously, in line with several other studies (e.g. Ross et al., 2008;Spriggs et al., 2018;Teyler et al., 2005), demonstrated robust modulation of VEP amplitudes using high-frequency visual stimulation in healthy adults . Corresponding with similar studies such as Elvsåshagen et al. (2012) and Zak et al. (2018) we found that the most robust modulation effect is observed in the P1-N1 peak-topeak amplitude of the VEP components. The current study is an extended analysis based on our earlier work  where neurocognitive assessment was not included, with a slightly higher n (101 versus 111). The main aim of the current study is to explore the association between LTP-like neural plasticity as indexed by modulation of VEP amplitudes, and higher-order learning and memory function defined as performance on standardized neuropsychological tests of visual and verbal learning and memory. If indeed the SRM protocol represents a proxy for measuring LTP-like plasticity, we would expect to find a positive association between the magnitude of VEP modulation and learning and memory performance indexed by neurocognitive assessment. We hypothesized that there would be a stronger association with visual memory performance compared to verbal memory performance, as the visual modality is modulated in the current SRM-experiment. We also hypothesized that the VEP modulation effect would be specifically associated to learning and memory function, and not to other domains of neuropsychological function.

Participants
A total of one hundred and eleven healthy subjects provided informed consent and participated in the study (69 females, 42 males; mean age 37.6 years, SD 13.9, range 17-71). Self-reported normal or corrected-to-normal vision, no ongoing substance abuse or use of psychoactive medication, and absence of any current or previous severe psychiatric or neurological condition was required. Participants were recruited through social media platforms (Facebook, Instagram), in addition to local advertisement. The regional ethics committee for medical research approved all procedures (ref. no: 2016/2003).

Data acquisition
Electroencephalographic (EEG) data were recorded using a 64-channel (Ag-AgCl electrodes) BioSemi ActiveTwo system (BioSemi B.V., Amsterdam). Electrodes were spatially positioned according to the international, extended 10-20 system (10-5; Oostenveld & Praamstra, 2001). Four additional external electrodes were positioned around the T.W. Rygvold et al. eyes; laterally, and inferior/superior to the right eye (corresponding to the 10-20 system locations of LO1, LO2, IO2, and SO2); and at each earlobe (locations A1 and A2). Raw data was recorded at a sampling rate of 1024 Hz. A hardware anti-aliasing filter was applied, no online filters were used. A 25-pin serial port was used to send the event markers from the MATLAB platform to the EEG acquisition software.

Experimental setup
The experimental protocol to demonstrate LTP-like modulation of sensory evoked potentials originally consisted of one visual and one auditory paradigm, run sequentially. In addition, a period of restingstate EEG recording and a loudness dependence of the auditory evoked potential (LDAEP) paradigm was included in the experimental setup. The session lasted approximately 50 min in total (see Fig. 1 for the protocol layout). We have previously reported less pronounced modulation effects of the auditory evoked potential paradigm (AEP), as well as demonstrating no significant within-subject correlations between modulation in the two sensory modalities ). In the current paper, we have therefore chosen to focus exclusively on the visual SRM paradigm.
Participants were comfortably seated 70 cm from a 24 ′′ LCD screen (BenQ, model ID: XL2420-B) on which visual stimuli were presented. The Psychtoolbox-3 environment was used to program the visual stimuli (Kleiner et al., 2007) and the MATLAB platform was used to run it (version 2015a; MathWorks, Natick, Massachusetts). Verbal instructions were given prior to the experimental session, as well as written paradigm-specific reminders on-screen before the onset of each paradigm. When not reading instructions, the participants were required to fixate on a red circular dot centrally positioned on the screen, both during stimulation and breaks.
The visual SRM paradigm consists of two baseline VEP blocks, one high frequency stimulation (HFS) block, and five post-HFS blocks that are identical to the baseline blocks. There are 40 trials in each pre-and post-HFS block, including five target trials. One trial corresponds to one reversal of a black and white checkerboard texture (check size approximately 1.0 • ), whereas the target trials cue the participants to press a response button when the color of the fixation dot briefly changes from red to green. These target trials were included to ensure attention and prevent fatigue and drowsiness during the experimental session. They were not included in the final VEP analyses. Trials in the pre-/post HFS blocks were separated by a random stimulus onset asynchrony (SOA) value in the 500-1500 ms range (mean reversal rate; 1 Hz). In the HFS block, the stimulus reversal frequency was locked to ~ 8.55 Hz, corresponding to a SOA of ~ 0.143 s customized to fit the monitor refresh frequency at 60 Hz. Each pre-and post HFS block had a duration of approximately 40 s, whereas the HFS block lasted 120 s. The post-HFS blocks were recorded approximately 2, 4, 6, 8 and 19 min after the baseline block, respectively. VEP components C1, P1 and N1 were identified for each subject individually by visual inspection. This was done by defining a time window for all peaks derived from the grand average for each participant (C1: 80-105 ms, P1: 110-145 ms, N1: 150-250 ms), and subsequently marking the amplitude peaks manually if they deviated in latency from the pre-set temporal windows. The block-specific peaks were then identified as the minimum/maximum amplitude data points inside this time window. In addition, a P1-N1 peak-to-peak amplitude was computed. The measurements were obtained from an occipital electrode cluster (mean amplitude of O1, Oz, and O2), selected to capture the maximum difference between baseline and post-HFS blocks.
The two pre-modulation blocks were averaged into one block, labelled baseline (BL). Post-HFS blocks were analyzed separately and labelled Post-HFS 1, 2, 3, 4 and 5 respectively.

EEG preprocessing
The EEG data were preprocessed in the EEGLAB (version 2021.1) environment  on the MATLAB platform (version 2019b). Continuous EEG data were resampled to 512 Hz and rereferenced to the average of the 64 EEG channels. EEG segments containing visual SRM data was extracted, and data with no relevance to the visual paradigm was excluded from further preprocessing. A lower bound 1 Hz high-pass filter (EEGLAB default, with data edge padding) was applied to remove the DC offset as well as low-frequency drifts. Channels with an amplitude SD outside an interval of 1-25 µV were removed from the reference signal iteratively. The ZapLine tool (de Cheveigné, 2020) and an upper bound 30 Hz low-pass filter (EEGLAB default, with data edge padding) were used to suppress line noise and high-frequency noise respectively. Segments of the data which displayed significant noise in > 50% of the channels were rejected. In addition, the remaining channels showing excessive noise in > 10% of the data points were removed. Signal artefacts attributable to eye blinks and facial movements were removed using independent component analysis . After EEG preprocessing, the average percentage of included epochs from the baseline recordings was 93.1%, and the average percentage of included epochs from all post-HFS blocks was 93.4%. The EEG preprocessing protocol is described in further detail in Rygvold et al. (2021) and attached in the supplementary materials.

Neurocognitive assessment
Verbal and visual memory performance was assessed using the Norwegian translation of Rey Auditory Verbal Learning Test (RAVLT) and the Aggie Figures Learning Test (AFLT) respectively. The RAVLT (Schmidt, 1996) is designed to assess verbal memory functions in adults (>16 years) and consist of a list-learning paradigm in which the participant is read a list of 15 nouns. During the learning phase this list is repeated five times, where the participant is asked to repeat as many words as possible after each reading. A second, "interference" word list is presented, and the participant is asked to recall as many words as T.W. Rygvold et al. Neurobiology of Learning and Memory 196 (2022) 107696 possible before they are asked to freely recall what is remembered from the principal list. 30 min after the principal list is first presented, the participant is again asked to freely recall the words. These trials represent immediate and delayed recall, respectively. Finally, the participant is presented with 50 words and asked to identify which words were in the principal list, which words were in the interference list, and which words have not been presented. This task provides an assessment of recognition memory. We report total number of immediately recalled words across the five learning trials to reflect verbal learning ability, and the immediate and 30 min delayed recall and recognition to reflect verbal memory recall and recognition.
The Aggie Figures Learning Test (AFLT) was designed as a nonverbal analogue to the RAVLT (Majdan, Sziklas, & Jones-Gotman, 1996). The AFLT structure is equivalent to the RAVLT and consists of five learning trials where the presented stimuli are abstract figures, one interference task, immediate and delayed recall, and recognition. As for the RAVLT, visual learning and memory performance is presented as the total number of correctly recalled figures across five learning trials, immediate and 30 min delayed recall and recognition.
The participants were also assessed on tests of attention, executive function, processing speed and verbal fluency, respectively. Assessment of attention included the digit forward task from the Wechsler Adult Intelligence Scale IV (WAIS-IV), and the total accuracy score from the Ruff 2&7 selective attention test (Ruff, Niemann, Allen, Farrow, & Wylie, 1992;Wechsler, 2008). Assessment of executive function consisted of the digit span backwards and digit span sequencing tasks from the WAIS-IV, as well as the trail making test 4, color-word interference test 1 and 2 and the verbal fluency category switching task from the Delis-Kaplan Executive Function System (D-KEFS) (Delis, Kaplan, & Kramer, 2001). Assessment of processing speed was done using the color-word interference test 1 and 2 and trail making test 2 and 3 from the D-KEFS battery (Delis, Kaplan, & Kramer, 2001), as well as the total speed score from the Ruff 2&7 selective attention test (Ruff et al., 1992). Verbal fluency was assessed by the Letter Fluency and Category Fluency tasks from the D-KEFS battery. Based on average t-scores, the test results were grouped into four composite domains of cognitive function: attention, executive function, processing speed and verbal fluency. The tests included in the executive function category corresponds to Miyake et al.'s definition of executive function, encompassing shifting, updating/monitoring and inhibition (Miyake et al., 2000).
The experimental protocol was identical for all participants. The ERP paradigm was administered prior to the neurocognitive assessment. After a brake, participants completed the neurocognitive assessment in a fixed order of test administration: the five learning trials of the AFLT as well as the interference list and immediate recall trial was administered first, followed by the five learning trials, interference list and immediate recall of the RAVLT. Subsequently, the neurocognitive tests spanning other domains were administered before the delayed recall and recognition trials of the AFLT and RAVLT were carried out 30 min following the learning trials.

Data reduction
Both previous  and current analyses showed that the modulation effect regressed to baseline levels in the fifth post-HFS block in all components. As the experimental setup included a loudness dependence of the auditory evoked potential (LDAEP) paradigm administered between post-HFS block four and five, we hypothesize that the lack of a modulation effect in the final post-HFS block may be attributed to interference from the auditory block, or alternatively reflect a fatigue effect. For this reason, only post-HFS block one through four is included in the current analysis. In addition to examining blockwise effects, average modulation scores across blocks were computed for all components; C1, P1, N1 and P1-N1 peak-to-peak. One participant had viable data from the post-HFS block five only and was therefore excluded from the VEP analyses. Due to time constraints, not all participants had the opportunity to complete the full neurocognitive assessment. The final n in the composite neurocognitive domains were as follows; attention n = 102, executive function n = 105, processing speed n = 104, and verbal fluency n = 105.

Statistical methods
We have previously shown a robust modulation effect of highfrequency visual stimulation evident as amplitude changes in all VEP components . As the current work is an extended analysis of this original study with a slightly larger n (111 vs 101), we reran analyses examining the main effect of the HFS block on the VEP component amplitudes by testing each component using repeated measures ANOVAs with block (BL, post-HFS 1, post-HFS 2, post-HFS 3, post-HFS 4, post-HFS 5) as the within-subject factor. Each component was subjected to post hoc paired samples t tests comparing each post-HFS block VEP amplitude to the associated BL block amplitude separately. The modulation effect was calculated by subtracting the BL amplitude from the post-HFS amplitudes on all components for each participant.
The distribution of learning and memory scores on AFLT and RAVLT as well as the distribution on the individual tests included in the composite score's attention, executive function, processing speed and verbal fluency were not normally distributed (Kolmogorov-Smirnov test of normality were all p <.05). Consequently, non-parametric rank-correlation analyses (Spearman's rho) were used to explore associations between VEP modulation and performance on all neurocognitive assessments. As a secondary analysis to further analyze the possible association between VEP modulation and neurocognitive function, the participants were split into two groups of higher and lower SRM responders. Higher responders were defined as participants showing a modulation effect more than one standard deviation above the mean. We have previously used the same classification criteria for higher versus lower responders in a recent publication (Rygvold, Hatlestad-Hall, Elvsåshagen, Moberget, & Andersson, 2022). Two independentsamples non-parametric Mann-Whitney U tests were conducted to compare the higher-modulation and the lower-modulation group on all VEP components separately (C1, P1, N1 and P1-N1 peak-to-peak) with visual and verbal learning and memory performance as well as the other neurocognitive composite test scores. Raw scores on tests of both visual and verbal learning and memory performance was used. As an additional exploratory analysis, the composite categories working memory, executive function, processing speed and verbal fluency (t-scores) were correlated with the averaged modulation of VEP amplitudes on all components, C1, P1, N1 and P1-N1 peak-to-peak. As the N1-component shows a negative polarity, an increase in modulation is expressed as an increased negativity. Significant negative correlation coefficients between the N1 and performance-based measures will hence signify an amplitude increase associated with higher performance scores. In cases where age correlated significantly with the neurocognitive performance scores, the correlation analyses were run as partial correlations controlling for age. Further, to examine possible differences between men and women, the partial correlation analyses were re-run split by sex. Two independent-samples Mann-Whitney U tests were also conducted to examine possible significant differences between men and women on visual and verbal learning and memory performance. Both Spearman's rho correlations and paired samples t tests were corrected for multiple comparisons using the Bonferroni correction. All correlation analyses between VEP modulation and visual learning and memory performance had a Bonferroni corrected α = 0.010, representing the correlation between a given post-HFS block and four learning and memory scores. Greenhouse-Geisser corrections were applied to all analyses of variance with repeated measures. Effect sizes are expressed as partial Eta-squared (ƞp 2 ) to facilitate for comparison across studies (Lakens, 2013). All statistical analyses were performed using IBM SPSS Statistics (version 27).  Fig. 2 for grand average VEP waveforms and the corresponding topographical maps at baseline and post-HFS. All block-wise analyses showed significant modulation in all post-HFS blocks in all components with the exception of post-HFS block 5. All analyses except post-HFS block 3 in the C1 performance survived Bonferroni correction (Bonferroni corrected α = 0.008).
Correlation analyses using Spearman's rho showed moderate, but significant, negative correlations between age and AFLT scores, both total learning score [r s ( 1 0 9 All correlations between age and AFLT scores except for the recognition score survived correction for multiple comparisons, whereas only the correlation between age and delayed recall on the RAVLT survived correction for multiple comparisons (Bonferroni corrected α = 0.01). As there were several significant correlations between age and memory and learning performance, further correlation analyses examining associations between SRM plasticity and neurocognitive performance were ran as partial correlations controlling for age.
Correlation analysis using Pearson's r showed a significant negative correlation between age and averaged C1 modulation score [r(1 0 8) = − 0.268, p =.005], indicating decreasing modulation with increasing age. There were no significant correlations between age and the remaining averaged VEP modulation scores, neither the P1, the N1 nor the P1-N1 peak-to-peak component. Table 1 shows partial, age-corrected correlations between memory and learning performance and P1-N1 modulation indicating which associations survived correction for multiple comparisons Results from correlation analyses in the following results section will refer to the partial, age-corrected correlation analyses. Partial correlation analysis corrected for age showed no significant correlations between modulation of neither the C1 nor the P1 component and visual learning and memory performance. There were significant correlations indicating an association between increased VEP amplitude modulation and visual learning and memory scores between modulation of the post-HFS 1 block of the N1 and total learning score  T.W. Rygvold et al. When splitting the results by sex while still controlling for age, there were significant correlations between the N1 post-HFS 1 block and both total learning score and recognition score in the female group. Further, the averaged N1 post-HFS 1-4 modulation showed significant correlations with total learning score. Significant correlations were also found between modulation of the P1-N1 post-HFS 2 block and total learning score as well as between modulation of the post-HFS 4 block and recognition score on the AFLT among the women. There were significant correlations between modulation of the N1 post-HFS block 2 and the total learning, immediate recall, and recognition score on the AFLT in the male group, as well as a significant correlation between the modulation of the post-HFS 1 block and delayed recall on the AFLT (all r s in the range 0.200 -0.393, all p-values in the range 0.012 -0.047).

Associations between block-wise VEP-modulation and verbal learning and memory performance
There were no significant correlations between modulation of neither the C1 nor P1 and verbal learning and memory performance. When splitting the results by sex while still controlling for age, there were significant correlations between modulation of N1 post-HFS blocks 1 and 2 and immediate recall on RAVLT, as well as a significant correlation between modulation of the N1 post-HFS block 2 and delayed recall on the RAVLT in women. Further, there were significant correlations between modulation of the P1-N1 post-HFS 2 and total RAVLT learning score as well as significant correlations between both modulation of the P1-N1 post-HFS 4 and modulation averaged across blocks 1-4 with the recognition score on the RAVLT among the women. There was a significant correlation between modulation of the P1-N1 post-HFS 1 and delayed recall on the AFLT in the male group. (all r s in the range 0.250 -0.283, all p-values in the range 0.019 -0.040).

Higher versus lower SRM response and performance on visual and verbal learning and memory
Secondary analyses using the Mann-Whitney U test showed    − 2.418, p=.016, ηp 2 = 0.053). There were no significant differences in visual learning and memory performance between higher and lower responders on the other VEP components; C1 (higher modulation n = 18, lower modulation n = 92) and P1 (higher modulation n = 14, lower modulation n = 96). See Fig. 6 for violin plots showing visual learning and memory performance in the higher-vs lower-modulation groups on all VEP components. There were significant differences between the higher and lower responding groups indicating greater learning and memory performance in participants with a higher VEP modulation effect on the P1-N1 peakto-peak on the total verbal learning score (U = 476.5, z = − 2.338, p=.019, ηp 2 = 0.049), delayed recall (U = 489.5, z = − 2.277, p=.023, ηp 2 = 0.047) and the recognition score (U = 557.5, z = − 2.067, p=.039, ηp 2 = 0.038) on the RAVLT. In addition, there were significant differences between the higher and lower responding groups on the N1 component on total learning score (U = 529.5, z = − 2.414, p=.016, ηp 2 = 0.052), and both immediate (U = 572, z = − 2.113, p=.035, ηp 2 = 0.040) and delayed recall (U = 542, z = − 2.364, p=.018, ηp 2 = 0.050) on the RAVLT. There were no significant differences in verbal learning and memory performance between higher responders and lower responders on the C1 and P1 components. See Fig. 7 for violin plots showing verbal learning and memory performance in the higher-vs lower-modulation groups on all VEP components.

Associations between block-wise VEP-modulation and other neurocognitive functions
Correlation analyses using Spearman's rho showed no significant correlations between any blocks on the VEP components C1, P1, N1 and P1-N1 peak-to-peak and the composite T-scores of neither attention, executive function, processing speed nor verbal fluency (see Table 2 for overview) after correcting for multiple comparisons. Further, there were no significant differences between the higher-modulation group and the lower-modulation group on neurocognitive test scores not related to memory on any VEP components.

Discussion
As presented in previous studies we report a robust SRM effect on all VEP components at different time intervals post-HFS Rygvold et al., 2022). The current study aimed to explore how the SRM effect relates to performance-based higher-order learning and memory function. We report modest, but significant, correlations between LTP-like neural plasticity indexed by modulation of VEP components and visual and verbal memory performance. We did not find associations between VEP modulation and other neurocognitive domains, including attention, executive function, processing speed and verbal fluency. This indicates that the plasticity indexed by this SRM protocol reflects processes more specific to performance-based learning and memory, thereby providing some support for the synaptic plasticity and memory hypothesis which argue that changes in synaptic efficacy underlie the formation of memories.

Effects of age and sex
Consistent with previous research (e.g. Golchert et al., 2019;Sundermann et al., 2016), we found significant differences between women and men in favor of the women on learning and memory performance, especially evident on verbal learning and memory. There were, however, no significant differences between the sexes on modulation of any of the VEP components after high-frequency stimulation. The current study found negative correlations between age and learning and memory performance, as previously well documented (e.g. Deary et al., 2009). Considering this, it is within reason to expect that modulation of LTP-like synaptic plasticity, assuming that this phenomenon reflects certain underlying aspects of learning and memory processes, also would decline with increasing age. However, both the current study and previous studies have shown conflicting results regarding this relationship. In the current study, only the C1 component displayed a negative correlation with age, consistent with a recent finding by Valstad et al. (2020) who additionally reported an increase of P1 modulation with increasing age (Valstad et al., 2020). de Porto et al. (2015) reported an increase of the N1b component in older participants post-HFS, whereas Spriggs et al. (2017) did not find this association (de Porto et al., 2015;Spriggs, Cadwallader, Hamm, Tippett, & Kirk, 2017). This lack of an association is hard to explain given the strong association seen between age and measures of neurocognitive performance. It is however unlikely that VEP modulation will reflect all aspects of learning and memory function. In addition, the current study sample had a relatively low mean age, which might have made any significant age-related changes in VEP amplitudes difficult to detect .

Modality-specific effects between visual memory performance and LTP-like visual synaptic plasticity
We hypothesized that there would be a stronger association between VEP modulation and visual learning and memory performance compared to verbal learning and memory performance, as the visual modality is modulated in the current experiment. Though there in some cases were stronger correlations between indices of VEP-modulation and visual learning and memory compared to verbal learning and memory, there were no statistically significant differences between the comparable correlation coefficients on learning and memory scores in the two modalities. Though this indicate that our a priori hypothesis cannot be confirmed and the associations with VEP modulation is not specific to learning and memory in the visual modality, there might still be certain shared mechanisms underlying sensory-induced synaptic plasticity in the visual cortex and higher-order visual learning and memory processes. This association may in turn reflect the involvement of the visual cortex in providing key input to the hippocampus that contribute to the construction of spatial memories (Tsanov & Manahan-Vaughan, 2008). Some evidence from animal studies demonstrate an interaction between synaptic changes in the visual cortex and the dentate gyrus in the hippocampus through parallel field recordings in both areas (Tsanov & Manahan-Vaughan, 2008). As such, the current results support the notion that experience-dependent plasticity processes in the visual system plays a pivotal part in memory-related processing of visual Fig. 6. Violin plots showing difference in total visual learning score on the AFLT between the higher-and lowerscoring modulation groups on all VEP components.
Most reported correlations between VEP plasticity and learning and memory function were in the low and moderate range. The lack of a stronger association between SRM-plasticity and memory performance may be due to the specific temporal characteristics of the paradigm. The final post-modulation block included in the current analysis (post-HFS block four) is recorded approximately-eight minutes after highfrequency stimulation and thusly represent so-called short-term potentiation rather than a later stage LTP-like effect expected to last several hours. Short-term potentiation represents the first phase of LTP induction, and only involves existing proteins. This effect is thought to decay about 15-30 min post-induction (Lisman, 2017). Spriggs et al. (2019) found that modulation of the visual evoked potential was a significant predictor of memory performance in a VEP block recorded approximately 40 min after HFS. The earlier VEP block recorded approximatelyfive minutes post-HFS did not significantly predict visual memory performance (Spriggs et al., 2019). The maintenance phase of LTP represent the continuation of synaptic efficacy supported by protein synthesis as well as insertion of additional α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) receptors, and can potentially last hours to days in vivo (Clapp et al., 2012). As induction and maintenance of LTP-like processes depend on different molecular processes, the maintenance phase of LTP-like plasticity might represent the temporal window containing the cellular properties most associated with memory performance (Spriggs et al., 2019). The brain-derived neurotrophic factor (BDNF), a member of the neurotrophin growth factor family, is secreted in an activity-dependent manner and thought to play a vital role in facilitating the expression of LTP (Bekinschtein, Cammarota, Izquierdo, & Medina, 2008). This phenomenon has been demonstrated both in vitro (Korte et al., 1995) and in vivo animal studies (Ying et al., 2002). In the extension of its association with LTP-expression, BDNF is thought to be implicated in human memory formation. Variations in the BDNF gene, specifically the val 66met polymorphism carried by approximately 25-50% of the human population, is associated with reduced performance on tests of declarative memory (Kambeitz et al., 2012). As the current study indicate that memory performance within both the visual and auditory domain is associated with an increase of VEP-amplitudes post-HFS, a presumed index of LTP-like synaptic plasticity, BDNF-expression might represent a shared factor underlying both these phenomena. As the expression of BDNF is rapid-acting, it is thought to play a part in short-term memory, i. e., the first stages of memory formation (Bekinschtein et al., 2008) and thus may potentially play a part in the early-stage short-term potentiation seen in the current study. As the BDNF val 66 met polymorphism has been shown to have a negative impact on both memory performance and magnitude of VEP-amplitudes post-HFS (Spriggs et al., 2019), the lowerresponders in the current study might hypothetically represent the percentage of the population that are carriers of the polymorphism. BDNF genotyping could be included in future studies of LTP-like plasticity and memory performance to further delineate this relationship.
Both learning and memory processes are complex phenomena not easily captured by one method of measurement, and we cannot exclude that activity-dependent synaptic modulation reflect other processes than learning such as a cellular preparation of new computational space (Bailey, Kandel, & Harris, 2015). While it is anticipated and plausible that neuroplasticity plays an important role in processes underlying learning and memory, these cognitive processes exist in multiple forms with multiple facets and thusly represent a level of complexity that likely cannot be reduced to LTP or LTP-like processes. However, the associations between synaptic plasticity and higher-order human learning and memory warrants further exploration, specifically regarding the clinical implications of impaired or disrupted plasticity in populations with learning and memory deficits. Impaired synaptic plasticity has been demonstrated in mood disorders such as major depressive disorder (Normann, Schmitz, Fürmaier, Döing, & Bach, 2007) and bipolar II disorder (Elvsåshagen et al., 2012;Zak et al., 2018), conditions that are also associated with cognitive impairments in learning and memory function (Landrø, Stiles, & Sletvold, 2001;Marvel & Paradiso, 2004). Additionally, explorations of disease-related modifications in plasticity in disorders of dementia and mild cognitive impairment (MCI), and their possible relationship to learning and memory deficits, should be the focus of further research.

Strengths and limitations of the current study
A strength of the current study is the high number of participants, which exceeds that of most related studies. The inclusion of a high Fig. 7. Violin plots showing difference in total verbal learning score on the RAVLT between the higher-and lowerscoring modulation groups on all VEP components. T.W. Rygvold et al. number of neurocognitive tests spanning several cognitive domains may also be considered a strength of this study as it allowed for assessing whether the association between LTP-like modulation and neurocognitive function is specific to memory performance. To ensure a higher degree of specificity, the current study could have included a group that did not receive the high-frequency stimulation to serve as a control group, and the lack of such is a limitation with the experimental design. A recent study by Jacob et al. (2021) did however control for this, and found that modulation of VEP amplitudes was only evident in participants receiving the high-frequency stimulation (Jacob et al., 2021). Another potential limitation with the current study concerns the paradigm design, where an auditory LDAEP paradigm was included between post-HFS block four and five. Though tapping into another modality, it is possible that exposure to this auditory paradigm contributed to the modulation effect reverting to baseline levels in post-HFS block five due to either fatigue or interference effects. To control for this, we could have run the experimental session without the LDAEP paradigm in a subgroup of participants. Finally, the experimental setup was rather extensive, encompassing both an ERP paradigm and neurocognitive assessment and spanning three hours. We cannot exclude that the total load of the experimental setup potentially affected the results negatively, as fatigue or reduced attention might have had an impact on the association between VEP plasticity and memory performance.

Conclusion
In summary, the current study found modest, but significant positive correlations between SRM, an index of LTP-like plasticity, and learning and memory performance. This association was specific to memory performance, as no correlations between SRM and measures of other neurocognitive domains was found. Though there is now a considerable body of evidence supporting that SRM is a viable method for assessing LTP-like plasticity in vivo, there is still a paucity of evidence concerning its association with higher-order learning and memory function. The specificity of the relationship between SRM and learning and memory performance seen in the current results is promising, though the moderate effect sizes warrant some caution in interpreting the results. Future studies might further explore the specificity of this relationship by assessing whether there is a stronger association between verbal learning and memory performance and indices of synaptic plasticity in the auditory modality.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by the Regional Ethics Committee, South-East Norway. ref. no: 2016/2003. The participants provided their written informed consent to participate in this study.

Author contributions
SA, CH-H, TE, and TM designed the experiment. TR collected data. CH-H designed the MATLAB script and pre-processed the data. TR analyzed data and wrote the manuscript under supervision of SA and revisions from CH-H and TM. All authors contributed to the article and approved the submitted version.

Funding
The University of Oslo covers article processing charges as part of an agreement for open access publishing with Elsevier.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.