Working memory (WM) or short-term memory actively retains a small amount of information to support ongoing mental processes (Baddeley, 2012). This core mental faculty relies upon distributed brain regions (Christophel et al., 2017; Eriksson et al., 2015), ranging from lower-level sensory areas (Harrison and Tong, 2009; but see Bettencourt and Xu, 2016) to higher-level frontoparietal networks (Bettencourt and Xu, 2016; Ester et al., 2015; Todd and Marois, 2004; Xu and Chun, 2006). This distributed neocortical network, however, often does not involve the medial temporal lobe (MTL), which is traditionally attributed to long-term episodic memory (Eichenbaum et al., 2007; Squire and Zola-Morgan, 1991). This distinction is grounded in the separation between WM and long-term memory in classic models (Atkinson and Shiffrin, 1968; Norris, 2017) and in early MTL lesion case studies (Milner et al., 1968; Scoville and Milner, 1957). Yet, this classic view is not free of controversy. A growing body of research has suggested that the MTL is involved in tasks that rely on information maintained in WM (Boran et al., 2022; Boran et al., 2019; Hannula and Ranganath, 2008; Johnson et al., 2018; Kamiński et al., 2017; Kornblith et al., 2017; Libby et al., 2014; Liu et al., 2020; Rissman et al., 2008; Xie et al., 2023a; Xie and Zaghloul, 2021). Furthermore, MTL lesions can disrupt WM task performance (Goodrich et al., 2019; Koen et al., 2017; Olson et al., 2006; Warren et al., 2014; Xie et al., 2023a). Despite these recent findings, however, major theories have not considered the MTL as a mechanism underlying WM (Jeneson and Squire, 2012; Sreenivasan and D’Esposito, 2019). First, it is unclear what computational process of the MTL is involved in WM (Sreenivasan and D’Esposito, 2019). Furthermore, the MTL tends to engage more in a WM task when long-term memory becomes relevant, for example when task loads are higher (Boran et al., 2022; Boran et al., 2019; Rissman et al., 2008) or when task stimuli are complex (Barense et al., 2007; Borders et al., 2022; Kamiński et al., 2017; Kornblith et al., 2017; Libby et al., 2014; Liu et al., 2020). As a result, contributions of the MTL to WM are often deemed secondary (Jeneson and Squire, 2012; Sreenivasan and D’Esposito, 2019).

Clarifying this issue requires specifying how the MTL contributes to WM representation and the extent to which this contribution holds even when WM task demand is minimized. Although WM and long-term memory are traditionally considered separate mental faculties, the functional parallels in both types of memory suggest potential shared neural mechanisms (Beukers et al., 2021; Cowan, 2001; Nee and Jonides, 2008; Ruchkin et al., 2003). For example, the ability to retain precise item-specific memory would require the computation to distinguish neural representations of similar information – a process known as pattern separation (Marr, 1971). This aspect of long-term memory is widely thought to emerge from various properties of the MTL’s entorhinal-DG/CA3 pathway (Aimone et al., 2011; Bakker et al., 2008; Cappiello et al., 2016; Ekstrom and Yonelinas, 2020; Korkki et al., 2021; Leal and Yassa, 2018; Marr, 1971; Reagh and Yassa, 2014; Yassa and Stark, 2011), such as abundant granule cells and strong inhibitory interneurons in the hippocampal DG, as well as powerful mossy fiber synapses between the DG and CA3 subfields (Aimone et al., 2011; Sahay et al., 2011). These properties make it possible to enable sparse coding to ensure a sufficient representational distance among similar information (Rolls, 2016; Rolls, 2013). As these hippocampal substructures communicate with other neocortical areas via the entorhinal cortex (Aimone et al., 2011; Leal and Yassa, 2018), there is a proposed gradian of pattern separation along the entorhinal-DG/CA3 pathway to support item-specific long-term episodic memory (Reagh and Yassa, 2014). These ideas are supported by evidence based on animal and human behaviors (Burke et al., 2011; Hunsaker et al., 2008; Ryan et al., 2012), electrophysiological recordings (Leutgeb et al., 2007; Lohnas et al., 2018; Sakon and Suzuki, 2019), and human fMRI (Bakker et al., 2008; Leal and Yassa, 2018; Montchal et al., 2019; Reagh and Yassa, 2014). However, the extent to which the entorhinal-DG/CA3 pathway is involved in WM, especially in humans other than animal models (Gilbert and Kesner, 2006), has remained unknown.

Several challenges faced in past research may add to this uncertainty. For example, it is difficult to infer signals from MTL substructures, especially those within the hippocampus, based on human fMRI using a standard spatial resolution (Bettencourt and Xu, 2016; Ester et al., 2015) or intracranial direct recording with limited electrode coverage (Boran et al., 2019; Johnson et al., 2018; Kamiński et al., 2017; Kornblith et al., 2017). Furthermore, the use of complex task designs with multiple memory items (Borders et al., 2022) might also be suboptimal to reveal item-specific WM information in MTL subregions without being too taxing on the WM storage limit. To investigate these issues, here, we leverage an established retro-cue orientation WM task (Bettencourt and Xu, 2016; Ester et al., 2015; Harrison and Tong, 2009) and a high-resolution fMRI protocol to test the key prediction that the MTL’s entorhinal-DG/CA3 pathway retains item-specific WM information of a simple surface feature. In this task, participants are directed to retain the orientation information of a cued stimulus from two sequentially presented orientation gratings (separated by >20°; Figure 1A). After a short delay (5 TRs; 1TR = 1.75 s), they try to reproduce the cued orientation grating as precisely as possible using the method of adjustment. As participants are retrospectively cued to retain only one item during the delay, they are expected to encode both items but then only keep one in mind during the delay period. This design imposes a task demand on the observer to correctly remember the cued orientation while resisting the interference from the internal representations of other similar orientation gratings. The retention of information selected after encoding over a short delay has been considered a hallmark of WM (Lorenc et al., 2021; Panichello and Buschman, 2021), regardless of the presence or absence of sustained neural activation (Lundqvist et al., 2018; Rose et al., 2016). If the MTL’s entorhinal-DG/CA3 pathway indeed supports this function, it is expected that the recorded delay-period activity should contain more information about the cued item, as compared with the uncued item, even though both items are initially remembered with an equal likelihood (Bettencourt and Xu, 2016; Ester et al., 2015; Harrison and Tong, 2009). If, however, information about the cued and uncued items is equally present during the delay period, the MTL may play a limited role in the representation of task-relevant information in WM but more during the initial encoding.

Figure 1 with 1 supplement see all

Download asset Open asset

Participants’ memory performance is quantified as recall error – the angular difference between the reported and the actual orientations of the cued item (Zhang and Luck, 2008). As the effective memory set size is low at one memory item, participants’ performance is high with an average absolute recall error of 12.01°±0.61° (mean ± s.e.m.). Furthermore, the recall error distribution is centered around 0° with most absolute recall errors smaller than 45° (~97% trials; Figure 1B). These behavioral data suggest that participants in general have remembered high-fidelity orientation information of the cued item during the delay period.

Fine discrimination of remembered WM content in the MTL

Of primary interest, we examined whether precise orientation information of the cued item is retained during WM retention in anatomically defined MTL regions of interest (ROIs; Figure 2A), including the entorhinal cortex (anterior-lateral, aLEC and posterior-medial, pMEC), the perirhinal cortex, para-hippocampus, and hippocampal DG/CA3, CA1, subiculum, as defined in the previous studies (Montchal et al., 2019; Reagh et al., 2017). Additionally, we chose the amygdala as a theoretically irrelevant but adjacent control region, because the involvement of the amygdala for emotionally neutral orientation information is expected to be minimal (Iwai et al., 1990). This allows us to gauge the observations in MTL ROIs while controlling for the signal-to-noise ratio in fMRI blood-oxygenation-level-dependent (BOLD) signals in deep brain structures.

Figure 2 with 2 supplements see all

Download asset Open asset

As recent neural theories of WM have proposed that information retained in WM may not rely on sustained neural activation (Ester et al., 2015; Kamiński and Rutishauser, 2020; Rose et al., 2016), we inspected how the multivoxel activity pattern in each subject-specific ROI is correlated with the retained WM content predicted by the cued orientation gating (Figure 2B). We found that certain voxels in an ROI could respond more strongly to a particular cued orientation, even when the average BOLD activity across voxels does not show preferred coding for a certain orientation (see an example in Figure 2—figure supplement 1). We then assessed the consistency of these stimulus-related multivoxel activity patterns in the MTL and the amygdala control region based on stimulus-based representational similarity analysis. In this analysis, we correlated the angular similarity of every pair of cued orientation gratings with the similarity of the evoked BOLD patterns in these trials. The rationale is that if orientation information is retained within an ROI, the recorded neural data should track the relative angular distance between any two cued orientation gratings (hence fine discrimination Kriegeskorte and Wei, 2021). Informed by the previous research (Ester et al., 2015; Harrison and Tong, 2009), we performed this analysis using the raw fMRI BOLD signals from the middle 3TRs out of the 5-TR retention interval to minimize the contribution of sensory process or anticipated retrieval, hence maximizing the inclusion of neural correlates of WM retention (Postle et al., 2000).

In line with our prediction, we found that stimulus similarity for the cued item was significantly correlated with neural similarity across trials as compared with the null in both the aLEC (t(15) = 4.29, p=6.48e-04, p_Bonferroni = 0.0052, Cohen’s d=1.11, p_boostrap <0.001) and the hippocampal DG/CA3 (t(15) = 3.64, p=0.0024, p_Bonferroni = 0.019, Cohen’s d=0.94, p_boostrap <0.001; Figure 2C). In contrast, stimulus similarity for the uncued item across trials could not predict these neural similarity patterns in these regions as compared with the null (aLEC: t(15) = –0.20, p=0.85, Cohen’s d=–0.05; DG/CA3: t(15) = 0.06, p=0.95, Cohen’s d=0.02; p_boostrap’s>0.50). Furthermore, the evoked neural similarity patterns in these regions were significantly more correlated with the cued item as compared with the uncued item (aLEC: t(15) = 2.66, p=0.018, Cohen’s d=0.69, p_boostrap = 0.015; DG/CA3: t(15) = 3.64, p=0.0024, Cohen’s d=0.94, p_boostrap = 0.0016). While the rest of the MTL showed similar patterns, we did not obtain significant evidence in other MTL ROIs following the correction of multiple comparisons (see Supplementary file 1a for full statistics). Furthermore, neural evidence related to the cued item in the aLEC and DG/CA3 was significantly stronger than that in the amygdala control ROI. This was supported by a significant cue (cued vs. uncued) by region (combined aLEC-DG/CA3 vs. amygdala) interaction effect on the correlation between stimulus and neural similarity patterns (F(1, 15)=4.97, p=0.042, p_boostrap = 0.036). Together, these results suggest that delay-period activity patterns in the entorhinal-DG/CA3 pathway are associated with retrospectively selected task-relevant information, implying the presence of item-specific WM representation in these subregions.

Reconstruction of item-specific WM information based on inverted encoding modeling

To directly reveal the item-specific WM content, we next modeled the multivoxel patterns in subject-specific ROIs using an established inverted encoding modeling (IEM) method (Ester et al., 2015). This method assumes that the multivoxel pattern in each ROI can be considered as a weighted summation of a set of orientation information channels (Figure 3A). By using partial data to train the weights of the orientation information channels and applying these weights to an independent hold-out test set, one can reconstruct the assumed orientation information channels to infer item-specific information for the remembered item – operationalized as the resultant vector length of the reconstructed orientation information channel normalized at 0° reconstruction error (Figure 3—figure supplement 1). As this approach verifies the assumed information content based on observed neural data, its results can be efficiently computed and interpreted within the assumed model even when the underlying neuronal tuning properties are unknown (Ester et al., 2015; Sprague et al., 2018). This approach, therefore, complements the model-free similarity analysis by linking representational geometry embedded in the neural data with item-specific information under a model-based framework (Kriegeskorte and Wei, 2021; Xie et al., 2023b). On the basis of this method, previous research has revealed item-specific WM information in distributed neocortical areas, including the parietal, frontal, and occipital-temporal areas (Bettencourt and Xu, 2016; Ester et al., 2015; Rademaker et al., 2019; Sprague et al., 2016), which are similar to those revealed by other multivariate classification methods (e.g. support vector machine, SVM, Ester et al., 2015). We have also replicated these IEM effects in the current dataset (Figure 3—figure supplement 2).

Figure 3 with 6 supplements see all

Download asset Open asset

Moving beyond these well-established observations in distributed neocortical structures, we found that the amount of reconstructed item-specific information for the cued item during WM retention was also significantly greater than chance level in two anatomically defined MTL subregions, aLEC (t(15) = 4.41, p=5.07e-04, p_bonferroni = 0.0041, Cohen’s d=1.14, p_boostrap <0.001) and the hippocampal DG/CA3 (t(15) = 4.73, p=2.68e-04, p_bonferroni = 0.0021, Cohen’s d=1.22, p_boostrap <0.001; Figure 3B). These effects were specific to the maintenance of the cued item, as information related to the uncued item was not statistically different from chance (aLEC: t(15) = –0.35, p=0.74, Cohen’s d=–0.09; DG/CA3: t(15) = 0.66, p=0.52, Cohen’s d=0.17; p_boostrap’s>0.50) and was significantly less than that for the cued item (aLEC: t(15) = 2.75, p=0.015, Cohen’s d=0.71, p_boostrap = 0.018; DG/CA3: t(15) = 3.83, p=0.0016, Cohen’s d=0.99, p_boostrap = 0.0023). Critically, the amount of information specific to the cued item in the aLEC and DG/CA3 was significantly greater than that in the amygdala control ROI, which is supported by a significant cue (cued vs. uncued) by region (combined aLEC-DG/CA3 vs. amygdala) interaction effect on IEM reconstruction outcomes (F(1, 15)=7.16, p=0.016, p_boostrap = 0.010).

Collectively, results from complementary analytical procedures suggest that the MTL’s entorhinal-DG/CA3 pathway retains precise item-specific WM content for a simple surface feature (e.g. orientation) to allow fine discrimination of different items in the feature space. As such, the stimulus-based prediction of neural similarity is highly correlated with the amount of reconstructed information based on IEM, even though these two analyses are based on different analytical assumptions (e.g. correlation between IEM and representational similarity analysis for the cued item, aLEC: r=0.87, p=0.000012, p_boostrap <0.001; DG/CA3: r=0.78, p=0.00037, p_boostrap <0.001; Figure 3—figure supplement 3).

Reconstruction of WM Item Information in the MTL is associated with recall fidelity

Next, we examined the extent to which WM information retained in the MTL’s aLEC-DG/CA3 circuitry is related to an observer’s subsequent recall behavior. As the angular resolution of the reconstructed orientation information is 20° in the current study, our data therefore suggest that the MTL can distinguish similar orientation information in WM that is at least 20° apart. This neural separation should be consequential for later recall performance, in that trials with greater item-specific information reconstructed from the MTL should be associated with higher WM recall fidelity. To test this prediction, we grouped the trials from each participant into two categories. The first category contained small recall error trials, where participants made an effective recall response within one similar item away from the cued item (absolute recall error <20°; 149±3 trials [mean ± s.e.m.]). Another category contained larger recall error trials (27±3 trials) with absolute recall errors that were greater than 20° but smaller than the 3 standard deviations (SD) of the aggregated recall error distribution (Figure 4A). These trials would capture participants’ imprecise recall responses for the cued item, instead of those with an extra-large recall error that could be attributed to other factors such as attentional lapses (deBettencourt et al., 2019). The two identified categories of trials together account for about 98% of the total trials (i.e. 176 out of 180 trials).

Figure 4

Download asset Open asset

We then performed the leave-one-block-out analysis to obtain trial-by-trial IEM reconstructions based on delay-period BOLD signals aggregated from the aLEC and DG/CA3. We averaged the IEM reconstructions from the small- and larger-error trials separately. Because trial counts between categories were not balanced, we resampled the data from the small-error trials based on the number of larger-error trials for 5000 times. We took the average of IEM reconstruction across iterations to obtain robust subject-level trial-average estimates with a balanced trial count across different behavioral trial types (Xie et al., 2020a; Yaffe et al., 2014). By contrasting these estimates at the subject level, we found that the small-error trials yielded significant IEM reconstructions for the cued item (t(15) = 4.50, p=4.21e-04, Cohen’s d=1.16, p_boostrap <0.001), whereas the larger-error trials did not (t(15) = 0.03, p=0.98, Cohen’s d=0.007, p_boostrap = 0.90; Figure 4B). Furthermore, the reconstructed WM information in the combined aLEC-DG/CA3 showed better quality in the small-error trials, as compared with that in the larger-error trials (t(15) = 2.45, p=0.027, Cohen’s d=0.61, p_boostrap = 0.032).

In addition to using an empirical criterion to separate in-memory trials from those extra-large error trials susceptible to occasional attentional lapses (deBettencourt et al., 2019), we have also tried another thresholding heuristic. As shown in Figure 1A, most trials from each participant fall within this 45° of absolute recall error (i.e. half of the 90° range), and the trials larger than this number are rare (~5 out of 180 trials). We, therefore, used 45° of absolute recall error as a cut-off to identify the imprecise recall trials that were greater than 20° but smaller than 45° of absolute recall error. We performed the same analysis to obtain trial-by-trial IEM reconstructions based on delay-period BOLD signals aggregated from the aLEC and DG/CA3 as outlined above, and then resampled the same number of trials to estimate the IEM reconstructions for the small-error and larger-error trials (<20° vs. 20° - 45° of absolute recall error). Consistent with the 3-SD heuristic, we found that the small-error trials identified by the 45° cut-off heuristic also yielded significant IEM reconstructions for the cued item (t(15) = 4.34, p=5.74e-04, Cohen’s d=1.12, p_boostrap <0.001), whereas the larger-error trials did not (t(15) = –0.69, p=0.50, Cohen’s d=–0.18, p_boostrap = 0.67). We then contrasted the difference in IEM reconstructions between these small- and large-error trials across participants. We found that IEM reconstruction for the cued item from the combined aLEC-DG/CA3 showed better quality in the small-error trials, as compared with that in the larger-error trials (t(15) = 3.41, p=0.004, Cohen’s d=0.88, p_boostrap = 0.008). Collectively, these results suggest that higher-quality WM representation in the entorhinal-DG/CA3 pathway during the delay period is associated with better subsequent recall fidelity and that this association is robust to the selection of cut-off scores for extra-large recall errors.

Based on high-resolution fMRI, this current study uncovers an often-neglected role of the MTL’s the entorhinal-DG/CA3 pathway in item-specific WM representation at a minimal task load. Our data suggest that the entorhinal-DG/CA3 circuitry retains item-specific information to allow fine discrimination of similar WM items across trials. The quality of item-specific WM information in the entorhinal-DG/CA3 pathway is associated with an observer’s subsequent recall fidelity. Together, these findings fill a missing link in the growing literature regarding the contribution of the MTL to item-level WM representation with a lower information load (Johnson et al., 2018; Sreenivasan and D’Esposito, 2019).

Theoretically, our findings are consistent with recent neural theories that highlight the involvement of distributed brain areas for WM (Christophel et al., 2017; Eriksson et al., 2015; Sreenivasan and D’Esposito, 2019), including mechanisms in the MTL that are traditionally deemed irrelevant for human WM (Beukers et al., 2021; Borders et al., 2022; Goodrich et al., 2019; Goodrich and Yonelinas, 2016). Our findings are built upon the established literature on the entorhinal-DG/CA3 circuitry and the formation of high-fidelity long-term episodic memory (Aimone et al., 2011; Bakker et al., 2008; Ekstrom and Yonelinas, 2020; Korkki et al., 2021; Leal and Yassa, 2018; Marr, 1971; Reagh and Yassa, 2014; Yassa and Stark, 2011). This function has been linked with various neuronal properties along the entorhinal-DG/CA3 pathway – such as abundant granule cells, strong inhibitory interneurons, and powerful mossy fiber synapses – which could enable sparse coding of information to minimize mnemonic interference (Aimone et al., 2011; Rolls, 2016; Rolls, 2013; Sahay et al., 2011). As such, similar information can be retained with a sufficient representational distance to support behavioral discrimination (Bakker et al., 2008; Burke et al., 2011; Hunsaker et al., 2008; Leal and Yassa, 2018; Leutgeb et al., 2007; Lohnas et al., 2018; Montchal et al., 2019; Reagh and Yassa, 2014; Ryan et al., 2012; Sakon and Suzuki, 2019). Our data suggest that the same MTL mechanism can also be used to support the quality of WM representation (Xie et al., 2020b). Conceptually, potential interference between items either across or within trials would place a demand on pattern separation even over a short delay (Oberauer and Lin, 2017). As such, the MTL circuitry involved in the resolution of mnemonic interference (Aimone et al., 2011) would play a key role in reducing inference between WM content and other similar information in the feature space. Our data suggest that this process would result in more similar and stable representations for the same remembered item across trials, as detected by multivariate correlational and decoding analyses. However, under certain task conditions (e.g. learning spatial routes in a naturalistic task over many repetitions), the MTL may maximally orthogonalize overlapping information to opposite representational patterns (hence ‘repulsion’) to minimize mnemonic interference (Chanales et al., 2017). It remains to be determined how these learning-related mechanisms in a more complex setting are related to MTL’s contributions to WM representation of simple stimulus features.

Empirically, our results have resolved an issue concerning the decodability of item-specific WM content in the MTL for simple stimulus features. Previously, MTL activity has been shown to scale with WM set size of letters and color squares without decodable item-specific WM content (Boran et al., 2022; Boran et al., 2019). When item information is shown, it often involves complex stimuli with rich information content (e.g. Kamiński et al., 2017). These observations raise the conceptual question concerning the extent to which the MTL responds to task difficulty or retains WM content. In other words, is the MTL not sensitive to simple stimuli or lower task demands at all? Here, with improved spatial resolution of MTL recordings and using a simple stimulus feature, our data suggest that the MTL retains item-level WM information even when the effective WM set size is one. The lack of significant observations in some previous studies using the same paradigm may be due to the lack of granularity in MTL recordings (e.g. Ester et al., 2015). To test this, we aggregated data from all the voxels in the hippocampus to examine whether blurred MTL signals would be sufficient to reveal item-specific WM content using the current IEM procedure. As CA1 and subiculum voxels contain less robust WM information (Figure 3B), we predicted that this aggregation procedure would attenuate the evidence for WM information due to the reduction in signal-to-noise ratio. Our data are in line with this prediction (Figure 3—figure supplement 4). These results, therefore, highlight the importance of fine-grained MTL signals in revealing item-specific WM content.

Alongside these theoretical and empirical contributions, our data also provide additional insights into the conditions under which the MTL is relevant for WM. First, our findings suggest that the MTL’s contribution to WM does not depend on whether task demands exceed a limited WM capacity (Jeneson and Squire, 2012), although this account has been proposed when interpreting some recent findings for WM tasks using complex stimuli or a higher memory set size (Boran et al., 2019; Jeneson and Squire, 2012; Kamiński et al., 2017; Kornblith et al., 2017; Libby et al., 2014). Second, our analysis has focused on the mid-delay activity (Postle et al., 2000) and hence our findings could not be explained by the MTL’s contribution to WM retrieval (Shrager et al., 2008). Furthermore, while our findings do not preclude the potential involvement of the MTL during perceptual encoding (Bonnen et al., 2021), perceptual involvement could not account for the results based on the comparison between the cued and uncued items (Bettencourt and Xu, 2016; Ester et al., 2015; Harrison and Tong, 2009). If the MTL primarily contributes to perceptual encoding instead of WM retention, we should have observed a comparable amount of information for both study items in the MTL, as they are presented in the same data acquisition TR before cue onset. Since participants do not know the cued item ahead of time, they need to initially remember both items. In line with this interpretation, a time-varying IEM analysis shows that aLEC-DG/CA3 indeed contains a comparable amount of information related to both the cued and uncued items at an earlier time point in the task (Figure 3—figure supplement 5). Yet, during the mid-delay period, aLEC-DG/CA3 contains significant information for the cued relative to the uncued item in a similar way as shown in the previous research (Ester et al., 2015; Harrison and Tong, 2009). Although it is well acknowledged that the current recording method has its inferential limitations in the time domain, these data unambiguously suggest that the entorhinal-DG/CA3 pathway supports the representation of a retrospectively selected memory item during a short delay – a hallmark of WM (Lorenc et al., 2021; Panichello and Buschman, 2021).

Several open questions remain to be addressed by future research. First, more data are needed to reveal how WM representation in the MTL is compared with and/or related to that retained in distributed neocortical areas (Christophel et al., 2017; Eriksson et al., 2015; Sreenivasan and D’Esposito, 2019). Although the IEM approach allows the reconstruction of information in neural signals, it is not well-suited to directly compare information reconstruction across brain regions. Such a comparison would be complicated by several issues, including the difference in the number of voxels involved and the lack of interpretability of null results when both brain regions contain some WM information. To improve interpretability, we have used the results based on the uncued item as a within-ROI control and contrasted how information specific to the cued item (cued vs. uncued) differs between MTL ROIs and a theoretically irrelevant control region (i.e. the amygdala). One additional potential approach is to examine how the representations of remembered items are correlated across brain regions (Figure 2—figure supplement 2A). The rationale is that delay-period neural patterns across trials should be correlated for two brain regions containing the same information (Pillet et al., 2019), as compared with brain regions that do not hold consistent information (Figure 2—figure supplement 2B). We tested this conjecture by examining the neural similarity across trials between the aLEC-DG/CA3 and a benchmark ROI in the superior temporal lobule (SPL) – a region that is consistently linked with item-specific information during visual WM retention both in the current data (Figure 3—figure supplement 4) and in the previous research (Bettencourt and Xu, 2016; Ester et al., 2015; Xu and Chun, 2006). Supporting this prediction, we found that the similarity of neural patterns between the aLEC-DG/CA3 and the SPL has increased from the pre-stimulus baseline to the WM retention period (Figure 2—figure supplement 2C), which contrasts with the lack of changes in the correlation of across-trial neural patterns between aLEC-DG/CA3 and the amygdala control ROI (Figure 2—figure supplement 2D). These data suggest that WM information in the entorhinal-DG/CA3 is similiar to that in a well-recognized neocortical WM-related area (Bettencourt and Xu, 2016; Ester et al., 2015; Xu and Chun, 2006). However, considering the limitation in temporal resolution of the current recording method, it remains unknown how the MTL contributes to the dynamic coding schemes underlying WM maintenance (Stokes, 2015). Future research with direct recordings from multiple brain areas would be more suitable to investigate the fine-scale temporal dynamic underlying these similar neural patterns across brain regions during WM.

Second, it remains unknown how the MTL circuitry is tuned to specific stimulus features such as orientations, although one of the analytical tools we used was inspired by findings based on neuronal tuning properties from the visual cortex (Brouwer and Heeger, 2009; Sprague et al., 2018). This is because the assumed orientation channels in IEM do not reflect the underlying neuronal tuning properties and are interpretable only within the assumed model (Liu et al., 2018; Sprague et al., 2018). Previous research using this method has therefore primarily focused on inferences related to the presence or absence of information content in the neural data (Bettencourt and Xu, 2016; Brouwer and Heeger, 2009; Ester et al., 2015; Rademaker et al., 2019; Sprague et al., 2016), instead of properties of neural tuning. In the current study, these IEM results are supported by the less assumption-laden results from stimulus-based representational similarity analysis (Kriegeskorte and Wei, 2021). These two approaches are therefore complementary to each other. Nevertheless, these analyses are correlational in nature. Hence, although fine-grained neural representations revealed by these analyses are associated with participants’ behavioral outcomes (Figure 4), it remains to be determined whether the entorhinal-DG/CA3 pathway contributes to the fidelity of WM representation or also to the process of information selection. Strategies for resolving this issue can involve generalizing the current findings to other WM tasks without an explicit requirement of retrospective information selection (Xie et al., 2023a) and/or further exploring how the frontal-parietal mechanisms related to visual selection and attention interact with the MTL system (Panichello and Buschman, 2021).

Third, the often-neglected role of the MTL in visual processing needs to be further explored. Our findings suggest that the entorhinal-DG/CA3 pathway in the MTL may play a role in retaining of task-relevant item-specific visual WM content, which could not be attributed to perceptual processing alone. These data adds to a growing body of literature that considers the MTL as an important part of the visual system, serving functions ranging from retinotopic coding (Knapen, 2021) to predictive coding (Hindy et al., 2016). Although retinotopic coding as a form of perceptual processing could underlie WM representation for orientation information, our data highlight that the MTL is sensitive to the retrospectively selected information – a hallmark of WM (Lorenc et al., 2021). Furthermore, while we have used orientation as a simple stimulus feature to minimize long-term memory influences, our results do not preclude the role of this MTL circuitry in remembering other stimulus features, such as colors (Xie et al., 2023a). To more precisely reveal the MTL mechanisms that are shared across WM and long-term memory, future research should examine the extent to which MTL voxels evoked by a long-term memory task (e.g., mnemonic similarity task, Bakker et al., 2008) can be directly used to directly decode mnemonic content in visual WM tasks using different simple stimulus features.

Non-identified data (e.g., MTL activities across ROIs and trial-by-trial behavior responses) and custom codes are available via the Open Science Framework repository (https://osf.io/zvdnr/).

1. 1. Xie W
   2. Cappiello M
   3. Yassa MA
   4. Ester E
   5. Zaghloul K
   6. Zhang W

   (2023) Open Science Framework

   ID zvdnr. The Entorhinal-DG/CA3 Pathway in the Medial Temporal Lobe Retains Visual Working Memory of a Simple Surface Feature.

   https://osf.io/zvdnr/

1. 1. Aimone JB
   2. Deng W
   3. Gage FH

   (2011) Resolving new memories: A critical look at the dentate gyrus, adult neurogenesis, and pattern separation

   Neuron 70:589–596.

   https://doi.org/10.1016/j.neuron.2011.05.010

   • PubMed
   • Google Scholar
2. 1. Atkinson RC
   2. Shiffrin RM

   (1968) Human memory: A proposed system and its control processes

   The Psychology of Learning and Motivation 2:89–195.

   https://doi.org/10.1016/S0079-7421(08)60422-3

   • Google Scholar
3. 1. Avants B
   2. Duda JT
   3. Kim J
   4. Zhang H
   5. Pluta J
   6. Gee JC
   7. Whyte J

   (2008) Multivariate analysis of structural and diffusion imaging in traumatic brain injury

   Academic Radiology 15:1360–1375.

   https://doi.org/10.1016/j.acra.2008.07.007

   • PubMed
   • Google Scholar
4. 1. Baddeley AD

   (2012) Working memory: theories, models, and controversies

   Annual Review of Psychology 63:1–29.

   https://doi.org/10.1146/annurev-psych-120710-100422

   • PubMed
   • Google Scholar
5. 1. Bakker A
   2. Kirwan CB
   3. Miller M
   4. Stark CEL

   (2008) Pattern separation in the human hippocampal CA3 and dentate gyrus

   Science 319:1640–1642.

   https://doi.org/10.1126/science.1152882

   • PubMed
   • Google Scholar
6. 1. Barense MD
   2. Gaffan D
   3. Graham KS

   (2007) The human medial temporal lobe processes online representations of complex objects

   Neuropsychologia 45:2963–2974.

   https://doi.org/10.1016/j.neuropsychologia.2007.05.023

   • PubMed
   • Google Scholar
7. 1. Bettencourt KC
   2. Xu Y

   (2016) Decoding the content of visual short-term memory under distraction in occipital and parietal areas

   Nature Neuroscience 19:150–157.

   https://doi.org/10.1038/nn.4174

   • PubMed
   • Google Scholar
8. 1. Beukers AO
   2. Buschman TJ
   3. Cohen JD
   4. Norman KA

   (2021) Is activity silent working memory simply episodic memory?

   Trends in Cognitive Sciences 25:284–293.

   https://doi.org/10.1016/j.tics.2021.01.003

   • PubMed
   • Google Scholar
9. 1. Bonnen T
   2. Yamins DLK
   3. Wagner AD

   (2021) When the ventral visual stream is not enough: a deep learning account of medial temporal lobe involvement in perception

   Neuron 109:2755–2766.

   https://doi.org/10.1016/j.neuron.2021.06.018

   • PubMed
   • Google Scholar
10. 1. Boran E
    2. Fedele T
    3. Klaver P
    4. Hilfiker P
    5. Stieglitz L
    6. Grunwald T
    7. Sarnthein J

    (2019) Persistent hippocampal neural firing and hippocampal-cortical coupling predict verbal working memory load

    Science Advances 5:eaav3687.

    https://doi.org/10.1126/sciadv.aav3687

    • PubMed
    • Google Scholar
11. 1. Boran E
    2. Hilfiker P
    3. Stieglitz L
    4. Sarnthein J
    5. Klaver P

    (2022) Persistent neuronal firing in the medial temporal lobe supports performance and workload of visual working memory in humans

    NeuroImage 254:119123.

    https://doi.org/10.1016/j.neuroimage.2022.119123

    • PubMed
    • Google Scholar
12. 1. Borders AA
    2. Ranganath C
    3. Yonelinas AP

    (2022) The hippocampus supports high-precision binding in visual working memory

    Hippocampus 32:217–230.

    https://doi.org/10.1002/hipo.23401

    • PubMed
    • Google Scholar
13. 1. Brouwer GJ
    2. Heeger DJ

    (2009) Decoding and reconstructing color from responses in human visual cortex

    The Journal of Neuroscience 29:13992–14003.

    https://doi.org/10.1523/JNEUROSCI.3577-09.2009

    • PubMed
    • Google Scholar
14. 1. Burke SN
    2. Wallace JL
    3. Hartzell AL
    4. Nematollahi S
    5. Plange K
    6. Barnes CA

    (2011) Age-associated deficits in pattern separation functions of the perirhinal cortex: A cross-species consensus

    Behavioral Neuroscience 125:836–847.

    https://doi.org/10.1037/a0026238

    • PubMed
    • Google Scholar
15. 1. Cappiello M
    2. Xie W
    3. David A
    4. Bikson M
    5. Zhang W

    (2016) Transcranial direct current stimulation modulates pattern separation

    Neuroreport 27:826–832.

    https://doi.org/10.1097/WNR.0000000000000621

    • PubMed
    • Google Scholar
16. 1. Chanales AJH
    2. Oza A
    3. Favila SE
    4. Kuhl BA

    (2017) Overlap among spatial memories triggers repulsion of hippocampal representations

    Current Biology 27:2307–2317.

    https://doi.org/10.1016/j.cub.2017.06.057

    • PubMed
    • Google Scholar
17. 1. Chen G
    2. Saad ZS
    3. Adleman NE
    4. Leibenluft E
    5. Cox RW

    (2015) Detecting the subtle shape differences in hemodynamic responses at the group level

    Frontiers in Neuroscience 9:375.

    https://doi.org/10.3389/fnins.2015.00375

    • PubMed
    • Google Scholar
18. 1. Christophel TB
    2. Klink PC
    3. Spitzer B
    4. Roelfsema PR
    5. Haynes JD

    (2017) The distributed nature of working memory

    Trends in Cognitive Sciences 21:111–124.

    https://doi.org/10.1016/j.tics.2016.12.007

    • PubMed
    • Google Scholar
19. 1. Cowan N

    (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity

    Behavioral and Brain Sciences 24:87–114.

    https://doi.org/10.1017/S0140525X01003922

    • Google Scholar
20. 1. Cox RW

    (1996) AFNI: software for analysis and visualization of functional magnetic resonance NeuroImages

    Computers and Biomedical Research 29:162–173.

    https://doi.org/10.1006/cbmr.1996.0014

    • PubMed
    • Google Scholar
21. 1. deBettencourt MT
    2. Keene PA
    3. Awh E
    4. Vogel EK

    (2019) Real-time triggering reveals concurrent lapses of attention and working memory

    Nature Human Behaviour 3:808–816.

    https://doi.org/10.1038/s41562-019-0606-6

    • PubMed
    • Google Scholar
22. 1. Eichenbaum H
    2. Yonelinas AP
    3. Ranganath C

    (2007) The medial temporal lobe and recognition memory

    Annual Review of Neuroscience 30:123–152.

    https://doi.org/10.1146/annurev.neuro.30.051606.094328

    • PubMed
    • Google Scholar
23. 1. Ekstrom AD
    2. Yonelinas AP

    (2020) Precision, binding, and the hippocampus: precisely what are we talking about?

    Neuropsychologia 138:107341.

    https://doi.org/10.1016/j.neuropsychologia.2020.107341

    • PubMed
    • Google Scholar
24. 1. Eriksson J
    2. Vogel EK
    3. Lansner A
    4. Bergström F
    5. Nyberg L

    (2015) Neurocognitive architecture of working memory

    Neuron 88:33–46.

    https://doi.org/10.1016/j.neuron.2015.09.020

    • PubMed
    • Google Scholar
25. 1. Ester EF
    2. Sprague TC
    3. Serences JT

    (2015) Parietal and frontal cortex encode stimulus-specific mnemonic representations during visual working memory

    Neuron 87:893–905.

    https://doi.org/10.1016/j.neuron.2015.07.013

    • PubMed
    • Google Scholar
26. 1. Feredoes E
    2. Postle BR

    (2007) Localization of load sensitivity of working memory storage: quantitatively and qualitatively discrepant results yielded by single-subject and group-averaged approaches to fmri group analysis

    NeuroImage 35:881–903.

    https://doi.org/10.1016/j.neuroimage.2006.12.029

    • PubMed
    • Google Scholar
27. 1. Gilbert PE
    2. Kesner RP

    (2006) The role of the dorsal CA3 hippocampal subregion in spatial working memory and pattern separation

    Behavioural Brain Research 169:142–149.

    https://doi.org/10.1016/j.bbr.2006.01.002

    • PubMed
    • Google Scholar
28. Book

    1. Good P

    (2013) Permutation Tests: A Practical Guide to Resampling Methods for Testing Hypotheses

    New York, NY: Springer Science & Business Media.

    https://doi.org/10.1007/978-1-4757-2346-5

    • Google Scholar
29. 1. Goodrich RI
    2. Yonelinas AP

    (2016) The medial temporal lobe supports sensing-based visual working memory

    Neuropsychologia 89:485–494.

    https://doi.org/10.1016/j.neuropsychologia.2016.07.011

    • PubMed
    • Google Scholar
30. 1. Goodrich RI
    2. Baer TL
    3. Quent JA
    4. Yonelinas AP

    (2019) Visual working memory impairments for single items following medial temporal lobe damage

    Neuropsychologia 134:107227.

    https://doi.org/10.1016/j.neuropsychologia.2019.107227

    • PubMed
    • Google Scholar
31. 1. Hannula DE
    2. Ranganath C

    (2008) Medial temporal lobe activity predicts successful relational memory binding

    The Journal of Neuroscience 28:116–124.

    https://doi.org/10.1523/JNEUROSCI.3086-07.2008

    • PubMed
    • Google Scholar
32. 1. Harrison SA
    2. Tong F

    (2009) Decoding reveals the contents of visual working memory in early visual areas

    Nature 458:632–635.

    https://doi.org/10.1038/nature07832

    • PubMed
    • Google Scholar
33. 1. Hindy NC
    2. Ng FY
    3. Turk-Browne NB

    (2016) Linking pattern completion in the hippocampus to predictive coding in visual cortex

    Nature Neuroscience 19:665–667.

    https://doi.org/10.1038/nn.4284

    • PubMed
    • Google Scholar
34. 1. Hunsaker MR
    2. Rosenberg JS
    3. Kesner RP

    (2008) The role of the dentate gyrus, ca3a,b, and ca3c for detecting spatial and environmental novelty

    Hippocampus 18:1064–1073.

    https://doi.org/10.1002/hipo.20464

    • PubMed
    • Google Scholar
35. 1. Iglesias JE
    2. Augustinack JC
    3. Nguyen K
    4. Player CM
    5. Player A
    6. Wright M
    7. Roy N
    8. Frosch MP
    9. McKee AC
    10. Wald LL
    11. Fischl B
    12. Van Leemput K
    13. Initiative ADN

    (2015) A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: application to adaptive segmentation of in vivo MRI

    NeuroImage 115:117–137.

    https://doi.org/10.1016/j.neuroimage.2015.04.042

    • PubMed
    • Google Scholar
36. 1. Irfanoglu MO
    2. Modi P
    3. Nayak A
    4. Hutchinson EB
    5. Sarlls J
    6. Pierpaoli C

    (2015) DR-BUDDI (diffeomorphic registration for blip-up blip-down diffusion imaging) method for correcting echo planar imaging distortions

    NeuroImage 106:284–299.

    https://doi.org/10.1016/j.neuroimage.2014.11.042

    • PubMed
    • Google Scholar
37. 1. Iwai E
    2. Yukie M
    3. Watanabe J
    4. Hikosaka K
    5. Suyama H
    6. Ishikawa S

    (1990) A role of amygdala in visual perception and cognition in macaque monkeys (Macaca fuscata and Macaca mulatta)

    The Tohoku Journal of Experimental Medicine 161:95–120.

    https://doi.org/10.1620/tjem.161.Supplement_95

    • Google Scholar
38. 1. Jeneson A
    2. Squire LR

    (2012) Working memory, long-term memory, and medial temporal lobe function

    Learning & Memory 19:15–25.

    https://doi.org/10.1101/lm.024018.111

    • PubMed
    • Google Scholar
39. 1. Johnson EL
    2. Adams JN
    3. Solbakk AK
    4. Endestad T
    5. Larsson PG
    6. Ivanovic J
    7. Meling TR
    8. Lin JJ
    9. Knight RT

    (2018) Dynamic frontotemporal systems process space and time in working memory

    PLOS Biology 16:e2004274.

    https://doi.org/10.1371/journal.pbio.2004274

    • PubMed
    • Google Scholar
40. 1. Kamiński J
    2. Sullivan S
    3. Chung JM
    4. Ross IB
    5. Mamelak AN
    6. Rutishauser U

    (2017) Persistently active neurons in human medial frontal and medial temporal lobe support working memory

    Nature Neuroscience 20:590–601.

    https://doi.org/10.1038/nn.4509

    • PubMed
    • Google Scholar
41. 1. Kamiński J
    2. Rutishauser U

    (2020) Between persistently active and activity-silent frameworks: novel vistas on the cellular basis of working memory

    Annals of the New York Academy of Sciences 1464:64–75.

    https://doi.org/10.1111/nyas.14213

    • PubMed
    • Google Scholar
42. 1. Knapen T

    (2021) Topographic connectivity reveals task-dependent retinotopic processing throughout the human brain

    PNAS 118:1–6.

    https://doi.org/10.1073/pnas.2017032118

    • PubMed
    • Google Scholar
43. 1. Koen JD
    2. Borders AA
    3. Petzold MT
    4. Yonelinas AP

    (2017) Visual short-term memory for high resolution associations is impaired in patients with medial temporal lobe damage

    Hippocampus 27:184–193.

    https://doi.org/10.1002/hipo.22682

    • PubMed
    • Google Scholar
44. 1. Korkki SM
    2. Richter FR
    3. Simons JS

    (2021) Hippocampal-Cortical encoding activity predicts the precision of episodic memory

    Journal of Cognitive Neuroscience 33:2328–2341.

    https://doi.org/10.1162/jocn_a_01770

    • PubMed
    • Google Scholar
45. 1. Kornblith S
    2. Quian Quiroga R
    3. Koch C
    4. Fried I
    5. Mormann F

    (2017) Persistent single-neuron activity during working memory in the human medial temporal lobe

    Current Biology 27:1026–1032.

    https://doi.org/10.1016/j.cub.2017.02.013

    • PubMed
    • Google Scholar
46. 1. Kriegeskorte N
    2. Diedrichsen J

    (2019) Peeling the onion of brain representations

    Annual Review of Neuroscience 42:407–432.

    https://doi.org/10.1146/annurev-neuro-080317-061906

    • PubMed
    • Google Scholar
47. 1. Kriegeskorte N
    2. Wei XX

    (2021) Neural tuning and representational geometry

    Nature Reviews. Neuroscience 22:703–718.

    https://doi.org/10.1038/s41583-021-00502-3

    • PubMed
    • Google Scholar
48. 1. Leal SL
    2. Yassa MA

    (2018) Integrating new findings and examining clinical applications of pattern separation

    Nature Neuroscience 21:163–173.

    https://doi.org/10.1038/s41593-017-0065-1

    • PubMed
    • Google Scholar
49. 1. Leutgeb JK
    2. Leutgeb S
    3. Moser MB
    4. Moser EI

    (2007) Pattern separation in the dentate gyrus and CA3 of the hippocampus

    Science 315:961–966.

    https://doi.org/10.1126/science.1135801

    • PubMed
    • Google Scholar
50. 1. Lewis-Peacock JA
    2. Postle BR

    (2008) Temporary activation of long-term memory supports working memory

    The Journal of Neuroscience 28:8765–8771.

    https://doi.org/10.1523/JNEUROSCI.1953-08.2008

    • PubMed
    • Google Scholar
51. 1. Libby LA
    2. Hannula DE
    3. Ranganath C

    (2014) Medial temporal lobe coding of item and spatial information during relational binding in working memory

    The Journal of Neuroscience 34:14233–14242.

    https://doi.org/10.1523/JNEUROSCI.0655-14.2014

    • PubMed
    • Google Scholar
52. 1. Liu T
    2. Cable D
    3. Gardner JL

    (2018) Inverted encoding models of human population response conflate noise and neural tuning width

    The Journal of Neuroscience 38:398–408.

    https://doi.org/10.1523/JNEUROSCI.2453-17.2017

    • PubMed
    • Google Scholar
53. 1. Liu J
    2. Zhang H
    3. Yu T
    4. Ni D
    5. Ren L
    6. Yang Q
    7. Lu B
    8. Wang D
    9. Heinen R
    10. Axmacher N
    11. Xue G

    (2020) Stable maintenance of multiple representational formats in human visual short-term memory

    PNAS 117:32329–32339.

    https://doi.org/10.1073/pnas.2006752117

    • PubMed
    • Google Scholar
54. 1. Lohnas LJ
    2. Duncan K
    3. Doyle WK
    4. Thesen T
    5. Devinsky O
    6. Davachi L

    (2018) Time-resolved neural reinstatement and pattern separation during memory decisions in human hippocampus

    PNAS 115:E7418–E7427.

    https://doi.org/10.1073/pnas.1717088115

    • PubMed
    • Google Scholar
55. 1. Lorenc ES
    2. Mallett R
    3. Lewis-Peacock JA

    (2021) Distraction in visual working memory: resistance is not futile

    Trends in Cognitive Sciences 25:228–239.

    https://doi.org/10.1016/j.tics.2020.12.004

    • PubMed
    • Google Scholar
56. 1. Lundqvist M
    2. Herman P
    3. Miller EK

    (2018) Working memory: delay activity, yes! persistent activity? maybe not

    The Journal of Neuroscience 38:7013–7019.

    https://doi.org/10.1523/JNEUROSCI.2485-17.2018

    • PubMed
    • Google Scholar
57. 1. Maass A
    2. Berron D
    3. Libby LA
    4. Ranganath C
    5. Düzel E

    (2015) Functional subregions of the human entorhinal cortex

    eLife 4:06426.

    https://doi.org/10.7554/eLife.06426

    • PubMed
    • Google Scholar
58. 1. Marr D

    (1971) Simple memory: A theory for archicortex

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 262:23–81.

    https://doi.org/10.1098/rstb.1971.0078

    • PubMed
    • Google Scholar
59. 1. Milner B
    2. Corkin S
    3. Teuber HL

    (1968) Further analysis of the hippocampal amnesic syndrome: 14-year follow-up study of H.M

    Neuropsychologia 6:215–234.

    https://doi.org/10.1016/0028-3932(68)90021-3

    • Google Scholar
60. 1. Montchal ME
    2. Reagh ZM
    3. Yassa MA

    (2019) Precise temporal memories are supported by the lateral entorhinal cortex in humans

    Nature Neuroscience 22:284–288.

    https://doi.org/10.1038/s41593-018-0303-1

    • PubMed
    • Google Scholar
61. 1. Nee DE
    2. Jonides J

    (2008) Neural correlates of access to short-term memory

    PNAS 105:14228–14233.

    https://doi.org/10.1073/pnas.0802081105

    • PubMed
    • Google Scholar
62. 1. Norris D

    (2017) Short-term memory and long-term memory are still different

    Psychological Bulletin 143:992–1009.

    https://doi.org/10.1037/bul0000108

    • PubMed
    • Google Scholar
63. 1. Oberauer K
    2. Lin HY

    (2017) An interference model of visual working memory

    Psychological Review 124:21–59.

    https://doi.org/10.1037/rev0000044

    • PubMed
    • Google Scholar
64. 1. Olson IR
    2. Moore KS
    3. Stark M
    4. Chatterjee A

    (2006) Visual working memory is impaired when the medial temporal lobe is damaged

    Journal of Cognitive Neuroscience 18:1087–1097.

    https://doi.org/10.1162/jocn.2006.18.7.1087

    • PubMed
    • Google Scholar
65. 1. Panichello MF
    2. Buschman TJ

    (2021) Shared mechanisms underlie the control of working memory and attention

    Nature 592:601–605.

    https://doi.org/10.1038/s41586-021-03390-w

    • PubMed
    • Google Scholar
66. 1. Pillet I
    2. Op de Beeck H
    3. Lee Masson H

    (2019) A comparison of functional networks derived from representational similarity, functional connectivity, and univariate analyses

    Frontiers in Neuroscience 13:1348.

    https://doi.org/10.3389/fnins.2019.01348

    • PubMed
    • Google Scholar
67. 1. Postle BR
    2. Zarahn E
    3. D’Esposito M

    (2000) Using event-related fmri to assess delay-period activity during performance of spatial and nonspatial working memory tasks

    Brain Research. Brain Research Protocols 5:57–66.

    https://doi.org/10.1016/s1385-299x(99)00053-7

    • PubMed
    • Google Scholar
68. 1. Rademaker RL
    2. Chunharas C
    3. Serences JT

    (2019) Coexisting representations of sensory and mnemonic information in human visual cortex

    Nature Neuroscience 22:1336–1344.

    https://doi.org/10.1038/s41593-019-0428-x

    • PubMed
    • Google Scholar
69. 1. Reagh ZM
    2. Yassa MA

    (2014) Object and spatial mnemonic interference differentially engage lateral and medial entorhinal cortex in humans

    PNAS 111:E4264–E4273.

    https://doi.org/10.1073/pnas.1411250111

    • PubMed
    • Google Scholar
70. 1. Reagh ZM
    2. Murray EA
    3. Yassa MA

    (2017) Repetition reveals ups and downs of hippocampal, thalamic, and neocortical engagement during mnemonic decisions

    Hippocampus 27:169–183.

    https://doi.org/10.1002/hipo.22681

    • PubMed
    • Google Scholar
71. 1. Reagh ZM
    2. Noche JA
    3. Tustison NJ
    4. Delisle D
    5. Murray EA
    6. Yassa MA

    (2018) Functional imbalance of anterolateral entorhinal cortex and hippocampal dentate/CA3 underlies age-related object pattern separation deficits

    Neuron 97:1187–1198.

    https://doi.org/10.1016/j.neuron.2018.01.039

    • PubMed
    • Google Scholar
72. 1. Rissman J
    2. Gazzaley A
    3. D’Esposito M

    (2008) Dynamic adjustments in prefrontal, hippocampal, and inferior temporal interactions with increasing visual working memory load

    Cerebral Cortex 18:1618–1629.

    https://doi.org/10.1093/cercor/bhm195

    • PubMed
    • Google Scholar
73. 1. Rolls ET

    (2013) A quantitative theory of the functions of the hippocampal CA3 network in memory

    Frontiers in Cellular Neuroscience 7:98.

    https://doi.org/10.3389/fncel.2013.00098

    • PubMed
    • Google Scholar
74. 1. Rolls ET

    (2016) Pattern separation, completion, and categorisation in the hippocampus and neocortex

    Neurobiology of Learning and Memory 129:4–28.

    https://doi.org/10.1016/j.nlm.2015.07.008

    • PubMed
    • Google Scholar
75. 1. Rose NS
    2. LaRocque JJ
    3. Riggall AC
    4. Gosseries O
    5. Starrett MJ
    6. Meyering EE
    7. Postle BR

    (2016) Reactivation of latent working memories with transcranial magnetic stimulation

    Science 354:1136–1139.

    https://doi.org/10.1126/science.aah7011

    • PubMed
    • Google Scholar
76. 1. Rosenthal R
    2. Rubin DB

    (1983) Ensemble-adjusted P values

    Psychological Bulletin 94:540–541.

    https://doi.org/10.1037/0033-2909.94.3.540

    • Google Scholar
77. 1. Ruchkin DS
    2. Grafman J
    3. Cameron K
    4. Berndt RS

    (2003) Working memory retention systems: a state of activated long-term memory

    Behavioral and Brain Sciences 26:709–728.

    https://doi.org/10.1017/S0140525X03000165

    • Google Scholar
78. 1. Ryan L
    2. Cardoza JA
    3. Barense MD
    4. Kawa KH
    5. Wallentin-Flores J
    6. Arnold WT
    7. Alexander GE

    (2012) Age-related impairment in a complex object discrimination task that engages perirhinal cortex

    Hippocampus 22:1978–1989.

    https://doi.org/10.1002/hipo.22069

    • PubMed
    • Google Scholar
79. 1. Sahay A
    2. Wilson DA
    3. Hen R

    (2011) Pattern separation: A common function for new neurons in hippocampus and olfactory bulb

    Neuron 70:582–588.

    https://doi.org/10.1016/j.neuron.2011.05.012

    • PubMed
    • Google Scholar
80. 1. Sakon JJ
    2. Suzuki WA

    (2019) A neural signature of pattern separation in the monkey hippocampus

    PNAS 116:9634–9643.

    https://doi.org/10.1073/pnas.1900804116

    • Google Scholar
81. 1. Saygin ZM
    2. Kliemann D
    3. Iglesias JE
    4. van der Kouwe AJW
    5. Boyd E
    6. Reuter M
    7. Stevens A
    8. Van Leemput K
    9. McKee A
    10. Frosch MP
    11. Fischl B
    12. Augustinack JC
    13. Alzheimer’s Disease Neuroimaging Initiative

    (2017) High-resolution magnetic resonance imaging reveals nuclei of the human amygdala: manual segmentation to automatic atlas

    NeuroImage 155:370–382.

    https://doi.org/10.1016/j.neuroimage.2017.04.046

    • PubMed
    • Google Scholar
82. 1. Scoville WB
    2. Milner B

    (1957) Loss of recent memory after bilateral hippocampal lesions

    Journal of Neurology, Neurosurgery, and Psychiatry 20:11–21.

    https://doi.org/10.1136/jnnp.20.1.11

    • PubMed
    • Google Scholar
83. 1. Shrager Y
    2. Levy DA
    3. Hopkins RO
    4. Squire LR

    (2008) Working memory and the organization of brain systems

    The Journal of Neuroscience 28:4818–4822.

    https://doi.org/10.1523/JNEUROSCI.0710-08.2008

    • PubMed
    • Google Scholar
84. 1. Sprague TC
    2. Ester EF
    3. Serences JT

    (2016) Restoring latent visual working memory representations in human cortex

    Neuron 91:694–707.

    https://doi.org/10.1016/j.neuron.2016.07.006

    • PubMed
    • Google Scholar
85. 1. Sprague TC
    2. Adam KCS
    3. Foster JJ
    4. Rahmati M
    5. Sutterer DW
    6. Vo VA

    (2018) Inverted encoding models assay population-level stimulus representations, not single-unit neural tuning

    ENeuro 5:ENEURO.0098-18.2018.

    https://doi.org/10.1523/ENEURO.0098-18.2018

    • PubMed
    • Google Scholar
86. 1. Squire LR
    2. Zola-Morgan S

    (1991) The medial temporal lobe memory system

    Science 253:1380–1386.

    https://doi.org/10.1126/science.1896849

    • PubMed
    • Google Scholar
87. 1. Sreenivasan KK
    2. D’Esposito M

    (2019) The what, where and how of delay activity

    Nature Reviews. Neuroscience 20:466–481.

    https://doi.org/10.1038/s41583-019-0176-7

    • PubMed
    • Google Scholar
88. 1. Stokes MG

    (2015) “ Activity-silent ” working memory in prefrontal cortex: a dynamic coding framework

    Trends in Cognitive Sciences 19:394–405.

    https://doi.org/10.1016/j.tics.2015.05.004

    • PubMed
    • Google Scholar
89. 1. Todd JJ
    2. Marois R

    (2004) Capacity limit of visual short-term memory in human posterior parietal cortex

    Nature 428:751–754.

    https://doi.org/10.1038/nature02466

    • PubMed
    • Google Scholar
90. 1. Warren DE
    2. Duff MC
    3. Cohen NJ
    4. Tranel D

    (2014) Hippocampus contributes to the maintenance but not the quality of visual information over time

    Learning & Memory 22:6–10.

    https://doi.org/10.1101/lm.037127.114

    • PubMed
    • Google Scholar
91. 1. Xie W
    2. Zhang W

    (2016) Negative emotion boosts quality of visual working memory representation

    Emotion 16:760–774.

    https://doi.org/10.1037/emo0000159

    • PubMed
    • Google Scholar
92. 1. Xie W
    2. Zhang W

    (2017a) Negative emotion enhances mnemonic precision and subjective feelings of remembering in visual long-term memory

    Cognition 166:73–83.

    https://doi.org/10.1016/j.cognition.2017.05.025

    • PubMed
    • Google Scholar
93. 1. Xie W
    2. Zhang W

    (2017b) Dissociations of the number and precision of visual short-term memory representations in change detection

    Memory & Cognition 45:1423–1437.

    https://doi.org/10.3758/s13421-017-0739-7

    • PubMed
    • Google Scholar
94. 1. Xie W
    2. Cappiello M
    3. Meng M
    4. Rosenthal R
    5. Zhang W

    (2018) ADRA2B deletion variant and enhanced cognitive processing of emotional information: A meta-analytical review

    Neuroscience and Biobehavioral Reviews 92:402–416.

    https://doi.org/10.1016/j.neubiorev.2018.05.010

    • PubMed
    • Google Scholar
95. 1. Xie W
    2. Bainbridge WA
    3. Inati SK
    4. Baker CI
    5. Zaghloul KA

    (2020a) Memorability of words in arbitrary verbal associations modulates memory retrieval in the anterior temporal lobe

    Nature Human Behaviour 4:937–948.

    https://doi.org/10.1038/s41562-020-0901-2

    • PubMed
    • Google Scholar
96. 1. Xie W
    2. Park HB
    3. Zaghloul KA
    4. Zhang W

    (2020b) Correlated individual differences in the estimated precision of working memory and long-term memory: commentary on the study by biderman, luria, teodorescu, hajaj, and goshen-gottstein (2019)

    Psychological Science 31:345–348.

    https://doi.org/10.1177/0956797620903718

    • PubMed
    • Google Scholar
97. 1. Xie W
    2. Zaghloul KA

    (2021) Visual and semantic contributions to visual short-term memory

    Trends in Cognitive Sciences 25:270–271.

    https://doi.org/10.1016/j.tics.2021.01.010

    • PubMed
    • Google Scholar
98. 1. Xie W
    2. Ye C
    3. Zhang W

    (2022) Negative emotion reduces visual working memory recall variability: A meta-analytical review

    Emotion 11:0001139.

    https://doi.org/10.1037/emo0001139

    • PubMed
    • Google Scholar
99. 1. Xie W
    2. Chapeton JI
    3. Bhasin S
    4. Zawora C
    5. Wittig JH
    6. Inati SK
    7. Zhang W
    8. Zaghloul KA

    (2023a) The medial temporal lobe supports the quality of visual short-term memory representation

    Nature Human Behaviour 33:01529-5.

    https://doi.org/10.1038/s41562-023-01529-5

    • Google Scholar
100. Book

     1. Xie W
     2. Wittig JH
     3. Zaghloul KA

     (2023b) What do I need to consider for multivariate analysis of ieeg data

     In: Axmacher N, editors. Intracranial EEG: A Guide for Cognitive Neuroscientists. Switzerland AG: Springer Nature: Springer Nature. pp. 1–10.

     https://doi.org/10.1007/978-3-031-20910-9_34

     • Google Scholar
101. 1. Xu Y
     2. Chun MM

     (2006) Dissociable neural mechanisms supporting visual short-term memory for objects

     Nature 440:91–95.

     https://doi.org/10.1038/nature04262

     • PubMed
     • Google Scholar
102. 1. Yaffe RB
     2. Kerr MSD
     3. Damera S
     4. Sarma SV
     5. Inati SK
     6. Zaghloul KA

     (2014) Reinstatement of distributed cortical oscillations occurs with precise spatiotemporal dynamics during successful memory retrieval

     PNAS 111:18727–18732.

     https://doi.org/10.1073/pnas.1417017112

     • PubMed
     • Google Scholar
103. 1. Yassa MA
     2. Stark CEL

     (2011) Pattern separation in the hippocampus

     Trends in Neurosciences 34:515–525.

     https://doi.org/10.1016/j.tins.2011.06.006

     • PubMed
     • Google Scholar
104. 1. Yushkevich PA
     2. Amaral RSC
     3. Augustinack JC
     4. Bender AR
     5. Bernstein JD
     6. Boccardi M
     7. Bocchetta M
     8. Burggren AC
     9. Carr VA
     10. Chakravarty MM
     11. Chételat G
     12. Daugherty AM
     13. Davachi L
     14. Ding S-L
     15. Ekstrom A
     16. Geerlings MI
     17. Hassan A
     18. Huang Y
     19. Iglesias JE
     20. La Joie R
     21. Kerchner GA
     22. LaRocque KF
     23. Libby LA
     24. Malykhin N
     25. Mueller SG
     26. Olsen RK
     27. Palombo DJ
     28. Parekh MB
     29. Pluta JB
     30. Preston AR
     31. Pruessner JC
     32. Ranganath C
     33. Raz N
     34. Schlichting ML
     35. Schoemaker D
     36. Singh S
     37. Stark CEL
     38. Suthana N
     39. Tompary A
     40. Turowski MM
     41. Van Leemput K
     42. Wagner AD
     43. Wang L
     44. Winterburn JL
     45. Wisse LEM
     46. Yassa MA
     47. Zeineh MM
     48. Hippocampal Subfields Group (HSG)

     (2015) Quantitative comparison of 21 protocols for labeling hippocampal subfields and parahippocampal subregions in in vivo MRI: towards a harmonized segmentation protocol

     NeuroImage 111:526–541.

     https://doi.org/10.1016/j.neuroimage.2015.01.004

     • PubMed
     • Google Scholar
105. 1. Zhang W
     2. Luck SJ

     (2008) Discrete fixed-resolution representations in visual working memory

     Nature 453:233–235.

     https://doi.org/10.1038/nature06860

     • PubMed
     • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "The Entorhinal-DG/CA3 Pathway in the Medial Temporal Lobe Retains Visual Working Memory of a Simple Surface Feature" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Floris de Lange as the Senior Editor. The reviewers have opted to remain anonymous.

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

Essential revisions:

1) The reviewers raised several issues of interpretation:

a) It is not clear whether the authors propose that the entorhinal-DG/CA3 circuitry retains specific visual feature (i.e., orientation and not other features) or enables discrimination among similar items (item1 vs. item2 independent of maintained feature) (see R1).

b) It is unclear whether the more distinct representation in MTL for the cued item may reflect hippocampal repulseion (see R2).

The authors should discuss these issues. We would welcome it if the authors could corroborate their interpretation with new data/analysis, but otherwise it would be important that the interpretational limitations are acknowledged in the manuscript.

2) The application of IEM methods should be motivated more clearly in the manuscript (see R2).

3) The behavioral data should be analyzed in greater detail (see R3).

Please see below for a full description of issues raised by the reviewers.

Reviewer #1 (Recommendations for the authors):

The study by Xie et al., investigates whether the entorhinal-DG/CA3 pathway is involved in working memory maintenance. The study consists of 16 participants who perform an orientation visual working memory task with a retro-cue. Briefly, participants are presented with two gratings each with a different orientation followed by a cue that indicates which orientation should be further maintained. The authors focused their analyses on the delay interval following the cue. The main findings include a correlation between stimulus and neural similarities that was specific for cued stimulus and entorhinal-DG/CA3 locations. The authors observed similar results (cuing and region specificity) using inverted encoding modeling approach. Finally, they also showed that trials in which participants made a smaller error showed a better reconstruction fidelity on the cued side (compared to un-cued). This effect was absent for larger-error trials.

The study challenges a widely held traditional view that working memory and episodic memory have largely independent neural implementations. Specifically, this traditional view posits that the MTL is critical for episodic memory but not for working memory. The study adds to a large body of evidence showing involvement of the hippocampus across a range of different working memory tasks and stimuli. Nevertheless, it still remains unclear what functions may the hippocampus play in working memory.

1) The authors interpret their findings as suggesting that the entorhinal-DG/CA3 circuitry retains item-specific information to allow fine discrimination of similar WM items across trials. It was not clear to me whether they propose that the entorhinal-DG/CA3 circuitry retains specific visual feature (i.e., orientation and not other features) or enables discrimination among similar items (item1 vs. item2 independent of maintained feature). One other alternative interpretation of their results is that the entorhinal-DG/CA3 circuitry participates in orienting attention to memory representation or otherwise supports selection between initially remembered orientations. The current approach cannot tell these alternatives apart which is the main limitation of the study.

2) If the authors propose that the hippocampus maintains item-specific information about orientation (p. 9), it would be important to show specificity and also discuss how hippocampal contribution to working memory differs from the contribution of the visual system (e.g., Harrison and Tong, 2009)? On the other hand if authors propose that hippocampus enables discriminating among similar items independent of maintained feature, it would be important to show whether this generalize to other features.

3) Furthermore, different features have been shown to elicit sustained or rhythmic neural activity in the hippocampus. To corroborate the functional interpretation put forth by the authors, it would thus be important to relate the current results to what is already known about hippocampal mechanisms of working memory maintenance.

Reviewer #2 (Recommendations for the authors):

1. It seems that a traditional way to quantify the performance in the retro-cue WM task is to analyze the reaction times. I wonder if the authors analyzed it along with the recall error and if it could be informative.

2. I would like to make sure if all the presented ROI results were Bonferroni-corrected for multiple comparisons, i.e. if the original α-value was divided by the number of analyses on the dependent variable, here the number of ROIs.

3. I was wondering if the authors could explain why they selected bilateral ROIs instead of more specific left and right hemisphere regions. Could there be some laterialization effects? Do I understand correctly that the ROI analyses were performed on an average signal from the left and right hemisphere?

4. Could the authors clarify if the representational similarity analysis and IEM were performed on the cued vs. uncued items, regardless of the recall error? The authors performed a nice follow-up control analysis of the IEM reconstructions across precise recall and imprecise recall trails, how about the representational similarity analysis?

5. Related to the control analysis of the IEM reconstructions across precise recall and imprecise recall trails – could the authors explain why did they combine data from aLEC and DG/CA3 for this analysis, if in all the other analyses they analyse these ROIs separately?

6. In the caption of Figure 2B the authors mention only correlating the feature similarity of every two cued items with the neural pattern similarity during the WM delay period, but in the analysis (described in C) the effect sizes were compared with the uncued items. Thus, the authors could mention analyzing the uncued items also in B, for the clarity.

Reviewer #3 (Recommendations for the authors):

Recommendations for improving the writing and presentation

– Check for consistent use of past tense (especially prevalent in Methods)

– Add a more detailed explanation of what model is fitted on which data during the knee-point thresholding, given that part of the results hinge on separation of high- and low-error trials and therefore the exclusion of faulty trials.

– Consider including this review in the introduction, Christophel, T. B., Klink, P. C., Spitzer, B., Roelfsema, P. R., and Haynes, J. D. (2017). The distributed nature of working memory. Trends in cognitive sciences, 21(2), 111-124.

– P.6 mid-page, note that significance and effect size are not the same.

– Consider putting paragraph 2 last since 2 is very speculative. Please make sure your wording does not suggest that your data support the role of MTL in reducing interference in WM.

– In the discussion consider picking up on concepts from the introduction on why you think you found this effect with simple stimuli whereas other studies claim MTL involvement only for more complex stimuli or high-demand tasks.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "The Entorhinal-DG/CA3 Pathway in the Medial Temporal Lobe Retains Visual Working Memory of a Simple Surface Feature" for further consideration by eLife. Your revised article has been evaluated by Floris de Lange (Senior Editor) and a Reviewing Editor.

All reviewers were happy with your revisions and recommended that the manuscript be accepted for publication. Reviewer #1 however had some small final comments with respect to the statistical analyses (suggesting to only use non-parametric tests when assumptions for parametric tests are not met; or skip the parametric tests altogether and only report non-parametric tests).

You will find the individual reviewer reports below.

Reviewer #1 (Recommendations for the authors):

The authors responded to most of my comments. Some concerns arise as a result of the revisions. Neither in the manuscript nor in the response the authors specified how they ensured assumptions of statistical tests are met. I do appreciate that their results are consistent when using non-parametric alternatives but it would be important to report these results in the manuscript.

I suggest to either explicitly test assumptions and wherever those are not met using non-parametric alternatives or skip the parametric statistics and use non-parametric tests to start with. Otherwise, the authors risk reporting tests that are not valid, particularly given their relatively small sample size.

Reviewer #2 (Recommendations for the authors):

I have now had the chance to read the revised manuscript and author responses. All my concerns, as well as the concerns of other reviewers, were addressed. The authors now included several clarifications in the methods and results sections, clarifications that facilitate a general understanding of the experiment and help future attempts to replicate the findings. Moreover, the authors extended the discussion section by commenting on interesting alternative interpretations of the results. Overall, the authors did an excellent job revising the manuscript and in the present form it presents compelling evidence for significant findings.

Reviewer #3 (Recommendations for the authors):

Thank you to the authors for their resubmission and responses to my comments.

Generally, the revised manuscript thoroughly addresses the comments I have made. The weaknesses I outlined have been acknowledged in the discussion. All of my recommendations have been followed, either by clarification in the text or addition of a figure.

Overall, the responses are convincing and their resubmission is satisfactory.

Essential revisions:

1) The reviewers raised several issues of interpretation:

a) It is not clear whether the authors propose that the entorhinal-DG/CA3 circuitry retains specific visual feature (i.e., orientation and not other features) or enables discrimination among similar items (item1 vs. item2 independent of maintained feature) (see R1)

b) It is unclear whether the more distinct representation in MTL for the cued item may reflect hippocampal repulseion (see R2).

The authors should discuss these issues. We would welcome it if the authors could corroborate their interpretation with new data/analysis, but otherwise it would be important that the interpretational limitations are acknowledged in the manuscript.

We thank the editor and reviewers for these cautious notes. We have now clarified that our results and interpretation of these results are grounded in the prior research concerning the role of the entorhinal-DG/CA3 circuitry in supporting the quality of long-term episodic memory. Our data highlight that this MTL circuitry can distinguish similar items from the same feature space across trials. As a result of this computation, item-specific information can also be retained in this pathway. Although we have used orientation as a simple stimulus feature to minimize long-term memory influences, our results do not preclude the role of this MTL circuitry in other stimulus features. For instance, we have recently generalized our findings to colors using intracranial EEG and confirmed that trial-by-trial mnemonic discrimination during a short delay is indeed associated with the fidelity of item-specific WM representation (Xie, Chapeton, et al., in press). Please see our response to Reviewer #1 for details on these issues.

Furthermore, we also would like to thank Reviewer #2 for raising an interesting and relevant alternative interpretation concerning hippocampal repulsion. We have now discussed this hypothesis in our revised manuscript. Please see our response to Reviewer #2 for details.

2) The application of IEM methods should be motivated more clearly in the manuscript (see R2).

We thank Reviewer #2 for raising this clarification question. We have addressed this in the response letter and the revised manuscript by highlighting the consistency between this method and other multivariate approaches as well as the empirical efficiency (model-based) in using this method to reveal item-specific information in the neural data. Further details of this method have been extensively discussed and used in the literature (e.g., Ester et al., 2015; Liu et al., 2018; Sprague et al., 2018). Here, we would like to emphasize the benefits of using this method in the current research to be able to compare our findings with that from the previous research (e.g., Ester et al., 2015).

3) The behavioral data should be analyzed in greater detail (see R3).

We thank Reviewer #3 for this note. We have included more details about the behavioral data in a supplementary figure (updated Figure 1—figure supplement 1).

Please see below for a full description of issues raised by the reviewers.

Reviewer #1 (Recommendations for the authors):

The study by Xie et al., investigates whether the entorhinal-DG/CA3 pathway is involved in working memory maintenance. The study consists of 16 participants who perform an orientation visual working memory task with a retro-cue. Briefly, participants are presented with two gratings each with a different orientation followed by a cue that indicates which orientation should be further maintained. The authors focused their analyses on the delay interval following the cue. The main findings include a correlation between stimulus and neural similarities that was specific for cued stimulus and entorhinal-DG/CA3 locations. The authors observed similar results (cuing and region specificity) using inverted encoding modeling approach. Finally, they also showed that trials in which participants made a smaller error showed a better reconstruction fidelity on the cued side (compared to un-cued). This effect was absent for larger-error trials.

The study challenges a widely held traditional view that working memory and episodic memory have largely independent neural implementations. Specifically, this traditional view posits that the MTL is critical for episodic memory but not for working memory. The study adds to a large body of evidence showing involvement of the hippocampus across a range of different working memory tasks and stimuli. Nevertheless, it still remains unclear what functions may the hippocampus play in working memory.

We thank the reviewer’s positive appraisal of the current research, which adds to the growing research interest in the MTL’s contribution to WM.

1) The authors interpret their findings as suggesting that the entorhinal-DG/CA3 circuitry retains item-specific information to allow fine discrimination of similar WM items across trials. It was not clear to me whether they propose that the entorhinal-DG/CA3 circuitry retains specific visual feature (i.e., orientation and not other features) or enables discrimination among similar items (item1 vs. item2 independent of maintained feature). One other alternative interpretation of their results is that the entorhinal-DG/CA3 circuitry participates in orienting attention to memory representation or otherwise supports selection between initially remembered orientations. The current approach cannot tell these alternatives apart which is the main limitation of the study.

We thank the reviewer for this insightful comment. We have tested whether the MTL keeps track of the feature distance of temporarily remembered items from a simple continuous feature space based on a trial-wise RSA. Our interpretation is that these results indicate the capability of the MTL to distinguish similar items across trials (and hence fine discrimination across items). Conceptually, this could involve the differentiation between items 1 and 2 within a trial, although the current recording method (e.g., slow fMRI signals) does not afford the temporal resolution to reveal this within-trial effect. Grounded in the theoretical relationship between representational similarity and neural coding (Kriegeskorte and Wei, 2021), these trial-wise RSA findings further indicate that there is item-specific information retained in MTL, which is supported by our IEM results. In essence, our current data suggest that the MTL contains information about a retrospectively selected memory item and can tell this item apart from other items in the feature space. Based on these results, the reviewer is correct that we could not distinguish the information retention versus selection processes in the MTL – an inferential limitation that also exists in previous research using similar methods (e.g., Bettencourt and Xu, 2015; Ester et al., 2015; Harrison and Tong, 2009). To resolve these issues, we have recently generalized our findings to colors and confirmed that mnemonic discrimination during a short delay period is indeed associated with the fidelity of item-specific WM representation (Xie et al., in press).

We have discussed these issues in our Discussion.

“… These two approaches are therefore complementary to each other. Nevertheless, these analyses are correlational in nature. Hence, although fine-grained neural representations revealed by these analyses are associated with participants’ behavioral outcomes (Figure 4), it remains to be determined whether the entorhinal-DG/CA3 pathway contributes to the fidelity of the selected WM representation or also to the selection of task-relevant information. Strategies for resolving this issue can involve generalizing the current findings to other WM tasks without an explicit requirement of information selection (e.g., intracranial stimulation of the MTL in a regular WM task without a retro-cue manipulation, Xie et al., in press) and/or further exploring how the frontal-parietal mechanisms related to visual selection and attention interact with the MTL system (Panichello and Buschman, 2021).”

2) If the authors propose that the hippocampus maintains item-specific information about orientation (p. 9), it would be important to show specificity and also discuss how hippocampal contribution to working memory differs from the contribution of the visual system (e.g., Harrison and Tong, 2009)? On the other hand if authors propose that hippocampus enables discriminating among similar items independent of maintained feature, it would be important to show whether this generalize to other features.

We thank the reviewer for this insightful comment. Motivated by the large body of literature on the entorhinal-DG/CA3 circuitry’s function in distinguishing similar information across different stimulus dimensions of episodic long-term memory (e.g., object details, Reagh and Yassa, 2014; temporal information, Montchal et al., 2019), our study is grounded in the hypothesis that the same circuitry may also distinguish similar representations of retained WM items. As a result of this computation, item-specific WM information can be retained along this pathway. These two aspects of findings (i.e., information discrimination and information individualization) are the two sides of the same coin (see some detailed theoretical discussions in Kriegeskorte and Wei, 2021).

We use orientation as a simple task content to minimize rich long-term memory associations often seen in complex task stimuli (e.g., objects and scenes). Our findings therefore should be generalizable to other simple stimulus features, such as colors, as shown in one of our recent studies (Xie et al., in press). Therefore, our findings should not be interpreted that the entorhinal-DG/CA3 circuitry retains only orientation information.

We have clarified this in our revised Discussion.

“Furthermore, while we have used orientation as a simple stimulus feature to minimize long-term memory influences, our results do not preclude the role of this MTL circuitry in remembering other stimulus features, such as colors (e.g., Xie et al., in press). To more precisely reveal the MTL mechanisms that are shared across WM and long-term memory, future research should examine the extent to which MTL voxels evoked by a long-term memory task (e.g., mnemonic similarity task, Bakker et al., 2008) can be directly used to directly decode mnemonic content in visual WM tasks using different simple stimulus features.”

3) Furthermore, different features have been shown to elicit sustained or rhythmic neural activity in the hippocampus. To corroborate the functional interpretation put forth by the authors, it would thus be important to relate the current results to what is already known about hippocampal mechanisms of working memory maintenance.

We thank the reviewer for this critical comment. However, as our findings are not limited to the hippocampus (e.g., also in aLEC) and as our current recording methods preclude inferences based precise timing information, our results cannot directly speak to sustained or rhythmic neural activity observed in the hippocampus during WM retention. Furthermore, as WM maintenance is dynamic, it may depend on multiple forms of neural activity (e.g., sustained, burst, or complex coupling). Hence, more will need to be done before we can link these complex neural dynamics of WM with MTL activity. We have discussed this issue in revised manuscript as a direction for future investigation.

“However, considering the limitation in temporal resolution of the current recording method, it remains unknown how the MTL contributes to the dynamic coding schemes underlying WM maintenance (Stokes, 2015). Future research with direct recordings from multiple brain areas would be more suitable to investigate the fine-scale temporal dynamic underlying these similar neural patterns across brain regions during WM.”

Reviewer #2 (Recommendations for the authors):

1. It seems that a traditional way to quantify the performance in the retro-cue WM task is to analyze the reaction times. I wonder if the authors analyzed it along with the recall error and if it could be informative.

We thank the reviewers for this suggestion. Reaction time in the current context is less informative because accuracy was emphasized during data collection and participants were given a long, fixed time window (3.5 seconds) to render a response based on the method of adjustment. As a result, participants can make multiple button presses when reproducing the remembered orientation gratings, rendering the response time data more complex and harder to interpret (e.g., a big chunk of the RT is the motor response time) than that in the prior research.

2. I would like to make sure if all the presented ROI results were Bonferroni-corrected for multiple comparisons, i.e. if the original α-value was divided by the number of analyses on the dependent variable, here the number of ROIs.

The reviewer is correct that the p values are Bonferroni-corrected through a multiplication of the original p values with the number of tests conducted (see Tables S1 and S2). Please note that minor discrepancy can occur due to rounding. This is equivalent to dividing the α-value by the number of tests conducted (Rosenthal and Rubin, 1983). Here, the number of tests is the number of ROIs.

3. I was wondering if the authors could explain why they selected bilateral ROIs instead of more specific left and right hemisphere regions. Could there be some laterialization effects? Do I understand correctly that the ROI analyses were performed on an average signal from the left and right hemisphere?

Some prior research has suggested that effects related to MTL pattern separation functions can emerge bilaterally (e.g., Montchal et al., 2019; Reagh et al., 2017). As there is not strong prior expectation about lateralized effects, we combine all voxels for bilateral ROIs in our multivariate analyses. We did not perform the analyses separated for each hemisphere and then average the results together, as that would increase the number of tests involved.

4. Could the authors clarify if the representational similarity analysis and IEM were performed on the cued vs. uncued items, regardless of the recall error? The authors performed a nice follow-up control analysis of the IEM reconstructions across precise recall and imprecise recall trails, how about the representational similarity analysis?

Yes, we retained all trials for RSA and IEM analyses, as we clarified in our Method section,

“Under an effective set size of one item, participants’ recall performance was high (Figure 1B), with most recall errors centered around ± 45° of the cued orientation (~97% of the trials) within the ± 90° range. Hence, we retained all trials when investigating the amount of WM information in the recorded neural data during the delay period for multivariate analyses.”

We performed a split-trial analysis based on IEM. This analysis is built upon a prediction-based framework, which allows us to reconstruct the information in each trial of a hold-out test set based on an independent training set from all other trials. That means, the IEM can produce a more balanced trial-level estimate (i.e., reconstruction of remembered information channel) through multiple iterations to allow subsequent averaging or grouping across trials. However, this is not feasible or less efficient based on the trial-wise representational analysis, considering that it is a correlational measure. The magnitude and variability of a correlational measure will be confounded by the number of trials used in the analysis (Schönbrodt and Perugini, 2013). As a result, it is conceptually and analytically challenging to address this confound when the number of smaller- and larger-error trials are too different in RSA.

5. Related to the control analysis of the IEM reconstructions across precise recall and imprecise recall trails – could the authors explain why did they combine data from aLEC and DG/CA3 for this analysis, if in all the other analyses they analyse these ROIs separately?

As we have shown that both the aLEC and DG/CA3 retain a statistically significant amount of information about the cued item (Figures 2 and 3), we have combined them together to reduce the number of tests involved in subsequent analyses. Below, we have shown the results of the split-trial analysis separately for smaller- and larger-error trials. As demonstrated in Author response image 1, the reconstructed information channel relative to the cued item shows less dispersion for the precise recall trials, as compared with the imprecise recall trials. Here, we did not perform separate statistical tests with post-hoc correction for multiple comparisons, considering that the aLEC-DG/CA3 pathway as a whole has been theoretically implicated in predicting behaviors associated memory fidelity (Aimone et al., 2011; Bakker et al., 2008; Ekstrom and Yonelinas, 2020; Korkki et al., 2021; Leal and Yassa, 2018; Marr, 1971; Reagh and Yassa, 2014; Yassa and Stark, 2011) and empirically shown to retained item-specific WM content (Figures 2 and 3).

Author response image 1

Download asset Open asset

6. In the caption of Figure 2B the authors mention only correlating the feature similarity of every two cued items with the neural pattern similarity during the WM delay period, but in the analysis (described in C) the effect sizes were compared with the uncued items. Thus, the authors could mention analyzing the uncued items also in B, for the clarity.

We thank the reviewer for this great suggestion. We have now added this information.

“Specifically, we correlated the similarity in evoked neural patterns during the WM delay period separately with the feature similarity of every two cued items and with that of every two uncued items.”

Reviewer #3 (Recommendations for the authors):

Recommendations for improving the writing and presentation

– Check for consistent use of past tense (especially prevalent in Methods)

Thanks for the suggestion. We have now proofread the revised manuscript to minimize grammatical errors.

– Add a more detailed explanation of what model is fitted on which data during the knee-point thresholding, given that part of the results hinge on separation of high- and low-error trials and therefore the exclusion of faulty trials.

In an earlier version of the manuscript, we attempted the knee-point thresholding heuristic to find high- vs. low-error trials. Although this is a common practice in machine-learning literature, feedback from colleagues has suggested us to use a more intuitive heuristic, for example, based on 3 SD of an empirical distribution. Therefore, in our previous submission, we have already replaced the knee-point heuristic to the 3 SD heuristic, which is detailed in the original submission,

“Next, to separate larger-recall errors based on less precise WM representation from those attributable to attention lapses (deBettencourt et al., 2019), we adopted a widely-used thresholding heuristic to find potentially different categories of data points based on the empirical SD of a distribution. Specifically, in our current data, we first calculated the empirical SD (17.33°) of the aggregated raw recall error distribution from all subjects across 2880 trials (ranging from -90° to +90°), which captures the overall variability in participants’ recall performance without a priori model assumption. We then retained the larger-recall error trials within 20° to 3 SD of the recall error distribution (27 ± 3 trials; Figure 4A).”

– Consider including this review in the introduction, Christophel, T. B., Klink, P. C., Spitzer, B., Roelfsema, P. R., and Haynes, J. D. (2017). The distributed nature of working memory. Trends in cognitive sciences, 21(2), 111-124.

Thanks for the suggestion. We have included this important citation in the introduction.

In the introduction,

“This core mental faculty relies upon distributed brain regions (Christophel et al., 2017; Eriksson et al., 2015), ranging from …”

– P.6 mid-page, note that significance and effect size are not the same.

We thank the reviewer for this cautious note. Given the same sample size, significance level is inversely related to effect size (Rosenthal and Rosnow, 2008). To avoid confusion, we have removed the original interpretation about effect size.

“While the rest of the MTL showed similar patterns, we did not obtain significant evidence in other MTL ROIs following the correction of multiple comparisons,…”

– Consider putting paragraph 2 last since 2 is very speculative. Please make sure your wording does not suggest that your data support the role of MTL in reducing interference in WM.

Thanks for this suggestion. We have revised a few places to avoid this confusion.

– In the discussion consider picking up on concepts from the introduction on why you think you found this effect with simple stimuli whereas other studies claim MTL involvement only for more complex stimuli or high-demand tasks.

Thanks for this suggestion. We have revised our discussion on the empirical contribution of this study. Intuitively, the lack of findings using simple stimuli in the past research could be attributed to a lack of sensitivity of the MTL to simple features or lower task loads. With improved spatial resolution in MTL recording, our data may be more sensitive to pick up the representation of simple stimulus features. We have also discussed and tested why some past research using the same paradigm as us has not discovered MTL’s relevance for WM representation.

In Discussion,

“Previously, MTL activity has been shown to scale with WM set size of letters and color squares without decodable item-specific WM … These observations raise the conceptual question concerning the extent to which the MTL responds to task difficulty or retains item-level information in WM. In other words, is the MTL not sensitive to simple stimuli or lower task demands at all? Here, with improved spatial resolution in MTL recordings and using a simple stimulus feature, our data suggest that … These results, therefore, highlight the importance of fine-grained MTL signals in revealing item-specific WM content.”

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

The authors responded to most of my comments. Some concerns arise as a result of the revisions. Neither in the manuscript nor in the response the authors specified how they ensured assumptions of statistical tests are met. I do appreciate that their results are consistent when using non-parametric alternatives but it would be important to report these results in the manuscript.

I suggest to either explicitly test assumptions and wherever those are not met using non-parametric alternatives or skip the parametric statistics and use non-parametric tests to start with. Otherwise, the authors risk reporting tests that are not valid, particularly given their relatively small sample size.

We thank the reviewer for this cautious note. We agree with the reviewer about the importance of ensuring statistical conclusion validity. However, it should also be noted that the common two-stage approach to first check statistical test assumptions and then evaluate the test outcome is not without controversy (García-Pérez, 2012; Wells and Hintze, 2007). Although we had initially reported only non-parametric results, some colleagues have suggested against this approach after our preprint was published online. To balance these different concerns and since our results hold regardless of which statistical approach is adopted, we therefore now report both the parametric and non-parametric test results in the main text. While we understand that there is no “the best” approach to address most statistical issues, it is the convergence of statistical conclusions from different approaches that adds to the solidity of a research finding.

In the Data Analysis section, we added,

“We evaluated statistical significance based on conventional within-subject statistical procedures, such as paired-sample t-tests, with two-tailed p values. Similar results were obtained and verified based on non-parametric statistics (e.g., bootstrapped p values) that have few analytical assumptions (Good, 2013). In particular, we resampled participants’ data with replacement over 1,000 iterations and calculated the empirical two-tailed p values (note: these p values can slightly vary across different iterations of resampling and those smaller than 0.001 are marked as p_bootstrap < 0.001).”

In the main text, we reported the p_boostrap value associated with each statistical test. Please see the main text for further details.

References

García-Pérez, M. A. (2012). Statistical conclusion validity: Some common threats and simple remedies. Frontiers in Psychology, 3, 1–11. https://doi.org/10.3389/fpsyg.2012.00325

Good, P. (2013). Permutation tests: A practical guide to resampling methods for testing hypotheses. New York, NY: Springer Science & Business Media.

Wells, C. S., & Hintze, J. M. (2007). Dealing with assumptions underlying statistical tests. Psychology in the Schools, 44(5), 495–502. https://doi.org/10.1002/pits.20241

Author details

1. Weizhen Xie

   1. Surgical Neurology Branch, National Institute of Neurological Disorders and Stroke, Bethesda, United States
   2. Department of Psychology, University of California, Riverside, Riverside, United States
   3. Department of Psychology, University of Maryland, College Park, United States

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   weizhen.xie@nih.gov

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4655-6496

2. Marcus Cappiello

   Department of Psychology, University of California, Riverside, Riverside, United States

   Contribution

   Data curation, Investigation, Writing – original draft, Project administration

   Competing interests

   No competing interests declared

3. Michael A Yassa

   Center for the Neurobiology of Learning and Memory, School of Biological Sciences, University of California, Irvine, Irvine, United States

   Contribution

   Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8635-1498

4. Edward Ester

   Department of Psychology, University of Nevada, Reno, United States

   Contribution

   Software, Formal analysis, Validation, Investigation, Visualization, Methodology

   Competing interests

   No competing interests declared

5. Kareem A Zaghloul

   Surgical Neurology Branch, National Institute of Neurological Disorders and Stroke, Bethesda, United States

   Contribution

   Resources, Formal analysis, Investigation, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8575-3578

6. Weiwei Zhang

   Department of Psychology, University of California, Riverside, Riverside, United States

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   weiwei.zhang@ucr.edu

   Competing interests

   No competing interests declared

• 391

  Page views

• 89

  Downloads

• 0

  Citations

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