Distinct spatiotemporal activity patterns reflecting similarity-based and category-based semantic processing in anterior temporal lobe and inferior frontal cortex revealed via magnetoencephalography

Distinct spatiotemporal activity patterns reflecting similarity-based and category-based semantic processing in anterior temporal lobe and inferior frontal cortex revealed via magnetoencephalography | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Distinct spatiotemporal activity patterns reflecting similarity-based and category-based semantic processing in anterior temporal lobe and inferior frontal cortex revealed via magnetoencephalography Norio Fujimaki, Atsushi Matsumoto, Takahiro Soshi, Aya S. Ihara This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6947110/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Semantic processing of similarity and category are closely related yet distinct. In a prior functional magnetic resonance imaging study, the neural correlates of similarity and category processing were examined using prime–target word pairs in which similarity and category were independently manipulated. Specifically, they included "marine mammal", which belongs to the mammal category,but is more similar tofish. In some trials, two animal names from the categories of fish, terrestrial mammals, marine mammals, and birds were presented sequentially to investigate the priming effect of the first name on the processing of the second. In other trials, the second word was an inanimate object. Participants classified the second word as living or nonliving in all trials. Neural activity in the left anterior temporal lobe reflected similarity processing, whereas activityin the left inferior frontal cortex (IFC) corresponded to task-relevant category processing, consistent with models positing distinct systems for semantic representation and semanticcontrol. The present study employed magnetoencephalography to measure neural activity during thesame task, achieving enhanced temporal resolution. Neural activity reflecting similarity processing was observed in the left IFC approximately 525 msafter target word onset. This finding suggests that category processing in the left IFC occurred after the living/nonliving decision, which had a reaction time near 600 ms. The present results thus reveal a dynamic temporal sequence in the retrieval of task-relevant semantic features. Semantic distance Similarity Category Animal name Anterior temporal lobe Inferior frontal cortex Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Humans rapidly and flexibly comprehend language, supported by a neural system comprising both context-independent semantic representations and semantic control mechanisms that modulate information from these representations according to context. Two influential models explain semantic processing: the hub-and-spoke model (Patterson et al. 2007 ) and the controlled semantic cognition (CSC) framework (Lambon Ralph et al. 2017 ). Drawing on neuropsychological and neuroimaging evidence, these models propose that the anterior temporal lobe (ATL) functions as a hub, integrating information from sensorimotor and other regions to provide semantic representations independent of task and context. In contrast, the prefrontal cortex (PFC) implements semantic control by retrieving and selecting task- or context-relevant semantic features. These and other regions form an interconnected distributed network responsible for semantic processing (Price 2012 ; Noonan et al. 2013 ; Rice et al. 2015 ; Lambon Ralph et al. 2017 ; Hoffman and Lambon Ralph 2018 ; Gonzalez Alam et al. 2019 ; Jefferies et al. 2020 ). Consistent with this division, our previous functional magnetic resonance imaging (fMRI) study (Matsumoto et al. 2021 ), using semantic priming with animal names, found that similarity-based processing engaged the left ATL, whereas task-associated category-based processing engaged the left ventrolateral prefrontal cortex (VLPFC). Among the core components of semantic processing, similarity and category constitute fundamental elements (Taylor 1995 ). Semantic distance, a measure of similarity between words, can be quantified using questionnaires on various semantic features, such as “has four paws” and “breathes” (De Deyne et al. 2008 ). Semantic distance is smaller (indicating greater similarity) between animals sharing more common features. In a previous study (Soshi et al. 2017 ), semantic distances among 75 animal names were measured using a questionnaire covering 195 features to demonstrate that category classification is grounded in semantic features. The results demonstrated that categorization closely relates to similarity evaluation. In a previous fMRI study (Matsumoto et al. 2021 ), neural activity patterns associated with semantic similarity and category were examined. Despite the close relationship between similarity and category, the fMRI results revealed distinct spatial activity patterns for these two semantic processes using the marine mammal category. Marine mammals (e.g., dolphins) belong to the mammal category but share various semantic features with fish, such as “lives in the sea” and “swims.” Consequently, marine mammals exhibit greater similarity to fish than to terrestrial mammals. In the mentioned study, participants viewed word pairs consisting of a prime word followed by a target word to investigate the influence of the prime on semantic processing of the target. Prime words were animal names drawn from fish, terrestrial mammals, marine mammals, or bird categories, while target words were animal names from the same lists or names of artificial objects. Participants judged whether the target was living or nonliving, requiring a semantic decision but not explicit biological classification among animal species. The fMRI results showed that activation in the left ATL and left VLPFC significantly depended on semantic distance for all category pairs except those involving marine mammal targets, with activation decreasing as semantic distance decreased. This neural priming effect was interpreted as repetition suppression, reflecting the processing of overlapping features between the prime and target words. That is, greater feature overlap between prime and target was associated with reduced activation, indicating increased priming. Thus, the observed dependence on semantic distance reflected similarity-based processing. A notable finding was that activation patterns for marine mammal targets differed between the two regions. In the left ATL, the semantic priming effect for marine mammal targets was greater when preceded by fish than by terrestrial mammals, suggesting that the ATL supports similarity-based processing. In contrast, in the left VLPFC, the priming effect was greater when marine mammal targets followed terrestrial mammals rather than fish, indicating a greater role for the VLPFC in categorical processing. These prior fMRI findings align with the hub-and-spoke and CSC models: ATL activity reflected the processing of similarity, that is, a broad range of relevant features independent of task demands, whereas VLPFC activity reflected category-relevant feature processing during the living/nonliving decision task. This pattern further suggests that the brain automatically engages in biological classification of animal targets, even in the absence of explicit categorization demands. However, as fMRI exhibits limited temporal resolution, when these distinct processes unfold during semantic processing remains unclear. The present study therefore aimed to clarify this issue using magnetoencephalography (MEG), which offers high temporal resolution, while employing the same stimuli, task, and experimental paradigm as in the previous fMRI study (Matsumoto et al. 2021 ). Previous electroencephalography (EEG) and MEG studies (Friederici 2011 ; Kutas and Federmeier 2011 ) have examined the temporal profile of semantic processing and identified the N400 component, which appears approximately 200–600 ms after stimulus onset, with a peak at around 400 ms. The N400 was originally observed in an EEG study in which participants read sentences ending in semantically incongruent words (Kutas and Hillyard 1980 ). Its amplitude is modulated by cloze probability, increasing as the predictability of a word decreases. Modulation of the N400 has also been observed in lexical priming paradigms, where the N400 amplitude for the target is reduced when it is semantically or categorically related to the prime. This priming effect is consistent with the repetition suppression observed in the previous fMRI study and reflects the influence of semantic similarity and category membership. Accordingly, modulation of the N400 component is expected to reflect the effects of semantic distance and category during target word processing. Additionally, MEG studies have reported early activation in the ATL and VLPFC during visual word processing within the 200–400 ms window (Halgren et al. 2002 ; Marinkovic et al. 2003 ; Fujimaki et al. 2009 ; Pylkkänen 2020 ). Supporting evidence also comes from a recent electrocorticography study that recorded neural activity directly from the cortical surface, which reported ventral ATL activation beginning approximately 250 ms after picture onset in a picture-naming task (Chen et al. 2016 ). Together, these findings suggest that semantic processing in the present study is expected to occur within the 200–600 ms latency range following stimulus onset. In this study, we investigated the relationship between neural processing of similarity and category. We hypothesized that activation in the left ATL reflects similarity-based, whereas that in the inferior frontal cortex (IFC) indicates category-based processing. Here, we use the term IFC to refer to regions including the VLPFC, due to the variability in the reported localization in the present study. Specifically, we predicted that the priming effect, corresponding to the repetition suppression of the N400 component, for marine mammal target words would be greater when preceded by fish than by terrestrial mammals in the left ATL, whereas the opposite pattern could be observed in the left IFC. Our findings support a functional dissociation between these regions: the left ATL and IFC were more engaged in task-independent similarity-based processing and task-relevant processing, respectively. These results suggest that the brain flexibly recruits distinct neural systems depending on whether a task emphasizes similarity or categorical relationships, shedding light on the neural architecture underlying semantic processing. Materials and methods Participants Nine healthy, right-handed native Japanese speakers (six men and three women; mean age [standard deviation, SD] = 26.3 [12.3] years, age range: 21–59) participated in the experiment. All had normal or corrected-to-normal vision and no history of neurological or psychological disorders. The study was approved by the ethics committee of the National Institute of Information and Communications Technology and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent for participation and publication of data was obtained from all participants prior to the study. Stimuli and task The MEG experiments used the same stimuli, task, and semantic priming paradigm as described in a previous fMRI study (Matsumoto et al. 2021 ), with a modification to the timing protocol. A total of 40 animal names—10 each from the biological categories of fish (F; e.g., carp), terrestrial mammals (TM; e.g., cat), marine mammals (MM; e.g., dolphin), and birds (B; e.g., chicken)—were used as prime words. These same words, along with 10 artificial object names (man-made objects, e.g., knife), were also used as target words. Semantic distances between animal names were derived from questionnaire data reported in the previous study (Online Resource, Table S1 ) and are expressed on a 0–16 scale, with lower values indicating greater semantic similarity. In each trial (epoch), participants were visually presented with a prime word selected from the animal lists, followed by a target word drawn either from the same animal lists (but not identical to the prime) or from the list of artificial objects (Fig. 1 ). Participants were instructed to judge whether the target word referred to a living or nonliving entity and to respond by pressing one of two buttons after the response cue appeared, based on their decision. Stimuli were presented and behavioral responses were recorded using Presentation software (Neurobehavioral Systems, Inc., Berkeley, CA, USA). The stimulus presentation protocol was identical to that of the previous fMRI study (Matsumoto et al. 2021 ), with the exception of the cue onset latency. In the MEG experiment, the response cue appeared 1,500 ms after the onset of the target word, compared to 1,000 ms in the fMRI experiment. This adjustment allowed participants to blink during the interval between the end of the MEG epoch (1,000 ms after target word onset) and the cue onset. Each participant was presented with all possible prime–target combinations, excluding identical prime–target pairs (1,960 combinations in total). These were randomized and divided into four blocks for use across four experimental runs. Block order was randomized for each participant. All participants completed four runs, except one (S1), whose second run was interrupted and subsequently divided into two runs, resulting in five total runs. Data acquisition MEG data were recorded from all participants using a 148-channel Magnes 2500WH system (4-D Neuroimaging, San Diego, CA, USA). Individual head shapes were measured before conducting four experimental runs, with rest periods between runs. Three markers (vitamin E) were affixed to the nose and both ears to enable coordinate alignment between MRI and MEG. Marker locations were measured at the start of each run to position the magnetic sensors relative to the head. MEG signals were recorded within a frequency bandwidth of 0–200 Hz, from 200 ms prior to 1,000 ms following the onset of the target word, at a sampling rate of 678.17 Hz. Structural MRIs were acquired using a 1.5-T MAGNETOM Vision system (Siemens AG, München, Germany) with a magnetization-prepared rapid acquisition gradient echo protocol, using the following parameters: TR, 9.7 ms; TE, 4 ms; FA, 12°; slice thickness, 1 mm; image matrix, 256 × 256; pixel size, 1 × 1 mm. These structural images were employed to localize potential neural sources within the cerebral cortex and to construct a lead field matrix describing the relationship between magnetic sensor outputs and source currents (dipole moments). Analysis Preprocessing of MEG data was performed to reduce noise and baseline drift. First, independent component analysis (ICA) was applied to the raw MEG data using EEGLAB v4.5b (Swartz Center for Computational Neuroscience, La Jolla, CA, USA). Components corresponding to heartbeats, trends, spikes, and those exhibiting contour maps characteristic of blinks and eye movements were removed based on visual inspection (Online Resource, Table S2). The remaining ICA components were subsequently back-transformed into MEG data. Second, the data were bandpass filtered between 0.1 and 40 Hz in both forward and backward directions and screened for artifacts using peak-to-peak thresholds (Fig. 2 (a)). Specifically, individual epochs were rejected if the peak-to-peak amplitude in any channel exceeded a subject-specific threshold (several pT). Approximately 10% of all MEG epochs were excluded. The resulting averaged MEG data were used to identify active locations (major dipole positions). Third, to further reduce trends and noise in estimating activation strengths (moments of major dipoles), the MEG data were bandpass filtered twice between 1 and 8 Hz in forward and backward directions (Fig. 2 (b)). Additionally, linear trends were removed from three types of averaged MEG data detailed later. Trends were estimated from latency windows of − 200 to 0 ms and 800 to 1,000 ms relative to target word onset for each channel. The bandpass filtering with a 1 Hz low-frequency cutoff combined with detrending effectively mitigated trends of up to a few hundred fT present in the data. As an additional preprocessing step, the MEG data from multiple runs were transformed to represent measurements at a single sensor location (Uutela et al. 2001 ) before applying a 0.1–40 Hz bandpass filter, since the data had been recorded using sensors positioned at different locations across runs. To convert MEG data from sensor location A to sensor location B, source activation was estimated from the MEG data at sensor location A using the pseudo-inverse of the lead field matrix, followed by derivation of MEG data at sensor location B through forward calculations. One run was selected as a reference, and the MEG data from all other runs were transformed to correspond to the sensor location of the reference run for each participant. The MEG data from the reference run were excluded from further analysis, as amplitude was slightly smaller for the transformed than for the non-transformed data, so that the inclusion of the latter would increase data variations. For each participant, an overall average MEG dataset (all-average MEG data) was obtained by averaging all artifact-free transformed MEG epochs associated with correct responses in the living/nonliving decision task. Additionally, subaverage and category-average MEG data were derived from artifact-free transformed epochs with correct responses as follows: To analyze dependence on semantic distance, consecutive series of 10 MEG epochs arranged in descending order of semantic distance were averaged to generate subaverage MEG data. The choice of 10 epochs per subaverage reflected a balance between statistical power for detecting semantic distance dependence and accuracy of detrending (see Appendix ). In total, 866 subaverages were obtained across all participants. For analysis of category dependence, MEG epochs obtained during TM–MM and F–MM pairs were averaged separately to yield category-specific averages for each participant. Each category pair comprised 40–80 MEG epochs per participant. Neural sources were estimated using the method described in a previous report (Fujimaki et al. 2014 ), except that the grouping procedure was omitted. Custom software implementing this method was developed in MATLAB R2008a (MathWorks Inc., Natick, MA, USA). The method first determined the major dipoles and their locations from the all-average MEG data, where each dipole represents an electrical model of a neural source. Moments of these dipoles were then fitted to three types of average MEG data to obtain corresponding moment estimates, which served as measures of neural activation. The original method included a grouping procedure in which moments of neighboring dipoles were summed. This was based on the fact that neighboring dipoles are subject to crosstalk; that is, their moments tend to become more dependent as spatial distance decreases, so the summed moment provides a better representation of the sources for the overall MEG signal than individual moments. In the present study, to focus on major dipoles located near fMRI foci, this procedure was omitted, and activation analysis employed moments of individual major dipoles. Using this source estimation method combined with regression analysis, the minimum detectable relative change in activation attributable to semantic distance dependence was estimated to be ± 19% for the parameters of the present study (see Appendix ). A previous fMRI study (Matsumoto et al. 2021 ) reported an observed relative change of approximately ± 40%, roughly twice the estimated minimum detectable level. Therefore, semantic distance dependence should be detectable, provided inter-participant variation remains within a residual margin of approximately ± 20%. Details of the source estimation procedures in the present study are as follows. First, dipole locations were obtained at ~ 3-mm intervals on the diluted cerebral cortex surface, along with a lead field matrix based on a realistic three-compartment boundary-element model of medium resolution, using CURRY V8 (Compumedics Ltd., Abbotsford, Australia) and participant-specific structural MRIs. At each location, two dipoles were placed with orthogonal moments aligned along the two strongest components derived from singular value decomposition of the lead field matrix, as described previously (Fujimaki et al. 2014 ). Dipole moment values were estimated by minimizing the blockwise l 1 -norm (Matsuura and Okabe 1995 ; Uutela et al. 1999 ; Haufe et al. 2008 ; Terazono et al. 2010 ) using the SeDuMi 1.2 solver (Lehigh University, Bethlehem, PA, USA), applied to all-average MEG data filtered with a 0.1–40 Hz passband. The major dipole selection method was similar to that in the previous report (Fujimaki et al. 2014 ), with an extension to accommodate time-varying data. For each dipole location, the peak value of the moment magnitude, weighted by the lead field matrix, was calculated over a latency range of 50–600 ms. Major dipoles were selected from local maxima of the weighted magnitudes, ensuring a minimum separation greater than 20 mm between dipoles, a threshold determined by a previous simulation study (Fujimaki et al. 2002 ). This procedure yielded 60–70 major dipoles per participant. The moments of the major dipoles were fitted to all-average, subaverage, and category-average MEG data at each sampling time by multiplying the average MEG data—filtered with a 1–8 Hz passband and detrended—by the pseudo-inverse of the lead field matrix. The major dipole moments of subaverage and category-average were projected onto those of all-average at each sampling time. The projected moments, rather than moment magnitudes, were used as a measure of neural activity to minimize noise components spatially orthogonal to signal components. The analysis focused on major dipoles located near active foci identified in fMRI studies, specifically the left ATL and VLPFC (Matsumoto et al. 2021 ), to assess spatiotemporal activation patterns associated with semantic processing of similarity and category. Due to inter-participant variability in the locations of selected major dipoles, the term IFC is used to denote active locations in and around the VLPFC. Talairach coordinates of these foci—(− 50, 3, − 27) mm for the left ATL and (− 53, 24, 6) mm for the left IFC—were obtained from the previous fMRI study (Matsumoto et al. 2021 ). According to Talairach Client Version 2.4.3 ( https://talairach.org/index.html/ , Research Imaging Institute, San Antonio, TX, USA), these coordinates correspond to Brodmann areas 21 and 45, respectively. For each participant and region (i.e., left ATL and IFC), major dipoles were selected based on one of the following two criteria: (1) the dipole located nearest to the corresponding fMRI focus or (2) slightly more distant dipoles demonstrating significant dependence on semantic distance in participant-level regression analysis (p < 0.05), provided they were within 35 mm of the fMRI focus and the nearest dipole lacked significant dependence. For analysis of semantic distance dependence, projected moment values of subaverage were averaged within 100 ms time windows, advancing in 25 ms intervals. After excluding outliers exceeding three SDs from the mean, a linear mixed-effects model was applied using the “lme4” package (Bates et al. 2015 ) in RStudio ( https://posit.co/download/rstudio-desktop/ , Posit PBC, Boston, MA, USA). Semantic distance was modeled as a fixed effect, while participant main effects and participant by semantic distance interactions were modeled as random effects. Variances were estimated via restricted maximum likelihood. For analysis of category dependence, projected moment values of category-average obtained during TM–MM and F–MM pair presentations were averaged separately within 100 ms time windows, advancing in 25 ms intervals, and compared using paired-samples t-tests in SPSS Version 19 (IBM Corporation, Armonk, NY, USA). In both analyses, significance levels were corrected for multiple comparisons across areas and latencies by the Benjamini–Hochberg method (Benjamini and Hochberg 1995 ), controlling the false discovery rate (FDR) at 0.05. Results The present MEG study employed the stimuli of animal names F, TM, MM, B, and of artificial objects, task (identification of a target word as living [F, TM, MM, or B] or nonliving [artificial object]), and priming paradigm (Fig. 1) to investigate differences in spatiotemporal neural activation patterns related to semantic processing of similarity and category. Behavior Prime words consisted of animal names from categories F, TM, MM, or B, whereas target words included these animal names (excluding the identical prime word) and names of artificial objects. Participants responded by pressing one of two buttons after a response cue, indicating whether the target word was living or nonliving. Response accuracy exceeded 97% in artifact-free epochs for all participants, demonstrating successful task execution. Active locations After preprocessing, participant-specific all-average MEG data were generated by averaging all epochs, excluding those with incorrect responses or artifacts (Fig. 2). Source estimation applied to these data identified 60–70 major dipoles per participant. The major dipoles near fMRI foci were selected for each participant (Fig. 3), according to criteria (1) and (2) described in the Materials and Methods section. For three participants in each region (out of nine), the dipoles selected were not the nearest but met significance criterion (2). For the remaining six participants, the nearest major dipoles were selected (criterion (1), see Online Resource, Table S3a and b). The mean Talairach coordinates of selected dipoles were (−49, −1, −23) mm for the left ATL and (−53, 18, 10) mm for the left IFC. The average Euclidean distances between selected dipoles and fMRI foci were 15.7 mm (SD 9.0) for the left ATL and 17.0 mm (SD 11.2) for the left IFC, yielding an overall average distance of 16.4 mm (SD 9.9). In contrast, an overall average distance between the nearest major dipoles and fMRI foci was 10.9 mm (SD 5.6) (Online Resource, Table S3a). Dependence on semantic distance For each participant, subaverage MEG data were generated by averaging consecutive 10-epoch segments ordered by semantic distance, excluding epochs with incorrect responses or artifacts. Moments of the major dipoles were fitted to these subaverages and projected onto moments fitted to the all-average MEG data to reduce orthogonal noise. The subaverages were further averaged within 100-ms time windows with 25-ms steps to enhance signal-to-noise ratio. A total of 866 subaverage values of projected moments were obtained across participants for each time window and used as a measure of neural activity after excluding outliers exceeding three SDs from the mean. To identify regions and latencies showing dependence on semantic distance, projected moments of subaverage were split into high and low semantic distance groups by dividing data into halves and averaged separately, yielding mean semantic distances of 12.6 and 8.9, respectively. Comparison revealed mainly two activity difference peaks between the two groups within the left ATL and IFC at latencies approximately 200 to 600 ms following target word onset (Fig. 4). Considering statistical power, focus was placed on the two largest difference peaks: center-latency of 550 ms and 325 ms for the left ATL, and 525 ms and 350 ms for the left IFC (Online Resource, Table S4a). A linear mixed-effects model analysis, conducted for each region and 100-ms time window with significance level corrected for multiple comparisons using the Benjamini–Hochberg procedure, demonstrated significant dependence on semantic distance (FDR = 0.05) at 350 ms (p = 0.0250) and 525 ms (p = 0.00952) for the left IFC; no significant effects were observed for the left ATL (Online Resource, Table S4b). Time windows exhibiting significant dependence are marked in Fig. 4. The significant dependences displayed positive regression line slopes, i.e., neural activation increased with greater semantic distance. Figure 5 illustrates a scatterplot of projected moments of subaverage versus semantic distance with a regression line for the left IFC at a center-latency of 525 ms. The regression slope, number of data points, and cross-correlation coefficient were 0.0104 nAm, 860, and 0.132, respectively (Online Resource, Table S4a). Dependence on biological category Dependence on biological category was tested using category-average MEG data obtained from each participant during the presentation of TM–MM and F–MM pairs, following the exclusion of epochs containing artifacts or incorrect responses. The moments of the major dipoles were fitted to the category-average MEG data and projected onto the moments fitted to the all-average MEG data to suppress orthogonal noise. The resulting signals were then averaged within 100-ms time windows at 25-ms intervals to further reduce noise. This process yielded one projected moment value for each category, area, participant, and time window. Figure 6 presents the projected category-average moments averaged across participants, along with the differences between conditions (MM followed by F or TM). Both the left ATL and left IFC exhibited mainly two activity difference peaks between TM–MM and F–MM pairs at center latencies ranging from approximately 200 to 600 ms. To ensure sufficient statistical power, analysis focused on the largest and second-largest peaks: at center latencies of 225 ms and 550 ms for the left ATL, and 325 ms, 525 ms, and 550 ms (the latter two showing nearly identical differences) for the left IFC. A paired-samples t-test, followed by the Benjamini–Hochberg procedure correcting multiple comparisons across areas and latencies, demonstrated that activation was significantly lower for F–MM pairs compared to TM–MM pairs (FDR = 0.05) at a center latency of 225 ms in the left ATL (p = 0.0198) and at 525 ms in the left IFC (p = 0.0190) (Online Resource, Table S5a and S5b). Time windows with significant differences are indicated in Fig. 6. Discussion Active locations and latencies Using blockwise l 1 -norm minimization (Terazono et al. 2010 ), 60–70 major dipoles were detected across the whole brain in each participant. To evaluate the dependence of neural activation on similarity and category, major dipoles located near activation foci identified in a previous fMRI study (Matsumoto et al. 2021 ) were selected. The mean distance between the fMRI-identified foci and the nearest major dipoles detected in the present MEG study was 10.9 mm (Online Resource, Table S3a). This distance is considered reasonable, as a prior simulation study using the same source estimation method reported a mean localization error of 6.3 mm (SD: 3.2 mm) between assumed source locations and the nearest detected major dipoles (Fujimaki et al. 2014 ). That simulation assumed simultaneous activation of five brain areas during visual word processing and an MEG signal-to-noise ratio (SNR) of 10. These conditions are comparable to those in the present study, which involved visual word stimuli and yielded a mean SNR of 9.02 across all participants in the all-average MEG data used to identify major dipole locations (Online Resource, Table S3b). Additional variability in the observed nearest dipole locations may reflect factors such as registration errors of several millimeters between MEG and fMRI coordinate systems. The mean distance between the fMRI activation foci and the selected major dipoles in the present MEG study was 16.4 mm (Online Resource, Table S3a), which is greater than the 10.9 mm distance observed for the nearest dipoles. This larger value may further reflect inter-individual variability in the locations of active foci. The present MEG study revealed significant dependence of activation on semantic distance in the left IFC at center latencies of 350 ms and 525 ms following target word onset (Fig. 4 ), and significant dependence of activation on category in the left ATL and IFC at center latencies of 225 ms and 525 ms, respectively (Fig. 6 ). These latencies, corresponding to semantic processing, are consistent with prior MEG studies that reported early activity in the left ATL and IFC within a latency range of 200–400 ms during visual word processing (Halgren et al. 2002 ; Marinkovic et al. 2003 ; Fujimaki et al. 2009 ; Pylkkänen 2020 ), as well as numerous EEG studies indicating that the N400 component, associated with semantic processing, typically occurs between 200 ms and 600 ms (Kutas and Federmeier 2011 ). Dependence on semantic distance and category Based on findings from a previous fMRI study (Matsumoto et al. 2021 ), we hypothesized that activation in the left ATL and IFC would reflect similarity- and category-based processing, respectively. However, the present MEG results showed that activation for the marine mammal target was reduced—or, equivalently, that the priming effect was larger—when preceded by a fish prime than by a terrestrial mammal prime in the left ATL and IFC at center latencies of 225 ms and 525 ms, respectively. This pattern indicates that activation in both regions reflected similarity-based processing. Therefore, the hypothesis regarding the role of the left IFC was not supported, as activation at 525 ms reflected similarity- rather than category-based processing. This discrepancy may be attributable to differences in the temporal resolution of MEG and fMRI. In our previous fMRI study, activation was averaged over several seconds due to the hemodynamic delay, potentially conflating similarity- and category-related signals. Additionally, later latency components may have been attenuated in the current MEG study due to the 1–8 Hz bandpass filtering applied to the data. Taken together, the MEG and fMRI findings suggest that activation in the left IFC reflects similarity-based processing up to approximately 600 ms following target word onset, with category-based processing emerging at later latencies. This delayed onset of category-related activity in the left IFC aligns with the following behavioral findings. In a prior behavioral study using the same stimuli and task as the fMRI study—except that participants were instructed to make a living/nonliving judgment immediately after target word presentation—the reaction time (RT) was approximately 600 ms (Matsumoto et al. 2021 ). These findings suggest that categorical processing in the left IFC occurred as a post-response process. In that study, RTs for the marine mammal target varied systematically with the semantic relationship between prime and target: responses were significantly faster for the fish–marine mammal pair (mean RT: 581 ms) than for the terrestrial mammal–marine mammal pair (589 ms), and significantly faster for the terrestrial mammal–marine mammal pair than for the bird–marine mammal pair (596 ms). This RT pattern paralleled the semantic distances between primes and the target; responses were faster when semantic distance was smaller. The priming effect on RT indicates greater facilitation when the prime and target shared more semantic features, reflecting greater repetition of semantic processing. Thus, RTs appear to be driven by similarity-based processing. This interpretation is plausible, as a living/nonliving judgment does not require explicit identification of the target’s specific category (e.g., mammal vs. fish). Therefore, the proposed interpretation that category-based processing in the left IFC occurs post-response is consistent with the view that RT primarily reflects similarity-based rather than category-based processing. Furthermore, the observed shift from similarity- to category-based processing suggests that the living/nonliving task entails both recognition of the target animal and access to its biological category. In the case of a marine mammal target, the left IFC may initially retrieve a broad set of relevant features from semantic representation, followed by more selective retrieval of features associated with the mammal category. This interpretation is consistent with the CSC model, which posits that the left IFC supports controlled semantic retrieval according to task demands. Limitations The present study has a few limitations. First, the sample size was small. Second, the short epoch length of 1.2 s (from 0.2 s before to 1.0 s after target word onset) may have precluded the capture of later neural events. Despite these constraints, the study successfully demonstrated sensitivity to both semantic distance and category effects. This was achieved by filtering the data within a narrow passband (1–8 Hz), removing linear trends to minimize noise and drift, and focusing on the largest and second-largest peaks in activation differences. Specifically, effects were analyzed for differences between high and low semantic distance conditions using a linear mixed-effects model, and between fish and terrestrial mammal primes using paired-samples t-tests. Further studies are warranted to more precisely characterize the temporal dynamics of semantic processing. Future work should (1) collect more data to allow finer-grained temporal analysis, (2) record continuous MEG signals instead of pre-segmented epochs to enable broader-band filtering (e.g., 0.1–40 Hz), which would reduce trends while preserving slowly evolving signals, and (3) extend the epoch duration to capture potential late-latency events time-locked to target onset. Conclusions In this MEG study, we used the same stimuli, task, and priming paradigm as our prior fMRI study (Matsumoto et al. 2021 ) to investigate the spatiotemporal characteristics of neural activity underlying semantic processing related to similarity and biological category. We hypothesized that the left ATL would be involved in similarity processing and the left IFC in category processing. While the role of the left ATL was supported, the left IFC was instead associated with similarity-based processing at approximately 525 ms post-target onset. These findings suggest that category-based processing in the left IFC occurs later, beyond approximately 600 ms. Although the living/nonliving decision task did not explicitly require identification of biological category, the results indicate that such categorical information was likely retrieved in the left IFC as a post-response process. Together with our previous fMRI findings, the present results support both the hub-and-spoke model and the CSC model, wherein the left ATL provides semantic representations and the left IFC facilitates controlled retrieval of task-relevant semantic features. Declarations Acknowledgments The authors would like to thank Editage for English language editing. Author contributions N.F., A.M., and A.S.I. designed the study. T.S. produced the materials. A.S.I. conducted the experiments. N.F. analyzed the data. The results were discussed by all authors. N.F. and A.S.I. wrote the first draft of the manuscript, and all authors revised the manuscript and approved the final version. Competing interests The authors declare no competing interests. Funding This work was supported by JSPS KAKENHI Grant Numbers JP16K01969 and JP22K12758. References Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1–48. https://doi.org/10.18637/jss.v067.i01 Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x Chen Y, Shimotake A, Matsumoto R, et al. (2016) The “when” and “where” of semantic coding in the anterior temporal lobe: temporal representational similarity analysis of electrocorticogram data. Cortex 79:1–13. https://doi.org/10.1016/j.cortex.2016.02.015 De Deyne S, Verheyen S, Ameel E, Vanpaemel W, Dry MJ, Voorspoels W, Storms G (2008) Exemplar by feature applicability matrices and other Dutch normative data for semantic concepts. Behav Res Methods 40:1030–1048. https://doi.org/10.3758/BRM.40.4.1030 Friederici AD (2011) The brain basis of language processing: from structure to function. Physiol Rev 91:1357–1392. https://doi.org/10.1152/physrev.00006.2011 Fujimaki N, Hayakawa T, Ihara A, et al. (2009) Early neural activation for Lexico-semantic access in the left anterior temporal area analyzed by an fMRI-assisted MEG multidipole method. Neuroimage 44:1093–1102. https://doi.org/10.1016/j.neuroimage.2008.10.021 Fujimaki N, Hayakawa T, Nielsen M, Knösche TR, Miyauchi S (2002) An fMRI-constrained MEG source analysis with procedures for dividing and grouping activation. Neuroimage 17:324–343. https://doi.org/10.1006/nimg.2002.1160 Fujimaki N, Terazono Y, Ihara AS, Hayakawa T, Matani A, Umehara H (2014) Accuracy of estimating strengths of dipole moments from a small number of magnetoencephalographic trial data. Brain Topogr 27:635–647. https://doi.org/10.1007/s10548-014-0363-9 Gonzalez Alam T, Karapanagiotidis T, Smallwood J, Jefferies E (2019) Degrees of lateralisation in semantic cognition: evidence from intrinsic connectivity. Neuroimage 202:116089. https://doi.org/10.1016/j.neuroimage.2019.116089 Halgren E, Dhond RP, Christensen N, Van Petten C, Marinkovic K, Lewine JD, Dale AM (2002) N400-like magnetoencephalography responses modulated by semantic context, word frequency, and lexical class in sentences. Neuroimage 17:1101–1116. https://doi.org/10.1006/nimg.2002.1268 Haufe S, Nikulin VV, Ziehe A, Müller KR, Nolte G (2008) Combining sparsity and rotational invariance in EEG/MEG source reconstruction. Neuroimage 42:726–738. https://doi.org/10.1016/j.neuroimage.2008.04.246 Hoffman P, Lambon Ralph MA (2018) From percept to concept in the ventral temporal lobes: Graded hemispheric specialisation based on stimulus and task. Cortex 101:107–118. https://doi.org/10.1016/j.cortex.2018.01.015 Jefferies E, Thompson H, Cornelissen P, Smallwood J (2020) The neurocognitive basis of knowledge about object identity and events: dissociations reflect opposing effects of semantic coherence and control. Philos Trans R Soc Lond B Biol Sci 375:20190300. https://doi.org/10.1098/rstb.2019.0300 Kutas M, Federmeier KD (2011) Thirty years and counting: finding meaning in the N400 component of the event-related brain potential (ERP). Annu Rev Psychol 62:621–647. https://doi.org/10.1146/annurev.psych.093008.131123 Kutas M, Hillyard SA (1980) Reading senseless sentences: brain potentials reflect semantic incongruity. Science 207:203–205. https://doi.org/10.1126/science.7350657 Lambon Ralph MA, Jefferies E, Patterson K, Rogers TT (2017) The neural and computational bases of semantic cognition. Nat Rev Neurosci 18:42–55. https://doi.org/10.1038/nrn.2016.150 Marinkovic K, Dhond RP, Dale AM, Glessner M, Carr V, Halgren E (2003) Spatiotemporal dynamics of modality-specific and supramodal word processing. Neuron 38:487–497. https://doi.org/10.1016/s0896-6273(03)00197-1 Matsumoto A, Soshi T, Fujimaki N, Ihara AS (2021) Distinctive responses in anterior temporal lobe and ventrolateral prefrontal cortex during categorization of semantic information. Sci Rep 11:13343. https://doi.org/10.1038/s41598-021-92726-7 Matsuura K, Okabe Y (1995) Selective minimum-norm solution of the biomagnetic inverse problem. IEEE Trans Bio Med Eng 42:608–615. https://doi.org/10.1109/10.387200 Noonan KA, Jefferies E, Visser M, Lambon Ralph MA (2013) Going beyond inferior prefrontal involvement in semantic control: evidence for the additional contribution of dorsal angular gyrus and posterior middle temporal cortex. J Cogn Neurosci 25:1824–1850. https://doi.org/10.1162/jocn_a_00442 Patterson K, Nestor PJ, Rogers TT (2007) Where do you know what you know? The representation of semantic knowledge in the human brain. Nat Rev Neurosci 8:976–987. https://doi.org/10.1038/nrn2277 Price CJ (2012) A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading. Neuroimage 62:816–847. https://doi.org/10.1016/j.neuroimage.2012.04.062 Pylkkänen L (2020) Neural basis of basic composition: what we have learned from the red-boat studies and their extensions. Philos Trans R Soc Lond B Biol Sci 375:20190299. https://doi.org/10.1098/rstb.2019.0299 Rice GE, Lambon Ralph MA, Hoffman P (2015) The Roles of Left Versus Right Anterior Temporal Lobes in Conceptual Knowledge: An ALE Meta-analysis of 97 Functional Neuroimaging Studies. Cereb Cortex 25:4374-4391. https://doi.org/10.1093/cercor/bhv024 Soshi T, Fujimaki N, Matsumoto A, Ihara AS (2017) Memory-based specification of verbal features for classifying animals into super-ordinate and sub-ordinate categories. Front Commun 2:1–14. https://doi.org/10.3389/fcomm.2017.00012 Taylor JR (1995) Linguistic categorization. Oxford University Press Terazono Y, Fujimaki N, Murata T, Matani A (2010) Point source reconstruction principle of linear inverse problems. Inverse Probl 26:115016. https://doi.org/10.1088/0266-5611/26/11/115016 Uutela K, Hämäläinen M, Somersalo E (1999) Visualization of magnetoencephalographic data using minimum current estimates. Neuroimage 10:173–180. https://doi.org/10.1006/nimg.1999.0454 Uutela K, Taulu S, Hämäläinen M (2001) Detecting and correcting for head movements in neuromagnetic measurements. Neuroimage 14:1424–1431. https://doi.org/10.1006/nimg.2001.0915 Additional Declarations No competing interests reported. Supplementary Files SupplementaryTableS1S5.zip Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6947110","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":476993437,"identity":"e17aa525-f66a-4d15-a134-d6a0f82ba633","order_by":0,"name":"Norio Fujimaki","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+ElEQVRIiWNgGAWjYJCCAww2DMxsDAzMDAk/bBjYmA8QoyUNqiWxJ42BjS2ZGHvSwCQzAwPbYSAmoMWcvffhgQ8Jh9n5xA4/NnjAcz6Pj43/mARDjU00Li2WPccNDs5IOMzMJp1mnJBgcbuYjY2ZTYLhWFpuAw4tBjfSGA7z/gBpSTA+kMBzO7FNvplNgrHhMH4tPGBb0j8fSGA7l9gGsoVILTlAh7EdIELLmWMMQL+kg7QUGyT2JIO0GFsk4PPL8TbmDx8SrJPlZ6dvlvzxwy5xfhvjwxsfamxwaoEBlLhgkUggoBwE7JA5zB+I0DEKRsEoGAUjBwAAOhBUUqnO0WIAAAAASUVORK5CYII=","orcid":"","institution":"National Institute of Information and Communications Technology","correspondingAuthor":true,"prefix":"","firstName":"Norio","middleName":"","lastName":"Fujimaki","suffix":""},{"id":476993438,"identity":"1fbe858e-b60c-4c94-a7e9-8d26d8fc9b90","order_by":1,"name":"Atsushi Matsumoto","email":"","orcid":"","institution":"Kansai University of Welfare Sciences","correspondingAuthor":false,"prefix":"","firstName":"Atsushi","middleName":"","lastName":"Matsumoto","suffix":""},{"id":476993440,"identity":"69e42d15-d1f5-4cb3-8424-71bf29544e34","order_by":2,"name":"Takahiro Soshi","email":"","orcid":"","institution":"Mejiro University","correspondingAuthor":false,"prefix":"","firstName":"Takahiro","middleName":"","lastName":"Soshi","suffix":""},{"id":476993443,"identity":"150431f2-1dd8-4e31-a25a-f2765ec63b21","order_by":3,"name":"Aya S. Ihara","email":"","orcid":"","institution":"National Institute of Information and Communications Technology","correspondingAuthor":false,"prefix":"","firstName":"Aya","middleName":"S.","lastName":"Ihara","suffix":""}],"badges":[],"createdAt":"2025-06-22 00:23:10","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6947110/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6947110/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85648978,"identity":"6ce7089d-e49d-453e-9577-a258cb0691a1","added_by":"auto","created_at":"2025-06-30 08:54:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":20621,"visible":true,"origin":"","legend":"\u003cp\u003eTime course of stimulus presentation.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/389b19d20a946a787120d369.png"},{"id":85648991,"identity":"68974221-8464-4807-9a71-b2a90456f786","added_by":"auto","created_at":"2025-06-30 08:54:39","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":62730,"visible":true,"origin":"","legend":"\u003cp\u003eWaveforms of all-average MEG data from a representative participant, with data from all channels superimposed.\u003cbr\u003e\n (a) Data filtered with a passband of 0.1–40 Hz and with artifacts rejected; these data were used to detect major dipoles and estimate their locations.\u003cbr\u003e\n (b) The same data further filtered with a 1–8 Hz passband and detrended; these data were used to compute the moments of major dipoles. The vertical axis indicates amplitude (fT), and the horizontal axis indicates latency (ms) from the onset of the target words.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/a7098cd7f9a5cbeff82b3dec.png"},{"id":85649057,"identity":"9e7f629c-6784-48d7-b14d-a93ef525e6c3","added_by":"auto","created_at":"2025-06-30 08:54:52","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":133419,"visible":true,"origin":"","legend":"\u003cp\u003eLocations of major dipoles obtained from the MEG data of nine participants. Filled red circles indicate major dipoles near the left anterior temporal lobe (ATL), and filled blue squares indicate those near the left inferior frontal cortex (IFC). Open symbols represent activation foci reported in our previous fMRI study (open circle: ATL; open square: IFC). They are overlaid on a standard brain image.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/16b4185768f2dd2ffa3100fd.png"},{"id":85649026,"identity":"628e113c-cbca-4250-a154-ca7bd869ab2b","added_by":"auto","created_at":"2025-06-30 08:54:45","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":49486,"visible":true,"origin":"","legend":"\u003cp\u003eTime courses of activation in the left ATL and IFC for the high- and low-semantic distance groups and their difference. Activations represent the projected moments of selected major dipoles, which were fitted to subaverage MEG data. These were averaged in 100-ms windows at 25-ms intervals across all nine participants and then separately averaged for each semantic distance group. The vertical axis indicates activation strength, and the horizontal axis indicates the center latencies of the 100-ms averaging windows from the onset of target words. Error bars represent standard errors. Symbols below the curves indicate the center latencies and time windows showing significant dependence on semantic distance (FDR = 0.05).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/c90159ba0039f69b6a16d430.png"},{"id":85649008,"identity":"c067f2ae-152e-40d7-8e78-2da11cb50b90","added_by":"auto","created_at":"2025-06-30 08:54:41","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":22233,"visible":true,"origin":"","legend":"\u003cp\u003eScatterplot showing the relationship between neural activation (vertical axis) and semantic distance (horizontal axis). Activations represent the projected moments of selected major dipoles, fitted to subaverage MEG data and averaged within a time window centered at 525 ms (475–575 ms) in the left IFC. The number of data points was 860; the correlation coefficient was 0.132, and the p-value for dependence on semantic distance was 0.00952.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/8fb2b9e30df59de46028e127.png"},{"id":85649056,"identity":"cc5359e6-d14c-4150-904c-af80e705abad","added_by":"auto","created_at":"2025-06-30 08:54:52","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":37242,"visible":true,"origin":"","legend":"\u003cp\u003eTime courses of neural activation in response to the prime–target word pairs \u003cem\u003eterrestrial mammal–marine mammal\u003c/em\u003e and \u003cem\u003efish–marine mammal\u003c/em\u003e, and their difference. Activations represent the projected moments of selected major dipoles fitted to category-average MEG data. These were averaged in 100-ms windows at 25-ms intervals across nine participants. The vertical axis indicates activation strength, and the horizontal axis indicates the center latencies of the 100-ms averaging windows from the onset of target words. Error bars represent standard errors. Symbols below the curves indicate time windows showing a significant difference between the two category pairs (FDR = 0.05).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/598390d0ee2cbad392123fc5.png"},{"id":85648983,"identity":"a6378ab3-9e88-4a79-8b69-4a37d2eb1758","added_by":"auto","created_at":"2025-06-30 08:54:35","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":19011,"visible":true,"origin":"","legend":"\u003cp\u003eTheoretical minimum detectable change in relative activation (\u003cem\u003eC\u003c/em\u003e\u003csub\u003e\u003cem\u003eth\u003c/em\u003e\u003c/sub\u003e (vertical axis)) as a function of the number of epochs in subaveraging (\u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003esa\u003c/em\u003e\u003c/sub\u003e (horizontal axis)). Equation (A6) was evaluated assuming nine participants (\u003cem\u003eK\u003c/em\u003e), 1,000 epochs per participant (\u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e), a signal-to-noise ratio of 6 for participant-level activations averaged across all epochs (\u003cem\u003esnr(N\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e)\u003c/em\u003e), and a significance level of α = 0.05.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/7a483df9f42ebbdf214e9263.png"},{"id":105921319,"identity":"e6c7055c-24bd-49ad-a70a-e4bc8b6a83c8","added_by":"auto","created_at":"2026-04-01 12:42:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":838676,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/c51c8505-f38a-4232-882a-949d81b7119e.pdf"},{"id":85649034,"identity":"923cc6e9-4c90-4ab6-b91e-e4cd0b3916c5","added_by":"auto","created_at":"2025-06-30 08:54:47","extension":"zip","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":708058,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS1S5.zip","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/999486b08c455281b7fb4761.zip"},{"id":85649016,"identity":"eb37f0c9-73ee-41b0-b275-4fbd4dcbaa6f","added_by":"auto","created_at":"2025-06-30 08:54:42","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":18118,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-6947110/v1/d71eb91ec95fea647a2671f5.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Distinct spatiotemporal activity patterns reflecting similarity-based and category-based semantic processing in anterior temporal lobe and inferior frontal cortex revealed via magnetoencephalography","fulltext":[{"header":"Introduction","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eHumans rapidly and flexibly comprehend language, supported by a neural system comprising both context-independent semantic representations and semantic control mechanisms that modulate information from these representations according to context. Two influential models explain semantic processing: the hub-and-spoke model (Patterson et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) and the controlled semantic cognition (CSC) framework (Lambon Ralph et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Drawing on neuropsychological and neuroimaging evidence, these models propose that the anterior temporal lobe (ATL) functions as a hub, integrating information from sensorimotor and other regions to provide semantic representations independent of task and context. In contrast, the prefrontal cortex (PFC) implements semantic control by retrieving and selecting task- or context-relevant semantic features. These and other regions form an interconnected distributed network responsible for semantic processing (Price \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Noonan et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Rice et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Lambon Ralph et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Hoffman and Lambon Ralph \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Gonzalez Alam et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Jefferies et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Consistent with this division, our previous functional magnetic resonance imaging (fMRI) study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), using semantic priming with animal names, found that similarity-based processing engaged the left ATL, whereas task-associated category-based processing engaged the left ventrolateral prefrontal cortex (VLPFC).\u003c/p\u003e\u003cp\u003eAmong the core components of semantic processing, similarity and category constitute fundamental elements (Taylor \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1995\u003c/span\u003e). Semantic distance, a measure of similarity between words, can be quantified using questionnaires on various semantic features, such as \u0026ldquo;has four paws\u0026rdquo; and \u0026ldquo;breathes\u0026rdquo; (De Deyne et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Semantic distance is smaller (indicating greater similarity) between animals sharing more common features. In a previous study (Soshi et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), semantic distances among 75 animal names were measured using a questionnaire covering 195 features to demonstrate that category classification is grounded in semantic features. The results demonstrated that categorization closely relates to similarity evaluation.\u003c/p\u003e\u003cp\u003eIn a previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), neural activity patterns associated with semantic similarity and category were examined. Despite the close relationship between similarity and category, the fMRI results revealed distinct spatial activity patterns for these two semantic processes using the marine mammal category. Marine mammals (e.g., dolphins) belong to the mammal category but share various semantic features with fish, such as \u0026ldquo;lives in the sea\u0026rdquo; and \u0026ldquo;swims.\u0026rdquo; Consequently, marine mammals exhibit greater similarity to fish than to terrestrial mammals. In the mentioned study, participants viewed word pairs consisting of a prime word followed by a target word to investigate the influence of the prime on semantic processing of the target. Prime words were animal names drawn from fish, terrestrial mammals, marine mammals, or bird categories, while target words were animal names from the same lists or names of artificial objects. Participants judged whether the target was living or nonliving, requiring a semantic decision but not explicit biological classification among animal species.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe fMRI results showed that activation in the left ATL and left VLPFC significantly depended on semantic distance for all category pairs except those involving marine mammal targets, with activation decreasing as semantic distance decreased. This neural priming effect was interpreted as repetition suppression, reflecting the processing of overlapping features between the prime and target words. That is, greater feature overlap between prime and target was associated with reduced activation, indicating increased priming. Thus, the observed dependence on semantic distance reflected similarity-based processing. A notable finding was that activation patterns for marine mammal targets differed between the two regions. In the left ATL, the semantic priming effect for marine mammal targets was greater when preceded by fish than by terrestrial mammals, suggesting that the ATL supports similarity-based processing. In contrast, in the left VLPFC, the priming effect was greater when marine mammal targets followed terrestrial mammals rather than fish, indicating a greater role for the VLPFC in categorical processing.\u003c/p\u003e \u003cp\u003eThese prior fMRI findings align with the hub-and-spoke and CSC models: ATL activity reflected the processing of similarity, that is, a broad range of relevant features independent of task demands, whereas VLPFC activity reflected category-relevant feature processing during the living/nonliving decision task. This pattern further suggests that the brain automatically engages in biological classification of animal targets, even in the absence of explicit categorization demands.\u003c/p\u003e \u003cp\u003eHowever, as fMRI exhibits limited temporal resolution, when these distinct processes unfold during semantic processing remains unclear. The present study therefore aimed to clarify this issue using magnetoencephalography (MEG), which offers high temporal resolution, while employing the same stimuli, task, and experimental paradigm as in the previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Previous electroencephalography (EEG) and MEG studies (Friederici \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Kutas and Federmeier \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) have examined the temporal profile of semantic processing and identified the N400 component, which appears approximately 200\u0026ndash;600 ms after stimulus onset, with a peak at around 400 ms. The N400 was originally observed in an EEG study in which participants read sentences ending in semantically incongruent words (Kutas and Hillyard \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e1980\u003c/span\u003e). Its amplitude is modulated by cloze probability, increasing as the predictability of a word decreases. Modulation of the N400 has also been observed in lexical priming paradigms, where the N400 amplitude for the target is reduced when it is semantically or categorically related to the prime. This priming effect is consistent with the repetition suppression observed in the previous fMRI study and reflects the influence of semantic similarity and category membership. Accordingly, modulation of the N400 component is expected to reflect the effects of semantic distance and category during target word processing. Additionally, MEG studies have reported early activation in the ATL and VLPFC during visual word processing within the 200\u0026ndash;400 ms window (Halgren et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Marinkovic et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Fujimaki et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Pylkk\u0026auml;nen \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Supporting evidence also comes from a recent electrocorticography study that recorded neural activity directly from the cortical surface, which reported ventral ATL activation beginning approximately 250 ms after picture onset in a picture-naming task (Chen et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Together, these findings suggest that semantic processing in the present study is expected to occur within the 200\u0026ndash;600 ms latency range following stimulus onset.\u003c/p\u003e \u003cp\u003eIn this study, we investigated the relationship between neural processing of similarity and category. We hypothesized that activation in the left ATL reflects similarity-based, whereas that in the inferior frontal cortex (IFC) indicates category-based processing. Here, we use the term IFC to refer to regions including the VLPFC, due to the variability in the reported localization in the present study. Specifically, we predicted that the priming effect, corresponding to the repetition suppression of the N400 component, for marine mammal target words would be greater when preceded by fish than by terrestrial mammals in the left ATL, whereas the opposite pattern could be observed in the left IFC.\u003c/p\u003e \u003cp\u003eOur findings support a functional dissociation between these regions: the left ATL and IFC were more engaged in task-independent similarity-based processing and task-relevant processing, respectively. These results suggest that the brain flexibly recruits distinct neural systems depending on whether a task emphasizes similarity or categorical relationships, shedding light on the neural architecture underlying semantic processing.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eNine healthy, right-handed native Japanese speakers (six men and three women; mean age [standard deviation, SD]\u0026thinsp;=\u0026thinsp;26.3 [12.3] years, age range: 21\u0026ndash;59) participated in the experiment. All had normal or corrected-to-normal vision and no history of neurological or psychological disorders. The study was approved by the ethics committee of the National Institute of Information and Communications Technology and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent for participation and publication of data was obtained from all participants prior to the study.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStimuli and task\u003c/h3\u003e\n\u003cp\u003eThe MEG experiments used the same stimuli, task, and semantic priming paradigm as described in a previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), with a modification to the timing protocol. A total of 40 animal names\u0026mdash;10 each from the biological categories of fish (F; e.g., carp), terrestrial mammals (TM; e.g., cat), marine mammals (MM; e.g., dolphin), and birds (B; e.g., chicken)\u0026mdash;were used as prime words. These same words, along with 10 artificial object names (man-made objects, e.g., knife), were also used as target words. Semantic distances between animal names were derived from questionnaire data reported in the previous study (Online Resource, Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e) and are expressed on a 0\u0026ndash;16 scale, with lower values indicating greater semantic similarity.\u003c/p\u003e \u003cp\u003eIn each trial (epoch), participants were visually presented with a prime word selected from the animal lists, followed by a target word drawn either from the same animal lists (but not identical to the prime) or from the list of artificial objects (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Participants were instructed to judge whether the target word referred to a living or nonliving entity and to respond by pressing one of two buttons after the response cue appeared, based on their decision. Stimuli were presented and behavioral responses were recorded using Presentation software (Neurobehavioral Systems, Inc., Berkeley, CA, USA). The stimulus presentation protocol was identical to that of the previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), with the exception of the cue onset latency. In the MEG experiment, the response cue appeared 1,500 ms after the onset of the target word, compared to 1,000 ms in the fMRI experiment. This adjustment allowed participants to blink during the interval between the end of the MEG epoch (1,000 ms after target word onset) and the cue onset. Each participant was presented with all possible prime\u0026ndash;target combinations, excluding identical prime\u0026ndash;target pairs (1,960 combinations in total). These were randomized and divided into four blocks for use across four experimental runs. Block order was randomized for each participant. All participants completed four runs, except one (S1), whose second run was interrupted and subsequently divided into two runs, resulting in five total runs.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eData acquisition\u003c/h3\u003e\n\u003cp\u003eMEG data were recorded from all participants using a 148-channel Magnes 2500WH system (4-D Neuroimaging, San Diego, CA, USA). Individual head shapes were measured before conducting four experimental runs, with rest periods between runs. Three markers (vitamin E) were affixed to the nose and both ears to enable coordinate alignment between MRI and MEG. Marker locations were measured at the start of each run to position the magnetic sensors relative to the head. MEG signals were recorded within a frequency bandwidth of 0\u0026ndash;200 Hz, from 200 ms prior to 1,000 ms following the onset of the target word, at a sampling rate of 678.17 Hz. Structural MRIs were acquired using a 1.5-T MAGNETOM Vision system (Siemens AG, M\u0026uuml;nchen, Germany) with a magnetization-prepared rapid acquisition gradient echo protocol, using the following parameters: TR, 9.7 ms; TE, 4 ms; FA, 12\u0026deg;; slice thickness, 1 mm; image matrix, 256 \u0026times; 256; pixel size, 1 \u0026times; 1 mm. These structural images were employed to localize potential neural sources within the cerebral cortex and to construct a lead field matrix describing the relationship between magnetic sensor outputs and source currents (dipole moments).\u003c/p\u003e\n\u003ch3\u003eAnalysis\u003c/h3\u003e\n\u003cp\u003ePreprocessing of MEG data was performed to reduce noise and baseline drift. First, independent component analysis (ICA) was applied to the raw MEG data using EEGLAB v4.5b (Swartz Center for Computational Neuroscience, La Jolla, CA, USA). Components corresponding to heartbeats, trends, spikes, and those exhibiting contour maps characteristic of blinks and eye movements were removed based on visual inspection (Online Resource, Table S2). The remaining ICA components were subsequently back-transformed into MEG data. Second, the data were bandpass filtered between 0.1 and 40 Hz in both forward and backward directions and screened for artifacts using peak-to-peak thresholds (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(a)). Specifically, individual epochs were rejected if the peak-to-peak amplitude in any channel exceeded a subject-specific threshold (several pT). Approximately 10% of all MEG epochs were excluded. The resulting averaged MEG data were used to identify active locations (major dipole positions). Third, to further reduce trends and noise in estimating activation strengths (moments of major dipoles), the MEG data were bandpass filtered twice between 1 and 8 Hz in forward and backward directions (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(b)). Additionally, linear trends were removed from three types of averaged MEG data detailed later. Trends were estimated from latency windows of \u0026minus;\u0026thinsp;200 to 0 ms and 800 to 1,000 ms relative to target word onset for each channel. The bandpass filtering with a 1 Hz low-frequency cutoff combined with detrending effectively mitigated trends of up to a few hundred fT present in the data.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAs an additional preprocessing step, the MEG data from multiple runs were transformed to represent measurements at a single sensor location (Uutela et al. \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) before applying a 0.1\u0026ndash;40 Hz bandpass filter, since the data had been recorded using sensors positioned at different locations across runs. To convert MEG data from sensor location A to sensor location B, source activation was estimated from the MEG data at sensor location A using the pseudo-inverse of the lead field matrix, followed by derivation of MEG data at sensor location B through forward calculations. One run was selected as a reference, and the MEG data from all other runs were transformed to correspond to the sensor location of the reference run for each participant. The MEG data from the reference run were excluded from further analysis, as amplitude was slightly smaller for the transformed than for the non-transformed data, so that the inclusion of the latter would increase data variations.\u003c/p\u003e \u003cp\u003eFor each participant, an overall average MEG dataset (all-average MEG data) was obtained by averaging all artifact-free transformed MEG epochs associated with correct responses in the living/nonliving decision task. Additionally, subaverage and category-average MEG data were derived from artifact-free transformed epochs with correct responses as follows: To analyze dependence on semantic distance, consecutive series of 10 MEG epochs arranged in descending order of semantic distance were averaged to generate subaverage MEG data. The choice of 10 epochs per subaverage reflected a balance between statistical power for detecting semantic distance dependence and accuracy of detrending (see \u003cspan refid=\"Sec17\" class=\"InternalRef\"\u003eAppendix\u003c/span\u003e). In total, 866 subaverages were obtained across all participants. For analysis of category dependence, MEG epochs obtained during TM\u0026ndash;MM and F\u0026ndash;MM pairs were averaged separately to yield category-specific averages for each participant. Each category pair comprised 40\u0026ndash;80 MEG epochs per participant.\u003c/p\u003e \u003cp\u003eNeural sources were estimated using the method described in a previous report (Fujimaki et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), except that the grouping procedure was omitted. Custom software implementing this method was developed in MATLAB R2008a (MathWorks Inc., Natick, MA, USA). The method first determined the major dipoles and their locations from the all-average MEG data, where each dipole represents an electrical model of a neural source. Moments of these dipoles were then fitted to three types of average MEG data to obtain corresponding moment estimates, which served as measures of neural activation. The original method included a grouping procedure in which moments of neighboring dipoles were summed. This was based on the fact that neighboring dipoles are subject to crosstalk; that is, their moments tend to become more dependent as spatial distance decreases, so the summed moment provides a better representation of the sources for the overall MEG signal than individual moments. In the present study, to focus on major dipoles located near fMRI foci, this procedure was omitted, and activation analysis employed moments of individual major dipoles. Using this source estimation method combined with regression analysis, the minimum detectable relative change in activation attributable to semantic distance dependence was estimated to be \u0026plusmn;\u0026thinsp;19% for the parameters of the present study (see \u003cspan refid=\"Sec17\" class=\"InternalRef\"\u003eAppendix\u003c/span\u003e). A previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) reported an observed relative change of approximately\u0026thinsp;\u0026plusmn;\u0026thinsp;40%, roughly twice the estimated minimum detectable level. Therefore, semantic distance dependence should be detectable, provided inter-participant variation remains within a residual margin of approximately\u0026thinsp;\u0026plusmn;\u0026thinsp;20%.\u003c/p\u003e \u003cp\u003eDetails of the source estimation procedures in the present study are as follows. First, dipole locations were obtained at ~\u0026thinsp;3-mm intervals on the diluted cerebral cortex surface, along with a lead field matrix based on a realistic three-compartment boundary-element model of medium resolution, using CURRY V8 (Compumedics Ltd., Abbotsford, Australia) and participant-specific structural MRIs. At each location, two dipoles were placed with orthogonal moments aligned along the two strongest components derived from singular value decomposition of the lead field matrix, as described previously (Fujimaki et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Dipole moment values were estimated by minimizing the blockwise \u003cem\u003el\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e-norm (Matsuura and Okabe \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1995\u003c/span\u003e; Uutela et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1999\u003c/span\u003e; Haufe et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Terazono et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) using the SeDuMi 1.2 solver (Lehigh University, Bethlehem, PA, USA), applied to all-average MEG data filtered with a 0.1\u0026ndash;40 Hz passband. The major dipole selection method was similar to that in the previous report (Fujimaki et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), with an extension to accommodate time-varying data. For each dipole location, the peak value of the moment magnitude, weighted by the lead field matrix, was calculated over a latency range of 50\u0026ndash;600 ms. Major dipoles were selected from local maxima of the weighted magnitudes, ensuring a minimum separation greater than 20 mm between dipoles, a threshold determined by a previous simulation study (Fujimaki et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2002\u003c/span\u003e). This procedure yielded 60\u0026ndash;70 major dipoles per participant.\u003c/p\u003e \u003cp\u003eThe moments of the major dipoles were fitted to all-average, subaverage, and category-average MEG data at each sampling time by multiplying the average MEG data\u0026mdash;filtered with a 1\u0026ndash;8 Hz passband and detrended\u0026mdash;by the pseudo-inverse of the lead field matrix. The major dipole moments of subaverage and category-average were projected onto those of all-average at each sampling time. The projected moments, rather than moment magnitudes, were used as a measure of neural activity to minimize noise components spatially orthogonal to signal components.\u003c/p\u003e \u003cp\u003eThe analysis focused on major dipoles located near active foci identified in fMRI studies, specifically the left ATL and VLPFC (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), to assess spatiotemporal activation patterns associated with semantic processing of similarity and category. Due to inter-participant variability in the locations of selected major dipoles, the term IFC is used to denote active locations in and around the VLPFC. Talairach coordinates of these foci\u0026mdash;(\u0026minus;\u0026thinsp;50, 3, \u0026minus;\u0026thinsp;27) mm for the left ATL and (\u0026minus;\u0026thinsp;53, 24, 6) mm for the left IFC\u0026mdash;were obtained from the previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). According to Talairach Client Version 2.4.3 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://talairach.org/index.html/\u003c/span\u003e\u003cspan address=\"https://talairach.org/index.html/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, Research Imaging Institute, San Antonio, TX, USA), these coordinates correspond to Brodmann areas 21 and 45, respectively. For each participant and region (i.e., left ATL and IFC), major dipoles were selected based on one of the following two criteria: (1) the dipole located nearest to the corresponding fMRI focus or (2) slightly more distant dipoles demonstrating significant dependence on semantic distance in participant-level regression analysis (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), provided they were within 35 mm of the fMRI focus and the nearest dipole lacked significant dependence.\u003c/p\u003e \u003cp\u003eFor analysis of semantic distance dependence, projected moment values of subaverage were averaged within 100 ms time windows, advancing in 25 ms intervals. After excluding outliers exceeding three SDs from the mean, a linear mixed-effects model was applied using the \u0026ldquo;lme4\u0026rdquo; package (Bates et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) in RStudio (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://posit.co/download/rstudio-desktop/\u003c/span\u003e\u003cspan address=\"https://posit.co/download/rstudio-desktop/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, Posit PBC, Boston, MA, USA). Semantic distance was modeled as a fixed effect, while participant main effects and participant by semantic distance interactions were modeled as random effects. Variances were estimated via restricted maximum likelihood. For analysis of category dependence, projected moment values of category-average obtained during TM\u0026ndash;MM and F\u0026ndash;MM pair presentations were averaged separately within 100 ms time windows, advancing in 25 ms intervals, and compared using paired-samples t-tests in SPSS Version 19 (IBM Corporation, Armonk, NY, USA). In both analyses, significance levels were corrected for multiple comparisons across areas and latencies by the Benjamini\u0026ndash;Hochberg method (Benjamini and Hochberg \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1995\u003c/span\u003e), controlling the false discovery rate (FDR) at 0.05.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe present MEG study employed the\u0026nbsp;stimuli of animal names F, TM, MM, B, and of artificial objects, task (identification of a target word as living [F, TM, MM, or B] or nonliving [artificial object]), and priming paradigm (Fig. 1) to investigate differences in spatiotemporal neural activation patterns related to semantic processing of similarity and category.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBehavior\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePrime words consisted of animal names from categories F, TM, MM, or B, whereas target words included these animal names (excluding the identical prime word) and names of artificial objects. Participants responded by pressing one of two buttons after a response cue, indicating whether the target word was living or nonliving. Response accuracy exceeded 97% in artifact-free epochs for all participants, demonstrating successful task execution.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eActive locations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter preprocessing, participant-specific all-average MEG data were generated by averaging all epochs, excluding those with incorrect responses or artifacts (Fig. 2). Source estimation applied to these data identified 60\u0026ndash;70 major dipoles per participant. The major dipoles near fMRI foci were selected for each participant (Fig. 3), according to criteria (1) and (2) described in the Materials and Methods section.\u003c/p\u003e\n\u003cp\u003eFor three participants in each region (out of nine), the dipoles selected were not the nearest but met significance criterion (2). For the remaining six participants, the nearest major dipoles were selected (criterion (1), see Online Resource, Table S3a and b). The mean Talairach coordinates of selected dipoles were (\u0026minus;49, \u0026minus;1, \u0026minus;23) mm for the left ATL and (\u0026minus;53, 18, 10) mm for the left IFC. The average Euclidean distances between selected dipoles and fMRI foci were 15.7 mm (SD 9.0) for the left ATL and 17.0 mm (SD 11.2) for the left IFC, yielding an overall average distance of 16.4 mm (SD 9.9).\u0026nbsp;In contrast, an overall average distance between the nearest major dipoles and fMRI foci was\u0026nbsp;10.9\u0026nbsp;mm (SD\u0026nbsp;5.6)\u0026nbsp;(Online Resource, Table S3a).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDependence on semantic distance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor each participant, subaverage MEG data were generated by averaging consecutive 10-epoch segments ordered by semantic distance, excluding epochs with incorrect responses or artifacts. Moments of the major dipoles were fitted to these subaverages and projected onto moments fitted to the all-average MEG data to reduce orthogonal noise. The subaverages were further averaged within 100-ms time windows with 25-ms steps to enhance signal-to-noise ratio. A total of 866 subaverage values of projected moments were obtained across participants for each time window and used as a measure of neural activity after excluding outliers exceeding three SDs from the mean.\u003c/p\u003e\n\u003cp\u003eTo identify regions and latencies showing dependence on semantic distance, projected moments of subaverage were split into high and low semantic distance groups by dividing data into halves and averaged separately, yielding mean semantic distances of 12.6 and 8.9, respectively. Comparison revealed mainly two activity difference peaks between the two groups within the left ATL and IFC at latencies approximately 200 to 600 ms following target word onset (Fig. 4). Considering statistical power, focus was placed on the two largest difference peaks: center-latency of 550 ms and 325 ms for the left ATL, and 525 ms and 350 ms for the left IFC (Online Resource, Table S4a). A linear mixed-effects model analysis, conducted for each region and 100-ms time window with significance level corrected for multiple comparisons using the Benjamini\u0026ndash;Hochberg procedure, demonstrated significant dependence on semantic distance (FDR = 0.05) at 350 ms (p = 0.0250) and 525 ms (p = 0.00952) for the left IFC; no significant effects were observed for the left ATL (Online Resource, Table S4b). Time windows exhibiting significant dependence are marked in Fig. 4. The significant dependences displayed positive regression line slopes, i.e., neural activation increased with greater semantic distance. Figure 5 illustrates a scatterplot of projected moments of subaverage versus semantic distance with a regression line for the left IFC at a center-latency of 525 ms. The regression slope, number of data points, and cross-correlation coefficient were 0.0104 nAm, 860, and 0.132, respectively (Online Resource, Table S4a).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDependence on biological category\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDependence on biological category was tested using category-average MEG data obtained from each participant during the presentation of TM\u0026ndash;MM and F\u0026ndash;MM pairs, following the exclusion of epochs containing artifacts or incorrect responses. The moments of the major dipoles were fitted to the category-average MEG data and projected onto the moments fitted to the all-average MEG data to suppress orthogonal noise. The resulting signals were then averaged within 100-ms time windows at 25-ms intervals to further reduce noise. This process yielded one projected moment value for each category, area, participant, and time window. Figure 6 presents the projected category-average moments averaged across participants, along with the differences between conditions (MM followed by F or TM). Both the left ATL and left IFC exhibited mainly two activity difference peaks between TM\u0026ndash;MM and F\u0026ndash;MM pairs at center latencies ranging from approximately 200 to 600 ms. To ensure sufficient statistical power, analysis focused on the largest and second-largest peaks: at center latencies of 225 ms and 550 ms for the left ATL, and 325 ms, 525 ms, and 550 ms (the latter two showing nearly identical differences) for the left IFC. A paired-samples t-test, followed by the Benjamini\u0026ndash;Hochberg procedure correcting multiple comparisons across areas and latencies, demonstrated that activation was significantly lower for F\u0026ndash;MM pairs compared to TM\u0026ndash;MM pairs (FDR = 0.05) at a center latency of 225 ms in the left ATL (p = 0.0198) and at 525 ms in the left IFC (p = 0.0190) (Online Resource, Table S5a and S5b). Time windows with significant differences are indicated in Fig. 6.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eActive locations and latencies\u003c/h2\u003e \u003cp\u003eUsing blockwise \u003cem\u003el\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e-norm minimization (Terazono et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2010\u003c/span\u003e), 60\u0026ndash;70 major dipoles were detected across the whole brain in each participant. To evaluate the dependence of neural activation on similarity and category, major dipoles located near activation foci identified in a previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) were selected. The mean distance between the fMRI-identified foci and the nearest major dipoles detected in the present MEG study was 10.9 mm (Online Resource, Table S3a). This distance is considered reasonable, as a prior simulation study using the same source estimation method reported a mean localization error of 6.3 mm (SD: 3.2 mm) between assumed source locations and the nearest detected major dipoles (Fujimaki et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). That simulation assumed simultaneous activation of five brain areas during visual word processing and an MEG signal-to-noise ratio (SNR) of 10. These conditions are comparable to those in the present study, which involved visual word stimuli and yielded a mean SNR of 9.02 across all participants in the all-average MEG data used to identify major dipole locations (Online Resource, Table S3b). Additional variability in the observed nearest dipole locations may reflect factors such as registration errors of several millimeters between MEG and fMRI coordinate systems. The mean distance between the fMRI activation foci and the selected major dipoles in the present MEG study was 16.4 mm (Online Resource, Table S3a), which is greater than the 10.9 mm distance observed for the nearest dipoles. This larger value may further reflect inter-individual variability in the locations of active foci.\u003c/p\u003e \u003cp\u003eThe present MEG study revealed significant dependence of activation on semantic distance in the left IFC at center latencies of 350 ms and 525 ms following target word onset (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), and significant dependence of activation on category in the left ATL and IFC at center latencies of 225 ms and 525 ms, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). These latencies, corresponding to semantic processing, are consistent with prior MEG studies that reported early activity in the left ATL and IFC within a latency range of 200\u0026ndash;400 ms during visual word processing (Halgren et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Marinkovic et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Fujimaki et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Pylkk\u0026auml;nen \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), as well as numerous EEG studies indicating that the N400 component, associated with semantic processing, typically occurs between 200 ms and 600 ms (Kutas and Federmeier \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eDependence on semantic distance and category\u003c/h2\u003e \u003cp\u003eBased on findings from a previous fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), we hypothesized that activation in the left ATL and IFC would reflect similarity- and category-based processing, respectively. However, the present MEG results showed that activation for the marine mammal target was reduced\u0026mdash;or, equivalently, that the priming effect was larger\u0026mdash;when preceded by a fish prime than by a terrestrial mammal prime in the left ATL and IFC at center latencies of 225 ms and 525 ms, respectively. This pattern indicates that activation in both regions reflected similarity-based processing. Therefore, the hypothesis regarding the role of the left IFC was not supported, as activation at 525 ms reflected similarity- rather than category-based processing. This discrepancy may be attributable to differences in the temporal resolution of MEG and fMRI. In our previous fMRI study, activation was averaged over several seconds due to the hemodynamic delay, potentially conflating similarity- and category-related signals. Additionally, later latency components may have been attenuated in the current MEG study due to the 1\u0026ndash;8 Hz bandpass filtering applied to the data. Taken together, the MEG and fMRI findings suggest that activation in the left IFC reflects similarity-based processing up to approximately 600 ms following target word onset, with category-based processing emerging at later latencies. This delayed onset of category-related activity in the left IFC aligns with the following behavioral findings.\u003c/p\u003e \u003cp\u003eIn a prior behavioral study using the same stimuli and task as the fMRI study\u0026mdash;except that participants were instructed to make a living/nonliving judgment immediately after target word presentation\u0026mdash;the reaction time (RT) was approximately 600 ms (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These findings suggest that categorical processing in the left IFC occurred as a post-response process. In that study, RTs for the marine mammal target varied systematically with the semantic relationship between prime and target: responses were significantly faster for the fish\u0026ndash;marine mammal pair (mean RT: 581 ms) than for the terrestrial mammal\u0026ndash;marine mammal pair (589 ms), and significantly faster for the terrestrial mammal\u0026ndash;marine mammal pair than for the bird\u0026ndash;marine mammal pair (596 ms). This RT pattern paralleled the semantic distances between primes and the target; responses were faster when semantic distance was smaller. The priming effect on RT indicates greater facilitation when the prime and target shared more semantic features, reflecting greater repetition of semantic processing. Thus, RTs appear to be driven by similarity-based processing. This interpretation is plausible, as a living/nonliving judgment does not require explicit identification of the target\u0026rsquo;s specific category (e.g., mammal vs. fish). Therefore, the proposed interpretation that category-based processing in the left IFC occurs post-response is consistent with the view that RT primarily reflects similarity-based rather than category-based processing. Furthermore, the observed shift from similarity- to category-based processing suggests that the living/nonliving task entails both recognition of the target animal and access to its biological category. In the case of a marine mammal target, the left IFC may initially retrieve a broad set of relevant features from semantic representation, followed by more selective retrieval of features associated with the mammal category. This interpretation is consistent with the CSC model, which posits that the left IFC supports controlled semantic retrieval according to task demands.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThe present study has a few limitations. First, the sample size was small. Second, the short epoch length of 1.2 s (from 0.2 s before to 1.0 s after target word onset) may have precluded the capture of later neural events. Despite these constraints, the study successfully demonstrated sensitivity to both semantic distance and category effects. This was achieved by filtering the data within a narrow passband (1\u0026ndash;8 Hz), removing linear trends to minimize noise and drift, and focusing on the largest and second-largest peaks in activation differences. Specifically, effects were analyzed for differences between high and low semantic distance conditions using a linear mixed-effects model, and between fish and terrestrial mammal primes using paired-samples t-tests. Further studies are warranted to more precisely characterize the temporal dynamics of semantic processing. Future work should (1) collect more data to allow finer-grained temporal analysis, (2) record continuous MEG signals instead of pre-segmented epochs to enable broader-band filtering (e.g., 0.1\u0026ndash;40 Hz), which would reduce trends while preserving slowly evolving signals, and (3) extend the epoch duration to capture potential late-latency events time-locked to target onset.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this MEG study, we used the same stimuli, task, and priming paradigm as our prior fMRI study (Matsumoto et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) to investigate the spatiotemporal characteristics of neural activity underlying semantic processing related to similarity and biological category. We hypothesized that the left ATL would be involved in similarity processing and the left IFC in category processing. While the role of the left ATL was supported, the left IFC was instead associated with similarity-based processing at approximately 525 ms post-target onset. These findings suggest that category-based processing in the left IFC occurs later, beyond approximately 600 ms. Although the living/nonliving decision task did not explicitly require identification of biological category, the results indicate that such categorical information was likely retrieved in the left IFC as a post-response process. Together with our previous fMRI findings, the present results support both the hub-and-spoke model and the CSC model, wherein the left ATL provides semantic representations and the left IFC facilitates controlled retrieval of task-relevant semantic features.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank Editage for English language editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eN.F., A.M., and A.S.I. designed the study. T.S. produced the materials. A.S.I. conducted the experiments. N.F. analyzed the data. The results were discussed by all authors. N.F. and A.S.I. wrote the first draft of the manuscript, and all authors revised the manuscript and approved the final version.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by JSPS KAKENHI Grant Numbers JP16K01969 and JP22K12758.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eBates D, M\u0026auml;chler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1\u0026ndash;48. https://doi.org/10.18637/jss.v067.i01\u003c/li\u003e\n \u003cli\u003eBenjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289\u0026ndash;300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x\u003c/li\u003e\n \u003cli\u003eChen Y, Shimotake A, Matsumoto R, et al. (2016) The \u0026ldquo;when\u0026rdquo; and \u0026ldquo;where\u0026rdquo; of semantic coding in the anterior temporal lobe: temporal representational similarity analysis of electrocorticogram data. Cortex 79:1\u0026ndash;13. https://doi.org/10.1016/j.cortex.2016.02.015\u003c/li\u003e\n \u003cli\u003eDe Deyne S, Verheyen S, Ameel E, Vanpaemel W, Dry MJ, Voorspoels W, Storms G (2008) Exemplar by feature applicability matrices and other Dutch normative data for semantic concepts. Behav Res Methods 40:1030\u0026ndash;1048. https://doi.org/10.3758/BRM.40.4.1030\u003c/li\u003e\n \u003cli\u003eFriederici AD (2011) The brain basis of language processing: from structure to function. Physiol Rev 91:1357\u0026ndash;1392. https://doi.org/10.1152/physrev.00006.2011\u003c/li\u003e\n \u003cli\u003eFujimaki N, Hayakawa T, Ihara A, et al. (2009) Early neural activation for Lexico-semantic access in the left anterior temporal area analyzed by an fMRI-assisted MEG multidipole method. Neuroimage 44:1093\u0026ndash;1102. https://doi.org/10.1016/j.neuroimage.2008.10.021\u003c/li\u003e\n \u003cli\u003eFujimaki N, Hayakawa T, Nielsen M, Kn\u0026ouml;sche TR, Miyauchi S (2002) An fMRI-constrained MEG source analysis with procedures for dividing and grouping activation. Neuroimage 17:324\u0026ndash;343. https://doi.org/10.1006/nimg.2002.1160\u003c/li\u003e\n \u003cli\u003eFujimaki N, Terazono Y, Ihara AS, Hayakawa T, Matani A, Umehara H (2014) Accuracy of estimating strengths of dipole moments from a small number of magnetoencephalographic trial data. Brain Topogr 27:635\u0026ndash;647. https://doi.org/10.1007/s10548-014-0363-9\u003c/li\u003e\n \u003cli\u003eGonzalez Alam T, Karapanagiotidis T, Smallwood J, Jefferies E (2019) Degrees of lateralisation in semantic cognition: evidence from intrinsic connectivity. Neuroimage 202:116089. https://doi.org/10.1016/j.neuroimage.2019.116089\u003c/li\u003e\n \u003cli\u003eHalgren E, Dhond RP, Christensen N, Van Petten C, Marinkovic K, Lewine JD, Dale AM (2002) N400-like magnetoencephalography responses modulated by semantic context, word frequency, and lexical class in sentences. Neuroimage 17:1101\u0026ndash;1116. https://doi.org/10.1006/nimg.2002.1268\u003c/li\u003e\n \u003cli\u003eHaufe S, Nikulin VV, Ziehe A, M\u0026uuml;ller KR, Nolte G (2008) Combining sparsity and rotational invariance in EEG/MEG source reconstruction. Neuroimage 42:726\u0026ndash;738. https://doi.org/10.1016/j.neuroimage.2008.04.246\u003c/li\u003e\n \u003cli\u003eHoffman P, Lambon Ralph MA (2018) From percept to concept in the ventral temporal lobes: Graded hemispheric specialisation based on stimulus and task. Cortex 101:107\u0026ndash;118.\u003c/li\u003e\n \u003cli\u003e\u003ca href=\"https://doi.org/10.1016/j.cortex.2018.01.015\"\u003ehttps://doi.org/10.1016/j.cortex.2018.01.015\u003c/a\u003e\u003c/li\u003e\n \u003cli\u003eJefferies E, Thompson H, Cornelissen P, Smallwood J (2020) The neurocognitive basis of knowledge about object identity and events: dissociations reflect opposing effects of semantic coherence and control. Philos Trans R Soc Lond B Biol Sci 375:20190300. \u0026nbsp;\u003ca href=\"https://doi.org/10.1098/rstb.2019.0300\"\u003ehttps://doi.org/10.1098/rstb.2019.0300\u003c/a\u003e\u003c/li\u003e\n \u003cli\u003eKutas M, Federmeier KD (2011) Thirty years and counting: finding meaning in the N400 component of the event-related brain potential (ERP). Annu Rev Psychol 62:621\u0026ndash;647. https://doi.org/10.1146/annurev.psych.093008.131123\u003c/li\u003e\n \u003cli\u003eKutas M, Hillyard SA (1980) Reading senseless sentences: brain potentials reflect semantic incongruity. Science 207:203\u0026ndash;205. https://doi.org/10.1126/science.7350657\u003c/li\u003e\n \u003cli\u003eLambon Ralph MA, Jefferies E, Patterson K, Rogers TT (2017) The neural and computational bases of semantic cognition. Nat Rev Neurosci 18:42\u0026ndash;55. https://doi.org/10.1038/nrn.2016.150\u003c/li\u003e\n \u003cli\u003eMarinkovic K, Dhond RP, Dale AM, Glessner M, Carr V, Halgren E (2003) Spatiotemporal dynamics of modality-specific and supramodal word processing. Neuron 38:487\u0026ndash;497. https://doi.org/10.1016/s0896-6273(03)00197-1\u003c/li\u003e\n \u003cli\u003eMatsumoto A, Soshi T, Fujimaki N, Ihara AS (2021) Distinctive responses in anterior temporal lobe and ventrolateral prefrontal cortex during categorization of semantic information. Sci Rep 11:13343. https://doi.org/10.1038/s41598-021-92726-7\u003c/li\u003e\n \u003cli\u003eMatsuura K, Okabe Y (1995) Selective minimum-norm solution of the biomagnetic inverse problem. IEEE Trans Bio Med Eng 42:608\u0026ndash;615. https://doi.org/10.1109/10.387200\u003c/li\u003e\n \u003cli\u003eNoonan KA, Jefferies E, Visser M, Lambon Ralph MA (2013) Going beyond inferior prefrontal involvement in semantic control: evidence for the additional contribution of dorsal angular gyrus and posterior middle temporal cortex. J Cogn Neurosci 25:1824\u0026ndash;1850. https://doi.org/10.1162/jocn_a_00442\u003c/li\u003e\n \u003cli\u003ePatterson K, Nestor PJ, Rogers TT (2007) Where do you know what you know? The representation of semantic knowledge in the human brain. Nat Rev Neurosci 8:976\u0026ndash;987. https://doi.org/10.1038/nrn2277\u003c/li\u003e\n \u003cli\u003ePrice CJ (2012) A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading. Neuroimage 62:816\u0026ndash;847. https://doi.org/10.1016/j.neuroimage.2012.04.062\u003c/li\u003e\n \u003cli\u003ePylkk\u0026auml;nen L (2020) Neural basis of basic composition: what we have learned from the red-boat studies and their extensions. Philos Trans R Soc Lond B Biol Sci 375:20190299. https://doi.org/10.1098/rstb.2019.0299\u003c/li\u003e\n \u003cli\u003eRice GE, Lambon Ralph MA, Hoffman P (2015) The Roles of Left Versus Right Anterior Temporal Lobes in Conceptual Knowledge: An ALE Meta-analysis of 97 Functional Neuroimaging Studies. Cereb Cortex 25:4374-4391. https://doi.org/10.1093/cercor/bhv024\u003c/li\u003e\n \u003cli\u003eSoshi T, Fujimaki N, Matsumoto A, Ihara AS (2017) Memory-based specification of verbal features for classifying animals into super-ordinate and sub-ordinate categories. Front Commun 2:1\u0026ndash;14. https://doi.org/10.3389/fcomm.2017.00012\u003c/li\u003e\n \u003cli\u003eTaylor JR (1995) Linguistic categorization. Oxford University Press\u003c/li\u003e\n \u003cli\u003eTerazono Y, Fujimaki N, Murata T, Matani A (2010) Point source reconstruction principle of linear inverse problems. Inverse Probl 26:115016. https://doi.org/10.1088/0266-5611/26/11/115016\u003c/li\u003e\n \u003cli\u003eUutela K, H\u0026auml;m\u0026auml;l\u0026auml;inen M, Somersalo E (1999) Visualization of magnetoencephalographic data using minimum current estimates. Neuroimage 10:173\u0026ndash;180. https://doi.org/10.1006/nimg.1999.0454\u003c/li\u003e\n \u003cli\u003eUutela K, Taulu S, H\u0026auml;m\u0026auml;l\u0026auml;inen M (2001) Detecting and correcting for head movements in neuromagnetic measurements. Neuroimage 14:1424\u0026ndash;1431. https://doi.org/10.1006/nimg.2001.0915\u003cstrong\u003e\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Semantic distance, Similarity, Category, Animal name, Anterior temporal lobe, Inferior frontal cortex","lastPublishedDoi":"10.21203/rs.3.rs-6947110/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6947110/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSemantic processing of similarity and category are closely related yet distinct. In a prior functional magnetic resonance imaging study, the neural correlates of similarity and category processing were examined using prime–target word pairs in which similarity and category were independently manipulated. Specifically, they included \"marine mammal\", which belongs to the mammal category,but is more similar tofish. In some trials, two animal names from the categories of fish, terrestrial mammals, marine mammals, and birds were presented sequentially to investigate the priming effect of the first name on the processing of the second. In other trials, the second word was an inanimate object. Participants classified the second word as living or nonliving in all trials. Neural activity in the left anterior temporal lobe reflected similarity processing, whereas activityin the left inferior frontal cortex (IFC) corresponded to task-relevant category processing, consistent with models positing distinct systems for semantic representation and semanticcontrol. The present study employed magnetoencephalography to measure neural activity during thesame task, achieving enhanced temporal resolution. Neural activity reflecting similarity processing was observed in the left IFC approximately 525 msafter target word onset. This finding suggests that category processing in the left IFC occurred after the living/nonliving decision, which had a reaction time near 600 ms. The present results thus reveal a dynamic temporal sequence in the retrieval of task-relevant semantic features.\u003c/p\u003e","manuscriptTitle":"Distinct spatiotemporal activity patterns reflecting similarity-based and category-based semantic processing in anterior temporal lobe and inferior frontal cortex revealed via magnetoencephalography","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-30 08:53:40","doi":"10.21203/rs.3.rs-6947110/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4034fbd6-36d4-4079-9772-0926ba4515d5","owner":[],"postedDate":"June 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T12:41:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-30 08:53:40","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6947110","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6947110","identity":"rs-6947110","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}