Skip to main content
NCBI home page
Human Brain Mapping logoLink to Human Brain Mapping
. 2024 Apr 24;45(6):e26681. doi: 10.1002/hbm.26681

Semantic context‐dependent neural representations of odors in the human piriform cortex revealed by 7T MRI

Toshiki Okumura 1,2, Ikuhiro Kida 2,3, Atsushi Yokoi 2,3, Tomoya Nakai 2, Shinji Nishimoto 2,3, Kazushige Touhara 1,4,✉, Masako Okamoto 1,✉
PMCID: PMC11041378  PMID: 38656060

Abstract

Olfactory perception depends not only on olfactory inputs but also on semantic context. Although multi‐voxel activity patterns of the piriform cortex, a part of the primary olfactory cortex, have been shown to represent odor perception, it remains unclear whether semantic contexts modulate odor representation in this region. Here, we investigated whether multi‐voxel activity patterns in the piriform cortex change when semantic context modulates odor perception and, if so, whether the modulated areas communicate with brain regions involved in semantic and memory processing beyond the piriform cortex. We also explored regional differences within the piriform cortex, which are influenced by olfactory input and semantic context. We used 2 × 2 combinations of word labels and odorants that were perceived as congruent and measured piriform activity with a 1‐mm isotropic resolution using 7T MRI. We found that identical odorants labeled with different words were perceived differently. This labeling effect was observed in multi‐voxel activity patterns in the piriform cortex, as the searchlight decoding analysis distinguished identical odors with different labels for half of the examined stimulus pairs. Significant functional connectivity was observed between parts of the piriform cortex that were modulated by labels and regions associated with semantic and memory processing. While the piriform multi‐voxel patterns evoked by different olfactory inputs were also distinguishable, the decoding accuracy was significant for only one stimulus pair, preventing definitive conclusions regarding the locational differences between areas influenced by word labels and olfactory inputs. These results suggest that multi‐voxel patterns of piriform activity can be modulated by semantic context, possibly due to communication between the piriform cortex and the semantic and memory regions.

Keywords: human olfaction, piriform cortex, semantic context, ultra‐high field fMRI

Associating distinct word labels with specific odorants not only differentiated perceptions of the same odorants but also differentiated multi‐voxel activity patterns within the primary olfactory cortex, as demonstrated by using ultra‐high field functional magnetic resonance imaging (fMRI).

graphic file with name HBM-45-e26681-g004.jpg

1. INTRODUCTION

Olfactory input is not the sole factor responsible for odor perception (Kaeppler & Mueller, 2013). Identical olfactory inputs can produce different odor perceptions depending on the smelling context, such as experience, satiety, and semantic information. Especially for humans, the semantic contexts formed by the language used are important because humans can experience the world based on their language, unlike other animals. Previous studies have shown that different labels presented with identical odors affect the perceived odor quality (Bae et al., 2019), pleasantness (Ayabe‐Kanamura et al., 1997; Cornell Kärnekull et al., 2021; de Araujo et al., 2005; Djordjevic et al., 2008; Lundström et al., 2006; Manescu et al., 2014), intensity (Cornell Kärnekull et al., 2021; Djordjevic et al., 2008; Manescu et al., 2014), edibility (Manescu et al., 2014), arousal (Djordjevic et al., 2008), and familiarity (Cornell Kärnekull et al., 2021). Human neuroimaging studies have shown that activities in the secondary and/or downstream olfactory areas, such as the anterior cingulate, hippocampus, and medial orbitofrontal cortex, differ when identical odorants are presented with different labels (Bensafi et al., 2014; de Araujo et al., 2005; Gottfried & Dolan, 2003). However, it is still unclear whether and how the primary olfactory cortex plays a role in the perceptual changes caused by semantic context.

Among the primary olfactory areas, the piriform cortex is well known for its role in olfactory perception. This region is located at and around the junction of the frontal and temporal gyri (Mai, 2015; Vaughan & Jackson, 2014) and is the largest recipient of projections from the olfactory bulb (Price, 2009), which relays odor information received by the olfactory receptors to the cerebral cortex (Gottfried, 2010; Lane et al., 2020). The piriform cortex also receives associative inputs from other primary olfactory cortices and their downstream regions (García‐Cabezas & Barbas, 2014; Johnson et al., 2000; Mohedano‐Moriano et al., 2005; Morán et al., 1987; Wang, Zhang, et al., 2020). Single‐cell recordings in rodents revealed that different olfactory inputs induce different ensemble activity patterns in the piriform cortex (Poo & Isaacson, 2009; Rennaker et al., 2007; Roland et al., 2017; Stettler & Axel, 2009; Tantirigama et al., 2017). In humans, functional magnetic resonance imaging (fMRI) studies have revealed that olfactory inputs activate the piriform cortex and that their magnitudes are associated with the intensity and/or pleasantness of odors (Bensafi et al., 2007; Donoshita et al., 2021; Grabenhorst et al., 2007; Rolls et al., 2003; Royet et al., 2003; Zelano et al., 2007). Furthermore, in line with rodent studies, multivariate pattern analysis showed that multi‐voxel activity patterns in the piriform cortex could differentiate finer perceptual differences, such as citrus versus minty odors (Fournel et al., 2016; Gottfried et al., 2006; Howard et al., 2009; Zelano et al., 2011). In addition to olfactory inputs, non‐olfactory semantic information associated with odors, such as videos of odorous items and their names, increased the activity in the piriform cortex, as shown by univariate analysis (Arshamian et al., 2013; González et al., 2006; Karunanayaka et al., 2015; Porada et al., 2019). Although these findings suggest that the piriform cortex receives non‐olfactory semantic information, it remains unclear whether multi‐voxel activity patterns in the piriform cortex are modulated when the semantic context changes odor perception.

When investigating the influence of semantic context on piriform cortex activity, it is important to consider regional differences within the piriform cortex. In rodents, the anterior and posterior parts of the piriform cortex exhibit distinct neuronal connections and functional roles. While both regions receive inputs from the olfactory bulb, the posterior part receives a greater number of inter‐cortical associative inputs and is likely to play a more significant role in representing associative information (Lane et al., 2020). In humans, the piriform cortex can be anatomically divided into frontal and temporal parts (Mai, 2015), and its functional roles have been examined by dividing the region into anterior and posterior parts (Fournel et al., 2016; Gottfried et al., 2002, 2006; Howard et al., 2009; Howard & Gottfried, 2014; Li et al., 2008; Zelano et al., 2011) and frontal and temporal parts (Arnold et al., 2020; Bensafi et al., 2007; Porter et al., 2005; Zelano et al., 2005; Zhou et al., 2019). However, neuronal connections at the cellular level are unknown and a consensus on the locations and boundaries of these subregions has yet to be reached. Because semantic information is transmitted via associative connections from regions related to semantic and memory processing, the effects of semantic context may only be observed in localized areas with dense associative inputs. In this regard, it is important to determine whether the locations affected by semantic context differ from those affected by olfactory input. Both olfactory inputs and semantic contexts modulate olfactory perception, but would evoke neural activity in the piriform cortex via different pathways. Therefore, this comparison would be helpful in understanding the top‐down influence of semantic context on piriform cortex activity.

Here, we raised three research questions: (1) whether the multi‐voxel patterns in the piriform cortex evoked by the same olfactory inputs differ when they are perceived differently in the presence of semantic labels; (2) whether the parts of the piriform cortex modulated by word labels differ from those modulated by olfactory inputs; and (3) whether the parts of the piriform cortex modulated by word labels communicate with regions related to semantic and memory processing. To address these questions, we used 2 × 2 combinations of word labels and odorants for the minty and citrus categories (Figure 1a) and performed a decoding analysis of the fMRI data. Given the relatively small size of the piriform cortex, we acquired fMRI data at an ultra‐high field of 7T with a 1‐mm isotropic resolution. This approach aimed to enhance the sensitivity of our decoding analysis while minimizing potential issues such as partial volume effects and signal leakage from neighboring brain regions. To validate the suitability of our experimental conditions for studying the piriform cortex activity, we initially conducted two analyses that replicated those of previous studies. Subsequently, we performed searchlight decoding analysis within the piriform cortex to classify either labels or odorants. While previous olfactory studies typically constructed decoding models for individual subjects (subject‐wise models) (Bao et al., 2016; Bhutani et al., 2019; Howard et al., 2009; Qu et al., 2016), we also created group models by pooling data from all available subjects, as each approach has its unique advantages (Wang, Cagna, et al., 2020). In addition to evaluating the activities in the piriform cortex, we examined the corresponding perceptual differences between conditions by obtaining dissimilarity and sensory ratings (Figure 1b).

FIGURE 1.

FIGURE 1

Open in a new tab

Experimental procedures. (a) The 2 × 2 combinations of word labels and odorants are shown for each category (minty and citrus). Abbreviations of odorants are written in parentheses. (b) The experimental task used in each session. During fMRI scanning, the intensity ratings were obtained and the order of the four choices presented on the computer monitor was randomized in each trial to decorrelate the odor intensities and finger motions. The other ratings were obtained after fMRI scanning. In all the sessions, stimulus durations were 2.5 s, and inter‐stimulus intervals (ISIs) were 7.5 s. The rating phase lasted until response or a maximum time (3 or 5 s). fMRI, functional magnetic resonance imaging; Max, maximum.

2. MATERIALS AND METHODS

2.1. Subjects

Subjects were recruited through advertisements posted online. The inclusion criteria were 20–39 years old, right‐handed (Edinburgh Handedness Test score > 50), height < 175 cm, body weight < 75 kg, non‐smokers, native Japanese speakers, who had previously participated in other fMRI experiments using 3T or 7T MRI, with no contraindications for MRI, not taking medications that could affect brain function, and having no history of head trauma, respiratory disorders, or psychiatric disorders.

Thirty‐seven subjects (mean, 22.9 years old; SD, 3.54 years old; 11 females) participated in this study with monetary compensation for their participation. Among them, 12 subjects were excluded for the following reasons: six subjects did not detect odors during fMRI scanning (see Section 2.8), five subjects had large head motions during fMRI scanning (see Section 2.7), and one subject did not complete the experiment because of a physical condition. Finally, 25 subjects remained (mean, 23.0 years old; SD, 3.85 years old; 7 females). Among them, five subjects did not participate in the dissimilarity rating session owing to time constraints; therefore, dissimilarity ratings were obtained from 20 subjects (mean, 23.0; SD, 4.14 years old; 5 females).

This study was approved by the ethics committees of the University of Tokyo, and ethics and safety committees of the National Institute of Information and Communications Technology, and was conducted in accordance with the Declaration of Helsinki. All subjects provided written informed consent after receiving a complete explanation of the study.

2.2. Odor delivery

The following odorants were diluted in dipropylene glycol (Sigma‐Aldrich, Darmstadt, Germany): l‐menthol (10% (v/v); Tokyo Chemical Industry Co., Ltd. [TCI], Tokyo, Japan), 1,8‐cineole (1% (v/v); TCI), citral (10% (v/v); TCI), and d‐limonene (10% (v/v); TCI). Odors were presented using a computer‐controlled olfactometer (Atrasuka OM‐1; Asuka Electric Co. Ltd., Osaka, Japan). The olfactometer provided a constant airflow (flow rate, 3.0 L/min) via a Teflon sleeve (length, 10 cm; inner diameter, 12 mm) placed 2–3 cm from the nostrils. The solenoid valves were controlled by a computer program that switched the base odorless airflow to an odorous airflow at the start of the odor presentation and vice versa at the end of the presentation.

2.3. Procedure for fMRI scanning

Each odorant was presented with one of two congruent words (Figure 1). We used odorless air, labeled as odorless, as a control. In each trial, one of the stimuli was presented for 2.5 s after a 2 s count‐down (Figure 1b). Subjects were then asked to rate the odor intensity using a 4‐point Likert scale (no smell, weak, moderate, and strong) by pressing response buttons (HHSC‐2 × 4‐C; Current Designs, Philadelphia, USA) using their right hands. To decorrelate the odor intensities and finger motions, the order of the four choices presented on the computer monitor was randomized in each trial. The time limit for the rating phase was 3 s. The intervals between stimulus offsets and rating onsets were jittered between 0.5 and 1.5 s. The inter‐stimulus interval (ISI) was 7.5 s. Each run consisted of 50 trials, with 5 trials per labeled odorant and 10 trials for odorless air labeled as odorless. Owing to script errors, citrus odorants were subjected to four or six trials per run, while the total number of trials across all runs remained constant (i.e., 20 trials per labeled odorant). Four runs were conducted for each subject. The order of the stimuli was semi‐randomized across subjects. To detect the onset and duration of inhalation, respiratory cycles were recorded using a breathing sensor with a pressure sensor (ASK‐BWT01; Asuka Electric Co., Ltd.) via cannulas (OX‐01L; NISSEI ECO Co., Ltd., Kanagawa, Japan) attached to the nose during fMRI scanning. To remove physiological noise from the BOLD signals, cardiac cycles were recorded using a pulse oximeter (Siemens Healthcare, Erlangen, Germany) attached to the left ring finger.

2.4. Procedure for dissimilarity ratings

Dissimilarity ratings were collected after fMRI scanning. Among 25 subjects, 5 subjects did not participate in the dissimilarity rating session owing to time constraints. We used the same eight labeled odorants as in the fMRI scanning. The subjects were asked to sit on a chair outside the MRI scanner. We presented the first labeled odorant for 2.5 s after a 2 s count‐down and the second labeled odorant for 2.5 s after a 5.5 s interval and a 2 s count‐down (Figure 1b). Then the olfactory dissimilarity between the first and second labeled odorant was rated using a Likert scale (0–10; anchor labels were provided for the endpoints as ‘similar’ for 0 and ‘dissimilar’ for 10). We asked subjects to rate ‘0’ if they thought they could not distinguish the odors. The anchor ‘10’ was explained as the degree of dissimilarity equivalent to that between most dissimilar odors that they encountered during the fMRI scanning. The intervals between the offsets of the second labeled odorant and the onsets of the first labeled odorant in the next trial were 7.5 s. Dissimilarity ratings were obtained for stimulus pairs within each category (Figure 2a). There were six pairs of labeled odorants with different labels or odorants in each category. Each was presented twice: once in forward order and once in reverse order. In addition, four pairs of identical odorants with identical labels were presented once for each category. Thus, a total of 32 trials were conducted. Subjects were asked not to rate when they did not detect either the first‐ or second‐labeled odorants. The number of subjects who did not provide ratings for each pair is shown in Figure S5.

FIGURE 2.

FIGURE 2

Open in a new tab

The effects of word labels on perceived dissimilarity among labeled odorants. (a) Grayscale indicates pair‐wise dissimilarity ratings averaged across subjects. Red, green, blue, and yellow dots indicate conditions as shown in (b). Labeled odorants presented firstly and secondly are indicated on horizontal and vertical axes, respectively. The order of labeled odorants is consistent for both axes. (b) Dissimilarity ratings averaged within each of four conditions. The colors correspond to those described in (a). Error bars indicate the standard error (SE). Black lines indicate the compared pairs, and red asterisks indicate significance (Student's t test, one‐tailed, FDR <0.05). Cit, citral; Euc, eucalyptol; eucal, eucalyptus; Lim, limonene; Men, menthol.

2.5. Procedure for congruency/pleasantness/familiarity ratings

Congruency/pleasantness/familiarity ratings were collected after fMRI scanning and dissimilarity ratings. We used the same eight labeled odorants as in the fMRI scanning. Subjects were asked to sit on a chair outside the MRI scanner. We presented the labeled odorants for 2.5 s after 2 s of count‐down. Then, the congruency between labels and odorants and pleasantness and familiarity of the labeled odorants were rated using a 4‐point Likert scale (anchor labels were provided for the endpoints only). The scales ranged from 0 to 4 labeled as ‘incongruent’ and ‘congruent,’ from −2 to 2 labeled as ‘unpleasant’ and ‘pleasant’ from 0 to 4 labeled as ‘unfamiliar’ and ‘familiar’ respectively (Figure 1c). The ISIs were 7.5 s. Two trials were performed for each labeled odorant.

2.6. MRI acquisition

All MRI data were acquired using a 7T whole‐body MRI scanner (MAGNETOM 7T; Siemens Healthcare, Erlangen, Germany) with a single‐channel transmit coil and a 32‐channel receiver head coil (Nova Medical, Wilmington, MA, USA).

To measure the BOLD signals in the piriform cortex, functional images were acquired using a multi‐band gradient‐echo echo‐planar imaging sequence (Moeller et al., 2010) with the following parameters: repetition time, 2.5 s; echo time, 21 ms; flip angle, 70°; iPAT, 4; multi‐band factor, 1; field‐of‐view, 204 × 204 mm2; matrix size, 204 × 204; slice thickness, 1‐mm with no gap; number of slices, 30 oblique slices; phase encoding direction, anterior to posterior. To reduce signal loss around the piriform cortex caused by intra‐voxel dephasing effects resulting from local magnetic field inhomogeneities, known as susceptibility noise, the axial slices were tilted 25° from the anterior commissure–posterior commissure line (Deichmann et al., 2003). A total of 212 volumes were acquired in each run. To correct for distortion of the functional images due to susceptibility noise, functional images with the same parameters but a reversed phase‐encoding direction (posterior to anterior) were acquired for five volumes.

T1‐weighted whole‐brain structural images were acquired using the magnetization‐prepared rapid acquisition gradient‐echoes (MP2RAGE) sequence, which was provided by the manufacturer (Siemens Healthcare, Erlangen, Germany) as a work‐in‐progress package, with the following parameters: repetition time, 5000 ms; echo time, 3.4 ms; flip angle, 4°/5°; inversion recovery time, 800/2600 ms; field‐of‐view, 209 × 209 mm2; matrix size, 300 × 300; slice thickness, 0.7 mm with no gap; number of slices, 280 sagittal slices. To coregister narrow‐range functional images to the T1‐weighted whole brain structural image more precisely, narrow‐range T2‐weighted structural images were acquired using turbo spin echo sequence with the following parameters: repetition time, 5.65 s; echo time, 44 ms; flip angle, 120°; iPAT, 2; field‐of‐view, 204 × 204 mm2; matrix size, 320 × 320; slice thickness, 2 mm with no gap; number of slices, 30 oblique slices; phase encoding direction, anterior to posterior.

2.7. fMRI data preprocessing

For each subject, the fMRI data was preprocessed using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/), ANTs 2.1.0 (http://stnava.github.io/ANTs/), FSL 6.0.5.1 (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki), and ITK‐SNAP 3.6.0 (http://www.itksnap.org/pmwiki/pmwiki.php). The first four functional volumes in each run were excluded from the main analysis. These volumes were only used for susceptibility distortion correction. Second, the slice timings of the functional volumes were corrected, and all functional volumes were motion‐corrected by referring to the first volume of the first run using SPM12. Third, susceptibility distortions of functional volumes were corrected using the FSL “topup” command. Specifically, the transformation matrix estimated using the discarded four volumes was applied all other volumes. Fourth, the T2‐weighted structural image was co‐registered to the mean functional volume, and the T1‐weighted structural image was then co‐registered to the T2‐weighted structural image. Fifth, T1‐weighted structural images were brain‐extracted using ANTs, veins around the piriform cortex were manually masked using ITK‐SNAP, and the structural image was transformed into the MNI space using ANTs. The transformation matrix was applied to all the functional volumes. Finally, for the second level GLM and functional connectivity analysis, functional volumes were smoothed with a 3 mm full width at half maximum (FWHM) Gaussian filter.

To assess head motion, we calculated framewise displacement (FD) based on motion parameters using the PhysIO toolbox (https://www.tnu.ethz.ch/en/software/tapas/documentations/physio-toolbox). It is known that FD would be underestimated when motion correction is conducted after slice timing correction (Power et al., 2017); therefore, we calculated the FDs using raw functional volumes. We rejected five subjects who had at least one run in which FDs exceeded 0.5 mm for more than 5% of the volumes (out of 208 volumes per run).

After preprocessing, all analyses including decoding analysis were performed in the MNI space.

2.8. First level GLM

To estimate brain activation elicited by each stimulus, we performed general linear modeling using SPM12. Two types of response variables were regressed: BOLD signals with smoothing for the second level GLM and beta series correlation, and BOLD signals without smoothing for decoding analyses. Run‐wise models were constructed for the second level GLM and group decoding analysis, where all regressors for all runs were included in a single model for each subject. Since we presented nine stimuli (eight labeled odorants, one odorless) per run, the number of regressors of interest was 36 (9 stimuli multiplied by 4 runs). They were coded as blocks and convolved with a canonical hemodynamic response function, and were included for each subject. Blocks were defined using the inhalation onset and duration within the stimulus presentation period. Only valid trials, where the responses to the task were any one of ‘weak’, ‘moderate’ or ‘strong’ for labeled odorants and ‘no smell’ for odorless air were included. Six subjects whose minimum number of valid trials per labeled odorant per run was less than half of the total trials for any run of any labeled odorant were excluded from the analysis. Six motion parameters, six cardiac phase parameters, one heart rate parameter, and four binary parameters indicating each run were used as nuisance regressors. Motion parameters were estimated from functional images using SPM12, and the cardiac phase and heart rate parameters were estimated from cardiac cycles using the PhysIO toolbox. For the subject‐wise decoding and beta series correlation analyses, trial‐wise models were constructed, where all regressors for all runs were included in a single model for each subject. The trial‐wise models were constructed in the same manner as run‐wise models, except that all trials were included as separate regressors, with the number of regressors of interest equal to the number of valid trials (Rissman et al., 2004).

2.9. Second level GLM

To identify the brain regions activated by odors, we compared the regression coefficients obtained from the first level GLM constructed based on the smoothed functional images for the labeled odorants with those for the odorless air labeled as odorless using Student's t‐test (one‐tailed). To determine the significance, a suprathreshold cluster test was conducted using a cluster‐based randomization approach with a maximal statistic accounting for the multiplicity of voxels (Nichols & Holmes, 2002). The cluster‐forming threshold was set at p < .001, and the alpha level for cluster size was set at 0.05. The null distribution of the maximum cluster size was estimated based on 50,000 permutations.

2.10. Decoding analysis

To investigate the differences in multi‐voxel activity patterns in the piriform cortex, we performed searchlight decoding analysis to classify the odor category or each pair of labeled odorants. In the decoding models, the explanatory variables or features, were the t‐values of the regression coefficients of the first level GLM constructed based on functional volumes without smoothing. We used t‐values instead of regression coefficients because the former would suppress the contribution of noisy voxels (Misaki et al., 2010) and mitigate differences in variance across conditions (Figure S6), which would be problematic in the decoding analysis (Görgen et al., 2018). The center voxels of the searchlight analysis were restricted to the piriform cortex, which was manually defined in the T1‐weighted images averaged across subjects in the MNI space based on the brain atlas (Mai, 2015). To reduce the smoothing effects of searchlight analysis, a relatively small number of voxels within 2 mm of the center voxels were analyzed. This resulted in 32 voxels, representing the number of features. In addition, a bias term was included in all models. The voxels outside the piriform cortex were included to maintain a constant number of features. The classification models were constructed using a support vector machine with a linear kernel using LIBLINEAR (https://www.csie.ntu.edu.tw/~cjlin/liblinear/). The optimal regularization parameters were selected from 29 candidates (10−7 to 107, log‐linearly spaced) based on the model performance on the training data using cross‐validation. Cross‐validations were conducted using a three‐fold (leave‐one‐run‐out) approach for subject‐wise models. For group models, we employed an eight‐fold approach, wherein data from 24 subjects were evenly divided into eight subsets, each containing data from 3 subjects.

Model accuracy was evaluated using leave‐one‐run‐out and leave‐one‐subject‐out cross‐validations for the subject‐wise and group models, respectively. In the subject‐wise models, the number of training and validation data varied across the subjects because we excluded invalid trials. For label or odorant classification, the maximum numbers were 30 and 10 (mean ± standard deviation, 29.2 ± 1.5 and 9.7 ± 0.7), respectively. For category classification, they were 120 and 40 (mean ± standard deviation, 116.3 ± 5.2 and 38.8 ± 2.0), respectively. In the group models, for label or odorant classification, there were 192 training data (24 subjects, four runs, and two labeled odorants) and eight validation data (one subject, four runs, and two labeled odorants). For category classification, 768 training data (24 subjects, four runs, and eight labeled odorants) and 32 validation data (one subject, four runs, and eight labeled odorants) were used.

For all models, the obtained accuracy maps were subtracted by chance (50%) and then smoothed using a 3 mm FWHM Gaussian filter. p‐Values for decoding accuracies against chance levels were calculated using a sign permutation test (one‐tailed). To determine significance, suprathreshold cluster tests were conducted using a cluster‐based randomization approach with a maximal statistic that accounted for the multiplicity of voxels (Nichols & Holmes, 2002). The cluster‐forming threshold was set at p < .001. When there was a multiplicity of pairs, the alpha level for cluster size was set at 0.05 divided by the number of pairs subject to the analysis. The null distribution of the maximum cluster size was estimated based on 50,000 permutations.

2.11. Beta series correlation analysis

To examine the functional connectivity between the activations in parts of the piriform cortex modulated by labels or odorants and those in the rest of the brain, we conducted a beta series correlation analysis. Each cluster that was significant in the decoding analysis was used as a seed region. Trial‐by‐trial regression coefficients (betas) were obtained as described in Section 2.8 and were used to construct the beta series. To avoid the possible influence of unrelated conditions, we only included betas from relevant trials when constructing the beta series. For example, when clusters detected in mint‐Men versus eucal‐Men were used as seed regions (red clusters in Figure 3), betas estimated for menthol labeled as mint or eucalyptus were used (Figure 4). Similarly, when clusters detected in lemon‐Cit versus lemon‐Lim were used as seed regions (blue cluster in Figure 3), betas estimated for citral and limonene labeled as lemon were used. The betas estimated for invalid trials (see Section 2.8) were not included. A beta series was constructed for each voxel by concatenating the beta values from relevant trials. The seed regions were averaged across voxels within each seed region to obtain the beta series per seed. Finally, Pearson's correlation coefficients between the beta series for each seed and each target voxel were computed and Fisher's z‐transformed. To test whether the beta series correlations were larger than 0, we performed a Student's t‐test (one‐tailed). To test whether the beta series correlations with parts of the piriform cortex modulated by word labels differed from those with a part modulated by olfactory inputs, comparisons were made for beta series correlations that used a cluster in the left temporal region modulated by odorants (blue in Figure 3) as a seed region against those that used the following seed regions: a cluster in the left junction modulated by labels (red in Figure 3) and two clusters in the right junction modulated by labels (red and green in Figure 3). Student's t‐tests (two‐tailed) were used for these comparisons. Significance was determined in the same manner as Section 2.9, except that clusters containing seed voxels were excluded when constructing null distributions for the suprathreshold cluster test. The multiplicity of seeds was corrected by setting the alpha level for the cluster size to 0.05 divided by the number of seeds.

FIGURE 3.

FIGURE 3

Open in a new tab

The effects of word labels and odorants on multi‐voxel activity patterns in the piriform cortex. Clusters modulated by word labels (red and green) and odorants (blue) are shown in the three‐dimensional rendering of the piriform cortex (left top panel) and coronal slices (right top panel). Red, green, and blue indicate z‐values for decoding accuracies against chance levels for mint‐Men versus eucal‐Men, mint‐Euc versus eucal‐Euc, and lemon‐Cit versus lemon‐Lim, respectively. The z‐values were calculated using the means and standard deviations of the null distributions of the decoding accuracies. Only voxels that belonged to significant clusters (sign permutation test, one‐tailed, p < .0125, cluster‐size corrected) were colored. Significant clusters were identified solely based on group models, with no significant clusters detected in subject‐wise models. In the middle panel, the piriform cortex is encircled with white lines in each coronal slice. In the right bottom panel, the slice locations are indicated by blue lines. L, left; P, posterior; S, superior.

FIGURE 4.

FIGURE 4

Open in a new tab

Functional connectivity with the clusters in the piriform cortex modulated by word labels. The significance map of the t‐values for the beta series correlation between the two red and one green clusters shown in Figure 3 and the rest of the brain within the range of functional images was overlaid on the structural image (Student's t‐test, one‐tailed, p < .0167, cluster‐size corrected). The yellow arrowheads indicate clusters located in (a) the bilateral temporal pole, (b) bilateral hippocampus, (c) left inferior and middle frontal gyri, and (d) bilateral orbitofrontal cortex. Red, green, and blue indicate the t‐values when the seed is a cluster in the left junction part (red cluster in Figure 3), right junction part (red cluster in Figure 3), or right junction part (green cluster in Figure 3) of the piriform cortex, respectively. The locations of each slice are indicated by blue lines.

2.12. Statistical tests

All statistical tests were conducted using MATLAB functions (Statistics and Machine Learning Toolbox, R2021b). Significance levels were set at p < .05 unless otherwise specified.

3. RESULTS

3.1. Modulation of olfactory perception by word labels

Four odors (two minty and two citrus odors) were used in this study. In each of minty and citrus odor category, there were 2 × 2 combinations of word labels and odorants (e.g., mint‐Men; Figure 1a). These stimuli were selected based on a pilot study so that all odorants were perceived as congruent with their labels. To assess whether word labels modulated olfactory perception, after fMRI scanning, we obtained dissimilarity ratings for all pairs of labeled odorants within each category by sequentially presenting the odorants with word labels, and asking the subjects to rate the dissimilarity with the perceived odors (Figure 1bii). Owing to time constraints, we were able to obtain dissimilarity ratings from 20 of 25 subjects (see Section 2). The rating scores averaged across subjects for each pair are shown in Figure 2a, and those averaged within the experimental conditions are shown in Figure 2b. When compared to pairs of identical odorants with identical labels (red), the rated dissimilarity for pairs with identical odorants presented with different labels (green) was significantly larger (N = 20, Student's t‐test, two‐tailed, p = 3.83 × 10−5, FDR corrected; Figure 2b). This suggests that identical odorants were perceived differently when presented with different labels than when they were presented with identical labels. In addition, rated dissimilarities for pairs with different odorants presented with different labels (yellow) were larger than those for pairs with different odorants presented with identical labels (blue; N = 20, Student's t‐test, two‐tailed, p = 6.70 × 10−6, FDR corrected; Figure 2b). In summary, odor perception was modulated by the labels used in this study such that odors presented with different labels were perceived differently, even when the odorants were identical.

3.2. Validating current fMRI condition by replicating previous studies

We measured blood oxygenation level‐dependent (BOLD) signals at 1‐mm isotropic resolution using 7T MRI while subjects were presented with labeled odorants and asked to rate the odor intensity (Figure 1bi). Before examining the effects of semantic context and olfactory inputs, we conducted two analyses to validate the appropriateness of our experimental conditions for evaluating piriform cortex activity. First, we confirmed the responses to odors in the piriform cortex and compared the brain responses to labeled odorants versus odorless air labeled as odorless by constructing a general linear model (GLM). To correct for the multiplicity of voxels, cluster‐level inferences were made using a randomization approach with maximal statistic (Nichols & Holmes, 2002). The responses to labeled odorants in the piriform cortex in both hemispheres were significantly greater than those to odorless air labeled as odorless (N = 25, Student's t‐test, one‐tailed, p < .05, cluster‐size corrected; Figure S1 and Table S1). This indicates that responses to odors in the piriform cortex could be obtained under our experimental conditions.

Next, we examined whether it is possible to classify citrus and minty odor categories based on multi‐voxel activity patterns in the piriform cortex, as has been shown in previous studies (Bao et al., 2016; Bhutani et al., 2019; Howard et al., 2009; Qu et al., 2016). In this analysis, differences in sniff volumes between odor categories could be a confounding factor, as sniffing itself could induce the activation of the piriform cortex (Zelano et al., 2005). Additionally, the levels of congruency between odorants and labels, which has been suggested to influence olfactory processing (Lundström et al., 2019; Porada et al., 2019; Sijben et al., 2018), may be another confounding factor. Following the previous studies that we aim to replicate, we evaluated the confounding factors using t‐tests. The results indicated that these factors did not differ significantly between categories (sniff volume, p = .15; congruency, p = .14). We evaluated the decoding accuracies using leave‐one‐run‐out and leave‐one‐subject‐out cross‐validation in the Montreal Neurological Institute (MNI) space for the subject‐wise and group models, respectively (see Section 2 for details). Significance against chance level was determined using a sign permutation test, and the multiplicity of voxels was corrected based on cluster‐level inference using a randomization approach with the maximum statistic (Nichols & Holmes, 2002). Significant clusters were found in both the subject‐wise (shown in red) and group (shown in blue) models (N = 25, sign permutation test, one‐tailed, p < .05, cluster‐size corrected; Figures S2, S3a; Table S2). In the subject‐wise model, clusters were found in the left temporal and right frontal parts, whereas in the group model, clusters were found in the bilateral junctions of the frontal and temporal parts and the left temporal part. Subject‐wise and group models have been shown to yield distinct significant regions (Wang, Cagna, et al., 2020), as the former can detect activity patterns unique to each subject, whereas the latter can offer higher sensitivity owing to its reliance on a larger dataset formed by pooling data across subjects. The differences in the cluster locations in the current analysis likely reflected these differences. These results suggest that the classification of odor categories based on multi‐voxel activity patterns in the piriform cortex is possible under our experimental conditions using both subject‐wise and group models.

3.3. Modulation of multi‐voxel activity patterns in the piriform cortex by word labels

Following the validation of our experimental conditions by replicating previous studies, we investigated whether multi‐voxel patterns in the piriform cortex would differ in response to identical odorants presented with different labels, reflecting the influence of the semantic context on odor perception. Prior to this analysis, we thoroughly examined the potential confounding factors for each odorant, specifically the sniff volume and the congruency level between words and odorants, for each odorant. In addition to using a t‐test, we computed Bayes factors (BFs), which quantify the evidence levels for the null (BF01) and alternative (BF10) hypotheses. Following conventional criteria, we considered BF >3 as an indication of “positive” evidence (Kass & Raftery, 1995). We found positive evidence supporting the consistency of the sniff volumes between pairs of labels for each odorant (BF01 > 3; Figure S4c and Table S6). Conversely, positive evidence supporting the consistency of congruency was found for two odorants, whereas positive evidence of a difference in congruency was observed for the other two (BF10 > 3; Figure S4c; Table S6). Given the potential influence of these confounding factors on the decoding results for these pairs, we assessed the decoding accuracy for each odorant separately in the subsequent analysis rather than averaging the results across all odorants. The effect of the labels on dissimilarity ratings was significant for all odorants (Figure S4a).

The decoding accuracies of the subject‐wise and group models were evaluated in the same manner as the category decoding. Because we examined the four odorants separately, we set the alpha level for the cluster size to 0.0125 to account for its multiplicity. In the subject‐wise model, no significant cluster was found for any pair (N = 25, sign permutation test, one‐tailed, p < .0125, cluster‐size corrected). In the group model, significant clusters were found for menthol (mint‐Men vs. eucal‐Men; shown in red) and eucalyptol (mint‐Euc vs. eucal‐Euc; shown in green) (N = 25, sign permutation test, one‐tailed, p < .0125, cluster‐size corrected; Figures 3, S3b; Table S3a). These clusters were situated at the junction of the frontal and temporal parts in both hemispheres for menthol and in the right hemisphere for eucalyptol, with partial overlapping in the right hemisphere. No significant clusters were observed for the remaining odorants. These results demonstrate that semantic context can differentiate multi‐voxel patterns in the piriform cortex evoked by identical olfactory inputs.

To investigate why significant decoding was achievable for certain pairs but not for others, we qualitatively examined the perceptual ratings (dissimilarity, intensity, pleasantness, and familiarity) as well as potential confounding factors (Figure S4; Table S6). However, we could not identify any shared factors that could account for the differentiation between the decodable and non‐decodable odorant pairs.

3.4. Modulation of multi‐voxel activity patterns in the piriform cortex by olfactory inputs

To investigate whether distinct regions of the piriform cortex are modulated by changes in olfactory perception arising from different olfactory inputs accompanied by identical labels, we analyzed multi‐voxel patterns within the piriform cortex. A summary of the perceptual ratings and potential confounding factors for each label is shown in Figure S4 and Table S6. The decoding accuracies were evaluated using the same method used to decode the word labels. In the subject‐wise model, no significant cluster was found for any pair (N = 25, sign permutation test, one‐tailed, p < .0125, cluster‐size corrected). In the group model, we found a significant cluster in the left temporal part for the lemon label (lemon‐Cit vs. lemon‐Lim) (N = 25, sign permutation test, one‐tailed, p < .0125, cluster‐size corrected; shown in blue in Figures 3 and S3b; Table S3b). No significant clusters were observed for the remaining labels. Among the perceptual ratings and potential confounding factors that we assessed, none could distinguish between the decodable and non‐decodable odorant pairs (Figure S4; Table S6).

The location of the cluster modulated by olfactory inputs differed from those modulated by labels (Figure 3). Considering the partial overlap of clusters influenced by the labels, identifying a cluster influenced by olfactory inputs at a distinct location is interesting. However, due to the limited number of decodable pairs, we could not draw a conclusion regarding whether the differences in olfactory inputs and word labels were represented in different subregions of the piriform cortex.

3.5. Functional connectivity with semantic and memory areas

Finally, to investigate whether parts of the piriform cortex modulated by word labels communicate with regions involved in semantic and memory processing, we examined the functional connectivity between each of the three clusters modulated by word labels (two red and one green clusters in Figure 3) and the rest of the brain within the range of functional images using a beta series correlation analysis (Rissman et al., 2004). To correct for the multiplicity of voxels, cluster‐level inferences were made using a randomization approach with maximal statistic (Nichols & Holmes, 2002). As we examined three seeds, we set the alpha level for the cluster size to 0.0167 to account for its multiplicity. All regions where significant functional connectivity were observed are summarized in Table S4 (N = 25, Student's t‐test, one‐tailed, p < .0167, cluster‐size corrected). Note that the clusters including the seed voxels were excluded from the statistical test (see Section 2). Therefore, in some regions close to the seeds, significance was only reported in the hemisphere opposite to that containing the seeds. It is well known that the temporal pole serves as a hub for semantic processing (Olson et al., 2007; Skipper et al., 2011), the hippocampus is related to memory (Duff et al., 2019), and the left inferior/middle frontal gyri processes language information (Friederici, 2011). In addition, the orbitofrontal cortex has been shown to be involved in semantic processing especially in olfaction (Olofsson & Gottfried, 2015). These regions showed a significant functional connectivity with at least two seeds (Figure 4, arrowheads), supporting our hypothesis that parts of the piriform cortex modulated by word labels can communicate with regions associated with semantic and memory processing.

One possible reason why some parts of the piriform cortex were modulated by word labels and others by odorants is that the parts modulated by word labels communicate with regions that process verbal information, whereas those modulated by olfactory input do not. To test this hypothesis, we examined the functional connectivity between the one cluster modulated by olfactory inputs (blue cluster in Figure 3) and the rest of the brain within the range of functional images. The seed region showed significant functional connectivity with the bilateral temporal pole, right hippocampus, and right orbitofrontal cortex, but not with the left inferior/middle frontal lobe (N = 25, Student's t‐test, one‐tailed, p < .05, cluster‐size corrected; Figure 5, arrowheads; Table S5). To examine functional connectivity specific to label effects, we compared the strength of the functional connectivity between clusters modulated by word labels and clusters modulated by olfactory inputs, but we did not find any significant differences (N = 25, Student's t‐test, two‐tailed, p < .0167, cluster‐size corrected; see Section 2). Therefore, these results do not support the hypothesis that the parts of the piriform cortex modulated by word labels have stronger functional connectivity with regions related to semantic and memory processing.

FIGURE 5.

FIGURE 5

Open in a new tab

Functional connectivity with the cluster in the piriform cortex modulated by odorants. The significance map of t‐values for the beta series correlation between the blue cluster shown in Figure 3 and the rest of the brain within the range of functional images was overlaid on the structural image (Student's t‐test, one‐tailed, p < .05, cluster‐size corrected). The yellow arrowheads indicate clusters located in (a) the bilateral temporal pole, (b) right hippocampus, and (d) right orbitofrontal cortex. In (c) the left inferior/middle frontal lobe, no clusters corresponding to Figure 4c were found. The corresponding locations are indicated by a yellow circle. Red indicates t‐values. The locations of each slice are indicated by blue lines.

4. DISCUSSION

This study demonstrated that identical odorants were perceived differently when presented with different word labels, and this perceptual shift was accompanied by changes in multi‐voxel activity patterns in the piriform cortex in some cases. We also found differences in multi‐voxel activity patterns in the piriform cortex between different olfactory inputs. Additionally, we observed significant functional connectivity between these modulated piriform areas, which were modulated by words or olfactory inputs, and semantic regions beyond the piriform cortex.

The main objective of this study was to determine whether the modulation of multi‐voxel activity patterns in the piriform cortex underlies changes in olfactory perception caused by semantic context. Previous human fMRI studies have suggested that olfactory perception is coded in the piriform cortex (Fournel et al., 2016; Gottfried et al., 2006; Howard et al., 2009; Zelano et al., 2011) and that activity is induced or enhanced by the presence of non‐olfactory semantic information (Arshamian et al., 2013; González et al., 2006; Karunanayaka et al., 2015; Porada et al., 2019). However, while differential labeling of odors can alter odor perception, corresponding changes in brain activity have only been observed in the secondary and/or downstream olfactory areas (Bensafi et al., 2014; de Araujo et al., 2005; Gottfried & Dolan, 2003). Our study revealed that differences between word labels were also present in multi‐voxel activity patterns in the piriform cortex, at least in some cases. Previously, the effects of other contexts, such as experience or satiety, were reported to modulate multi‐voxel activity patterns in the piriform cortex by comparing patterns before and after associative learning or eating (Howard & Gottfried, 2014; Li et al., 2008; Qu et al., 2016; You et al., 2022). In contrast, our study showed that the multi‐voxel patterns evoked by the same olfactory inputs changed trial‐by‐trial, suggesting that odor representations in the piriform cortex could change rapidly and flexibly based on the semantic context. Compared to previous fMRI studies examining the effects of semantic labels (Bensafi et al., 2014; de Araujo et al., 2005; Gottfried & Dolan, 2003), our study had two methodological advantages. First, we conducted a multivariate decoding analysis, which has higher detection power than univariate analysis. Second, the 1‐mm isotropic spatial resolution was over 27 times higher than that in previous studies. This should have allowed for more precise localization of piriform cortex activity, reduced partial volume effects, and improved the ability to detect subtle differences in the piriform activity that reflect semantic context. Nevertheless, our approach could detect only the effects of word labels on multi‐voxel activity patterns in two of four odorants. We could not conclude whether there are no effects in the remaining odorants or whether there are effects which we could not detect in the current study. To address this, it may be necessary to construct more powerful decoding models by increasing the number of subjects or trials, or by obtaining fMRI data with higher signal‐to‐noise ratios.

Areas in the piriform cortex modulated by word labels showed significant functional connectivity with various brain regions, including the temporal pole, hippocampus, left inferior/middle frontal gyri, and orbitofrontal cortex. These regions have been implicated in semantic and memory processing in olfaction and other modalities (Duff et al., 2019; Friederici, 2011; Olofsson & Gottfried, 2015; Olson et al., 2007; Skipper et al., 2011). Functional connectivity between the temporal part of the piriform cortex and the aforementioned regions has also been demonstrated using resting‐state fMRI in previous research (Zhou et al., 2019). While functional connectivity does not indicate structural connections, studies in rodents and monkeys have shown that all these regions, except for the left inferior/middle frontal gyri, project axons to the piriform cortex (Johnson et al., 2000; Mohedano‐Moriano et al., 2005; Morán et al., 1987; Wang, Zhang, et al., 2020). Therefore, similar structural connections may exist among humans. The piriform cortex may receive semantic information from these semantic and memory‐related areas, which may in turn modulate odor representation in the piriform cortex.

Another objective of this study was to investigate whether the areas modulated by word labels differed from those modulated by olfactory inputs within the piriform cortex. Our results showed differences in multi‐voxel activity patterns between pairs with word differences and a pair with differences in olfactory inputs, and the locations of the modulated areas varied between them. However, we could detect such modulations only for limited pairs, making it difficult to conclude whether the observed locational differences were solely based on whether the cluster was modulated by odorants or labels. Notably, there were no significant differences in functional connectivity between conditions. Because both conditions involved combinations of words and odors, and perceptual differences were present in both conditions, it is possible that the modulated areas represented perception arising from integrated word‐odor information. The observed locational differences may be attributed to variations in nuanced perceptual characteristics across the decoding pairs. Although the current study did not provide evidence of regional differences within the piriform cortex between top‐down and bottom‐up processing, considering previous monkey studies, there are likely laminar differences in neuronal connections and regional variations in layer thickness within the piriform cortex (Carmichael et al., 1994; Mohedano‐Moriano et al., 2005; Turner et al., 1978). Future studies using advanced imaging and analytical techniques may reveal regional or laminar differences between top‐down and bottom‐up processing in the human piriform cortex.

The fact that the group models hold indicates that at a 1‐mm isotropic resolution, multi‐voxel activity patterns in the piriform cortex representing odor information are partially common across subjects. This finding may seem surprising, considering that rodent studies have shown the absence of a common organization in the projections from the olfactory bulb to the piriform cortex among subjects (Miyamichi et al., 2011). A previous fMRI study reported a lack of topographical consistency in multi‐voxel activity patterns for odor categories across human subjects within the posterior piriform cortex (Howard et al., 2009). However, no quantitative analysis of the commonality of neural patterns across subjects was conducted. As group models do not require complete consistency of multi‐voxel activity patterns across subjects, our results do not contradict those of the previous study. In our study, more significant clusters were observed in the group model compared to the subject‐wise model when decoding odor categories, and significant clusters were observed only in the group model when decoding labels and odorants. This could be due to the higher detection power of the group model, which benefits from a larger training dataset size (approximately N times larger than the subject‐wise model, where N indicates the number of subjects) (Wang, Cagna, et al., 2020). The usefulness of group models in olfactory studies should be further explored in future research.

In conclusion, by leveraging ultra‐high field fMRI, this study provides evidence that multi‐voxel activity patterns in the piriform cortex representing odors exhibit flexible changes based on word labels associated with the odors. The influence of semantic context on piriform cortex activity might be mediated by communication between the piriform cortex and external regions involved in semantic and memory processing, as evidenced by the observed functional connectivity. These findings advance our understanding of how semantic information shapes odor perception as well as the underlying neural mechanisms. Further investigations are necessary to understand the potential regional differences within the piriform cortex in the processing of top‐down and bottom‐up information.

AUTHOR CONTRIBUTIONS

T. O., I. K., A. Y., T. N., S. N., K. T., and M. O. conceived and designed the study. T. O., I. K., and A. Y. performed the experiments. T. O., I. K., A. Y., T. N., S. N., K. T., and M. O. analyzed and interpreted data. T. O. and M. O. drafted the manuscript with critical inputs from all authors.

FUNDING INFORMATION

This work was supported by a Grant‐in‐Aid for JSPS Research Fellow to T. O. (Grant Number JP21 J13615), JST ERATO to S. N. (Grant Number JPMJER1801), JSPS KAKENHI to S. N. (Grant Number JP18H05522), JST Mirai Program to K. T. (Grant Number JPMJMI19D1), and JSPS KAKENHI to M. O. (Grant Numbers JP18H04998 and JP21H05808).

CONFLICT OF INTEREST STATEMENT

The authors declare no competing interests.

PATIENT CONSENT STATEMENT

All subjects provided written informed consent after receiving a complete explanation of the study.

Supporting information

APPENDIX S1: Supporting information.

HBM-45-e26681-s001.docx (3.7MB, docx)

ACKNOWLEDGMENTS

We thank Mses. Rumi Iwasaki, Akiko Yamamoto, and Rumiko Miyatake for their help in conducting the experiments, and Dr. Tobias Kober for developing the MP2RAGE sequence and providing technical support. This work was supported by a Grant‐in‐Aid for JSPS Research Fellow to T. O. (Grant Number JP21 J13615), JST ERATO to S. N. (Grant Number JPMJER1801), JSPS KAKENHI to S. N. (Grant Number JP18H05522), JST Mirai Program to K. T. (Grant Number JPMJMI19D1), and JSPS KAKENHI to M. O. (Grant Numbers JP18H04998 and JP21H05808).

Okumura, T. , Kida, I. , Yokoi, A. , Nakai, T. , Nishimoto, S. , Touhara, K. , & Okamoto, M. (2024). Semantic context‐dependent neural representations of odors in the human piriform cortex revealed by 7T MRI . Human Brain Mapping, 45(6), e26681. 10.1002/hbm.26681

Contributor Information

Kazushige Touhara, Email: ktouhara@g.ecc.u-tokyo.ac.jp.

Masako Okamoto, Email: masakookamoto3@gmail.com.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in Open Science Framework at https://osf.io.

REFERENCES

  1. Arnold, T. C. , You, Y. , Ding, M. , Zuo, X. N. , de Araujo, I. , & Li, W. (2020). Functional connectome analyses reveal the human olfactory network organization. eNeuro, 7(4), ENEURO.0551‐19.2020. 10.1523/eneuro.0551-19.2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Arshamian, A. , Iannilli, E. , Gerber, J. C. , Willander, J. , Persson, J. , Seo, H. S. , Hummel, T. , & Larsson, M. (2013). The functional neuroanatomy of odor evoked autobiographical memories cued by odors and words. Neuropsychologia, 51(1), 123–131. 10.1016/j.neuropsychologia.2012.10.023 [DOI] [PubMed] [Google Scholar]
  3. Ayabe‐Kanamura, S. , Kikuchi, T. , & Saito, S. (1997). Effect of verbal cues on recognition memory and pleasantness evaluation of unfamiliar odors. Perceptual and Motor Skills, 85(1), 275–285. 10.2466/pms.1997.85.1.275 [DOI] [PubMed] [Google Scholar]
  4. Bae, J. , Yi, J. Y. , & Moon, C. (2019). Odor quality profile is partially influenced by verbal cues. PLoS One, 14(12), e0226385. 10.1371/journal.pone.0226385 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bao, X. , Raguet, L. L. , Cole, S. M. , Howard, J. D. , & Gottfried, J. (2016). The role of piriform associative connections in odor categorization. eLife, 5, e13732. 10.7554/eLife.13732 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bensafi, M. , Croy, I. , Phillips, N. , Rouby, C. , Sezille, C. , Gerber, J. , Small, D. M. , & Hummel, T. (2014). The effect of verbal context on olfactory neural responses. Human Brain Mapping, 35(3), 810–818. 10.1002/hbm.22215 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bensafi, M. , Sobel, N. , & Khan, R. M. (2007). Hedonic‐specific activity in piriform cortex during odor imagery mimics that during odor perception. Journal of Neurophysiology, 98(6), 3254–3262. 10.1152/jn.00349.2007 [DOI] [PubMed] [Google Scholar]
  8. Bhutani, S. , Howard, J. D. , Reynolds, R. , Zee, P. C. , Gottfried, J. , & Kahnt, T. (2019). Olfactory connectivity mediates sleep‐dependent food choices in humans. eLife, 8, e49053. 10.7554/eLife.49053 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Carmichael, S. T. , Clugnet, M. C. , & Price, J. L. (1994). Central olfactory connections in the macaque monkey. The Journal of Comparative Neurology, 346(3), 403–434. 10.1002/cne.903460306 [DOI] [PubMed] [Google Scholar]
  10. Cornell Kärnekull, S. , Gerdfeldter, B. , Larsson, M. , & Arshamian, A. (2021). Verbally induced olfactory illusions are not caused by visual processing: Evidence from early and late blindness. Iperception, 12(3), 20416695211016483. 10.1177/20416695211016483 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. de Araujo, I. E. , Rolls, E. T. , Velazco, M. I. , Margot, C. , & Cayeux, I. (2005). Cognitive modulation of olfactory processing. Neuron, 46(4), 671–679. 10.1016/j.neuron.2005.04.021 [DOI] [PubMed] [Google Scholar]
  12. Deichmann, R. , Gottfried, J. A. , Hutton, C. , & Turner, R. (2003). Optimized EPI for fMRI studies of the orbitofrontal cortex. NeuroImage, 19(2 Pt 1), 430–441. 10.1016/s1053-8119(03)00073-9 [DOI] [PubMed] [Google Scholar]
  13. Djordjevic, J. , Lundstrom, J. N. , Clément, F. , Boyle, J. A. , Pouliot, S. , & Jones‐Gotman, M. (2008). A rose by any other name: Would it smell as sweet? Journal of Neurophysiology, 99(1), 386–393. 10.1152/jn.00896.2007 [DOI] [PubMed] [Google Scholar]
  14. Donoshita, Y. , Choi, U. S. , Ban, H. , & Kida, I. (2021). Assessment of olfactory information in the human brain using 7‐Tesla functional magnetic resonance imaging. NeuroImage, 236, 118212. 10.1016/j.neuroimage.2021.118212 [DOI] [PubMed] [Google Scholar]
  15. Duff, M. C. , Covington, N. V. , Hilverman, C. , & Cohen, N. J. (2019). Semantic memory and the hippocampus: Revisiting, reaffirming, and extending the reach of their critical relationship. Frontiers in Human Neuroscience, 13, 471. 10.3389/fnhum.2019.00471 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Fournel, A. , Ferdenzi, C. , Sezille, C. , Rouby, C. , & Bensafi, M. (2016). Multidimensional representation of odors in the human olfactory cortex. Human Brain Mapping, 37(6), 2161–2172. 10.1002/hbm.23164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Friederici, A. D. (2011). The brain basis of language processing: From structure to function. Physiological Reviews, 91(4), 1357–1392. 10.1152/physrev.00006.2011 [DOI] [PubMed] [Google Scholar]
  18. García‐Cabezas, M. , & Barbas, H. (2014). A direct anterior cingulate pathway to the primate primary olfactory cortex may control attention to olfaction. Brain Structure & Function, 219(5), 1735–1754. 10.1007/s00429-013-0598-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. González, J. , Barros‐Loscertales, A. , Pulvermüller, F. , Meseguer, V. , Sanjuán, A. , Belloch, V. , & Avila, C. (2006). Reading cinnamon activates olfactory brain regions. NeuroImage, 32(2), 906–912. 10.1016/j.neuroimage.2006.03.037 [DOI] [PubMed] [Google Scholar]
  20. Görgen, K. , Hebart, M. N. , Allefeld, C. , & Haynes, J. D. (2018). The same analysis approach: Practical protection against the pitfalls of novel neuroimaging analysis methods. NeuroImage, 180(Pt A), 19–30. 10.1016/j.neuroimage.2017.12.083 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gottfried, J. A. (2010). Central mechanisms of odour object perception. Nature Reviews. Neuroscience, 11(9), 628–641. 10.1038/nrn2883 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Gottfried, J. A. , Deichmann, R. , Winston, J. S. , & Dolan, R. J. (2002). Functional heterogeneity in human olfactory cortex: An event‐related functional magnetic resonance imaging study. The Journal of Neuroscience, 22(24), 10819–10828. 10.1523/jneurosci.22-24-10819.2002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Gottfried, J. A. , & Dolan, R. J. (2003). The nose smells what the eye sees: Crossmodal visual facilitation of human olfactory perception. Neuron, 39(2), 375–386. 10.1016/s0896-6273(03)00392-1 [DOI] [PubMed] [Google Scholar]
  24. Gottfried, J. A. , Winston, J. S. , & Dolan, R. J. (2006). Dissociable codes of odor quality and odorant structure in human piriform cortex. Neuron, 49(3), 467–479. 10.1016/j.neuron.2006.01.007 [DOI] [PubMed] [Google Scholar]
  25. Grabenhorst, F. , Rolls, E. T. , Margot, C. , da Silva, M. A. , & Velazco, M. I. (2007). How pleasant and unpleasant stimuli combine in different brain regions: Odor mixtures. The Journal of Neuroscience, 27(49), 13532–13540. 10.1523/jneurosci.3337-07.2007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Howard, J. D. , & Gottfried, J. A. (2014). Configural and elemental coding of natural odor mixture components in the human brain. Neuron, 84(4), 857–869. 10.1016/j.neuron.2014.10.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Howard, J. D. , Plailly, J. , Grueschow, M. , Haynes, J. D. , & Gottfried, J. A. (2009). Odor quality coding and categorization in human posterior piriform cortex. Nature Neuroscience, 12(7), 932–938. 10.1038/nn.2324 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Johnson, D. M. , Illig, K. R. , Behan, M. , & Haberly, L. B. (2000). New features of connectivity in piriform cortex visualized by intracellular injection of pyramidal cells suggest that "primary" olfactory cortex functions like "association" cortex in other sensory systems. The Journal of Neuroscience, 20(18), 6974–6982. 10.1523/jneurosci.20-18-06974.2000 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kaeppler, K. , & Mueller, F. (2013). Odor classification: A review of factors influencing perception‐based odor arrangements. Chemical Senses, 38(3), 189–209. 10.1093/chemse/bjs141 [DOI] [PubMed] [Google Scholar]
  30. Karunanayaka, P. R. , Wilson, D. A. , Vasavada, M. , Wang, J. , Martinez, B. , Tobia, M. J. , Kong, L. , Eslinger, P. , & Yang, Q. X. (2015). Rapidly acquired multisensory association in the olfactory cortex. Brain and Behavior: A Cognitive Neuroscience Perspective, 5(11), e00390. 10.1002/brb3.390 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kass, R. E. , & Raftery, A. E. (1995). Bayes factors. Journal of the American Statistical Association, 90(430), 773–795. 10.2307/2291091 [DOI] [Google Scholar]
  32. Lane, G. , Zhou, G. , Noto, T. , & Zelano, C. (2020). Assessment of direct knowledge of the human olfactory system. Experimental Neurology, 329, 113304. 10.1016/j.expneurol.2020.113304 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Li, W. , Howard, J. D. , Parrish, T. B. , & Gottfried, J. A. (2008). Aversive learning enhances perceptual and cortical discrimination of indiscriminable odor cues. Science, 319(5871), 1842–1845. 10.1126/science.1152837 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lundström, J. N. , Regenbogen, C. , Ohla, K. , & Seubert, J. (2019). Prefrontal control over occipital responses to crossmodal overlap varies across the congruency spectrum. Cerebral Cortex, 29(7), 3023–3033. 10.1093/cercor/bhy168 [DOI] [PubMed] [Google Scholar]
  35. Lundström, J. N. , Seven, S. , Olsson, M. J. , Schaal, B. , & Hummel, T. (2006). Olfactory event‐related potentials reflect individual differences in odor valence perception. Chemical Senses, 31(8), 705–711. 10.1093/chemse/bjl012 [DOI] [PubMed] [Google Scholar]
  36. Mai, J. r. K. M. M. P. G. (2015). Atlas of the human brain. http://www.vlebooks.com/vleweb/product/openreader?id=none&isbn=9780128028018
  37. Manescu, S. , Frasnelli, J. , Lepore, F. , & Djordjevic, J. (2014). Now you like me, now you don't: Impact of labels on odor perception. Chemical Senses, 39(2), 167–175. 10.1093/chemse/bjt066 [DOI] [PubMed] [Google Scholar]
  38. Misaki, M. , Kim, Y. , Bandettini, P. A. , & Kriegeskorte, N. (2010). Comparison of multivariate classifiers and response normalizations for pattern‐information fMRI. NeuroImage, 53(1), 103–118. 10.1016/j.neuroimage.2010.05.051 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Miyamichi, K. , Amat, F. , Moussavi, F. , Wang, C. , Wickersham, I. , Wall, N. R. , Taniguchi, H. , Tasic, B. , Huang, Z. J. , He, Z. , Callaway, E. M. , Horowitz, M. A. , & Luo, L. (2011). Cortical representations of olfactory input by trans‐synaptic tracing. Nature, 472(7342), 191–196. 10.1038/nature09714 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Moeller, S. , Yacoub, E. , Olman, C. A. , Auerbach, E. , Strupp, J. , Harel, N. , & Uğurbil, K. (2010). Multiband multislice GE‐EPI at 7 tesla, with 16‐fold acceleration using partial parallel imaging with application to high spatial and temporal whole‐brain fMRI. Magnetic Resonance in Medicine, 63(5), 1144–1153. 10.1002/mrm.22361 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Mohedano‐Moriano, A. , Martinez‐Marcos, A. , Muñoz, M. , Arroyo‐Jimenez, M. M. , Marcos, P. , Artacho‐Pérula, E. , Blaizot, X. , & Insausti, R. (2005). Reciprocal connections between olfactory structures and the cortex of the rostral superior temporal sulcus in the Macaca fascicularis monkey. The European Journal of Neuroscience, 22(10), 2503–2518. 10.1111/j.1460-9568.2005.04443.x [DOI] [PubMed] [Google Scholar]
  42. Morán, M. A. , Mufson, E. J. , & Mesulam, M. M. (1987). Neural inputs into the temporopolar cortex of the rhesus monkey. The Journal of Comparative Neurology, 256(1), 88–103. 10.1002/cne.902560108 [DOI] [PubMed] [Google Scholar]
  43. Nichols, T. E. , & Holmes, A. P. (2002). Nonparametric permutation tests for functional neuroimaging: A primer with examples. Human Brain Mapping, 15(1), 1–25. 10.1002/hbm.1058 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Olofsson, J. K. , & Gottfried, J. A. (2015). The muted sense: Neurocognitive limitations of olfactory language. Trends in Cognitive Sciences, 19(6), 314–321. 10.1016/j.tics.2015.04.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Olson, I. R. , Plotzker, A. , & Ezzyat, Y. (2007). The enigmatic temporal pole: A review of findings on social and emotional processing. Brain, 130(Pt 7), 1718–1731. 10.1093/brain/awm052 [DOI] [PubMed] [Google Scholar]
  46. Poo, C. , & Isaacson, J. S. (2009). Odor representations in olfactory cortex: "sparse" coding, global inhibition, and oscillations. Neuron, 62(6), 850–861. 10.1016/j.neuron.2009.05.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Porada, D. K. , Regenbogen, C. , Seubert, J. , Freiherr, J. , & Lundström, J. N. (2019). Multisensory enhancement of odor object processing in primary olfactory cortex. Neuroscience, 418, 254–265. 10.1016/j.neuroscience.2019.08.040 [DOI] [PubMed] [Google Scholar]
  48. Porter, J. , Anand, T. , Johnson, B. , Khan, R. M. , & Sobel, N. (2005). Brain mechanisms for extracting spatial information from smell. Neuron, 47(4), 581–592. 10.1016/j.neuron.2005.06.028 [DOI] [PubMed] [Google Scholar]
  49. Power, J. D. , Plitt, M. , Kundu, P. , Bandettini, P. A. , & Martin, A. (2017). Temporal interpolation alters motion in fMRI scans: Magnitudes and consequences for artifact detection. PLoS One, 12(9), e0182939. 10.1371/journal.pone.0182939 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Price, J. L. (2009). Olfactory higher centers anatomy. In Squire L. R. (Ed.), Encyclopedia of neuroscience (pp. 129–136). Academic Press. 10.1016/B978-008045046-9.01692-2 [DOI] [Google Scholar]
  51. Qu, L. P. , Kahnt, T. , Cole, S. M. , & Gottfried, J. A. (2016). De novo emergence of odor category representations in the human brain. The Journal of Neuroscience, 36(2), 468–478. 10.1523/jneurosci.3248-15.2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Rennaker, R. L. , Chen, C. F. , Ruyle, A. M. , Sloan, A. M. , & Wilson, D. A. (2007). Spatial and temporal distribution of odorant‐evoked activity in the piriform cortex. The Journal of Neuroscience, 27(7), 1534–1542. 10.1523/jneurosci.4072-06.2007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Rissman, J. , Gazzaley, A. , & D'Esposito, M. (2004). Measuring functional connectivity during distinct stages of a cognitive task. NeuroImage, 23(2), 752–763. 10.1016/j.neuroimage.2004.06.035 [DOI] [PubMed] [Google Scholar]
  54. Roland, B. , Deneux, T. , Franks, K. M. , Bathellier, B. , & Fleischmann, A. (2017). Odor identity coding by distributed ensembles of neurons in the mouse olfactory cortex. eLife, 6, e26337. 10.7554/eLife.26337 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Rolls, E. T. , Kringelbach, M. L. , & de Araujo, I. E. (2003). Different representations of pleasant and unpleasant odours in the human brain. The European Journal of Neuroscience, 18(3), 695–703. 10.1046/j.1460-9568.2003.02779.x [DOI] [PubMed] [Google Scholar]
  56. Royet, J. P. , Plailly, J. , Delon‐Martin, C. , Kareken, D. A. , & Segebarth, C. (2003). fMRI of emotional responses to odors: Influence of hedonic valence and judgment, handedness, and gender. NeuroImage, 20(2), 713–728. 10.1016/s1053-8119(03)00388-4 [DOI] [PubMed] [Google Scholar]
  57. Sijben, R. , Hoffmann‐Hensel, S. M. , Rodriguez‐Raecke, R. , Haarmeier, T. , & Freiherr, J. (2018). Semantic congruence alters functional connectivity during olfactory‐visual perception. Chemical Senses, 43(8), 599–610. 10.1093/chemse/bjy048 [DOI] [PubMed] [Google Scholar]
  58. Skipper, L. M. , Ross, L. A. , & Olson, I. R. (2011). Sensory and semantic category subdivisions within the anterior temporal lobes. Neuropsychologia, 49(12), 3419–3429. 10.1016/j.neuropsychologia.2011.07.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Stettler, D. D. , & Axel, R. (2009). Representations of odor in the piriform cortex. Neuron, 63(6), 854–864. 10.1016/j.neuron.2009.09.005 [DOI] [PubMed] [Google Scholar]
  60. Tantirigama, M. L. , Huang, H. H. , & Bekkers, J. M. (2017). Spontaneous activity in the piriform cortex extends the dynamic range of cortical odor coding. Proceedings of the National Academy of Sciences of the United States of America, 114(9), 2407–2412. 10.1073/pnas.1620939114 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Turner, B. H. , Gupta, K. C. , & Mishkin, M. (1978). The locus and cytoarchitecture of the projection areas of the olfactory bulb in Macaca mulatta . The Journal of Comparative Neurology, 177(3), 381–396. 10.1002/cne.901770303 [DOI] [PubMed] [Google Scholar]
  62. Vaughan, D. N. , & Jackson, G. D. (2014). The piriform cortex and human focal epilepsy. Frontiers in Neurology, 5, 259. 10.3389/fneur.2014.00259 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Wang, Q. , Cagna, B. , Chaminade, T. , & Takerkart, S. (2020). Inter‐subject pattern analysis: A straightforward and powerful scheme for group‐level MVPA. NeuroImage, 204, 116205. 10.1016/j.neuroimage.2019.116205 [DOI] [PubMed] [Google Scholar]
  64. Wang, L. , Zhang, Z. , Chen, J. , Manyande, A. , Haddad, R. , Liu, Q. , & Xu, F. (2020). Cell‐type‐specific whole‐brain direct inputs to the anterior and posterior piriform cortex. Frontiers in Neural Circuits, 14, 4. 10.3389/fncir.2020.00004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. You, Y. , Novak, L. R. , Clancy, K. J. , & Li, W. (2022). Pattern differentiation and tuning shift in human sensory cortex underlie long‐term threat memory. Current Biology, 32(9), 2067–2075.e2064. 10.1016/j.cub.2022.02.076 [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Zelano, C. , Bensafi, M. , Porter, J. , Mainland, J. , Johnson, B. , Bremner, E. , Telles, C. , Khan, R. , & Sobel, N. (2005). Attentional modulation in human primary olfactory cortex. Nature Neuroscience, 8(1), 114–120. 10.1038/nn1368 [DOI] [PubMed] [Google Scholar]
  67. Zelano, C. , Mohanty, A. , & Gottfried, J. A. (2011). Olfactory predictive codes and stimulus templates in piriform cortex. Neuron, 72(1), 178–187. 10.1016/j.neuron.2011.08.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Zelano, C. , Montag, J. , Johnson, B. , Khan, R. , & Sobel, N. (2007). Dissociated representations of irritation and valence in human primary olfactory cortex. Journal of Neurophysiology, 97(3), 1969–1976. 10.1152/jn.01122.2006 [DOI] [PubMed] [Google Scholar]
  69. Zhou, G. , Lane, G. , Cooper, S. L. , Kahnt, T. , & Zelano, C. (2019). Characterizing functional pathways of the human olfactory system. eLife, 8, e47177. 10.7554/eLife.47177 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

APPENDIX S1: Supporting information.

HBM-45-e26681-s001.docx (3.7MB, docx)

Data Availability Statement

The data that support the findings of this study are openly available in Open Science Framework at https://osf.io.

Articles from Human Brain Mapping are provided here courtesy of Wiley

RESOURCES