Perceptual differences in olfactory fat discrimination are not detected in neural activation

Abstract

Olfaction is involved in detecting, identifying, and discriminating dietary fat within foods, yet the underlying neural mechanisms remain uncharted. Our functional magnetic resonance imaging (fMRI) study investigated the neural correlates of olfactory fat perception and their association with discrimination ability in a complex food matrix. We measured brain activation resulting from orthonasal exposure to an ecologically relevant fat-related odor source—dairy milk, manipulated to contain 0%, 3.5%, or 14% fat. Twenty-six healthy, normosmic adults underwent olfactory fat content discrimination testing, followed by an fMRI task during which the 3 odor stimuli were delivered via an olfactometer (25 times/fat level) and rated on perceived intensity and liking. Participants discriminated between all fat levels, with fat level influencing perceived odor intensity and liking. These perceptual differences, however, were not reflected in differential brain activation. Brain activation differences were observed only when comparing odor exposure with no exposure. Specifically, in response to any odor, activation occurred in the anterior part of the supplementary motor area (SMA) while deactivating parts of the hippocampus, putamen, superior temporal gyrus, anterior cingulate cortex, insula, and posterior part of the SMA. Exposure to the 0% fat odor also activated the thalamus. No associations were found between perceived intensity and liking and neural responses. Results reaffirm the human ability to distinguish food fat content using solely olfactory cues and reveal a divergence between sensory perception and neural processing. Subsequent research should replicate and extend these findings onto retronasal fat perception while also examining potential effects of hunger, genetics, and dietary habits.

fat perception, odor, discrimination, fMRI, brain activity

1 Introduction

Dietary fat is an indispensable macronutrient in the human diet, playing a vital role in maintaining and promoting optimal health (Lichtenstein et al. 1998). The nutritional significance of dietary fat is often overshadowed by its pleasure-inducing sensory characteristics (Drewnowski 1997; Drewnowski and Almiron-Roig 2009), which can promote overconsumption, thereby contributing toward the development of obesity (Golay and Bobbioni 1997; Blundell and Macdiarmid 1997; Bray et al. 2004). In light of this, gaining a deeper understanding of the sensory perception of dietary fat could offer valuable insights for developing effective public health strategies targeting the reduction of fat intake.

Sensory perception of dietary fat is multimodal (Guichard et al. 2018), with olfaction playing a considerable role. It contributes to its detection, identification, and discrimination, even when fat or its constituents are embedded within complex food matrices. Olfactory discrimination of food fat content has been corroborated by our previous work, demonstrating that fat levels in dairy milk can be distinguished using solely ortho- or retronasal cues (see Pirc et al. 2022). Moreover, fat-related odors have been shown to have the capacity of altering the perception of various olfactory as well as nonolfactory sensory qualities, such as mouthfeel (see review by Pirc et al. 2023). These findings affirm the notion that humans possess a functional olfaction-based mechanism for detecting and discriminating food fat content. Nevertheless, despite convincing perceptual evidence, the underlying mechanisms remain to be fully understood, particularly in the neurobiological domain. Considering that orthonasal odors relate to food source detection and appetite arousal prior to eating, with retronasal odors playing a vital role in flavor perception, potentially affecting intake and satiation (Bojanowski and Hummel 2012; Boesveldt and Lundstrom 2014), investigating neural mechanisms underpinning olfactory fat perception could further our understanding of feeding behavior surrounding fatty foods. In the context of curbing excessive fat intakes, exploring differences in brain activation resulting from olfactory exposure to varying food fat levels is particularly relevant.

In contrast to oral fat perception (i.e. taste and/or mouthfeel), which has been investigated in numerous neurobiological studies (Verhagen et al. 2003; De Araujo and Rolls 2004; Grabenhorst et al. 2010; Eldeghaidy et al. 2011; Wistehube et al. 2018; Andersen et al. 2020), to our knowledge, no study investigated how exposure to fat-related odors is processed in the human brain. Notwithstanding the lack of such studies, oral fat perception research has identified several brain areas which might be of relevance in the neural processing of fat-related odors as well. Activation in response to oral fat stimulation has been observed in the insula (Eldeghaidy et al. 2011), orbitofrontal cortex (OFC), anterior cingulate cortex (ACC), hypothalamus, and ventral striatum (VS) (De Araujo and Rolls 2004). Given the overlap in neural processing between smell and taste perception in regions such as the insula (Small et al. 2005; Veldhuizen et al. 2010; Lundström et al. 2011; Yeung et al. 2018; Torske et al. 2022), OFC, and ACC (Small et al. 2004), which are crucial in the processing of flavor perception, investigating brain activation in these regions in response to fat-related odor exposure seems promising. Moreover, it is widely acknowledged that energy-dense foods, including those rich in fat, are potently rewarding (Drewnowski 1997). A crucial brain system involved in the processing of reward, specifically the generation of desire and pleasure in response to rewarding stimuli such as fat-containing foods, is the mesolimbic dopamine system (Volkow et al. 2011; Berridge and Kringelbach 2015; Berridge and Robinson 2016; Higgs 2016; Robinson et al. 2016). It encompasses overlapping reward networks, which include regions such as the prefrontal cortex, including portions of the OFC, the insula, and ACC, as well as subcortical limbic structures such as the VS (including the nucleus accumbens and ventral pallidum), ventral tegmental area, and amygdala (Berthoud 2002; Berridge and Kringelbach 2015; Higgs 2016). Taken together, brain areas likely to be involved in the processing of fat-related odors include reward- and oral fat perception-related areas, both contained within the broader reward system. Olfaction-related areas, as per Lundström et al. (2011), Seubert et al. (2013), and Fjaeldstad et al. (2017), might be implicated as well.

The current study aimed to map brain activation in response to olfactory (orthonasal) exposure to varying levels of dietary fat embedded within an ecologically relevant food source (dairy milk), and exploring potential associations between brain activation, olfactory fat content discrimination ability, and perceived odor intensity and liking of the utilized samples. We hypothesized that exposure to higher fat levels would lead to increased activation in regions associated with reward and olfactory perception. We also expected to confirm previous findings demonstrating participants’ ability to discriminate between sample fat levels using only orthonasal cues. Since high fat foods tend to be perceived as pleasurable (Blundell and MacDiarmid 1997; Guichard et al. 2018), we expected positive correlations between fat content and perceived liking, and between perceived liking and brain activation in various regions of the reward system.

2 Materials and methods

The study was conducted according to the Declaration of Helsinki and approved by the Medical Ethics Committee Oost Nederland (NL78261.091.21).

2.1 Participants

Twenty-six volunteers (mean age 24.9 ± 5.4 years; mean BMI 22.6 ± 2.1 kg/m²; 6 males), recruited from Wageningen (The Netherlands) and its surroundings, participated in the study. All were healthy, nonsmoking, regular dairy milk consumers (self-reported) and met the following eligibility criteria: being between 16 and 55 years of age (as smell function decreases with age; Doty and Kamath 2014); having a BMI between 18.5 and 25 kg/m² (as BMI may negatively influence olfactory functioning; Peng et al. 2019); being normosmic; being right-handed (as handedness is associated with differences in odor-related processing; Hummel et al. 1998; Royet et al. 2003); not dieting currently or in the past 2 months; not having any dairy-related allergies or intolerances; not being pregnant or lactating (as pregnancy may negatively influence olfactory performance; Ochsenbein-Kolble et al. 2007); not suffering from claustrophobia; and having no other MRI-related contraindications (e.g. nonremovable ferromagnetic implants and piercings, epilepsy). All participants provided written informed consent and received financial compensation for taking part in the study.

2.2 Olfactory stimuli

Dairy milk samples, containing 0%, 3.5%, and 14% fat (henceforth referred to as low fat (L), medium fat (M), and high fat (H), respectively), served as olfactory stimuli throughout the study. They were chosen to represent ecologically relevant fat sources with comparable fat content, namely skimmed milk, whole milk, and reduced-fat cooking cream. Fresh, pasteurized skimmed milk (0% fat, AH Magere melk, Albert Heijn B.V.) served as the L sample (containing traces of fat), whereas M and H samples were made from fresh, pasteurized skimmed milk and fresh, pasteurized full-fat cream (35% fat, AH Verse Slagroom, Albert Heijn B.V.), and homogenized with a dispersing apparatus (T 25 digital Ultra-Turrax, IKA - Werke GmbH & Co. KG). All samples were prepared freshly prior to each experimental session. See Table A1 in the supplementary material for sample ingredients and corresponding nutritional composition.

2.3 Study design and procedures

The study was carried out in 2 sessions, conducted on separate days. During the first session, participants underwent screening and training, along with discrimination testing aimed at assessing their orthonasal fat content discrimination ability. The second session was aimed at measuring brain responses resulting from orthonasal exposure to the 3 milk samples, by means of functional magnetic resonance imaging (fMRI). Perceived intensity and liking of said odor stimuli were assessed during this session as well. See Fig. 1 for an overview of the study design.

Fig. 1.

Study design overview. First, participants underwent screening and training, along with discrimination testing aimed at assessing their orthonasal fat content discrimination ability. The fMRI session was aimed at measuring brain responses resulting from orthonasal exposure to the three milk samples.

Participants were instructed to refrain from using strong-scented cosmetic products and consuming anything apart from water 2 h prior to all sessions.

2.3.1 Screening and training.

Screening involved questionnaires aimed at assessing study eligibility criteria, body height and weight measurements, and olfactory function assessment. The latter was carried out using the Sniffin’ Sticks 16-item odor identification test, with a score of ≥12 indicating normosmia (Hummel et al. 2007; Oleszkiewicz et al. 2019).

If deemed eligible, participants underwent a short training procedure, to get accustomed to handling odor delivery containers used during discrimination testing. Additionally, a full-size, nonfunctioning, MRI scanner replica was used to familiarize them with the scanning environment and fMRI task-related procedures. While in the replica scanner, participants performed a short practice sequence, similar to the experimental one, which involved seeing visual stimuli displayed on a screen through a head coil-mounted mirror, experiencing the odor delivery system (cream odor, delivered via an olfactometer; Lundstrom et al. 2010), hearing prerecorded MRI noises, and learning how to use a response box to answer visual analogue scale (VAS) questions in the scanner.

2.3.2 Discrimination testing.

Discrimination testing took place in sensory booths and followed the dual reminder A-not A (DR A-not A) (see Mun et al. 2019) and pairwise design (Hautus et al. 2018). In this version of the test, participants are first required to smell a reference stimulus twice, before smelling the test stimulus once, and deciding whether the test stimulus is the reference or not. The following response options were available: “test sample is the reference – I am sure,” “test sample is the reference – I am unsure,” “test sample is the reference – I am guessing,” “test sample is not the reference – I am guessing,” “test sample is not the reference – I am unsure,” and “test sample is not the reference – I am sure.” Testing was carried out in 3 blocks of 6 trials, with 30-s and 3-min breaks implemented in-between tests and blocks, respectively. Two fat content levels were compared within a block: either 0% and 3.5% (L-M); 0% and 14% (L-H); or 3.5% and 14% (M-H), with the lower fat level of each pair always serving as a reference. Two possible sample presentation sequences were utilized: S_A – S_A – S_A or S_A – S_A – S_{not A}. Block and trial orders were randomized across participants and each block started with a sample familiarization procedure, to stabilize participants’ cognitive decision criteria (Lee et al. 2007) (see Pirc et al. 2022 for details on familiarization). Olfactory stimuli were presented in 60-mL amounts at 20 ± 1 °C, using specialized odor delivery containers, as described by Pirc et al. (2022).

2.3.3 MRI scanning.

Upon arrival to the MRI facility, participants’ adherence to experimental and MRI safety protocols were assessed. If necessary, MRI-safe clothing and glasses were provided. Scanning commenced with a 5-min anatomical scan, followed by a brief practice procedure, to familiarize participants with the use of the response box, before concluding with the olfactory fMRI task which spanned 3 consecutive 14-min functional runs.

A scanner-mounted computer display, receiving input from a computer running E-Prime 3.0 (Psychology Software Tools, Sharpsburg, USA) stimulus presentation software, was used to present instructions, questions, and other visual stimuli to participants in the scanner. Specifically, an orange fixation crosshair indicated imminent odor release, a green fixation crosshair indicated odor release and prompted participants to inhale, while a white fixation crosshair indicated the waiting period. All visual stimuli were presented on a dark gray background.

Throughout the fMRI task, odors were delivered by means of an 8-channel olfactometer (Burghart Messtechnik GmbH, Holm, Germany) and presented orthonasally to both nostrils using a small nasal canula. During scanning, the odors were embedded in a constant nonodorous airflow (8 L/min without odor, 4 L/min with odor) with a relative humidity of 80% and a temperature of 36 °C. Odor release was controlled using E-Prime 3.0 stimulus presentation software.

Participants were randomly assigned to 1 of 2 odor presentation sequences to avoid potential order effects. In both sequences, each odor (L, M, or H) was presented 25-times for 2 s. The interstimulus interval varied between 27 and 32 s, including time for perceptual ratings (for some of the trials), a jittered waiting period (“REST”), to avoid expectation effects, and a 1-s cue indicating that the odor would be released, in order for participants to prepare and align their inhalation. In total, the fMRI task consisted of 75 odor trials. In 18 trials, 100-unit VAS questions about perceived odor intensity and liking (3 of each per odor stimulus level, left anchor = “not at all”; right anchor = “extremely”) were included. Participants were allotted 11 s to provide their responses using the response box. See Fig. 2 for a schematic overview of an fMRI trial.

Fig. 2.

Overview of fMRI trial. A trial started with a cue that the odor would be released, followed by the odor presentation. In some of the trials, participants were asked to rate intensity or liking of the odor stimuli, in all trials there was a waiting period.

MRI image acquisition was performed using a 3-Tesla MRI scanner (Elition X, Koninklijke Philips N.V., Amsterdam, the Netherlands) using a 32-channel head coil. A high-resolution T₁-weighted 3D TFE anatomical scan was conducted with the following parameters: repetition time (TR) of 10 ms, echo time (TE) of 4.6 ms, flip angle of 8°, field of view (FOV) of 256 × 243 × 180 mm, acquisition of 450 sagittal slices, scanning voxel size of 0.8 × 0.8 × 0.8 mm, and reconstructed voxel size of 0.4 × 0.4 × 0.4 mm. Functional scans were conducted using a T₂-weighted gradient echo 2D-EPI sequence. The following parameters were used: TR = 1152 ms, TE = 25 ms, a flip angle of 57°, SENSE factor = 2.2 (AP), multiband factor = 3, FOV = 230 × 230 × 139 mm, acquisition = 63 axial slices in an ascending order, scanning voxel size = 2.2 × 2.2 × 2.2 mm, and reconstructed voxel size = 1.8 × 1.8 × 2.2 mm.

2.4 Data analysis

2.4.1 Perceptual data.

Discrimination ability was assessed with R-index analyses carried out in accordance with the protocols described by Lee and van Hout (2009). To account for replicated testing, R-indices were computed based on weighted means of individual R-index values (derived from 6 signal/noise tests per participant) (Bi 2015). Statistical significance was established by calculating the R-index critical value, using R statistical software (R-Core Team 2020) and the code provided by Bi and O’Mahony (2020). The R-index critical value for a one-sided test, involving 78 control and 78 test samples, at significance levels of 0.05 and 0.001, amounts to 57.52 and 63.67, respectively.

Effects of odor fat levels on ratings of perceived intensity and liking were analyzed using linear mixed model (LMM) analyses in SPSS Statistics version 29 (IBM Corp.), using intensity or liking as dependent variables, odors as a fixed factor, and subjects as a random factor. Post-hoc pairwise comparisons with Bonferroni corrections were applied to compare ratings between the odors. Potential effects of olfactory adaptation due to repeated exposure to the odors throughout the 3 functional runs were assessed by adding intensity as dependent variables to the model, fMRI task progress (functional runs) as a fixed factor, and subjects as a random factor. Statistical significance was set at P < 0.05.

2.4.2 fMRI data.

Preprocessing and analysis of the imaging data were performed using the SPM12 (revision version 7771) software package (Wellcome Department of Imaging Neuroscience, London, UK) implemented in MATLAB R2021a (The Mathworks, Inc., Natick, MA, USA).

Preprocessing involved realigning and slice-time correcting functional images before coregistering anatomical images with the mean functional image. Following coregistration, both the anatomical and functional images were normalized to the standard Montreal Neurological Institute (MNI) space. After this, functional images underwent smoothing, using a 3D Gaussian smoothing kernel ([3.6, 3.6, 4.4] mm full width at half-maximum). To assess motion-related artifacts, a volume-wise check was performed using ArtRepair software (Mazaika et al. 2009). Data from all participants were deemed suitable for analysis, as none of the participants exceeded the motion exclusion threshold (more than 20% of the total volumes exceeding 0.5 mm/TR).

Statistical parametric maps were generated per participant as part of subject level analyses. This was done by fitting a boxcar function to the time series and convolving it with the canonical hemodynamic response function. To remove low-frequency noise, data were filtered using a high-pass filter (128-s cutoff). Individual general linear models included 5 conditions per functional run: (1) “Prepare to smell” (orange fixation crosshair); (2) L odor; (3) M odor; (4) H odor; and (5) subjective ratings (VAS). Motion-related variance was accounted for by adding the realignment parameters as regressors to the model. Despite odor presentation lasting 2 s, it was modeled as a 3-s event, due to the potential lingering nature of the utilized odors, which were expected to have an impact beyond the duration of the direct exposure. Ultimately, contrast images comparing various conditions were created.

Contrasts compared odors with each other (L–M, M–H, and L–H) and with the “REST” condition (L–REST; M–REST; and H–REST), during which participants viewed a white crosshair in the absence of odor stimulation, subjective ratings, or preparatory cues. As further exploration, comparisons were also made between exposure to samples containing no fat and samples containing fat (L–MH). Parametric modulation, using odor stimulus levels as modulators, was performed as well. However, as the obtained results closely replicated those of the other analysis approach, further discussion of these findings is omitted.

The different contrast images were analyzed on a group level using one sample t-tests. Whole-brain and region of interest (ROI) analyses were performed. For the ROI analysis, an ROI mask image was created by combining Automated Anatomic Labeling (AAL) atlas (Tzourio-Mazoyer et al. 2002) masks of reward-related brain areas, as described by Smeets and de Graaf (2019), as well as olfaction-related brain areas, as described by Seubert et al. (2013) and Fjaeldstad et al. (2017) with the use of the WFU PickAtlas toolbox (Maldjian et al. 2003). Brain areas included were the insula, anterior and middle cingulate cortex, supplementary motor area (SMA), OFC (including the inferior, middle, and superior frontal gyri), thalamus, striatum (caudate, putamen, and pallidum), amygdala, hippocampus, piriform cortex, entorhinal cortex, parahippocampal gyrus, gyrus rectus, and the superior temporal pole. Statistical significance for ROI analyses was determined using a cluster-forming threshold of P = 0.001 (uncorrected), with a cluster voxel extent (k) of k > 19. Clusters were deemed significant when the cluster-level quantitative false-discovery rate (q_FDR) was less than 0.05. Whole-brain analyses were executed using an FWE-corrected threshold of P = 0.05, with a cluster voxel extent threshold of k > 5. The rationale for choosing a less stringent multiple comparison correction for ROI analysis was to avoid missing potentially relevant neural landmarks (Lieberman and Cunningham, 2009). Because the primary classical analysis did not support the existence of differences in brain activation between the odor stimuli, we conducted a Bayesian analysis to assess the strength of evidence for a difference in odor-related activation following the guidance of Han and Park (2018). Given the lack of differences observed with the classical analysis approach a Cohen’s d of 0.2 (small effect size) and a log-odds threshold of 3 (indicating the presence of weak evidence) were used.

Potential associations between brain activation in response to utilized odors and perceptual ratings were investigated by correlating mean parameter estimates from significant clusters identified during ROI analyses with mean perceived odor intensity and liking ratings, using Spearman correlations (in SPSS Statistics version 29, IBM Corp.). Mean parameter estimates were extracted from significant clusters using the MarsBar toolbox (http://marsbar.sourceforge.net/) run in MATLAB R2021a. Correlation analyses were carried out per contrast and corresponding odor ratings. Due to the absence of significant correlations within the significant ROI clusters and the lack of theoretical rationale, we did not extend the correlation analysis to include broader brain regions outside of these clusters.

Brain activation clusters were overlaid onto a mean anatomical image of all participants and identified with a combination of the use of the AAL brain atlas in MRIcron, version 1.40, build 1 (https://www.nitrc.org/projects/mricron) and Neuromorphometrics in SPM.

3 Results

3.1 Discrimination testing

Results of R-index analyses (Fig. 3) indicate that all 3 odors could be discriminated orthonasally: L-M (M_R-index = 82.4 ± 23.7, P < 0.001), L-H (M_R-index = 84.7 ± 24.2, P < 0.001), and M-H (M_R-index = 69.6 ± 21.4, P < 0.001).

Fig. 3.

Boxplots of R-index analysis results. Dashed line indicates the discrimination cutoff at P < 0.001. L-M = low fat compared with medium fat samples; L-H = low fat compared with high fat samples; M-H = medium fat compared with high fat samples. All 3 odors could be discriminated from each other.

3.2 Perceptual ratings

LMM analyses indicate that odor stimulus fat level had a main effect on perceived odor intensity (F(2, 206) = 3.86, P = 0.023) and liking (F(2, 206) = 14.29, P < 0.001) (see Fig. 4).

Fig. 4.

Boxplots of odor intensity (left) and liking ratings (right) for the 3 odors. Asterisks denote statistically significant (P < 0.05) differences between means (L = low fat, M = medium fat, H = high fat). Intensity of the L odor was rated significantly lower than that of the M odor, but not than that of the H odor. Odor liking ratings differed significantly between all the 3 odors and increased with fat level.

Intensity of the L odor (M_L = 42.9 ± 25.8) was rated significantly lower (P = 0.020) than that of the M odor (M_M = 51.5 ± 21.1). No significant differences in intensity were observed between L and H odors (M_H = 48.4 ± 18.8) (P = 0.244) nor between M and H odors (P = 0.963).

Odor liking ratings differed significantly between all the 3 odors (P_L-M = 0.029; P_L-H < 0.001; P_M-H = 0.021), and increased with fat level (M_L = 44.4 ± 20.3; M_M = 50.8 ± 18.0; M_H = 57.4 ± 16.6).

No effects of repeated exposure to the 3 odors were observed on intensity ratings (F(2, 206) = 2.103, P = 0.125), indicating that olfactory adaptation had not occurred with repeated exposure to the 3 odors during the fMRI task.

3.3 Neuroimaging results

ROI, as well as whole-brain analyses comparing brain activation resulting from exposure to the different odor stimuli (L-M, L-H, M-H, and L-MH), did not show any significant differences. In the Bayesian analysis, there was not even weak evidence for a difference in activation between the 3 stimuli suggesting that neural responses to the different fat concentrations were equivalent. These results remained the same when n = 5 nondiscriminators were omitted. Consequently, investigation into potential relationships between discrimination ability for these odor comparisons and brain activation was not pursued.

Significantly activated ROI areas per odor condition (L, M, H, vs. REST) are presented in Fig. 5, with corresponding mean parameter estimates being presented in Fig. 6.

Fig. 5.

Color-coded T-maps of ROI activation per odor condition compared with REST, overlaid onto the mean anatomical image at T = 3.45 (P = 0.001). Named ROIs denote significant clusters (qFDR < 0.05). Signification activations (indicated in yellow) for all odors were found in the SMA. Significant deactivations (indicated in blue) for all odors were found in STG, hippocampus, putamen, ACC, insula, and SMA.

Fig. 6.

Boxplots of parameter estimates associated with ROI brain activation per odor condition compared with REST. Yellow colored boxes indicate positive parameter estimates (i.e. activation vs. rest); blue boxes indicate negative parameter estimates (i.e. deactivation vs. rest).

An overview of the activated ROI areas in the L-REST comparison is presented in Table 1. ROI analyses comparing exposure to the L odor versus the REST condition revealed greater activation in the right SMA and left thalamus. Conversely, deactivation was observed in the left superior temporal gyrus (STG), bilateral hippocampus, bilateral putamen, left ACC, left insula, and right SMA.

Table 1.

ROI brain activation in response to the L odor, compared with rest.

Contrasta	Brain region	Side	k	Peak voxel coordinate (MNI)			Z-score
				X	Y	Z
L > REST	SMA	R	157	4	16	66	5.03
	Thalamus	L	83	−1	−22	11	4.52
L < REST	STG	L	256	−57	−15	0	5.49
	Hippocampus	L	701	−30	−22	−17	5.27
		R	404	26	−42	0	4.99
	Putamen	R	897	28	−2	14	5.12
		L	411	−28	−9	5	4.60
	ACC	L	1055	−8	39	7	4.82
	Insula	L	136	−35	−22	18	4.67
	SMA	R	245	8	−24	58	4.65

^aSignificant at cluster-level q_FDR < 0.05; L = left, R = right; k = cluster extent.

An overview of the activated ROI areas in the M-REST comparison is presented in Table 2. ROI analyses comparing exposure to the M odor versus the REST condition showed greater activation in the right SMA, and deactivation in the bilateral STG, bilateral hippocampus, left putamen, left ACC, bilateral insula, and right SMA.

Table 2.

ROI brain activation in response to the M odor, compared with rest.

Contrasta	Brain Region	Side	k	Peak voxel coordinate (MNI)			Z-score
				X	Y	Z
M > REST	SMA	R	175	4	16	66	4.87
M < REST	Hippocampus	L	722	−17	−8	−22	5.24
		R	523	28	−2	−19	5.11
	SMA	R	230	6	−22	60	5.20
	STG	L	276	−59	−13	0	5.13
		R	703	49	−9	−13	4.59
	Putamen	L	445	−28	−9	5	5.12
	ACC	L	920	−5	41	−4	4.72
	Insula	L	159	−35	−22	18	4.45
		L	95	−34	14	−15	3.98
		R	101	38	14	−15	4.25

^aSignificant at cluster-level q_FDR < 0.05.

An overview of the activated ROIs in the H-REST comparison is presented in Table 3. ROI analyses comparing exposure to the H odor versus the REST condition showed greater activation in the right SMA, and deactivation in the left STG, bilateral hippocampus, left putamen, left ACC, bilateral insula, and right SMA.

Table 3.

ROI brain activation in response to the H odor, compared with rest.

Contrasta	Brain region	Side	k	Peak voxel coordinate (MNI)			Z-score
				X	Y	Z
H > REST	SMA	R	172	4	16	66	4.80
H < REST	Hippocampus	L	672	−23	−8	−17	5.38
		R	243	29	−2	−19	4.60
		R	196	26	−40	3	4.78
	SMA	R	200	6	−22	60	5.30
	Putamen	L	418	−30	−9	3	5.20
	STG	L	245	−55	−13	−2	4.91
	ACC	L	645	−7	39	−4	4.72
	Insula	R	595	42	−17	5	4.56
		L	150	−39	−22	20	4.41

^aSignificant at cluster-level q_FDR < 0.05.

An overview of brain regions activated in various comparisons during exploratory whole-brain analyses is presented in Table 4. Whole brain analyses showed activation in the left Cerebellum Crus1 region and deactivation in the left SMA in the L odors versus REST comparison; activation in the left Cerebellum Crus1 region and deactivation in the left Fusiform gyrus in the M versus REST comparison; and activation in the left Lingual gyrus, with deactivation in the left Putamen and right Cerebellum IX region in the H versus REST comparison.

Table 4.

Identified brain activation (whole-brain analysis) in response to the 3 odors, compared with rest.

Contrasta	Brain region	Side	k	Peak voxel coordinate (MNI)			Z-score
				X	Y	Z
L > REST	Cerebellum Crus1	L	10	−7	−83	−15	5.56
L < REST	STG	L	14	−57	−15	0	5.49
M > REST	Cerebellum Crus1	L	7	−7	−83	−15	5.42
M < REST	Fusiform gyrus	L	13	−32	−49	−15	5.81
H > REST	Lingual gyrus	L	6	−5	−78	−11	5.25
H < REST	Putamen	L	12	−30	−9	3	5.20
	Cerebellum IX	R	7	4	−49	−37	5.46

^aSignificant at cluster level P_FWE-corr < 0.05.

No significant correlations between significant ROI clusters and corresponding perceptual ratings of odor intensity and liking were observed in any condition (see Appendix A, Table A2 in the supplementary material for correlation analysis outputs).

4 Discussion

The current study investigated brain responses resulting from olfactory exposure to varying fat concentrations embedded within an ecologically relevant food source—dairy milk. In addition to mapping neural correlates of olfactory fat perception, it also aimed at exploring potential associations between brain activation, olfactory fat content discrimination ability, and perceived odor intensity and liking of the odor stimuli. Despite samples being perceptually distinguishable, there was no differential brain activation between the odors differing in fat content. Perceived intensity and liking differences between the odors could not be linked to specific neural responses either.

In line with previous work on olfactory fat perception (Boesveldt and Lundstrom 2014; Pirc et al. 2022), all 3 odor stimulus fat levels could be discriminated using solely olfactory cues. Whereas discrimination between 3.5% and 14% fat was previously only reported for retronasal cues (Pirc et al. 2022), the current study also confirmed it for orthonasal ones. Considering that orthonasal detection thresholds tend to be lower than retronasal ones (Chale-Rush et al. 2007; Goldberg et al. 2018), this is an expected finding. Perceptual differences between the 3 odors were also reflected in perceived odor intensity and liking. Notably, while the M odor was perceived as more intense compared with both L and H odors, liking consistently increased with fat concentration. This aligns with observations from our earlier work (Boesveldt and Lundstrom 2014; Pirc et al. 2022). The consistent discrepancy between discrimination ability and perceived odor intensity observed across our experiments reaffirms the notion that olfactory fat content discrimination is underpinned by factors other than intensity differences. It is likely that odor quality differences between fat levels play a more relevant role when it comes to olfactory fat content discrimination. Several studies support the notion that even minute odor stimulus concentration alterations may affect perceived odor quality (Gross-Isseroff and Lancet 1988; Le Berre et al. 2008; Stevenson 2011).

Contrary to expectations, and despite perceptual differences, no differential ROI brain activation was observed when comparing exposure to the 3 odors. Likewise, no differential activation was observed when comparing exposure to nonfat odors from the lower fat range to exposure to odors from the higher fat range (L–MH). Odor exposure resulted in changes in brain activity only when compared with the rest condition (no odor exposure). Specifically, exposure to either odor activated the SMA, while exposure to the L odor also activated the thalamus. Additionally, all 3 odors led to the deactivation in the STG, hippocampus, SMA, putamen, ACC, and insula.

The thalamus, hippocampus, and insula are all involved in olfactory processing (Lundström et al. 2011; Seubert et al. 2013; Roy-Côté et al. 2021). The thalamus, while traditionally not considered as an olfactory relay (Kay and Sherman, 2007), receives input from primary olfactory sensory areas (Lundström et al. 2011). It has also been implicated in modulating odor-related attention (Sabri et al. 2005; Plailly et al. 2008; Tham et al. 2009). Considering the latter, thalamic activation in response to the L odor suggests that its relatively low intensity might have required more attention from participants anticipating a percept during odor release. However, given the absence of intensity differences between L and H odors, one might also anticipate a thalamic response during H odor exposure. Like the thalamus, the hippocampus plays a role in integrating information from various sensory inputs, including olfaction (Zhou et al. 2021). Moreover, this area is known for its role in the formation of odor-related memories (Eichenbaum and Otto 1992; Eichenbaum 1998) and has been shown to be activated by orthonasally presented food odors, such as chocolate (Small et al. 2005). While we can speculate that the hippocampus’s role in this study relates to exposure to food odors, the absence of a nonfood modality and the lack of comparable studies hinders definitive conclusions regarding its deactivation. Lastly, the insula acts as a junction for chemosensory inputs integral to food flavor perception (Seubert et al. 2013; Roy-Côté et al. 2021), responding to various food odors (Small et al. 2005; Sorokowska et al. 2016). A pivotal factor in insular activation appears to be stimulus valence (Roy-Côté et al. 2021), with the right insula purportedly responding to pleasant odors (Fulbright et al. 1998; Heining et al. 2003) and the left to unpleasant ones (Lombion et al. 2009; Bensafi et al. 2012; A Sorokowska et al. 2016). The right insula has also been found to be more activated by food odors than nonfood ones (Sorokowska et al. 2017). While these insights into insular function provide valuable context, they do not directly elucidate the insular deactivation observed in our study. The bilateral insular response to M and H odors challenges the typical lateralization based on valence. Furthermore, the notion that the left insula primarily responds to unpleasant odors is at odds with the liking ratings for the L odor, which suggest that it was perceived as neutral rather than negative. The reasons for insular deactivation in this study therefore remain unclear.

Two distinct clusters of activation were observed in the supplementary motor cortex. Specifically, in response to all odors, activation occurred in the anterior SMA also referred to as pre-SMA, with concurrent deactivation in the posterior SMA. The SMA is known for its role in motor planning and execution (Nachev et al. 2008; Makoshi et al. 2011). This region also overlaps with reward-related regions, enabling approach and avoidance behaviors (Hollmann et al. 2012). The anterior SMA tends to engage during the planning phase of movement, while the posterior portion becomes active during movement execution (Lee et al. 1999; Nachev et al. 2008). Our analysis approach modeled odor release as a 3-s event, with actual odor release lasting 2 s. This was done to account for any potential lingering of the odors beyond the duration of direct exposure. Although odor release was preceded by an orange crosshair indicating an imminent requirement to sniff an odor, part of the preparatory stage of the act of sniffing might have carried over into odor release. We can therefore speculate that observed anterior SMA activation was due to participants preparing to respond (i.e. sniff the odor) in the initial stages of odor release, while posterior SMA deactivation reflected sniffing inhibition once the odor release stage concluded and participants received the cue that sniffing was no longer required (white cross).

In addition to being involved in the processing of food rewards (Weltens et al. 2014), the ACC and putamen are also both involved in odor processing (Seubert et al. 2013). Both regions tend to exhibit greater activation in response to food odors compared with nonfood ones (Sorokowska et al. 2017). Moreover, while the putamen has been associated with encoding odor pleasantness (Torske et al. 2022), the ACC is likely involved in mediating odor–taste interactions in flavor perception (Small et al. 2004). Specifically, when a taste is perceived simultaneously with a retronasal odor, the ACC has been shown to activate, whereas when presented alongside an orthonasal odor, deactivation occurs (Small et al. 2004). This observation lends support to the idea that ACC involvement in our task might have been the result of orthonasal exposure to food-related odors. Given that fat odor-flavor associations can be learned (Sundqvist et al. 2006) and that the odors utilized here are typically experienced retronasally in combination with other modalities involved in dairy milk flavor perception such as taste, ACC deactivation might stem from learned cross-modal associations. The reason behind the putamen’s deactivation, as that of the STG—a region with no apparent relevance in the context of our study, remains to be elucidated.

All in all, brain activation within the ROIs does not directly reflect the perceptual findings of our current study. Not only are there no activation differences between the odors, but also odor exposure led to decreased activity in most identified regions. This seemingly contrasts with studies on oral fat perception (De Araujo and Rolls 2004; Grabenhorst et al. 2010; Eldeghaidy et al., 2011), although some also failed to find fat-concentration-related neural activation (Eldeghaidy et al., 2012; Stice et al., 2013) or have used ecologically implausibly high fat concentrations (De Araujo and Rolls 2004; Eldeghaidy et al. 2011). Our results do not align with earlier work on neural processing of odor intensity and valence. Specifically, Anderson et al. (2003) observed that perceived intensity correlates with amygdala and piriform cortex activation; Rolls et al. (2003) observed correlations between perceived pleasantness and activation in the OFC and ACC; while Winston et al. (2005) showed that the amygdala reflects perceived intensity of pleasant or unpleasant odors, but not neutral ones. However, these studies not only employed relatively small sample sizes and artificial odorants (e.g. anisole, citral acid, valeric acid, geranyl acetate), they were also not replicated. The possibility remains that our sample size was not sufficient to detect neural activation differences between the odors on an ROI level. Although our perceptual findings argue otherwise, we cannot authoritatively conclude that the absence of differences in this case means an absence of neural effects.

Whole brain analyses identified 4 additional potentially relevant brain activation clusters: in the fusiform gyrus, in the crus-1 region of the cerebellum, in the IX region of the cerebellum, and in the lingual gyrus. Both the lingual and fusiform gyri are visual areas known to respond to high calorie food cues, as highlighted by the meta-analysis of Yang et al. (2021). With repeated exposure, our participants might have formed associations between the visual cue signifying odor delivery (green crosshair) and M and H milk odors as high-calorie food cues. Similarly, although the cerebellum has traditionally been associated with motor and coordination control, the crus-1 region has been shown to respond to visual food cues (such as the green crosshair in our study) (Berman et al. 2013; Iosif et al. 2023). Lastly, the IX cerebellar region is part of the so-called default mode network (Stephen et al. 2018). Since default mode network areas deactivate during externally focused tasks, the observed deactivation in this region might have resulted from active engagement with the fMRI task (Menon 2023).

Assuming that discrimination ability in our experiment was underpinned by quality, rather than intensity differences between the samples, as suggested above, then it is reasonable to expect differential activation within the posterior piriform cortex, as shown by Howard et al. (2009), Gottfried et al. (2006), and Li et al. (2010). One could argue, however, that our analysis approaches were not fine-grained enough to detect differential brain activation resulting from subtle odor quality differences. Even though previous studies on oral fat perception have found differential brain activation in response to different fat concentrations using univariate approaches (Grabenhorst 2010; Eldeghaidy 2011), multivariate analysis techniques which focus on distributed patterns of voxel activity across regions rather than individual voxel activations, such as employed by Howard et al. (2009), might be more suitable to detect minor differences in physical and perceptual features than the current approach. Including additional perceptual ratings, especially those relating to fat-related odor quality differences (as in Howard et al. 2009 and Gottfried et al. 2006) would further refine the fMRI task.

Within the context of olfactory fat perception, the current study is the first of its kind and employed ecologically relevant odor stimuli. The lack of benchmark studies, however, makes the interpretation of our findings challenging. Nevertheless, the study serves as a foundation for subsequent work on the topic and yields several relevant considerations. To deepen our understanding of olfactory fat perception, it is essential to not only corroborate our findings but to explore fat perception using other olfactometer-compatible fat-related food sources. For instance, it might be interesting to compare vegetable oil emulsions varying in fat or oils/fats varying in origin (e.g. olive oil, sunflower oil, and lard). Considering that fatty acids can be discriminated using solely olfactory cues (Bolton and Halpern 2010; Kallas and Halpern 2011), it seems relevant to explore the neural correlates of olfactory perception of food sources with distinct fatty acid profiles. Moreover, given that retronasal olfaction plays a central role in food perception (Boesveldt and de Graaf 2017) and is arguably more ecologically relevant for food odors than orthonasal olfaction, future studies should consider an experimental design primarily focused on retronasal fat perception. Such an approach could reveal distinct neural pathways associated with retronasal fatty odor processing. The influence of hunger state on olfactory fat perception and discrimination and associated neural activity should be assessed as well. It is plausible that brain activation differences between food odors varying in fat might become more apparent in a depleted state, such as with protein (Griffioen-Roose et al. 2014). Lastly, it might be worthwhile exploring potential genetic predispositions or dietary influences on olfactory fat perception. The former appears especially intriguing. Not only were genetics identified as a factor in fat taste sensitivity (Running and Mattes 2016), but also our own observations across different experiments hint that genetics might play a role in olfactory fat perception. Specifically, a subset of participants could not distinguish between any fat levels, while the majority could—raw discrimination data from the current and past experiments is available on the Open Science Framework Repository (Pirc et al. 2021; Pirc et al. 2023).

5 Conclusion

Our work reaffirms the notion that food fat content can be distinguished using solely olfactory cues. This ability and underlying perceptual differences, however, were not reflected by brain activation in the current experiment. While this study has paved the way in understanding the neural underpinnings of olfactory fat perception, the absence of neural correlates to the perceptual results raises further questions and also highlights the complex dynamics between olfaction and brain responses. Moreover, it underscores the need for further neuroimaging studies and the use of more fine-grained analysis approaches such as pattern analysis, to characterize the brain processes underlying food odor processing in general and olfactory fat perception in particular.

Acknowledgments

The use of the 3T MRI facility has been made possible by Wageningen University Shared Research Facilities. During the writing process, ChatGTP 4.0 was utilized to facilitate readability and language. All generated content was reviewed and edited prior to inclusion in the manuscript. The authors take full responsibility for the content of the publication.

Author contributions

PM: Conceptualization, Methodology, Investigation, Data curation, Writing—Original Draft, Writing—Review & Editing, Project Administration; KC: Investigation, Data curation, Writing—Review & Editing; SP: Conceptualization, Supervision, Writing—Review & Editing; BS: Conceptualization, Supervision, Writing—Review & Editing, Project Administration.

Funding

This work was supported by the Division of Human Nutrition and Health of Wageningen University.

Conflict of interest statement

The authors declare no conflicts of interest.

Data availability

Parameter estimates extracted from relevant brain regions, along with perceptual data, are available on the Open Science Framework Repository, at https://osf.io/qwy65.