The standard methodology for determining the sensory quality of virgin olive oil (VOO) is the sensory assessment by a normalized methodology, the so-called panel test (IOC, 2015). However, results of the sensory assessment are questioned by members of the olive oil sector because of its poor repeatability when oils are submitted to different panel tests that sometimes result in the classification of the same olive oil indistinctly as extra-virgin (without sensory defects) or virgin (with slight defects) according the panel that evaluates it. Studies have been published on the variability of the panels when evaluating virgin olive oils (García-González et al., 2007).

Volatile compounds are directly responsible for aroma (“I smell therefore there are volatiles”) and they can produce desirable or undesirable sensory perceptions, later expressed by panelists as sensory descriptors. Thus, the ‘Gordian knot’ is the connection between volatiles and sensory descriptors that have to be expressed through standard panel tests, which are questioned by many members of the virgin olive oil sector.

The solution should come from a chemical analysis of volatile compounds. In order to obtain a full understanding of sensory quality from the composition of volatile compounds, a better knowledge of the olfaction process is needed, including the processes that occur after olfactory receptor activations. Gaining knowledge on this aspect entails studying the activity of the neurons that receive direct input from the olfactory bulbs responsible for the perception of virgin olive oil aroma. This approach would mean to explore the cognitive process related to olfaction tasks by using the complex aromas of different virgin olive oils as study material.

The high spatial and temporal resolution of functional magnetic resonance imaging (fMRI) has significantly contributed to the understanding of visual (Bedny et al., 2010), working memory (Kim et al., 2011) and olfaction tasks (Cerf-Ducastel and Murphy, 2001) among many others. The study of the neural response induced by olfaction involves, however, more difficulty than the study of other cognitive tasks (Qureshy et al., 2000). The high number of variables in the presentation of the odorants (e.g. time, concentration of odorant, carrier gas flow and humidity), together with the heterogeneity of the magnetic field in the primary olfactory cortex (POC) because of the near cerebral bone structures, partially explain this difficulty. Furthermore, the samples may be presented to induce an orthonasal and/or retronasal stimulation (Cerf-Ducastel and Murphy, 2001), which introduces an additional variability source. Therefore, the studies of the neural activities induced by olfaction require the use of an optimized system of odorant delivery to be reproducible throughout all the experiments (Tabert et al., 2007; García-González et al., 2011). In general terms, the functional studies of olfaction have described activations in the primary olfactory cortex (POC), which receives direct input from the olfactory bulb and it comprises piriform cortex, entorhinal cortex and periamygdaloid cortex. Other areas as orbitofrontal cortex, insula, thalamus, hippocampus, cerebellum and occipital lobe, including the visual cortex, have been pointed out as brain anatomic zones that are also involved in olfaction processes probably generating mental visual images for comestibility judgments (Cerf-Ducastel and Murphy, 2001). The activations registered in higher order olfactory areas are commonly robust unlike the activations in the primary olfactory cortex, which are inconsistent in some studies. This inconsistency may be due to artifacts or the habituation effect and it must be considered in the data treatment. In fact, it has been demonstrated that the primary cortex habituates very quickly, and it shows inconsistent activation when the stimulus time (ON period) is long enough (Poellinger et al., 2003). In addition, artifacts can be induced by susceptibility differences between brain tissue and the underlying bone and air, which result in signal loss from orbitofrontal cortices that have strong activations when subjects smell highly aversive odors (Royet et al., 2003).

Data from fMRI experiments are commonly analyzed with a contrast analysis assuming the hypothesis of a greater activity during a cognitive process compared to the rest state (Bifone et al., 2010). Thus, the stimulus is sequentially presented to a subject (ON period) alternating with a rest time (OFF period) according to a block design. In addition to assuming a lower brain activity in rest periods, another important assumption is that the timing at which the neural responses are registered matches the time specified in the paradigm (Tabert et al., 2007). Furthermore, the stimulus needs to be presented to the individual enough times to get a reliable statistical significance (Cerf-Ducastel and Murphy, 2004).

Despite the extensive studies on functional neuroimaging of olfaction, little is known about the modulation of brain activities in the particular case of complex aromas. Most of the studies are based on the delivery of substances that are strongly odorant, such as geranyl acetate (Rolls et al., 2003), lavender oil (Savic et al., 2000) etc. Few studies have been focused on food aroma produced by a complex mixture of volatile compounds producing a mild olfactory perception (Verhagen and Engelen, 2006). Furthermore, the information on the neural mechanism by which a food aroma results in a pleasant or unpleasant perception depending on the sensory quality is incomplete. In this respect, Royet et al. (2003) studied the brain activity of 126 odorants, some of which being food aromas. These odorants were classified as pleasant and unpleasant and some differences were found in the brain activities stemming from these two groups.

In a previous study we also addressed a study on brain activities when individuals smelled virgin olive oil from different qualities (García-González et al., 2011). However, a further step in this study is necessary to examine the different activities by means of a 2^nd level analysis (two samples t-test) that consider the variability among different groups of subjects. This study would allow confirming previous results centered on fMRI studies of virgin olive oil aroma (García-González et al., 2011) and would allow obtaining more detailed information between the activations caused by pleasant and unpleasant aromas. Thus, the aim of this paper is to study the brain activity in response of smelling the aroma of virgin olive oils from different sensory qualities and considering different individuals. Assuming the brain activity is different according to the pleasantness and unpleasantness of the aroma, it is expected to find differences in the neuroimages across different quality degrees of a virgin olive oil. The flavor of virgin olive oil is among the most studied from a sensory/chemical perspective (García-González et al., 2010), which provides a further advantage in the interpretation of fMRI results in regards to the olfactory perception and the volatile compounds that are responsible for the aroma.

The samples under study were subjected to a sensory assessment to determine the median of defects (Md) –unpleasant aroma- and the median of the fruity sensory attribute (Mf) –pleasant aroma-, thereby certifying their inclusion within the categories of extra virgin olive oil (EVOO) and lampante virgin olive oils (LVOO). Three EVOOs were qualified with a median of fruity attribute of 4–4.5 with a relative standard deviation (% RSD) of 6.9%. The three LVOOs, qualified with sensory defects (rancid, fusty and winey/vinegary), were qualified, as expected, with a high value of Md (2.9, 4.0, 7.7 respectively) with a % RSD of 12.9%.

The volatiles quantified in the samples are presented in Table 1, together with the odor activity value (OAV) and sensory descriptors. The OAV of each compound, expressed as the ratio between the concentration and the odor threshold, revealed that 20 volatiles contribute to EVOOs with pleasant aromas (OAV>1). Volatiles with high OAVs that contribute to pleasant sensory attributes – like fruity, sweet and green - are 3-methyl-butanal, pentanal, 1-penten-3-one, hexanal, E-2-hexenal and Z-3-hexen-1-ol while volatiles contributing with unpleasant aromas are some aldehydes such as hexanal (responsible for a fatty perception at a high concentration) (Morales et al., 2005), heptanal, octanal, E-2-heptenal, nonanal, E-2-decenal and E-2-undecenal, together with some alcohols (e.g. 2-butanol), ketones (e.g. 2-octanone) and all of the organic acids.

The most significant volatiles contributing to the unpleasant aroma depend on the kind of defect. Thus, high concentrations of aldehydes were particularly remarkable in the rancid LVOO, as a consequence of oxidation (Morales et al., 2005): hexanal, heptanal, E-2-heptenal, octanal, nonanal, 2,4-hexadienal, E-2-decenal and E-2-undecenal. The rancid LVOO also showed the highest concentrations of hexanoic and octanoic acids. The winey/vinegary LVOO showed a high concentration of acetic acid and ethyl acetate. The fusty LVOO was characterized with a high concentration of some aldehydes such as nonanal and, overall, for the high concentration of acids such as butanoic, pentanoic, hexanoic and octanoic acids. The high concentration of all these compounds responsible for sensory defects explained that no panellist disagreed in the description of the defects, and the samples were considered to be adequate as representative of unpleasant aromas in defective virgin olive oils for the fMRI study. Thus, it was expected that all the subjects smelling these off-flavors could identify the associated negative perceptions in contrast to pleasant aromas provided by extra virgin olive oil.

The aromas of the virgin olive oils were presented to the subjects during the fMRI experiments. The results of the two-sample t test study allow comparing human brain areas that are activated in an inter-subject study (Figure 2), unlike a 1^st level study, which is more appropriate for intra-subject studies (García-González et al., 2011). Table 2 shows several brain regions that were activated in the course of the experiments and they are split into left and right hemispheres because of their different locations (x y z MNI coordinates: mean±standard error of the mean) and the lateralization of the olfactory system. The activated regions (Table 2) included the piriform cortex, the amygdala, the entorhinal cortex, the parahippocampal gyrus, the orbitofrontal cortex (OFC), the insula, the cingulate gyrus and the cerebellum. Some of these areas are illustrated in Figure 2 as areas highlighted in the contrast analysis of pleasant vs. unpleasant aromas. The first three regions may be considered as parts of the primary olfactory cortex as they receive direct projections from the olfactory bulb.

A great number of previous studies on olfaction report strong activation in the heterogeneous orbitofrontal cortex (García-Falgueras et al., 2006) that we have also found with the bilateral activations of inferior and middle frontal gyrus (Table 2). Thus, Figure 3 shows the MRI signal of the orbitofrontal cortex averaged across 14 subjects after stimulation with virgin olive oil aroma. Also other regions are profusely described in the literature and found in this study such as the cerebellum, the insula and the cingulate gyrus (Poellinger et al., 2001). In contrast, the activations of the amygdala, the entorhinal, parahippocampal gyrus, middle temporal gyrus, precentral gyrus and posterior cingulated gyrus have scarcely been reported in olfactory studies, and these regions were activated by the odor of virgin olive oils.

The bilateral activation of the orbitofrontal cortex was found in different parts (inferior and middle frontal gyrus) either in the left or the right hemisphere. The lateralization of orbitofrontal activation in this study seems to relate to the place on a hedonic scale of the two kinds of samples that are placed at opposing extremes of the scale: pleasant and appreciated odors of EVOOs at the top of the scale, and high intensity of defective sensory descriptors of LVOOs at the bottom although the activation was more robust in the left hemisphere in the case of oils qualified with unpleasant aromas, which explains the activation observed in this region when comparing pleasant and unpleasant aroma smelled by subjects (Figure 2). This result agrees with Royet et al. (2003), who found a left-dominant network involved in the hedonic valence. Furthermore, Royet et al. (2004) also reported that the right orbitofrontal cortex is much more activated by familiar odors, which was also observed for the selected subjects who were habitual consumers of EVOOs. These authors also associated the activation of the left amygdala with the emotional processing of olfactory stimuli from unpleasant odors (Royet et al., 2003) which agrees with a strong left orbitofrontal activation at the individual level when subjects smell aversive odorants (rancid, fusty and winey/vinegary virgin olive oils).

In all the fMRI experiences, high activation values were registered on orbitofrontal, frontal and temporal lobes, in the BA 10, 11 and 20 (Table 3). Bilateral activations on BA 10 and 11 were associated with the olfaction process itself (García-Falgueras et al., 2006), which explains their activation in response to both pleasant and unpleasant samples. The activation of these areas in the orbitofrontal cortex is justified by the projection received by the olfactory primary cortex (Royet et al., 1999). In particular, Zatorre et al. (1992) found a high activation of right orbitofrontal cortex at BA 11 during olfaction. Royet et al. (1999) also observed an increment in cerebral blood flow at BA 11 (right medial frontal gyrus) during the smelling of a large set of odorants, and this area was clearly associated with the familiarity of the odors. The activation of this area by familiar odors would explain the high intensities observed in all cases, where subjects were regular consumers of virgin olive oils. The activation of BA 11 was more intense in most of the unpleasant aromas, probably because of the relation of this area as a consequence of unexpected outcome and emotional arousal (Yamasaki et al., 2002), which would explain the observed activation of this area when pleasant and unpleasant aromas were compared with a two-sample t test (Figure 2). This higher activation in unpleasant aromas is ultimately related to the occurrence of volatile compounds that normally are not present in good quality virgin olive oil (Table 1). These results reveal the importance of the orbitofrontal cortex in higher-order olfactory processing, supported by the association between physiological damages in this zone and anomalies in odor perception and identification (Zatorre et al., 1992).

Activation at the inferior temporal lobe (BA20) was also found in pleasant and unpleasant aromas. This activation was bilateral in some odors, while only one of the sides (right side) was activated in the others. The activation of this area was related to recognition and working memory tasks (Cutting et al., 2006), probably associated with the consumption habits of the subjects. It is important to note that subjects were asked to name the odors after the fMRI experiments, so it may cause the activation of those areas related to memory retrieval and naming. On the other hand, although fusty and winey/vinegary aromas might be unknown odors to the subjects, the rancid attribute is among the most commonly known unpleasant aromas. Thus, people are usually familiarized with a rancid odor produced by a rise in hexanal, nonanal and other aldehydes originating from oxidation (Table 1), and this defect may evoke negative memories for them.

The importance of memory integration and recognition tasks in the olfaction process was also pointed out by the maximum responses in pleasant odors. In the olfaction of pleasant aromas of virgin olive oils characterized by positive green and fruity aromas - due to the presence of volatiles as E-2-hexenal, hexanal at low concentration, and Z-3-hexen-1-ol, among many others - the maximum responses were mostly localized at the inferior frontal gyrus (BA 47) (Table 3, Figure 2). The activation of this area is also related to the recognition of familiar odors (Savic and Berglund, 2004; Royet et al., 1999). On the contrary, the maximum responses recorded for unpleasant odors (rancid, fusty and winey/vinegary virgin olive oils) were localized in other foci. Among them, in most samples the maximal response corresponded to inferior parietal lobule in the assigned BA 40, which is bilaterally activated. This area is activated during aversive feelings (Uher et al., 2005), which would explain its activation during the smelling of off-flavors in defective virgin olive oil. No activation of this area was observed when smelling good quality virgin olive oils where these compounds were not present or they were present at low concentration (e.g. acids). On the other hand, the lack of activation of the brain areas related to familiar odors may be involved in the cognitive process of unpleasant aromas since neophobia appears to be a key factor in hedonic responses (Rouby and Bensafi, 2002). Thus, Royet et al. (1999) found a high correlation between familiarity and pleasantness in the smelling of odors.

Other areas related to negative feelings were activated in most of the unpleasant aromas, with no activation in pleasant samples. Thus, superior temporal gyrus was activated in BA 38, in the left side or bilaterally (Table 3, Figure 2). The location of this area explains that it participates in emotion processing (Pelletier et al., 2003). In fact, this activation could point out aversion or dislike, according to previous studies that have related this area with negative emotions (Eugéne et al., 2003). Kosslyn et al. (1996) also observed more activation in left middle temporal gyrus when perceiving aversive stimuli in comparison to neutral stimuli.

Intense activations were also found in the anterior cingulate gyrus when virgin olive oils with off-flavors were presented to the subjects. The activated voxels comprised the Brodmman areas numbered 24, 32 and 33 (Figure 2). The activations in the anterior cingulate gyrus are related to familiar odors (Royet et al., 1999). As these areas are within the limbic system; they are associated with emotions (Britton et al., 2006). A previous work has shown that the activation of anterior cingulate may depend on whether the emotion is social or non-social. Thus, Britton et al. (2006) observed activations of anterior cingulate in negative emotions when they were affected by a social component (e.g. sadness). Nevertheless, the effect of the social dimension on negative and positive emotions seems not to be clear and it requires further studies. The interaction between sociality and emotion makes the cingulate area more difficult to interpret. Concerning olfaction, activation in the anterior cingulate is detected during odor naming, which involves a semantic work (Qureshy et al., 2000) and during the smelling of familiar odors (Royet et al., 1999). However, the activation in this area observed during the presentation of defective virgin olive oils seems to be related to the averse, negative emotions produced by those volatiles that were present in defective oils at high concentrations (Table 1), rather than familiarity of odors, since no activation was observed in most of normal virgin olive oils. In particular, some areas in the anterior cingulate gyrus (e.g. BA 32) are clearly involved in negative emotions produced by aversive olfactory stimuli. Thus, Fulbright et al. (1998) detected significant activation in BA 32 (lateralized to left) in response to a compound responsible for an unpleasant aroma (isovaleraldehyde). The activation of BA 32 may be strongly affected by emotion since it is a zone richly connected with the limbic system (Fulbright et al., 1998), which could explain something which is not distinctly associated with off-flavor. The effect of emotion and familiarity of odors in the activation of brain areas involved in olfaction explains the difficulty in interpreting the hedonic response to odors (Rouby and Bensafi, 2002). In order to check the hypothesis that the activation of the cingulate area in the BA 32 is due to an unpleasant perception, two volatile compounds characterized with undesirable aromas, heptanal and propanoic acid were presented to the subjects. These compounds were diluted in odorless refined oil up to a concentration of 100 mg/kg. Both volatiles are produced during the oxidation and fermentation processes and they are commonly present in virgin olive oils with sensory defects (Morales et al., 2005). The concentrations quantified in the defective samples for these compounds were higher than their odor thresholds (OAV>1, Table 1) and, in consequence, they were contributors to the aroma perceived by the panelists. In particular, the contribution of heptanal to the aroma is significant in rancid oils, where the saturated aldehydes originated during oxidation are mainly responsible for the rancid and fatty attributes (Morales et al., 2005). On the other hand, propanoic acid was a volatile marker of the fusty sensory defect and it is also typically found in oils with other sensory defects. During these fMRI experiments all the subjects described the aroma of the dilution of these compounds as unpleasant and an intense activation was observed in BA 32 (Figure 4). A high activation was also detected in BA 38, as it was previously observed in defective virgin olive oil samples and in the two-sample t test study (Figure 2).

In addition to areas that are activated in response to off-flavor, other brain areas located on the medial frontal lobe seem to act as modulators in the cognitive processing of odors although they are not activated consistently in all the samples. Thus, strong odors (both pleasant and unpleasant) elicited an intense activation in BA 6 (Figure 2). In most of cases, the activation of this area was observed in the defective samples (e.g. rancid and fusty oils) and in those good quality samples characterized with an intense green-fruity aroma. Thus, Table 1 shows that 29 volatile compounds in defective samples had OAV>1 compared to only 21 compounds in extra virgin olive oil, which explains the higher intensity in aroma of the defective samples. The activation of this area in the left hemisphere is associated with strong odors, as was reported by Miyanari et al. (2007) who did not observe activation when weak odorants were presented to the subjects. Although this area is also activated by other tasks such as object naming (Hirsh et al., 2001), their activation may modulate the pleasantness of the aroma, since the perceived intensity has an important role in rising/diminishing pleasantness.

The results obtained after presenting the aroma of pleasant and unpleasant VOOs show that the hedonic component of olfactory perception is explained by the combined effect of three groups of brain areas. The first group (BA 9, 10, 11, 13, 20, 47) is activated in both pleasant and unpleasant samples and their activations are clearly related to the olfaction process itself. A second group of BAs (mainly 32 and 38) points out adversity and they are activated in most of the unpleasant samples. Finally, the third group of brain areas would act as modulators of the perception indicating strength of the aroma (BA 6) and familiarity (BA 9, 40, 47).

The results, which are being checked with more individuals and other kinds of virgin olive oils, will help in the building of a complete neuroscientific model to identify problems in panelists evaluating virgin olive oils and to understand consumer preferences for some particular varietal oils, integrating multimodal features of oil perception such as taste (bitterness, pungency, astringency) and color (Verhagen and Engelen, 2006). This procedure is also useful to verify the importance of volatile markers in producing off-flavor and therefore validating their utility as quality markers.