Abstract
Sensory systems enable organisms to detect and respond to environmental signals relevant for their survival and reproduction. A crucial aspect of any sensory signal is its intensity; understanding how sensory signals guide behavior requires probing sensory system function across the range of stimulus intensities naturally experienced by an organism. In olfaction, defining the range of natural odorant concentrations is difficult. Odors are complex mixtures of airborne chemicals emitting from a source in an irregular pattern that varies across time and space, necessitating specialized methods to obtain an accurate measurement of concentration. Perhaps as a result, experimentalists often choose stimulus concentrations based on empirical considerations rather than with respect to ecological or behavioral context. Here, we attempt to determine naturally relevant concentration ranges for olfactory stimuli by reviewing and integrating data from diverse disciplines. We compare odorant concentrations used in experimental studies in rodents and insects with those reported in different settings including ambient natural environments, the headspace of natural sources, and within the sources themselves. We also compare these values to psychophysical measurements of odorant detection threshold in rodents, where thresholds have been extensively measured. Odorant concentrations in natural regimes rarely exceed a few parts per billion, while most experimental studies investigating olfactory coding and behavior exceed these concentrations by several orders of magnitude. We discuss the implications of this mismatch and the importance of testing odorants in their natural concentration range for understanding neural mechanisms underlying olfactory sensation and odor-guided behaviors.
Significance Statement
Understanding how environmental signals guide behavior requires studying sensory system function at natural stimulus intensities. In olfaction, experimental studies typically present olfactory stimuli at empirically chosen concentrations without reference to those encountered naturally. We review the concentrations of odorants measured in diverse natural settings in comparison with those used in experimental studies in rodents and insects, as well as with detection thresholds measured in rodents. This comparison reveals that most experimental studies in olfaction use odorant concentrations several orders of magnitude higher than their occurrence in most natural settings. Additionally, our analysis suggests that odorants are often encountered at concentrations near or below their detection threshold. We discuss the implications of these conclusions for understanding the neurobiology of olfactory function.
Introduction
Establishing a range of odor concentrations in natural settings
Understanding the neural basis of sensory perception requires describing how sensory neurons represent external stimuli, how neural circuits transform these representations, and how ensuing neural activity patterns relate to sensory-guided behavior or perceptual measures. A key determinant of these processes is the physical intensity of a sensory stimulus. Exploring the representation and processing of stimuli across the range of intensities experienced by an organism in natural environments has been fundamental to understanding the neural mechanisms underlying sensation. In vision and audition, for instance, referencing experimental results to stimulus intensities from specific environmental settings is common practice, as it allows experimental results to be interpreted in a more meaningful context (Mennitt et al., 2014; Faust and Logan, 2018).
Studies of the olfactory system stand out as an exception to this approach, in that the physical intensities (i.e., concentrations) of olfactory stimuli are not often—and not easily—referenced to those found in an animal's natural environment. Indeed, olfaction may be the only sensory modality in which experimental studies routinely use stimulus intensities with no stated relationship to those occurring naturally. To be fair, there are good reasons for this lack of context. First, experimentalists in olfaction face a daunting sampling problem. The olfactory system is charged with detecting a vast and extremely diverse array of chemical stimuli (i.e., “odorants”) that occur in natural sources as complex mixtures. Odor detection occurs via large repertoires of differentially sensitive olfactory receptor proteins, each of which defines the specific response profile of individual olfactory sensory neurons (OSNs); the number of receptor genes ranges from dozens in insects to over a thousand in rodents and other mammals (Niimura and Nei, 2006; Robertson, 2019). As a result, the probability of activating a given receptor using an arbitrarily chosen monomolecular odorant is low but increases at higher stimulus concentrations (as noted below).
Second, knowing what stimulus concentrations occur in natural odor environments is a challenge in itself. Odorant concentrations can be expected to vary over a vast range that depends on the chemical properties of each odorant compound and the sources from which they derive. Moreover, in both air and water, the propagation of olfactory stimuli from their source occurs largely by bulk flow in a turbulent fluid, leading to complex odorant concentration profiles that vary over time and space (Moore and Crimaldi, 2004; Riffell et al., 2008; Celani et al., 2014). This complexity makes accurate extrapolations of odorant concentrations even a small distance away from the source very challenging (Yu et al., 2010).
Additionally, directly measuring odorant concentrations in natural settings—especially airborne ones—is extremely difficult. Whereas the intensity or frequency of sound or light is easily measured with submillisecond resolution using inexpensive electronics, identifying which compounds are present in a mixture and at what concentrations requires complex and expensive equipment such as a gas chromatograph and mass spectrometer, which generally provide information with low temporal resolution (Capelli et al., 2013).
Neural mechanisms underlying olfactory sensation depend on intensity
As with any sensory modality, stimulus intensity is a crucial determinant of how olfactory information is represented by OSNs and processed by the brain (Mainland et al., 2014). Indeed, nearly all aspects of olfactory function are impacted by stimulus intensity (i.e., odorant concentration). First, odorant concentration impacts the nature of odor identity coding. The canonical view from experiments using odorant concentrations in the range of single parts per million (ppm) or higher is that odor identity is encoded combinatorially across OSN subtypes with distinct but overlapping molecular receptive ranges (Malnic et al., 1999; de Bruyne et al., 2001; Kajiya et al., 2001; Hallem and Carlson, 2006). Experiments using odorants at concentrations in the ppm range suggest that OSNs are activated by structurally diverse odorants, and shared physicochemical features of odorants are not reliable predictors of OSN co-tuning (Hallem and Carlson, 2006; Pelz et al., 2006; Kreher et al., 2008; Ma et al., 2012; Chae et al., 2019). In contrast, studies presenting odorants at concentrations that are several orders of magnitude lower—in the range of parts per billion (ppb)—have found that OSNs are narrowly tuned to a few structurally similar compounds and that individual odorants are represented by sparse patterns of activation across the OSN population that tend to reflect relatively straightforward chemical features of olfactory stimuli (Bhandawat et al., 2007; Kreher et al., 2008; Zhang et al., 2013; Cichy et al., 2019; Burton et al., 2022).
This dependence of odor representations on stimulus concentration is obvious at the first synaptic level in the CNS, where OSNs expressing the same receptor converge onto the same glomerulus in the olfactory bulb or antennal lobe, thereby generating a map of odorant receptor identity (Mori et al., 1999; Fulton et al., 2024). Here, increasing odorant concentration substantially increases the number of excited glomeruli (Fig. 1A) and broadens their tuning cross chemical space (Fig. 1B; Sato et al., 1994; Wachowiak and Cohen, 2001; Hallem and Carlson, 2006; Kreher et al., 2008; Zhang et al., 2013; Hu et al., 2020). In this higher concentration regime, odorant physicochemical features are poor predictors of the similarity of glomerular activation patterns or the location of activated glomeruli (Ma et al., 2012; Chae et al., 2019). Thus, increases in concentration lead to a qualitative change in the population-level representation of an odor.
Impact of odorant concentration on primary odor representations. A, Maps of olfactory sensory neuron (OSN) responses as measured using GCaMP6s in glomeruli across the mouse olfactory bulb, evoked by three concentrations of the same odorant (methyl tiglate), spanning four orders of magnitude. Responses are from the same mouse. Arrows in bottom panel indicate most-sensitive glomerulus (see Burton et al., 2022 for experimental details). ppb, parts per billion; ppm, parts per million. B, Tuning curves of olfactory sensory neuron (OSN) inputs to single OB glomeruli evoked by perithreshold (∼0.2–10 ppb) versus suprathreshold (∼0.2–10 ppm) concentrations of the same 59 odorants. Data from S. Burton, M. Wachowiak.
Other features of odorant-evoked neural response patterns emerge or are amplified at higher odorant concentrations—these include adaptation at the level of OSNs (Duchamp-Viret et al., 2000; Lecoq et al., 2009; Conway et al., 2023), a loss of coherence with the respiratory cycle (Moran et al., 2021), and persistent firing following the offset of odorant (Matsumoto et al., 2009; Patterson et al., 2013). Concentration likely also impacts how neurons represent natural odor mixtures. At the level of OSNs, interactions between components in a mixture have been proposed to enhance the coding of natural odors by generating unique, mixture-specific patterns of activity across the OSN population. However most studies have delivered individual odorants in the ppm concentration range (Duchamp-Viret et al., 2003; Lovitz et al., 2012; Zak et al., 2020) or, when delivered in liquid phase, at even higher concentrations (Xu et al., 2020). Mixture interactions may be less widespread at lower concentrations, and individual components may sum more linearly in generating representations of natural odor objects, at least in the periphery (Lin et al., 2006; Davison and Katz, 2007; Schubert et al., 2014).
Olfactory stimulus intensity also strongly impacts the function of neural circuits underlying the central processing of olfactory information. In the olfactory bulb and antennal lobe, the first sites of central olfactory processing, feedforward inhibition shapes the response of projection neurons to odorant-evoked sensory input in an intensity-dependent manner; this inhibition is hypothesized to filter responses to weaker inputs driven by lower-intensity stimuli (Gire and Schoppa, 2009; Cleland and Linster, 2012; Barth-Maron et al., 2023). At the same time, feedback inhibition regulates transmitter release from OSNs, resulting in intensity-dependent gain control of sensory inputs (McGann et al., 2005; Murphy et al., 2005; Olsen and Wilson, 2008; Root et al., 2008). Circuits mediating inhibition between glomeruli also profoundly shape odor representations in the olfactory bulb and antennal lobe, and the strength of this lateral inhibition depends on input intensity and odorant concentration (Yokoi et al., 1995; Arevian et al., 2008; Olsen et al., 2010; Cleland and Linster, 2012; Banerjee et al., 2015; Hong and Wilson, 2015; Economo et al., 2016). Even the monosynaptic transmission of excitatory input from OSNs to second-order neurons is nonlinearly dependent on input intensity. Weak stimulus inputs are amplified by the high convergence ratio of OSNs at the level of the glomerulus, while strong inputs (i.e., high-intensity stimuli) drive depletion of neurotransmitter release via synaptic depression (Murphy et al., 2004; Kazama and Wilson, 2008). Thus, odorant concentration influences how neural circuits receive and process incoming olfactory information.
Finally, odorant concentration can alter olfactory perception and odor-driven behaviors. In humans, perceptual descriptors associated with both odorant “quality” as well as hedonic value (i.e., “pleasantness”) change with concentration (Moskowitz et al., 1976; Laing et al., 2003). In rodents and insects, innate behavioral responses to odorants tend to transition from attraction (or no response) to avoidance with increasing concentrations (Stensmyr et al., 2003; Kreher et al., 2008; Horio et al., 2019; Manoel et al., 2021; Lopes et al., 2023; Ray et al., 2023).
All of these examples point to the importance of referencing experimental approaches to an appropriate concentration range. Probing olfactory function at unnaturally high (or low) stimulus intensities risks mischaracterizing how neural circuits represent and transform sensory inputs and may motivate theoretical models to explain neural or behavioral responses to odorant concentrations that an organism may rarely, if ever, experience. Including a naturalistic intensity range within experimental studies allows for concentration-dependent phenomena to be examined in a stimulus regime that the nervous system evolved to process. Further, referencing odorant concentrations to particular environmental settings, when possible, is essential for interpreting neurophysiological and behavioral results in an appropriate ethological context.
Comparing odorant concentrations in experimental neuroscience and natural settings
In contrast to the ease with which datasets for naturally occurring light and sound intensities in different settings are accessed (Mennitt et al., 2014; Faust and Logan, 2018; DecibelPro, 2024; EngineeringToolBox, 2024), similar resources for odorant concentrations are not readily available. Defining naturalistic concentration ranges for odorants is difficult given the tremendous diversity of odorant compounds arising from natural sources and the variety of settings or natural situations in which these might be encountered.
One approach to this problem is to define particular settings relevant to olfaction, which include the extreme ends of odorant concentrations an animal might be expected to encounter and measure odorant concentrations in each setting. A breakdown of different settings for olfaction is illustrated in Figure 2 and includes ambient concentrations of odorants in an organism's “typical” environment, concentrations encountered in close proximity to or in direct contact with an odor source (“direct sampling”), and concentrations encountered at a distance from the source. We also separately consider odorant concentrations in the headspace of a container enclosing a natural source, since these have been measured from many sources. Data regarding odorant concentrations in each of these settings exist but originate from disparate fields including food chemistry, atmospheric science, and environmental safety and, to our knowledge, have not been compiled into formats that can be compared directly.
Different settings for defining relevant ranges of odorant concentrations.
In the sections that follow, we first survey the range of odorant concentrations used in experimental studies of olfaction, focusing on widely used animal models (rodents and insects), and then review studies measuring odorant concentrations in different settings of relevance for these organisms. To provide additional context, we also survey literature reporting psychophysical measures of odorant sensitivity in rodents. After presenting the results of these surveys, in the final section of this review we discuss their implications for considering the neuroscience of olfaction in a more naturalistic context.
Odorant Concentrations Used in Experimental Olfactory Neuroscience
To evaluate the range of odorant concentrations used in laboratory studies of olfactory function in terrestrial animals, we surveyed experimental studies in rodents and insects, two commonly studied groups in experimental olfaction. We began by reading all primary studies that include odor stimulation cited in two exceptionally comprehensive reviews of odor coding in mammals and insects (Galizia and Sachse, 2010; Fulton et al., 2024). We supplemented this list with several more recent studies, selected for their use of large, well-defined odor panels or high level of citations.
Our survey revealed, first, that absolute stimulus concentrations are often undefined in olfactory experimental literature: approximately one-third of rodent studies and ∼80% of insect studies either did not report delivered odorant concentration or did not deliver odorants in a manner that allowed its estimation (in vapor phase). Ultimately our analysis of experimental odorant concentrations considered 51 studies, 34 of which were from rodents and 17 from insects.
Second, our survey showed that odorant concentrations used in experimental studies vary tremendously, ranging over ∼7 orders of magnitude, both within and across studies (Fig. 3A). Some of this variation is intentional (i.e., odorant concentration is used as an independent variable), but much of the variance in odorant concentrations reflects differences in vapor pressure across odorants that were not factored into the experimental design. Indeed, only 8 of 51 studies attempted to deliver different odorants at the same concentration in vapor phase. Overall, however, experimental odorant concentrations were in the range of parts per million in air (ppm) or higher: the median odorant concentration across stimuli exceeded 1 ppm in 45 of 51 studies, with 28 studies exceeding 10 ppm and 11 exceeding 100 ppm. We next asked how these concentrations compare with those found in different natural settings of relevance for rodents and insects.
Comparison of odorant concentrations in experimental studies, headspace of natural sources, and detection thresholds in rodents. A, Left, “experimental odor concs”, Delivered odorant concentrations in 51 experimental studies. Shading indicates insect (light) or rodent (dark) study. Middle, “source concs”, Distributions of 90th percentile odorant concentrations in the vapor phase (“headspace”) or in the source itself (“in-source”), across sources, per odorant. Right, “threshold”, Distribution of odorant detection thresholds in rodents. All box plots show median (red) and quartiles, whiskers indicate 10–90th percentile. See text for details. List of studies as indexed in plot: 1 (Markopoulos et al., 2012); 2 (Ogg et al., 2018); 3 (Yamada et al., 2017); 4 (Medinaceli Quintela et al., 2020); 5 (Gadziola et al., 2015); 6 (Devaud et al., 2001); 7 (Wang et al., 2020); 8 (Pashkovski et al., 2020); 9 (Barnes et al., 2008); 10 (Boyd et al., 2015); 11 (Chapuis and Wilson, 2013); 12 (Rothermel et al., 2014); 13 (Bhandawat et al., 2007); 14 (Petzold et al., 2009); 15 (Rennaker et al., 2007); 16 (Chae et al., 2022); 17 (Bolding and Franks, 2017); 18 (Perez-Orive et al., 2004); 19 (Parnas et al., 2013); 20 (Otazu et al., 2015); 21 (Scott et al., 2014); 22 (Ma et al., 2012); 23 (Wilson and Laurent, 2005); 24 (Poo and Isaacson, 2011); 25 (Schoonover et al., 2021); 26 (Iurilli and Datta, 2017); 27 (Davison and Katz, 2007); 28 (Miura et al., 2012); 29 (Doucette et al., 2011); 30 (Shang et al., 2007); 31 (Johnson et al., 2002); 32 (Tan et al., 2010); 33 (Singh et al., 2019); 34 (Chae et al., 2019); 35 (Shen et al., 2013); 36 (Soucy et al., 2009); 37 (A. Dehaqani et al., 2024); 38 (Moran et al., 2021); 39 (Wilson et al., 2004); 40 (Schlief and Wilson, 2007); 41 (Olsen and Wilson, 2008); 42 (Wang et al., 2004); 43 (Esquivelzeta Rabell et al., 2017); 44 (Pelz et al., 2006); 45 (Nagappan and Franks, 2021); 46 (Endo et al., 2020); 47 (Tully et al., 1994); 48 (Frechter et al., 2019); 49 (Illig and Haberly, 2003); 50 (Nakayama et al., 2022); 51 (Burton et al., 2022a). B, Difference between experimental and maximal headspace concentrations measured by PTR-MS, plotted for each compound and expressed as the log of the ratio between each experimental concentration and the maximal headspace concentrations (90th percentile across sources) measured in the PTR-MS study survey (see text). Positive values indicate higher experimental concentrations relative to headspace concentrations. C, Difference between rodent detection thresholds and maximal headspace concentrations, plotted for each compound as in B. Positive values indicate higher headspace concentrations relative to detection threshold. Each point indicates threshold data from a single study. Inset shows cumulative probability distribution of log ratios. List of studies included the following: Davis (1973); Pierson (1974); Walker and O’Connell (1986); Slotnick and Schoonover (1993); Bodyak and Slotnick (1999); Youngentob and Margolis (1999); Larson et al. (2003); Kraemer and Apfelbach (2004); Vedin et al. (2004); Clevenger and Restrepo (2006); Joshi et al. (2006); Laska et al. (2006, 2009); Smith et al. (2008); Larsson and Laska (2011); Can Güven and Laska (2012); Løtvedt et al. (2012); Wallén et al. (2012); Sarrafchi et al. (2013); Laska (2014); Sarrafchi and Laska (2017); Dewan et al. (2018); Peixoto et al. (2018); Blount and Coppola (2020); Williams and Dewan (2020); Jennings et al. (2022); Williams et al. (2024).
Odorant Concentrations in Different Natural Settings
We surveyed a diverse body of literature to assess the range of odorant concentrations encountered in different environmental settings, as described above. Table 1 shows concentration ranges of characteristic odorants associated with select sources, or in select environmental settings, to provide touchstones for the reader; details of these surveys are described in the subsections below. A more comprehensive reporting of survey results is presented in the Appendix.
Ambient odorant concentrations
Studies from the field of atmospheric science have measured the concentration of biogenic volatile organic compounds (VOCs) in various environments, helping to constrain the range of ambient, “background” odorant concentrations experienced by animals like mice or flies in a natural setting. For example, terpenes are the most abundant class of biogenic VOCs emitted by plant sources and their emission has global-scale environmental impacts (Curci et al., 2009; Guenther et al., 2012). Measurements of ambient levels of terpenes and other VOCs have been made in different settings of relevance for rodents or insects, including forests and fields of cultivated crops (Antonelli et al., 2020). Biogenic VOC concentrations have also been measured in “non-natural” high-intensity environments such as industrial-scale greenhouses and animal processing facilities, which could be used as a proxy for the high end of ambient odorant concentrations expected to arise from natural sources. In each of these environments, overall concentrations of individual volatiles are well below the parts per million range. For example, in forests of monoterpene-emitting tree species (i.e., oak, pine, tropical tree spp.), levels of total monoterpenes range from <1–5 parts per billion (ppb), with occasional peaks up to 30 ppb (averaged over intervals ranging from 5–30 min; Antonelli et al., 2020); concentrations of the most abundant single monoterpene typically account for 30–50% of this total (Saxton et al., 2007; Mielke et al., 2010; Křůmal et al., 2016; Bach et al., 2020; Mermet et al., 2021). Other highly emitted compounds (e.g., methyl vinyl ketone and methacrolein) range from <1 to 3 ppb (Kesselmeier et al., 2000; Mielke et al., 2010). Even in high-intensity environments like greenhouses and commercial cultivation facilities for cannabis plants, which are high terpene emitters (de Ferreyro Monticelli et al., 2022), maximal ambient concentrations of the most abundant terpenes range from ∼60 to ∼500 ppb (Urso et al., 2023).
Animal waste (i.e., feces, urine) is another important natural odor source, comprising different chemicals from those emitted by plants. While we are not aware of reports of ambient levels of waste-related volatiles in natural settings (e.g., a rodent nesting area), such measurements have been made in anthropogenic settings—for example, in commercial animal production facilities (Schiffman et al., 2001; Ni et al., 2012; Sintermann et al., 2014). While these sites represent the high end of concentrations for animal-related compounds (i.e., those creating potential health and environmental concerns), ambient concentrations of the majority of waste-related volatiles are still generally in the single ppb range. Even in these cases, only a few biogenic VOCs are reported at levels above 10 ppb; these tend to be highly volatile compounds like hydrogen sulfide, short-chain carboxylic acids, and small amines (Tables 1 and 2).
Concentration ranges of select volatiles in natural and high-intensity ambient settings and in the enclosed headspace of various sources
Peak ambient concentrations of select volatiles in natural and high-intensity settings
Overall, these data suggest that ambient concentrations of most volatiles encountered by the olfactory system in common natural settings are present at several orders of magnitude below the concentrations used in the majority of experimental studies. Examples from aquatic olfaction suggest a similar conclusion: ambient concentrations of free amino acids in aquatic environments are reported to range from ∼10−9 to 10−7 M (Jørgensen, 1987; Uhrman, 2006), while at least some experimental studies present amino acid stimuli at concentrations ranging from 10−6 to 10−4 M (Caprio and Byrd, 1984; Friedrich and Korsching, 1998; Junek et al., 2010).
Concentrations at an odor source
While ambient VOC levels provide a common working range for olfaction, animals often sample odors at close range to an odor source: rodents, for example, often sniff directly at the surface of an odor object, while flies may make odor-based decisions while walking on the object itself. Estimates of odorant concentrations in this regime can effectively define an upper limit to odorant concentrations an animal could reasonably be expected to encounter in nature. We are not aware of direct measurements of vapor phase odorant concentrations at the surface of odor sources, although these concentrations could be estimated from measurements of the emission (flux) rates of VOCs, which have been reported for a variety of relevant sources including soil, many species of flowers, and plants that are rodent food sources (Davison et al., 2007; Abis et al., 2018; Farré-Armengol et al., 2020; Sarkar et al., 2020).
An alternative approach is to measure the volatiles in the air surrounding an odor source in an enclosed volume (without air exchange) at equilibrium (i.e., “headspace analysis”; Liberto et al., 2019). Such measurements have been made for many odor sources, both natural and human-manufactured, and can be a useful resource for assessing the composition and concentrations of individual compounds emitted by a source. While these measurements do not reflect naturalistic concentrations (animals do not typically encounter odorants in an enclosed, vapor-saturated headspace), they can be used to estimate a near upper limit to the concentrations an organism might encounter in close proximity to an odor source.
To assess the range of headspace concentrations from natural sources, we first surveyed studies using direct-sampling techniques (i.e., PTR-MS), which measure volatile concentrations directly in their native vapor phase (Lindinger et al., 2005). We began by reviewing all references included in a comprehensive review of all PTR-MS analyses of food sources from 1995 to 2022 (Mazzucotelli et al., 2022) and selected all studies that reported quantitative measurements of identified compounds in the headspace of natural sources, excluding human-processed or heated sources such as baked goods, beer and liquors. We included additional studies published since 2022 or arising from nonfood sources, including animal urine and soil. In total we surveyed 20 studies analyzing headspace concentrations of 383 compounds from 24 sources of relevance for olfaction, including various fruits, nuts, and plants (including flowers and spices); animal products including milk, cheese, and honey; animal (canine) urine; and soil (Table 3).
Analysis of headspace volatile concentrations from example natural sources measured with PTR-MS
In any given source, headspace concentrations of individual compounds were in the low-ppb range (range of median concentrations per source, 1–37 ppb; range of 75th percentile of concentrations; 2–507 ppb). Even the maximal headspace concentration for a given compound across all sources is still in the ppb range (median, 6.0 ppb; 10th–90th percentiles, 0.3–146 ppb, n = 383 unique compounds), with only 4% of compounds present at concentrations above 1 ppm (Fig. 3A). These compounds tended to be low-molecular weight, high-volatility compounds that are common products of plant metabolism or bacterial action (Table 4).
High-concentration outliers in headspace of natural sources, as measured by PTR-MS
To assess the range of odorant concentrations across a broader array of sources, we analyzed a large compilation of (human) food headspace profiles (Volatile Compounds in Foods database, v16.10; VCFonline, 2024). The database includes the volatile chemical composition of 1,260 food products compiled from >4,500 literature references. Measurements from human food sources allow estimation of the range of odorant concentrations likely to be important to (nonhuman) animals, especially human commensals like domestic mice, rats, and fruit flies. Restricting the survey to only natural food sources (i.e., excluding cooked or processed foods) yielded a final dataset containing 1,942 unique compounds measured from 479 food product sources spanning nine major food groups (see Appendix for details). Plant sources—fruits (∼33%) and vegetables (∼14%)—were most heavily represented.
Values in the VCF database consist of the estimated concentration of each volatile within the source (typically this is determined by calibration to known concentrations of the identified volatile; Liberto et al., 2019). These in-source concentrations thus represent an extreme upper limit of what might be encountered in natural olfaction—i.e., through direct contact with the source (usually, an artificially homogenized source). In the case of insects, this might be achieved by antennal touch or by larvae crawling through liquefied food (Kim et al., 2017); in the case of rodents, this would likely never occur as it requires inhalation of the source itself. Nonetheless we compared in-source concentrations to those in our experimental survey, on an odorant-by-odorant basis. To avoid the long tail of low concentrations for compounds present at trace levels in certain sources, we compared each instance of experimental concentration (Cexp) to the 90th percentile of all in-source concentrations (C90) for the same compound (compounds measured in fewer than 10 sources were excluded). The analysis yielded comparisons for 228 unique compounds and included 1,206 instances in experimental studies.
Notably, the overall range of in-source concentrations was similar to that of vapor-phase concentrations delivered in experimental studies, with a median value of 4.1 ppm (Fig. 3A). Over half of all experimental instances of an odorant were higher than the C90 for that odorant in natural sources; ∼20% of Cexp values were >100 times higher than the in-source C90. We also estimated the vapor-phase, headspace concentrations of odorants from natural sources using their vapor pressures and reported in-source values (see Appendix Figure 5 for details). The range of estimated headspace concentrations aligned well with directly measured headspace concentrations, with the large majority of compounds estimated in the low ppb range (Fig. 3A); the high end (C90) of headspace concentrations for an odorant across all sources ranged from 0.000048–180 ppb (10th–90th percentile), with a median of 0.084 ppb. Even the total concentration of all headspace volatiles emitted by a source, while variable, was below 1 ppm for most sources (median, 322 ppb; 10th–90th percentile, 0.02 ppb–1,724 ppm).
Odorant concentrations in naturalistic settings: key takeaways
The totality of these comparisons is sobering. First, they suggest that for most natural sources, and in nearly any plausible natural setting, individual odorants are encountered at concentrations in the range of a few or few tens of parts per billion. This perhaps surprising finding is bolstered by the agreement between odorant concentrations measured using independent methods derived from diverse fields such as environmental chemistry and food science. Second, they indicate that most experimental studies present odorants at concentrations exceeding their naturally occurring range by several orders of magnitude (Fig. 3B), and, in fact, often present odorants at concentrations higher than what is found within the source itself.
Are there any environmental settings yielding odorant concentrations that might be matched to the ppm-or-higher range of concentrations that are most heavily represented in experimental studies? The only settings where we found multiple compounds occurring at ppm or higher levels involved decomposing animal or plant material in an enclosed space—for example, a composting bioreactor (Schiavon et al., 2017) and in the headspace of rotting seafood (Kim et al., 2009; Table 1). Overall, odorant concentrations exceeding 1 ppm in the vapor phase represent an extreme high end of the naturally occurring range for most odorants, with such levels often associated with aversive or toxic environments.
Referencing Olfactory Stimulus Intensities to Detection Thresholds
Another touchstone for the relevant concentration range of an odorant is to consider its behavioral detection threshold. Human psychophysics studies often characterize the abundance of individual components of an odor source in units of multiples of the detection threshold (i.e., “odor activity value”; Grosch, 1994); expressing odorant concentrations in such terms gives an intuitive sense of the strength of an odorant relative to a well-defined perceptual measure. Having an analogous reference for odorants presented in experimental studies in animal models would be useful, but, to our knowledge, this has not been done. Odorant detection thresholds are more difficult to determine in animal models and are far scarcer in the literature as compared with human thresholds. In insects, we are aware of only a few reports of detection threshold using behavioral measures (Kaissling and Priesner, 1970; Kaissling, 1996; Kreher et al., 2008), effectively precluding characterizing experimentally delivered odorant concentrations relative to insect thresholds.
For rodents, we were able to compile data from 27 studies reporting 123 detection thresholds for 103 unique compounds (Fig. 3A). The majority of reported thresholds were in the single ppb range (median = 2.0 ppb); threshold values for 55% of odorants fell within the range of 0.1–10 ppb. Notably, this range is similar to that of concentrations surveyed from different ambient settings and even comparable with the range of headspace concentrations of many odor sources. For example, a recent study in mice found detection thresholds for various monoterpenes ranging from 0.7–18 ppb (Williams et al., 2024)—a range similar to that of ambient concentrations measured in terpene-emitting natural environments (Tables 1 and 2). Comparing detection thresholds with headspace concentrations from our survey of direct-sampled headspace analyses (see above) on an odorant-by-odorant basis, ∼55% of all headspace concentrations are within one log unit of their detection threshold in rodents (Fig. 3C).
This comparison implies that volatile compounds emitted by natural odor sources are often encountered at concentrations near, or below, their detection threshold. This conclusion is not as surprising as it may seem, given the long tail to zero that arises from the physical nature of volatilization and dispersion of a compound from its source. Analyses of the contribution of individual volatiles to food odors in humans suggest that as few as 3% may reach perceptible levels (Dunkel et al., 2014). Similar conclusions have been made for human-detected odorants in settings involving animal waste (Schiffman et al., 2001; Blanes-Vidal et al., 2009).
Discussion
Olfactory function in the context of natural odorant concentrations
Accepting that a common operating regime for olfaction in a natural setting is at or near behavioral thresholds has important implications for considering the evolution of olfactory systems and the neural mechanisms underlying odor coding, odor perception, and odor-guided behaviors. With respect to the evolution of olfactory systems, recent behavioral studies have suggested that odorant detection thresholds may be determined by the single, highest-affinity receptor for a given compound (Sato-Akuhara et al., 2016; Dewan et al., 2018). These studies have thus far been limited to only a few odorants and receptors in mice, but they have been surprising given the canonical view that odor identity is encoded via a combinatorial code across multiple receptors (Malnic et al., 1999; Kajiya et al., 2001; Hallem et al., 2004). The unexpected alignment between naturally encountered concentrations and behavioral detection thresholds may indicate that evolutionary selection pressure on the majority of olfactory receptors is driven by their ability to detect a narrow set of odorants at natural concentrations (Dewan et al., 2018; Giaffar et al., 2023). In this view, the well-characterized variation in the sizes and compositions of OR gene repertoires across species may be regulated, in part, by evolutionary constraints on olfactory sensitivity. A strong reliance on perithreshold performance may also explain one of the most highly conserved features of olfactory systems: the convergence of OSNs expressing the same odorant receptor to a single glomerulus, and the role of OSN convergence in enhancing sensitivity in typical olfactory environments (van Drongelen et al., 1978; Duchamp-Viret et al., 1989; Bhandawat et al., 2007).
Functioning in a perithreshold concentration regime also has implications for the role of active sensing in olfaction. Though the contribution of active sampling to odor-guided behaviors remains unclear, high-frequency sniffing is a hallmark of active odor investigation in mammals. In experimental settings, repeated sniffing of an odorant does not appear important for odor perception, since behavioral discrimination and robust neural encoding of odorant identity occurs rapidly after only a single inhalation (Uchida and Mainen, 2003; Bathellier et al., 2008; Wesson et al., 2008; Cury and Uchida, 2010; Miura et al., 2012; Resulaj and Rinberg, 2015). In many cases, rodents appear to actively cease high-frequency sniffing as soon as the odorant is detected (Kepecs et al., 2007; Wesson et al., 2008, 2009). However, odorant concentrations used in these studies were typically far above threshold (∼1–250 ppm) and also higher than the range of naturally occurring concentrations for the settings we have surveyed above. In any case, high-frequency sniffing can increase the amount of odorant deposition onto the epithelium up to 10-fold (Yang et al., 2007; Staymates et al., 2016; Rygg et al., 2017). While this may confer no advantage to olfactory function at high experimental odorant concentrations, active sniffing in a natural environment may strongly impact olfactory function by allowing odorant concentrations to rise above detection threshold levels for a larger number of odor components. A role for sniffing in odor detection is consistent with the fact that animals often exhibit high-frequency sniffing during search behavior or trail following and when investigating an object of interest (Welker, 1964; Khan et al., 2012).
We note that certain compounds, in certain natural settings, are more abundant than the ppb levels typical of most natural sources. These tend to fall into a few classes: low molecular weight products of cellular metabolism (e.g., acetaldehyde, methanol, ethanol, and acetone), highly volatile products of bacterial action associated with waste or protein decomposition (e.g., sulfides, small amines and small acids), and compounds generated by damaged cellular membranes in plants (e.g., “C6” compounds such as cis-3-hexenal and cis-3-hexenol). Exploring neural and behavioral responses to such compounds may be interesting given that they constitute common cues signaling environmental processes relevant to survival.
Considerations for choosing experimental odorant concentrations
Olfactory neuroscientists in laboratory settings often use monomolecular odorants as simple and reproducible stimuli. Odorant concentration is chosen based on what evokes a measurable neural or behavioral response. Given that responses to odorants are sparse in the naturalistic ppb concentration range but broader in response to higher concentrations, this empirical approach will almost always bias experimentalists toward the use of the latter.
A key question is whether this tradeoff matters. One justification for the use of higher odorant concentrations in experimental settings is that the olfactory system can clearly “work” in this regime: higher odorant concentrations are more robust at eliciting innate behavioral responses (Manoel et al., 2021), and rodents and flies readily learn odor-guided operant tasks, including odor discrimination, detection, and source localization using high-concentration stimuli (Tully et al., 1994; Bodyak and Slotnick, 1999; Devaud et al., 2001; Parnas et al., 2013; Esquivelzeta Rabell et al., 2017). However, in both insects and rodents, increasing odorant concentration tends to change the valence of innate responses (Stensmyr et al., 2003; Kreher et al., 2008; Manoel et al., 2021; Lopes et al., 2023; Ray et al., 2023). A recent analysis of innate behavioral responses of mice to odorants of varying concentrations found that the lowest concentrations only elicited either neutral or attractive behaviors, while higher concentrations often elicited behaviors reflecting avoidance or stress (Manoel et al., 2021). Higher odorant concentrations can also alter sampling behavior via the suppression of sniffing (Johnson et al., 2003; Cenier et al., 2013), which itself may impact odor-driven behavioral performance. In flies, experiments using high-concentration odorants have concluded that chronic exposure to odors in early life—which reduces behavioral aversion to an intense familiar odor—leads primarily to behavioral habituation (Devaud et al., 2001) and accompanying plasticity in second-order olfactory neurons (Sachse et al., 2007; Das et al., 2011). Yet analogous experiments using natural odor sources and odorants at naturally occurring concentrations show that early-life exposure increases behavioral responses to an odor—in this case, attraction to the familiar natural odor—with little or no change in olfactory sensitivity of second-order neurons (Dylla et al., 2023; Gugel et al., 2023).
Thus, we argue that the disparity between experimentally delivered and naturally occurring odorant concentrations has important implications for considering experimental approaches to understanding odor-guided behaviors, whether innate or learned. For studies squarely aimed at understanding the mechanisms governing odor-guided ethological behaviors—for instance egg-laying, food-seeking, mating, or predator avoidance—the use of odorants at realistic concentrations encountered in natural settings seems especially important (Zawistowski and Richmond, 1985; Stensmyr et al., 2012; Ebrahim et al., 2015; Wang et al., 2018; Lopes et al., 2023). We also argue that, to the extent that principles of odor-guided perceptual abilities have been defined using high odorant concentrations—for example, odor reaction times (Uchida and Mainen, 2003; Rinberg et al., 2006; Wesson et al., 2008), odor mixture analysis (Abraham et al., 2004; Kadohisa and Wilson, 2006a; Shen et al., 2013; Rokni et al., 2014), and odor generalization (Campbell et al., 2013; Pashkovski et al., 2020)—it is equally important to characterize these phenomena at natural stimulus intensities. Another justification for the use of higher (i.e., ppm) odorant concentrations is that many neurons appear to respond within their dynamic range at these levels (Hallem and Carlson, 2006; Bhandawat et al., 2007), and high concentrations of a monomolecular odorant may substitute as an approximation of natural stimuli (i.e., mixtures of multiple odorants at low concentrations), as both elicit broad activity patterns. In this view, principles of odor identity coding inferred from testing single odorants at high concentrations may generalize to a naturalistic regime. We argue that this rationale is conceptually flawed.
First, other sensory modalities show a similar dependence of sensory afferent tuning on stimulus intensity (audition is a classic example; Evans, 1972), yet a single frequency high-intensity stimulus is not generally considered a proxy for broadband low intensity stimuli. At the simplest level, response features of OSNs do not scale linearly with concentration; instead, increasing odorant concentrations lead to saturation, adaptation, or even inactivation by depolarization block (Duchamp-Viret et al., 2000; Conway et al., 2023). As a result, monomolecular odorants presented at abnormally high concentrations may drive abnormally distributed patterns of activity across OSNs or glomeruli (Fig. 4A): those OSNs most sensitive to the odorant will be saturated, while additional OSNs are recruited that may rarely, if ever, respond to that odorant in a natural setting (but see Stevens, 2016). In contrast, multicomponent odor mixtures arising from natural sources likely evoke distributed patterns of activation across multiple receptor populations at nonsaturating levels, which may be important in robustly encoding odor identity across naturally varying intensities of the source odor (Gschwend et al., 2016).⇓
Potential impact of concentration on odor coding by olfactory neuron populations. A, Idealized concentration-response function for a single olfactory sensory neuron (OSN), and responses across the OSN population for a single odorant presented at ppb and ppm concentrations, compared with responses to a natural odor source. B, Hypothesized population representations of monomolecular and natural odor sources at the level of OSNs or glomeruli of the olfactory bulb (OB) or antennal lobe (AL; top row) and third-order neurons in piriform cortex (PCx) or mushroom bodies (MB; bottom row). Schematic of canonical neuronal projections from primary to tertiary levels is shown at left. For monomecular odorants, perithreshold (ppb) concentrations evoke sparse activity with individual units narrowly tuned to structurally similar odorants. Third-order representations are even more sparse, with most stimuli failing to evoke appreciable activity. Unnaturally high (ppm) concentrations evoke broad activity with many first-order units saturated; third-order representations reflect broad chemical classes inherited from the broad tuning of first-and second-order neurons. For natural sources, individual sources (or naturalistic odor mixtures) evoke sparse but distributed activity across the first-order population. Third-order representations are sparse and optimally tuned to specific odor objects.
Estimated volatile concentrations in the headspace of natural food sources. A, Distribution of volatile compound concentrations across all measurements (references) in the complete and filtered VCF dataset. Each combination of compound x measurement contributes one count. B, Distribution of the 90% percentile concentrations of odorants across measurements in a product category in the filtered dataset, representing the distribution of the upper bounds of odorant concentrations that could be expected to be encountered in natural sources. Each combination of compound × product contributes one count. C, Distribution of concentrations of total volatiles of the consensus “maximal” headspace for all sources in the filtered dataset. Each product category (e.g., “apple”) contributes one count.
Second, we hypothesize that the difference in activity patterns evoked by high-intensity monomolecular odorants versus natural odor sources is important for how odor representations manifest at higher-order stages of processing. For example, unnaturally high odorant concentrations may underlie response features in second-order neurons that have been proposed to play important roles in odor identity coding, such as “afterimages” of mitral/tufted activity after odorant offset (Patterson et al., 2013) or the reformatting of mitral/tufted cell odorant response patterns with repeated inhalations (Eiting and Wachowiak, 2020); whether such response features are relevant to olfactory processing in a naturalistic concentration range remains to be tested.
The dominant model for odor object encoding in neurons of third-order olfactory regions (e.g., mammalian piriform cortex or insect mushroom body or lateral horn) relies on the integration of multiple coactive second-order inputs (M/T cells or projection neurons; Kadohisa and Wilson, 2006b; Masse et al., 2009; Gottfried, 2010; Gruntman and Turner, 2013; Jeanne et al., 2018). The nature of this integration with respect to the number and receptive ranges of second-order inputs remains unclear. While monomolecular odorants can effectively drive activation of third-order neurons, they may be poorly suited to probe odor representations at this stage: odorants presented at naturalistic (i.e., ppb) concentrations may fail to activate sufficient third-order neurons due to their sparse activation of sensory neurons (Davison and Ehlers, 2011; Gruntman and Turner, 2013), while unnaturally high concentrations (i.e., ppm or higher) may drive third-order neurons in a manner that reflects the receptive range of their strongest inputs (Pashkovski et al., 2020) rather than how they encode natural odor objects (Fig. 4B).
Recalibrating olfaction to the range of naturally occurring concentrations
We offer several recommendations to better align experimental studies of olfactory system function to the regime of naturally occurring olfactory stimuli. First, despite inherent difficulties in accurately delivering odorants in vapor phase at a given concentration (Cometto-Muñiz et al., 2003; Thiele et al., 2023), reporting estimates of delivered concentrations based on ideal assumptions and, optimally, calibration measurements is critical for interpreting experimental results in any context. Delivered concentrations were either unreported or impossible to determine in a substantial fraction of the experimental studies we surveyed; this included several studies that have generated foundational datasets used to characterize the neural representation of olfactory stimuli (Takahashi et al., 2004b; Hallem and Carlson, 2006). Clearly, important advances in olfaction have been made with experimentation using relative stimulus concentration, but transitioning to using naturally relevant odorant concentrations will be important for advancing our understanding of olfactory sensation and behavior in natural settings.
Another recommended practice is to reference the choice of concentration to a natural source or setting based on reports from the literature. Data such as those compiled here can be used as a general guide for establishing appropriate concentration ranges. As a field, we need more data regarding naturally encountered odorant concentrations for settings and sources relevant to common experimental models like flies and rodents. While perhaps less frequent, encounters with odorants at substantially higher concentrations than those surveyed here may still be crucial for survival, and we note that some settings in which this may occur are not well represented in the surveyed literature. For example, urine from conspecifics or predators may emit particular compounds at concentrations higher than those measured from food or plant sources and emitted concentrations of some compounds may increase as urine dries on a substrate (Kwak et al., 2013). Other compounds may be emitted at high concentrations in a very localized fashion via secretions from animal glands or plant trichomes or during important behavioral events (Jefson et al., 1983; Kimoto et al., 2005; Kilpinen et al., 2012). Finally, some transport processes may preserve packets of exceptionally high concentration by aggregation of molecules onto aerosols or dust (Bottcher, 2001; Jami et al., 2020). New studies measuring the concentrations of ethologically important odorants in their relevant natural settings would be extremely helpful in placing past and future experimental efforts in an ecologically relevant context.
Experimental approaches aimed at understanding the function of higher-order olfactory areas or investigating phenomena at the whole-organism level, such as innate and learned odor-guided behaviors, may be better served by using odorant mixtures as stimuli. Naturalistic odorant mixtures may uniquely engage central circuits tuned by evolution or experience to optimally represent odors based on their specific chemical signatures (Takahashi et al., 2004a; Mathis et al., 2016; Dweck et al., 2018; Lahondère et al., 2020; Coureaud et al., 2022). A better understanding of the statistical structure of the occurrence of monomolecular odorants that comprise different natural odors could be useful for the rational design of artificial odor mixtures that reasonably approximate natural odor stimuli.
Another recommendation is to consider using natural odor sources themselves as stimuli (Lin et al., 2006; Vincis et al., 2012; Schubert et al., 2014). This approach has limitations, including the complexity of the odor stimulus and variability across samples over time. However, natural stimuli can be crucial in verifying responsiveness of target neurons thought to be tuned to particular compounds known to be present in a given source (Lin et al., 2006; Dewan et al., 2013; Dweck et al., 2013; Mansourian et al., 2016). In such cases, responses to a natural source can be used as a reference point for analyses using monomolecular odorants or defined mixtures. While it is tempting to use extracts or essential oils from natural sources to generate more naturalistic stimuli, we caution that these tend to be highly concentrated and thus have the same potential for using unnaturally high concentrations.
In certain experimental settings—particularly behavioral experiments—the variability of odor stimuli provided via natural sources might be a design feature, rather than a confound. Response consistency in the face of such variability provides information about the extent to which the system generalizes and assigns noisy stimuli to the correct odor source. In this sense, the use of natural sources better models the tasks that olfactory systems must solve in the environments in which they evolved. We argue that, despite some additional challenges, placing a greater emphasis on defining appropriate stimulus concentration will yield significant benefits to our understanding of olfactory system function.
Appendix
Calculating and comparing vapor phase odorant concentrations across studies
Behavioral and experimental studies rarely measure delivered vapor phase concentrations directly. Instead, these are calculated from the estimated headspace concentration of pure odorant (based on the vapor pressure), its dilution by air (or other carrier) in the vapor phase, and its dilution in the liquid phase assuming a linearly proportional reduction in vapor pressure by liquid dilution that follows ideal gas laws (i.e., Raoult’s law). However, many, if not most, odorants deviate from ideal behavior due to molecular interactions between the odorant and solvent. While negative deviations are possible, positive deviations are more common, resulting in vapor odorant concentrations that can be significantly higher than ideal predictions (Jennings et al., 2022; Cometto-Muñiz et al., 2003). While such deviations (described by activity coefficients or Henry’s law coefficients) can be corrected, these have been empirically determined for only a small number of the odorant/solvent combinations. In the few cases where such corrections were made, or in which delivered concentrations were directly measured and reported, we used these corrected values; otherwise, we included the values reported based on ideal assumptions.
To compare datasets across studies and settings, all reported concentration values were converted into a common unit of parts per billion (ppb) in the vapor phase—i.e., molecules of odorant per molecules of air—which was the most commonly used unit across different study types. Molarity (M, mol/L) was converted into ppb using the conversion: ppb = M * 1 × 109 / 0.0446, where 0.0446 is approximately the number of moles per liter of air. Units of mass per volume (e.g., g/m3) were converted to ppb assuming a mass of 1.2 g/L of air. Odorants were indexed by their PubChem Chemical Identifier Number (CID; pubchem.org).
Survey of experimental studies
A complete survey of odorant concentrations used in experimental studies of olfactory function was impractical. To achieve a reasonably broad survey and avoid bias in study selection, we selected studies cited in two exceptionally comprehensive reviews of odor coding in mammals and insects (Fulton et al., 2024; Galizia and Sachse, 2010). We initially included all studies using at least six odorants to probe some aspect of olfactory function in vivo in rodents and insects; this criterion included 46 studies from rodents and 139 from insects. Areas of focus included odor-evoked neural responses at the level of primary sensory inputs, the olfactory bulb or antennal lobe, higher-order olfactory areas including piriform cortex, amygdala, mushroom bodies and lateral horn, and behavioral measures of odor perception. For experimental studies where delivered concentrations were not reported, we estimated the delivered concentration using ideal assumptions, as described above, and the details provided regarding liquid and vapor phase dilutions. We excluded studies which lacked the details necessary to estimate delivered concentration or in which odorant concentration was not sufficiently well controlled; this was the case in 16/46 (35%) of rodent studies and 109/139 (78%) of insect studies. Finally, several additional studies were included beyond those cited in the reviews based on their recent publication or high level of citations. For the analyses shown in Figure 3, we included all unique combinations of odorant and concentration used in a given study.
Ambient concentrations of volatiles in different settings
Table 2 shows reported ambient concentrations for select compounds measured in different settings. When possible, we selected studies using proton transfer reaction-mass spectrometry (PTR-MS) as a measurement method, as it is a direct sampling method that allows for time-resolved measurements. In many settings, ambient concentrations vary over time—for example, ambient concentrations of monoterpenes show diurnal variations depending on temperature and wind speed. For this compilation, we included the peak levels reported in a given study. For studies with measurements made over long time periods (weeks or months), we include the highest mean concentrations reported. When individual compounds are included in the table, these were selected based on their high abundance relative to other measured compounds.
Survey of directly measured odorant concentrations in headspace vapor
We compiled data from studies that used PTR-MS to quantitatively analyze the concentrations of compounds directly from the headspace (without an adsorbent intermediate). The 20 studies surveyed included 383 unique compounds. For some sources, different instances of a source were analyzed—for example, intact versus cut apple; these were compiled independently, for a total of 32 source instances. Median and 75th percentile of concentrations of all identified compounds in a given source instance are reported in Table 3. To estimate the upper end of naturally encountered concentrations for a given compound across all potential odor sources, we used the 90th percentile of all measured headspace concentrations, per compound, across all source instances; however, 121 of the compounds occurred in only a single instance. The majority of compounds were identified in only one or a few instances (median # of sources = 2), although 10% of compounds were common to at least 10 of the 32 instances, consistent with previous meta-analyses indicating that odor chemical “space” includes both “specialized” (i.e., source-specific) and “generalist” compounds (Dunkel et al., 2014).
PTR-MS has some limitations in its ability to unambiguously identify certain compounds, especially those with same mass/charge ratio (Pang, 2015; Majchrzak et al., 2018), and it is common to report concentrations of compound classes, such as monoterpenes, or molecular fragments. We excluded such data from this compilation. In cases where a compound was identified as one of two or three possibilities (e.g., stereoisomers such as trans-2-hexenol or cis-3-hexenol), we included an entry for each compound with the same concentration value assigned to each, in order to facilitate comparison with other datasets.
Analysis of data from the Volatile Compounds in Food database
The Volatile Compounds in Food (VCF) database (v16.10, https://www.vcf-online.nl/VcfHome.cfm) is a subscription-based compilation of published volatile compounds in 1,260 human food products. Food sources belonged to one of nine major food groups: wine and beer; fruits; vegetables; fungi; grains; nuts; dairy; meat and poultry; and seafood. To focus on odor sources likely to be encountered in natural environments, we filtered the database to exclude cooked or heavily processed foods (e.g., smoked meats, baked goods, oil extracts, liquors, cocoa or coffee beverages, etc.), while retaining fermented food products such as wine and yogurt. We also only included entries with quantitative concentration data (e.g., samples where compound abundance was given qualitatively as present, absent, trace, etc. were dropped). These criteria brought the number of food sources considered in our analysis to 479.
Entries in the VCF database are reported as their estimated concentrations in the source, typically after headspace analysis using absorptive methods that involve static sampling of volatiles in the headspace for a fixed time interval, followed by desorption and odorant identification and quantification using gas chromatography-mass spectrometry. In-source concentrations are then calculated based on calibrations to standards of known concentration, sampled equivalently (Ioffe, 1984).
To derive estimates of headspace vapor concentrations from such data (Fig. 5), we converted the reported in-source concentrations to headspace concentrations using the vapor pressures of each odorant, assuming ideal gas law behavior (Cometto-Muñiz et al., 2003). An important caveat to these calculations is that this conversion undoubtedly introduces error into estimates of headspace concentrations due to nonideal interactions, in addition to nonideal effects arising from the nonliquid matrix of the food source. These errors likely differ for different sources and odorants. Nevertheless, the overall structure and distribution of odorants across the natural sources in the dataset was similar to that observed from direct PTR-MS measurements: most compounds occurred in a small number of food sources (median number of sources = 1; mean number of sources = 3.4), and a small fraction of compounds (∼7%) occurred frequently in many sources (>15% of food sources). Fruits—particularly citrus fruits—and wines tended to have the highest total volatile concentrations, while nuts, meat, and poultry tended to cluster towards the lower end.
For calculations of the total concentration of all volatiles in different natural source headspaces (Fig. 5C), we generated a consensus headspace for each food product. We merged all instances (references) of the food product into a single product category (e.g., “apple”) by compiling the union of odorants (observed in two or more measurements of the product), using the 90% percentile concentration for each odorant across all measurements in which it occurs. This method resulted in a consensus “maximal” headspace profile that represents the upper concentration bound for all volatile compounds that can be expected to be reliably encountered for each product.
Survey of psychophysical detection thresholds
Rodent detection thresholds were obtained from the literature using the following criteria: (1) the study used psychophysical methods and (2) delivered odorants using an olfactometer that allowed for controlled delivery of a known (or estimated) vapor phase concentration. These criteria resulted in 28 published studies reporting 125 detection thresholds for 103 unique odorants; these are listed below. We defined detection threshold as the average sensitivity for all experimental subjects within a study, according to the study’s definition of threshold. One study (Blount and Coppola, 2020) was excluded as their reported detection thresholds for carvone and limonene were extreme outliers, differing from other studies testing the same odorant by 6-13 orders of magnitude, leaving 123 detection thresholds from 27 studies. All detection thresholds were converted to ppb and include studies that assume ideal gas behavior (87 detection thresholds) and those which have accounted for the nonideal behavior of diluted odorants or used flow dilution olfactometry (36 detection thresholds). For the 10 odorants tested across multiple studies, the lowest reported detection threshold for each odorant was utilized in our analyses.
Footnotes
We thank the National Science Foundation (NSF) Neuronex Odor2Action research group for facilitating discussion of the ideas discussed here. We also thank Shawn Burton, Don Katz, and members of the Wachowiak and Hong labs for providing critical feedback. We thank Kirstyn Grams for generating Figure 2 (created with BioRender.com). This work was supported in part by NSF Neuronex Award DBI2014217 (M.W. and E.J.H.), the Center for Evolutionary Science at Caltech (E.J.H.), Brain Initiative U19 U19NS112953 (T.B.) and R01DC020720 (A.D.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Matt Wachowiak at matt.wachowiak{at}utah.edu or Elizabeth Hong at ejhong{at}caltech.edu.
