4.1. Selective adaptation determines which odors in mixtures can be identified
Odors come and go. Hidden when stimulus compounds are presented at the same time in a mixture background, characteristic odors of compounds momentarily appear when newly added to the mix. To imitate this scenario, our subjects were presented with a stimulus to sniff before switching about 5 seconds later to sniff a stimulus containing one extra compound X. The brief selective adaptation greatly enhanced identification of characteristic odors of un-adapted X and reduced recognition of odors of previously presented compounds. We propose that odors of newly appearing compounds are often perceived for a few moments above the concealed odors of ambient compounds. The transient, ever-changing odor percepts are the consequence of two remarkably important and unexpectedly rapid physiological processes, mixture suppression and sensory adaptation.
Selective adaptation’s impact on identification was large, about 50 % of the 80 % practical range of our subjects’ ability to identify the odors. While testing stimulus pairs, correct identification of single compounds’ odors averaged 90 % and false positives averaged 10 % (). When compounds defined as extra or ambient were presented with no adaptation in control sessions, identification of component odors averaged 55 % (the average of extra and ambient control bars in ); however, after selective adaptation, identification of an extra compound’s odor increased to 75 % and identification of odors of ambient compounds fell to 38 % (). The 20 % increment for extra compounds plus 17 % decrement for ambient compounds sums to 37 %, nearly 50 % of the subjects’ 80 % identification range.
The strong influence of selective adaptation on odors suppressed in mixtures demonstrates dynamic olfactory coding, a rapid focusing of the olfactory system on newly introduced and stronger odors at the expense of veridical identification of the odors of all compounds that are present. To demonstrate this, we compared veridical identification (scoring 1 point for identifying odors of present compounds and 1 point for not identifying odors of absent compounds) with identification predicted by dynamic odor coding (scoring 1 point for identifying odors of newly added extra compounds and 1 point for not identifying odors of any other compounds, ambient or absent) (). Veridical coding fell far short of dynamic odor coding (F (1,3) = 39, p <.0001). Furthermore, veridical odor identification worsened going from 3 to 4 component mixtures (p <.0001) but identification focused on the extra compound did not (F (1,13) = 16, p <.001). Our subjects were equally good at identifying un-adapted extra odors in 3 component or 4 component mixtures (cf., ). Humans may detect characteristic odors of newly added compounds no matter how many ambient suppressed/adapted compounds are already present.
Identification of odors in mixtures
4. 2. Mixture component identification reveals analytic odor processing
We studied the processing of odor information in humans by asking subjects to identify an odor embedded in a mixture immediately after presenting other mixture components. Substantial effects of mixture suppression [45
] and selective adaptation [41
] on identification demonstrated the value of the method both technically and theoretically. By using 16 mixture permutations of 4 component compounds (), a robust statistical treatment based simply on correct identification was possible [24
We used water soluble compounds with distinctive, recognizable and agreeable odors, compounds that had overlapping as well as distinct molecular features. The subjects easily learned to use odor labels for single components before testing sessions. The 4 compounds, with concentrations adjusted to moderate odor intensities consistent with mixture suppression [39
], exhibited mixture suppression during testing. Control identification proved to be reduced in 3 and 4 component mixtures (F(1,13) = 28.80, p < 0.0001) even though odors of the 4 single compounds were not recognized with equal frequency (F(3,39) = 4.27, p < 0.009).
Studies of anosmic humans [12
] have shown that phenethyl alcohol and vanillin mostly activate the olfactory nerve, CN I; but isopropyl alcohol [60
] and l
-menthol, like many olfactory stimuli, also activate the trigeminal nerve, CN V. Compared to extra controls (control bars at left of bars for extra stimuli in ), correct identification of characteristic odors of extra compounds, with or without trigeminal epitope, increased by 18 % when added to the adapted mix. However, compared to ambient controls (control bars at left of bars for ambient stimuli, ), the 5-sec adaptation reduced correct identification of vanilla and rose odors by 22 %, more than the alcohol and mint odors, which were reduced by 8 % (F (1,13) = 5.33, p = 0.04). We are uncertain whether this difference reflects fundamental differences in trigeminal and olfactory adaptation/identification [59
] or slight mismatches in stimulus intensity, distinctiveness or familiarity. Nonetheless, identification of odors of the four component compounds all showed similar improvement after selective adaptation, providing evidence for analytical processing of odors not available from other experimental paradigms [25
We propose that by concealing ambient odor identity, mixture intensity suppression and adaptation allow for recognition of dominant characteristic odors. The odor of a very complex mixture without recognizable attributes would have little utility because mixtures typically lack repeatability and therefore identity, especially on variable backgrounds. Consistent with mixture suppression, odors of many common foods and fragrances appear to be dominated by notes of single compounds that impart characteristic, albeit crude, identities [20
]. Because odor-component identification is suppressed in mixtures [44
], we surmise that olfactory notes originate from very few components that dominate odors of mixtures due to their relative strength at a given point in time [39
]. “Top note,” “middle note” and “bottom note” [62
] are terms used to specify the timing of successive odors emerging from fragrant stimulus mixtures. Such identifiable odors are similar in principle to the odors that appear following selective adaptation that we report experimentally.
4.3. Characteristic odors rapidly wane after a sniff or two
In our experimental configuration, component odor recognition was already suppressed in mixtures. Yet, compared to controls, our subjects’ abilities to identify ambient odors declined an additional 31 % after 5 seconds of sampling, which changed relative odor saliency [38
], making an extra odor easier to identify. In adaptation sessions we did not simply measure declines due to waning odors, but tested ambient-odor identification after an extra compound had been introduced into the mix. Because identification in adaptation sessions was compared to control sessions when subjects received all test
compounds without adaptation
, we attribute declines in ability to identify ambient odors to a rapid sensory adaptation, not suppression. A few seconds of adaptation had quite significant consequences on odor identification, revealing an important odor coding strategy: a rapid reduction of ambient “noise”.
Although some human psychophysical experiments having subjects rate stimulus intensity suggest rapid adaptation/recovery and partial adaptation after a single sniff [53
]; other rating experiments found odor adaptation may be incomplete even after several minutes or more, depending on the type of stimulus and its concentration [4
]. Furthermore, cross-adaptation, effects of prior presentation of one compound on the odor intensity of another compound, may not consistently reflect molecular or odor similarity [53
]. However, we demonstrated rapid selective adaptation with a clear loss in adapted-odor identification, an effect that may be associated with intensity reduction [41
Animal studies also provide experimental data on the time course of olfactory adaptation and recovery from adaptation that are quite varied; loci and mechanisms of adaptation remain unclear [15
]. Newt and salamander olfactory sensory neurons (OSN) adapt within 1 sec. and recover from adaptation within 10 sec. [31
]. Rat and mouse OSN also partially adapt in a few seconds [5
]; however, rapid OSN adaptation is not easily reconciled with olfactory bulb and olfactory cortex recordings showing adaptation taking a minute or more [29
]. The mismatch may come, in part, from how sensory adaptation is defined. Some recordings from brain sites address decreased responding with repeated stimulus presentation (habituation), which likely involves different mechanisms than decreased responding with continuous stimulus exposure (adaptation). In fact, if OSN were to rapidly adapt and recover (as OSN may), responses to repeated stimulus pulses would not decline peripherally and the habituation observed would require central circuits. Also, inconstant odors due to sources that fluctuate or actions, such as varying head position and sniffing [47
], may be experienced for long periods due to varying stimulus concentration. Nevertheless, under the controlled conditions of our experiment, the ability to correctly identify ambient compounds’ characteristic odors by associated labels [10
] rapidly declined to a level sufficient to decisively promote recognition of single added compounds. Determination of how the adaptation we measured by identification quantitatively relates to decreases in intensity measured by ratings [41
] requires further study.
4.4. Dynamic odor coding assures odor-quality constancy
Our finding that odor mixture components will reclaim their individual identities after selective adaptation raises important issues about mechanisms of odor quality coding. A general lack of a one-to-one correspondence between particular stimuli and individual receptors has led to a consensus that olfactory coding is based on combinatorial molecular feature detection [48
]. One specific theory argues that OSN “deconstruct” an olfactory stimulus into its individual molecular features [7
] and, according to this model, each individual molecular feature of the compounds smelled by our subjects would be separately coded by a distinct set of OSN. The compounds that we used shared some molecular features (). The olfactory responses to shared molecular features, theoretically, should have been adapted for extra as well as ambient compounds. Subtraction of responses to a key molecular feature would likely obscure odor identity; yet characteristic odors of extra compounds were more easily
detected above the background. Subjects identified odors as the same as characteristic odors of the distinguishable individual compounds for which they had learned “veridical” labels [10
]. Thus, the key constellation of molecular features of an individual odorous compound appears to remain irresolvable in mixtures.
There are studies that support a combinatorial model of odor stimulus recognition by overlapping and diverse receptor arrays [6
]. Two common themes in these studies are that olfactory receptors are responsive to a wide range of chemical stimuli and most odorants activate multiple receptors. A given receptor type could be activated by a range of ligands, even those having different functional groups and presumably different odor qualities [7
]. Octanal, a simple aliphatic aldehyde, was an effective stimulus for 55 different receptors, implying that octanal recognition might require activation of this large array [18
]. Furthermore, some of the same receptors also responded to octanol, an alcohol analog with a different odor quality.
There are reasons why a broad spectrum of chemicals at high concentrations might activate receptors. For one, specificity of tuning may be greatest at the lowest concentrations [2
] and the 10 to 100 micromolar stimuli used could result in non-specific binding [43
]. Although it is hard to estimate what concentrations are “physiological,” many odorants are recognized in the gas phase at 1 part per million or less, corresponding to a vapor concentration of about 40 nanomolar, or roughly 1000 times weaker than that used in the liquid phase experiments noted above. Also, certain compounds, particularly aldehydes that can form Schiff bases with amino groups in proteins, might not be acting as true ligands, but rather as “reagents” at high concentrations. Another notable example is the use of high concentrations of dicarboxylic acids as olfactory ligands in mammals [48
], when dicarboxylic acids as a group have negligible vapor pressure and no odor [52
The olfactory receptors (OR) can display a high degree of specificity, as reflected by the ability of humans to discriminate optical isomers [40
] and by the repeated ligand-driven isolation of the same OR in single-cell PCR experiments [64
]. The original portrayal of rat OR I7 as an “octanal receptor” [71
] further suggested specificity of ligand-receptor interactions but this concept may need reappraisal in view of subsequent identification of many receptors sensitive to octanal [18
]. Specific anosmias in humans [1
] are consistent with codes for particular odors representing restricted activation of some part of the olfactory neural circuit. The fact that the component stimuli we used did not cross-adapt implies that they did not elicit functionally significant overlap in receptor-activation patterns.
OSN display highly specific features. Each one of the hundreds of G-protein coupled ORs [8
] is exclusively expressed in a subset of OSN that target a specific few glomeruli located in stereotyped positions in the olfactory bulb [51
]. Furthermore, there is growing evidence that olfactory bulb output neurons, the mitral cells, are specific chemical feature detectors [43
]. Each mitral cell obtains direct input exclusively from a single glomerulus to convey to the olfactory cortex. A hallmark of mitral cell responses is a sharpening of tuning attributable to inhibitory inputs from neighboring glomeruli, which receive direct input from different OSN and respond to different olfactory stimuli [69
]. Overall, mixtures would suppress mitral cell activity while enhancing differential responses to a new stimulus. The masking of components in mixtures and their emergence following selective adaptation demonstrated by our study are logical perceptual consequences of olfactory bulb inhibitory circuits. Ambient odors fade quickly over time, allowing identification of the appearance of a new stimulus. However, the odor of the new stimulus will be identified only as long as it remains stronger than odors of other mixture components: its odor too quickly adapts. This changeability is the hallmark of dynamic odor coding
An important corollary of our odor-identification study is that it demonstrates odor-quality “constancy” in the presence of different backgrounds, a necessary and sufficient condition for olfactory stimulus recognition. Taste-quality “constancy” in mixtures is also observed [3
] and the unmasking of a taste or olfactory mixture component following selective adaptation has been briefly noted previously [33
]. A principal difference between taste and smell lies in numbers of human taste and odor qualities. A total of 5 taste qualities [28
] is sufficient to identify gustatory attributes of many different foods and drinks. Molecular receptors for the sweet and bitter tastes [23
] are linked to heterodimeric GPCR complexes and multiple GPCRs, respectively, each expressed in single taste-bud receptor cells [55
]. The hundreds of ORs devoted to olfaction, singly expressed in dedicated OSN, may likewise each be associated with one quality or odor note. If pure substances were to each have just two resolvable, distinguishable dominant odor notes of similar strength [20
], an array of 350 different human olfactory receptors [19
] could theoretically allow the recognition of more than 100,000 different chemicals. Predictably, because, unlike tastes, they are so numerous, odor qualities are difficult to bring to mind unless linked to actual objects in one’s personal experience [10
] and appear countless and difficult, if not impossible, to classify [9
Although we ordinarily associate smells with objects, the idea that we recognize distinct coffees and wines by patterns generated by extremely complex “cocktails” of olfactory stimuli is implausible, given mixture suppression. Coffee is a highly variable commodity and only a few key odor notes will likely be common to all coffee preparations, allowing us to call them all “coffee.” In a similar way, wine odor is likely dominated by a few chemicals, invariably including ethanol, with distinctive attributes of different wines recognizable [9
], but a particular wine would not be expected to have a very large number of odor qualities [20
]. Recognition, characterization and categorization of the complexes of odor qualities of variable objects may require some of the constellation of hierarchical-parallel circuits available in the central olfactory system for odor memory, cognition and affect [9
In evolutionary terms, receptors are unlikely to have evolved for complex substances such as coffee and wine because both are variable and these popular examples are largely artificial as well. The receptors and possibly the associated neural circuits in the olfactory bulb [69
], more likely evolved to detect specific constellations of chemical features, more or less representative of single compounds [42
] —not complex and variable objects. It is likely for this reason that so many receptor classes are needed to characterize the multiple odor qualities.