Richard J Stevenson. Encyclopedia of Perception. Editor: E Bruce Goldstein. Sage Publication. 2010.
Olfactory imagery is the capacity to experience a smell, intentionally, when the appropriate stimulus is absent (e.g., sniff and imagine the smell of lemon). This is referred to as phenomenal imagery. Phenomenal imagery is difficult to study, as it relies upon self-report, which is hard to validate. Consequently, many investigators explore whether imagining an odor has measurable effects on behavior that are akin to those observed when its real equivalent odor is smelled (e.g., does imagining a smell help one to detect its real equivalent?). This is referred to as performance imagery. Performance imagery assumes the presence of an odor image, but as described in this entry, other causes may produce apparently similar results. Finally, an olfactory image may be evoked automatically and can come to form an indistinguishable part of the percept (e.g., smell vanilla—if you judge this as “sweet” smelling then you have just experienced this form of imagery). In this case, volition has no part to play, there is an odor stimulus, but the resulting percept is a combination of image and stimulus. This is referred to as cue-driven imagery.
People report that they can imagine odors. However, they also report that their images are not very vivid and are hard to generate, relative to their experience of visual and auditory images. Some more specific findings have also emerged. Olfactory experts report being better able to imagine odors than nonexperts and factor analysis of self-report measures of imagery in all modalities indicates that there is a separate olfactory/gustatory (taste) dimension. That these two modalities should appear together is no surprise, as they routinely combine to form flavor.
The difficulty in generating olfactory images and their lack of vividness may result from a number of causes. It has been suggested that smell perception may be less vivid and distinct than sensory experience in the other modalities, and so this may be reflected in imagined experiences too. However, certain smells can clearly capture attention (e.g., burning or leaking gas) and others may readily elicit strong emotions or distant memories. Consequently, this may not provide a convincing explanation for the reported lack of vividness and difficulty of evocation.
A further reason is that normal participants may simply be unpracticed at forming olfactory images, and the finding that experts are better in this regard suggests this possibility. Similarly, many imagery experiments begin in the following way: “Imagine the smell of lemon.” To form the image, some connection must exist between the word lemon and a memory of what lemon smells like. Interestingly, normal participants are very poor at naming odors (unlike olfactory experts) that are going from the odor percept to the name. It has been suggested that this is the flipside of going from a name to an odor memory, and that the impoverished nature of the odor-name link—which may result from lack of practice or poor neural connections—may explain why our capacity to imagine odors is not as good as imaging visual or auditory objects. Support for this notion comes from several sources. Hard to name odors are also hard to imagine (unlike their visual equivalents—try to visualize the smell of peanuts versus the sight of peanuts), but learning the odor’s names improves self-report measures of imagery. More interestingly, nearly all of the successful performance imagery studies described in the next section involved participants learning the target odor names prior to imagining them.
Three problems make interpretation of the performance imagery literature difficult. The first is that if a perceptual condition (where the odors are really experienced) reveals similar performance to an imagery condition (where the same odors are imagined), the two conditions, real and imagined, will not appear to differ. Unfortunately, this type of outcome, termed a null result, is very hard to interpret, as there are almost an infinite number of possible explanations for why something did not differ. The second is that many studies do not adequately exclude more compelling alternative explanations for any observed similarity. People may have considerable semantic knowledge about the way odors behave and their properties. Semantic knowledge is usually able to be verbalized and reflects what one knows about a particular thing (e.g., lemons are yellow and taste sour). This semantic knowledge might underpin performance, resulting in equivalent outcomes between real and imaginary experimental conditions. For example, people may know that a stronger odor masks a weaker one, thus, if they are asked to imagine a strong smell while actually sniffing a weak one, they may report that the weak one has been masked. This problem has been effectively overcome in some of the most recent reports considered here. Finally, even if one convincingly demonstrates that performance under imaginary and real conditions is equivalent, and that it does not result from the use of semantic knowledge, how can one be sure that it does result from experiencing an odor image? Attempts to tackle these thorny problems are persistent themes throughout the performance imagery literature.
An impressive variety of procedures have been deployed to examine whether olfactory imagery and perception have similar effects. Two examples illustrate this diversity. First, participants asked to imagine smelling foul odors demonstrate an observably different pattern of sniffing compared to when they imagine pleasant smelling odors. This mirrors the way people behave with real equivalent odors. Second, qualitative similarity (i.e., what something smells like) has been examined for sets of imagined and real odors. There are substantial similarities between real and imagined sets of odors (e.g., lemon and orange odors are judged more alike than boot polish when odors are imagined and when they are smelled). However, as critics such as Hendrick Schifferstein have pointed out, such effects might be obtained by simply asking participants to use their semantic knowledge of odors.
An apparently more definitive finding came from Brian Lyman and Mark McDaniel, who found that the ability to recognize whether one had (or had not) smelled an odor before could be enhanced by imaging its smell. Although others have failed to obtain results akin to this with conceptually similar designs, notably Rachel Herz, the finding that image clarity ratings in Lyman and McDaniel’s study (obtained while participants attempted to imagine the odors) were correlated with recognition memory performance suggests otherwise. It is difficult to see why better self-reported imagery performance should correlate with recognition memory performance if the process was mediated semantically.
The most compelling evidence for performance imagery comes from a recent study conducted by Jelena Djordevic and colleagues. Participants were trained so that they could readily name the two odors used in the study. After establishing participants’ threshold for these odors, they were assigned to one of three experimental conditions, olfactory imagery, visual imagery, or a no imagery control. Each group was then given forced-choice detection trials (i.e., is an odor present?). Participants in the imagery conditions were asked to imagine one of the two stimuli just prior to the detection trials. On half of the detection trials, the imagined stimulus corresponded with the target to be detected, and on the other half it did not. They found that detection accuracy was impaired in the olfactory imagery condition, relative to the visual and control conditions, when participants were asked to imagine one odor (e.g., rose) while detecting another (e.g., lemon). These findings are not based upon a null result, as only the olfactory imagery condition affected performance. They also exclude a semantic mediation effect, as participants in the visual imagery condition would have had access to the same semantic knowledge about the odors as participants in the olfactory imagery condition. Finally, they also obtained some evidence for a relationship between participants’ self-report of imagery ability and performance on the task—but only in female participants.
A further approach to studying performance imagery has been to establish whether brain areas involved in olfaction are activated when participants imagine odors. While favorable evidence has been obtained by Robert Henkin, the findings are still open to interpretation. In particular, one recent study revealed activation of the primary olfactory cortex simply via reading olfactory-related words (e.g., cinnamon). Participants were never instructed to form images. Thus, the repeated concern about the role of semantic, as opposed to perceptual processes, arises here as well.
Two different forms of cue-driven imagery have been demonstrated. The most compelling example of the first form, where the cue comes from another sensory modality, is from an experiment on wine tasting. George Morrot and colleagues had participants describe the characteristics of white wine and white wine that had been colored red. Participants described the odor of the red-colored white wine in a manner consistent with descriptions of real red wine. A large number of similar findings have also been obtained in which color has striking effects on participants’ judgments of odor intensity, naming, quality, and hedonics. One explanation for these effects is that the color cue automatically triggers odor perceptual memories that are consistent with this visual information, and that this top-down perceptual memory overwrites information derived from bottom-up stimulus driven processing. However, evidence favoring this type of explanation has not been forthcoming, and the alternative, in which participants utilize semantic knowledge to generate their evaluations, is better supported. In a recent study, Richard Stevenson had participants engage in an odor discrimination task, where the fluids carrying the odor were incidentally colored. Participants were poorer at discriminating similar odors (e.g., cherry and strawberry) when these were presented in inappropriate colors (e.g., in green fluid). Importantly, this effect was eliminated when participants had to engage in a task to prevent them using verbally based (i.e., semantic) information.
A second form of cue-driven imagery is where the cue comes from the olfactory modality. This can be conceptualized as a form of fragment completion, where the completion appears to draw upon recovering a memory of the whole. The following type of experiment serves as a good example of this. Participants are exposed to an odor-taste mixture such as lychee and sucrose. Later they smell lychee alone and are asked to judge its characteristics. They will typically report lychee as smelling sweet—a property of the taste system. This type of finding has been confirmed in several different studies and suggests that what people experience when they smell an odor is often composed of both stimulus-driven and memory-based components. This memory-based component can be conceptualized as an example of cue-driven imagery.
It is undoubtedly harder to imagine the odor of a lemon than it is to imagine the sight of one. Nonetheless, the evidence to date suggests that we do have some capacity to imagine smells. However, a contrarian perspective is also possible, whereby our performance on a real olfactory task can be affected by imagining a smell, but the imaginary smell may not always be consciously experienced. This alternative interpretation would require researchers to ask why certain people can consciously imagine smells and others cannot.