marisa carrasco and joseph b. ridout - new york …...288 marisa carrasco and joseph b. ridout is,...

Journal of Experimental Psychology:Human Perception and Performance1993, Vol. 19, No. 2,287-301

Copyright 1993 by the American Psychological Association, Inc.0096-1523/93/$3.00

Olfactory Perception and Olfactory Imagery:A Multidimensional AnalysisMarisa Carrasco and Joseph B. Ridout

Whether a system of imagery for olfaction exists is currently an unsettled issue. Moreover, thedimensions underlying odor perception have eluded researchers for many years. Two experimentsbearing on these issues are presented. In one experiment, a group of 32 undergraduates rated thesimilarity of pairs of 16 commonplace odorants (e.g., chocolate and leather) they perceived usingscratch and sniff stimuli; in another, a different group of 44 undergraduates was asked to imagineand then rate the similarity of the same pairs of odors. Multidimensional scaling of the datasuggests that three-dimensional solutions with similar stimulus dimensions, such as fruitiness,strength, and familiarity, underlied the ratings of both perceived and imagined odors. By findingthat similar dimensions define the psychological space of both the imagery and the perceptiontasks, this study suggests that imagery does indeed exist for olfaction.

Many have remarked on the uncanny ability of smells torevive long-forgotten memories (e.g., Engen, 1987; Laird,1935). Some researchers have shown a relationship betweenodor and emotion experimentally (e.g., Ehrlichman & Halp-ern, 1988; Hvastja & Zanuttini, 1989; Kirk-Smith, Van Tol-ler, & Dodd, 1983). Neuroanatomically, compared with theother senses, the olfactory system is not nearly as connectedwith the neocortex and has many more direct connectionswith the limbic system, where emotion and affects are reg-ulated (e.g., Kandel & Schwartz, 1985). This and other factshave given rise to a debate over whether olfaction is fun-damentally different from the other sensory modalities.There are somewhat contradictory findings with regard tothe similarities and differences between olfaction and theother senses. Furthermore, whereas for some modalities theexistence of a corresponding imagery system is acknowl-edged, the existence of olfactory imagery is still an unsettledissue.

The experiments reported in this article address threemain issues. First, if olfaction is similar to other sensorymodalities, the characteristics or dimensions of stimuli per-tinent to odor perception should be identifiable, as they havebeen for other modalities. Nonetheless, this fundamentalgoal of olfactory research has been elusive, in part, no doubtbecause of the relative paucity of our vocabulary for smells.To circumvent this vocabulary problem, in one experimentwe made use of multidimensional scaling (MDS) in anattempt to identify the dimensions that emerge when a group

Marisa Carrasco and Joseph Ridout, Department of Psychology,Wesleyan University.

This study was funded by a Faculty Project Grant of WesleyanUniversity awarded to the Marisa Carrasco.

We thank John Seamon, Jim Fernandez, Harry Lawless, and ananonymous reviewer for helpful comments on drafts of this article,and Svetlana Katz for her useful assistance at different stages ofthis study.

Correspondence concerning this article should be addressed toMarisa Carrasco, Department of Psychology, Wesleyan University,Middletown, Connecticut 06459-0408. Electronic mail may besent to mcarrasco.wesleyan.

of subjects perceive and compare a broad sample of odors.A second unresolved issue is whether olfactory imageryexists. To obtain data on this issue, in another experimentwith another group of subjects we again used MDS to elicitthe relevant dimensions in an olfactory imagery task. Third,by comparing the results of these two experiments, we wereable to study whether common dimensions underlie oursubjects' perception and imagery of odors; this is relevant tothe question concerning the (im)possibility of imagining("olfactizing") odors.

Perception

Recent work on odor perception has begun to emphasizethe similarities between the sense of smell and other senses.For example, whereas it was previously believed that odorperception possessed a differential threshold (JND) of about25%, recent findings suggest that this belief was mainly dueto physical noise of the stimuli produced by random fluc-tuations in the vapor-phase concentrations and not to aninherent inability to discriminate between odors. Using anolfactometer, the difference threshold has been found to beabout 11%, which is comparable to that of brightness invision and loudness in audition (Cain, 1977a). It should benoted, however, that odorant type also influences the differ-ential thresholds (see Cain, 1977a, 1977b; Stone & Bosley,1965; Wenzel, 1949). Olfaction is also similar to the visualand auditory modalities with regard to age-related changes:Olfactory sensitivity deteriorates with age (e.g., Cain & J.Stevens, 1989; J. Stevens & Cain, 1987), and the effect ofage becomes more pronounced through repeated testing(Cain & Gent, 1991). Peak performance in olfactory iden-tification occurs between the third and fourth decades oflife, and the overall age decay curves for this ability areanalogous to those of vision and audition (Doty et al., 1984).

Likewise, olfactory memory was thought to differ fromother forms of human memory. For instance, it was consid-ered to be immune to interference from backward countingtasks, to lack a short-term memory store (Engen, Kuisma, &Eimas, 1973), and to have a unique flat forgetting curve, that

287

288 MARISA CARRASCO AND JOSEPH B. RIDOUT

is, poor initial scores and high retention over time (Engen &Ross, 1973). Recent findings, however, indicate that distrac-tor odors interpolated between odor learning and testingdecrease memory performance (Walk & Johns, 1984), ashort-term store for odor intensity seems to exist (Barker &Weaver, 1983), and the forgetting curves of both odors andambiguous, undefined shapes are similar (Lawless, 1978).For a recent review of odor memory, see Schab (1991).

Given that perception and memory for smells seem to besimilar to those of other modalities, identification of per-ceptual dimensions in the case of olfaction would be desir-able. Nonetheless, this task has proved to be vexing. Someof the pertinent dimensions of visual stimuli are known to bebrightness, hue, and saturation; of auditory stimuli, pitchand loudness. Even taste has a rough model (sweetness-bitterness and saltiness-sourness), with a possible neuralbasis for at least the sweetness-bitterness elements (Bou-dreau, Oravec, & White, 1981). However, even thoughmany researchers over the last century have proposedframeworks to relate similar odors to common chemicalproperties, there is still no structural model to account forthe varying characteristics of odors. Early classifications ofodors were proposed by Henning (1916) and by Crocker andHenderson (1927). The former considered the six elementsof his smell prism (flowery, putrid, fruity, spicy, tarry, andresinous) analogous to the building blocks of the primarycolors, and the latter proposed four characteristics: burned,acid, fragrant, and caprylic. Amoore (1970) proposed a ste-reochemical ("lock and key") schema as a model for themolecular basis of the seven "primary" odors. Althoughnone of these categorizations has achieved wide acceptance,it is generally agreed that stereochemistry plays at leastsome part in determining odor quality (Cain, 1988). Theseattempts to analytically reduce to a set of basic qualitieswhat is seen as "unitary" perceptual events (Engen & Ross,1973) have some support. The notion that primary odorsmay exist is strengthened by findings that certain individu-als suffer from specific anosmias; for example, the inabilityto detect a musklike odorant may be genetically determined(Wysocki & Beauchamp, 1984).

One of the most serious obstacles to the identification ofolfaction's dimensions is the fact that there exists no accept-ably reliable verbal classification scheme; there is a lack ofa universally endorsed system of odor classification (e.g.,Engen, 1982; Harper, Bate Smith, & Land, 1968; Lawless,1986, 1989; Levine & McBurney, 1983). Lawless (1989),however, has pointed out that consistencies found in a va-riety of applied studies, which used different odor evalua-tion methods and different stimuli sets, suggest that devel-oping a general system of classification is possible. Odorterms such as fruity, spicy, floral, and green have beenidentified in studies dealing with multivariate analyses ofperfumery compounds and terms (Chastrette, Elmouaffek,& Zakarya, 1986), cluster analysis of perfumery terms(Chastrette, Elmouaffek, & Sauvegrain, 1988), empiricallyestablished descriptive systems for fragrance evaluation ofhousehold products (Civille & Lawless, 1986), and factoranalysis of fragrance compounds (Jeltema & South-wick, 1986).

It is certain, though, that in contrast to visual stimuli,which are described by a rich vocabulary, odor stimuli aredescribed by a rather limited set of adjectives. As Engen(1987) has commented:

Considering this difficulty in naming, it is curious that odorclassification has dominated research in olfaction. The work-ing assumption is that stimuli cause sensations of certain qual-ities which subjects can describe and categorize. What isbasically wrong with this approach is its emphasis on words,which have but a tenuous connection to odor perception,(p. 498)

One way to avoid this problem is to use MDS techniques.Because MDS analyses are not dependent on verbal input(e.g., Kruskal & Wish, 1978; Young & Hamer, 1987), theyare an ideal application to olfactory data. No judgmentcriteria are specified so as to encourage a more preciserecreation of the way in which odors occupy the "psycho-logical space" in a particular subject's perception. Instead ofsemantic responses, MDS uses similarity judgments to or-ganize the position of each stimulus in a spatial configura-tion. Experimenter bias is eliminated, and unanticipated cri-teria are allowed to emerge. Of course, we are well awarethat the psychological space derived from a given experi-ment is affected by the sample of stimuli used as well as bythe expertise of the subjects. After the psychological spacehas been outlined, experimenters are free to interpret thedimensions as they see fit, and other researchers may rein-terpret them, because any given interpretation is only a bestestimate.

Previous studies that have used MDS to study odor qual-ity have found the following dimensions to be relevant:fruity (Moskowitz & Barbe, 1976; Schiffman, 1974a,1974b), foodlike-nonfoodlike (Moskowitz & Barbe, 1976),sharpness or burning (Schiffman, Robinson, & Erickson,1977), pleasantness-unpleasantness (B. Berglund, Ber-glund, Engen, & Ekman, 1973; Jones, Roberts, & Holman,1978; Schiffman, 1974a, 1974b; Schiffman et al., 1977;Woskow, 1968), spiciness (Schiffman, 1974a, 1974b)woody, citrus, and intensity (Lawless, 1989). Some of thesestudies, however, have used sets of odorants with obviousgroupings between them. For instance, in Moskowitz andBarbe's study, all of the odorants were household foododors, and half of them were either fruits or meats, and inLawless's study each odorant was selected a priori on thebasis of its proximity to a woody or citrus odor quality. Thedimensions gleaned from such experiments seem restrictedto these limited odor sets. The use of a broader range ofodors could provide more general dimensions relevant toolfactory classification; such is a purpose of this study.

Whereas many of the experiments that have used MDShave aimed to identify trends leading to the eventual isola-tion of the physicochemical building blocks of odor quality,this study is not concerned with the identification of chem-ically mediating factors in odor judgments. Rather, by usinga relatively heterogeneous stimuli set, the degree to whichthe dimensions underlying a group of subjects' odor per-ception could be explained by odor attribute ratings (e.g.,pleasantness, familiarity, spiciness) was explored.

OLFACTORY PERCEPTION AND OLFACTORY IMAGERY 289

Perception and Imagery

Most general theories of imagery acknowledge the exist-ence of imagery systems for modalities other than vision(e.g., Farah, 1989b; Paivio, 1986; Shepard, 1975). Althoughresearch has not always supported the existence of an ol-factory image system (Engen, 1982; Schab, 1990), someresearch suggests that such a system may exist (Algom &Cain, 1990, 1991; Andreani, 1988; Crowder & Schab, inpress; Lyman & McDaniel, 1986, 1990; McDaniel, 1988).Lynian and McDaniel (1986) suggested that the internalrepresentations of odors are stored in an independent olfac-tory imagery system that reflects the perceptual features ofthe stimuli. More recently, they have found that subjectswho used olfactory imagery performed better on an olfac-tory recognition test than those who used visual imagery,and inversely, those who used visual imagery performedbetter on a picture recognition task (Lyman & McDaniel,1990). These researchers attributed the transfer-appropriateeffects they found to the existence of "a functional equiva-lence between imagery and perception" (p. 662) and inter-preted their results as support for the idea of modality-specific imagery systems (e.g., Brooks, 1967, 1968; Farah,1989b; Segal & Fusella, 1970); moreover, they extended theevidence for this modality-specific imagery system to theolfactory domain.

In a recent article, Schab (1991) has stated that "there isno unambiguous empirical evidence supporting an ability toimage odors" (p. 242). However, he does consider somefindings to be consistent with an ability to image odors. Forinstance, odor memory is impaired following odor imageryinstructions for distractor stimuli (Crowder & Schab, inpress), and psychophysical power functions for perceivedand remembered odor intensities are parallel; that is, theymap onto physical stimuli in the same way, supporting theview that memory and imagery processes may mirror per-ceptual processes (Algom & Cain, 1990, 1991).

Much of the recent work on imagery has focused on thesimilarities between perception and imagery. Imagery isfunctionally equivalent to perception to the extent that sim-ilar mechanisms in the visual system are activated whetherobjects or events are imagined or actually perceived (Farah,1989a, 1989b; Finke, 1989). Shepard (1975) reported thatsimilarity judgments between real and imagined objects arefound to be essentially the same whether the objects areactually presented or only named. This was the case, forinstance, for shapes of 15 of the states of the United Statesthat were compared using MDS (Shepard & Chipman,1970), as well as for photographs of faces of well-knownpolitical figures (Gordon & Hay ward, 1973).

Though most of the research has been within the visualmodality, it is plausible that a close perception-imageryrelationship could be present in other modalities as well.The use of MDS in the comparison of the underlying struc-tures of perceived and imagined stimuli has been recom-mended (Finke & Shepard, 1986). Studying whether com-mon dimensions underlie perception and imagery is relevantto the question concerning the (im)possibility of imaginingodors. MDS provides an ideal means of testing the relation-ship between olfactory perception and olfactory imagery,

because it allows for a comparison of the interstimulusspace in both conditions. This study explored the relation-ship between olfactory perception and olfactory imagery.Subjects made similarity judgments of pairs of perceived orimagined odorants. Similarity judgments are the primarymeans for recovering the underlying structure of relation-ships among a group of stimuli (Schiffman et al., 1981).They are also considered psychologically fundamental andless prone to biases resulting from language (e.g., Carroll &Wish, 1974; Tversky, 1977). Using MDS, we obtained boththe psychological space that describes one group of sub-jects' perception of the stimuli (Experiment 1) and the psy-chological space describing another group's imaging ofthose same stimuli (Experiment 2); finally the two MDSspaces were compared. It stands to reason that if subjects areable to imagine the odors, a correspondence would be foundbetween the psychological spaces derived from the similar-ity ratings of the perceived odors and those of the imaginedodors. Because previous MDS research has shown femalesubjects to be more consistent than male subjects in odorjudgments (Gilbert, Greenberg, & Beauchamp, 1989), wealso hypothesized that the female group may be more con-sistent in both the perceptual and the imagery conditions.

Experiment 1

The purpose of this experiment was to obtain the psycho-logical space within which a given group of subjects orga-nize the odorants they perceived. In Phase 1, to assist ininterpretation of the psychological space derived from thesimilarity judgments of Phase 2, a group of subjects ratedeach of the physical odorants used in terms of 12 descrip-tors. In Phase 2 a second group of subjects made similarityjudgments of pairs of perceived odorants, and MDS wasused to investigate what dimensions were relevant.

Phase 1

To assist in interpretation of the psychological space de-rived from the similarity judgments, ratings of 12 descrip-tors for each of the 16 odorants used in Phase 2 wereobtained.

Method

Subjects. Thirty Wesley an University undergraduates (17women and 13 men) participated in fulfillment of a requirementfor an introductory psychology course. Subjects were nonsmokersand had not recently been ill. The subjects were naive as to thepurpose of the experiment.

Materials. Subjects were given 16 odorants, microencapsu-lated strips chosen from among those used in a "scratch andsniff' test (University of Pennsylvania Smell Identification Test[UPSIT], commercially termed the Smell Identification Test, Sen-sonics, Inc., Haddonfield, NJ): bubble gum, menthol, cherry,mint, banana, clove, leather, coconut, fruit punch, licorice, ched-dar cheese, gasoline, strawberry, cedar, chocolate, and ginger-bread. There is a large body of normative data regarding the char-acteristics of these stimuli, which include ratings of intensity(strength), familiarity, irritation, pleasantness, and coolness, for


both women and men of different age groups (Doty, Shaman, &Dann, 1984).'

Subjects were also given 16 lists (one per odorant) that con-tained 12 descriptors in a randomized order: familiarity, pleasant-ness, frequency of exposure, recency, sharpness, strength, irrita-tion, warm-cool, foodlike, fruitiness, spiciness, and the locationof the referent object (the likelihood of it being found inside-outside the home). Some of these descriptors are nonsensorial(e.g., familiarity), but because they play an important role in odorrecognition, and given that early efforts to find the perceptual di-mensions for olfaction have not been all that fruitful, these de-scriptors were included to allow for common dimensions toemerge in both conditions, perception and imagery.

Familiarity has been considered the most basic level of odorrecognition (Schab, 1991): It is related to more accurate identifi-cation (Desor & Beauchamp, 1974; Doty et al., 1984; Goldman& Seamon, 1992), quicker association to an odor (Lawless & En-gen, 1977), slower forgetting (Rabin & Cain, 1984), better qual-ity of label, which in turn leads to more accurate identification(Rabin, 1988; Rabin & Cain, 1984), and greater ability to link theodor in a paired association task (Davis, 1975); pleasantness waschosen as a descriptor term based on its successful use in otherodor MDS research (e.g., B. Berglund et al., 1973; Jones et al.,1978; Schiffman, 1974a, 1974b; Schiffman et al., 1977; Woskow,1968), and frequency and recency were chosen because it wasthought that they might play a mediating role in either or both ofthese other two attributes. Sharpness, strength, and irritation wereincluded so as to provide a good semantic approximation of stim-ulation of the common chemical sense, the "tickling" or "burn-ing" nasal experience mediated by the trigeminal nerve, which isrelated to touch and pain as well as to olfactory sensation.2 Eventhough its neural pathways are essentially dissociated from theolfactory nerve, the trigeminal nerve is consistently found to playan important role in olfactory judgment (cf. Cain, 1988; Engen,1982); furthermore, sharpness-burning was found to be a relevantodor quality dimension (Schiffman et al., 1977), and perceptualdata for strength, irritation, and warm-cool have been compiledby Doty et al. (1984). Intensity or strength has been found to bea relevant dimension for fragrance categories and ambiguousodors (Lawless, 1989), as well as for flavor perception of fruitsand vegetables (D. Stevens & Lawless, 1981). The attributes offoodlike (Moskowitz & Barbe, 1976), fruitiness (Moskowitz &Barbe, 1976; Schiffman, 1974a, 1974b), and spiciness (Schiff-man, 1974a, 1974b) had each been suggested as dimensions byprevious MDS studies. Fruitiness and spiciness have also beenidentified in a series of applied studies (Chastrette et al., 1988;Chastrette et al., 1986; Civille & Lawless, 1986; Jeltema &Southwick, 1986; Lawless, 1989). The "location" of the odor-emitting objects (inside-outside the home) was used to test theproposition that "in general, odor perception is not organized bynouns but around the similarity of objects causing odors and es-pecially the contexts in which odors usually occur"(Engen, 1987, p. 502), as well as the idea that the natural func-tion of our olfactory experience is to identify the sources of indi-vidual odors, not the individual odors per se (Richardson &Zucco, 1989).

Procedure. Participants were instructed to scratch and sniffeach odorant before rating it on the basis of each of the attributesof the list. For each odorant, subjects were instructed to indicatehow they found the to-be-smelled odorant, using a scalefrom 1 to 7, with 1 being the least and 7 being the most

. Both the order of the 16 odorants and the order of the 12ratings of each list were randomized for each subject. Subjectsspent about 3^4 mm completing each list and rested for 1 min be-fore proceeding to smell the following odorant. There were three4-min resting periods after 4, 8, and 12 odorants had been rated.

After the subjects had finished rating all the odorants, they wereasked to identify them in the following manner. Each subject wasgiven a bag containing a complete set of the 16 odorants (whichwere coded with a number from 1-16) and a 16-slot fill-in-the-blank form. The subject randomly drew each stimulus from thebag, smelled it, and then wrote an answer in the correspondingblank of this form.

Results

To see whether our data replicated previous perceptualratings obtained with the same UPSIT stimuli, we comparedour ratings with a second group of measures drawn from astudy conducted by Doty et al. (1984, indicated by the suffixD): familiarityD, irritationD, p!easantnessD, strengthD, andwarm-coolD. Both estimates of all attributes correlated sig-nificantly: familiarity (r = .67, p < .005), irritation (r =.90, p < .001), pleasantness (r = .91, p < .001), strength(r = .50, p < .05), and warm-cool (r = .58, p < .02).Hence, the present results successfully replicated Doty etal.'s findings. Interattribute correlations including both setsof ratings appear in Table 1.

Among these perceptual ratings, the measures of irrita-tion, irritationD, pleasantness, pleasantnessD, fruitiness,foodlike, and sharpness all show close correspondence (p <.05, except for foodlike-sharpness). These measures mayrepresent a more general underlying quality. Although pre-vious studies have found familiarity and pleasantness tocorrelate significantly (Doty et al., 1984; Engen & Ross,1973), in the present experiment no significant relationshipemerged.

The odorants were identified correctly by the followingnumbers of subjects (out of 30): gingerbread and fruitpunch, 1; leather, 4; gasoline, 5; cedar and banana, 8; straw-berry, 9; cherry, 11; bubble gum, clove, coconut, and ched-dar cheese, 12; menthol, 13; chocolate and licorice, 22; andmint, 24 (x = 11; 8 = 6.9). Thus, even though all subjectshad already performed 16 ratings on each odorant, on av-erage each odorant was identified by only a third of them. Ithas been shown previously that people are not good atnaming even very familiar odors (e.g., Engen, 1987). Dur-ing the debrief period, we learned that while attempting toidentify the stimuli practically all subjects experienced the"tip of the nose" phenomenon, that is, knowing that an odoris familiar but being unable to name it (Lawless & Engen,1977).

1 All stimuli are familiar; they are used to quantitatively assesssmell function in a standardized manner. Only stimuli that havebeen correctly identified by the majority of a large number ofnormal subjects are included in this smell identification test (Sen-sonics, Inc., Haddonfield, NJ)

2 The trigeminal (or fifth cranial) nerve is anatomically distinctfrom the first (or olfactory) cranial nerve. It is composed of freenerve-endings that are activated by substances that irritate mucousmembranes (such as the eyes or the nose). Certain substancesderive their distinctive effects from stimulation of the trigeminalnerve (the "cool" sensation of menthol is one example).


•a,o

D t^

3 'S£3

es

a.

•E-S~

W "O

£.8

-8.

e.i

s >>cd

CL,


Phase 2

Subjects made similarity judgments of the odorants; MDSwas used to investigate what dimensions were relevant forthe perceived odors used here. The ratings gathered fromPhase 1 aided in the identification of dimensions.

Method

Subjects. Thirty-two Wesleyan University undergraduates (16women and 16 men) participated in fulfillment of a course re-quirement for introductory psychology. Subjects were nonsmok-ers and had not recently been ill. The subjects had not partici-pated in Phase 1 and were naive as to the purpose of theexperiment.

Materials. The same 16 odorants used in Phase 1 were usedin this phase.

Procedure. Subjects were told that the experiment involvedsmelling various odorants and rating them in terms of similarity.All 120 nonduplicate pairings of the stimuli were presented, indifferent random order for each subject, every 18-20 s. Subjectswere given the two stimuli of each pair and asked to scratch andsniff them, one after the other. The pairs were rated on the basisof similarity judgments, on a scale of 1 (most dissimilar) to 7(most similar). Subjects were told that they could use whatevercriteria they wished in order to make their similarity judgments;in the visual domain, for instance, the ratings might be made interms of size, color, use of the object, shape, and so forth. Someexamples were provided to illustrate this scale; for instance, abook and a notebook would have a higher similarity rating than abook and a boat, an elephant and a chicken would be more simi-lar than an elephant and an asparagus. There were four 4-minresting periods after 24, 48, 72, and 96 trials had been completed.

Results

To create an MDS spatial configuration, pair similarityjudgments were analyzed using the SPSS version of theALSCAL program (Young, Takane, & Lewyckyj, 1980),which assumes that subjects will vary in the importance theyassign to each criterion (Young & Hamer, 1987). A nonmet-ric, weighted Euclidean distance model was specified, alongwith ordinal, matrix conditional, untied data. Two-, three-,and four-dimensional stress were .16, .10, and .06, and therespective R2 values were .87, .93, and .96. Stress measuresthe square root of the residual variance from the differencein the fit between the model and the scaled data (the "bad-ness of fit"), and R2 represents the squared correlation indistances that measures the proportion of variance of thedata in the matrix. Based on both stress and R2 values, threedimensions were selected as the most appropriate solution;they are shown in Figure 1, and the two-dimensional com-ponents of the graph are broken down in Figure 2, where thetop panel depicts Dimensions 1 and 2 and the bottom paneldepicts Dimensions 2 and 3.

A multiple regression analysis was performed to aid ininterpretation of the MDS configuration (e.g., Kruskal &Wish, 1978; Lawless, 1989; D. Stevens & Lawless, 1981).The mean attribute ratings for each perceived odorant ob-tained in Phase 1 served as dependent variables and wereregressed against the MDS coordinates. Table 2 shows themultiple correlation coefficients and the optimum regression

DM2

stmbeiry

OM1

Figure I . A three-dimensional space representation of the loca-tion of perceived odors (Experiment 1) based on multidimensionalscaling similarity judgments. (Dim denotes dimension.)

weights corresponding to each multiple correlation (direc-tion cosines; the regression coefficients are normalized sothat their sum of squares equals 1.0 for every scale). Thesedirection cosines represent the angles at which vectors cor-responding to the best fitting projection of the attributeratings map or scale onto the MDS models. The vectorsuseful in interpreting the solutions were chosen based on thesignificance of the multiple correlation, the significance ofthe individual regression weights, and the orthogonality toother vectors; the chosen vectors are plotted in Figure 2.

The three MDS dimensions correlated well with the per-ceived adjective ratings. Dimension 1 highly correlated withthe perceptual attributes of fruitiness followed by pleasant-ness, which are significantly correlated with each other(Table 1), as well as with irritation, sharpness, and foodlike;the second group of attributes, however, did not achieveorthogonality with the other vectors. Dimension 1, thus, wasbest characterized as fruitiness and pleasantness. As can beseen in Figure 2a, the fruits and fruit-related odorants(strawberry, cherry, banana, bubble gum, and fruit punch)cluster at the positive extreme of the fruitiness vector,whereas the nonfruit-related items (e.g., cedar, leather, mint,gasoline) are at the negative end of this vector. In a similarway, the fruits and other items such as chocolate and coco-nut are at the positive end of the pleasantness vector, anditems such as mint, cedar, licorice, and clove are at thenegative end of this vector.

The best fit of Dimension 2 with the perceptual attributeswas accomplished by strength followed by warm-cool andsharpness; this last attribute was not orthogonal with Di-mension 1. Hence, Dimension 2 was best represented bystrength. As can be seen in Figure 2a, licorice, mint, clove,and gingerbread are at one end of this vector, whereasgasoline, leather, and coconut are at the other end.

Dimension 3 highly correlated with the perceptual at-tributes of familiarity, followed by spiciness, frequency, andinside-outside; foodlike and irritation, which also correlated


Fig2a

DIM 2

Strength

/

DIM1 0-

^ n*X •

X

-.leathergasoline

licorij

xctove

X~giOgeAre

menthol

tax.

1 chc

1

f fruK punch bubble

* ' ian^ -». •

X cRecHet. •<*lel

Xx '"**'"

3nut x

X

fdate

FruftliMM

Fig2bDIM 3

^

1-

DIM 2 0 -

-1 -

3-

X \

Sx x ™

leainer v ^

strawbefKcoconut •

cherry

chocolate• gingei

ithol

tgjlt punch

cedarbanana *'bubbteaum

\ X mint. X • .^ X licorice\ dgveX^

5read\ X\ Xi

0 \1 2\

Familiarity

SplclrwM

Figure 2. A representation of the location of perceived odors(Experiment 1), based on multidimensional scaling similarity judg-ments, in two-dimensional spaces. (Panel a depicts Dimension 1and Dimension 2. Panel b depicts Dimension 2 and Dimension 3.These dimensions were best described by the perceptual attributesof fruitiness, pleasantness, strength, familiarity, and spiciness. Dimdenotes dimension.)

with Dimension 3, did not attain orthogonality with Dimen-sion 1. Dimension 3 was then best characterized as famil-iarity and spiciness. At the positive end of the familiarityvector (Figure 2b), one can see licorice and chocolate to-gether with a cluster of spicelike odorants (mint, clove, andgingerbread), whereas at the negative end of this vector, wefind odorants like gasoline, cheddar, leather, and menthol.Except for chocolate, the same odorants are found at therespective ends along the spiciness vector.

The analysis performed on each individual subject matrix,using the SPSS version of INDSCAL (Young et al., 1980),yielded the same three dimensions as the group matrix anal-ysis, plus the overall importance of each dimension: Dimen-sion 1 = 40%, Dimension 2 = 35%, and Dimension 3 =25%. In addition to having the best fit to an axis, fruitinesswas the greatest determinant of subjects' psychologicalspace. Analysis of the separated women's and men's datawithin INDSCAL showed no significant differences either

in stress or R2 values, f(15) = 0.19, ns, and f(15) = 0.63, ns,indicating that women did not achieve greater consistencyin their judgments of odor perception.

Discussion

The attribute-based correlations suggested dimensionsof perceived fruitiness, pleasantness, strength, familiarity,and spiciness underlying judgments of perceived odor sim-ilarity. Of these dimensions, perceived fruitiness had thebest fit to an axis. The identification of these dimensionscorresponds to some previous MDS findings: Fruitiness isone of the dimensions that has been identified in MDSstudies (Moskowitz & Barbe, 1976; Schiffman, 1974a,1974b). In the applied literature, different methodologieshave yielded the conclusion that fruitiness is a basic term inodor classification (Chastrette et al., 1988; Chastrette et al.,1986; Civille & Lawless, 1986; Jeltema & Southwick, 1986;Lawless, 1989). Pleasantness has been identified in severalstudies (B. Berglund et al., 1973; Jones et al., 1978; Schiff-man, 1974a, 1974b; Schiffman et al., 1977; Woskow, 1968).Intensity or strength of the odorants is one of the dimensionsthat has been found to underlie MDS scaling of fragrancecategories and ambiguous odors (Lawless, 1989), as well asflavor perception of fruits and vegetables (D. Stevens &Lawless, 1981). Although the role of familiarity has notbeen previously examined in an MDS context, there is alarge body of evidence for an influence of familiarity onodor judgments, recognition, and identification (e.g., Desor& Beauchamp, 1974; Doty et al., 1984; Goldman &Seamon, 1992; Lawless & Engen, 1977; Rabin, 1988; Rabin& Cain, 1984; Schab, 1991). Spiciness has also emerged inanother MDS study (Schiffman, 1974a, 1974b) and, likefruitiness, it has been convergently identified as a basic termin a variety of odor classification studies using diversemethodologies (Chastrette et al., 1988; Chastrette et al.,1986; Civille & Lawless, 1986; Jeltema & Southwick, 1986;Lawless, 1989).

Experiment 2

This experiment was conducted to explore the relevantdimensions in an imagery condition of olfactory qualityjudgments. There were two phases analogous to those ofExperiment 1: In Phase 1, one group of subjects rated eachof the nominal odorants in terms of 12 descriptors. In Phase2, another group of subjects made similarity judgments ofnominal pairs of odorants, to explore what dimensions arerelevant for imagined odors, as well as to test whether thesame dimensions seem to underlie odor judgments underboth perceived and imagined conditions.

Phase 1

To assist in interpretation of the psychological space de-rived from the similarity judgments, ratings of the 12 de-scriptors were obtained for each of the 16 imagined stimuli,corresponding to those used in Experiment 1.


Table 2Multiple Regression Coefficients and Regression Weights (Direction Cosines) for Predicting Attribute Ratings FromMultidimensional Scaling Dimensions in Experiment 1 (Similarity of Perceived Odors) and Experiment 2 (Similarityof Imagined Odors)

Experiment 1 Experiment 2

Attribute Multiple Rs Dimension 1 Dimension 2 Dimension 3 Multiple Rs Dimension 1 Dimension 2 Dimension 3

FamiliarityFoodlikeFrequencyFruitinessInside-outsideIrritationPleasantnessRecencySharpnessSpicinessStrengthWarm-cool

.70*94**.62*.96**.69*.86**.87**.57.91**.86**.80**.71*

.09

.63*

.47

.97**

.47-.73**

.78**

.42-.70*-.29-.41

.14

.39

.24

.22

.10

.40

.07

.11

.12

.70*

.61

.87**-.80**

-.92**-.74*-.86**

.24-.79**

.68*-.61-.90**

.18-.74*-.29-.59

-.71*.79**.46.73*.50.71*.47.43.76**.78**.82**.59

-.20.71*.46.96**

-.57-.55

.17-.75**_79**-.96**-.60-.42

-.61.64*

-.04-.09-.70*-.76**

.98**

.43-.51-.29-.71*

.71*

.77**-.29

.89**-.28

.42

.35-.10-.50

.35-.06

.36

.56*p<.01. **/><.001.

Method

Subjects. Fourteen volunteer Wesleyan University undergrad-uates, who had not participated in Experiment 1, and who werenaive as to the purpose of the study, served as subjects.

Materials. The odorants and the attributes were the same asthose used in Experiment 1: bubble gum, menthol, cherry, mint,banana, clove, leather, coconut, fruit punch, licorice, Cheddarcheese, gasoline, strawberry, cedar, chocolate, and gingerbread.Subjects were given 12 lists in a random order, one per attribute.Each list had the name of 16 odorants printed in a random order.Subjects were asked to rate each odorant on the basis of the 12attributes.

Procedure. Participants were instructed to imagine the smellof each stimulus and to rate the imagined odorant according tothe specified attribute measure. The experimenter read aloud thelist of odorants and asked subjects to report whether they werefamiliar with the stimuli; all subjects were familiar with all theodorants used. Subjects were told that to imagine-does not alwaysmean to visualize; rather, in this particular case, we wanted themto recreate the experience of smelling (because no specific wordexists to refer to this recreation of the original experience ofsmelling something, we drew the analogy between visualizationand olfactization and instructed the subjects to attempt to performthe latter). Rating scales ranged from 1 (least) to 7 (most). Boththe order of the 12 attribute lists and the order of the names ofthe 16 odorants were randomized for each subject. Subjects weretold to spend about 4-5 min completing each list. There werethree 4-min resting periods after 3, 6, and 9 lists had been rated.

Results

Because no previous ratings for imagery had been ob-tained for the attributes used in this study, and for purposesof reliability, ratings were obtained from another group of15 students for 4 of the 12 descriptors: familiarity, fre-quency, pleasantness, and recency. High correlations werefound for these four attributes, ranging from .81 for famil-iarity to .94 for frequency (all p < .001).

Rank correlations for the perceptual (Experiment 1, PhaseI) and imagery attribute ratings were obtained: The highestcorrelations were achieved by irritation and sharpness (p <

.001), followed by fruitiness and spiciness (p < .002),inside-outside and warm-cool (p < .005), and foodlike (p< .05). Pleasantness, strength, recency, frequency, and fa-miliarity did not reach significance; whereas the first two ofthese attributes fell just short of significance, the last threewere far from being significant. The perceptual ratings fromDoty et al.'s (1984) study followed the same patterns as thecorresponding ratings from this study; that is, irritationD (p< .001) and warm-coolo (p < .005) significantly corre-lated with the respective imagery ratings, whereas pleasant-nessD, strengthD, and familiarityD did not reach signifi-cance. This was expected in light of the high correlationsachieved when comparing the attribute ratings of Experi-ment 1 with Doty's ratings (see above).

Interattribute correlations are presented in Table 3.Among the imagery ratings, the measures of irritation,strength, spiciness, and sharpness all showed close corre-spondence (p < .005, all but spiciness-strength, p < .10).Fruitiness and foodlike are also highly related to these mea-sures; perhaps they all represent a more general underlyingquality.

Phase 2

Subjects made similarity judgments of all possible pairsof 16 imagined odorants; MDS was used to obtain thegroups' psychological space of the odors they imagined. Theratings gathered from Phase 1 aided in the identification ofdimensions.

Method

Subjects. Forty-four Wesleyan University undergraduates (23women and 21 men) participated to obtain credit for an introduc-tory psychology course. They had not participated in the previousphases of this study and were naive as to the purpose of this ex-periment.

Materials. A list of the names (no physical stimuli) of the 16odorants used in Experiment 1 and in Phase 1 of this experimentwas used for the imagery task.


Table 3Attribute Intercorrelations for Imagined Ratings: Experiment 2, Phase 1

Attribute

AttributeFamil- Food- Fre- Fruit- Inside- Irrit- Pleasant- Sharp- Spici- Warm-iarity like quency iness outside ability ness Recency ness ness Strength cool

FamiliarityFoodlikeFrequencyFruitinessInside-outsideIrritationPleasantnessRecencySharpnessSpicinessStrengthWarm-cool

-.28.49

-.22.33.21.02

-.34.17.00.51*.23

.06

.38-.77***-.63**

.34-.20-.58*-.31-.72**

.05

-.20.28

-.18.21

-.87***-.19-.37

.22

.21

-.13-.64**

.27-.04-.64**-.48-.60*-.37

—.34

-.06-.07

.34

.07

.54*-.11

—-.57*

.36

.89***

.72**

.70**-.04

—-.19-.46-.21-.17

.64**

—.38.50*

-.07-.18

—.83*** —.68** .46 —

-.08 -.01 .25 —*p<.os. **p<m. ***p<.ooi.

Procedure. Subjects were read the names of pairs of odorantsand instructed to rate the pairs in terms of similarity of the imag-ined smells, on a scale of 1 (most dissimilar) to 7 (most similar).Instructions and examples given to subjects were the same as inExperiment 1. As in Phase 1, subjects were told that to imaginedoes not always mean to visualize; rather, in this particular case,we wanted them to recreate the experience of smelling (to olfac-tize the odors). After having ascertained that the subjects were fa-miliar with the odorants, the to-be-imagined odorants were readaloud in pairs by the experimenter (e.g., cedar-gasoline, mint-banana), with each pair approximately 12 s apart. The order of all120 nonduplicate pairing combinations was randomized in fourdifferent ways, one for each group of 10-12 subjects. As in Ex-periment 1, there were four 4-min resting periods after 24, 48, 72,and 96 trials had been completed.

Results

Specifications for the ALSCAL program were identical tothose of the previous experiment. The two-, three-, andfour-dimensional alternatives yielded Kruskal's Stress For-mula 1 values of .31, .17, and .14, and the respective R2

values were .40, .61, and .70. Based on both the stress andthe R2 values, the three-dimensional representation was cho-sen as the optimum solution; it is shown in Figure 3, and thetwo-dimensional components of the graph are broken downin Figure 4, where the top panel depicts Dimensions 1 and2 and the bottom panel depicts Dimensions 2 and 3.

As in Experiment 1, a multiple regression analysis wasperformed to aid in interpretation of the MDS configuration(e.g., Kruskal & Wish, 1978; Lawless, 1989; D. Stevens &Lawless, 1981). The mean attribute ratings for each imag-ined odorant obtained in Phase 1 served as dependent vari-ables and were regressed against the MDS coordinates. Ta-ble 2 shows the multiple correlation coefficients and theoptimum regression weights corresponding to each multiplecorrelation. These direction cosines represent the angles atwhich vectors corresponding to the best fitting projection ofthe attribute ratings scale into the MDS models. The vectorsuseful in interpreting the solutions were chosen based on thesignificance of the multiple correlation, the significance ofthe individual regression weights, and the orthogonality to

other vectors; the chosen vectors are plotted in Figure 4.The three MDS dimensions correlated well with the imag-

ined adjective ratings. Dimension 1 correlated with fruiti-ness, spiciness, sharpness, and foodlike; this last adjectivewas not orthogonal to Dimension 2. Fruitiness and spicinessbest represent this dimension. Figure 4a illustrates that theimagined fruits (banana, strawberry, coconut) and the fruit-related imagined odorants (fruit punch and bubble gum)cluster at the positive end of the fruitiness vector, whereasthe nonfruit-related imagined odorants (leather, ginger-bread, cedar, and clove) are at the negative end of thisvector. For the spiciness vector, the spicelike imagined odor-ants (clove, mint, gingerbread, and menthol) as well asleather and cedar, which ranked high in this vector, are atone end, whereas the fruits (strawberry, coconut, banana,cherry) and chocolate are at the other end.

The best fit of Dimension 2 with the perceptual attributeswas accomplished by strength and irritation, as well asfoodlike; this last attribute was not orthogonal to the vector

DM3

DM2

DM1

Figure 3. A three-dimensional space representation of the loca-tion of imagined odors (Experiment 2) based on multidimensionalscaling similarity judgments. (Dim denotes dimension.)


Fig4a

DIM 2

Spteirwss -1

gingerbread

cedar

"̂ ~~ -~ licoriceleather -O^

..- ""*

mint

menthol

chocoiate

cherry

cheddar

COCOpUt-"

_^ •— * 8tr8wbetry

b»»%~.fgiH punch

bubble gumFrultl

Fig4b

DIM 3

DIM 2 0-

gasojne ^

bananamentnol

dove•

fruit punch

bubblMurnx 7

X/'//nit

/

"*•/,'r ''<'\f^

gingerbread .

codnu, chocotete

cherry

licorice

Familiarity

Strength, Irritation

Figure 4. A representation of the location of imagined odors(Experiment 2), based on multidimensional scaling similarity judg-ments, in two-dimensional spaces. (Panel a depicts Dimension 1and Dimension 2. Panel b depicts Dimension 2 and Dimension 3.These dimensions were best described by the imagined attributesof fruitiness, spiciness, strength, irritation, and familiarity. Dimdenotes dimension.)

in Dimension 1. These three attributes are highly intercor-related {Table 3). Dimension 2 was best characterized asstrength and irritation. Dimension 3 highly correlated withthe imagined attribute of familiarity, which thus character-izes this dimension. The imagined odorants that would beconsidered high in strength, irritation, and familiarity arelocated toward one end of these vectors, for example, thespices (clove, menthol, mint), bubble gum, and licorice.The other ends include those odorants that ranked low instrength, irritation, and familiarity, for example, cheddarcheese, chocolate, cedar, and strawberry (Figure 4b).

The analysis performed on each individual subject matrixusing the SPSS version of INDSCAL (Young et al., 1980)yielded the same three dimensions as the group matrix anal-ysis, plus the overall importance of each dimension: Dimen-sion 1 = 37%, Dimension 2 = 35%, and Dimension 3 =

28%. Analysis of the separated female and male data withinINDSCAL showed the female group as having lower stressand higher R2 values, r(20) = -4.32, p < .001, and ?(20) =4.19, p < .001, respectively. That is, when no physicalodorants were present, women were more consistent thanmen in their odor judgments. This finding corresponds to aprevious MDS odor perception study, in which women weremore consistent than men in their odor judgments (Gilbertet al., 1989), but not to Experiment 1 of this study, whichfound neither sex to be more consistent.

Discussion: Compared Dimensions of Perceptionand Imagery

The degree of similarity between the relevant dimensionsfrom the perception (fruitiness, pleasantness, strength, fa-miliarity, spiciness) and the imagery (fruitiness, spiciness,strength, irritation, familiarity) experiments was examined.Indeed, the similarity ratings for all possible pairs of odor-ants obtained for perception (Experiment 1, Phase 2) and forimagery (Experiment 2, Phase 2; see Shepard & Chipman,1970) were highly correlated (p < .001). Another measureof global configurational similarity that is independent ofscaling and rotation is that of interpoint distance correlation(Lawless, 1986), which was significant for both the simi-larity judgments (input data) and the MDS solutions (outputconfiguration) (r = .80, p < .001, and r = .52, p < .001,respectively). The perception condition was substantiallylower in group stress and higher in R2 values than theimagery condition; that is, imagery judgments were morediffuse than those performed in the perceptual context. Onthe other hand, the similar R2 values in the within-subjectsconditions indicate that subjects showed similarity in theirproportional use of existing dimensions.

How similar were the underlying dimensions for the per-ception and the imagery MDS spaces that emerged in thisstudy? In both conditions, the dimension that accounted forthe largest amount of variability of the data, Dimension 1, isrelated to the attribute of fruitiness. On the one hand, pleas-antness was relevant for the perceptual condition but not forthe imagery one; on the other hand, spiciness was relevantfor the imagery condition but not for the perceptual one. Amore direct comparison shows that Dimension 1 for per-ception and imagery strongly correlate (r = .80, p < .001).Notice that although there are some dissimilarities in Di-mension 1 of the groups' perception and the imagery psy-chological spaces, the general pattern is the same: In boththe perception (Figure 2a) and imagery (Figure 4a) condi-tions, banana, cherry, strawberry, bubble gum, fruit punch,cheddar cheese, coconut, and chocolate are clearly sepa-rated on the right half of the graph, and the other eightstimuli congregate on the other half. More specifically,leather, cedar, and clove are at the negative end, whereasbanana, strawberry, and fruit punch are at the positive end.Thus, in addition to both being described by the same ad-jective descriptor, the two Dimension 1 elements are similarto each other in a more direct fashion. Fruitiness seems to bea cross-conditional mediator of judgments.

The Dimension 2 for perception and that for imagery didnot significantly correlate with each other (p > .10). They


are, however, related to a common dimension, namelystrength. In both perception and imagery (Figures 2a and4b), chocolate and leather are at one end and licorice, mint,bubble gum, and fruit punch are at the other end of thisvector. Dimension 3 also showed significant consistencyacross both perception and imagery experiments. Dimen-sion 3 from the perception and from the imagery experi-ments correlated with each other (r = .61, p < .01) as wellas with their respective familiarity attribute. For the percep-tion experiment, spiciness was also a relevant dimension. Inboth dimensions, for instance, leather and cheddar cheeseare at one end, whereas mint and licorice are at the other endof these vectors (Figures 2b and 4b).

The attributes of fruitiness, strength, familiarity, and spic-iness, thus, are cross-conditional dimensions, and they rep-resent sources of consistency between perception and im-agery-based odor judgments. It is worth noting that irritation(imagery, Dimension 2) and pleasantness (perception, Di-mension 1) are interrelated in the perceptual ratings (p <.001; Table 1), the imagined ratings (p < .05; Table 3), andcross-conditionally (p < .01). It must be pointed out thatlabels assigned in MDS interpretations, though potentiallyof great explanatory value, are always best guesses and arenever applied with absolute certainty on the part of theresearchers (Young & Hamer, 1987).

General Discussion

Experiment 1 found that one group of subjects' perceptualjudgments of the odors we used seem to have been mediatedby the perceptual attributes of fruitiness, strength, familiar-ity, pleasantness, and spiciness. Experiment 2 found thatanother group of subjects' imagery judgments of odors seemto have been mediated by imagined attributes of fruitiness,strength, familiarity, irritation, and spiciness. In both exper-iments, fruitiness seems to have been the dominant dimen-sion. The results from both experiments correspond to di-mensions found in other MDS studies on odor perception.

The comparison of the perceptual and imagery multidi-mensional spaces obtained here points to ways in whichthey may be similar, as well as to the areas in which thereappears to be less correspondence. First we address thesimilarities. The most obvious similarity was the use offruitiness in both conditions. This attribute has appeared inprevious MDS analyses of odor perception (Moskowitz &Barbe, 1976; Schiffman, 1974a, 1974b), as well as in con-verging odor classification studies based on different meth-odologies (Lawless, 1989), such as fragrance evaluation(Civille & Lawless, 1986), factor analysis (Jeltema &Southwick, 1986), multivariate analysis (Chastrette et al.,1986), and cluster analysis (Chastrette et al., 1988), but thisis the first time such a dimension has been found in animagery context. The respective Dimension 1 for perceptionand imagery showed a close relationship to each other, aswell as to the attribute-based label they were assigned.

For Dimension 2 there was not a direct correlation be-tween perception and imagery conditions; however, bothwere described by strength. This is suggestive of a secondconsistent cross-conditional dimension. Intensity or strengthof the odorants is one of the dimensions that has been found

to underlie MDS scaling of fragrance categories and am-biguous odors (Lawless, 1989), as well as flavor perceptionof fruits and vegetables (D. Stevens & Lawless, 1981).Perceptual data for strength as well as for irritation, alsofound to be relevant for Dimension 2 in the imagery exper-iment, have been compiled by Doty et al. (1984).

Significant consistency across both perception and imag-ery experiments was also found in Dimension 3. Perceptualand imagery Dimension 3 not only significantly correlatedwith each other, but the same label was most applicable forboth: familiarity. This is suggestive of a third consistentcross-conditional dimension. A variety of studies provideevidence for an influence of familiarity on odor judgments,recognition, and identification (e.g., Desor & Beauchamp,1974, Doty et al., 1984; Goldman & Seamon, 1992; Lawless& Engen, 1977; Rabin, 1988; Rabin & Cain, 1984; Schab,1991), but its role had not been previously identified in anMDS context, neither in a perceptual condition nor in animagery one.

Whereas in the imagery experiment pleasantness did notemerge as a significant dimension, in the perception exper-iment this attribute was relevant to define Dimension 1.Pleasantness has been identified in several perceptual stud-ies (B. Berglund et al., 1973; Jones et al., 1978; Schiffman,1974a, 1974b; Schiffman et al., 1977; Woskow, 1968). Al-though being a relevant attribute, it was not found to be thedominant one. This finding contrasts with those of a numberof perceptual studies, in which this dimension has beenfound to dominate odor discrimination. This difference maybe due to the use of a relatively heterogeneous stimuli set,because many of the "pleasantness" findings have emergedfrom experiments in which limited odorant lists were used.For example, Schiffman et al. (1977) found pleasantness tobe the dominant dimension of their two-dimensional solu-tion; nevertheless, the fact that almost half of their stimulihad been defined by the Merck Index on the basis of wordssuch as agreeable, unpleasant, or revolting, may have bi-ased their results toward identification of a pleasantnessdimension. Similarly, as Jones et al. (1978) admitted, afterfinding pleasantness as a unidimensional solution for theMDS space of 11 spices, "No doubt the relatively homoge-neous set of odors which we used as stimuli contributed tothe amount of variance assignable to this dimension" (p. 5).The present results suggest that with a relatively heteroge-neous group of odorants, pleasantness may not be the pri-mary means of discrimination, whereas in more homoge-neous samples it may become dominant. This hypothesismerits exploration.

Spiciness was also found to be relevant for both percep-tion (Dimension 3) and imagery (Dimension 1). This at-tribute has emerged in another MDS study (Schiffman1974a, 1974b), and, like fruitiness, it has been convergentlyidentified as a basic term in odor classification studies bystudies using diverse methodologies (Chastrette et al., 1988;Chastrette et al., 1986; Civille & Lawless, 1986; Jeltema& Southwick, 1986; Lawless, 1989). Interestingly, thesetwo attributes correspond to two of the six elements ofHenning's (1916) smell prism.

The similarities uncovered in this study between the per-ception and imagery conditions provide correlational evi-


dence supporting the existence of olfactory storage, andthey suggest that in general the way in which one group ofpeople perceives odors can be predicted by, or predictive of,how a separate group will organize the same odors whenimagining (olfactizing) them. In both perception and imag-ery conditions, for example, a cluster of the fruity (fruitpunch, banana, coconut, strawberry, and bubble gum) stim-uli and other clusters of odorants (gasoline, leather, andmenthol; cedar, clove, and gingerbread) appeared in roughlyequivalent locations. These stimuli contain enough salientfeatures to yield comparable processing in both perceptionand imagery. Our subjects' attribute correlations suggestthat these clusters are mediated by fruitiness, strength, fa-miliarity, and spiciness. However, these items are all slightlymore dispersed in the imagery condition, as indexed byhigher stress values and lower R2 values in imagery than inperception. For instance, banana and bubble gum, two odor-ants grouped very closely in the perception condition, arelocated relatively far apart in the imagery condition, andcherry, which appears in the fruity grouping in the percep-tion condition, seems to break off from the cluster in theimagery condition (see Figures 2 and 4).

All of these findings concerning the similarities betweenthe dimensions underlying both olfactory perception andolfactory imagery, as well as the findings that odor memoryis impaired following odor imagery instructions for distrac-tor stimuli (Crowder & Schab, in press), and that psycho-physical power functions for perceived and rememberedodor intensities are parallel (Algom & Cain, 1990, 1991),ultimately support the idea of a second-order isomorphismbetween the internal representations and their external ob-jects. Unlike a concrete, first-order isomorphism betweeneach individual representation and its corresponding object,a second-order isomorphism is a more abstract kind of cor-respondence, between the functional relations among theinternal representations on the one hand and the structuralrelations among the corresponding external objects on theother hand (Shepard, 1975; Shepard & Chipman, 1970).3

Whatever neurophysiological events are taking place whileone is merely imagining the external process in question,these events have much in common with the internal eventsthat occur when one is actually perceiving the externalprocess itself (Algom, Wolf, & Bergman, 1985; Shepard &Cooper, 1982).

Along with the broad similarities emerging from MDS,there were also disparities between the odor perception andodor imagery conditions. For example, pleasantness-foodlike and frequency-familiarity were highly correlatedin the perception condition but not in the imagery condition;moreover, although females seemed to be more consistent inthe imagery condition, this was not the case in the percep-tion condition. This, along with the higher stress values andlower R2 values in imagery than in perception and the morediffuse imagery groupings, leads us to question how stronga relationship there may be between olfactory perceptionand olfactory imagery.

In contrast to the findings of Gilbert et al. (1989), therewas no evidence for greater female consistency in the per-ception of odors. Their finding may have been a function ofthe limited stimuli set they used; there were only five odors

in their MDS analysis, two of which were fruits and two ofwhich were "woodlike." The present results suggest thatwhen a more heterogeneous set of stimuli is used, the sexdifference in odor judgment consistency may disappear.However, the fact that the imagery task showed positiveresults for greater female consistency precludes a definitiveconclusion regarding the issue. The hypothesis needs furtherexamination.

The aforementioned differences between the perceptionand imagery conditions do not refute the existence of olfac-tory imagery. Had the same subjects participated in bothconditions, a larger degree of perceptual-imagery corre-spondence may have occurred. This, however, was not theaim of this study; its goal was to look for the broadestpossible similarities between perceptual and imagery odorjudgments. Furthermore, in a within-design study, had theperception condition been conducted before the imagerycondition, similarity ratings in the latter condition couldhave been affected by subjects' memories of the former,especially if they had identified the odors (i.e., if they hadlabeled them). On the other hand, if the imagery conditionhad been conducted before the perception condition in awithin-design study, the name labels from the former couldhave affected the similarity ratings of the latter and verballymediated the response. This is something we wanted toavoid; that is precisely why we (a) chose to use MDS, (b)asked subjects to identify the odorants only once they hadfinished rating all the attributes of the perceptual stimuli(Experiment 1, Phase 1), and (c) did not ask subjects toidentify the odorants when making the similarity judgmentsabout the perceptual odorants (Experiment 1, Phase 2).Moreover, the criteria used in the first task, whatever theymay have been, could have also confounded the criteriaused in the second one. Even with the between-subjectsdesign used in this study, consistencies between perceptionand imagery were found.

The consistencies between olfactory perception and ol-factory imagery could be related to the two following hy-potheses: greater salience in semantic mediation or a visualmediation effect. First, consider that these consistencieswere a function of semantic mediation. Whereas MDS al-lows us to minimize the explicit use of words, perhapssubjects, while making similarity judgments in both theperception and imagery experiments, were referring to se-mantic memory and thus exerting a common influencethrough verbal mediation. In other words, maybe the se-mantic store (which would likely contain information thatbanana, coconut, strawberry, etc., are all fruit related) ratherthan perceptual and imagined qualities per se determined thecommon groupings. Such an interpretation would be sup-ported by recent research concerning a facilitatory role ofverbal encoding in odor recognition: Subjects who madeverbal associations to an odor demonstrated superior odor

3 Recently, we discovered that Lyman (1988), in an unpublishedexperiment of his doctoral dissertation, found that when perceivedand imagined odors were modeled separately, the relationshipamong members of the set of the actual odors was preservedamong the members of the set of imagined odors, even though theimagined and actual odors did not exactly resemble each other.


recognition and recall performance when tested 24 hr (Ly-man & McDaniel, 1990) or 7 days (Lyman & McDaniel,1986) later; similarly, verbal suppression during presenta-tion of olfactory stimuli can significantly worsen subse-quent recall and recognition performance (Perkins & Cook,1990).

The semantic mediation hypothesis is more difficult tosustain, however, in the case of other groupings, for in-stance, in the groupings of cedar, gingerbread, and clove, orgasoline, leather, and menthol, because common semanticknowledge about these stimuli would not seem likely. Fur-thermore, verbal labels are not associated with odors aseasily as they are with visual or other verbal stimuli (Davis,1975; Engen, 1987; Lawless & Cain, 1975; Lawless &Engen, 1977; Richardson & Zucco, 1989). It should also benoted that because in the perception condition (Experiment1) subjects were not asked to identify or label the odorantsin the similarity judgments (Phase 2) and asked to identifythem only at the end of the attribute rating task (Phase 1),semantic mediation was not necessarily elicited. Moreover,as mentioned earlier, subjects' performance on identificationwas rather poor; on average, only a third of the subjectscorrectly identified each odorant, even though they hadalready rated them in terms of all 12 attributes. It has beenshown previously that people are not good at naming evenvery familiar odors. An average person with a normal senseof smell can identify only about 50% of common odors,indicating that the link between names and odors is inher-ently weak (Engen, 1987). It has been suggested that be-cause subjects do not have access to a well-articulated rep-resentational system for labeling and classifying odors interms of their perceptual properties, verbal encoding andrehearsal play a small role in the processing of olfactoryexperience (Richardson & Zucco, 1989). Furthermore, therole of semantic mediation in remembering odors has beenquestioned (Engen, 1987). Lyman and McDaniel (1990)maintain that verbal codes do not seem to have any specialelaborative properties for odors, because visual elaborationhas been found to be equally effective in enhancing recog-nition memory. These researchers have suggested that dif-ferential modality encoding, not verbal encoding per se,seems to be an important factor for effective elaboration ofodors; subjects using olfactory imagery perform better on anolfactory recognition test than those who use visual imag-ery, and inversely, subjects who use visual imagery performbetter on a picture recognition task.

Additionally, a visual mediation effect for odor recogni-tion has been suggested. In one study, visual suppression ledto inferior recognition performance (Perkins & Cook,1990), and in another, a visual cue aided later odor recall(Lyman & McDaniel, 1990). Other findings by these re-searchers, nevertheless, clearly support a modality-specificimagery system for the olfactory domain: Subjects whoused visual imagery performed better on a picture recogni-tion task, and inversely, those who used olfactory imageryperformed better on an olfactory recognition task. In anycase, the visual mediation effect could not adequately beused here to account for the groupings that were consistent.The attribute that is the most related to visual mediation,location of the odor-emitting object (inside-outside), was

not related to the dimensional axes. Second, although someof the odorants in the fruity cluster share a common visualproperty (e.g., strawberry and cherry are colored red), thereis no specific visual commonality among other clusters(e.g., leather, gasoline, and menthol) that other stimuli donot also share. Thus, none of these hypotheses, semanticmediation or visual mediation, seem to be able to fullyaccount for the present findings.

In addition to our findings concerning the existence anddimensions of olfactory imagery and their relationship toolfactory perception, the experiments reported here alsoyielded secondary findings which could be of general inter-est to olfaction research. For instance, the difficulty of ap-plying verbal labels to an olfactory experience is clearlyborne out by our results. The relationships found betweenthe perceptual and imagined attribute ratings are instructivein this regard: perception pleasantness failed to reach sig-nificance with imagery pleasantness (r = .38, ns) but had avery strong correspondence to imagery irritation (r = -.77,p < .002) and an almost perfect relationship with imageryfruitiness (r = .92, p < .002). What is described as "pleas-ant" during odor perception may be best understood as "notirritating" or "fruity" during an odor imagery task. The nearperfect correlation between perception pleasantness and per-ception irritation ratings is even more startling (r = -.95and r = -.93, p < .001, for data obtained in this study andby Doty et al., 1984, respectively). Perhaps "pleasantness,"when mentioned in the olfactory context, can be best un-derstood as "nonirritating," and low or high pleasantnessscores are largely determined by a high or low degree oftrigeminal irritation, respectively.

Furthermore, the present results could help orient futureresearch in odor recognition, odor imagery, and odor recall.It is known, for example, that greater intralist dissimilarityleads to superior recall performance (Lawless, 1978) andthat vice versa, performance can be decreased by choosingodorants that are closely associated in odor space (Engen &Ross, 1973). Given that some odorants appear close to eachother along each dimension in both the perception and im-agery conditions (such as banana and strawberry, bubblegum and fruit punch, leather and gasoline, or gingerbreadand clove), the use of these pairs of odors in the same taskshould be discouraged when distinguishability of odors isdesired.

In conclusion, Levine and McBurney (1983) haveclaimed that "Of the many systems of classifying odors thathave been proposed over the centuries, none has ever donemuch for the field except decorate textbooks" (p. 183). Wedo not wish to fall into the perennial trap of overstating thecase for neat, easily identifiable classes of odors. The di-mensions gleaned from MDS studies, like those reportedhere, are affected by the sample of stimuli used as well as bythe expertise of the subjects. Likewise, the labels assigned tothose dimensions are always best guesses subject to rein-terpretation (Young & Hamer, 1987). Nonetheless, as Law-less (1989) has suggested, it might be possible to identifythe psychological dimensions for olfactory processing; it isour belief that MDS will be an invaluable tool for thatenterprise. And ultimately, the experiments reported here,regardless of our ability to nail down the specific names of


the dimensions, shed light on an important issue: The factthat the psychological space derived from having one groupof subjects imagine a set of odors is predictive of, or pre-dicted by, the space that organizes another group's percep-tion of the same set of odors, lends support to the existenceof olfactory imagery.

References

Algom, D., & Cain, W. S. (1990). Chemosensory representation inperception and memory. In S. J. Bolanowski & G. A. Gescheider(Eds.), Ratio scaling of psychological magnitude (pp. 183-198).Hillsdale, NJ: Erlbaum.

Algom, D., & Cain, W. S. (1991). Remembered odors and mentalmixtures: Tapping reservoirs of olfactory knowledge. Journal ofExperimental Psychology: Human Perception and Performance,17, 1104-1119.

Algom, D., Wolf, Y., & Bergman, B. (1985). Integration of stim-ulus dimensions in perception and memory: Composition rulesand psychophysical relations. Journal of Experimental Psychol-ogy: General, 114, 451^71.

Amoore, J. E. (1970). Molecular basis of odor. Springfield, IL:Thomas.

Andreani, O. (1988). Imagery and memory. In M. Denis, J.Engelkamp, & J. T. E. Richardson (Eds.), Cognitive and Neuro-psychological approaches to mental imagery (pp. 57-65).Amsterdam: Martinus Nijhoff.

Barker, L. M., & Weaver, C. A. Ill (1983). Rapid, permanent, lossof memory for absolute intensity of taste and smell. Bulletin ofthe Psychonomic Society, 21, 281-284.

Berglund, B., Berglund, U., Engen, T., & Ekman, G. (1973). Mul-tidimensional analysis of twenty-one odors. Scandinavian Jour-nal of Psychology, 14. 131-137.

Boudreau, J. C., Oravec, J., & White, T. D. (1981). A peripheralneural correlate of the human fungiform papillae sweet, bittersensations. Chemical Senses, 6, 130-141.

Brooks, L. R. (1967). The suppression of visualization by reading.Quarterly Journal of Experimental Psychology, 19, 289-299.

Brooks, L. R. (1968). Spatial and verbal components of the act ofrecall. Canadian Journal of Psychology, 22, 349-368.

Cain, W. S. (1977a). Differential sensitivity for smell: "Noise" atthe nose. Science, 195, 796-798.

Cain, W. S. (1977b). Odor magnitude: Coarse vs. fine grain. Per-ception & Psychophysics, 22, 545-549.

Cain, W. S. (1988). Olfaction. In R. C. Atkinson, R. J. Herrnstein,G. Lindzey, & R. D. Luce (Eds.), Stevens' handbook of experi-mental psychology (2nd ed., pp. 409^59). New York: Wiley.

Cain, W. S., & Gent, J. F. (1991). Olfactory sensitivity: Reliability,generality, and association with aging. Journal of ExperimentalPsychology: Human Perception and Performance, 17, 382-391.

Cain^ W. S., & Stevens, J. C. (1989). Uniformity of olfactory lossin aging. Annals of the New York Academy of Sciences, 561,29-38.

Carroll, J. D., & Wish, M. A. (1974). Multidimensional perceptualmodels and measurement methods. In E. C. Carterette & M. P.Friedman (Eds.), Handbook of perception, Volume 2: Psycho-physical judgment and measurement (pp. 391^47). New York:Academic Press.

Chastrette, M., Elmouaffek, A., & Sauvegrain, P. (1988). A mul-tidimensional statistical study of similarities between 74 notesused in perfumery. Chemical Senses, 13, 295-305.

Chastrette, M., Elmouaffek, A., & Zakarya, D. (1986). Etude statis-tique multidimensionalle des similarites entre 24 notes utiliseesen perfumerie [A multidimensional statistical study of similari-ties between 24 notes used in perfumery]. Correspondence de

Academie Science Paris, 303 (Serie II), 1209-1214.Civille, G. V., & Lawless, H. T. (1986). The importance of lan-

guage in describing perceptions. Journal of Sensory Studies, I,203-215.

Crocker, E. C., & Henderson, L. F. (1927). Analysis and classifi-cation of odors. American Perfumer and Essential Oil Review,22, 325-327.

Crowder, R. G., & Schab, F. R. (in press). Imagery for odors. InR. G. Crowder & F. R. Schab (Eds.), Memory for odors. Hills-dale, NJ: Erlbaum.

Davis, R. G. (1975). Acquisition of verbal associations to olfactorystimuli of varying familiarity and to abstract visual stimuli.Journal of Experimental Psychology: Human Learning andMemory. I, 134-142.

Desor, J. A., & Beauchamp, G. K. (1974). The human capacity totransmit olfactory information. Perception & Psvchophvsics, 16,551-556.

Doty, R. L., Shaman, P., Applebaum, S., Giberson, R., Sikorski, L.,& Rosenberg, L. (1984). Smell identification ability changeswith age. Science, 226, 1441-1443.

Doty, R. L., Shaman, P., & Dann, M. (1984). Development of theUniversity of Pennsylvania Smell Identification Test: A stan-dardized microencapsulated test of olfactory function. Physiol-ogy & Behavior, 32, 489-502.

Ehrlichman, H., & Halpern, J. N. (1988). Affect and memory:Effects of pleasant and unpleasant odors on retrieval of happyand unhappy memories. Journal of Personality and Social Psy-chology, 55, 769-779.

Engen, T. (1982). The perception of odors. New York: AcademicPress.

Engen, T. (1987). Remembering odors and their names. AmericanScientist, 75, 497-503.

Engen, T, Kuisma, J. E., & Eimas, P. D. (1973). Short-term mem-ory of odors. Journal of Experimental Psychology, 99, 222-225.

Engen, T, & Ross, B. M. (1973). Long-term memory of odors withand without verbal descriptions. Journal of Experimental Psv-chology, 100. 221-227.

Farah, M. J. (1989a). Mechanisms of imagery-perception interac-tions. Journal of Experimental Psychology: Human Perceptionand Performance, 15, 203-211.

Farah, M. J. (1989b). The neuropsychology of mental imagery. InJ. W. Brown (Ed.), Neuropsychology of visual perception (pp.183-202). Hillsdale, NJ: Erlbaum.

Finke, R. A. (1989). Principles of mental imagery. Cambridge,MA: MIT Press.

Finke, R. A., & Shepard, R. N. (1986). Visual functions of mentalimagery. In K. R. Boff, L. Kaufman, & J. Thomas (Eds.), Hand-book of perception and human performance (Vol. 2, pp. 1-55).New York: Wiley.

Gilbert, A. N., Greenberg, M. S., & Beauchamp, G. K. (1989). Sex,handedness, and side of nose modulate human odor perception.Neuropsychologia, 279, 505-511.

Goldman, W. P., & Seamon, J. G. (1992). Very long term memoryfor odors: Retention of odor-name associations. American Jour-nal of Psychology, 105, 549-563.

Gordon, I. E., & Hayward, S. (1973). Second order isomorphismof internal representations of familiar faces. Perception & Psy-chophysics, 14, 334-336.

Harper, R., Bate Smith, E. C., & Land, D. G. (1968). Odourdescription and odour classification: A multidisciplinary exam-ination. New York: American Elsevier.

Henning, H. (1916). Der Geruch. Leipzig: Barth.Hvastja, L., & Zanuttini, L. (1989). Odour memory and odour

hedonics in children. Perception, 18, 391-396.Jeltema, M. A., & Southwick, E. W. (1986). Evaluation and ap-


plications of odor profiling. Journal of Sensory Studies, 1, 123-136.

Jones, F. N., Roberts, K., & Holman, E. W. (1978). Similarityjudgments and recognition memory for some common spices.Perception & Psychophysics, 24, 2-6.

Kandel, E. R., & Schwartz, J. H. (1985). Principles of neuralscience. New York: Elsevier.

Kirk-Smith, M. D., Van Toller, S., & Dodd, G. H. (1983). Uncon-scious odour conditioning in human subjects. Biological Psy-chology, 17, 221-231.

Kruskal, J. B., & Wish, M. (1978). Basic concepts of multidimen-sional scaling. Beverly Hills, CA: Sage.

Laird, D. (1935). What can you do with your nose? ScientificMonthly, 41, 126-130.

Lawless, H. T. (1978). Recognition of common odors, pictures,and simple shapes. Perception & Psychophysics, 24, 493^495.

Lawless, H. T. (1986). Multidimensional scaling. In H. L. Meisel-man & R. S. Rivlin (Eds.), Clinical measurement of taste andsmell (pp. 86-103). New York: Macmillan.

Lawless, H. T. (1989). Exploration of fragrance categories andambiguous odors using multi-dimensional scaling and clusteranalysis. Chemical Senses, 14, 349-360.

Lawless, H. T, & Cain, W. S. (1975). Recognition memory forodors. Chemical Senses and Flavor, 1, 331-337.

Lawless, H., & Engen, T. (1977). Associations to odors: Interfer-ence, mnemonics, and verbal labeling. Journal of ExperimentalPsychology: Human Learning and Memory, 3, 52-59.

Levine, J. M., & McBurney, D. H. (1983). The role of olfaction insocial perception and behavior. In C. P. Herman, M. P. Zanna, &E. T. Higgins (Eds.), Physical appearance, stigma, and socialbehavior: The Ontario Symposium (Vol. 3, pp. 179-217). Hills-dale, NJ: Erlbaum.

Lyman, B. J., & McDaniel, M. A. (1986). Effects of encodingstrategy on long-term memory for odors. Quarterly Journal ofExperimental Psychology, 38A, 753-765.

Lyman, B. J., & McDaniel, M. A. (1990). Memory for odors andodor names: Modalities of elaboration and imagery. Journal ofExperimental Psychology: Learning, Memory, and Cognition,16, 656-664.

McDaniel, M. A. (1988). Empirical approaches to a functionalanalysis of imagery and cognition. In M. Denis, J. Engelkamp,& J. T. E. Richardson (Eds.), Cognitive and neuropsychologicalapproaches to mental imagery (pp. 281-291). Amsterdam: Mar-tinus Nijhoff.

Moskowitz, H. R., & Barbe, C. D. (1976). Psychometric analysisof food aromas by profiling and multidimensional scaling. Jour-nal of Food Science, 41, 567-571.

Paivio, A. (1986). Mental representation: A dual coding approach.New York: Oxford University Press.

Perkins, J., & Cook, N. M. (1990). Recognition and recall ofodours: The effects of suppressing visual and verbal encodingprocesses. British Journal of Psychology, 81, 221-226.

Rabin, M. D. (1988). Experience facilitates olfactory quality dis-crimination. Perception & Psychophysics, 44, 532-540.

Rabin, M. D., & Cain, W. S. (1984). Odor recognition: Familiarity,identifiability, and encoding consistency. Journal of Experimen-tal Psychology: Learning, Memory, and Cognition, 10, 316-325.

Richardson, J. T. E., & Zucco, G. M. (1989). Cognition and ol-faction: A review. Psychological Bulletin, 105, 352-360.

Schab, F. R. (1990). Odors and the remembrance of things past.

Journal of Experimental Psychology: Learning, Memory, andCognition, 16, 648-655.

Schab, F. R. (1991). Odor memory: Taking stock. PsychologicalBulletin, 109, 242-251.

Schiffman, S. S. (1974a). Contributions to the physicochemicaldimensions of odor: A psychophysical approach. Annals of theNew York Academy of Sciences, 242, 164^183.

Schiffman, S. S. (1974b). Physicochemical correlates of olfactoryquality. Science, 785, 112-117.

Schiffman, S. S., Reynolds, M. L., & Young, F. W. (1981). Intro-duction to multidimensional scaling: Theory, methods and ap-plications. New York: Academic Press.

Schiffman, S. S., Robinson, D. E., & Erickson, R. P. (1977).Multidimensional scaling of odorants: Examination of psycho-logical and physicochemical dimensions. Chemical Senses andFlavor, 2, 375-390.

Segal, S. J., & Fusella, V. (1970). Influences of imaged picturesand sounds on detection of visual and auditory signals. Journalof Experimental Psychology, 83, 458^464.

Shepard, R. N. (1975). Form, formation, and transformation ofinternal representations. In R. Solso (Ed.), Information process-ing and cognition: The Loyola symposium (pp. 87-122). Hills-dale, NJ: Erlbaum.

Shepard, R. N., & Chipman, S. (1970). Second order isomor-phisms of internal representations: Shapes of states. CognitivePsychology, 1, 1-17.

Shepard, R. N., & Cooper, L. (1982). Mental images and theirtransformations. Cambridge, MA: MIT Press.

Stecher, P. E. et al. (Eds.). (1968). The Merck index (8th ed.).Rahway, NJ: Merck.

Stevens, D. A., & Lawless, H. T. (1981). Age-related changes inflavor perception. Appetite, 2, 127-136.

Stevens, J. C., & Cain, W. S. (1987). Old-age deficits in the senseof smell gauged by thresholds, magnitude matching, and odoridentification. Psychology and Aging, 2, 36-42.

Stone, H., & Bosley, J. J. (1965). Olfactory discrimination andWeber's law. Perceptual and Motor Skills, 20, 657-665.

Tversky, A. (1977). Features of similarity. Psychological Review,84, 327-352.

Walk, H. A., & Johns, E. E. (1984). Interference and facilitation inshort-term memory for odors. Perception & Psychophysics, 36,508-518.

Wenzel, B. M. (1949). Differential sensitivity in olfaction. Journalof Experimental Psychology, 39, 129-143.

Woskow, M. H. (1968). Multidimensional scaling of odors. In N.Tanyolac (Ed.), Theories of odors and odor measurement (pp.147-188). Istanbul: Robert College Research Center.

Wysocki, C. J., & Beauchamp, G. K. (1984). Ability to smellandrosterone is genetically determined. Proceedings of theNational Academy of Sciences (USA), 81, 4899^902.

Young, F. W., & Hamer, R. M. (1987). Multidimensional scaling:History, theory, and applications. Hillsdale, NJ: Erlbaum.

Young, F. W., Takane, Y, & Lewyckyj, R. (1980). ALSCAL: Amultidimensional scaling program with several individual dif-ference options. The American Statistician, 34, 117-118.

Received June 24, 1991Revision received July 2, 1992

Accepted July 6, 1992 •

marisa carrasco and joseph b. ridout - new york …...288 marisa carrasco and joseph b. ridout is,...

Documents