In a dirty virtual room: exposure to an unpleasant odor increases the senses of presence, reality, and realism

Abstract

A recent study found that in a virtual room devoid of obvious visual cues linking visual and olfactory stimuli, exposure to an unpleasant odor (but not to a pleasant one) led to statistically significant increases in the sense of Presence. A second study obtained similar results when concordant visual cues linking the visual scene and pleasant odor were presented. Both studies also reported that neither exposure to an unpleasant or pleasant odor influenced the sense of Realism. However, in the second study, the sense of Reality was statistically significantly higher when participants were exposed to the pleasant odor (but not to the unpleasant one). The goal of the current (and third) study was to clarify these relationships in a virtual environment where the visual scene was linked to an unpleasant odor. To this end, 60 participants were immersed in a filthy virtual kitchen, unaware of the potential exposure to an olfactory stimulus. Depending on their experimental condition, they were exposed to either the ambient odor in the laboratory, a pleasant odor, or an unpleasant odor. The results revealed that exposure to the unpleasant odor increased the senses of Presence and of Reality in a statistically significant manner. Furthermore, the results regarding odor detection rates suggest that visual/olfactory concordance may have facilitated the detection of the unpleasant odor. Overall, the results suggest that, in the case of an unpleasant odor, visual/olfactory concordance can notably enrich the quality of the user experience in virtual reality.

Keywords:

• virtual reality
• presence
• olfaction
• odors
• reality
• realism

The potential benefits of integrating olfactory stimuli in virtual reality (VR) have been outlined by various authors (Baus & Bouchard, 2010; Flavián et al., 2021; Ischer et al., 2014), including in clinical practice (e.g., Herz, 2021). However, despite a consensus surrounding the potentially positive effect of taking VR beyond sight, sound, and touch, empirical studies examining the effects of olfactory stimuli (i.e., odorants/physiochemical molecules) on the quality of user experience are scarce.

1. Olfaction in virtual reality

Few authors have published results relating to the potential effect of olfactory cues on user experience in VR. Of these, Dinh et al. (1999) completed an ambitious project examining the effects of tactile, olfactory, audio, and visual sensory cues on the sense of Presence (the perceptual illusion of “being there” in the virtual environment [VE], Heeter, 1992). To this end, they randomly assigned 322 participants to 16 experimental conditions. During the immersion, which lasted about five minutes, participants were made to tour office spaces. The effect of the olfactory stimulus on Presence was investigated via exposure/non-exposure to a coffee smell near the coffee pot in the reception area. All participants were informed that they may be exposed to odors (i.e., the perception of odorants by the nervous system). The analyses revealed that exposure to the odor did produce higher Presence scores, but that the differences were not statistically significant.

The results of Munyan et al. (2016) revealed that Presence significantly declined for participants first exposed to odors concordant with visual stimuli followed by no odors, whereas it significantly increased for participants exposed to the condition without odors followed by the odors concordant with the visual stimuli. The sense of Presence of participants exposed to the same condition twice (i.e., concordant odors—concordant odors or no odors—no odors) remained relatively stable. These results suggest that participants experienced higher Presence when they received odors than when they did not.

Considering the important role played by the sense of smell in everyday life (Baus & Bouchard, 2010), Dinh et al.’s (1999) somewhat counterintuitive results motivated Baus and his collaborators to re-examine the issue (Baus & Bouchard, 2017; Baus et al., 2019, and the current study, which concludes a set of three integrated studies). In two already published studies, the authors examined the effects of exposure to olfactory stimuli on the senses of Presence, Reality, and Realism. In the first study (Baus & Bouchard, 2017), they devised a protocol intended to improve on Dinh et al.’s (1999) methodology. Participants in the experimental conditions were exposed either to a pleasant or an unpleasant odor in a virtual kitchen where the visual scene was devoid of stimuli that could be the obvious source of the odors. The second study’s (Baus et al., 2019) methodology was identical, except that visual stimuli associated with the pleasant odor were presented in the virtual kitchen (i.e., cinnamon apple pies, the ingredients needed to make them, and cutlery).

The results of the first study (Baus & Bouchard, 2017) suggested that exposure to the unpleasant odor, but not to the pleasant one, had a significant effect on Presence. However, exposure to neither odor had an effect on the senses of Reality or Realism. Interestingly, in virtuo explicit and conscious detection rates of the olfactory stimuli for the pleasant and unpleasant odors were as low as 15% and 60%, respectively. The authors suggested that the low explicit and conscious detection rate of the pleasant stimulus may have contributed to its lack of effect on Presence. The results of the second study (Baus et al., 2019) revealed that despite the visual scene being concordant with the pleasant odor, exposure to the odor did not have a significant effect on the senses of Presence or Realism. However, it did have a significant effect on the sense of Reality. The conscious detection rates were also higher than those in the first study (i.e., 75% and 80% for the pleasant and unpleasant odors, respectively).

In one of two studies reported by Braun (2019), participants reported a significantly higher sense of Presence and found the VE more realistic when they received pleasant odors concordant with the visual stimuli. In a second study (Braun, 2019), participants were immersed with pleasant odors not concordant with the flowers in the VE presented (i.e., odors of chocolate chip cookies and of cherries). Participants reported feeling confused by the pleasant odors not being concordant with the visual stimuli. The results also indicated that the non-concordant odors did not significantly increase participants’ sense of Presence. In other words, it seems that simply exposing participants to odors per se is not sufficient to enhance their sense of Presence.

In a more recent study, Archer et al. (2022) found that participants experienced significantly higher Presence when they received unpleasant odors concordant with visual stimuli than when they did not receive any odors. Interestingly, Realism, or the ecological validity and naturalness of the VE, was not higher when the unpleasant odors were presented than when they were not. Overall, the findings of these studies suggest that presenting unpleasant odors concordant with the VE significantly enhances participants’ Presence.

1.1. Reality, realism, and presence

Quantifying the potential benefits of integrating olfactory cues in VR can be accomplished via a variety of concepts, including the senses of Reality, Realism, and Presence. While the sense of Reality is defined as the degree to which the user perceives the immersion as actually occurring (Baños et al., 2000; e.g., having the impression of actually baking when baking a virtual cake), the sense of Realism is defined as the similarity between the virtual and the corresponding physical stimuli (Baños et al., 2000; e.g., a virtual apartment resembles one from the physical environment). Although the exact definition of the sense of Presence varies depending on the authors, Heeter’s (1992) definition stating that the term refers to the impression of being in the VE was adopted here and in Baus and Bouchard (2017, 2019). Some researchers (e.g., Slater, 2009) have argued that a user’s experience in VR includes feeling as if one is actually in the VE, but also feeling as if events in the VE are actually occurring and the virtual stimuli are similar to their physical versions. To reflect these Baus and Bouchard (2017, 2019) and the current study assessed participants’ senses of Presence, Reality, and Realism.

A number of individual (user-related) and environmental (system-related) variables have been found to affect Presence. Because it is difficult to influence individual factors, these are more of theoretical interest. Previous studies have shown that individual factors such as immersive tendencies (Witmer & Singer, 1998) and immersion-related negative side effects (Kennedy et al., 2000) can impact Presence. However, few studies examining the effect of olfactory cues on Presence seem to have used measures to control for individual differences in immersive tendencies and immersion-related negative side effects, which could be a methodological limitation. Environmental factors (e.g., level of visual realism, Waterworth & Waterworth, 2003), however, can be readily manipulated, and accordingly are of particular interest to VR developers. It has also been suggested that an increase in the number of stimulated senses can lead to higher levels of Presence. More specifically, it has been reported that exposure to sound (Dinh et al., 1999), spatialized sound (Hendrix & Barfield, 1996b), and the use of tactile augmentation (Dinh et al., 1999; Hendrix & Barfield, 1996a; Hoffman et al., 1996; Meehan et al., 2002) can increase the sense of Presence. On the basis that engaging a greater number of a user’s senses increases their sense of Presence, it would be expected that exposure to olfactory stimuli will also increase the sense of Presence in VR, especially if implicit and explicit information processing leads to concordant multisensory integration (Berthiaume et al., 2021).

2. Objectives and hypothesis

It was the primary objective of the current study to examine further the relationship between conscious detection rates, the concordance of olfactory and visual stimuli, and the sense of Presence. This time in a VE visually consistent with the unpleasant odor. Exploratory analyses were also conducted to examine whether exposure to a visual/olfactory-concordant VE would be associated with higher levels of Reality and Realism. Lastly, it was of interest how, compared to the results obtained by Baus and Bouchard (2017) and Baus et al. (2019), the detection rates of the unpleasant odor would be affected by the high level of concordance with the visual scene.

The first research hypothesis for this experiment stated that, in a VE that includes visual elements normally associated with an unpleasant odor, exposure to an unpleasant odor would be associated with a statistically significantly higher level of Presence compared to the control condition. Based on the results of Baus et al. (2019), which indicated that the sense of Reality statistically significantly increased when participants were exposed to a pleasant odor concordant with the visual scene, a second hypothesis stipulated that the sense of Reality would be statistically significantly higher when the odor is concordant with the visual scene (i.e., unpleasant odor in the dirty virtual kitchen) compared to the control condition. In addition, it was expected that concordance between the unpleasant odor and visual scene would statistically significantly increase the sense of Realism (i.e., participants exposed to the unpleasant odor would find the VE more realistic because it matches the dirty kitchen) compared to the control condition. Finally, based on the results of Baus and Bouchard (2017) and Baus et al. (2019), a fourth hypothesis stated that a high level of concordance with the visual scene would lead to higher detection rates of the unpleasant odor compared to the pleasant condition.

3. Method

3.1. Participants

This study complies with the Declaration of Helsinki for Medical Research involving Human Subjects, and the Canadian Tri-Council Policy Statement for Ethical Conduct for Research Involving Humans and was approved by the institutional ethical review boards. Informed consent was obtained once before the start of the experiment, then a second time when the real purpose of the study was revealed to the participants.

Seventy-two participants were recruited in 2015 on local university and college campuses via adverts and class visits. Participants had to be between the ages of 18 and 60 and functionally fluent in French. Potential participants were excluded if they met any of the following criteria: (1) a history of severe motion sickness, (2) taking medication for anxiety or depression, (3) having extensive experience using VR, or (4) meeting post-experiment one of the following four criteria established a priori: poor stereoscopic vision, cybersickness, experiences that would disturb the immersion (technical problem or colliding with the laboratory’s physical walls) and considering the pleasant odor as unpleasant (or vice-versa). One potential female participant was excluded during recruitment to maintain a balance in gender in each group (see below).

Selected participants were randomly assigned to one of three groups (balanced with regards to gender): (1) AMB group (exposure to the ambient air of the laboratory), (2) PLE group (exposure to a pleasant odor: apple pie/cinnamon), or (3) UNP group (exposure to an unpleasant odor: urine). The pleasant and unpleasant odors were selected based on the results of a pilot study, which showed a consensus regarding the pleasantness of the smell of apple pie/cinnamon and the unpleasantness of the smell of urine.

Eleven participants were excluded post-experiment and from the sample/statistical analyses for the following reasons: (1) poor stereoscopic vision according to the Randot test (n = 1), (2) excessive cybersickness (n = 4), (3) came into contact with the physical world during the immersion (e.g., colliding with the laboratory’s walls; n = 1), (4) assigned to the pleasant group but reported the odor as unpleasant during the immersion (n = 3), and (5) did not detect an odor during the immersion and, when re-exposed to the odor post-immersion, reported the pleasant odor as unpleasant (n = 2). To obtain three gender-balanced groups of 20 participants, recruitment was continued, and participants excluded post-experiment were replaced by the next eligible participant of the same gender.

The final sample consisted of 60 participants (21 males and 39 females), with each group composed of 20 participants (7 males, 13 females). The group’s mean age, dominant cultural influence (by continent), and highest level of education completed are shown in Table 1 for the final sample of participants.

Table 1. Sociodemographic description of participants in the three odorant conditions (N = 60)

Measure	Ambient (AMB) n = 20	Pleasant (PLE) n = 20	Unpleasant (UNP) n = 20
Age	25.55 (6.50)	23.10 (5.54)	23.70 (5.35)
Dominant culture
African	5%	0%	15%
North American	95%	100%	80%
South American	0%	0%	5%
Highest level of education completed
Master’s/Ph.D.	0%	10%	5%
Bachelor	10%	0%	20%
Pre-university college	75%	60%	55%
High school	15%	30%	20%

Note. Standard deviations of age are shown in brackets.

3.2. Material

The material included a VE displayed in an NVIS nVisor SX60 HMD head-mounted display combined with an InterSense InertiaCube3 motion tracker. Odorants were distributed by an SDS100 Scent Palette (Biopac Systems, Inc.) using two scent cartridges (Biopac Systems, Inc.). For a detailed description, see, Baus and Bouchard (2017). The VE (see, Figure 1) was identical to that used by Baus and Bouchard (2017) except that the virtual kitchen was very untidy and dirty (see, Figure 2).

Figure 1. Virtual rooms searched prior to the virtual kitchen.

Note. a) Bathroom, b) Living room, c) Office, d) Bedroom.

Figure 2. Virtual kitchen and immersion room.

Note. a) Dirty kitchen, b) participant in immersion. For the purpose of the photograph, the lights in the immersion room were left on. During the actual protocol, the immersion is carried out in the dark. The scent delivery system is the apparatus seen above the participant, in the center of the immersion area in b.

3.3. Instruments and measures

3.3.1. General measures

Sociodemographic variables were collected via a questionnaire. The Randot^R Stereotest (Stereo Optical Company, Inc.; www.stereooptical.com) was used to ensure that participants could detect, at worst, a depth difference of 70 seconds of arc. The participants’ interpupillary distance was measured using a Topcon PD-5 pupilometer (Topcon Medical Systems, www.topconmedical.com), and the Quick Smell Identification^TM Test (Q-SIT; Sensonics, Inc., Haddon Heights, NJ) was used to verify that the participants could, at worst, identify two of the three test odors. The Quick Smell Identification^TM Test was administered to participants at the very end of the protocol to test their olfactory capabilities. They had to correctly identify at least two of the three odors or otherwise they were excluded from the study. A fuller description of these instruments can be found in Baus and Bouchard (2017).

3.3.2. Measures of senses of presence, reality, and realism

The sense of Presence was measured, per-immersion, via a brief verbal measure, and post-immersion. For the brief verbal measure of Presence during the immersion (Bouchard et al., 2004), a pre-recorded audio instruction requested participants to rate the level of Presence on a 0 to 100% scale in the bedroom (control condition without odorant for all participants) and the kitchen (experimental condition). See Baus and Bouchard (2017) for psychometric information. The sense of Presence was measured with a more comprehensive instrument post-immersion using the Spatial Presence subscale of the Independent Television Commission Sense of Presence Inventory (ITC-SOPI; Lessiter et al., 2001).

The sense of Reality was measured per-immersion using a brief verbal measure administered during the immersion with an audio recording asking participants to verbally rate on a 0 to 100% scale to what extent does their experience in the room seem real (see Baus & Bouchard, 2017). During the immersion, the sense of Reality was measured in the bedroom (control condition) and in the kitchen (experimental condition). The sense of Realism was measured post-immersion using the Ecological Validity/Naturalness subscale of the ITC-SOPI (Lessiter et al., 2001).

3.3.3. Measures of control variables

Several measures were taken to potentially control for individual differences that can affect the level of Presence. The Immersive Tendencies Questionnaire (ITQ; Witmer & Singer, 1998) measured the ease by which the participants could be immersed in a VE. The Modified Tellegen Absorption Scale (MODTAS; Jamieson, 2005) measured the ease by which the participants could feel absorbed. The Simulator Sickness Questionnaire (SSQ; Kennedy et al., 1993) and the Negative Effects subscale of the ITC-SOPI (Lessiter et al., 2001) measured the participants’ level of immersion-related discomfort. These instruments are the most widely used to assess immersive tendencies and cybersickness. A fuller description of these instruments can be found in Baus and Bouchard (2017).

3.3.4. Measures of olfactory perception

The in virtuo detection of olfactory stimulus questionnaire was administered post-immersion to all participants. Its function was to verify whether an odor had been detected in virtuo, and if so, to collect the perceived characteristics of that odor. A test was also conducted in vivo with an olfactory stimulus questionnaire administered only to the participants of the PLE and UNP groups, once all other questionnaires had been completed. It consisted of a 30-second in vivo re-exposure to the olfactory stimulus and aimed to verify the characteristics of the odor, as perceived in the non-synthetic world. A fuller description of these instruments can be found in Baus and Bouchard (2017). In addition to these measures, during the in vivo exposure to the olfactory stimulus, the time elapsed between the start of the exposure and the participant’s detection of the olfactory stimulus was also measured.

3.4. Overview of the protocol

In order to allow for inter-experiment comparisons among the set of three studies by Baus and Bouchard (2017), 2019, current study), this study followed to the letter the protocol used by Baus and Bouchard (2017), except for the following points: (a) while Baus and Bouchard (2017) virtual kitchen was devoid of objects that could be the obvious source of the experimental odors, the virtual kitchen used in this latest experiment was filled with a group of items normally associated with an unpleasant smell; and (b) an in vivo test measuring the detection time for each of experimental odors was added (see, Figure 3).

Figure 3. The 30-second in vivo exposure to the olfactory stimulus previously used in the virtual kitchen.

Note. The time needed for participants to detect the odor is measured, and they rate the characteristics of the odor.

Pre-immersion, participants completed the Immersive Tendencies Questionnaire and the Simulator Sickness Questionnaire and read instructions on the task to complete in VR. The actual objective of the study was not immediately revealed to participants (i.e., they were unaware that they would potentially be exposed to odors) as this could bias the results. Instead, participants were told that each room in the VE had a different level of visual detail and they had to determine whether this affected the senses of Presence, Reality, and Realism. Their specific task involved finding a virtual knife used in a murder in an apartment (no actual knife in the apartment). They had to search the apartment’s bathroom, living room, home office, bedroom, and kitchen, which was filled with items associated with an unpleasant smell (e.g., piles of dirty dishes, garbage). To increase participants’ stress levels, the VE included suspense-type sounds and participants were told that the murderer’s location was unknown.

Participants started the immersion in a hall outside the apartment to familiarize themselves with navigation in VR and to accommodate to the VE. Once participants indicated being ready to begin the experiment, a series of pre-recorded instructions reminded them of the scenario and gave them directions to the first room to search, the bathroom. They searched for the knife for one minute and then a pre-recording asked them to verbally rate their senses of Presence and Reality using the brief verbal measures (this information was not used but collected to prevent raising attention to this procedure). Once completed, another pre-recording directed them to the next room. The procedure was repeated for all the rooms. Participants searched, in this sequence, the living room, home office, bedroom, and kitchen; they were only exposed to ambient air until they reached the kitchen. Once in the kitchen, they were exposed to one of the following odors (according to their randomly assigned group): (1) ambient air of the laboratory (AMB group/control group), (2) pleasant apple pie/cinnamon-like odor (PLE group; not concordant with the visual scene of the dirty kitchen), or (3) unpleasant urine-like odor (UNP group; concordant with the visual scene). The immersion lasted approximately 15 minutes.

Post-immersion, participants completed the Spatial Presence and Ecological Validity/Naturalness subscales of the ITC-SOPI, as well as the Simulator Sickness Questionnaire. The experimenter administered the in virtuo detection questionnaire and re-exposed participants in vivo to the odors from the VE to confirm, via the detection of olfactory stimulus questionnaire, that those from the PLE group found the odor pleasant and those from the UNP group found the odor unpleasant. The experimenter then administered the Quick Smell Identification^TM Test and debriefed the participants (including confirming their informed consent). For additional details about the protocol, see Baus and Bouchard (2017).

4. Results

IBM SPSS Statistics 28 was employed for all statistical analyses. Assumptions were verified prior to each analysis and non-parametric tests were carried out, if applicable. A Bonferroni was applied to the four measures of Presence, Reality and Realism to control for Type 1 error among conceptually related variables (critical alpha set at .05/4 = .0125). Analyses in Table 3 were double checked for accuracy, leading to minor changes during the review process.

4.1. Analysis of control variables and manipulation checks

4.1.1. Pre-immersion differences between groups

The Kruskal-Wallis test and multiple one-way ANOVAs revealed that the groups did not differ in terms of age, immersive tendencies, or absorption tendencies (for complete results, see Supplementary Material).

4.1.2. Immersion-induced negative side effects

A combination of the Kruskal-Wallis and Wilcoxon tests confirmed that the groups did not differ in terms of immersion-induced negative effects (for complete results, see Supplementary Material).

4.1.3. Characteristics of olfactory stimuli detected in virtuo

The detection rates of the PLE and UNP odors were, respectively, 35% and 95% (X² = 15.83, p < .001). The characteristics of the PLE and UNP odors and a comparative analysis of these are presented in the Supplementary Material.

4.1.4. Characteristics of olfactory stimuli detected in vivo

A t-test confirmed that there was no statistically significant difference between the number of seconds required to detect the pleasant (M = 5.03, SD = 1.72) and unpleasant odors (M = 4.40, SD = 1.46; t = 1.23, p = .23). The detection rates and characteristics of the PLE and UNP odors, as well as a comparison of these characteristics, can be found in the Supplementary Material.

4.2. The sense of presence

4.2.1. Brief measure of presence

In order to verify the effect of the exposure to odors on Presence, planned contrasts interaction for repeated-measures designs were used; in this case involving three conditions (AMB, PLE, UNP) by two times (bedroom-control, kitchen-experimental).¹ Mean values and standard deviations are shown in Table 2, results of the contrasts are shown in Table 3, and an illustration of the interaction is shown in Figure 4. The a priori contrast revealed that in terms of Condition, there was no statistically significant difference in Presence either between AMB and PLE, or between AMB and UNP. There was a statistically significant Time effect on Presence. Finally, while there was a statistically significant Interaction effect between AMB and UNP, there was none between AMB and PLE.

Table 2. Brief measures of presence and reality collected in virtuo

Measure		Bedroom (control)		Kitchen (experimental)
		M (SD)	95% CI	M (SD)	95% CI
Brief Presence
	Ambient	79.50 (10.99)	[74.36, 84.64]	80.00 (15.81)	[72.60, 87.40]
	Pleasant	75.10 (15.02)	[68.07, 82.13]	77.75 (17.53)	[69.54, 85.96]
	Unpleasant	75.90 (14.67)	[69.04, 82.76]	85.45 (12.07)	[79.80, 91.10]
Brief Reality
	Ambient	76.75 (12.80)	[70.76, 82.74]	76.00 (16.67)	[68.20, 83.80]
	Pleasant	64.85 (23.90)	[53.67, 76.03]	66.40 (23.92)	[55.21, 77.59]
	Unpleasant	75.40 (14.78)	[68.48, 82.32]	83.95 (13.47)	[77.64, 90.26]

Note. CI = confidence interval.

Table 3. Planned contrasts for measures of presence, reality, and realism collected in virtuo

Measure
		t	p	η^2
Brief Presence
	AMB vs. PLE	.79	.43	.01
	AMB vs. UNP	−.22	.83	.00
	Time	−2.88	.01	.13
	Interaction AMBxPLE	.60	.55	.01
	Interaction AMBxUNP	2.51	.01	.10
Brief Reality
	AMB vs. PLE	1.97	.05	.06
	AMB vs. UNP	−.60	.55	.01
	Time	−2.15	.04	.07
	Interaction AMBxPLE	.65	.52	.01
	Interaction AMBxUNP	2.62	.01	.11
		t	p	Cohen’s d
ITC Spatial Presence
AMB vs. PLE		.71	.24	.23
AMB vs. UNP		−.72	.24	−.23
	ITC Ecological Validity
	AMB vs. PLE	.38	.35	.12
	AMB vs. UNP	−2.34	.01	−.74

Note. AMB = ambient (control) condition; PLE = pleasant condition; UNP = unpleasant condition.

Figure 4. Illustration of the brief measure of presence x time interaction effect, as shown by mean change from the control to the experimental condition.

Note. Standard errors are represented in the figure by the error bars attached to each column.

4.2.2. Spatial presence subscale of the ITC-SOPI

To verify the effect of the exposure to odors on Spatial Presence, planned t-tests comparisons were used. The contrasts revealed that neither the level of Spatial Presence of the UNP group (M = 3.46, SD = .60) nor that of the PLE group (M = 3.17, SD = .69) differed in a statistically significant manner from the level of Spatial Presence of the AMB group (M = 3.31, SD = .66). Results of the contrasts are shown in Table 3, and mean values and standard errors are shown in Figure 5.

Figure 5. Sense of presence (as measured by the spatial presence subscale of the ITC-SOPI).

Note. Standard errors are represented in the figure by the error bars attached to each column.

4.2.3. relationship between presence and the characteristics of odors

In order to verify how the individual characteristics of an odor perceived in virtuo may have affected the sense of Presence, correlations were calculated. To measure change in Presence, Presence residual gain scores were calculated (Manning & Du Bois, 1962; Tracy & Rankin, 1967) using the brief measures of Presence in the bedroom and in the kitchen. The results of the Spearman correlations are shown in Table 4. Two correlations were statistically significant.

Table 4. Relationship between presence and characteristics of odors

Measure	Concordance	Pleasantness	Intensity	Familiarity
Change in Presence	.22	−.23	−.04	−.41*
ITC Spatial Presence	.47*	−.09	.37	.15

Note. Correlations are based on the PLE and UNP participants who detected an odor in virtuo (PLE = 7; UNP = 19).

* p < .05.

4.3. The sense of reality

4.3.1. Brief measure of reality

The a priori contrast² revealed that in terms of Condition, there was no statistically significant difference in Reality, either between AMB and PLE or between AMB and UNP. The Time effect on Reality was statistically significant. Finally, there was a statistically significant Interaction effect between AMB and UNP but not between AMB and PLE. Mean values and standard deviations are shown in Table 2, results of the contrasts are shown in Table 3, and an illustration of the interaction is presented in Figure 6.

Figure 6. Illustration of the brief measure of reality X time interaction effect, as shown by mean change from the control to the experimental condition.

Note. Standard errors are represented in the figure by the error bars attached to each column.

4.4. The sense of realism

4.4.1. Ecological validity subscale of the ITC-SOPI

In order to verify the effect of the exposure to odors on Ecological Validity, planned t-tests comparisons were used. The first contrast revealed that the Realism experienced was statistically significantly higher for participants in the UNP condition (M = 3.71, SD = .69) than the AMB condition (M = 3.17, SD = .77). The second contrast showed no statistically significant difference between the Realism of the exposure to the PLE (M = 3.07, SD = .89) and AMB groups (M = 3.17, SD = .77). Results of the contrasts are shown in Table 3, and mean values and standard errors are shown in Figure 7.

Figure 7. Sense of realism (as measured by the ecological validity subscale of the ITC-SOPI).

Note. Standard errors are represented in the figure by the error bars attached to each column.

5. Discussion

This study concludes a series of three related experiments conducted with different participants. The methodology was similar in each study, except for the visual stimuli in the virtual kitchen, which this time were visually congruent with an unpleasant smell. To provide a comprehensive picture, results will first be briefly summarized, then interpreted in the context of the other two related studies (Baus & Bouchard, 2017; Baus et al., 2019) before being interpreted in the broader context of the literature on VR and olfaction. The first hypothesis of the current experiment stated that, in a VE that includes visual elements that could be associated with an unpleasant smell, the exposure to an unpleasant odor would be associated with a statistically significantly higher sense of Presence. This was confirmed only for the brief measure of presence, not for the Spatial Presence subscale of the ITC-SOPI. The second hypothesis, which stipulated that the sense of Reality would be statistically significantly higher when the unpleasant odor is concordant with the visual scene, was confirmed. The third hypothesis stating that concordance between the unpleasant odor and visual scene would statistically significantly increase the sense of Realism was also confirmed. Finally, the hypothesis stipulating that a high level of concordance with the visual scene would lead to higher detection rates of the unpleasant odor was confirmed, although the results warrant further investigation. All statistically significant differences confirming the hypotheses met the significance level corrected for the number of comparisons.

5.1. Sense of presence

No statistically significant difference was detected between the levels of Presence of the pleasant and ambient conditions with both measures of presence. This result is similar to that reported by Baus and Bouchard (2017) and Baus et al. (2019). However, it must be pointed out that in the later experiment, there was a particularly low level of concordance between the pleasant smell of apple pie and the dirty visual scene.

In the unpleasant odor condition, the brief measure of presence administered during the immersion revealed a statistically significant Condition X Time interaction effect; the size of the effect was small. This result is also in line with the results found by Baus and Bouchard (2017). The slightly higher effect size detected in the current experiment may be attributable to the notably higher concordance of the unpleasant odor with the visual scene. Berthiaume et al. (2021) also suggested that higher concordance between stimuli would increase Presence in VR. As mentioned previously, the concordance effect between the visual stimuli and the unpleasant odor was not statistically significant on the ITC-SOPI Spatial scale. The inconsistence between the measures was unexpected but interesting. The multiple items of the ITC-SOPI questionnaire give it an edge in terms of validity compared to the brief measure. However, the per-immersion brief repeated measures of Presence have the advantage of being taken immediately during the immersion to capture changes in Presence as it occurs over time. Considering the notable number of individual predispositions that have been found to potentially affect Presence (Sacau et al., 2008), a within-between repeated measures design probably provides a more accurate picture of the impact of the experimental manipulation on Presence than only one post-immersion assessment with a questionnaire (Jones et al., 2004). A repeated-measures design allows to control for potential differences in participants after randomization and better isolate the impact of the experimental manipulation. This designed was used by Braun (2019), and Munyan et al. (2016) confirmed the increase and decrease in Presence when odors were added or withdrawn from the virtual environment. Such a design was unfortunately ruled out for the current study to avoid revealing the purpose of the study to participants; a brief measure administered during the immersion was used for within-between subject analyses and the questionnaire was used only for the between-subject analyses. Archer et al. (2022) used a between-subject design and did find a significant difference in Presence, but they used significantly more varied and frequent unpleasant odorants than what was done in our study. The inconsistency in our results on Presence illustrate the pros and cons of both options. Nevertheless, with Archer et al.’s (2022) and Munyan et al.’s (2016) results, evidence is converging to support the notion that unpleasant virtual stimuli that are concordant with the content of the VE contribute to an increase in Presence.

5.2. Sense of reality

Although a secondary measure in this study, the results pertaining to the sense of Reality are quite interesting. In the first experiment in this series of three article, Baus and Bouchard (2017) used a VE devoid of stimuli associating the visual scene to any of the odors and found that exposure to neither odor had an effect on the sense of Reality. This time, using a very dirty virtual kitchen that “looks to smell bad,” exposure to the unpleasant odor led to a statistically significantly higher Reality score than the ambient air. The effect size was small. No such difference was detected between the pleasant and ambient groups. These results suggest that the concordance between the dirty kitchen and the unpleasant odor may have contributed to a higher sense of Reality.

5.3. Sense of realism

As measured via the Ecological Validity subscale of the ITC-SOPI, exposure to the pleasant odor had no statistically significant effect on the sense of Realism, which is consistent with the results of Baus and Bouchard (2017), Baus et al. (2019), and Archer et al. (2022). Exposure to the unpleasant odor, however, did increase the sense of Realism in a statistically significant manner in the current study; the effect size was large. This result differs from the findings of Baus and Bouchard (2017) and Baus et al. (2019) and suggests that the match between the dirty kitchen and the unpleasant odor may have contributed to a higher sense of Realism. It would be interesting to further investigate the relationship between the sense of Realism and visual/olfactory-concordant VEs as Baus et al. (2019) found that exposure to the pleasant odor, even though it matched the visual scene, did not statistically significantly increase the sense of Realism. The difference in results from Archer et al. (2022) is likely to be related to the fact the current study involved an actual immersion in VR but not Archer et al. (2022), although other methodological differences could also be invoked. Participants in the current study were excluded if they failed the Quick Smell Identification^TM Test, but the device used to distribute the odorants was more refined in Archer et al.’s (2022) study.

5.4. Detection rates

There were important differences between the in virtuo conscious detection rates of the pleasant and unpleasant odors (35% for PLE and 95% for UNP in the current study). Interestingly, during the explicit in vivo test conducted after the immersion in VR and the completion of the questionnaires, there was no statistically significant difference between the times to detect these odors when the participants were informed of the impending exposure to an odor. A comparison of the pleasant and unpleasant detection rates reported by Baus and Bouchard (2017) revealed that, in the current study where the visual scene was loaded with potentially odorous objects, the in virtuo conscious detection rates were notably higher for both the pleasant (35% versus 15%) and unpleasant (95% versus 60%) groups. In their 2019 study, where the visual scene was concordant with the pleasant odor, Baus et al. reported conscious detection rates of 75% and 80% for the pleasant and unpleasant groups, respectively. Although the detection rates for both odors were lower than in the current study, the concordant visual scene and pleasant odor led to a higher detection rate for the pleasant group (75% versus 35% and 15%). The results of these studies suggest that the visual scene presented in the VE may have contributed to a higher awareness of the olfactory stimuli dispensed during the immersion. This impression is in line with the findings of previous studies reporting that higher degrees of semantic concordance between an odor and a picture facilitate olfactory detection (Gottfried & Dolan, 2003).

At the same time, the control group’s non-negligible 15% rate of false positives perception of an odor is reminiscent of the wine tasting experiment (Morrot et al., 2001) that demonstrated how visual information can mislead the olfactory sense. It is important to note that participants were not informed they would potentially be exposed to odors during the immersion to prevent potential biases. In their study, Munyan et al. (2016) also found that many participants reported detecting odors in the No odors condition. Many participants also stated having experienced a change in ambient room temperature, even though the authors did not manipulate it. It would be worth further exploring the impact of visual information on detection rates in further studies. It would also be interesting to use different pleasant and unpleasant odors in future studies to further examine the impact of odor concordance and intensity on detection rates.

5.5. Limitations

Two limitations that have been already addressed in this Discussion is how to measure Presence without creating a break in Presence when questioning the participant per-immersion and the importance of repeated-measure designs to control for potential interpersonal differences among participants after randomization. Objective measures of Presence include mostly physiological and behavioral responses to virtual stimuli (Souza et al., 2021), but they are proxies based on emotional reactions to stimuli that address more the senses of Reality and Realism than Presence. Research is ongoing to improve these measures (e.g., Dey et al., 2020; Ochs et al., 2022). Limitations related to the persistence of odors also need to be acknowledged. The device used to release odorants distributes odors from the ceiling to the entire room (see, Figure 2). Because odors linger in the air and the exact concentration of odorant reaching participants’ olfactory system could not be controlled, it was not possible to reliably quantify the amount of olfactive immersion. Alternative solutions to distribute odorants must allow users immersed in VR to move freely in the environments and without emitting cues revealing when odorants are being released.

The study focused on users’ experience instead of formally quantifying the amount of immersion provided by the olfactory stimuli. Therefore, the perceived unpleasant or pleasant valence of the stimuli was required for the experimental manipulation. Some participants were excluded because they reported the pleasant odor as unpleasant. This, and the inclusion of participants who did not formally detect the odors during the immersion, illustrate the limits of experimental manipulations of subjective experiences versus objective stimuli. As studies are including more and more complex and varied odorants (e.g., Archer et al., 2022), it will become increasingly challenging to conduct studies formally quantifying olfactory stimuli.

6. Conclusions and recommendations

The results of this study suggest that, in a VE that includes visual elements normally associated with an unpleasant smell, the effect of exposure to an unpleasant odor can increase the senses of Presence, Reality, and Realism. Furthermore, when previous results from two related studies are taken into account, the results of this last study of the trilogy suggest that, in a VE, a higher level of visual/olfactory concordance can increase the conscious detection rate of an unpleasant odor. Further confirmation remains necessary.

While the results of this experiment give support to the possibility of increasing the quality of user experience in VEs by matching visual and olfactory cues, further research is needed to address this issue in the context of a pleasant odor, as some results remain unclear. For example, exposure to the unpleasant odor and concordant visual scene had a statistically significant effect on the sense of Presence in the current study, but this was not the case for the pleasant odor in Baus et al. (2019). Is the pleasant odor’s lack of effect on Presence related to a lower conscious detection rate, a low visual/olfactory concordance, or does a pleasant odor simply have only minimal effect on Presence in VR? Other independent studies (e.g., Archer et al., 2022; Dinh et al., 1999; Munyan et al., 2016) did not focus on pleasant odors. Braun (2019) studied pleasant odors and visual concordance but did not compare the impact of unpleasant odors. It would be interesting to conduct studies with immersions in VR using a three (no added odor, pleasant odor, and unpleasant odor) by two (neutral visual content and visual content congruent with a pleasant odor) design and quantify the effect of the visual scene, the odor (pleasant versus unpleasant), and their interaction on the senses of Presence, Reality, and Realism. Further studies should also explore if including odors in VEs could improve user experience during psychotherapy to elicit memories and emotions (Herz, 2021).

Disclosure statement

Stéphane Bouchard is the President of, and owns equity in, Cliniques et Développement In Virtuo, a spin-off company from the university that distributes virtual environments designed for the treatment of mental disorders. The terms of these arrangements have been reviewed and approved by the Université du Québec en Outaouais in accordance with its conflict of interest policies. The remaining authors report no financial relationships with commercial interests.

Data Availability Statement

The data that supports the findings of this study is available upon request addressed directly to the relevant Research Ethics Boards ([email protected] and [email protected]). The dataset is not publicly available due to privacy and ethical restrictions.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/23311908.2022.2115690

Additional information

Funding

This research was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) scholarship awarded to the first author, by grants from the NSERC, the Canada Foundation for Innovation (CFI) and the Canada Research Chairs awarded to the second author, and a post-doc grant awarded by the Fonds de Recherche du Québec Société et Culture (FRQSC) to the third author.

Notes

1. As Presence was not normally distributed for the PLE and UNP groups, these results were confirmed through a combination of Mann-Whitney and Wilcoxon tests.

2. However, the sense of Reality measures were not normally distributed in either the bedroom or the kitchen, and variances in those two rooms were uneven. Thus, these results were confirmed through a combination of Mann-Whitney and Wilcoxon tests.