One of the leading claims of aromatherapy is that it can affect the olfactory system directly, because the molecules from the aroma material can penetrate the blood-brain barrier through the nose (Price 1993). From basic human physiology, we know there is some validity to this claim, since certain molecules can attach to receptor cells on the cilia inside the nose, which are connected to the olfactory bulb of the brain (Goldstein 2002). The olfactory bulb has neural pathways that further link it to the limbic system, which is responsible for emotional processes (Goldstein 2002). Therefore, it is theoretically possible that certain molecules could excite or inhibit certain pathways in the limbic system, leading to feelings of anxiety or arousal (Goldstein 2002). However, it is unclear whether or not the essential oils in aromatherapy actually contained molecules compatible with the receptor sites on the cilia of the olfactory bulb. This study attempts to specifically examine whether biological effects from the chemicals in aromatherapy lead to a reduction of anxiety.

There are clinical studies that use anxiety-related questionnaires to assess the effectiveness of aromatherapy. Ballard et al. (2002) used double-blind placebo-controlled procedures to test whether or not aromatherapy reduces agitation on subjects with severe dementia. Using the Cohen-Mansfield Agitation Inventory (CMAI) questionnaire, they found that 60% of the experimental group experienced a 30% reduction in CMAI score, which correlates to a reduction in agitation. The subjects did not know about the aromatherapy, which minimizes subjects’ response bias. The essential oil, Melissa, was mixed into a hand and face lotion, and nurses applied it to the patients’ skin at certain times of the day.

Some studies have tried to demonstrate a reduction in anxiety from aromatherapy by taking indirect physiological measurements (Saeki 2000). The most common of these measurements are blood pressure and heart rate, since these are affected by the autonomic nervous system. There have been mixed results in these studies. Saeki (2000) believes that the "resolution" of blood pressure and heart rate measurements is too low to show significant results even though it is possible for aromatherapy to be affecting the autonomic nervous system.

Other indirect measurements have been made using electroencephalograms. The theory is that alpha-wave patterns in the brain are associated with inactivity and relaxation. Lorig and Schwartz (1987), and many subsequent researchers, have demonstrated an increase in alpha-waves after the inhalation of an anxiety-reducing aroma, such as lavender derived from Lavendula angustifolium. Other studies using lavender have shown it to have spasmolytic or muscle relaxing activity in humans, rats, and guinea pigs, which suggests that the essential oil lavender contains substances that induce relaxation (Lis-Balchin and Hart 1999).

Diego et al. (1998) demonstrated that lavender increases beta waves in the brain, which suggests increased drowsiness. Subjects also had less depressed moods, reported feeling more relaxed, and performed mathematical computations faster and more accurately after the application of aromatherapy. However, the mechanism producing this phenomenon remains unexplained. Other studies have suggested that the pleasantness of the aroma may be the controlling factor (Baron and Bronfen 1994).

Figure 1. Study Design

The present study proposes a possible mechanism for this increase in mathematical or overall cognitive performance. The theory is that a person performs best at an optimum amount of arousal or stress according to the well-known Yerkes-Dodson inverted-U law (Yerkes and Dobson 1908) (Figure 1). According to this theory, if a person’s anxiety or arousal level is too low or too high, his performance will be lower compared to a person operating at the optimum level of stress or at the peak of the inverted-U curve. Since certain aromatherapy products, such as lavender and ylang-ylang, often claim to reduce anxiety, it is possible that this decrease in anxiety is what is causing the increase in mathematical computation performance from the study by Diego et al. (1998).

In the present study, we use this theory to indirectly examine the anxiety-reducing effects of aromatherapy. If aromatherapy truly reduces anxiety, then it should lead to increased cognitive performance. We hypothesize that the anxiety-reducing effects of aromatherapy can be effectively measured by the increase in cognitive and memory tasks performance. A secondary purpose of this study is to examine how variations in individual optimal stress levels due to personality characteristics can skew the inverted-U curve, affecting the experimental results. In addition, the effects of perceived pleasantness of aroma and tasks performance will also be examined.

Ylang-ylang: The essential oil ylang-ylang (Cananga odorata) was chosen for its previously identified relaxant quality (Price 1993). It was preferred over lavender because ylang-ylang was claimed to be more specific in reducing anxiety and stress, whereas lavender is more associated with an increase of general "well-being" (Price 1993). The brand name of the essential oil is The Body Spa, and is available at most supermarkets.

The experimenter gave the aromatherapy to the subjects using a "mock sensory experiment." The subjects were asked to smell 5 vials containing different dilutions of the aromatherapy product and were asked to determine the vial that did not have any odor. This procedure was performed prior to every cognitive and memory task. The subjects were also asked to rate the pleasantness of the odor prior to the administration of the anxiety inventories. The experimenter demonstrated the "wafting method" as a means of smelling the vials, which is a safe and effective technique that is often used in chemistry labs. The subject waves his or her hand over the vial, toward the nose (Appendix A). In the control condition, vegetable oil, which is odorless, was used instead of the aromatherapy product, ylang-ylang.

Digit Symbol Test: The digit symbol test was chosen to measure cognitive performance because it is a test that is sensitive to anxiety and stress. In fact, the performance on the Digit Symbol Test is so affected by anxiety that it is one of the three tests associated with the "Anxiety Triad" (Groth-Marnat 1997). The other two tests of the "Anxiety Triad" are the Arithmetic and Digit Span tests. The Digit Span Test was also used in this experiment.

The Digit Symbol Test is a subtest of the Wechsler Adult Intellignece Scale (WAIS). It is highly reliable and valid in measuring psychomotor speed, attention, and visual short-term memory (Groth-Marnat 1997). Subjects are required to enter the correct digit beneath each given symbol in an array of symbols as quickly as possible. The number of correct digits generated in 60 seconds is recorded.

The WAIS version of the digit symbol test was not used in this experiment because the test cannot be administered by undergraduates (Psychology resource center 2003). Instead, another version of the digit symbol test was created by the experimenters (Appendix B).

Digit Span Test: The Digit Span Test is another of the so-called "Anxiety Triad" tests, since a subject’s performance on the test is easily affected by his or her anxiety level (Groth-Marnat 1997). Subjects are required to recall and repeat numbers in the proper sequence.

The Digit Span Test is also part of the WAIS. It is a well-known, highly reliable and valid measure of short-term memory and attention. The test-retest reliability of the Digit Span Test is generally cited in the range of 0.80 (Groth-Marnat 1997).

The WAIS version of the Digit Span Test was not used for the same reason as the Digit Symbol Test could not be used. Instead, the experimenters created another version of the Digit Span Test (Appendix C). The criteria for making the test are very similar to that of the WAIS version. The sequences of digits that were used were randomly generated with the restriction that the numbers were not repeated and did not fall in numerical sequence (e.g., 4-7-2-9-3-5-1). The sequences of numbers were recorded on a cassette and played to the subjects during the experiment.

Pleasantness rating: The subjects rated the pleasantness of the aroma using a Likert scale of 0 to 4. A rating of 0 is very unpleasant, while 4 is very pleasant, and 2 is neutral (Appendix C).

State-Trait Anxiety Inventory (STAI): The STAI is a questionnaire that was used as a pre- and post-experiment test for measuring the anxiety levels of the subjects. State anxiety refers to the anxiety level during a particular situation, such as under experimental condition. Trait anxiety refers to the general anxiety level of the subject, which is usually more stable than State anxiety.

Eysenck Personality Questionnaire (EPQ). The EPQ consists of four scales: psychoticism, extroversion, neuroticism, and lie. In this experiment, the psychoticism scale was not used. A low score of extroversion indicates a tendency toward introversion. A high score in the lie scale means that the subjects have the tendency to give socially desirable responses, which may be used to generalize the likelihood to give socially desirable responses on the other scales. If the lie scale scores were too high, the EPQ test would be invalid. The EPQ scores of the subjects were compared with the population mean scores. The test-retest reliability for the EPQ is quite high. (Extroversion scale: α = .89; Neuroticism scale: α = 0.86; Lie scale: α = 0.84)

Personal information form. This form considered the possible quasi-variables, which are independent variables over which the experimenter has no control, and, that may affect the overall outcome of the experiment (Appendix D). For this experiment, the quasi-variables considered were age, sex, allergy, cold symptoms, medication, and hand dominance (or handedness).

Design: A single-blind, 2x2 mixed factorial design was used in this experiment. A 2x2 mixed factorial design consists of one within-subject variable (the aroma) with two levels (control and experimental conditions), and one between-subjects variable (the order of testing) with two levels (Group 1 and Group 2 arrangements). Single-blind refers to the deception technique used on the subjects about the true nature of the experiment. The subjects were randomly assigned into one of two repeat-measure conditions: Group 1 (control/experimental) and Group 2 (experimental/control). In the control condition, no aroma was used; in the experimental condition, the aroma of ylang-ylang was used. Both Group 1 and Group 2 were subjected to the same treatments.

Procedures: Participants were tested in groups. Prior to the experiment, subjects were informed about the experiments to be conducted without giving away the nature of the study. They were told that a sensory experiment involving smell, cognitive, and memory tests, and a few questionnaires would be performed.

The order of tests given in the Control Condition of Group 1 was: 1) Mock sensory experiment; 2) Digit Symbol Test; 3) A second Mock Sensory experiment; 4) Digit Span Test; 5) Pleasantness rating; 6) State-Trait Anxiety Inventory (Figure 2).

Figure 2. Yerkes-Dobson Inverted-U Law for task difficulties.

In Group 1, the Experimental Condition occurred immediately after the Control Condition. The procedures were the same as the Control Condition, except ylang-ylang was used instead of vegetable oil for the mock sensory experiments and the pleasantness rating.

In Group 2, the second half of the subjects (n=25) was experimented as a counterbalance for Group 1. This means that the Experimental Condition (with the ylang-ylang aroma) was given before the Control Condition (with no aroma). The order of testing was the same as Group 1 (Figure 2).

Both Group 1 and Group 2 were given the EPQ and personal information form after all the experiments were conducted. At the end of the study, all subjects were fully debriefed on the condition that they did not reveal to others the main purpose of the study.

Table 1 . Yerkes-Dobson Inverted-U Law for task difficulties.

Independent t-tests were used to investigate the order effect between Group 1 and Group 2. The results show that there is a significant difference between Group 1 and Group 2 for both the control conditions [t(48) = -3.814, p < .0001] and experimental conditions [t(48) = 3.215, p < .002].

The relationships between Digit Symbol scores and anxiety level and personality were investigated using Pearson r-correlation analysis, which showed that there was no relationship between the Digit Symbol and STAI scores. Analysis of variance (ANOVA) demonstrated that there were no relationships between Digit Symbol scores and EPQ scores on the extroversion, neuroticism, and lie scale for both Group 1 and Group 2.

As for the quasi-variables, subjects with allergies scored significantly higher than subjects without allergies in both the control conditions [t(46) = 2.190, p < .034] and experimental conditions [t(46) = 2.069, p < .044]. Otherwise, independent t-tests indicated that there were no relationships between Digit Symbol scores and age, gender, cold symptoms, or handedness. ANOVA indicates that there was also no relationship with the pleasantness of the smell [F(4,31) = .826, p < .519 (Control); F(4,31) = .607, p < .661 (Experimental)].

Digit Span Test: Paired-sample t-tests performed from the combined Digit Span scores of Group 1 and Group 2 yield a significant result [t(49) = 2.983, p < .004]. Although the differences of the mean correct Digit Span responses appear to be small (Table 1), the differences between control and experimental conditions were statistically significant.

Independent t-tests were used to investigate the order effect between Group 1 and Group 2. The results showed that there is no significant difference between Group 1 and Group 2 for the control condition [t(48) = .102, p < .919]. However, for the experimental conditions, there appears to be a small but significant difference between Group 1 and Group 2 [t(48) = 2.016, p < .049].

The relationships between Digit Span scores, and anxiety level and personality were investigated using Pearson r-correlation analysis, which demonstrated that there was no relationship between the Digit Span and STAI scores. ANOVA of Digit Span scores and EPQ scores on the extroversion, neuroticism, and lie scale demonstrated that there were no relationships for all subjects.

For the quasi-variables, independent t-tests indicated that there were no relationships between Digit Span scores and age, gender, cold symptoms, or handedness. Also, there was no correlation between the pleasantness of the smell and Digit Span performance [F(4,31) = .107, p < .979 (Control); F(4,31) = .673, p < .616 (Experimental)].

Participants Differences: The EPQ scores for all subjects were compared with the mean scores of the population of the same age group and gender (Eysenck and Eysenck 1975). One-sample t-test statistics indicated that there were no significant difference between the subjects and the population in the EPQ scores for the extroversion, neuroticism, and lie scales. Also, there was no significant difference in State anxiety [t(48) = .757, p < .453] and Trait anxiety [t(48) = .408, p < .685] between the subjects in Group 1 and Group 2.

Figure 3. Digit Span Test. Both Group 1 and Group 2 scored significantly worse in the experimental condition as compared to the control condition.

In the Digit Span Test, the subjects of both Group 1 and Group 2 did worse in the experimental condition than in the control condition (Figure 3). This was not anticipated, because we expected that the relaxing aroma would decrease anxiety, which would then lead to an increase in cognitive performance. These results, however, appear to be supported by the Yerkes-Dobson inverted-U Law for task difficulties (Figure 1). When arousal is too low, performance is impaired. Because the Digit Span Test is a relatively easy task, higher arousal was needed to produce better performance. Therefore, one of the assumptions of our hypothesis regarding the decrease in anxiety or arousal transpiring into better cognitive performance appears to have been incorrect. As it turns out, any significant change, whether it is an increase or decrease in cognitive and memory performance, can be viewed as an indication that the relaxing aroma was having an affect on the subjects. However, when the Digit Span scores were compared to the STAI and EPQ scores, there were no relationships between them. Therefore, we cannot directly attribute the worsening of the Digit Span performance during the aroma condition to either the decrease in anxiety level or certain personality characteristics, such as neuroticism.

One possible reason why there was no relationship between the Digit Span and STAI scores is that the changes in anxiety were very small, and the questionnaire was unable to measure such small changes. This lack of resolution from questionnaires in detecting anxiety changes seems to be a valid idea. Previous physiology research has suggested that small changes in anxiety are difficult to detect because they may not transpire into autonomic responses, such as heart rate and blood pressure (Saeki 2000). However, it is possible that these small changes in anxiety level, which the body seems impermeable to, may have transpired into a reduction in cognitive and memory task performance.

In contrast to the Digit Span Test, our version of the Digit Symbol Test was negatively affected by the order effect, meaning that the order in which the tests were conducted affected the results. The scores between Group 1 and Group 2 were significantly different, which means the subjects always did better in the post-test; it did not matter if it was during the control (no aroma) or experimental (aroma) conditions. This is also where the study by Diego et al. (1998) may be flawed. They did not take into account the order effect; therefore, they would always find a significant increase in mathematical computational score in the aroma condition, which always occurred in our post-testing. However, there was no such order effect for the Digit Span Test in our experiment; both Group 1 and 2 did better in the control (no aroma) condition, which also contradicts the Diego group’s findings. There are a few possible explanations for this: 1) The Diego study may be inherently inconsistent, since it did not counterbalance the subjects and did not provide a theoretical reason why subjects should perform better with aromatherapy. 2) The Diego study used a different cognitive measure (math computation) compared to our Digit Symbol Test. It is possible that aromatherapy may affect mathematical computation abilities differently. 3) Although the Diego group did not not take the order effect into account, the difference between the control and experimental groups may be large enough that the order effect could mask some of the effects of the aroma.

Similar to the Digit Span Test, there were no relationships between Digit Symbol scores and anxiety level (STAI) or between Digital Symbol scores and personality traits (EPQ). This may be explained by the same problem of low resolution with the self-report questionnaires in detecting small anxiety changes.

Figure 4. Allergies and Digit Symbol Test. Subjects with allergies scored significantly better on the Digit Symbol Test than subjects without allergies.

Interestingly, one of the quasi-variables was presence of allergies. The subjects with allergies (n=16) did significantly better than the subjects without allergies (Figure 4). This may be due to the smaller sample size (16 subjects); however, the experiment design was repeated-measure, which means that it was less prone to individual differences affecting results. Possiblely, subjects with allergies performed better because they were less able to smell the aroma. Another possiblity is that the subjects with allergies were on medication, which can lead to an increase in performance. One common ingredient in allergy medication is pseudoephedrine, which is banned from use in the Olympics because it of its suspected performance-enhancing capabilities (Olympic movement anti-doping code 2003).

One important purpose of our study was to assess the effects of the aroma chemical on the dependent variables (Digit Symbol and Span Test scores and anxiety levels) and not the psychological effects associated with the pleasantness of the smell. For both the Digit Symbol Test and Digit Span Test, there was no relationship with the subjects’ perception of the pleasantness of the aroma. This suggests that the subject’s performances were not based on their preconceived idea that a more pleasant-smelling environment would lead them to a better mood (Baron and Bronfen 1994). Instead, the study is suggesting that some chemicals in the aroma may be causing the changes in anxiety level.

There are sound theoretical reasons and experimental evidence from the present study that cognitive and memory tests are useful for indirectly measuring the anxiety effects of aromatherapy. However, it appears that not all cognitive and memory tests are sensitive to anxiety and stress. For example, this study has shown that the Digit Span Test is more sensitive to low arousal level than then the Digit Symbol Test.

For future study, our findings should be compared to an experiment with an aroma that has an opposite effect. If the aroma of ylang-ylang truly causes a reduction in arousal, as shown by a decrease in Digit Span performance, an interesting study would be to see whether a stimulating aroma, such as rosemary (Price 1993), causes an increase in arousal, and hence an increase in Digit Span scores.

Another possible research avenue is the investigation of the psychological effects that are always associated with consumer products. In our study, the subjects were blinded to the nature of the experiment; however, it would be interesting to see whether knowing beforehand that aromatherapy is being applied would affect the performance of the subjects.

Finally, the key limitation to the present study is the unavailability of the commercial cognitive and memory tests, which are under the more sensitive categories of IQ tests. Our versions of the Digit Span and Digit Symbol Tests appear to be reliable, but may be less effective than the more standardized tests, such as the WAIS. This was especially the case with the Digit Symbol Test, since the task of designing symbols was relatively more complex than the Digit Span Tests, which required only recalling and repeating sequences of numbers.

Appendix A. Wafting Method

(Click to view enlarged image)

Appendix B. Digit Symbol Test

(Click to view enlarged image)

Appendix C. Digit Span Test

(Click to view enlarged image)

Appendix D. Personal Information Form

(Click to view enlarged image)

Baron RA, Bronfen MI. (1994). A whiff of reality: Empirical evidence concerning the effects of pleasant fragrances on work-related behavior. Journal of Applied Social Psychology. 24, 1179-203.

Cohen S et al. (1993). Negative life events, perceived stress, negative affects, and susceptibility to the common cold. Journal of Personality and Social Psychology. 64, 131-40.

Diego MA et al. (1998). Aromatherapy positively affects mood, EEG patterns of alertness and math computations. International Journal of Neuroscience. 96(3-4), 217-24.

Eysenck HJ, Eysenck SBG (1975). Manual of the Eysenck Personality Questionnaire. Edits/educational and industrial testing service: San Diego.

Goldstein EB. (2002) Sensation and perception, 6th ed. Wadsworth-Thomson: Pacific Grove, CA.

Groth-Marnat G. (1997). Handbook of Psychological Assessment (3^rd ed.) John Wiley & Son: New York.

Lis-Balchin M, Hart S. (1999). Studies on the mode of action of the essential oil of lavender (Lavadula augustifolia P. Miller). Phytotherapy Research. 13, 540-542.

Lorig TA, Schwartz GE. (1988). Brain and odour: alteration of human EEG by odour administration. Psychobiology. 16, 281-284.

Olympic movement anti-doping code. (January 1, 2003). Prohibited classes of substances and prohibited methods. Accessed April 13, 2003: http://multimedia.olympic.org/pdf/en_report_542.pdf

Price S. (1993). The aromatherapy workbook. Thorsons: London.

Psychology resource center. (n.d.) Psychological tests. Accessed April 13, 2003: http://www.yorku.ca/psycentr/tests.html

Saeki Y. (2000). The effect of foot-bath with or without the essential oil of lavender on the autonomic nervous system: a randomized trial. Complement Therapeutic Medicine. 8(1);2-7.

Weiten W. (1998). Psychology: Themes & Variations (4^th ed.) Brooks/Cole: Washington, D.C.

Yerkes RM, Dodson JD. (1908). The relationship of strength of stimulus to rapidity of habit-formations. Journal of Comparative Neurology and Psychology. 18, 459-482.