Every participant signed a consent form after we explained the study design and possible risks and benefits to the participants. We gave the participants a plastic freezer bag that contained the materials for the study. This included a white, extra-large, 100% cotton t-shirt, half of a piece of unscented soap, a small bottle of unscented shampoo, and one unscented laundry detergent pack. Each freezer bag was numbered, and we assigned each participant a random number. We asked them to wash their sheets with the detergent and take a shower with the provided materials before wearing the t-shirt for two nights. We asked participants to refrain from using deodorant or perfumes, eating, or sharing their beds with other people or pets, engage in sex with another person, avoid spicy and pungent foods (e.g., onions, garlic, etc.), or to smoke while wearing the t-shirts. We also gave the participants a link for a questionnaire on Qualtrics that included several surveys to measure stress levels and emotional states of the t-shirt wearers described in detail in the next section.
After the first group of participants wore the t-shirts, we asked them to return the t-shirt in the plastic bag, which we then stored in a freezer at 0℉. Due to the bacterial activity involved in the samples, freezing is a common practice in t-shirt studies to limit further bacterial action post-collection which may alter the perceptions of odors. Following the procedure of previous studies, we removed the t-shirts from the freezer and allowed them to thaw before participants rated them. Lenchova, Roberts, & Havlicek (2009) have demonstrated that these freeze-thaw cycles do not have an impact on the ratings of the t-shirts.
The second group of participants responded to a Qualtrics survey which cycled randomly through all the t-shirt numbers. For each t-shirt, we asked the participant the next number on the survey, retrieved that t-shirt, opened and folded down the sides of the bag, and had them smell the t-shirt for 10 seconds. The participant then rated each shirt on pleasantness (1 = very unpleasant, 10 = very pleasant), intensity (1 = not intense, 10 = very intense), and disgust (1 = not disgusting at all, 10 = very disgusting).
We also collected skin conductance data from each participant while they smelled the t-shirts using eSense Skin Response sensors (Mindfield Biosystems, Gronau, Germany). Before the procedure started, we attached the sensors to their index and ring fingers between the basal and first joint on their non-dominant hand and collected skin conductance data each time they smelled a new shirt.
We also collected any comments the participants had about each t-shirt, though they are not analyzed in the current study. After they had smelled and rated all the t-shirts, the participants completed a survey with demographic information.
Some participants received extra credit in classes for participating, but students were offered alternative extra credit activities as well. All methods were approved by the University of Alabama Institutional Review Board (Protocol ID #17-11-693).
Questionnaire And Inventories
The questionnaire administered to the t-shirt wearers and raters both began with the same demographic questions, which included age and two questions on gender and biological sex. The biological sex of the participant was measured as a separate question and was combined with information on hormonal contraceptive use by providing participants the prompt to “Please select which applies to you” with the following three options: 1) female taking birth control (this includes IUD’s), 2) Female not taking birth control, 2) male. We collected data on Hormonal contraceptive use information, as previous research has found that this can impact the scent of t-shirts (Singh & Bronstad, 2001; Trouton, Guitar, Carmen, Grandis, & Geher, 2012). We also asked about the date of last menstrual cycle and the average length of menstrual cycle with the intention of calculating the estimated ovulatory phase. However, due to a small sample size and the inaccuracy of predicting ovulation without a biomarker, ovulatory phase was not included in the analyses. Moreover, as discussed in the results section, ratings of the t-shirts did not vary based on birth control use or gender, further limiting the utility of information about menstrual cycle in the current analysis.
To measure stress and emotional states in the t-shirt wearers, we used a variety of scales that asked about different periods of time. Since the t-shirt wearers wore the t-shirt over the course of two nights, we chose to use scales that assess stress and emotional states over longer periods of time. For example, from the State-Trait Anxiety Scale, we used the Trait Anxiety Scale, which inventories anxiety levels in general, but not the State Anxiety Scale, which only inventories anxiety at the moment (Spielburger et al., 1983). The Trait Anxiety Scale asks about emotional states at this moment and has respondents rate on a 1–4 scale how well they agree with statements such as “I feel calm” and “I am jittery” (Spielburger et al., 1983). To look at stress and affect over the past month, we used the perceived stress question from the PANAS-X inventory. The PANAS-X inventory asks respondents to rate on a 1–5 scale the extent to which they have felt, for example, “afraid,” “nervous,” enthusiastic,” or “strong” in the past few weeks (Watson and Clark, 1993). The Pittsburgh Sleep Quality Index was also administered to assess sleep quality, which asks respondents to describe their sleeping habits and rate how often they have trouble sleeping for a number of reasons, such as because they “cannot breathe comfortably” or “feel too hot” (Buysse et al., 1989).
Data Analysis
In this study, we were interested in how stress impacts perceptions of scent, which we tested through the stinky t-shirt model, in which t-shirts were worn by individuals over the course of two nights and then rated by a separate group of t-shirt smellers. We used four main mood inventories to investigate the stress of the t-shirt wearers in this study: the State-Trait Anxiety Scale, Perceived Stress, Pittsburgh Sleep Quality Index, and the PANAS-X inventory. From these inventories we extracted the following independent variables for each t-shirt wearer: trait anxiety, perceived stress, positive affect, negative affect, and sleep quality. We measured the t-shirt smeller’s perceptions of the shirts through Likert scale ratings of disgust, intensity, and pleasantness. Additionally, we collected the t-shirt smeller’s electrodermal activity (EDA) as they smelled each shirt through eSense monitors. The dependent variables for all analyses were the respective averages of the Likert scales and the EDA data.
To analyze the results, we scored the wearer responses for perceived stress, state anxiety, trait anxiety, and positive and negative affect. Scoring involved adding up the wearers’ responses to the questions from each inventory and comparing them to each inventory’s scoring guide to produce a numerical score for each participant and inventory. We then ran a regression analysis between the wearer responses and the smellers’ ratings of them. Then we ran separate linear regression analyses for each factor of the wearer responses to determine if state anxiety, trait anxiety, perceived stress, negative affect, or sleep quality were related to average ratings of disgust and intensity in separate models. Additionally, we ran a linear regression to examine if positive affect predicted ratings of pleasantness for the t-shirts.
To analyze the EDA data, we calculated the standard deviations for each t-shirt smeller and then used the average of these across the participants to provide a standardized measurement of how each t-shirt evoked EDA changes in the smellers. These averages were then used as the dependent variable in regression analyses with the same independent variables mentioned above (state anxiety, trait anxiety, perceived stress, positive affect, negative affect, and sleep quality). Values were considered significant if p < .05.