Warm Water and Loneliness Redux: Rejoinder to Bargh … to Bargh and Shalev (2014) 2 Abstract Bargh and Shalev (2014) replied to our work and summarized results from three new studies

Rejoinder to Bargh and Shalev (2014) 1

Warm Water and Loneliness Redux: Rejoinder to Bargh and Shalev (2014)

M. Brent Donnellan, Richard E. Lucas, Joseph Cesario

Michigan State University

Draft Date: 11 November 2014

Address correspondence to:

M. Brent Donnellan

Department of Psychology

Texas A & M University

4235 TAMU

College Station, TX 77843

Note: Ryne Sherman and Eric-Jan Wagenmakers provided helpful comments. This acknowledgement does not mean that they agree with our interpretations or analytic strategy. Brent Donnellan is now at Texas A & M University. This was recently just accepted for publication by Emotion (http://www.apa.org/pubs/journals/emo/). Once we sign all of the papers APA will own the copyright. This article may not exactly replicate the final version published in the APA journal. It is not the copy of record.


Abstract

Bargh and Shalev (2014) replied to our work and summarized results from three new

studies concerning the associations between trait loneliness and showering/bathing habits. We

clarify a few issues and provide a foundation for future work by conducting a meta-analysis of

the relevant studies. The inclusion of new data does little to change our basic conclusions. There

are no indications of strong connections between trait loneliness and showering/bathing habits.

Additional studies are needed to test moderators of these associations and to evaluate possible

cross-cultural differences in the connection between loneliness and physical warmth extraction

from baths and showers.

Word Count: 98

Key words: Loneliness, Effect Sizes, Statistical Precision, Replication, Meta-Analysis


Do lonely individuals use warm showers and baths to compensate for a lack of social

connection? This provocative question has now generated a surprising number of empirical

investigations. Bargh and Shalev (2014) replied to our paper and summarized results from three

new studies. An inspection of the size of the correlations from these new studies suggests their

new effect size estimates are closer to our estimates than to those reported in their 2012 paper.

Thus, their reply reinforces the concerns motivating our replication studies and does little to

answer questions about the anomalies in their Study 1a. In this rejoinder, we respond to a few

issues and quantitatively summarize the results from the existing studies in this exchange.

Is the Physical Warmth Extraction Index Variable the Most Valid Test of the

Substitutability Hypothesis?

Bargh and Shalev (2012) created a physical warmth extraction variable by taking the

average of the standardized frequency, warmth, and duration items. They argue that this

summary variable captures how much net warmth participants extract from bathing activity (see

their p. 156). Bargh and Shalev (2014) justified this index by suggesting that their composite

measure is akin to a snacking index. However, we believe their snacking index analogy breaks

down in this context. Pies, candy bars, and ice cream all contain calories and they all belong to a

superordinate snack category. In contrast, the frequency and duration items refer to different

aspects of showering/bathing activity and they are conceptually unrelated to warmth. Moreover,

these items are not strongly correlated with the temperature variable (Frequency: r = .04 in the

combined file for our Studies 1 to 4, r = .02 in the combined file for our Studies 5 to 9; Duration:

r = .23 in the combined file for our Studies 1 to 4, r = .14 in the combined file for our Studies 5

to 9). Participants who report taking longer showers/bath also report taking warmer


showers/baths but it is not the case that participants who report taking more showers/baths also

report taking warmer showers/baths. The frequency and duration items are negatively correlated

(r = -.10 in our Studies 1 to 4, r = -.12 in the combined file for our Studies 5 to 9). Participants

who take longer shower/baths have a slight tendency to take fewer baths and showers per week.

Ultimately, their arguments did not change our perspective on the relative merits of the

warmth item versus the physical warmth extraction index when testing the relevant hypothesis.

As it stands, we believe it is reasonable to place more emphasis on the warmth item when

evaluating the evidence for the ideas presented in Bargh and Shalev (2012). The warmth item is

explicitly about water temperature and is therefore directly relevant to the substitutability

hypothesis. We found virtually no support for the idea that lonelier people take warmer showers

and baths using a straightforward approach for testing the underlying ideas in the Bargh and

Shalev paper. Likewise, frequency should be positively correlated with loneliness if one follows

the logic of the Bargh and Shalev arguments; however, there was no indication that frequency

was positively correlated with loneliness in our studies or in their Study 1b. At best, there

appears to be a small positive correlation between trait loneliness and the duration of showers

and baths.

On the other hand, it is possible that at least among those who take warm showers,

variability in frequency and duration are the primary means by which these individuals extract

warmth. This possibility suggests an interaction between temperature and the other two

variables: Among those who take warmer showers, duration and frequency should be positively

associated with loneliness. However, these interactions did not reach statistical significance in


either of our combined files (the warmth by duration interaction: p = .497 for Studies 1 to 1, p =

.667 for Studies 5 to 9; the warmth by frequency interaction: p = .196 and p = .808, respectively).

What are the Effect Size Estimates in their New Data?

Bargh and Shalev graciously shared the raw data from their new studies. To facilitate

comparisons with the existing studies, we averaged the separate bathing and showering items in

each of their new studies to calculate overall frequency, warmth, and duration composite

variables. We then standardized these composite variables within each sample to create the

overall index composite variable in each sample. We also combined their data into a single file to

compute overall correlations across their new data (akin to the approach they used in Table 3 of

their online supplement). The correlations along with sample sizes are reported in Table 1. The

correlation between loneliness and the overall index variable was statistically significant at p <

.05 (r = .14) as was the correlation for the duration variable (r = .23). However, there was no

overall effect for the warmth variable (r = -.01, p = .806) or for the frequency variable (r = -.01,

p = .774). The near zero overall correlation for the physical warmth variable is particularly

striking and weakens support for the idea that lonely people prefer warm temperatures for baths

and showers using their own data.

In addition, we have concerns about their mTurk sample from India given differences in

the relative amount excluded data from the samples collected from the United States versus

India. Bargh and Shalev (2014) excluded 33.6% of the data collected from India compared to

9.6% of the data from the United States. As we noted in our original paper, we agree that it is

reasonable to try to identify attentive from non-attentive respondents when using mTurk samples

(see also Berinsky, Margolis, & Sances, 2014; Goodman, Cryder, & Cheema, 2013). However,


we believe researchers should provide the explicit rationale for exclusions and report what

happens when all participants are included in the analyses (see also Berinsky et al., 2014, p.

751). Bargh and Shalev did not describe the precise criteria they used to discard participants and

they did not report key results with all participants included. In contrast, we were explicit about

our exclusion procedures and also reported results with all participants throughout our report. In

short, we believe the mTurk sample from India should be viewed with some caution until the

results are replicated and supplemented with additional cross-cultural samples beyond those

derived from mTurk. This is especially important given possible differences between mTurk

workers from the United States and India beyond just their country of origin (Ipeirotis,

2010). The potential for cross-cultural differences is intriguing and it would be ideal to have

more data on this issue.

What Happened with the Frequency Variable In Study 1a?

We are still perplexed by the frequency distribution reported for their Study 1a. Bargh

and Shalev (2012) asked participants “How often you usually take a bath/shower?” for both

studies and yet the resulting distributions are substantially different in their Studies 1a and 1b.

Bargh and Shalev explain this anomaly by suggesting that the participants in their first study

misunderstood the question that was being asked. It is difficult to understand how undergraduate

participants from an elite university could have misunderstood this item in such a way as to

produce a distribution whereby 90% of the participants take less than 1 shower/bath per week.

This was our biggest reservation with their Study 1a and none of the frequency distributions for

samples from the United States look anything like the distribution in their Study 1a. Moreover,

none of the available studies have been able to duplicate their initial correlation of .48 between


trait loneliness and frequency of showering/bathing from Study 1a. In fact, the correlation

between loneliness and the frequency item composites was negative and consistently near zero in

all three of their new studies (see Table 1). Accordingly, we think it would be prudent to discard

Study 1a from the scientific record.

Are Our Studies Really That Different from Their Studies?

Bargh and Shalev (2014) suggest that differences in measures and procedures across our

nine studies are a concern that precludes taking a meta-analytically-informed perspective on the

existing literature. We disagree strongly with this perspective and we doubt that most experts in

meta-analysis would endorse this view. Moreover, we think the practice of embedding these

particular questions within larger surveys would actually mitigate their concerns about

participant awareness. Taking a step back, however, it is unclear why awareness of the purpose

of the study would systematically bias the results one way or the other. It is entirely possible that

participants would provide responses with an acceptable degree of fidelity to their actual

thoughts, feelings, and behaviors even if they were aware that researchers were truly interested in

testing whether there is an association between loneliness and showering/bathing habits.

Nonetheless, Bargh and Shalev (2014) correctly noted that we did not question our

participants about their awareness of the study hypothesis.1 To address this gap, we conducted an

additional study (see online supplement) to evaluate how removing any aware participants would

impact our results. As it turned out, removing any participant suspected of study awareness by at

1 We did not previously think this was an issue because no one in their 2012 paper expressed accurate awareness of the purpose of their questionnaire studies


least one of the three authors did not change our effect size estimates to any appreciable degree.

Thus, we do not think participant awareness is a critical confound in our studies.

Ultimately, we believe that the any differences across our studies should be seen as a

virtue rather than a weakness especially as it concerns the generalizability of the basic findings.

We should also clarify a point about our procedures so readers can better understand our

decisions. Some of the differences in measures stem from the timing of our email exchanges.

We received an initial copy of their lifestyles questionnaire on 28 February 2012 but that survey

asked participants to respond to the warmth item using picture of a thermometer rather than a

survey item as described in the 2012 paper. Given the discrepancy, we asked for clarification but

the response time did not work for our research goals. We wanted to collect data before the end

of our Spring semester so we created an item modeled on the description in their published paper

(i.e., our warmth item was measured on a 7-point scale ranging from cold to very hot, p. 156).2

We received a copy of the correct survey on 30 May 2012 but this was after we had already

collected data for our Study 1. At that point we decided to keep using our survey questions given

the similarities across the measures (see the Appendix of Donnellan et al., 2014). To be sure,

there is no theoretical reason to expect that differences in the two sets of measures would impact

the ability of researchers to find a connection between showering/bathing habits and trait

loneliness.

2 The description of the items in the published paper does not consistently refer to showers/baths but we

assumed this was an oversight.


What Happens When All Data Are Combined?

In light of our preference for meta-analytic approaches to research questions, we

combined the correlations from Table 1 with the 9 studies in our original article along with 3

more studies posted on the Psychology File Drawer (Donnellan & Lucas, 2014; Ferrell, Gosling,

& Donnellan, 2014; McDonald & Donnellan, 2014) and the new study described in the online

supplement for this rejoinder. The correlations and overall meta-analytic effects sizes are

displayed in Table 2. As seen in the bottom rows of Table 2, the aggregate effect size estimates

were very similar to the ones reported in our original paper. Lonely people appear to take longer

showers/baths but they also take fewer showers/baths per week. The most hypothesis relevant

estimate for the warmth item was not significantly different from zero. Thus, our original

conclusions about the connections between bathing habits and trait loneliness are essentially

unchanged when incorporating this new information.

There are at least two other aspects of Table 2 that are worth emphasizing. First, the .48

correlation between frequency and loneliness from Bargh and Shalev (2012) is a noticeable

outlier given that the next largest positive correlation is .08 (our Study 3). The Q-statistic is not

significant for the remaining 17 studies once this .48 effect is discarded (Q = 17.587, df = 16, p =

.349). These considerations reinforce our concerns about the anomalous results from Study 1a.

Second, the correlations from the University of Texas study are especially informative

because they were collected by outside researchers who administered surveys to a large live-

broadcast online course (Ferrell et al. 2014). There was an approximately 1-week gap between

the lifestyle questions and the loneliness measure. College students completed the surveys

simultaneously wherever they watched the broadcast lectures (i.e., they were not all in the same


room when they answered the questions). Donnellan analyzed the data for the final write up for

the Psychology File Drawer. In other words, an independent research group obtained results

more in line with those reported in Donnellan et al. (2014) than those reported in Bargh and

Shalev (2012). Ultimately, we believe that independent inquiry into these questions from

additional labs is necessary to clarify the empirical relations between trait loneliness and

showering/bathing habits.

Summary and Conclusion

The new data presented by Bargh and Shalev (2014) does not change our conclusions

about the connections between loneliness and showering/bathing habits. Their new studies used

larger sample sizes compared to their original studies but yielded smaller effect size estimates.

When we combined the relevant data using meta-analytic procedures, the overall effect size

estimates for the correlation between loneliness and the warmth variable was trivial and not

statistically distinguishable from zero. Thus, we believe extreme caution is warranted before

asserting that lonely people take warm showers to compensate for a lack of social connection.

Lonelier people report taking fewer showers/baths but they also report taking longer showers and

baths. As is always the case, additional research by outside groups is needed. In particular, future

studies should evaluate cross-cultural differences and test theoretically-relevant potential

moderators of these relations.


References

Bargh, J. A., & Shalev, I. (2012). The substitutability of physical and social warmth in daily life.

Emotion, 12, 154-162.

Berinsky, A. J., Margolis, M. F., & Sances, M. W. (2013). Separating the shirkers from the

workers? Making sure respondents pay attention on self‐administered surveys. American

Journal of Political Science,58, 739-753.

Donnellan, M. B. & Lucas, R. E. (2014). Showering and Loneliness Take 10. (2014, April 29).

Retrieved 12:58, July 01, 2014 from

http://www.PsychFileDrawer.org/replication.php?attempt=MTg2

Ferrell, J. D., Gosling, S. D., & Donnellan, M. B. (2014) Showering and Loneliness: U of T

Replication. (2014, April 30). Retrieved 12:58, July 01, 2014 from

http://www.PsychFileDrawer.org/replication.php?attempt=MTg3

Goodman, J. K., Cryder, C. E., & Cheema, A. (2013). Data collection in a flat world: The

strengths and weaknesses of Mechanical Turk samples. Journal of Behavioral Decision

Making, 26(3), 213-224.

Ipeirotis, P. (2010). Demographics of Mechanical Turk. CeDER-10-01 working paper. New

York University.

McDonald, M. & Donnellan, M. B. (2014). Showering and Loneliness: Participants from Israel.

Retrieved 10:10, October 07, 2014 from

http://www.PsychFileDrawer.org/replication.php?attempt=MTk5


Russell, D. W. (1996). UCLA Loneliness Scale (Version 3): Reliability, validity, and factor

structure. Journal of Personality Assessment, 66, 20-40.


Table 1: Correlations between Trait Loneliness and Bathing/Showering Items – Original and New Bargh and Shalev Studies

Frequency Temperature Duration Index N

Original Studies Study 1a .48* .26 .29* .57* 51 Study 1b .03 .34* .33* .37* 41

New Studies

Israel -.005 .046 .179* .121 148 India -.009 .207* .026 .136 199

United States -.028 .068 .254* .160* 206 Overall -.012 -.010 .230* .137* 553 95% CI -.089 to .064 -.106 to .077 .147 to .306 .056 to .215 Note: The overall correlation was based on an aggregate data file that combined the three new studies into a single file. 95% Confidence Intervals constructed with Biased-Corrected Bootstrapping Procedures with 1,000 resamples using the aggregated data file. * p < .05.


Table 2: Meta-Analytic Results

Frequency Temperature Duration Index Bargh and Shalev

Studies Bargh and Shalev 1a .48 .26 .29 .57 Bargh and Shalev 1b .03 .34 .33 .37

Israel -.01 .05 .18 .12 India -.01 .21 .03 .14

United States -.03 .07 .25 .16 Donnellan et al. Studies

Study 1 -.15 -.03 .07 -.06 Study 2 -.11 .02 .08 -.01 Study 3 .08 -.02 .17 .13 Study 4 -.20 .01 .02 -.10 Study 5 -.07 .06 .18 .10 Study 6 -.13 .07 .16 .06 Study 7 -.10 .03 .11 .02 Study 8 -.08 .01 .10 .02 Study 9 -.08 -.12 -.01 -.13

MSU Spring 2014 -.06 .00 .13 .04 MSU Summer 2014 -.03 -.07 .01 -.05 University of Texas -.07 -.08 .09 -.03

Israel M & D -.15 -.05 .16 -.03 N 5,289 5,285 5,290 5,293

Random-Effects Model

Point Estimate -.067 .017 .115 .043 95% CI -.107 to -.027 -.020 to .055 .081 to .149 -.006 to .092

p .001* .356 .000* .086 Fixed-Effect Model

Point Estimate -.076 .015 .116 .031 95% CI -.103 to -.049 -.012 to .042 .089 to .142 .004 to .058

p .000* .273 .000* .026* Note: Sample sizes of 51 and 41 were used for Bargh and Shalev Studies 1a and 1b, respectively. Effect sizes for Study 6 and the MSU Summer 2014 study were calculated using the average of the z-scored UCLA Loneliness measure and the Bargh and Shalev Loneliness measure. University of Texas data are from Ferrell et al. (2014) and the Israel M & D data are from McDonald and Donnellan (2014).


Supplemental Online Study

Testing the Impact of Participant Awareness and Lay Theories

Sample

Participants were 323 college students (66.6% women; M age = 20.66 years, SD = 2.27)

who received course credit as part of the Michigan State University Psychology Subject Pool

during the First and Second Summer Sessions of 2014 (12 May 2014 through 14 August 2014).

The majority of participants self-identified as white (71.2%). All data were collected online

throughout the entire summer session period.

Procedure

We randomly assigned participants to complete the lifestyle items first (n = 156) or the

Bargh and Shalev loneliness questions first (n = 167). Thus, the first part of the study was a

close replication of Study 1a in Bargh and Shalev (2012) using an online platform. Participants

also completed the 36-item Experiences in Close Relationships-Revised questionnaire (Fraley,

Waller, & Brennan, 2000) with items presented in a random order to each participant. This

measure was used to test whether attachment dimensions moderated connections between

loneliness and showering in line with recent findings that attachment styles moderate the impact

of temperature manipulations on prosocial outcomes in children (e.g., IJzerman, Karremans,

Thomsen, & Schubert, 2013; see also Ferrell et al., 2014). However, no interaction model

produced a significant overall F value at p < .05 out of the 12 models we estimated. Given the

number of analyses we conducted and the null results of Ferrell et al. (2014), we did not pursue

these interactions further


After answering attachment questions, participants completed the UCLA Loneliness scale

(Russell, 1996) and demographic items. We included a directed response item embedded in the

demographics questionnaire that asked participants to select the “rarely” response option. We

also included the “I responded to this survey honestly” item right before the question about the

purpose of the study. Only those participants who passed both of these checks were included in

these analyses (approximately 85% of the data collected) following procedures in Donnellan et

al. (2014).

At the bottom of the demographic questions, participants were asked “What do you think

was the purpose of this study?” and presented with a text box to record their answers (34

participants left this box blank). Participants were then presented with a page break and asked

three closed ended question with Yes, Maybe, or No response options: “Do you think there is a

connection between loneliness and a preference for water temperature?”, “Do think lonely

people prefer cold temperatures for baths/showers?”, and “Do you think lonely people prefer

warm temperatures for bath/showers”. Participants were then asked a final open-ended question

“Why might there be a connection between showering/bathing habits and loneliness?” and

presented with a text box to record their answers (30 participants left this box blank).

Coding of Awareness

All three authors independently coded responses to the purpose question using a simple

three category system (Guessed Purpose; Unsure/Potential Awareness; No). The authors coded

these responses using a file stripped of responses to the quantitative questions. No awareness was

the most common code for all three raters (MBD: 93.8%, REL: 98.5%; JC: 91.0%). Agreement

was formally indexed using kappa across all three pairs (MBD and REL: .389; MBD and JC:


.525; REL and JC: .220). Given that these kappas were not terribly impressive, we adopted the

most inclusive approach possible by flagging any participant suspected of awareness by at least

one coder (32 participants or 9.1% of the sample).

Results

There was no evidence for a statistically detectable association between loneliness as

assessed with the Bargh and Shalev measure (M = 2.02; SD = .59, alpha = .89) and the physical

warmth index (r = -.07, p = .234, n = 323). This was also the case evaluating the correlation for

the UCL A loneliness measure (M = 2.06; SD = .49, alpha = .93; r = -.03, p = .640, n = 323).

The same null results were obtained when we used all participants with relevant data (Bargh and

Shalev measure: r = -.10, p = .054, n = 374; UCLA measure: r = -.06, p = .251, n = 367).

Separate analyses with each shower/bath item also did not support the substitutability predictions

(see Supplemental Table 1). Thus, we did not replicate any of the results from Bargh and Shalev

(2012) in these data. The correlation between the two loneliness measures was .71 so we created

a composite loneliness measure for use in the meta-analysis.

The ordering of survey questions did not moderate the reported association between the

extraction index variable for the Bargh and Shalev loneliness measure (p = .833 for the

interaction term). Correlations for each condition are reported in Supplemental Table 1. There

was no evidence of moderation for any of the specific items (minimum p = .068 for duration) and

the same null results were obtained for the UCLA loneliness measure (minimum p = .077 for

duration) and the composite loneliness measure (minimum p = .052 for duration). Critical

readers may observe that the interaction for duration were close to the alpha of .05. However, an


inspection of Table 1 indicates that although the direction of effect was different for the two

conditions, the correlations were not statistically significant in either condition.

Additional Analyses

We computed correlations for the 291 participants who were not suspected of any

awareness by any coder. The same null results for the overall sample were obtained for this

sample as show in Supplemental Table 1. The effect size estimates were quite similar to the

overall results. Thus, there were no indications that participant awareness distorted conclusions

drawn from this study.

We also considered the distributions for questions about participant intuitions regarding

connections between water temperature and loneliness. These are displayed in Supplemental

Table 2. Approximately 12% of these participants answered “Yes” to the question about a

connection between loneliness and a preference for water temperatures. Perhaps the most

informative distribution occurred for responses to the question about whether lonely people

prefer warm temperatures for showers/baths. This distribution was not significantly different

from a distribution that would be expected if participants responded at random to this question

(i.e., a data generating process whereby each category had a 1/3 chance of being endorsed; chi-

square value = 3.981, df = 2, p = .137). This is consistent with the idea that college students are

largely unaware of an explicit connection between warm water and loneliness as argued in Bargh

and Shalev (2012).

A critical question was whether intuitions about the underlying research topic would

impact the correlational findings. Accordingly, we examined the correlations between the


aggregate loneliness measure and the frequency, temperature, duration, and index variables

separately by responses to the intuition questions. (We used the aggregate measure to reduce the

number of comparisons as a way to curb Type I errors.) For example, we computed the

correlation between the composite loneliness measure and the physical warmth index separately

for those who indicated “yes” to the question about a connection between loneliness and a

preference for water temperature (n = 38), those who indicated “maybe” (n = 154), and those

who indicated “no” (n = 131). Across all comparisons, there were two positive correlations that

reached statistical significance. Duration and loneliness were positively correlated (r = .65, p =

.002) within the sample of 21 people who answered yes to the question about whether lonely

people prefer cold temperatures. Likewise, duration and loneliness were positively correlated (r

= .22, p = .015) within the sample of 121 people who answered maybe to the question about

whether lonely people prefer warm temperatures. Both of these results are consistent with the

meta-analytic results albeit with exaggerated effect sizes. Accordingly, we concluded there was

little reason to worry that participant intuitions were a major confound. An excel file with these

correlations is available upon request.


Supplemental Table 1: Correlations between Trait Loneliness and Bathing/Showering Items –MSU Summer Session 2014 Study

Frequency Temperature Duration Index

Overall Sample (n = 323)

Bargh and Shalev

-.034 -.054 -.023 -.066

UCLA -.017 -.067 .041 -.026

Composite

-.028

-.066

.010

-.050

Lifestyle First (n = 156) Bargh and Shalev

-.082 -.111 .090 -.064

UCLA .008 -.087 .151 .046

Composite -.039 -.106 .130 -.009

Bargh and Shalev Loneliness First (n = 167)

Bargh and Shalev

-.015 -.022 -.108 -.083

UCLA -.061 -.061 -.053 -.101

Composite -.041 -.045 -.088 -.100

No Awareness (n = 291)

Bargh and Shalev

-.049 -.037 -.021 -.064

UCLA

-.039 -.040 .056 -.013

Composite -.042 -.042 .019 -.042 Note: Bargh and Shalev refers to the loneliness measure created by Bargh and Shalev and UCLA refers to the Russell (1996) loneliness measure. * p < .05.


Supplemental Table 2

Intuitions about Water Temperatures and Loneliness

Connection Cold Warm

Yes 11.8% 6.5% 28.5% Maybe 47.7% 31.9% 37.5%

No 40.6% 61.6% 34.1% Note: N = 323 Connection: Do you think there is a connection between loneliness and a preference for water temperatures? Cold: Do you think lonely people prefer cold temperatures for baths/showers? Warm: Do you think lonely people prefer warm temperatures for baths/showers?


Distributions of Showering/Bathing Items

How often do you usually take a bath/shower?

Value Response Attentive All

1 More than 3 times a day 0.0% 0.3%

2 3 times a day 0.3% 1.1%

3 2 times a day 9.6% 10.7%

4 Once a day 68.1% 66.6%

5 Once every other day 18.6% 17.9%

6 2-3 times a week 3.1% 3.2%

7 Once a week 0.3% 0.3%

8 Less than once a week - -

Sample Size 323 374

What temperature do you use for the water when you take a bath/shower?


1 Very hot 7.4% 8.3%

2 Hot 50.8% 47.3%

3 Warm 36.2% 38.8%

4 Lukewarm 5.0% 5.1%

5 Cold 0.6% 0.5%

6 Very Cold 0.0% 0.0%

Sample Size 323 374


About how much time do you spend in the bath/shower?


1 Less than 2 minutes 0.0% 0.3%

2 2-5 minutes 1.9% 1.9%

3 5-10 minutes 23.2% 23.0%

4 10-15 minutes 38.7% 39.6%

5 15-20 minutes 25.1% 23.8%

6 20-30 minutes 8.4% 8.0%

7 Over 30 minutes 2.8% 3.5%

Sample Size 323 374

Documents

Warm Water and Loneliness Redux: Rejoinder to Bargh … to Bargh and Shalev (2014) 2 Abstract Bargh and Shalev (2014) replied to our work and summarized results from three new studies