I wanna hold your hand: Handholding is preferred over gentle stroking for emotion regulation

Haran Sened; Chen Levin; Manar Shehab; Naomi Eisenberger; Simone Shamay-Tsoory

doi:10.1371/journal.pone.0284161

. 2023 Apr 6;18(4):e0284161. doi: 10.1371/journal.pone.0284161

I wanna hold your hand: Handholding is preferred over gentle stroking for emotion regulation

Haran Sened ^1,^*, Chen Levin ¹, Manar Shehab ¹, Naomi Eisenberger ², Simone Shamay-Tsoory ¹

Editor: Rochelle Ackerley³

PMCID: PMC10079127 PMID: 37023083

Abstract

Social touch is an important form of interpersonal emotion regulation. In recent years, the emotion regulation effects of two types of touch have been studied extensively: handholding and stroking (specifically of skin with C-tactile afferents on the forearm, i.e. C-touch). While some studies compare their effectiveness, with mixed results, no study to date has examined which type of touch is subjectively preferred. Given the potential bidirectional communication provided by handholding, we hypothesized that to regulate intense emotions, participants would prefer handholding. In four pre-registered online studies (total N = 287), participants rated handholding and stroking, presented in short videos, as emotion regulation methods. Study 1 examined touch reception preference in hypothetical situations. Study 2 replicated Study 1 while also examining touch provision preferences. Study 3 examined touch reception preferences of participants with blood/injection phobia in hypothetical injection situations. Study 4 examined types of touch participants who have recently given birth recalled receiving during childbirth and their hypothetical preferences. In all studies, participants preferred handholding over stroking; participants who have recently given birth reported receiving handholding more than stroking. This was especially evident in Studies 1–3 in emotionally intense situations. These results demonstrate that handholding is preferred over stroking as a form of emotion regulation, especially in intense situations, and support the importance of two-way sensory communication for emotion regulation via touch. We discuss the results and possible additional mechanisms, including top-down processing and cultural priming.

Introduction

Touch is an important form of interaction in humans as well as in other species [1]. Interpersonal touch serves a variety of functions, including emotion communication [2] and modulating interpersonal bonds [3]. One major reason humans touch one another is to help regulate subjective experiences such as negative emotions and pain, sometimes referred to as consoling or comforting touch. Importantly, people can attempt to use different types of touch to regulate each other’s emotions; each type of touch might be performed on different body locations, in different social contexts, and may be associated with different emotional experiences [4].

Fotopoulou and colleagues [5] argue that one of the functions of touch is homeostatic regulation. In homeostatic regulation, touch helps reset biological functioning to a fixed setpoint after an acute disturbance in homeostasis, as in warming up a cold baby. Shamay-Tsoory and Eisenberger [6] suggest that this kind of regulation involves a feedback loop: one person experiences an acute, unwanted experience (e.g. pain, or a negative emotion) while another person senses that and is motivated to provide comforting touch. The first person, in turn, perceives this touch, which serves to diminish emotional intensity. This feedback loop is homeostatic as it includes a corrective mechanism: feedback from a distressed target allows the toucher to adapt the touch accordingly. Thus, for homeostatic regulation to occur, both parties must establish two-way communication. This suggests that forms of touch which allow for better two-way feedback would be preferable over others, when trying to regulate intense emotional experiences and pain.

Two types of touch have been thoroughly examined in recent years using traditional paradigms testing the effectiveness of touch in distress regulation: handholding and gentle stroking.

Handholding entails grasping the other person’s hand in varying degrees of strength. Studies have found that handholding can reduce pain [7], reduce pre-surgical anxiety [8], and lessen the emotional pain of recalling negative emotional experiences [9]. Handholding has been shown to attenuate pain-related activation in the posterior insula, the anterior cingulate cortex, the orbitofrontal cortex and the prefrontal cortex [10].

Gentle stroking entails slowly stroking skin regions with C-tactile afferents [11], such as hairy skin on the forearm. Although slow stroking activates all types of mechanoreceptive afferents, it is also referred to as C-touch (as C fibers respond optimally to this type of touch). At normal skin temperature (but not at very high or low temperatures), CT afferent firing is correlated with touch pleasantness [12]. Other studies have found that this kind of stroking, activates regions in the posterior insular cortex and the mid-anterior orbitofrontal cortex more than other types of touch [13,14]. Moreover, stroking activates social reward mechanisms [15], is associated with reduced feelings of social exclusion [16], although some of its effects may be attenuated by attachment anxiety[17,18]. It has also been suggested that stroking may be associated with elevated oxytocin levels [19].

Regarding terminology, “Stroking” can mean many things–stroking can be performed in many different parts of the body at various speeds and forms. As the current study aims to match effectiveness studies, unless explicitly stated otherwise, we use “stroking” to mean using one’s hand to stroke another person’s forearm at a speed of 1–10 cm/s, the optimal speed found in most studies. As for “C-touch”, recent studies have found some C-tactile afferents in the palm skin as well, albeit at a much lower density [20]; other studies show that these afferents can be activated by deep pressure and not only by stroking, which might mean that they are activated by handholding as well [21]. As such, we avoid using the term C-touch for stroking in the rest of the manuscript as both types of touch might be activating C-tactile afferents.

The studies cited above have demonstrated many cases in which both handholding and stroking are associated with distress regulation (creating a sense of pleasantness, and reducing pain, anxiety and feelings of social exclusion). However, as detailed above, homeostatic regulation is driven by two-way feedback. While any form of touch allows for two-way sensing, the tactile sensitivity of the palms and fingers is larger by orders of magnitude than that of the forearms [22,23]. In handholding, as opposed to stroking, both participants’ palms and fingers are involved, allowing both of them to sense each other with optimal sensitivity. This two-way tactile communication should help close the feedback loop required for homeostatic regulation by providing bi-directional feedback, as supported by findings showing that handholding promotes synchrony in skin conductance response (SCR) and brain activity [7,24]. Thus, handholding may be more suitable for homeostatic regulation (i.e., regulation of intense, short-term experiences) than gentle stroking.

The importance of handholding as a bidirectional or reciprocal form of touch can also be viewed in the broader context of reciprocity in nonverbal interaction [25]. In general, people are motivated to reciprocate others’ actions if those actions create pleasant arousal [26]. For example, one study found that couples are more likely to reciprocate each others’ touch the more established their relationship is [27]. The use of handholding in emotion regulation is somewhat distinct from classic reciprocity as defined in the literature as the touching person is not reciprocating the touched person’s action (i.e., they are not initiating handholding in response to touch by the other person); instead, they are responding to the other person’s emotion in a manner which, in the case of handholding, has a reciprocal nature. As reciprocity in considered rewarding [28] there could be a general positive reaction to reciprocity; for example, in one study [29] it was found that people who performed reciprocal forms of touch were more highly regarded than people who perform non-reciprocal forms of touch.

The existing studies in the field detailed above have demonstrated the effects of handholding and stroking separately. However, studies comparing the effectiveness of one of these types of touch with the other are scarce, with studies comparing two types of touch often comparing stroking of the forearm to stroking or touching of the hand, but not to handholding. Studies that compared general pleasantness of touch between touching forearm and palm skin found little difference [30]. In one of the only studies directly comparing the emotion regulation properties of stroking of the forearm and handholding, Reddan and colleagues [24] compared participants who received painful thermal stimulation to their legs, while their romantic partners supported them via gentle stroking of the forearm, handholding or no touch. When measuring pain subjectively and through skin conductance response, no significant differences were found between handholding and gentle stroking, although effects of handholding were greater in every case. In sum, studies on the effectiveness of touch on distress regulation has shown that handholding and gentle stroking are effective forms of emotion regulation, with the few studies comparing the two finding no major differences.

While effectiveness studies attempt to explore how people experience various types of touch, they do not examine which touch people actually prefer or choose to provide, when they encounter the need to regulate emotions in real-life. Touch choice may be based on representations of touch effectiveness, but it might also have other considerations. For example, people could be choosing types of touch they have seen others perform or which were depicted culturally (i.e., in various forms of media), or types of touch which require less physical effort. Thus, effectiveness studies are not enough to understand which types of touch people choose to use. The current study aims to address this gap in the literature by examining subjective preference—presenting people with various situations and asking them which types of touch they would prefer to receive or to provide.

As detailed above, the feedback-loop and homeostatic regulation models suggests that handholding might be preferred over stroking. While we could find no studies which directly examined preference, some studies provided indirect evidence. One study examined adolescents undergoing cancer treatment and found that adolescents see handholding as an extremely effective coping method [31]. Another study which developed a scale concerning support in face of pain found that an item asking about handholding was indicative of general partner support in painful situations [32]. Notably, neither study explicitly presented stroking as a valid coping method (We could not find the initial list of items for the scale development paper by Krahé and colleagues [32]–it could be the case that one of the initial items which were discarded due to low factor loadings involved stroking). Still, this indirect evidence, alongside the theoretical model, led us to hypothesize that handholding would be preferred over gentle stroking.

Beyond exploring our main hypothesis regarding preference, we examined several possible moderators. First, we examined whether the hypothesized preference for handholding would be stronger in more intense situations. Based on the homeostatic regulation model it was reasoned that if two-way sensory communication is indeed crucial for regulating acute disturbances in homeostasis, handholding may be more effective, and therefore preferred, as a way to regulate emotions in intense situations. Thus, we hypothesized that the preference for handholding would be stronger in intense situations.

Second, we examined in a more exploratory manner whether situation valence (i.e., whether the experience is positive or negative) and physicality (i.e., whether the experiences involved physical as opposed to purely emotional pain or pleasure) would moderate the effect. Effectiveness studies have examined the effects of touch in both positive and negative contexts (e.g., pleasantness [13] versus pain [17]), and when regulating both emotional [9] and physical [33] pain, but to the best of our knowledge have not compared these contexts to one another. As such, we did not have a specific directional hypothesis concerning valence and physicality.

Third, Study 3 aimed to examine whether including a situation that was specifically relevant to participants–namely, an injection for participants with blood/injection phobia–would induce a different pattern of results. We examined whether the level of phobia would moderate results. Finally, as we could not find effectiveness studies which performed cultural comparisons, we sought to examine whether touch preference would be different in different cultures. To do so, in Study 4 we compared the touch preferences of Arab and Jewish women. Importantly, these last two moderators–blood/injection phobia and culture–build on less established literature. As such, they were only included in one study each (Studies 3 and 4, respectively), and our investigations of them are intended to inform future research, rather than to reach solid final conclusions.

As detailed above, the aim of the current study was to connect effectiveness research with subjective preferences. While many types of stroking exist, stroking researchers have focused on gentle stroking of the forearm by the hand at a velocity of 1–10 cm/s as the optimal stroking speed; in fact, stroking at higher speeds is often used as a control condition (e.g., in a study by von Mohr and colleagues [16]). As our hypothesis was that handholding would be preferred over stroking, we sought to present stroking in the optimal way possible so that our design tests the hypothesis effectively. Therefore, in Studies 1–3 and in the second part of Study 4 we used looping videos instead of verbal labels for representing different types of touch. In the first part of Study 4, we asked participants which types of touch were provided to them during childbirth and provided standard verbal labels–“stroking” and “handholding”. As both optimal and non-optimal forms of stroking would count towards “stroking” in this paradigm, we could be confident that the prevalence of optimal stroking would, if anything, be overestimated.

Notably, the use of videos somewhat resembles paradigms involving vicarious touch–the study of the effects of seeing people touching one another [34]. However, the current study does not ask people about their feelings while watching the videos, or their thoughts about the videos or about the people depicted in them, which are the questions explored by vicarious touch research. Instead, the current study asks participants about hypothetical or recalled scenarios and only uses the videos to describe the types of touch. While some vicarious effects might be triggered by watching the videos, the same is true for actual touch in real life: when touching or being touched by another person, we usually also see the act of touching. Additionally, to the best of our knowledge, no study of vicarious touch has compared observing handholding to observing stroking.

Overview

In four pre-registered studies, we examined whether handholding would be subjectively preferred over gentle stroking, and whether this preference would be stronger in intense emotional situations. In each study participants rated how helpful they would find stroking or handholding in various hypothesized or recalled situations. We also examined whether participants’ choice was moderated by situation intensity, touch reception versus provision, situation valence, emotional versus physical situations, and cultural differences, as detailed above.

Study 1 examined the type of touch people would prefer to receive from a romantic partner in hypothetical situations by comparing ratings as well as discrete choices of each type of touch. Study 2 replicated most of the results of Study 1 while correcting some methodological issues as well as examining the types of touch participants would prefer to provide.

Study 3 examined touch preferences in the context of a more severe stressor. Participants with some level of blood/injection phobia [35] were asked to imagine themselves receiving an injection.

Finally, in addition to hypothesized situations, Study 4 examined recalled situations and behavior. We asked women who recently gave birth to recall types of touch provided by a close person who was present during their labor, which is usually accompanied by intense pain [36], and to rate which type of touch they preferred. Study 4 also examined cultural differences between Arabic-speaking and Hebrew-speaking participants.

Hypotheses and pre-registration

The hypotheses, study design and sample analysis code for all studies were pre-registered prior to analysis (and in the case of Study 1, prior to data collection), and final study data, methods and code were posted on the Open Science Framework (see below for links). The current study examined the following hypotheses:

Subjective rating (All studies): Participants will rate handholding as more helpful than stroking, especially in intense situations.
Subjective choice (Study 1 only): When instructed to choose one type of touch over the other, participants will prefer handholding over stroking, especially in intense situations.
Recollections of touch received during childbirth (Study 4 only): During childbirth, participants will have received handholding more often than stroking.

The main effect (handholding preferred over stroking overall) was pre-registered only in Studies 3 and 4. The interaction effect (handholding preferred more over stroking as intensity increases) was pre-registered in all studies. We also exploratorily examined the effects of valence, physicality, and cultural differences with no specific directional hypothesis, and exploratorily examined the effects of blood/injection phobia expecting blood/injection phobia to be associated with a stronger preference for handholding, especially in intense situations. Results regarding pre-registered hypotheses concerning loneliness in Studies 1 and 2 will be reported as part of a separate project alongside additional studies.

General method

Ethics statement

All procedures were done in accordance with the principles expressed in the declaration of Helsinki for the treatment of human participants. Procedures were approved by the University of Haifa Department of Psychology IRB, approval 048–21. All participants were over 18 years old. Participants’ written consent was obtained and recorded by completing a digital consent form.

Open data and preregistration

We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study. All four studies were preregistered on the Open Science Framework:

Study 1: https://osf.io/x95f4/?view_only=e272eb9ccc2a46668114685dcb8988e5

Study 2: https://osf.io/nzg6v/?view_only=b9cf1e802d894ce387c80c47ffd7bc97

Study 3: https://osf.io/8n3pu/?view_only=f23a2e3d485946ffb95d32c02e469b85

Study 4: https://osf.io/74szg/?view_only=4b12d6d2a8a04483abbd6ea0c8c225ca.

Study data, analysis code and touch videos for Studies 1–3 are available at the link below, except for touch videos for Study 4 which were not posted as participants who filmed the videos did not consent to wide distribution.

https://osf.io/nju48/?view_only=81dadc972f6545f19d107c83d8dba368

Power analyses

Power analyses were conducted by simulation, using the R package paramtest [37]. Data were simulated with the planned number of participants and study designs, and the theoretical effect sizes detailed below. According to analyses conducted before the project began, a sample size of 100 for Studies 1 and 2 would suffice to detect a main effect with a standardized beta of .25 with very high power (Power > .999) and would have adequate power to detect an interaction effect of .0625 (Power = .812). Studies 3 and 4 were planned with a sample size of 60, with similar power (Power > .999) to detect a .25 main effect but only enough power to detect an interaction effect of .125 (Power = .86 and .904, respectively). Unfortunately, these populations proved to be difficult to recruit and sample sizes were lower than planned.

Study 1

Method

Participants

We recruited a sample of Amazon Mechanical Turk (MTurk) workers from the United States and the UK. All participants had a masters qualification and declared that they were over 18 years old, and in romantic relationships that lasted six months or more. Participants who completed the study received $1.50 in compensation. One participant did not correctly answer an attention check and was removed. Two additional participants were erroneously removed from the dataset before pre-registration as they made a technical error while submitting their questionnaires (We re-ran all analyses including these participants and all results remain the same. Results for these analyses are provided in the Robustness Tests section in the S1 Appendix).

Of 99 participants, 45 identified as female and 54 identified as male. Mean relationship length was 11.23 years (SD = 10.94). Age data was not collected in this study.

Procedure

All studies were conducted online via the Qualtrics survey platform. Participants began by providing informed consent and indicating the length of their romantic relationship. In the first part of the study, participants were asked to imagine themselves in 8 hypothetical situations in which their romantic partner touched them in order to make them feel better. For each situation, participants rated its intensity and positivity/negativity. They were then asked to select one type of touch which they would prefer to receive. The available types of touch were presence (i.e., no touch), handholding, and optimal stroking. To ensure that the participants understood which type of touch is discussed, the types of touch were not labelled with words (e.g., “handholding”), but instead with looping 5-second videos showing the type of touch (see Fig 1, and details below).

Fig 1 — The types of touch depicted are, from left to right, stroking, handholding, and no touch. The types of touch were depicted with videos, so the figure shows a still image from each video. Touch type order was randomized.

In the second part of the study, participants saw 8 different situations, and similarly rated intensity and positivity/negativity. However, this time they were asked to provide a rating for each type of touch. Again, the types of touch were labelled with looping videos. As detailed below, each participant saw the same 16 situations which were assigned randomly to one of the two parts of the study. The touch labels were presented in random order.

After these two parts, they completed a loneliness questionnaire for a separate project.

Measures

Situations. A list of hypothetical situations was generated using a pilot study. We showed 21 participants a list of 33 emotional or physical situations generated by the researchers and asked them to rate the valence and intensity of each situation. For each context (emotional or physical), we selected the two situations rated the most positive and intense as the positive intense situations for that context. We similarly selected three additional pairs of situations (negative intense, positive not intense, negative not intense). This process yielded 16 situations classified as intense/not intense (intensity), positive/negative (valence), and physical/emotional (context). Two situations were available for each of the eight possible configurations of these variables. In the actual study, two positive situations originally designated as intense (one physical and one emotional) were each rated less intense than one of the corresponding non-intense positive situations. Thus, the designations of these two situations were switched.

In each part of the study, participants saw 8 situations. Situation order was randomized such that in each part of the study participants saw one situation of each designation (e.g., one physical, non-intense, negative situation; one physical, non-intense, positive situation, etc.).

Situation Intensity and Valence Ratings. Each participant rated the intensity of each situation by responding to the prompt “How intense does this situation feel?” on a 4-point Likert scale ranging from “not intense at all” to “very intense”. They also rated the valence of each situation by responding to the prompt “How positive or negative does this situation feel?” on a 7-point Likert scale ranging from “very negative” to “very positive”.

Touch Preference. In the first part of the study, touch preference was measured using a direct choice. Participants were asked “Which kind of touch would you prefer to receive from your partner in that situation?” and selected one of the types of touch on the screen. Types of touch were presented using looping videos (Fig 1).

In the second part of the study, touch preference was measured more granularly by comparing participants’ ratings of each type of touch. Participants read the following instructions: “Please rate how would that kind of touch affect your feeling?” and responded on a 5-point Likert scale ranging from “would make me feel much worse” to “would make me feel much better”. Types of touch were presented using looping videos (Fig 2– note that this figure includes an additional type of touch which was only added in Study 2).

Fig 2 — The types of touch depicted are, from top to bottom, no touch, handholding, full palm stroking, and standard stroking. The types of touch were depicted with videos, so the figure shows a still image from each video. Note that full palm stroking was only included in Studies 2 and 3.

Touch Label Videos. The touch literature specifies exact parameters for optimal stroking–stroking of the arm using the index and middle fingers, at speeds of 1–10 cm/s [11,38,39]. As describing this optimal procedure could confuse participants, videos were used instead of verbal labels to depict the different types of touch (Fig 1). The videos showed only the actors’ hands against a flat white background. The videos depicted presence (i.e., two hands not touching each other), stroking, and handholding. In the video depicting handholding the actors were instructed to hold their hands statically. In the stroking videos, the actors were asked to stroke their partner’s arm covering an area of nine cm over a period of three seconds. The videos were then evaluated by judges to confirm that they look natural. Follow-up inspection of videos revealed a range of speeds between 4–12 cm/s.

Videos were recorded by two heterosexual couples who had been in their romantic relationship for over a year. Each couple recorded one video for each type of touch. Videos by one couple were used in the first part of the study, and videos by the other couple were used in the second part, chosen randomly for each participant.

Statistical analysis

All analyses were conducted using R [40]. Analyses were conducted twice, with intensity measured as a rating–the intensity of the situation as self-reported by participants–or as a dichotomous variable–the pre-assigned intensity of the situations. For brevity, we report here only the findings for intensity rating; full results for dichotomous intensity are reported in the S2 Table in S1 Appendix. We state whenever there was a difference in significance between results using dichotomous intensity or intensity rating.

The direct touch type choices performed In the first part of the study were analyzed in a mixed logistical regression using the R package lme4 [41], using the following equation:

{C h o i c e}_{i j} \sim b_{0} + b_{1 i} + b_{2} * {I n t e n s i t y}_{i j}

The choice of participant i in situation j–stroking or handholding was estimated using a fixed intercept (b₀), a random intercept for each participant (b_1i), and an intensity effect (b₂). Occasions on which participants chose presence over the other types of touch were not analyzed.

The touch ratings collected in the second part of the study were analyzed in a mixed linear model using the R package nlme [42], using the following equation:

{R a t i n g}_{i j} \sim b_{0} + b_{1 i} + b_{2} * {I n t e n s i t y}_{i j} + b_{3} * {T o u c h t y p e}_{i j} + b_{4} * {I n t e n s i t y}_{i j} * {T o u c h t y p e}_{i j}

The rating of participant i in situation j was estimated using a fixed intercept (b₀), a random intercept for each participant (b_1i), an intensity effect (b₂), a touch type effect (b₃), and an interaction effect between intensity and touch type (b₄). Touch type was coded 0.5 for handholding and -0.5 for stroking, so that coefficients for other effects would reflect average effects across the two touch types. In exploratory analyses with additional independent variables, fixed slopes were included for each main effect and for every possible interaction. All dependent variables were person mean-centered.

All analyses are accompanied by partial f² effect sizes, calculated using the procedures outlined by Selya and colleagues [43]. To calculate f² for a specific predictor, R² was calculated for the complete model (R²_Total), and for the model without the predictor (R²_Omitted); f² was calculated as (R²_Total—R²_Omitted) / (1–- R²_Total).

To ensure that the effects are not due to specific statistical choices, we have re-run analyses for the main hypotheses using different statistical methods. These include modeling random slopes for all variables, using cumulative link models [44] implemented in the R package ordinal [45] (also with random slopes for all variables), and using simple repeated ANOVA tests to test effects with dichotomous intensity ratings. All of the results stayed the same–no significant effect became non-significant and vice versa. Results for the cumulative link models are provided in Tables 1 and 2, results for the other analyses are provided in the Robustness Tests section in the S1 Appendix. Note that the fully saturated cumulative link model did not converge for the second part of Study 2 (touch provision) when using participant intensity ratings (the model using dichotomous situation classifications converged successfully). Thus we ran that model without a random slope for touch type (but with random slopes for intensity and the interaction between intensity and touch type); this was the only way to have a converging model while removing only one random slope. We performed additional robustness tests on this specific study which are detailed in the S1 Appendix.

Table 1. The effect of intensity measured by intensity ratings and touch type on touch preference analyzed using cumulative link models.

		b(SE)	z	p
Study 1	Intensity	.404(.086)	4.69	< .001***
	Touch Type	.698(.141)	4.97	< .001***
	Intensity*Touch Type	.402(.133)	3.015	.003**
Study 2 touch provision	Intensity	0.439(.096)	4.556	< .001***
	Touch Type	1.267(.205)	6.169	< .001***
	Intensity*Touch Type	0.51(.128)	3.981	< .001***
Study 2 touch reception	Intensity	0.286(.088)	3.262	.001**
	Touch Type	1.484(.184)	8.053	< .001***
	Intensity*Touch Type	0.495(.12)	4.125	< .001***
Study 3	Intensity	1.058(.259)	4.092	< .001***
	Touch Type	3.885(.627)	6.191	< .001***
	Intensity*Touch Type	1.406(.339)	4.146	< .001***
Study 1 with omitted participants	Intensity	.393(.084)	4.705	< .001***
	Touch Type	.651(.14)	4.64	< .001***
	Intensity*Touch Type	.426(.131)	3.258	.001**

Open in a new tab

Table 2. The effect of intensity measured by dichotomous situation classifications and touch type on touch preference analyzed using cumulative link models.

		b(SE)	z	p
Study 1	Intensity	.109(.115)	0.949	.343
	Touch Type	.694(.136)	5.098	< .001***
	Intensity*Touch Type	.817(.208)	3.933	< .001***
Study 2 touch provision	Intensity	0.072(.112)	0.638	.523
	Touch Type	1.221(.193)	6.34	< .001***
	Intensity*Touch Type	0.624(.186)	3.364	.001***
Study 2 touch reception	Intensity	0.112(.122)	0.92	.358
	Touch Type	1.424(.179)	7.939	< .001***
	Intensity*Touch Type	0.548(.187)	2.927	.003**
Study 3	Intensity	2.207(.547)	4.033	< .001***
	Touch Type	3.847(.685)	5.617	< .001***
	Intensity*Touch Type	2.977(.779)	3.822	< .001***
Study 1 with omitted participants	Intensity	.093(.113)	0.824	.410
	Touch Type	.649(.137)	4.74	< .001***
	Intensity*Touch Type	.822(.204)	4.025	< .001***

Open in a new tab

Results

Descriptives

Descriptive values for situation intensity, situation valence, and touch type rating are presented in Table 3 and in Fig 3; A figure depicting all data points is included in the S2 Fig in S1 Appendix. As for choice questions, out of 395 situations classified as intense, handholding was chosen 216 times (54.7%), stroking 87 times (22%), and no touch 92 times (23.3%). Out of 389 situations classified as non-intense, handholding was chosen 133 times (34.2%), stroking 164 times (42.2%) and no touch 92 times (23.7%). These frequencies are presented in Fig 4. The slight difference in the number of situations (389 vs 395) is due to us switching the classification of two of the situations post randomization, as detailed above. The significance of none of the results regarding direct choice changed due to the switch.

Table 3. Descriptive statistics.

		All Situations		Intense Situations		Non-Intense Situations
		Person-level Mean(SD)	Within-Person SD	Person-level Mean(SD)	Within-Person SD	Person-level Mean(SD)	Within-Person SD
Study 1	Intensity	2.4(0.5)	0.77	2.86(0.56)	0.53	1.99(0.58)	0.6
	Valence	4.11(0.34)	2.12	3.93(0.81)	2.24	4.03(0.7)	1.79
	Touch rating–- Handhold	3.87(0.59)	0.64	3.97(0.66)	0.59	3.8(0.64)	0.57
	Touch rating–- Presence	2.79(0.44)	0.43	2.72(0.5)	0.42	2.81(0.52)	0.36
	Touch rating–- Stroke	3.65(0.63)	0.63	3.6(0.66)	0.65	3.71(0.7)	0.54
Study 2 touch reception	Intensity	2.51(0.49)	0.81	2.93(0.54)	0.58	2.09(0.61)	0.62
	Valence	4.18(0.53)	2.05	4.12(0.6)	2.22	4.25(0.62)	1.82
	Touch rating–Full palm stroke	3.67(0.83)	0.64	3.67(0.86)	0.61	3.68(0.89)	0.55
	Touch rating–- Handhold	4.13(0.58)	0.6	4.18(0.61)	0.58	4.09(0.63)	0.51
	Touch rating–- Presence	2.8(0.57)	0.48	2.75(0.61)	0.5	2.86(0.6)	0.39
	Touch rating–- Stroke	3.6(0.79)	0.65	3.57(0.82)	0.62	3.64(0.86)	0.58
Study 2 touch provision	Intensity	2.38(0.54)	0.78	2.77(0.61)	0.57	2(0.63)	0.6
	Valence	4.17(0.49)	1.94	4.11(0.53)	2.14	4.23(0.56)	1.68
	Touch rating–Full palm stroke	3.81(0.89)	0.65	3.76(0.95)	0.64	3.86(0.94)	0.57
	Touch rating–- Handhold	4.25(0.8)	0.74	4.32(0.82)	0.74	4.18(0.88)	0.61
	Touch rating–- Presence	2.79(0.61)	0.5	2.76(0.68)	0.5	2.82(0.6)	0.41
	Touch rating–- Stroke	3.73(0.86)	0.64	3.67(0.9)	0.69	3.78(0.9)	0.52
Study 3	Intensity	2.58(0.48)	.85	3.07(0.59)	.3	1.59(0.73)	²
	Valence	2.82(0.79)	.98	2.41(1)	.52	3.63(1.2)	²
	MBIPI¹	66.94(31.06)
	Touch rating–Full palm stroke	3.14(1.05)	.65	3.25(1.24)	.36	2.9(1.08)	²
	Touch rating–- Handhold	4.14(0.82)	.59	4.38(0.9)	.18	3.65(1.09)	²
	Touch rating–- Presence	2.84(0.83)	.38	2.83(0.95)	.18	2.86(0.87)	²
	Touch rating–- Stroke	3(1.1)	.6	3.06(1.29)	.36	2.88(1.07)	²
Study 4	Intensity	0.01(0.63)	0.69	0.4(0.68)	0.37	-0.46(0.82)	0.26
	Touch rating–- Handhold	62.84(28.83)	17.09	64.14(28.41)	17.88	64.43(31.2)	12.91
	Touch rating–- Presence	22.86(26.22)	8.4	20.46(26.1)	5.88	26.1(31.09)	7.07
	Touch rating–- Stroke	38.2(35.16)	14.2	37.12(34.85)	15.58	44.14(36.97)	11.16

Open in a new tab

¹Person-level variables do not differ between situations and as such have no within-person SD or situation-specific values.

²Study 3 had only one non-intense situation.

Fig 3 — The figure shows how many times each type of touch was chosen as the preferred type of touch for a situation.

Fig 4 — The figure plots the distribution of subject mean ratings for each type of touch for each situation (e.g., the mean of one specific subject’s ratings of stroking for 4 intense situations would be one data point). Touch ratings refer to the way participants thought their feelings would change if they received this type of touch.

Direct choice

We performed a Chi-square test for goodness of fit to compare the overall number of times each type of touch was chosen (349 times handholding, 251 times stroking) to a chance (equal) distribution. The test confirmed that the distribution was significantly different from chance, meaning that handholding was significantly preferred over stroking overall (χ2(1) = 16.007, p < .001). We then performed a chi-square test for independence between the type of touch chosen and situation intensity. Touch type chosen and intensity were significantly associated with one another beyond chance (i.e., not statistically independent; χ2(2) = 43.317, p < .001), meaning that handholding was preferred in intense situations. We tested simple effects by comparing the distribution of touch type choices to chance, separately for intense and non-intense situations. Tests for simple effects found that in intense situations handholding was chosen more than stroking beyond chance (χ2(1) = 54.921, p < .001), whereas the difference between stroking and handholding in non-intense situations was not significantly different from chance (χ2(1) = 3.236, p = .072).

To examine whether participants would usually choose handholding over stroking, especially in situations which they rated as more intense (as opposed to the Chi-square tests which compared preferences between situations which were pre-classified as intense or not intense), we conducted a logistic mixed linear regression using the R package lme4 [41] to examine the effect of intensity on touch choice, as detailed above (situations in which “no touch” was chosen were removed). The results confirmed our hypothesis that handholding would be chosen significantly more often stroking (i.e., the intercept was positive and significant; b = .329(SE = .096), z = 3.427, p < .001) and that this effect would be stronger in more intense situations (i.e., an intensity effect, b = .593(SE = .101), z = 5.847, p < .001). As a robustness check, we also ran the model with saturated random slopes. Again, the results confirmed our hypothesis that handholding would be chosen significantly more often stroking (i.e., the intercept was positive and significant; b = .302(SE = .105), z = 2.861, p = .004) and that this effect would be stronger in more intense situations (i.e., an intensity effect, b = .77(SE = .16), z = 4.809, p < .001). This pattern of results also held when including the two omitted participants.

Rating

To examine whether participants would rate handholding as better at regulating their emotions than stroking, especially in situations which they rated as more intense, we conducted a mixed linear regression analysis as detailed above. Handholding was rated significantly higher than stroking, and the interaction between intensity and touch type was significant such that the difference between handholding and stroking was even larger in more intense situations (Full numerical results are provided in Table 4; results are demonstrated in Fig 5).

Table 4. The effect of situation intensity as measured by participant ratings and of touch type on touch rating.

		b(SE)	95% CI	t(df)	p	f²
Study 1	Intercept	3.759(.057)	3.65,3.87	66.231(1482)	< .001***	0
	Intensity	0.117(.022)	0.07,0.16	5.2(1482)	< .001***	.011
	Touch Type	0.226(.035)	0.16,0.30	6.375(1482)	< .001***	.016
	Intensity*Touch Type	0.127(.045)	0.04,0.22	2.829(1482)	.005**	.003
Study 2 touch provision	Intercept	3.886(.066)	3.76,4.02	58.613(2320)	< .001***	0
	Intensity	0.105(.02)	0.07,0.14	5.359(2320)	< .001***	.007
	Touch Type	0.389(.031)	0.33,0.45	12.401(2320)	< .001***	.036
	Intensity*Touch Type	0.12(.039)	0.04,0.20	3.047(2320)	.002**	.002
Study 2 touch reception	Intercept	3.885(.067)	3.75,4.02	58.129(2320)	< .001***	0
	Intensity	0.062(.02)	0.02,0.10	3.104(2320)	.002**	.002
	Touch Type	0.495(.033)	0.43,0.56	15.101(2320)	< .001***	.055
	Intensity*Touch Type	0.113(.04)	0.04,0.19	2.859(2320)	.004**	.002
Study 3	Intercept	3.603(.115)	3.38,3.83	31.34(405)	< .001***	0
	Intensity	0.193(.052)	0.09,0.30	3.74(405)	< .001***	.019
	Touch Type	1.069(.092)	0.89,1.25	11.636(405)	< .001***	.18
	Intensity*Touch Type	0.249(.103)	0.05,0.45	2.404(405)	.017*	.008
Study 4	Intercept	51.108(4.677)	41.88,60.34	10.926(185)	< .001***	.001
	Intensity	3.155(2.378)	-1.54,7.85	1.327(185)	.186	0
	Touch Type	22.627(3.456)	15.81,29.45	6.546(185)	< .001***	.1
	Intensity*Touch Type	0.729(4.686)	-8.52,9.97	0.156(185)	.877	0

Open in a new tab

Fig 5 — The figure shows the regression results for the association between the ratings for each type of touch and situation intensity.

We performed simple slope analyses via contrasts using the simple_slopes function of the R package reghelper [46]. Handholding was rated significantly higher than stroking at both high and low intensity levels, i.e., one standard deviation above and below mean intensity (when using dichotomous intensity classification, for intensity one standard deviation below the mean, the difference was in the same direction but not significant, p = .059). Ratings for handholding significantly increased with situation intensity. Ratings for stroking did not significantly change with intensity (when using dichotomous intensity classification, ratings for stroking significantly decreased with intensity). Full simple slope results are reported in Table 5.

Table 5. Simple slope analyses for the interaction between intensity measured as a self-reported rating and touch type in Studies 1–3.

		b(SE)	t(df)	p
Study 1	Higher ratings for handholding (vs. stroking) at low intensity (-1 SD)	0.13(0.05)	2.51(1482)	.012*
	Higher ratings for handholding (vs. stroking) at high intensity (+1 SD)	0.33(0.05)	6.51(1482)	< .001***
	Higher ratings for handholding as intensity increases	0.18(0.03)	5.68(1482)	< .001***
	Higher ratings for stroking as intensity increases	0.05(0.03)	1.68(1482)	.094^†
Study 2 touch provision	Higher ratings for handholding (vs. stroking) at low intensity (-1 SD)	0.29(0.04)	6.61(2320)	< .001***
	Higher ratings for handholding (vs. stroking) at high intensity (+1 SD)	0.48(0.04)	10.92(2320)	< .001***
	Higher ratings for handholding as intensity increases	0.17(0.03)	5.15(2320)	< .001***
	Higher ratings for stroking as intensity increases	0.05(0.02)	2.00(2320)	.045*
Study 2 touch reception	Higher ratings for handholding (vs. stroking) at low intensity (-1 SD)	0.40(0.05)	8.65(2320)	< .001***
	Higher ratings for handholding (vs. stroking) at high intensity (+1 SD)	0.59(0.05)	12.70(2320)	< .001***
	Higher ratings for handholding as intensity increases	0.12(0.03)	3.65(2320)	< .001***
	Higher ratings for stroking as intensity increases	0.00(0.02)	0.21(2320)	.832
Study 3	Higher ratings for handholding (vs. stroking) at low intensity (-1 SD)	0.85(0.13)	6.52(405)	< .001***
	Higher ratings for handholding (vs. stroking) at high intensity (+1 SD)	1.29(0.13)	9.92(405)	< .001***
	Higher ratings for handholding as intensity increases	0.32(0.08)	3.76(405)	< .001***
	Higher ratings for stroking as intensity increases	0.07(0.06)	1.16(405)	.248

Open in a new tab

Exploratory analyses

We performed exploratory analyses on moderation of these effects by valence and physicality in the second part of the study. Participants rated all types of touch higher in negative situations than in positive situations, especially when those situations were physical. The difference between participants’ ratings of handholding as opposed to stroking was larger in positive than in negative situations and larger in emotional as opposed to physical situations. Full analysis tables are provided in the S1 Appendix.

Study 2

Study 2 aimed to replicate the results of Study 1, while also examining the types of touch participants thought would be better to provide to their partners, and controlling for the amount of skin touching by including a full palm stroking touch type.