Exploratory Hand Movements Enhance the Liking Effect in Haptics

Introduction

When people choose one from multiple items, they usually move their eyes to compare the items’ appearances. Eye movements are essential to bring the targets into focus on the retina to observe their detail. However, Shimojo, Simion, Shimojo, and Scheier (2003) show that eye movements are also related to preference formation. In their study, participants looked at two images alternately and selected one they preferred. The images were alternately presented at the left and the right of a screen, requiring participants to move their eyes from side to side to look at them. When one image was presented longer than the other, participants preferred that image more frequently than chance. People have been found to prefer items presented repeatedly, which is known as the mere-exposure effect (Bornstein, 1989; Zajonc, 1968). The bias in choosing an image observed by Shimojo et al. (2003) might be attributed to this effect. However, when the two images were presented alternately at the center of the screen, removing the need for participants to move their eyes to see both images, the bias was not found.

In typical studies of the mere-exposure effect, participants look at multiple images on a screen presented a pre-determined number of times before evaluating their preference for each image. During their exposure to the images, participants usually do not move their eyes. Therefore, Shimojo et al.’s (2003) observation of higher preference for images presented for longer when participants’ eyes moved (compared with no eye movement) seems a different phenomenon from the mere-exposure effect. In this study, we call the tendency to choose an item presented for longer or more frequently in two-alternative forced-choice tasks the liking effect.

Shimojo et al. (2003) also reported participants’ tendency to look at the image they would choose just before deciding in a two-alternative forced-choice task. This tendency has also been observed in many subsequent studies, which have discussed the relationship between eye movements and preference formation considering the liking effect (Glaholt & Reingold, 2009; Krajbich, Armel, & Rangel, 2010; Mitsuda & Glaholt, 2014; Nittono & Wada, 2009; Schotter, Berry, McKenzie, & Rayner, 2010).

This tendency for contact with an item a person will choose just before deciding has also been observed in auditory (Lindsen, Moonga, Shimojo, & Bhattacharya, 2011), odor (Mitsuda, 2015), and haptic preference tasks (Mitsuda & Yoshioka, 2015). Therefore, this tendency must be a general phenomenon regardless of sensory modalities.

Mitsuda and Yoshioka (2018) also recently found the liking effect in a haptic preference task. With participants asked to choose between handkerchiefs in a two-alternative forced-choice task, the liking effect manifested through the difference not in touch duration but in touch frequency: The likelihood of choosing the more frequently touched handkerchief was higher than chance. This finding suggests that the liking effect observed in visual preference tasks is a general phenomenon regardless of sensory modalities. To verify the liking effect in haptic preference, which only Mitsuda and Yoshioka (2018) have previously reported, this study examines the liking effect by the difference in touch frequency between two handkerchiefs, hypothesizing that the effect increases with a greater difference in the number of touches (Hypothesis 1).

In addition, by analogy with eye movement in visual tasks (Shimojo et al., 2003), this study hypothesizes that the liking effect in a haptic task will appear when participants actively move their hand to objects for evaluation, but not when participants do not move their hand (Hypothesis 2). To test this hypothesis, we adopt a two-alternative forced-choice task using handkerchiefs and compare the likelihood of choosing the more frequently touched item when participants moved their hand to the items (as in Mitsuda & Yoshioka, 2018) and when they did not. The hypothesis predicts that the liking effect will appear only in the former condition.

All three experiments in this study were approved by the university ethics review committee and conform with the Declaration of Helsinki; all participants provided written informed consent.

Experiment 1

Experiment 1 examined the liking effect in a two-alternative forced-choice task, comparing conditions in which one handkerchief was touched two and four times more than the other handkerchief. According to Hypothesis 1, the likelihood of choosing the more frequently touched handkerchief is expected to be greater in the latter condition.

Apparatus

Participants seated on a chair touched handkerchiefs placed on a working plate (depth 30 cm, width 75 cm, height 2 cm) fixed to a linear actuator (SCLG6-020-800, Dyadic Systems Co., Ltd., Kanazawa, Japan). In Experiments 2 and 3, the linear actuator moved the working plate from side-to-side to switch the handkerchief, but in Experiment 1 the actuator did not move. A hand rest (depth 30 cm, width 10 cm, height 1 cm) was fixed at the center of the working plate, and two handkerchiefs were placed on both sides of the hand rest. Each participant’s arm was supported by a wheeled mounting (depth 13 cm, width 20 cm, height 12 cm) to decrease fatigue. Participants placed their aim inside a box (depth 30 cm, width 50 cm, height 37 cm) that hid the handkerchiefs from them. The movement of each participant’s hand was monitored by a position sensor (Patriot, Polhemus, Colchester, VT) at 60 Hz. A small receiver for the position sensor (depth 2 cm, width 2.5 cm, height 1.5 cm) was attached to the back of the hand used for touching. Figure 1(a) shows the experimental setup.

OPEN IN VIEWER

Figure 1.

(a) Experimental setup for haptic preference judgment tasks using two handkerchiefs. (b) Hand movements in all arm-move condition trials. (c) Movement of handkerchiefs by linear actuator in all arm-fixed condition trials. Location signposts labeled “L” or “R” at the top, which participants could see above the box, were used only in Experiments 2 and 3. A monitor displaying the handkerchief for touching (i.e., “Left” or “Right”) was placed to the left side of the box in Experiments 2 and 3 (not shown in the figure).

Procedure

Every trial started with the participant placing their dominant hand on the hand rest. After a cue signal, each participant touched the two handkerchiefs in a pre-determined order and reported which they preferred by pressing the “Z” key for the left handkerchief and the “X” key for the right one using their non-dominant hand, after first returning their dominant hand to the hand rest. Participants were instructed to spread out their palm and stroke the handkerchiefs to sense their touch.

In the first condition, participants touched the right handkerchief for 4 seconds, the left handkerchief for 8 seconds, and then the right handkerchief again for 4 seconds. Participants returned their hand to the hand rest after each touch period and waited in that position for 1 second before the next touch period began. In this series of hand movements, participants touched the right handkerchief twice and the left handkerchief once, with a total touch duration of 8 seconds for each handkerchief.

In the second condition, participants touched the handkerchiefs as follows: right for 2 seconds, right again for 2 seconds, left for 8 seconds, right for 2 seconds, and right again for 2 seconds. As in the first condition, they waited for 1 second between each touch period with their hand on the hand rest and returned their hand there after the final touch period. In this series of hand movements, participants touched the right handkerchief four times and the left handkerchief once, with a total touch duration of 8 seconds for each handkerchief.

The cue for each hand movement was a beep sound created by a personal computer. When a participant moved their hand before the beep or did not reach the hand rest or designated handkerchief within 1 second of the beep, an alarm sounded and the participant repeated the same trial. Participants practiced the hand movements using dummy handkerchiefs before commencing recorded trials. The experiment took about 40 minutes in total for each participant.

Experiment 2

In Experiment 2, the liking effect was compared between a condition where participants moved their arm to touch the handkerchiefs and a condition where their arm remained static and the linear actuator carried the handkerchiefs to their hand. Considering the small magnitude of the liking effect in Experiment 1, participants were allowed in Experiment 2 to pinch and touch the handkerchiefs freely, as in Mitsuda and Yoshioka (2018).

Results and Discussion

In the arm-move condition, the likelihood of choosing the more frequently touched handkerchiefs (referred to as the likelihood for readability in the following sentences) was significantly greater than chance, M = .63, t(15) = 5.0, p = .0002, d_z = 1.3. The likelihood was greater than that in Experiment 1’s Condition 2 (M = .57), in which participants stroked the handkerchiefs, although without statistical significance, t(15) = 2.0, p = .065, d_z = 0.50. In the arm-fixed condition, the likelihood was also significantly greater than chance, M = .56, t(15) = 2.6, p = .020, d_z = 0.65, but was significantly smaller than that in the arm-move condition, t(15) = 2.4, p = .028, d_z = 0.61.

Figure 2 shows the likelihoods by handkerchief pair (i.e., two identical vs. two different) and by condition (i.e., arm-move vs. arm-fixed). A 2 × 2 repeated measures ANOVA crossing the handkerchief pair and the condition detected a significant main effect of the condition, F(1, 15) = 6.0, p = .028, η_p² = .28. However, the main effect of handkerchief pairs was not significant, F(1, 15) = 2.4, p = .15, η_p² = .14, which differed from the result of Experiment 1. The interaction between handkerchief pairs and conditions was not significant (F < .01).

OPEN IN VIEWER

Figure 2.

The likelihood of choosing the more frequently touched handkerchief by arm movement condition and by handkerchief pair in Experiment 2.

Supporting Hypothesis 2, the liking effect was greater when participants moved their arm than when they did not. This reinforces that the difference in the liking effect between active and passive sampling, as observed by Shimojo et al. (2003), is a general phenomenon regardless of sensory modalities. However, whereas Shimojo et al. (2003) found no liking effect when participants did not move their eyes, Experiment 2 found a significant liking effect when participants did not move their arm. In the arm-fixed trials, participants nonetheless moved their fingers to touch the handkerchiefs, which might produce a similar effect to that of arm movement.

Experiment 3

In Experiments 1 and 2, participants touched the right handkerchief more frequently in all trials, with the intention of avoiding confusion regarding hand movements. However, a difference in preference between left and right might have affected the results in Experiments 1 and 2, though Mitsuda and Yoshioka (2018) reported a significant liking effect in trials where handkerchief positions were counterbalanced. Experiment 3 examined this issue by alternating the position (left or right) of the handkerchief that participants touched more frequently in every trial.

Results and Discussion

Consistent with the results in Experiment 2, the likelihood of choosing the more frequently touched handkerchiefs was significantly greater than chance in the arm-move condition, M = .65, t(15) = 4.2, p = .0007, d_z = 1.1, and in the arm-fixed condition, M = .60, t(15) = 5.3, p < .0001, d_z = 1.3. The likelihood was greater in the arm-move condition than in the arm-fixed condition, but not significantly, t(15) = 1.8, p = .098, d_z = 0.44.

A 2 × 2 repeated measures ANOVA crossing the handkerchief pair (i.e., two identical vs. two different) and the condition (i.e., arm-move vs. arm-fixed) detected a significant main effect of handkerchief pairs, F(1, 15) = 9.4, p = .008, η_p² = .39, inconsistently with Experiment 2. However, the main effect of conditions was weak, F(1, 15) = 3.6, p = .078, d_z = 0.19, and there was no interaction between handkerchief pairs and conditions (F = 0.59).

The likelihood of choosing the more frequently touched handkerchiefs was significantly greater than chance for two identical handkerchiefs (M = .70 in the arm-move condition; M = .63 in the arm-fixed condition) and two different handkerchiefs (M = .61 in the arm-move condition; M = .57 in the arm-fixed condition), ps < .023. The likelihood was significantly greater for two identical handkerchiefs than for two different handkerchiefs—arm-move condition: t(15) = 2.4, p = .031, d_z = 0.59; arm-fixed condition: t(15) = 2.5, p = .027, d_z = 0.61. When participants touched two identical handkerchiefs, the liking effect was significantly greater in the arm-move condition than in the arm-fixed condition, t(15) = 2.9, p = .011, d_z = 0.73. However, when participants touched two different handkerchiefs, the liking effect did not differ significantly between the conditions, t(15) = 0.89, p = .39, d_z = 0.22.

Figure 3 shows the liking effects by the position of the more frequently touched handkerchiefs (i.e., left vs. right) and by condition (i.e., arm-move vs. arm-fixed), for all the trials (both identical and different handkerchief pairs). A 2 × 2 repeated measures ANOVA crossing the handkerchief position and the condition did not detect any significant main effect of handkerchief position, F(1, 15) = .66, p = .43, η_p² = .042. A weak interaction between handkerchief position and condition was detected, but without statistical significance, F(1, 15) = 3.8, p = .07, η_p² = .20. The liking effect in the arm-move condition was significantly greater when the more frequently touched handkerchiefs were placed on the right than when placed on the left—M = .69 for right placement; M = .61 for left placement; t(15) = 2.4, p = .029,d_z = 0.60. By contrast, no significant effect of the handkerchief position was observed in the arm-fixed condition—M = .61 for right placement; M = .59 for left placement; t(15) = 0.39, p = .70, d_z = 0.10.

OPEN IN VIEWER

Figure 3.

The likelihood of choosing the more frequently touched handkerchief by arm movement condition and by first-touched handkerchief position for all the trials (both identical and different handkerchief pairs) in Experiment 3.

The results of Experiment 3 verify that the bias of choosing the more frequently touched handkerchief exists regardless of handkerchief positions. Consistent with Experiment 2, the liking effect was greater when participants moved their arm than when they did not; however, the difference only appeared when participants touched the right handkerchief more frequently than the other.

General Discussion

This study examined two hypotheses. Hypothesis 1 was that the liking effect observed in visual preference tasks is a general phenomenon regardless of sensory modalities and so also exists in haptic preference tasks. The results of this study’s three experiments support this hypothesis, as participants tended to choose the more frequently touched handkerchiefs. Mitsuda and Yoshioka (2018) previously found bias in a preference judgment task but not in a smoothness judgment task. Therefore, the bias observed in this study must be attributed to preference-level difference.

In Experiment 1, a significant difference in touch frequency was not found to enhance the liking effect because the bias was small, which we attribute to handkerchiefs being touched differently (i.e., stroking) compared with Mitsuda and Yoshioka’s (2018) study (i.e., pinching and touching freely). However, when the latter way of touching handkerchiefs was adopted, the liking effects when participants touched one handkerchief four times and the other once (Experiment 2: M = .63; Experiment 3: M = .65) were greater than that when participants touched one handkerchief twice and the other once (M = .56) in Mitsuda and Yoshioka (2018). Therefore, a greater difference in touch frequency between the two handkerchiefs seems likely to enhance the liking effect, as in the mere-exposure effect (Bornstein, 1989; Zajonc, 1968).

One possible limitation of this study is that the serial order of touching handkerchiefs may have affected preference. Many studies have reported the serial order’s effect on choosing one from two or more items such as beverages, musical pieces, and landscapes (Biswas, Grewal, & Roggeveen, 2010; Carney & Banaji, 2012; Li & Epley, 2009; Mantonakis, Rodero, Lesschaeve, & Hastie, 2009; Pandelaere, Millet, & Van den Bergh, 2010). However, whether coming first or last has the greater effect has varied between different items. In addition, these studies presented each item only once in a series; no paper has examined the effect of serial order in the trial condition of presenting items repeatedly, as in this study. Mitsuda and Yoshioka (2018) found no bias in choosing a handkerchief when participants touched two handkerchiefs either once or twice each, suggesting that touching first or last did not affect preference. However, it is possible that the effects of touching first and last canceled each other because the first handkerchief and last handkerchief differed when participants touched them alternately at the same frequency. This issue should be examined in future studies. In any event, the effect of serial order on preference would not explain that arm movement and more frequent touching enhanced the liking effect in this study.

Hypothesis 2 was that the liking effect appears only when people actively touched objects, as in visual preference tasks. Contrary to its prediction, the results of Experiments 2 and 3 show that the liking effect also appeared when participants did not move their arm, instead touching handkerchiefs carried to their hand by a machine. However, the magnitude of the liking effect when participants moved their arm to touch the handkerchiefs was greater than that when their arm was static. In other words, the liking effect existed regardless of arm movement. At the same time, moving the arm to touch handkerchiefs only enhanced the magnitude of the liking effect when the right handkerchief was touched more frequently.

Shimojo et al. (2003) did not observe a significant liking effect when two images appeared on a screen alternately and participants did not need to move their eyes to see them. By contrast, this study did find the liking effect for the (passive) arm-fixed condition. This contradiction shows that the liking effect observed in haptics may have different source from that in vision. Therefore, further studies should explore the liking effect in visual preference and its relation to that in haptic preference.

This study observed a significant liking effect regardless of whether participants moved their arm to touch the handkerchiefs. Therefore, the mere-exposure effect, which has also previously been observed in haptics (Jakesch & Carbon, 2012), might induce the liking effect. The arm movement could increase attention to the handkerchief and the input of haptic sense to the brain, in turn enhancing the liking effect.

Another possible source of the liking effect observed in this study is the emergence of ownership from touching items (Peck & Shu, 2009). Previous studies have shown that touch increases preference for an object when considering a purchase (Grohmann, Spangenberg, & Sprott, 2007; McCabe, & Nowlis, 2003). In this study, the handkerchiefs were invisible, so participants could only choose by the feel of touch. This condition differs from that in the studies of ownership. However, this study’s participants were made aware before the trials of the object they would be touching, which might have made them consider the handkerchiefs as items for purchase, leading to the emergence of ownership. Accordingly, the greater liking effect in the arm-move condition may be due to arm movements inducing greater ownership.

Enhancement of the liking effect when participants moved their arm and touched the right handkerchief more frequently could be related to reachability. All the study’s participants reported that their right hand was dominant. Consequently, they touched the handkerchiefs in every trial with their right hand. The position of each seated participant relative to the apparatus meant that it was always easier for them to touch handkerchiefs on the right than those on the left. Bar-Hillel (2015) suggested that the item easiest to reach will be favored. Position effects in choosing an item have been extensively studied; although the results have been inconsistent between studies, reachability cannot be discounted as a possible explanation for the effect (see a review by Bar-Hillel, 2015, and the comment by Rodway, Schepman, & Thoma, 2016). Conversely, the unnatural behavior of touching the left handkerchief using the right hand might spoil the consciousness of the active movement (by causing participants to feel forced to touch the handkerchief), leading to a smaller liking effect or smaller ownership.

Heuristic decision-making is another possible cause of the effects observed in this study. People sometimes make decisions based on their metacognitive beliefs, such as the “center-stage heuristic” that important people sit in the middle (Raghubir & Valenzuela, 2006). In this study, similar metacognitive beliefs might have affected participants’ decisions when they had insufficient information to guide their judgment.

One limitation of this study is that the experimental results cannot determine whether the observed choosing bias was due to the liking or the disliking effect. The longer duration of a single touch could have increased aversion to those handkerchiefs. This issue should be examined in future studies.

Finally, this study found that different ways of touching objects affected preference. Experiments using other sensory modalities are required to check if the liking effect is a general phenomenon, as determining the influence of sampling behavior (such as hearing and smelling) on preference is essential for understanding the decision-making process. It would also be valuable to examine the liking effects in haptics with visual information—which may generate a multiplier effect—as people rarely select an item without looking at it in daily activity.