Because of the broad variety of social and cultural interpretations of fashion and clothing, it was decided to examine the firsthand accounts of the subjects from the investigator interviews. Specifically the focus was to gain insight into people’s perceptions of the opening and closing actions as performed by Zipperbot. The video data of the subjects’ responses from the investigator’s interviews were reviewed, segments were transcribed and grammar and language analyzed for repeating themes and ideas.
Based on the analysis of the interviews, precise interpretations of Zipperbot varied and covered a broad range but responses generally fell into binary states, representing positive and negative perceptions. The following sections will develop the various themes that emerged from the participants’ post conversation interviews.
Positives and Negatives in a Broad Range of Perceptions
Based on our findings from this study one significant emerging theme pertained to positive and negative perceptions by the participants. The Zipperbot unzipping/open-collar was perceived as a more pleasant or positive state, while the Zipperbot zipping up/closed-collar represented a more unpleasant or negative state. We can directly observe the use of positive and negative language in the post conversation study interviews. The following quotes are taken from the transcripts of the post conversation interviews with the investigator:
Participant 0046: The first time it did it, I thought it was like ‘boring not gonna listen to you’, or like protecting myself, not hearing you. And then like one time it opened too and we were having a good conversation and it’s like ‘oh relaxing.’
Participant 0053: And when it undid I thought she was, she’s going to sit with me; she’s opening up.
Participant 0053: The zipper was closing off the scene… it’s closing up, the situation is tightening, it’s like over, the person is not, the openness is no longer there it must be like we’re coming to the end of something.
Participant 0048: When it was up I thought she (confederate) looked a bit uncomfortable and then when it came down she looked a bit more relaxed.
Subjects expressed negative perceptions with the zipping up/closed-collar condition as represented by phrases such as “boring not gonna listen to you” or “the situation is tightening, it’s like over” in contrast to positive phrases associated with the unzipping/open-collar condition like “she looked a bit more relaxed” and “oh relaxing.” These comments indicate an understanding of the collar states as performed by the Zipperbot action. Relaxed/comfortable as opposed to uncomfortable was the primary theme.
However a broad range of perceptions was noted. Reflecting the range of participant perceptions to Zipperbot’s unzipping action, one Participant 0055 discussed the sensual aspect of the action which is consistent with dressing or undressing. While at another extreme Participant 004 stated that they thought the zipper opening and closing was an environmental cue that the wearer was perhaps too hot or cold. This range reflects the spectrum of individual human response to any given situation.
Participant 0055: Oh well the going down was I mean it was more sensual, l, you know, the zipper top is, I don’t know, for me for it has like a very sensual feeling, especially visually.
Participant 004: some sort of like um environmental cue like she was comfortable or it was hot or something.
Perceived Body Posture
A variation on the positive and negative response theme was that of perceived body posture. Based on the review of the video, the confederate made neither hand gestures nor significant modifications to posture during the interactions. In spite of this, two participants recounted a leaning in posture which is considered a positive or engaging gesture when Zipperbot unzipped to the open-collar condition.
Participant 0046: I almost feel like but I don’t know if this is true, that there was times when it was open that maybe [confederate] leaned forward but I don’t know if she did?
Participant 0050: Maybe when it was down she was more comfortable cause she (confederate) was sort of leaning forward when it, she [confederate] was more comfortable as it went down.
This perception of positive body posture reinforces the theme of Zipperbot creating a positive effect for the observer when it unzips.
Questions of Control
Another evolving theme revolved around questions regarding ownership and control of the actions. For example, when Zipperbot would unzip the jacket, subjects were unsure of whether the confederate or the robot was in control. Some participants described their perceptions in terms of the conversation or overall situation providing descriptions such as “the situation is tightening.” This comment avoids ascribing actions to either the confederate or the robot and objectifies the action in an impersonal manner. Others ascribed those perceptions more towards the confederate referencing the pronoun she as in, “I just assumed she (confederate) was just making it go up and down.” Further still, participants mixed pronoun usage referring both to she and it when referencing the sartorial action, “And when it undid I thought she was, she’s going to sit with me, she’s opening up.” This seems to attribute the motivating action to Zipperbot but a change in state to the confederate, albeit an active state change in that she is going to sit or open up, both describing activity. This seems to indicate that the participants perceived Zipperbot as more of its own entity acting on the wearer. This underlying question of control and intention ascribed to Zipperbot, as to whether the robotic zipper was acting autonomously as its own entity or as an expressive physical behavior driven by or connected to the wearer, was evident in the following comments:
Participant 0055: It was kind of like, just had a life of its own, like kind of happening.
Participant 006: I just assumed she (confederate) was just making it go up and down.
Participant 0050: The first time it was funny because she (confederate) couldn’t do anything about it and just closing.
Along with the references to the pronoun it in descriptions and recollections of the zipper action, some participants’ statements reflected the autonomous nature of Zipperbot, “It was kind of like, just had a life of its own,” and that the confederate, in reference to the zipper was not in control, “she couldn’t do anything about it and just closing.” These observation indicates a perception of Zipperbot as some sort of an independent entity. In contrast one participant indicated the opposite, “I just assumed she was just making it go up and down,” but overall the range of conflicting interpretations points to a general uncertainty in regards to control.
Analysis of Body Language and Verbal Utterances
Considering that subjects had a variety of interpretations of the Zipperbot actions it was not clear how these perceptions might influence a conversation. So we looked at body language, examining positive nonverbal communications such as smiling, head nods and leaning in posture and negative nonverbal communications such as frowning, head shakes and leaning back posture. The video data was segmented based on zipper action and open-collar and closed-collar conditions then annotated using Anvil Video Annotation software in order to code physical gestures related to nonverbal communication and verbal interactions[Kip01]. Individual physical gestures were counted and because the time of participant responses and interactions varied, frequency of physical gesture was calculated into number of gestures per minute. A series of two factor without replication analysis of variance (ANOVA) was then conducted comparing the various gesture frequencies between conditions. In the analysis we did not determine statistically significant differences between the overall various conditions for gesture frequency. However, when we examine the specific time frame when the Zipperbot is in motion, we do observe an effect with combinations of body language. For example, one participant smiled, looked at Zipperbot and then tilted their head to the side in a quizzical manner. Another smiled and laughed, while yet another had the opposite reaction and frowned, then leaned back away from the confederate. These reactions again reflect the wide range of human responses when confronting a new experience.
Paired two sample t test
t(7) = 2.5
p < .05
Figure 7-17. Frequency of verbal utterances indicating a reduction from an open-collar state to a zipped up closed-collar state.
Verbal statements and verbal utterances such as yeah and mm are backchannel responses that have been shown to be an indicator of listener engagement[Lam11]. It was considered that the open-collar condition with no Zipperbot movement would be a more comfortable state and act as the neutral control. A paired two-sample t-test was performed comparing the mean frequency of verbal utterance during the conversation between the open-collar condition with no Zipperbot movement and the closed-collar with Zipperbot movement. The mean frequency of utterances for the open-collar condition scored a (M= 7.39, SD = 4.40) and for the closed-collar with Zipperbot movement a (M= 3.78, SD = 3.29). There was a significant effect on utterance frequency, t(7) = 2.48, p < .05, with the closed-collar with Zipperbot movement condition recording a lower frequency of utterances than the open-collar condition. This suggests that subjects became less engaged when Zipperbot activated and closed the collar of the wearer. While Condition 1 and Condition 2 showed statistically significant differences, a two factor without replication ANOVA of all the conditions was not statistically significant. However, the frequency of verbal utterances data did show a pattern consistent with our hypothesis. Utterances drop when the zipper closes the collar, then increase when the zipper opens the collar and then drop again when it closes the final time, Figure 7-17. While this represents a single measure, this would seem to support the data from the interviews. In addition it could be that the Zipperbot motion was distracting, as one subject described their thoughts:
Participant 0050: I’m not sure when it’s going to start moving so I keep looking down at the zipper and so um I feel like I’m not being a good conversationalist because I keep you know being distracted by something else.
Missing Gesture Phases
In the interviews after the study, when discussing the conversation and the actions of Zipperbot, subjects consistently used their hands to recreate the gesture of zipping or unzipping, including the very specific action of pinching with their fingers indicating the grasping of a zipper slide. This was noteworthy, as during the study conversations the confederate never used her hands to actuate the zipper, nor did the interviewer use this gesture when asking questions. This co-speech gesture suggests that there is an associated commonly understood arm/hand gesture to describe zipping and unzipping a garment. While this may seem obvious, it is an indication that arm/hand gestures associated with our clothing may be far more ingrained in our understanding of body language than previously thought. Some gestures, like co-speech gestures, can be described by a specific series of movement phases consisting of resting, preparation, stroke, hold, retraction and then back to resting[Kit98]. With autonomous action, Zipperbot lacks distinct movement phases associated with traditional gestures from the wearer. This means no arm/hand movement to indicate the beginning of the overall gesture. This begins to explain some of the ambiguity of the Zipperbot between being perceived as its own entity or as connected to the wearer. With no direct gesture from the wearer, combined with autonomous movement, control of the clothing has not been firmly established by the wearer, and as we’ve noted throughout this thesis, control of one’s body and as an extension, one’s clothing, is critical in social performance.
However, subjects still reported positive and negative perceptions in relation to the open-collar and closed-collar states but the gesture or lack of gesture was uncertain. This leads to the conclusion that there is a distinction between sartorial state, which is more the static appearance of clothing and sartorial gesture, which represents the action carried out in order to manipulate sartorial elements to achieve the sartorial state. While Zipperbot was successful in transitioning between sartorial states, the sartorial gesture was perceived more ambiguously.
Within study designs are known to be affected by carryover effects, that is, when conditions of one condition carryover to the next condition. In relation to the Zipperbot robot, when seen for the first time it tends to elicit strong reactions from observers and as such could have caused carryover effects which influenced the subsequent conditions and the study results. The carryover effect could go in either a positive or negative direction depending on whether the observer was initially amused by Zipperbot or if they found it jarring. Another limitation would be that the confederate had trained to respond with minimum affect, but being human, an absolute repeatedly consistent response is difficult to obtain. This also could have influenced the results.
There are certainly many applications and potential benefits for sartorial robotics or robots that are situated in close proximity with our clothing and bodies. However, from the Zipperbot study we learned that the implications for these robotic systems in terms of social interactions and communication is particularly complex. While sartorial cues might mainly fit within the appearance classification of nonverbal communication, the hand/arm movement to achieve the appearance seems to fit more within a gesture classification. Zipperbot allowed us to specifically remove the preparation and retraction phases of gesture and isolate the stroke phase. Subjects recognized the sartorial state change and the stroke phase of the sartorial gesture, but the absence of preparation and retraction phases seemed to put in to question the control or ownership of the overall gesture. This in turn creates some apprehension in interpreting social meaning.
Both the intimate proximity to the wearer’s body and the prominent status placed on our clothing as representations of our identity create a challenging space for robotics. In this space the robot, the clothing and the person represent a triadic social performance but seemingly performed by one singular entity. What is the robot? What is the clothing? What is the person? As such the repeating idea of control throughout this thesis becomes even more critical when considering the design of these systems. The individual roles during this performance should be either clearly delineated or seamlessly orchestrated to ensure proper social interpretations. This is a significant finding which will help inform design for future developments in wearable computing, wearable robots or in general ‘smart clothing’. Zipperbot represents the first steps towards introducing this level of control into a robotic sartorial system and building upon the knowledge generated from studying its human-sartorial robot interactions, we can now begin to utilize Zipperbot as a platform to explore and design more sartorial gestures.
The work in this thesis established sartorial robotics both as a practice and as a unifying premise for understanding the merger of soft-architecture robotics and social robotics. This demonstrates its potential as a beneficial field of study. In order to build upon this work it is advantageous to outline directions for future research. The next steps for future research should focus on wearable autonomous or semi-autonomous robotic surfaces in relation to social performance.
Agency and Control
When considering the prospects of wearable autonomous or semi-autonomous robotics in relation to social performance, agency and control were repeated key concepts. Independent or self-directed mechanical movement generally elicits thoughts of agency. However, whether or not agency is defined as an involuntary behavior or a purposeful action will depend on the context of the action and situation. The wearable element, either for people or robots, will imply a connection between the wearer and the robotic surface. The confederate, unlike in the Zipperbot study where the robot acted autonomously, also had the opportunity to wear the Hobble Dress and directly control the robot herself. The confederate had the unique experience of both controlling Zipperbot and having no control over Zipperbot. As an anecdotal note, she expressed completely different perspectives during both experiences. During the Zipperbot study she found that not being in control of her clothing was an unsettling experience, but she also found that its actions did make her feel either more closed off during the conversation when it zipped up, or more relaxed and open to the conversation when it unzipped, independent of the actual conversation. She commented on how she felt her personal space was affected by the actions. This mirrored the participants’ perceptions during the study. However when the confederate wore the Hobble Dress and was given direct control over the Zipperbot’s actions, she described experiencing a sense of empowerment. Obviously from a physical perspective, changing the fit of your clothing can change your attitude and have a psychological effect altering social interactions. It is noteworthy that these feelings came about not from the wearer’s actions but by the autonomous actions of Zipperbot. This distinction is important and leads us into the concept of performativity. Philosopher Judith Butler, in thinking through gender and identity in terms of performativity, states that:
…gender is in no way a stable identity or locus of agency from which various acts proceede; rather, it is an identity tenuously constituted in time-an identity instituted through a stylized repetition of acts. Further, gender is instituted through the stylization of the body and, hence, must be understood as the mundane way in which bodily gestures, movements, and enactments of various kinds constitute the illusion of an abiding gendered self[But88].
This concept of performativity expands beyond gender and suggests that identity is not the sole source of expression. That instead of an internalized notion of identity, identity is constructed through the external semiotics of our performative actions, such as speech acts, bodily gestures and of course in our judgment sartorial actions. From this perspective an autonomous robot acting on or as our clothing could have a profound influence on forming or shaping our identities. While Zipperbot is a playful sounding robot, it presents a far more complex system of social interactions than it originally implies. Given our observations of the wearer’s experiences regarding agency and control in both applications of Zipperbot and our understanding of performativity, it would be beneficial to run further studies in which volunteers would wear Zipperbot or other robotic clothing in order to measure the effects that control or lack of control might have on their own identities.
When we consider both the examples of fashion designers Alexander McQueen and Hussein Chalayan’s work, we see robots or machines acting on the garment, independent of the wearer. This may work for the fashion runway in creating a visual spectacle, but when we tested a similar experience with the autonomous actions in the Zipperbot study, this caused apprehension during human-human social interaction for the observer. Analysis of study interviews indicated that control of Zipperbot was in question. However when control over Zipperbot was added via an Android device in the Hobble Dress application, the model wearing the dress noted a heightened sense of empowerment. Through having more direct control, the wearer exhibits a mastery over technology, over the robotic surface and to an extent the wearer now becomes a virtuoso of identity play via clothing. For the observer, the simple addition of the touch gesture on the Android device creates a connection to the robotic surface and helps establish control and ownership of the sartorial robotic gesture. Just as there are subtleties to human sartorial gesture, there are subtleties to sartorial robotic gestures, and understanding all the nuances is a challenging task. Returning to Erving Goffman’s view of social interaction, in our attempt to influence how others perceive us by controlling our settings, manners and appearance, our clothing is situated somewhere between an object or prop and an extension of our bodies, if not considered an actual part of our bodies at times. Winnicott’s transitional object, generally defined as an object which can provide psychological comfort, also exists as a merger or link of our inner selves and the external world; clothing occupies this same space as being self and not-self at the same time [Win71][Ben00]. Through our clothing we play with this boundary and introducing robotic elements presents a profound modification to our customary social performances. In order to learn more about this we might break from the more conventional investigation methods and perhaps employ performers such as actors, dancers and performance artists who would be able to shed light on the dynamic range of body and sartorial gesture.
Sartorial State versus Sartorial Gesture
The evaluation of Zipperbot drew attention to the role of pre-gesture, the motions before the main stroke gesture, during sartorial transformations, and it also defined the importance of a sartorial state, which is more the static appearance of clothing and sartorial gesture, which is the more dynamic actions associated with transitioning from state to state. From the Zipperbot study, the open-collar would be considered one state and the closed-collar is the other state, the Zipperbot action changing the collar state would be the sartorial gesture. The evaluation of Zipperbot drew attention to the role of pre-gesture, the motions before the main stroke gesture, during sartorial transformations. Because of the importance of pre-gestures, future studies should examine adding either robotic pre-gestures or human pre-gestures to the robotic stroke gesture. Adding the pre-gesture would assist in establishing control of the stroke gesture and thus its ownership. A robotic pre-gesture may be as simple as flashing LEDs prior to beginning the stroke or a human initiated pre-gesture, based on capacitive sensing built into the robot, allowing a simple touch of the Zipperbot to activate and begin its stroke gesture. Either scenario would help in establishing ownership, but whether the pre-gesture is initiated by the robot or the human would be critical for the interpretation of the gesture. Further still, the specific type of pre-gesture could have an effect. A glance of the eyes down at Zipperbot to begin its movement as compared to a touch of the finger might be perceived quite differently but would also allow for more human improvisation during control. This future work would help to harmonize the sartorial state and sartorial gesture.
Another open question worthy of further investigation might be how Zipperbot would be perceived acting on the surface of another robot, perhaps a humanoid robot. As robots are generally thought of as electromechanical in nature, the addition of Zipperbot, also electromechanical, may be perceived as an extension of the robot much more easily than when experienced on a person’s body.
Designing with Gestures
Given the distinction we have made between sartorial state and sartorial gesture, future work should investigate principles for designing the sartorial robotic gesture. The development of the design principles for sartorial robotics helped us create robotic surfaces. In a similar manner, we’ll need a set of principles or tools for apparel designers and roboticists to design the specific gestures. This would initially require analysis of specific gestures, since, for example, zipping up a jacket would be very different than loosening a tie. Consider the entire process of movement and gesture involved in fully dressing one’s self from putting on socks to buttoning shirts; it is a highly sophisticated orchestration and in the field of sartorial robotics requires further study and consideration. As a general principle, actuated robotic surfaces seem to remove pre-gestures so understanding the meaning of this gesture phase in relation to clothing would be critical. Future work towards developing a taxonomy of sartorial gestures and an analysis of the parameters for understanding the gestures would be essential for designers. In an effort to get sartorial robotics into the hands of fashion designers and roboticists, the support(S) strategy developed in this thesis benefits the production of do-it-yourself (DIY) kits which would provide standardized systems that could be incorporated into a variety of projects in the same manner Zipperbot was applied to both a jacket and a dress. The development of DIY kits would allow researchers to experiment with sartorial robotics in their own work exploring the sartorial gesture phases and to contribute to the knowledge base of sartorial gesture.
For this thesis multiple Zipperbot prototypes were constructed with a variety of technical abilities. We were able to demonstrate a simple action of two Zipperbots working in tandem, but it is easy to imagine several working in an orchestrated manner for some more complex task. We can imagine multiple Zipperbots working together in complex self-assembling soft-architectures, for example portable structures like tents or highly technical clothing like an astronaut’s space suit which may have multiple parts to attach during donning or doffing. In terms of fashion, these multiple Zipperbot scenarios generate a variety of creative opportunities for performance and performativity. However in order to realize this level of coordination, scaling for multi-robot control will require better robot-robot communication, or a more intuitive interface for human operators to control multiple Zipperbots, perhaps focusing on reading natural movements and gestures of the wearer. From a technical research perspective for multi-robot control, Zipperbot would require wireless communications built into the main circuit board and an established communication protocol for multiple robots. Research in this area would help to facilitate the adoption of multiple Zipperbot applications and usage.
This dissertation makes contributions to several different areas including soft-architecture robotics, interdisciplinary dialogues for the design of robotic systems and nonverbal communication for both social robotics and wearable computing and fashion. What follows is a statement of the thesis’s contributions to each of these areas.
Within the emerging field of soft-architecture robotics, textiles, electronic textiles and fiber electronics have yet to be thoroughly explored particularly in the area of social robotics. The demonstration projects that are presented within this thesis such as Zipperbot and the Group Identity Surface help to illustrate and to extend the means of integrating these materials and processes into the field of robotics. While there is still much more work to be done in this area, this thesis has shown the possibilities and benefits of sartorial methodologies and practices to the field of soft-architecture robotics. The design framework which was developed as part of this thesis and which was used in the development of the sartorial robot demonstration projects will hopefully provide guidelines and methodologies to other roboticists in developing soft-architecture robots particularly those that might have a social interaction component. In particular the formulaic approach adapted from Barthes can assist in defining fashion and sartorial states in a manner which can help with designing sartorial interactions and assisting with robot aesthetics as viewed through the sartorial lens.
Interdisciplinary dialogues between robotics, textile design and apparel design will act as an initial step in further opening the field of robotics to outside disciplines. As more robots and robotic products enter the marketplace, more diverse fields of study will be required to meet the full spectrum of public needs. As robotics moved out of the industrial settings of the factories and into people’s homes, psychology began playing a role in designing the human-robot interactions. When researchers began to explore robotics for educational settings, we saw roboticists reaching out to teachers and educators for guidance. In the development of this thesis, we experienced with the Honda ASIMO project the utilization of electronic-textiles to sense and explore social touch gestures. Working with soft-architecture robotics, the expertise of apparel designers and textile designers was crucial. Apparel design in particular is instilled with much tradition and as an artisan pursuit has drastically different tools and language than those found in robotics. Bridging this difference was a challenge, but more an opportunity as the fields of both soft-architecture robotics and wearable computing will need to fuse successfully with apparel for widespread adoption.
Sartorial Nonverbal Communication
While the overall aesthetic of robots is often considered in the field of robotics, the social and cultural role of clothing as a nonverbal communication system has been largely overlooked within robotics. Even when robots distinctly lack clothing, people tend to view the robot form through a sartorial lens. Clothing is a form of nonverbal communication among people and this thesis work explored how that language can transition to social robotic interactions. From the results of the Sartorial Robot Gender Study, we observed increases and decreases in the perceived masculine and feminine traits indicating that sartorial cues representing gender did in fact have an impact on people’s perceptions of gender traits for robots. There are many different types of fashion and sartorial signaling, and while this study represents one particular sartorial cue for one particular social construct, that of gender, its results show that sartorial signaling as nonverbal communication for human-human interactions can translate to human-robot interaction and should be considered a fertile area for future work. This thesis has explored the role of sartorial cues in robotics and has laid the foundation for its further development and inclusion into research.
Through the development and use of Zipperbot as a sartorial robot and the resulting behavioral analysis, we were able to explore the distinction between the sartorial state, which is more the static appearance of clothing and sartorial gesture, which represents the action carried out in order to manipulate clothing elements to achieve the sartorial state. Zipperbot allowed us to remove phases of hand/arm movement within the gesture which in turn disrupted the ownership of the gesture and helped to illuminate the importance of the sartorial gesture in using clothing as nonverbal communication. This suggests that clothing is not only a form of nonverbal communication based on appearance but can be a combination of both appearance and gesture.
Previous researchers and literature have postulated that control of one’s body is crucial to being perceived as a competent social performer. This thesis established that this same concept extends to clothing and our wearable surfaces. When Zipperbot performed a gesture with the wearer’s clothing, ownership of that gesture was uncertain and there was a question over control of the clothing. This in turn reflects on the wearer in terms of their ability to socially perform. This level of understanding will be essential in the design of wearable robotics or smart clothing that intends to do any form of actuation.
The world of fashion and the field of robotics are socially and culturally diverse. This thesis established a foundation for merging the two fields into sartorial robotics so that both can contribute to each other. This synthesis will enhance the social aspects of human-robot interactions and assist in how we situate robotics in our lives and cultures.
Emi12: , (Glazer, 2012),
Gra04: , (Scambler, 2004),
Bar90: , (Barthes, 1990),
Ahm01: , (Ahmed, 2001),
Gra97: , (Granville Stanley Hall, 1897),
Gof59: , (Goffman, 1959),
Car: , (DiSalvo C. , 2009),
Pau69: , (Hamid, 1969),
Kar11: , (Pine, Fletcher, & Howlett, 2011),
For88: , (Forsythe, 1988),
Forwn: , (Forlizzi, Zimmerman, Mancuso, & Kwak, 2007),
Lee08: , (Lee, 2008),
Pra07: , (Pratt, Hauser, Ugray, & Patterson, July, 2007),
Nas05: , (Nass & Brave, 2005),
Mar11: , (Marx & Engels, 2011),
The28: , (The Harvard Crimson, 1928),
Ele28: , (Popular Mechanics, 1928),
Sch07: , (Schaut, 2007),
Lis94: , (Harvey, 1994),
Lan78: , (Winner, 1978),
JBa68: , (Bauldrillard, 1968),
Jam84: , (Cameron, 1984),
REP97: , (Post, 1997),
Pan07: , (Panhuis, 2007),
HuL10: , (Hu, 2010),
Abo07: , (Abouraddy, 2007),
Sor07: , (Sorin, 2007),
Eli09: , (Thomson, 2009),
Bra09: , (Keim, 2009),
EHa10: , (Hawkes, 2010),
Don11: , (Melanson, 2011),
Fer09: , (Feraeus & Jacobsson, 2009),
Mit08: , (Mitchell, 2008),
McQ: , (McQueen, 1999),
Hus10: , (Chalayan, 2007),
Tho98: , (Thorpe, 1998),
Kni09: , (Knight, Toscano, Stiehl, Chang, Wang, & Breazeal, 2009),
Mar04: , (Martin, Ambrose, Diftler, Platt, & Butzer, 2004),
Sti05: , (Stiehl, Lieberman, Breazeal, Basel, Lalla, & Wolf, 2005),
Ori11: , (Hirata, 2011),
Cra01: , (Crane, 2001),
Hol98: , (Holt & Ellis, 1998),
Wei81: , (Weil, "Bag" Radio, 1981),
PEk78: , (Friesen, 1978),
Pau02: , (Ekman, 2002),
Mil11: , (Miller, Van den Berg, Fritz, Darrell, Goldberg, & Abbeel, 2011),
Wan11: , (Wang, Miller, Fritz, Darrell, & Abbeel, 2011),
Mil111: , (Miller, Fritz, Darrell, & Abbeel, 2011),
Mlo13: , (Mlot, 2013),
Dav09: , (Davidson, 2009),
Mor70: , (Mori, 1970),
Ays11: , (Saygin, Chaminade, Ishiguro, Driver, & Frith, 2011),
Kid11: , (Kiderra, 2011),
Say12: , (Saygin A. P., 2012),
Byr96: , (Reeves & Nass, 1996),
Ema10: , (Emami, 2010),
Gra11: , (Gray, Knobe, Sheskin, Bloom, & Barrett, 2011),
Wei83: , (Weiner L. , 1983),
Pre98: , (Presley, 1998),
Lam11: , (Lambertz, 2011),
Kip01: , (Kipp, 2001),
Kit98: , (Kita, van Gijin, & van der Hulst, 1997),
But88: , (Butler, 1988),
Win71: , (Winnicott, 1971),
Ben00: , (Benson, 2000),