To begin, we conducted a scoping review focused on the specific experience of embodiment of technologies worn or carried, for example, mobile phones or smart watches, searching for terms describing the experience of embodiment with a tool (Textbox 1). The scoping review included: (1) identifying the research question; (2) identifying relevant studies; (3) study selection; (4) charting the data; (5) collating, summarizing, and reporting the results; and (6) consultation [37,38]. Our review resulted in a total of 80 papers in disciplines that included electronic health, neuroscience, mobile computing, wearable computing, ubiquitous computing, psychology, sociology, and philosophy. Although some areas specifically describe the embodiment of technology, other research described the embodiment of tools such as Merleau Ponty’s example of a blind person embodying their walking cane [39]. We believe these examples also apply to the embodiment of technologies. Therefore, studies which included the embodiment of a technical or nontechnical tool were included in the review. In total, 20 of the papers were discarded because they described embodiment without the use of technology or tools and therefore did not help in developing the concept. The 60 remaining papers were organized and analyzed for their description of the embodiment of technology/tools.

A structure was utilized to organize the study characteristics of the 60 papers, including: (1) authors and publication year, (2) main research findings, (3) research design details (ie, experimental or nonexperimental design), (4) item embodied (ie, mobile and wearable technology or physical prosthetic), (5) description of the specific kind of embodiment, (6) measurement items, and (7) embodiment dimensionality. Using this literature review, we could establish the entity (the whom/what) and the property (the relevant process/aspect) of the construct, which created the foundation for our construct definition [31,35,36,40]. The entity here is obviously a person. We described the property as experiencing wearable technology and perceiving an extension of oneself [4,17,41,42].

After having defined the property and the entity of wearable technology embodiment, we explored the dimensionality of the concept [43-45]. Although most of the examined literature suggested the existence of multiple dimensions, there seemed to be no overall consensus on either the number or naming of the dimensions. We therefore decided to uncover the most plausible categorization by systematically collecting, juxtaposing, and comparing possible dimensions as mentioned in previous studies. This process, which is also known as structured conceptualization [44], resulted in the emergence of 3 clear themes: body extension, cognitive extension, and self-extension. These themes were not only the most popular topics but also covered the full spectrum of the individual (ie, body, mind, and sense of self). Body extension refers to a physical addition or replacement of the body. For example, a robotic hand that communicates touch to the end of the human limb, improving dexterity [46]. Cognitive extension, sometimes called cognitive prostheses [17,47], refers to the experienced extension of one’s cognitive capabilities such as navigation assistance or knowledge of the number of steps taken during a day. Finally, self-extension refers to an object being perceived as part of a person’s identity or sense of self. For example, experiencing a mobile phone as a representation or extension of yourself and personalizing the technology to be congruent with your self-image [11,15,48,49].

We then combined the entity (a human individual), property (experiencing wearable technology and perceiving an extension), and dimensionality of the construct (body extension, cognitive extension, and self-extension) to concisely define our construct [31]. This led to the following definition of wearable technology embodiment: A person experiencing technology worn on or near the body perceiving a certain extension of the body, cognition, or (sense of) self.

To come up with a first set of measurement items, we started the process of item generation, sorting, and selection (for a schematic overview of the item sorting and selection process, see Multimedia Appendix 1.) From our literature review, a first set of 24 preliminary items (12 body extensions, 7 cognitive extensions, 5 self-extensions; see Multimedia Appendix 2) was generated from the embodiment descriptions [43-45]. Some scales from the prosthesis literature were adapted to describe the extension of the body, which resulted in more items within this dimension. We then followed Hinkin and Tracey’s [50] content validity assessment approach to adapt and refine the wording of the 24 items. In total, 2 members of the research team independently assessed the items within each dimension and then compared their assessments to determine whether they tied in with our definition, sat in the correct dimension, and read clearly and concisely [31,51,52]. During this process, 6 items were removed (items 8 to 12 and 19, see Multimedia Appendix 2), which led to an updated item pool of 18 items.

To judge the content validity of the 18 measurement items, we made use of 8 expert interviews [31,43,52,53] to examine each item for comprehension, applicability, and fit into the construct’s dimension [54]. In total, 2 of the experts worked as researchers at universities within human-computer interaction, one expert was an information systems researcher at a university and specialized in measurement scale development, 3 experts worked in wearable technology innovation and development, and 2 experts were individuals who were recently given wearables for the first time.

The experts were interviewed either in person or through a Skype conference call. After receiving a brief description and definition of wearable technology embodiment, they completed a Web-based survey, rating the applicability of each item from 1 (very inapplicable) to 5 (very applicable) and whether the item fits in the dimension. The experts chose whether they believed the item should be: (1) kept as is, (2) modified in a minor way, (3) modified in a major way, or (4) omitted [55]. Each item included an open field for suggestions or considerations [56]. The interviews ended with a brief discussion regarding the classification of 3 dimensions and possible suggestions for any new measurement items. The experts approved the dimensionality of our scale (ie, body extension, cognitive extension, and self-extension) and no new items were recommended. On the basis of their input, a few items were slightly reworded to improve their linguistic clarity. Using the feedback of the experts, 5 items were removed (items 5 to 7, 18, and 24; see Multimedia Appendix 2) because of repetition or lack of fit with the construct. This resulted in an updated measurement item pool of 13 items, which were subsequently used for empirical testing.

To test the dimensionality and further refine the scale items, we ran an exploratory factor analysis (EFA, principal component analysis with varimax rotation) on the set of 13 measurement items [57]. Aggregating the data of the 3 vignettes, each with 182 respondents, led to a sample that contained 546 responses. In total, 3 items were removed from the analysis (items 4, 16, and 17; see Multimedia Appendix 2) as they loaded substantially on 2 dimensions or more [58]. Rerunning the EFA with the 10 remaining items confirmed the 3 dimensions of wearable technology embodiment (Kaiser-Meyer-Olkin Measure of sampling adequacy ((0.80) Bartlett’s test of sphericity 2390, P<.001) and accounted for 71.22% of the variance. All items loaded significantly on only 1 dimension and all factor loadings were above the recommended threshold value of 0.50 [58]; therefore, providing first evidence of the convergent and discriminant validity of the measurement instrument. As the largest factor within the EFA explained less than 50% of the variance (28.5%), evidence for common method bias was not found [31,59].

To further test the measurement instrument, we performed a confirmatory factor analysis (CFA) [60,61] using the software package IBM, SPSS Amos 23 [57] (maximum likelihood estimation). We tested the EFA solution of 3 dimensions as intercorrelated first-order factors [61]. See Table 3 for: Chi square degrees of freedom calculated probability, Minimum Discrepancy Degrees of Freedom (CMIN/ df), Goodness of Fit (GFI), Adjusted Goodness of Fit (AGFI)), Normed Fit Index, Incremental Fit Index, Tucker Lewis Index (TLI), Comparative Fit Index (CFI), root mean square error of approximation (RMSEA), Akaike information criterion, Browne- Cudeck Criterion (BCC), Bayesian Information Criterion (BIC). After removing one item (item 23, see Multimedia Appendix 2) to improve the model fit, the found 9-item solution demonstrated a good fit with the data (CMIN/df) <5; GFI, AGFI, Normed Fit Index Incremental Fit Index (NFI), TLI, CFI >0.90; RMSEA <0.08; Table 3).

To further test the applicability of our dimensions, we tested 2 alternative models [52]: a model of 3 uncorrelated first-order factors and a model treating the 9 items as indicators of 1 first-order factor. The CFA results of the alternative models clearly showed that the alternative models did not have a good fit with the data (CMIN/df>5; GFI, AGFI, NFI, TLI, CFI<0.90; RMSEA<0.08; Akaike information criterion, BCC, BIC>scores of 3 first-order correlated model). These outcomes confirm that wearable technology embodiment is best modeled as a set of 3 correlated, first-order factors. In addition, the very poor fit of the 1 first-order factor reconfirmed the absence of common method bias.

The convergent validity of the 3 dimensions and their reliability (Table 4) was confirmed via the CFA factor loadings (>0.70) [58], average variance extracted (AVE) values (>0.50) [53], and minimum item to total correlations (>0.40) [62], which exceeded the established threshold value. The reliability of the dimensions was also confirmed as Cronbach alpha and composite reliability exceeded 0.70. The discriminant validity (Table 5).could be confirmed as the AVE of each construct exceeded the values of the crossconstruct squared correlations [63,64] .

To test the predictive validity of the wearable technology embodiment instrument, we utilized 5 dependent variables related to technology adoption: trust [55], involvement [65], perceived usefulness [66], attitude toward use [67], and continuous intention [68]. Structural equation modeling was used to test the extent to which the 3-dimensional wearable technology embodiment construct explained each of the 5 dependent variables by making use of the software package IBM SPSS Amos 23 [57] (maximum likelihood estimation).

The results of the analyses (see Multimedia Appendix 5) confirm a good fit with the data as the GFI, AGFI, NFI, TLI, and CFI exceed the recommended value of 0.90 and the RMSEA does not surpass the value of 0.08 [58]. The 3 dimensions of wearable technology embodiment explain considerable (trust and perceived usefulness) to acceptable amounts of the variance (involvement, attitude, and continuous intention) of the dependent variables [69]. Except for the influence of body extension on the attitude (beta=.17: P=.004), the standardized paths imply that the found influences account for a substantial proportion of the variance [70] (beta>.20). Overall, the results confirm the predictive validity of both the multidimensional construct and individual dimensions of the wearable technology embodiment instrument.