Fransella (2005) argued that the nature of elements depends on the context being explored. Elements must be “within the range of convenience of the constructs used” and “representative of the area being investigated” (Fransella, 2005) (p. 18). They must also be discrete, evaluative (Stewart et al., 1981), homogenous, and relevant to all participants (Easterby-Smith, 1980). The selection of elements has two feasible alternatives:

1) Supplied elements, which are provided by the researcher;

According to Reger (1990), several factors support a researcher’s decision to supply elements. First, elements may be provided if a researcher’s aim is to learn about a given set of elements from several participants. Second, elements may be supplied if a researcher wants an existing theory to guide the selection. Third, elements may be given if the researcher is interested in comparing the responses of different participants to a standard set of elements. Such comparison could be within a homogenous group of participants or across different ones.

Meanwhile, researchers can ask participants to elicit elements that they themselves consider relevant. Elements can be elicited in different ways (Easterby-Smith, 1980):

1) Researchers can define role or situation descriptions in which participants provide their own specific examples that fit these descriptors.

A “good” construct expresses participants’ meaning precisely and completely, has a clear contrast, is detailed appropriately, and has a direct link to the research topic (Jankowicz, 2005). RGT supports several construct elicitation methods. Minor variations and combinations can be applied to some of these methods depending on the research aim (Easterby-Smith, 1980; Beail, 1985; Reger, 1990). The triadic sort method is believed to be the most convenient when exploring participants’ constructs (Kelly, 2003).

1) Supplied constructs, in which the researcher provides the constructs.

As with all methods for eliciting personal constructs, the resulting constructs are not required to be used in some form of repertory grid (Fransella, 2005). The obtained information can be of sufficient use in itself. However, elements are linked to constructs in three ways: dichotomizing, ranking, and rating (Easterby-Smith, 1980; Tan and Hunter, 2002).

1) In dichotomizing, elements are either close to the left construct pole or the right construct pole and are labeled accordingly.

The intensive nature of RGT often means the sample size is relatively small but it can be also large using the new rep:grid systems. This helps in generating quantitative data to understand a large population. However for qualitative data, a sample size of 15–25 within a population will frequently generate sufficient constructs to approximate the “Universe of meaning” regarding a given domain of discourse (Dunn, 1986; Ginsberg, 1989). That is, no new constructs are normally added despite an increase in sample size. Studies using RGT with a small sample size can develop items for larger sample instruments such as closed-ended preprepared constructs.

According to Scopus, more than half of the studies in our sample (432 of 782) were published in more than one subject area. One record contributed to six different subject areas: social sciences, health professions, agriculture and biological sciences, arts and humanities, psychology, and nursing. Another contributed to five different subject areas. Of the 432 records, 14 contributed to four subject areas, 36 three subject areas, and 315 two subject areas. The remaining 350 of the 782 records contributed to only one subject area, in which the subject areas of 72 studies were classified as “others.” Figure 3 illustrates the distribution of these studies among subject areas defined by Scopus.

As highlighted in Figure 3, a total of 30 records identified within the built environment scope also contributed to one or more subject areas such as social sciences (n = 19), environmental science (n = 2), engineering (n = 2), agricultural and biological sciences, decision sciences, and economics (n = 1 each). The subject areas of four studies were classified as “others.”

The included records were published in several parts of the world as indicated in Figure 4. To a greater extent, and since 1967, research has been conducted in the United Kingdom (n = 290), the United States (n = 77), Australia (n = 67), Spain (n = 36), Canada and Germany (n = 35 each), Taiwan (n = 28), Italy (n = 21), Sweden (n = 20), the Philippines (n = 14), New Zealand and Turkey (n = 12 each), China and Netherlands (n = 11 each), and Hong Kong and Switzerland (n = 10 each).

On a smaller scale, studies have been published in Belgium (n = 9); Denmark, Ireland, and Japan (n = 6 each); India and South Africa (n = 5 each); France, Malaysia, and the Russian Federation (n = 4 each); Norway (n = 3); Argentina, Brazil, Czech Republic, Finland, Greece, Pakistan, Portugal, and other countries (n = 2 each); and Bahrain, Bangladesh, Chile, Cyprus, Iceland, Indonesia, Iran, Luxembourg, Mexico, Namibia, Singapore, South Korea, Thailand, and the United Arab Emirates (n = 1 each).

As for the records within the built environment field, research has been conducted in the United Kingdom (n = 10); Australia (n = 5); Malaysia (n = 3); Japan and Switzerland (n = 2 each); and Bahrain, Greece, Hong Kong, Luxembourg, Taiwan, Turkey, Spain, and the United States (n = 1 each). One record was classified as “others.”

Figure 5 illustrates the 782 studies published between 1967 and 2020 as a line graph. Overall, the publication rate improved during that period. From 1967 to 1994, the publication rate remained consistent, from one to seven publications per year. The rate increased dramatically afterward, peaking at 36 publications in 1997 and again in 2000. Between 2001 and 2015, the publication rate fluctuated, reaching a low of 17 studies published in 2003 and 2008 and a high of 31 studies published in 2012 and 2014. Later, the number of publications increased to 44 in 2017 but slightly fell to 38 in 2020. Meanwhile, studies within the built environment field emerged in 1978. Studies published in 2012 and 2014 peaked at four (out of 30).

Within the built environment scope, Table 2 lists the research aims of studies that have adopted RGT.

Based on the definition of the built environment cited earlier, these studies can be classified as either direct or indirect contributors to the field. Direct contributors included record 234 (Dayaratne, 2016), which used RGT to examine the applicability of physiological techniques in a housing project, and records 133 (Tu et al., 2018) and 135 (Cheah et al., 2018), which adopted RGT to promote sustainable behavior practices or avoid extra energy consumption. Additionally, records 26 (Ratnasingam et al., 2020), 209 (Lallemand and Koenig, 2017), 712 (Vassiliadis and Fotiadis, 2007), 1078 (Mitchell and Kiral, 1999), and 1088 (Mitchell and Kiral, 1998) used RGT to explore human–space interaction, user experience, and spatial experience; assess images of space, or determine material selection. Records 301 (Wan and Shen, 2015), 541 (Abdul-Rahman et al., 2011), 606 (Home et al., 2010), 740 (Home et al., 2007), 875 (Selby, 2004), 1503 (Potter, 1986), 1505 (Potter and Coshall, 1986), and 1652 (Taylor and Stough, 1978) used RGT to elicit users’ perceptions of key attributes of urban green spaces, understand the meanings of urban green spaces, provide planners with the necessary tools, identify spatial inequalities, or perform multivariate analysis in geography.

Meanwhile, indirect contributors included record 1062 (Okoroh and Torrance, 1999), which used RGT to analyze the subcontractor’s risk elements in constructing refurbishment projects, and records 24 (Kislali et al., 2020), 736 (Byrne and Skinner, 2007), 745 (Pike, 2007), 751 (Naoi et al., 2007), 797 (Naoi et al., 2006), 885 (Hankinson, 2004), 917 (Pike, 2003), 922 (Waitt et al., 2003), 933 (Caldwell and Coshall, 2002), 1025 (Coshall, 2000), 1174 (Botterill and Crompton, 1996), and 1411 (Embacher and Buttle, 1989), which adopted RGT to evaluate users’ experience, perception, and destination image and its assessment. Finally, the reviewers grouped 30 records into six main topics according to the research aim and keywords: architecture and meaning, building sciences, interior architecture, management and decision-making, tourism and city branding, and urban planning and design. Table 3 summarizes these topics and their typologies.

To explain how RGT can be methodologically practiced within the built environment field, this section presents the results of the seven records categorized as direct contributors to the field and published after 2010. The main components of RGT are reported for each record. Table 4 summarizes the use of RGT design components for these records.

Records in the above table used RGT to explore the personal constructs of end users, subcontractors, managers, architects, and urban designers with the aim to evaluate users’ experience, perception, and destination image; promote sustainable behavior practices; manage energy consumption; and elicit key attributes of urban green spaces.

To develop and construct a housing neighborhood in Sri Lanka (Dayaratne, 2016), residents’ input was used over two tasks. The first sorting task was conducted with 12 families to achieve a profound understanding of their existential consciousness and conceptualization of their setting. The authors asked participants to elicit elements by writing the locations they recognize of the present setting on sorting cards. These sorting cards were then presented to each participant as elements to sort into several groups and elicit 10 positive and negative constructs. The sorting process was conducted until no further groups could be generated. The second sorting task was conducted with 20 participants to divulge their conceptualizations and fantasies about the places they wished to see in the new neighborhood. The previous process was repeated, and 28 spatially relevant constructs were used.

In (Tu et al., 2018), two phases were used to promote sustainable behavior practices. The first phase used RGT to understand participants’ views of the Design for Sustainable Behavior (DfSB) approach, develop an assessment tool in the form of a grid, and propose design strategies for improving the sustainable behavior of household appliances. The second phase used a questionnaire to verify the developed tool. In the first phase, four designers were given six design approaches written on cards as elements. The participants were asked to randomly select three cards, elicit the similarities between two cards, and elicit the difference between those and the third card. The participants repeated the process until no new constructs could be extracted or a certain period had elapsed. A total of 187 constructs falling under 23 typologies were obtained. The 10 most common constructs elicited and the six supplied elements were used to build a 6 × 10 grid tool for understanding the perceptions and experiences of 38 new young designers, who rated each element based on the construct poles using a five-point Likert scale. To establish the basic concepts of DfSB approaches among participants in the two parts of the first phase, the participants read a short descriptive essay before the interview sessions.

To investigate motives and challenges in preventing electricity wastage, the participants in (Cheah et al., 2018) were asked to list up to six household electricity users. They were supplied with two additional elements and were then asked to randomly select three cards representing three household members and group two cards that were similar but different from the third. The selection criteria were based on members’ motivations for not wasting electricity. Once the participants began answering, the constructs were elicited. In every triad process, the participants rated family members according to the constructs in the grid and using a seven-point scale. The elements were placed back into the pile and reshuffled once the participant could no longer elicit new constructs from the same set of cards. Laddering was used, and the process was repeated until the participants either could not generate new constructs or became visibly tired.

To determine the most preferred attributes of overlay materials used in wood-based panel furniture, the participants examined three wood samples (Ratnasingam et al., 2020) by describing the characteristics and attributes in which one sample was different from the other two. The participants rated all three samples with regard to the elicited constructs on a five-point Likert scale.

To explore students’ perception of learning environments, 26 students were given 37 printed Web pictures representing different learning spaces during three focus sessions (Lallemand and Koenig, 2017). They followed standardized instructions on how to compare triads and elicit constructs and had 20 min to complete this task individually and repeat the process until saturation was reached. The authors then clustered all contract pairs into six distinct and relevant categories under an affinity diagram format.

To elicit users’ perceptions and assessments of key attributes of urban green spaces (UGS) in Hong Kong, 21 participants were asked to elicit nine UGS that fit a given set of descriptors (Wan and Shen, 2015). Eight descriptors were assigned to a UGS they recall visiting in a certain period, and a ninth descriptor labeled “ideal urban green space” was not assigned but was for their imagination. Each participant was presented with at least six different triad combinations of the elicited elements to generate 131 constructs, which were later used to develop a questionnaire.

To understand the meanings of UGS and provide planners with a tool to match urban natural resource management with residents’ needs (Home et al., 2010), 17 participants were supplied with nine photographs of UGS representing different green spaces in Zurich and one imaginary ideal landscape. They were presented with random groups of three elements to elicit construct pairs. A total of 118 constructs were generated and linked using a Likert scale.

Elements can be supplied if the research aim is to learn about a given set of elements from several participants. For example, in (Ratnasingam et al., 2020), the researcher provided participants with a set of overlay materials to learn more about their characteristics and attributes from the participants’ perspectives. In (Lallemand and Koenig, 2017), the researcher supplied participants with 37 printed Web pictures of learning spaces to explore their perceptions of learning environments. In (Lallemand and Koenig, 2017), the researcher presented participants with nine photographs of UGS that represent different green spaces in Zurich. Elements can also be given if the researcher is interested in allowing an existing theory guide element choice or in comparing the responses of several respondents given a standard set of elements. None of the seven records followed these criteria. Meanwhile, elements can be elicited if they are implied by the study or if they must be relevant to the participants (Easterby-Smith, 1980). This method is particularly effective if the researcher’s aim is to involve laypeople’s perceptions in the design process. For example, in (Dayaratne, 2016), the researcher (designer) asked participants to elicit places they recall in their current neighborhood and places they wish to have in a future development. In (Wan and Shen, 2015), because of the large number of elements that meet the research criteria and their relevance to the participants, researchers asked participants to elicit places they have visited relying on their real-life experience. They were provided with descriptors to guide their elicitation process. However, in (Cheah et al., 2018), because the research sought to study specific behaviors toward energy consumption, participants were asked to elicit names of household users as elements for the experiment.

Whether supplied or elicited, elements refer to stimuli that are significant in a research study. They can be in the form of printed or digital photographs (Home et al., 2010; Lallemand and Koenig, 2017), text, people (Cheah et al., 2018), or objects (Ratnasingam et al., 2020) or any other suitable format based on research purpose. In cases where real experiences are difficult to reproduce, researchers can use pictures or objects without introducing particular biases. When a large number of elements is relevant to the study or when the identification of a common list for all participants is difficult, researchers can provide participants with a set of descriptors, roles, or situations from which they can elicit their own specific examples that fit these categories. For cases where participants might be unfamiliar with the topic, researchers can provide them with short descriptive essays before interview sessions as in (Tu et al., 2018).

RGT supports several methods for eliciting constructs to which minor variations and combinations can be applied. The minimum context form, also known as the triadic sort method, is considered the most convenient and common method for exploring a participant’s constructs. This method involves the selection of three random elements from a full set of elements. Participants then identify how two elements are similar yet different from a third. In (Home et al., 2010; Wan and Shen, 2015; Lallemand and Koenig, 2017; Cheah et al., 2018; Tu et al., 2018; Ratnasingam et al., 2020), researchers asked participants to elicit constructs through the minimum context form.

Dyads can also be used to elicit constructs if participants find it difficult to generate constructs via the triadic method or if the elements themselves are complex (Easterby-Smith, 1980; Keen and Bell, 1980). None of the records in Table 4 implemented this approach.

Another method to elicit constructs is the full context form. In this method, participants sort a full set of elements into groups using any similarity criteria of their choice (Easterby-Smith, 1980). In (Dayaratne, 2016), the full context form was adopted, and the participants were asked to elicit constructs in the form of short descriptive titles for each pile of elements. Other methods include supplied constructs or group construct elicitation. Again, none of the studies in Table 4 used group construct elicitation or supplied the constructs, as the methods did not elicit the required data to answer the research question.

As with all methods for eliciting personal constructs, the resulting constructs are not required to be used in some form of repertory grid (Fransella, 2005); the information gained can be deemed sufficient in itself. Therefore, not all researchers will perform linking. As Table 4 indicates, in (Wan and Shen, 2015; Dayaratne, 2016; Lallemand and Koenig, 2017), researchers were satisfied with the information they obtained and did not use any linking method.

In most cases, elements are linked to constructs through ratings. When rating elements, participants exercise greater freedom and are not compelled to make discriminations where none exist. In (Home et al., 2010; Cheah et al., 2018; Tu et al., 2018; Ratnasingam et al., 2020), ratings were used. When the range of rating values exceeds the number of elements, participants’ freedom is maximized. For example, in (Ratnasingam et al.), the participants rated three samples on a five-point scale; in (Tu et al., 2018), the participants rated six elements using a five-point scale; and in (Cheah et al., 2018), the participants rated eight elements on a seven-point scale.

It is important to mention that not all the practice of RGT is responding effectively to the original theory of Kelly especially in the case of group elicitation and rating approaches.

It is important to note that RGT can be used as a scientific research method solely or in conjunction with other approaches. It can also help validate other methods or serve as a preliminary phase for further research. This was evident in most studies such as (Wan and Shen, 2015), where researchers used the construct elicitation results as a base to develop a closed-ended constructs for their research purpose.

Furthermore, Dayaratne (2016) believed that in conventional surveys, participants mostly express negativity. However, RGT allows participants to freely express both negative and positive conceptualizations that are real and strongly experienced.

Also worth noting is that in some cases, elements and constructs can be elicited by a participant group and then supplied to another participant group for research purposes. For example, in (Tu et al., 2018), four designers elicited elements and constructs that were then rated by another group of 38 young designers. However, the data analysis of individual RGT can be studied in groups of individuals to understand how they perform in a certain task.