492 The Professional Counselor | Volume 10, Issue 4 Data Collection We collected data through a demographic questionnaire and semi-structured individual interviews. The demographic questionnaire consisted of nine questions focused on general demographic characteristics (i.e., gender, age, race, and education). Additionally, we asked questions focused on participants’ experiences as researchers (i.e., professional organization affiliations, service, conference presentations, publications, and grant experience). These questions were used to triangulate the data. The semi-structured interviews consisted of eight open-ended questions asked in sequential order to promote consistency across participants (Heppner et al., 2016) and we developed them from existing literature. Examples of questions included: 1) How would you describe your research identity? 2) Identify or talk about things that happened during your doctoral program that helped you think of yourself as a researcher, and 3) Can you talk about any experiences that have created doubts about adopting the identity of a researcher? The two doctoral students on the research team conducted the interviews via phone. Interviews lasted approximately 45–60 minutes and were audio recorded. After all interviews were conducted, a member of the research team transcribed the interviews. Data Analysis and Trustworthiness We followed grounded theory data analysis procedures outlined by Corbin and Strauss (2008). Prior to data analysis, we recorded biases, read through all of the data, and discussed the coding process to ensure consistency. We followed three steps of coding: 1) open coding, 2) axial coding, and 3) selective coding. Our first step of data analysis was open coding. We read through the data several times and then started to create tentative labels for chunks of data that summarized what we were reading. We recorded examples of participants’ words and established properties of each code. We then coded line-by-line together using the first participant transcript in order to have opportunities to check in and share and compare our open codes. Then we individually coded the remainder of the participants and came back together as a group to discuss and memo. We developed a master list of 184 open codes. Next, we moved from inductive to deductive analysis using axial coding to identify relationships among the open codes. We identified relationships among the open codes and grouped them into categories. Initially we created a list of 55 axial codes, but after examining the codes further, we made a team decision to collapse them to 19 axial codes that were represented as action-oriented tasks within our theory (see Table 1). Last, we used selective coding to identify core variables that include all of the data. We found that two factors and four subfactors most accurately represent the data (see Figure 1). The auditor was involved in each step of coding and provided feedback throughout. To enhance trustworthiness and manage bias when collecting and analyzing the data, we applied several strategies: (a) we recorded memos about our ideas about the codes and their relationships (i.e., reflexivity; Morrow, 2005); (b) we used investigator triangulation (i.e., involving multiple investigators to analyze the data independently, then meeting together to discuss; Archibald, 2015); (c) we included an internal and external auditor to evaluate the data (Glesne, 2011; Hays &Wood, 2011); (d) we conducted member checking by sending participants their complete transcript and summary of the findings, including the visual (Creswell & Miller, 2000); and (e) we used multiple sources of data (i.e., survey questions on the demographic form; Creswell, 2007) to triangulate the data.