Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Sexual harassment in academe is underreported, especially by students in the life and physical sciences

  • Stephen J. Aguilar ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Rossier School of Education, University of Southern California, Los Angeles, California, United States of America

  • Clare Baek

    Roles Data curation, Writing – review & editing

    Affiliation Rossier School of Education, University of Southern California, Los Angeles, California, United States of America


What factors predict the underreporting of sexual harassment in academe? We used logistic regression and sentiment analysis to examine 2,343 reports of sexual harassment involving members of university communities. Results indicate students were 1.6 times likely to not report their experiences when compared to faculty. Respondents in the life and physical sciences were 1.7 times more likely to not report their experiences when compared to respondents in other disciplines. Men represented 90% of the reported perpetrators of sexual harassment. Analysis of respondents’ written accounts show variation of overall sentiment based on discipline, student type, and the type of institution attended, particularly with regard to mental health. Our results suggest that institutional and departmental barriers driven by power asymmetries play a large role in the underreporting sexual harassment among students—especially those in STEM disciplines.


Sexual harassment in the workplace is pervasive and disproportionally experienced by women. The majority of sexual harassment reports made to the Equal Employment Opportunities Commission (EEOC), for example, have been initiated by women [1] and 60% to 75% of women report having experienced “unwanted sexual attention or sexual coercion” in the workplace [2]. Recent studies have shown similar patterns for college students—50% of female students experienced some kind of sexual harassment during their college years, and women of color have been shown to experience especially high rates of harassment [34]. In the academy, the majority of reported sexual harassment involves unwelcome physical contact (e.g., groping, sexual assault, and domestic abuse-like behaviors) by repeat offenders [5]. Women in the academy are also more likely to be the targets of inappropriate sexual comments from supervisors during fieldwork [6], and are often targeted when they are trainees [7].

Those who experience sexual harassment in post-secondary settings suffer mental, psychological, physical, academic, and work-related consequences—including lower GPA for students and, in the case of faculty, potentially leaving academe altogether [811]. The negative effects of sexual harassment, moreover, can affect an entire scholarly community; recent work has shown that an environment where sexual harassment is pervasive can lead to faculty burnout among women, even if they do not experience sexual harassment directly [12].

Yet, despite the pervasiveness of sexual harassment in academic settings, evidence indicates that reporting rates in academe are generally low, potentially due to fear of retaliation—especially when the perpetrator is a prominent scientist [13]. Such examples signal the role of power imbalances between perpetrator and those whom they victimize. Undergraduate students, for example, report sexual harassment less often when the perpetrator is a faculty member vs. a fellow student [14], and also report sexual harassment less overall compared to their graduate student counterparts [3].

Reporting rates are also influenced by a process that often compels university employees to report student sexual assault disclosures, even when students wish to maintain confidentiality and despite limited evidence that compulsory reporting is beneficial to the students themselves [15]. Such policies may explain recent evidence that shows that students who are knowledgeable about the Title IX reporting process still choose to not report incidents of sexual harassment in hypothetical scenarios [16]. Compulsory reporting, moreover, has been argued to be in conflict with professional ethics codes and practices, such as those of the American Psychological Association [17].

The attitudes of university leadership, a university’s prior response to sexual harassment reports, and prior outcomes for those who reported sexual harassment also play a prominent role in reporting behaviors [1820]. An organizational climate that exhibits tolerance of sexual harassment, for example, may have leaders with dismissive attitudes that undermine formal (or informal) reporting processes. This perceived dismissiveness has been shown to deter victims and whistleblowers from reporting sexual harassment [21]. In contrast, evidence suggests that the presence of a clear no-tolerance policy increases reporting [19].

Given the high rate of sexual harassment in the academy, there is a renewed and vigorous call for action to address—and prevent—it [22]. Faculty accused of sexually harassing their graduate students have been placed on leave pending investigation [2325]. Those found guilty are sometimes stripped of academic honors [22, 26], or recommended for termination [2325]. Funding agencies have also taken steps to address sexual harassment; the National Science Foundation, for example, has established new reporting requirements that will enable it to track sexual harassment committed by its grantees [27]. Despite these positive steps, a report by the National Academies of Science, Engineering, and Medicine found sexual harassment in the academy to be pervasive, and concluded that organizational climate is a potent predictor of it [28].

Addressing and preventing sexual harassment in the academy requires a better understanding of what factors contribute to reporting behaviors so that sound policy can be written to address it. The present study responds to calls for more research in this area (e.g., [17]), by examining predictors of non-reporting behaviors and providing a descriptive analysis of the sentiment used in narrative statements describing sexual harassment experiences gathered through a crowed-sourced survey of sexual harassment in higher education.

Sentiment analysis

Sentiment analysis uses natural language processing to analyze writing that contains opinions, emotions, and attitudes towards topics of interest like products, events, and organizations. The process quantifies the degree of positive, negative, or neutral sentiment contained within the text, and can be applied to varying text segment lengths [2930]. Document-level analysis, for example, classifies a whole document as generally positive, negative or neutral, whereas sentence-level analysis does the same for a sentence [29]. This process helps to reveal the intensity of the writers’ opinions on a topic of interest (e.g., [3033]). Chen, Zhu, Kifer, and Lee [34], for example, used sentiment analysis to compare the opinions of Republican and Democratic senators with respect to a variety of topics, focusing on similarities and dissimilarities between the two groups’ opinions. Other work has found differences between men and women when it comes to the emotional content of work-place email [35]. Sentiment analysis has also been used to investigate the mental health status reflected through written language (e.g., [36]), and has been used with geographic information system data to identify communities in need of mental health services in the wake of natural disasters [37].

Current study

We contribute to studies on sexual harassment (e.g., [19, 16, 3841]) by focusing on reporting behaviors in the academy, and the sentiment present when describing harassment. Specifically, we analyzed unique reports of sexual harassment that occurred in higher education settings across diverse disciplines and institution types. These reports were collected through a crowdsourced survey, where the respondents shared the accounts of perceived sexual harassment through their answers to open-ended questions. We focus on characteristics that predict the reporting behaviors in order to identify potential trends across academic disciplines, and our analysis answers the following research questions:

RQ1: Which respondent, perpetrator, and/or institutional characteristics to predict the non-reporting behavior of sexual harassment?

RQ2: What is the sentiment of respondents’ answers to open-ended questions about sexual harassment experiences?


Survey instrument

This study was approved in accordance with the University of Southern California’s IRB guidelines (UP-18-00179). The data used for our analysis was gathered through an anonymous survey written and disseminated by Dr. Karen Kelsky—a consultant who focuses on helping individuals navigate the Ph.D. job market and advocate for marginalized groups within the academy. Her stated goal for the survey was “…for the academy as a whole to begin to grasp the true scope and scale of [sexual harassment] in academic settings,” [42]. The survey was launched in November 2017 on Google Forms, and shared through social media posts. The survey did not have a formal close date; responses were made publicly available and were downloaded February 2018. The crowdsourced survey captured responses from many individuals (n = 2,343) from a variety of institutions all around the country, and beyond. The survey captures direct and second-hand experience of sexual harassment in the academy at various points in time in a person’s academic career (e.g., undergraduate, graduate, faculty).

Demographic items.

Respondents answered demographic questions that included their status during an incident of sexual harassment (e.g., graduate student); the status of the perpetrators (e.g., tenured professor); the gender of the perpetrator; the type of institution where the harassment occurred (e.g., “R1 institution”); and the respondent’s field of study.


Respondents answered one item that asked whether they reported the incident to their institution, and the institution’s response if the incident was reported.

Open-ended questions.

Respondents also answered four open-ended questions that focused on: 1) what happened, and when it happened—what we term the “harassment narrative”; 2) the impact the harassment had on their career; 3) the impact the harassment had on their mental health; and 4) the impact the harassment had on their choices or life trajectory.


The majority of respondents were from U.S. based institutions. The data captures instances of sexual harassment that may not have been reported to the institution where it occurred, as well as respondents’ reflections on the impact the incident had on various dimensions of their lives. The data reflect 2,343 unique respondents and 9,372 harassment, mental health, career impact, and life trajectory narratives (combined).


The survey did not formally ask respondents to indicate their sex or gender. Many of the respondents, however, used feminine pronouns (e.g., she/her) or otherwise described those who experienced sexual harassment as “women” or “female” in their description of sexual harassment. These responses were coded for the presence of such pronouns in the narrative statement. A second rater coded a sub-sample (n = 434) to calculate agreement; Cohen's κ was run to determine if there was agreement between the two raters’ judgments. There was moderate agreement (73%), κ = .570, p < .001. Given thus, we infer that approximately 43% of respondents reported that women were the targets of the sexual harassment experience. The majority of respondents (70%) were students at various stages of their academic careers during the alleged harassment, (e.g., undergraduate or graduate); 20% were faculty, and 10% were staff (see Table 1 for further breakdown). Respondents came from different fields, including the humanities (47%), the social sciences (26%), the physical sciences (16.9%), and professional schools (7%) (see Table 2).

Reported perpetrators.

The vast majority of reported perpetrators were men (92%), with 73% of them also identified as faculty. Faculty perpetrators had various supervisory roles (e.g., advisors), and 35% of perpetrators were identified as being tenured (see Table 3 for more detailed breakdown of perpetrator statistics).


Harassment took place across various institution types: 25% of respondents reported experiencing sexual harassment at R-1 research institutions, 40% at Ivy or “Elite” institutions, and 33% at other intuitions (see Table 4 in for a full list).

Data preparation

Sentiment scores generation for open-ended responses.

Text data (RQ2) were processed in R using the Text Mining (TM) package [4345]. This process enables quantitative analysis of text, and also removes common stop word (e.g., the, it, a) to facilitate analysis. We then used the “SentimentAnalysis” R package [46] to conduct sentiment analysis on each of open-ended responses of the data set; this process generates a sentiment score between -1 (negative) and 1 (positive) for each of the narratives using existing dictionaries. Sentiment variables were generated for each of the passages below.

Harassment Narrative. Respondents to the survey were asked: “What Happened and When?” The mean word length of responses was 57 (SD = 70).

Impact on Career Narrative. Respondents to the survey were asked: “The Impact of the Harassment on Your Career.” The mean word length of responses was 11 (SD = 15)

Impact on Mental Health Narrative. Respondents to the survey were asked: “The Impact of the Harassment on Your Mental Health.” The mean word length of responses was 11 (SD = 18)

Impact on Career Trajectory Narrative. Respondents to the survey were asked: “The Impact of the Harassment on Your Life Choices/Trajectory.” The mean word length of responses was 10 (SD = 17).

Manually coded variables.

We manually coded discipline and institutional response variables since many of them (e.g., the respondent’s status within the university) were written as free responses in the survey, instead of pre-determined categories that respondents could select from. These variables were manually coded using the process below.

Discipline. Disciplines reported by respondents were aggregated into 6 different categories based on the National Science Foundation’s categories of earned doctorates [47] (see Table 5.)

Table 5. Disciplines reported by respondents, assigned category by research team, and coded value for analysis.

Institutional response. The survey question read: “Institutional Responses to the Harassment (If Any).” Responses were coded into six categories based on existing literature, e.g., retaliation, taking actions to redress the situation, and not taking any action (e.g., [18, 48]). A second rater coded a random-sample of 10% of the responses to calculate agreement; Cohen's κ was run to determine if there was agreement between the two raters’ judgments. There was moderate agreement (64%), κ = .512, p < .001. These coded categories were (see Table 6 for more examples):

  • None: Respondents did not report a formal response by the institution, e.g., “None,” “Nothing has been done,” and “No formal response.”
  • Did not report: Respondents did not report the incident, e.g., “Never reported it,” “I didn’t say anything,” and “I never told anyone.”
  • Action taken: Respondents reported an institutional response after the incident was reported, e.g., “Title IX investigation,” “He was fired,” “He was removed.”
  • Unclear, Not sure: Respondents’ indicated an unclear institutional response, e.g., “Unclear,” “Unknown,” “I think there was an investigation but I never heard any results.”
  • Retaliation (against the respondent):. Respondents reported retaliation against them for reporting the incident, e.g., “I was silenced and punished,” “I was fired,” “Retaliation.”
  • Other: Responses that do not fall into the above categories, e.g., “There were conflicting responses from faculty,” “bathroom renovation over a year after the fact.”

Respondent status. Respondent status at the time of the incident was coded into 16 categories that ranged from graduate student to tenured faculty. These categories were based on the open-ended responses provided on the survey by the respondents (see Table 7 for categories and examples).

Table 7. Respondent status codes at time of reported harassment.

Perpetrator’s status. The reported perpetrator’s status was coded into 8 categories, based on the open-ended responses provided on the survey by the responders. Some perpetrators had multiple roles (ex: a professor AND a dean) or relationship with the responder (e.g., a professor AND an advisor, see Table 3.)

Data analysis

Logistic regression model specification (RQ1).

We used respondent, perpetrator, and institutional characteristics to estimate the likelihood of subgroups’ choice to not report sexual harassment. We specified a logistic regression model (Eq 1, below) with perpetrator characteristics (β1–9), respondent characteristics (β10–15), and institution type (β16). Variables were simultaneously entered into a logistic regression model using Stata 15.


The primary coefficients of interest were β10–15, which represent disciplinary affiliations of respondents. Coefficients of secondary interest were β1–9, which represent perpetrator characteristics (e.g., tenure status). Institutional type was operationalized as a control variable (β16), since the original survey categories do not follow traditional Carnegie Classifications for institutions of higher education [49]. We are, moreover, unaware of extant literature that speaks to a pattern of institutional differences regarding sexual harassment reporting and would thus warrant a different approach.

Sentiment analysis (RQ2).

Our analysis used dictionary-based semantic annotations to assign a sentiment score ranging from -1 (negative) to 1 (positive) to each of the open-ended responses. This approach was the most appropriate, given that we do not know of a collection of text responses that we could use as a reference category to use more sophisticated methods [50]. Different sentiment analysis scores suggest different emotional states and intensity of language used. Descriptive heatmaps were generated by plotting sentiment scores along two dimensions: 1) institution type, and 2) demographic characteristics of respondents. Heatmaps in the results section were generated using R. (See Table 8 for summary statistics of sentiment scores.)


Institutional response

Institutional responses to respondents reporting sexual harassment included: some action taken (9%), no action taken (34%), retaliation (8%), and unclear if action taken (6.5%). Choosing to not report the incident at all was the most common response (36%). This decision varied across disciplines, with roughly 50% of respondents in the physical sciences and engineering choosing to not report the incident. In contrast, only 18% of staff chose to not report the incident (see Table 9).

Logistic regression

Results from logistic regression analysis indicate that students were 1.6 times more likely to not report their experiences when compared with faculty and staff respondents (p < .001). Students in the physical sciences were 1.7 times more likely to not report their experiences when compared to students in other disciplines (p < .05). If the perpetrator was identified as a faculty member, then respondents were 1.5 times more likely to not report the incident compared to respondents who identified the perpetrator as a graduate student, postdoc, or other non-faculty (p < .001). Institution type (e.g., Elite/Ivy, Other R1) was not predictive of respondents choosing to not report sexual harassment (see Table 10 for full logistic regression results).

Descriptive sentiment analysis

Descriptive analysis of the open-ended responses (RQ2) indicated suggests variation in the mental health narratives, especially for students attending elite/Ivy League institutions as compared to other respondents (Fig 1C). A further breakdown analysis indicated that this pattern varied most among undergraduate and graduate students who attended elite/Ivy League institutions (Fig 2C). Sentiment scores of the mental health narrative were more negative for students in the social sciences who attended elite/Ivy institutions when compared to students attending other institution types of institutions, regardless of discipline (Fig 3C).

Fig 1. Sentiment score heatmaps of harassment narrative by student/non-student.

(A), career narrative (B), mental health narrative (C), and life trajectory narrative (D), groped by student and non-student and institution type.

Fig 2. Sentiment score heatmaps of harassment narrative by student status.

(A), career narrative (B), mental health narrative (C), and life trajectory narrative (D), grouped by student type and institution type.

Fig 3. Sentiment score heatmaps of harassment narrative by discipline.

(A), career narrative (B), mental health narrative (C), and life trajectory narrative (D), groped by student type and institution type.


We note that the data that underpin our analysis were taken from a crowdsourced survey that was primarily distributed via social networks (e.g., Facebook), and as such likely suffers from selection bias. Specifically, it is likely that respondents to the survey have a tendency to use social networks regularly and to engage in social networks. Respondents were also likely familiar with Dr. Karen Kelsky’s service and/or somehow connected to one or more of her social network accounts. This suggests that, as a group, they may have a similar set of experiences [51]. This is born out in the data itself; nearly half of the respondents are from Humanities (Table 2) and other members of academia were underrepresented (e.g., administrative staff).

Some respondent characteristics (e.g., race, sexual orientation, gender, age at the time of incident) and key dates (e.g., the year of the incident), moreover, were not captured by the survey. This is unfortunate, since such information would have enabled us to examine how reported experiences differed as a function of individual’s demographic characteristics and/or policies that would have varied over time (e.g., the 1980’s vs. today). Respondents’ decision to report—and the reporting procedures available to them—have likely differed over time as a result of changes to EEO and Title IX policies and interpretations. In addition, retrospection bias (e.g., [5253]), affects the reliability of the respondents’ accounts. As such, we note that we do not make strong claims regarding the intensity of respondents accounts; our logistic regression analysis, for example, is limited to a subset of reporting behaviors (i.e., not reporting). Our covariates were chosen similarly so that we could mitigate against the influence of any retrospection bias.

There are some challenges associated with the use of sentiment analysis as there are with all methods. One of the challenges pertains to the context of the text such that a particular word’s meaning might vary depending on the context (e.g., a word that is positive in one context might be negative in a different context), or a lack of context in texts [54]. It is also difficult to detect subtle and implicit social cues such as humor and sarcasm, as well as inconsistencies in written texts—such as contradictory statements [54]. We emphasize that our sentiment analysis is descriptive, as we did not use scores in subsequent inferential statistical analysis.


Our study supports to the notion that when sexual harassment occurs in the academy, it often does so via power asymmetries, i.e., scenarios which the perpetrator of sexual harassment is in a position of power relative to the person who is sexually harassed. Specifically, student respondents were more likely to avoid reporting sexual harassment. Similarly, despite being underrepresented in our sample, respondents who identified as staff reported more retaliation (Table 9). Previous work has shown that power imbalances can elicit situations associated with high rates of harassment and underreporting [1, 10, 55]. Newins & White [16], for example, found that over a third of students (36%) were unsure whether they would tell a faculty member about sexual assault, and about a fifth (16% to 22% percent) were not willing to disclose at all.

The variability of underreporting behaviors across disciplines also suggests differences in disciplinary culture—the fact that only 50% of respondents in STEM fields chose to report their incident suggests that there are potential barriers to doing so, and may also suggest a process that makes reporting undesirable, a point underscored by the #MeTooSTEM movement and previous work [22, 56]. Representation likely plays a role, as decades of work shows that women are underrepresented in STEM (e.g., [5760]). Low representation of women, in addition to higher rates of experiencing sexual harassment, may thus create a difficult environment for women in STEM.

Sentiment analysis of written accounts helps to uncover additional information within a data set ([3637]), and our descriptive analysis suggests institutional differences among the mental health narratives (Figs 13). These differences were most apparent when examining sentiment scores of students who attended social science programs within elite/Ivy League institutions (Fig 3). We infer from this analysis that one must also attend to institutional differences when investigating sexual harassment and its impact on the academic community. Such studies should attend to differences in implementing and reinforcing policies regarding sexual harassment [48, 6163]. However, our study cannot definitely state that the apparent institutional differences in respondents’ sentiments regarding the mental impact of sexual harassment were due to true institutional differences or were in fact an artifact of how data were collected. Regardless, it is imperative for each institution to provide pathways that can best support those impacted by sexual harassment and provide necessary services, such as counseling. Compulsory reporting practices, if in place, should also be examined so that the well-being of those who report is protected.

It is important to address sexual harassment in the academy; those who choose to sexually harass their students and/or colleagues are responsible for a host of negative outcomes among those they victimize. Chronic sexual harassment has been shown to be predictive of anxiety, depression, and substance abuse; negative health outcomes include hypertension and poor sleep [6465]. Sexual harassment is also associated with psychological duress and lower academic satisfaction [66]. Men are not immune from sexual harassment in academia, but experience it at lower rates compared to women [4].

Recent work has highlighted the fact that women feel more supported and/or empowered to speak of when they experience sexual harassment [67]. Notably, one of the themes that emerged in the Keplinger et al. [67] indicates that the #MeToo movement has helped many feel comfortable with sharing sexual harassment experiences. Troublingly, however, while sexual harassment has decreased in the wake of the #MeToo movement, gender harassment (defined as negative views of women and gender hostility) has increased [67]. Universities should thus continue to pay close attention to gender harassment to ensure that it does not increase as sexual harassment decreases. Future work in this area might also examine differences between the reported sexual harassment experiences of women before and after the #MeToo movement. It is also crucial for future work to examine the reported sexual harassment experiences of populations may be particularly vulnerable to sexual harassment, such as LGBTQ community, people with disabilities, and people of color. Such work could reveal whether the institutions have made effective and tangible changes to the ways they handle sexual harassment. Taking such steps will help to ensure that the next generation of scientists are trained in an environment where safety and respect are the norm.


First and foremost, we thank the thousands of women who were brave enough to share (and potentially relive) their experiences. We also thank K. Kelsky for collecting these stories, and important outcome measures that made our analyses possible.


  1. 1. Shaw E, Hegewisch A, Hess C. Sexual harassment and assault at work: Understanding the costs. Institute for Women’s Policy Research Publication, IWPR B. 2018;376.
  2. 2. Feldblum CR, Lipnic VA. Select task force on the study of harassment in the workplace. Washington: US Equal Employment Opportunity Commission. 2016 Jun.
  3. 3. ‌Thakur MB, Paul P. Sexual Harassment in Academic Institutions: A Conceptual Review. Journal of Psychosocial Research. 2017 Jan 1;12(1).
  4. 4. Witze A Sexual harassment is rife in the sciences, finds landmark US study. Nature. 2018 Jun 1;558(7710):352–4. pmid:29921850
  5. 5. Cantalupo NC, Kidder WC. A systematic look at a serial problem: Sexual harassment of students by university faculty. Utah L. Rev. 2018:671.
  6. 6. Bohannon J. Survey of peers in fieldwork highlights an unspoken risk. Science. 2013; 340(6130), 265–265. pmid:23599455
  7. 7. Clancy KB, Nelson RG, Rutherford JN, Hinde K. Survey of academic field experiences (SAFE): Trainees report harassment and assault. PloS one. 2014 Jul 16;9(7).
  8. 8. Jordan CE, Combs JL, Smith GT. An exploration of sexual victimization and academic performance among college women. Trauma, Violence, & Abuse. 2014 Jul;15(3):191–200.
  9. 9. Mengo C, Black BM. Violence victimization on a college campus: Impact on GPA and school dropout. Journal of College Student Retention: Research, Theory & Practice. 2016 Aug;18(2):234–48.
  10. 10. St. John K, Riggs E, Mogk g. Sexual harassment in the sciences: a call to geoscience faculty and researchers to respond. Journal of Geoscience Education. 2016 Nov 2;64(4):255–7.
  11. 11. Yoon E, Stiller Funk R, Kropf NP. Sexual harassment experiences and their psychological correlates among a diverse sample of college women. Affilia. 2010 Feb;25(1):8–18.
  12. 12. Takeuchi M, Nomura K, Horie S, Okinaga H, Perumalswami CR, Jagsi R. Direct and indirect harassment experiences and burnout among academic faculty in Japan. The Tohoku journal of experimental medicine. 2018;245(1):37–44. pmid:29760353
  13. 13. Clery D. Shining a light on sexual harassment in astronomy. Science, 2015;350(6259), 364–365. pmid:26494736
  14. 14. Shepela ST, Levesque LL. Poisoned waters: Sexual harassment and the college climate. Sex roles. 1998 Apr 1;38(7–8):589–611.
  15. 15. Holland KJ, Cortina LM, Freyd JJ. Compelled disclosure of college sexual assault. American Psychologist. 2018 Apr;73(3):256. pmid:29355356
  16. 16. Newins AR, White SW. IX sexual violence reporting requirements: knowledge and opinions of responsible employees and students. Journal of aggression, conflict and peace research. 2018 Apr 9.
  17. 17. Newins AR. Ethical considerations of compelled disclosure of sexual assault among college students: Comment on Holland, Cortina, and Freyd (aaaaaaaaaaa).
  18. 18. Bergman ME, Langhout RD, Palmieri PA, Cortina LM, Fitzgerald LF. The (un) reasonableness of reporting: Antecedents and consequences of reporting sexual harassment. Journal of Applied Psychology. 2002 Apr;87(2):230. pmid:12002952
  19. 19. Jacobson RK, Eaton AA. How organizational policies influence bystander likelihood of reporting moderate and severe sexual harassment at work. Employee Responsibilities and Rights Journal. 2018 Mar 1;30(1):37–62.
  20. 20. Knapp DE, Faley RH, Ekeberg SE, Dubois CL. Determinants of target responses to sexual harassment: A conceptual framework. Academy of Management Review. 1997 Jul 1;22(3):687–729.
  21. 21. Clarke HM. Predicting the Decision to Report Sexual Harassment: Organizational Influences and the Theory of Planned Behavior. Journal of Organizational Psychology. 2014 Nov 1;14(2).
  22. 22. Clark-Parsons R. # ThisEndsHere: Confronting Sexual Assault and Harassment in Our Campus Communities. Communication Culture & Critique. 2018 Sep 1;11(3):475–8.
  23. 23. Mangan, K. Professor Who Complained of Vigilante Justice Is Found Responsible for Harassment. The Chronicle of Higher Education. [Cited 2019 March 4]. Available from:
  24. 24. Thomason, A. 2 Sexual-Harassment Cases Draw Renewed Attention to Gender Bias in Astronomy. The Chronicle of Higher Education. [Cited 2019, March 4]. Available from:
  25. 25. Thomason, A. Boulder Philosophy Professor Accused of Sexual Harassment Retires. The Chronicle of Higher Education. [Cited 2019 March 4]. Available from:
  26. 26. Wadman M. National Academy of Sciences will vote on ejecting sexual harassers. Science. 2019.
  27. 27. Mervis J., Kaiser J. NSF issues sexual harassment policy as NIH promises action. Science. 2018;361(6409), 1299–1300. pmid:30262477
  28. 28. National Academies of Sciences, Engineering, and Medicine. Sexual Harassment of Women: Climate, Culture, and Consequences in Academic Sciences, Engineering, and Medicine. Washington, DC: The National Academies Press. 2018. pmid:29894119
  29. 29. Liu B. Sentiment analysis and opinion mining. Synthesis lectures on human language technologies. 2012 May 22;5(1):1–67.
  30. 30. Yadollahi A, Shahraki AG, Zaiane OR. Current state of text sentiment analysis from opinion to emotion mining. ACM Computing Surveys (CSUR). 2017 May 25;50(2):1–33.
  31. 31. Cambria E. Affective computing and sentiment analysis. IEEE Intelligent Systems. 2016 Mar 17;31(2):102–7.
  32. 32. Liang Y, Zheng X, Zeng DD. A survey on big data-driven digital phenotyping of mental health. Information Fusion. 2019 Dec 1;52:290–307.
  33. 33. Sharma A, Dey S. An artificial neural network based approach for sentiment analysis of opinionated text. In Proceedings of the 2012 ACM Research in Applied Computation Symposium 2012 Oct 23 (pp. 37–42).
  34. 34. Chen B, Zhu L, Kifer D, Lee D. What is an opinion about? exploring political standpoints using opinion scoring model. InTwenty-Fourth AAAI Conference on Artificial Intelligence 2010 Jul 4.
  35. 35. Mohammad SM, Yang TW. Tracking sentiment in mail: How genders differ on emotional axes. InProceedings of the 2nd workshop on computational approaches to subjectivity and sentiment analysis 2011 Jun 24 (pp. 70–79). Association for Computational Linguistics.
  36. 36. Spinczyk D, Nabrdalik K, Rojewska K. Computer aided sentiment analysis of anorexia nervosa patients’ vocabulary. Biomedical engineering online. 2018 Dec 1;17(1):19. pmid:29394945
  37. 37. Gruebner O, Lowe SR, Sykora M, Shankardass K, Subramanian SV, Galea S. A novel surveillance approach for disaster mental health. PLoS one. 2017;12(7).
  38. 38. Adams JW, Kottke JL, Padgitt JS. Sexual harassment of university students. Journal of College Student Personnel. 1983 Nov.
  39. 39. Carr PL, Ash AS, Friedman RH, Szalacha L, Barnett RC, Palepu A, Moskowitz MM. Faculty perceptions of gender discrimination and sexual harassment in academic medicine. Annals of internal medicine. 2000 Jun 6;132(11):889–96. pmid:10836916
  40. 40. Finn J. A survey of online harassment at a university campus. Journal of Interpersonal violence. 2004 Apr;19(4):468–83. pmid:15038885
  41. 41. Reilly ME, Lott B, Gallogly SM. Sexual harassment of university students. Sex Roles. 1986 Oct 1;15(7–8):333–58.
  42. 42. Kelsky, K. A Crowdsourced Survey of Sexual Harassment in the Academy. The Professor Is In. [Cited 2019 February 25]. Available from
  43. 43. Feinerer I. An introduction to text mining in R. The Newsletter of the R Project Volume 8/2, October 2008. 2008 Oct;8:19.
  44. 44. Feinerer I. Introduction to the tm Package Text Mining in R. Accessible en ligne: 2013 Jan 28.
  45. 45. Meyer D, Hornik K, Feinerer I. Text mining infrastructure in R. Journal of statistical software. 2008 Mar 31;25(5):1–54.
  46. 46. Feuerriegel S, Proellochs N. SentimentAnalysis: Dictionary-based sentiment analysis. R package version 1.3–2.
  47. 47. Survey of Earned Doctorates. NCSES | NSF(2017). [Cited October 2, 2019]. Available from: website:
  48. 48. Ellman-Golan E. Saving Title IX: Designing more equitable and efficient investigation procedures. Mich. L. Rev. 2017;116:155.
  49. 49. Carnegie Classifications. Carnegie Classifications, Basic Classification. [Cited 2019 April 11]. Available from:
  50. 50. Grimmer J, Stewart BM. Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political analysis. 2013;21(3):267–97.
  51. 51. Dykstra-DeVette TA, Tarin C. Isolating Structures of Sexual Harassment in Crowdsourced Data on Higher Education. Women's Studies in Communication. 2019 Jul 3;42(3):371–93.
  52. 52. Levine LJ, Lench HC, Karnaze MM, Carlson SJ. Bias in predicted and remembered emotion. Current opinion in behavioral sciences. 2018 Feb 1;19:73–7.
  53. 53. Pudney S. Perception and retrospection: The dynamic consistency of responses to survey questions on wellbeing. Journal of Public Economics. 2011 Apr 1;95(3–4):300–10.
  54. 54. Vinodhini G, Chandrasekaran RM. Sentiment analysis and opinion mining: a survey. International Journal. 2012 Jun;2(6):282–92.
  55. 55. McLaughlin H, Uggen C, Blackstone A. Sexual harassment, workplace authority, and the paradox of power. American sociological review. 2012 Aug;77(4):625–47. pmid:23329855
  56. 56. Wadman M. This neuroscientist is fighting sexual harassment in science—but her own job is in peril. Science. [Cited 2019, March 6]
  57. 57. Armstrong MA, Jovanovic J. Starting at the crossroads: Intersectional approaches to institutionally supporting underrepresented minority women STEM faculty. Journal of Women and Minorities in Science and Engineering. 2015;21(2).
  58. 58. Ellis J, Fosdick BK, Rasmussen C. Women 1.5 times more likely to leave STEM pipeline after calculus compared to men: Lack of mathematical confidence a potential culprit. PloS one. 2016 Jul 13;11(7).
  59. 59. Nimmesgern H. Why are women underrepresented in STEM fields?. Chemistry–A European Journal. 2016 Mar 7;22(11):3529–30.
  60. 60. Noonan, R. Women in STEM: 2017 Update. ESA Issue Brief# 06-17US Department of Commerce. 2017.
  61. 61. Fusilier M, Penrod C. University employee sexual harassment policies. Employee Responsibilities and Rights Journal. 2015 Mar 1;27(1):47–60.
  62. 62. Hobson CJ, Guziewicz J. Sexual harassment preventive/protective practices at US colleges and universities. College Student Affairs Journal. 2002 Apr 1;21(2):17.
  63. 63. Streng TK, Kamimura A. Sexual Assault Prevention and Reporting on College Campuses in the US: A Review of Policies and Recommendations. Journal of Education and Practice. 2015;6(3):65–71.
  64. 64. McGinley M, Wolff JM, Rospenda KM, Liu L, Richman JA. Risk factors and outcomes of chronic sexual harassment during the transition to college: Examination of a two-part growth mixture model. Social science research. 2016 Nov 1;60:297–310. pmid:27712687
  65. 65. Thurston RC, Chang Y, Matthews KA, von Känel R, Koenen K. Association of sexual harassment and sexual assault with midlife women’s mental and physical health. JAMA internal medicine. 2019 Jan 1;179(1):48–53. pmid:30285071
  66. 66. Huerta M, Cortina LM, Pang JS, Torges CM, Magley VJ. Sex and power in the academy: Modeling sexual harassment in the lives of college women. Personality and Social Psychology Bulletin. 2006 May;32(5):616–28. pmid:16702155
  67. 67. Keplinger K, Johnson SK, Kirk JF, Barnes LY. Women at work: Changes in sexual harassment between September 2016 and September 2018. PloS one. 2019;14(7).