Team Learning and Psychological Safety Survey

1. Abstract

The Team Learning and Psychological Safety Survey, developed by Amy C. Edmondson in her seminal 1999 investigation published in Administrative Science Quarterly, represents an indispensable psychometric framework designed to measure interpersonal climate and collective learning dynamics in organizational work teams. The assessment operationalizes four distinct, interdependent constructs: Team Psychological Safety (the shared belief held by team members that the team is safe for interpersonal risk-taking), Internal Team Learning Behaviors (deliberate actions aimed at reflecting, monitoring, and experimenting within the immediate team boundary), External Team Learning Behaviors (cross-boundary information scanning, stakeholder engagement, and feedback acquisition), and Team Learning Outcomes (perceived collective skill acquisition, knowledge transfer, and process innovation). Comprising 23 to 24 items scored on a 7-point Likert-type scale ranging from 1 (“Strongly Disagree”) to 7 (“Strongly Agree”), the instrument employs a multi-level referent-shift model where items reference the collective unit rather than idiosyncratic individual experiences.

Psychometrically, Edmondson’s battery displays robust structural integrity, consistently exhibiting high internal consistency across subscales with Cronbach’s alpha coefficients commonly exceeding .75 to .85 across industrial, healthcare, educational, and technological sectors. Confirmatory factor analyses support a multidimensional architecture, while within-group agreement indices ($r_{wg}$) and intraclass correlation coefficients (ICC[1] and ICC[2]) justify the statistical aggregation of individual responses to the group level of analysis. By delineating psychological safety not as a general climate of unconditional warmth or low accountability, but rather as an enabling affective-cognitive state that mitigates the interpersonal risks of speaking up, asking for help, and learning from mistakes, this scale has catalyzed foundational research in organizational learning, team reflexivity, and high-reliability organizing.

2. Keywords

Team psychological safety, team learning behavior, organizational learning, Amy Edmondson, interpersonal risk, boundary spanning, single-loop learning, double-loop learning, team reflexivity, psychometrics, error management, collective climate.

3. Authors

The Team Learning and Psychological Safety Survey was developed and validated by:

• Amy C. Edmondson, Ph.D.: Novartis Professor of Leadership and Management at Harvard Business School (Boston, Massachusetts, USA). Edmondson’s research focuses on psychological safety, organizational learning, teaming, and leadership within complex, fast-changing environments. Her pioneering field research across manufacturing plants, hospital intensive care units, and executive suites established the primary conceptual and operational metrics of team-level psychological safety used globally.

4. Purpose

In modern knowledge economies, organizations rely heavily on cross-functional, multidisciplinary teams to solve ambiguous problems, innovate rapidly, and maintain operational reliability. However, human beings possess a pervasive social vulnerability: an innate reluctance to engage in behaviors that could jeopardize their perceived competence, status, or belonging within their peer group. Initiating questions, acknowledging personal errors, offering dissenting opinions, or seeking clarification carries perceived interpersonal costs—namely, looking ignorant, incompetent, intrusive, or disruptive. Edmondson conceived the Team Learning and Psychological Safety Survey to illuminate and quantify these tacit psychological dynamics.

The primary purpose of the instrument is twofold: to diagnose the interpersonal conditions that suppress or facilitate risk-taking, and to capture the actual behavioral manifestations of learning in intact work groups. Many historical models of organizational learning treated learning abstractly at the institutional level (e.g., through knowledge repositories and standard operating procedures) or reduced it strictly to individual cognition. Edmondson bridges this divide by focusing on the team as the primary learning engine of the modern organization. Teams operate as micro-climates; even within the same corporate culture, two teams in adjacent hallways can possess vastly different social contracts regarding error admission and candid discourse.

In research contexts, the scale serves as a foundational assessment battery for examining the antecedents, mediators, and moderators of collective performance, continuous improvement, and innovation. It has been extensively deployed in studies exploring leader behavior, transformational leadership, diversity, psychological capital, and high-reliability systems. In clinical and applied organizational development settings, the tool functions as a predictive diagnostic. By measuring psychological safety alongside internal and external learning behaviors, organizational consultants and team leaders can identify whether suboptimal team performance stems from psychological fear and self-censorship, operational boundary insularity, or ineffective process reflection.

5. Psychological Construct

The survey operationalizes team dynamics through four interrelated constructs that capture psychological climate, intra-team behaviors, inter-team behaviors, and subjective outcomes:

Team Psychological Safety

Psychological safety is conceptualized as a shared belief held by members of a team that the team is safe for interpersonal risk-taking. Crucially, it is not an individual personality trait (such as extraversion or neuroticism), nor is it equivalent to interpersonal trust or warmth. While trust focuses on the cognitive expectation of an individual about another person’s reliability or goodwill, psychological safety describes a collective, tacit climate regarding the likely interpersonal consequences of taking risks in the group. In a psychologically safe team, individuals anticipate that if they admit a blunder, pose an unconventional hypothesis, or identify a procedural flaw, their peers will not mock, humiliate, marginalize, or punish them. Examples include feeling confident that asking a basic question will not be seen as a sign of incompetence and believing that constructive criticism will be embraced rather than penalized.

Internal Team Learning Behaviors (TLBI)

Internal team learning behaviors encompass the observable, iterative processes through which team members actively examine assumptions, test operational procedures, monitor progress against milestones, and correct deviations within their unit. This construct integrates both single-loop learning (detecting errors and adjusting behaviors within an existing framework of rules and norms) and double-loop learning (questioning underlying assumptions, systemic paradigms, and values that guide decision-making). Specific behavioral indicators include holding candid post-mortem discussions, openly examining errors to uncover root causes, scheduling time to systematically refine work workflows, and welcoming internal dissent during project planning.

External Team Learning Behaviors (TLBE)

Teams rarely function in complete isolation; their success hinges on navigating interdependent organizational ecosystems. External team learning behaviors—frequently described in the organizational sociology literature as boundary spanning—represent the intentional efforts of team members to solicit feedback, acquire information, benchmark best practices, and coordinate with stakeholders outside the immediate team boundary. This involves scanning the broader organization for strategic changes, inviting external subject-matter experts to review team outputs, gathering customer requirements, and ensuring institutional buy-in from lateral departments and upper management.

Team Learning Outcomes (TLO)

Team learning outcomes capture the self-reported developmental dividends accrued by the team and its individual members as a consequence of their collective workflow. Unlike objective performance metrics (e.g., project completion dates, defect rates), TLO measures the degree to which working within the group has enriched members’ specialized cross-functional expertise, expanded their cognitive repertoires, and yielded innovative intellectual property, such as new protocols, products, or methodologies.

6. Theoretical Framework

The Team Learning and Psychological Safety Survey synthesizes multiple paradigms from social psychology, organizational behavior, and behavioral economics:

Theories of Action and Defensive Routines (Argyris & Schön)

A central pillar of Edmondson’s model is Chris Argyris and Donald Schön’s (1978) conceptualization of organizational defensive routines. Argyris and Schön posited that human beings frequently operate under “Model I” theories-in-use: cognitive mindsets aimed at maximizing unilateral control, minimizing negative affect, and appearing competent at all costs. When faced with errors or threats, individuals instinctively suppress unpleasant facts and engage in bypass strategies, resulting in systemic error accumulation and superficial single-loop corrections. To achieve generative double-loop learning, organizations must construct environments where individuals can safely challenge assumptions without triggering defensive retaliation. Edmondson operationalizes this dynamic, showing that team psychological safety is the social-structural antidote to Model I defensive routines.

Psychological Safety in Human Change (Schein & Bennis)

The term “psychological safety” was originally articulated by Edgar Schein and Warren Bennis (1965) in their study of organizational change. Schein asserted that transformative organizational change inherently induces “learning anxiety”—the dread of appearing inadequate or losing personal identity while adopting new patterns of behavior. For individuals to overcome learning anxiety, leaders must establish psychological safety, which lowers the perceived risk of change without reducing the standard of accountability. Edmondson transplanted this concept from macro-level change consulting into the micro-level dynamics of task-performing teams.

Personal Engagement Framework (Kahn)

William Kahn’s (1990) qualitative examination of personal engagement and disengagement at work provided critical micro-foundations for the construct. Kahn defined psychological safety as “feeling able to show and employ one’s self without fear of negative consequences to self-image, status or career.” Edmondson adapted Kahn’s individual-level perception into a shared, emergent property of the team unit, arguing that when individuals interact within a stable collective, their shared experiences crystallize into a common normative assessment of interpersonal safety.

Multi-Level Climate Theory (Kozlowski & Klein)

The architectural logic of the survey relies on multi-level theory (Kozlowski & Klein, 2000). The instrument employs a “referent-shift consensus model” (Chan, 1998). Instead of asking individual team members about their personal traits (e.g., “I feel comfortable taking risks”), items are systematically framed around the collective unit (e.g., “It is completely safe to take a risk on this team”). This conceptual framing ensures that when responses are aggregated, the resulting score represents a legitimate group-level construct (team climate) rather than a mere aggregate of individual self-efficacy.

7. Validity

Empirical evaluations of the Team Learning and Psychological Safety Survey have established strong construct, convergent, discriminant, and criterion-related validity across diverse organizational settings:

Construct and Convergent Validity

In her foundational 1999 study conducted across 51 work teams in an office design and manufacturing company, Edmondson established construct validity through rigorous multi-method cross-validation. Survey-derived measures of psychological safety converged closely with qualitative behavioral observations collected via longitudinal ethnography. Psychometric subscales displayed high convergent validity with related organizational constructs: team psychological safety correlated positively with team efficacy ($r = .35, p < .05$) and leader coaching behaviors ($r = .55, p < .001$), demonstrating that supportive leadership practices directly foster safe interpersonal climates.

Discriminant Validity

Discriminant validity is critical to demonstrate that psychological safety is not merely a rebadged version of general team cohesion, groupthink, or positive affect. Edmondson demonstrated that team psychological safety and team efficacy, while positively related, occupy distinct cognitive domains. Team efficacy reflects a shared belief in the team’s technical capacity to execute task demands successfully (“can we do it?”), whereas psychological safety reflects the climate regarding interpersonal vulnerability (“will we be penalized for asking or stumbling?”). Factor analytic models reliably distinguish psychological safety items from measures of group cohesion, affective commitment, and perceived organizational support.

Predictive and Criterion-Related Validity

The instrument displays exceptional predictive validity for performance outcomes. In Edmondson’s original validation sample, team psychological safety accounted for significant incremental variance in internal team learning behaviors ($R^2 = .49, p < .001$). In turn, team learning behavior mediated the relationship between psychological safety and team performance (measured both through external manager ratings and objective production metrics). Subsequent cross-industry studies, including Google’s widely cited internal research initiative (Project Aristotle), reaffirmed that psychological safety, measured via adaptations of Edmondson’s scale, serves as the single strongest predictor of high-performing, innovative teams, outperforming individual academic credentials, personality blends, or tenure distribution.

8. Reliability

The Team Learning and Psychological Safety Survey exhibits robust internal consistency and psychometric reliability across various international and occupational populations:

Internal Consistency (Cronbach’s Alpha)

Across empirical studies, the internal reliability coefficients consistently exceed accepted psychometric thresholds ($lpha ge .70$):

• Team Psychological Safety: In Edmondson’s 1999 sample, the psychological safety subscale demonstrated a Cronbach’s alpha of $lpha = .82$. Subsequent cross-sectional and longitudinal studies across healthcare, software engineering, and aviation have mirrored these findings, reporting coefficients typically ranging from $.80$ to $.88$.
• Internal Team Learning Behaviors (TLBI): The internal learning behavior subscale exhibited a Cronbach’s alpha of $lpha = .78$ in the original validation study, with subsequent replications yielding alphas between $.76$ and $.84$.
• External Team Learning Behaviors (TLBE): Edmondson reported an internal consistency of $lpha = .80$ for the external boundary-spanning subscale.
• Team Learning Outcomes (TLO): The 3-item outcome subscale achieved an alpha of $lpha = .75$ to $.78$.

Aggregation Statistics and Within-Group Agreement

Because the survey assesses group-level phenomena, psychometricians must verify that members within the same team agree with one another significantly more than they agree with members of different teams. Edmondson established group-level reliability using standard multi-level psychometric tests:

• Within-Group Agreement ($r_{wg}$): For team psychological safety, the median $r_{wg}$ in Edmondson’s study was $.88$, comfortably exceeding the conventional $.70$ benchmark and verifying strong perceptual convergence among team peers.
• Intraclass Correlation Coefficients: The ICC(1)—which estimates the proportion of variance in individual ratings attributable to team membership—yielded values ranging from $.20$ to $.35$ ($p < .001$), demonstrating substantial group-level effects. The ICC(2)—which measures the reliability of the team means—routinely exceeds $.70$, establishing statistical justification for aggregating individual Likert ratings to team-level mean scores.

9. Factor Analysis

The factorial validity of the Team Learning and Psychological Safety Survey has been repeatedly confirmed via both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA):

Confirmatory Factor Modeling

In structural evaluations, a multi-factor nested model exhibits superior fit compared to single-factor or two-factor alternatives. CFA models testing the four distinct dimensions—Psychological Safety, Internal Learning Behaviors, External Learning Behaviors, and Team Learning Outcomes—consistently demonstrate adequate goodness-of-fit indices:

• Comparative Fit Index (CFI): Values typically range from $.92$ to $.96$, exceeding the accepted standard of $.90$.
• Tucker-Lewis Index (TLI): Coefficients routinely range between $.91$ and $.95$.
• Root Mean Square Error of Approximation (RMSEA): Estimates consistently fall between $.045$ and $.068$, well below the $.08$ cutoff for acceptable model fit.
• Standardized Root Mean Square Residual (SRMR): Values generally remain below $.060$.

Factor Loadings and Structural Distinctiveness

Standardized factor loadings across items consistently display substantial magnitude, with the vast majority of items loading onto their designated latent constructs at $lambda ge .55$, and frequently exceeding $lambda = .70$. The reverse-scored items within the psychological safety scale (e.g., “When someone makes a mistake in this team, it is often held against him or her”) occasionally demonstrate slightly lower factor loadings due to method effects common in negatively phrased psychometric items, yet they remain statistically robust and contribute meaningfully to attenuating acquiescence response bias.

10. Instrument / Measurement Tool

The Team Learning and Psychological Safety Survey is structured as a self-report diagnostic battery optimized for intact work groups. The operational characteristics of the measurement tool are summarized below:

• Instrument Type: Multi-dimensional self-report survey using a referent-shift consensus model.
• Target Population: Members of defined, interdependent work teams across corporate, clinical, technical, and educational environments.
• Total Item Count: 23 items (distributed across four specific subscales: 6 items for Psychological Safety, 8 items for Internal Team Learning Behaviors, 6 items for External Team Learning Behaviors, and 3 items for Team Learning Outcomes).
• Response Format: 7-point Likert scale (1 = Strongly Disagree, 2 = Disagree, 3 = Slightly Disagree, 4 = Neither Agree Nor Disagree, 5 = Slightly Agree, 6 = Agree, 7 = Strongly Agree).
• Reverse-Scored Items: Several items contain negative phrasing to control for acquiescence response bias (marked with “(R)” in the scale items). These items must be linearly transformed prior to calculating composite scores ($Score_{reversed} = 8 – Score_{raw}$).
• Scoring and Aggregation Protocol:
  • Individual Subscale Scores: Computed by calculating the arithmetic mean of all items belonging to a given subscale for an individual respondent following reverse-scoring adjustments.
  • Team Subscale Scores: Computed by aggregating individual subscale averages to generate a group-level mean. Aggregation should strictly be conducted only after verifying within-group agreement ($r_{wg} ge .70$) and significant between-group variance ($F$-test in one-way ANOVA; ICC[1] > 0).

11. Permissions & Fee and Test Year

The Team Learning and Psychological Safety Survey was published by Amy C. Edmondson in 1999 within the peer-reviewed scholarly journal Administrative Science Quarterly. Under academic fair-use guidelines, the item inventory is publicly accessible in the academic literature for non-commercial scholarly research, educational purposes, and scientific replication studies without requiring financial licensing fees, provided formal academic attribution is preserved.

Commercial applications, software implementations, and proprietary corporate diagnostic deployments may require formal permission or licensing arrangements through the author, the original publisher (SAGE Publications), or affiliated commercial intellectual property representatives. Researchers and organizational practitioners planning extensive organizational interventions are advised to consult the author’s institutional office at Harvard Business School for authorized guidelines and benchmarking datasets.

12. References

• Argyris, C., & Schön, D. A. (1978). Organizational learning: A theory of action perspective. Addison-Wesley.
• Chan, D. (1998). Functional relations among constructs in the same content domain at different levels of analysis: A typology of composition models. Journal of Applied Psychology, 83(2), 234–246. https://doi.org/10.1037/0021-9010.83.2.234
• Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383. https://doi.org/10.2307/2666999
• Edmondson, A. C. (2018). The fearless organization: Creating psychological safety in the workplace for learning, innovation, and growth. John Wiley & Sons.
• Kahn, W. A. (1990). Psychological conditions of personal engagement and disengagement at work. Academy of Management Journal, 33(4), 692–724. https://doi.org/10.5465/256287
• Kozlowski, S. W., & Ilgen, D. R. (2006). Enhancing the effectiveness of work groups and teams. Psychological Science in the Public Interest, 7(3), 77–124. https://doi.org/10.1111/j.1529-1006.2006.00030.x
• Kozlowski, S. W., & Klein, K. J. (2000). A multilevel approach to theory and research in organizations: Contextual, temporal, and emergence processes. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations (pp. 3–90). Jossey-Bass.
• Schein, E. H., & Bennis, W. G. (1965). Personal and organizational change through group methods: The laboratory approach. John Wiley & Sons.

13. Items of the Scale