This study is concerned with the construct, convergent and divergent validity of a new short measure of Psychological Safety (PS) at work. This area has attracted a great deal of research ( Amoadu et al., 2024 ; Edmondson & Mogelof, 2005 ; O’Donovan et al., 2020 ; Sasaki et al., 2022 ). While the theory can be found in the writings of Carl Rogers and management writers like Kahn (1990) , the concept of PS is usually attributed to Edmondson (1999) .

The concept has been defined as the “belief that one will not be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes”. It is both an individual and group-shared concept, held by members of a team, that they are safe within the group to conduct many types of interpersonal risk-taking activities. Thus, it can be measured at an individual, group/team, and organizational level ( Newman et al., 2017 ). Further, it is suggested that it is state—rather than trait-like, in that it may vary considerably with work circumstances, though it is conceivable that it has consistent individual difference correlates. That is, some people may seem more sensitive to, and concerned with, safety issues. Perhaps the most important claim in the literature is that PS has a direct relationship with work output measures, like satisfaction and productivity ( Kim et al., 2020 ); that is, PS is an important factor in understanding business success.

Over ten years ago, Edmondson and Lei (2014) noted that most of the research has taken place in specific industry contexts, most notably in healthcare delivery. Further, the “coherence and amount of the cumulative research suggests that creating a climate of psychological safety should be near the top of the leadership agenda for organizations around the world” (p. 72). There was, however, little mention of the measurement of the concept.

The concept of PS is made up of facets or factors, though there is no clear definition of what they are. According to Plouffe et al. (2023) , PS has various components: the confidence to share knowledge and request feedback without fear of negative (interpersonal risk‐taking); mutual trust and respect between team members; the availability of organizational and structural supports, which renders them less vulnerable to resource depletion, insecurity, and defensiveness; understanding personal identity, importance, and security within the team. Indeed, these are the factors in the Psychological Safety Inventory (PSI), one of the few measures in the area.

In a review, Newman et al. (2017) noted:

In the majority of studies we reviewed, psychological safety was the mechanism through which the effects of these supportive environments were transmitted to desirable outcomes, such as increased knowledge sharing, engagement, creativity, innovation, and ultimately performance…psychological safety is a valuable resource, especially important in hazardous work contexts where speaking up and providing feedback is imperative in order to reduce errors and improve safety. (p. 530)

Research on this topic has appeared in a number of papers this millennium ( Bienefeld & Grote, 2014 ; Carmeli, 2007 ; Chandrasekaran & Mishra, 2012 ; Ferrère et al., 2022 ; Nembhard & Edmondson, 2006, 2011 ; O’Donovan & McAuliffe, 2020a, 2020b ; Roussin et al., 2016 ). There are probably now in excess of 100 published papers looking very specifically at the topic. Indeed, there have been a number of reviews of the literature and meta-analyses of the literature ( Dong et al., 2024 ; Dong et al., 2025 ; Edmondson & Bransby, 2023 ; Frazier et al., 2017 ).

It is often the case that multiple instruments are devised to measure the same psychological concept. There are a number of PS measures ( Brown & Leigh, 1996 ; Tynan, 2005 ). A quarter of a century ago, Edmondson (1999) developed the seven-item Psychological Safety Scale (PSS), which evaluates team‐level psychological safety as a unidimensional construct reflecting interpersonal risk‐taking.

In this study, we first look at the concurrent validity of a new, short measure of PS, which we compare to the PSI. This measure was devised and validated over a long period of time. There are longer versions of the one we used here, but given the fact that both researchers and work clients favour shorter measures, we chose to examine the shorter one ( Allen et al., 2022 ).

We also examine the discriminant validity by assessing the relationship of PS with four sets of variables. First, we explore personality, as it is likely that safety perceptions are linked to traits. Thus, we attempt to understand the place of PS within Five Factor personality. Based on an understanding of the five personality factor models, we predict PS is positively correlated with Extraversion (H1) because of the social confidence and skills associated with extraversion, and Agreeableness (H2), because of the trusting nature of high scorers. In contrast, it would be negatively associated with Neuroticism (H3) (low Adjustment) because of the general social anxiety associated with this trait.

Second, we also examine intelligence correlates of PS, as it is a fundamental aspect of all individuals. We did, however, not expect significant correlations (H4), though it is possible that brighter people are better educated and more articulate and hence happier, where appropriate to speak up.

Third, we explore Self-Compassion and its relationship to PS. Neff (2003) conceptualized self-compassion as consisting of three main elements: self-kindness (being kind and understanding towards oneself), common humanity (recognizing that suffering is part of the human shared experience), and mindfulness (observing thoughts and feelings for what they are). We assumed that high scorers would experience more PS (H5).

Fourth, we look at the relationship between PS and the concept of Attachment, which was made famous by Ainsworth et al. (1978) and Bowlby (1979) . A central tenet of the concept of Attachment relationship is that the caregiver serves as a “secure base” from which the infant explores his/her environment. Over time, three types of attachment relationships manifest: secure, avoidant, and anxious. Attachment theory posits that attachment occurs across the lifespan but is not immutable. A number of different research groups have attempted to devise robust and reliable measures of workplace attachment ( Scrima, 2015, 2020 ).

Neustadt et al. (2011) demonstrated that adult attachment styles have been linked to leadership, trust, satisfaction, performance, and other work-related outcomes ( Harms, 2011 ). There have been a number of recent studies in this area, such as that of Mostafa et al. (2025) , who showed that perceptions of abusive supervision induce employees’ attachment insecurity at work (attachment avoidance and attachment anxiety), and thus undermine employees’ service performance. We assume that individuals who are Securely Attached (H6) would report higher levels of PS at work while those who had Avoidant (H7) and Anxious (H8) Attachment patterns would show negative correlations with PS at domain and facet levels.

All of our hypotheses concerning the correlates of our short PS measure were confirmed. What this study highlighted was the relationship between patterns of attachment at work and PS at the facet level. It appears that these two literatures have developed relatively separately. The former remains anchored in clinical and developmental literature, though there has been an interest in understanding it within work settings ( Scrima et al., 2017 ). The PS literature, though much more recent, is predominantly found in the I/O (industrial/ organizational) work psychology world.

One issue of importance remains the definition and measurement of PS. One question is the components/ facets of PS. If one examines the original definition of PS, presented by Edmondson (1999) , it seems two factors from this study appear to overlap, namely mutual trust and risk-taking. Indeed, Identity Clarity and Structural Support seem less relevant. Also, it could be argued that Identity Clarity reduces flexibility and organizational citizenship behaviour. It is also worth noting that the PSI total variance explained is much higher than the individual factors, which is a potential risk for common method bias.

It was interesting that neither sex, age, nor intelligence was consistently related to PS at the domain or facet level, though there was some evidence that females had higher scores than males on a few facets. Whilst it might be assumed that these variables would be related to safety and perceptions of “psychological danger” (i.e., younger, more inexperienced people feeling more threatened), there seems to be very little evidence of this. This raises the question of the utility of individual trait assessment for group settings over relational ones, which may be more context-specific.

What was also clear, at least from the multiple regressions, was that several of the traits we measured were weakly related to PS, though the correlations did indicate some association. These included Self-Compassion, Agreeableness, Openness and Conscientiousness. It is no surprise that Emotional Stability was a significant predictor in a number of the regressions as it is associated with instability, moodiness and anxiety: anxious people are less inclined to feel safe and speak up. Similarly, Extraverts who are more socially confident and more likely to feel secure in social settings

Perhaps the major clear finding from, this study was the relationship between the PS facets and the three attachment styles. What was striking was that in five of the six regressions, the Secure Attachment at Work variable was by far the strongest correlate. It has been established that people with Secure Attachment patterns can and do develop and maintain strong, happy, and healthy relationships across a wide variety of settings: home, work, and leisure. Evidently, the opposite is true of those insecurely attached (Avoidant and Anxious), who may therefore be particularly sensitive to issues in the workplace. This suggests that in all work environments securely attached individuals would experience more aspects of PS than those who were either Avoidant or Anxiously attached. In this sense, people come to work with a particular propensity for experiencing PS

Three questions arise from this study. The first is the extent to which a person’s work attachment style is a relatively immutable trait that transfers within work settings and, as such, largely determines their feelings of PS. It was argued by the originators of the attachment style literature that these interpersonal patterns of attachment are established early in life and relate to many relationships the person has throughout their lifetime, but that they are indeed open to change; they are not immutable. However, that process is far from simple or easy.

The second is whether the same work environment is perceived quite differently by workers in terms of PS. In this sense, the perception of the PS environment is at least as much a function of personality as it is a product of interaction patterns in the environment. Hence, there are individual differences in perceptions about safety: the same psychological work-space is experienced quite differently by people as a function of their personality.

The third is, what interventions are best for those who feel unsafe, and what are their consequences for those who do feel safe and the business as a whole? Presumably, the first step is to identify those who feel most unsafe and the reasons for their perceptions. Furthermore, it seems a good idea to monitor the PS of the organisation at regular intervals, attempting to determine what the drivers of change are. Equally, given the importance of attachment, it may be wise for an organisation to assess all staff to get an indication of where problems might arise.

Like all studies, this one had limitations. We were restricted mainly to short self-report variables with all their well-known limitations in terms of internal reliability, bias and impression management. We also note that the alphas/correlations for the personality measure were low, but not unusual for the 10-item scale which is extensively used. The Alpha for the Anxious Attachment scale was .67 which is below the usually acceptable cut-off point of .70. Further, there is some doubt about the factor structure of the attachment measure we used and whether the two measures, Avoidant and Anxious attachment can be clearly distinguished. This has been highlighted by previous research with this measure. Hence, it may be more parsimonious to differentiate more simply between Securely and Insecurely Attached.

We also had no details of the type of work a person did, nor the details of their environment. It is abundantly clear that some working environments are much more aggressive and hostile than others where issues of safety are ignored. They may attract and select people who are less concerned with PS. On the other hand, some working environments are highly sensitive to PS issues which they believe to be major factors in the success of the organization.

In this sense, this can be considered a pilot study with the issues explored worthy of replication and further research.

In conclusion, this study aimed to examine the correlates of a new short scale of PS. In doing so it highlighted the association between the concept of attachment and PS, which has not been emphasized in the previous literature. It suggests that those involved in trying to increase a sense of PS in a particular workplace may consider determining an individual’s work attachment profile to get a clearer understanding of how and why they tend to feel unsafe at various levels.