Hindrance demands are a set of demands that can be framed within the Job Demands-Resources (JD-R) model. According to Bakker and Demerouti (2007), demands within this model are considered as stressors. The concept of hindrance demands relates to negative demands that are likely to impact personal growth and trigger negative emotions, leading to distress (LePine, Podsakoff, & LePine, 2005).

The JD-R model is a theoretical framework that integrates two independent research traditions: stress and motivation research (Demerouti & Bakker, 2011). It was developed by Demerouti, Bakker, Nachreiner, and Schaufeli (2001). This model postulates that every occupation (or organisation) is likely to have its own work characteristics associated with well-being. The model integrates and elaborates upon previously developed job characteristics models such as the Demand Control Model (Karasek, 1979) and the Effort Reward Imbalance Model (Siegrist, 1996).

A growing interest exists to understand the positive experiences of employees, which include their optimal functioning (Luthans, 2002; Seligman & Csikzenthmihalyi, 2000). However, in the context of this study, it is apparent that there is a need to determine the aspects of work that may hinder optimal functioning of academics in HEIs. Stressors in the workplace can reduce optimal functioning of employees. However, it should be noted that even though stressors exist in an occupation, they may not always present harmful reactions. According to Podsakoff, LePine, and LePine (2007), some stressors may potentially spur personal growth and achievement. These are referred to as challenge stressors and can be differentiated from the stressors that employees consider as constraining their personal development or work accomplishment.

Hindrance stressors influence employee turnover intention, employee loyalty, job attitudes and satisfaction (Boswell, Olson-Buchanan, & Lepine, 2004; Podsakoff et al., 2007) as well as task performance and withdrawal behaviour (Lepine et al., 2005; Podsakoff et al., 2007). Challenge stressors consist of job or role demands, quantitative and subjective workload, time pressures and pressure to complete tasks, job scope and responsibility. Hindrance stressors consist of situational constraints, hassles, organisational politics, resource inadequacies, role ambiguity, role conflict, role overload and concerns about job security (Cavanaugh, Boswell, Roehling, & Boudreau, 2000; Lepine et al., 2005; Podsakoff et al., 2007).

It must be acknowledged that job demands may act as hindrance or challenge demands (Bakker & Sanz-Vergel, 2013). The JD-R model postulates that there are job demands which may hinder employees by draining their energy. According to Tims, Bakker, and Derks (2013), prolonged exposure to work that has high job demands will deplete employees’ levels of well-being. Certain demands can cause negative emotions and inhibit employees’ work goals and well-being. In such instances, hindrance demands act as a barrier, preventing employees from achieving their goals (Siu, 2013).

This study is based on the premise that hindrance demands exist and affect the work performed by academic staff. It goes beyond simply measuring whether academic staff experience work overload and goes on to include the various types of demands that cause stress. This brings to the fore the need for such an instrument to be developed and validated within a South African context.

Ideally, this measure contributes to the most recent occurrences at universities and current events in South African HEIs. The scale consists of 29 items of which 7 items capture workload, 4 items capture higher education unrest and 5 items capture each of the following: change management, decolonisation and online teaching and learning. The last dimension, psychological safety, is captured through three items. All items in the scale were measured on a seven-response format. This response format is in line with methodological recommendations and is regularly used in studies (Weijters, Cabooter, & Schillewaert, 2010). A seven-point scale contains a mid-point, which ensures measurement quality (Nowlis, Kahn, & Dhar, 2002). According to Weijters et al. (2010), a seven-response format is less problematic as it provides more categories and is best used for populations who rank high on cognitive ability.

In terms of scale development, a series of steps were followed:

In this study, the following dimensions can be considered as a hindrance demand and have been delineated as items measured on the HEHDS.

The scales were developed based on the unique nature of HEIs in South Africa and its associated challenges faced at the time of development of the scale. Notably, there are dimensions from existing scales which were adapted from the JD-R scale, namely, workload. Although the JD-R scale exists, the dimensions are not unique to higher education. The intention of developing such a scale was not to create many items or even a lengthy questionnaire. The HEHDS is seen as a brief measure and we acknowledge that brief measures have certain psychometric limitations (Widaman, Liitle, Preacher, & Sawalani, 2011). The nature of the scale is meant to be narrow. We acknowledge that there are a few items in the scale, however, where there is a possibility that higher order factors could exist.

In terms of occupational stress, the concept of workload relates to job demands that may result in or contribute to the development of stress or injury. This is characterised as a psychosocial hazard (MacDonald, 2003). It is noted that the job demands of academic staff are ascribed to various factors, one of which is work overload (Barkhuizen & Rothmann, 2008). Working long hours and the intensification of work have become a common factor in the modern-day workplace (Sang, Powell, Finkel, & Richards, 2015). A frequent occurrence amongst academic staff is the blurring of boundaries between work and home (Wright, Williamson, Schauder, & Stockfeld, 2003). Similar patterns of work have been observed internationally (Bagilhole & White, 2013). The work environment of universities has changed over the last decade and the work performed by academics has been altered from secure and autonomous to insecure and invisible (May, Strachan, & Peetz, 2013). This has resulted in significant changes in the working conditions of academia (Rainnie, Goods, Barns, & Burgess, 2013). Hence, it is critical to assume that workload has the likelihood to hinder academic work.

Workload was measured by seven items on a seven-response format, which ranged from 1 = strongly disagree to 7 = strongly agree, and 1 = never to 7 = always. An example of an item in this dimension is ‘I work under time pressure’. The items in this dimension were adapted from the workload scale of Rothmann and Jordaan (2006), which scored an acceptable alpha of 0.76, and the workload dimension of Roberts, Lapidus, and Chonko (1997), which scored an acceptable alpha of 0.82.

Several manifestations of student protests have occurred over recent years, particularly from 2015 to 2017. These acted as a mechanism for students to have their voices and concerns raised on decisions made by their institutions (Kamsteeg, 2016). South African HEIs were marred by multiple student protests, such as the ‘Rhodes Must Fall’ campaign and the ‘Fees Must Fall’ campaign (Dell, 2015; Rantao, 2015). Such unrest caused major damage to institutions, namely, disruption of the academic calendar and revision of exam dates (Pijoos, 2016), damage to buildings and disruption of examinations (Petersen, 2016). Public HEIs across South Africa were subjected to unrest and violence (Hall, 2016), where artwork, statues and buildings were damaged. Hall (2016) further states that this brought about brutal clashes and confrontations between the police and student factions.

In light of the ‘Fees must fall’ campaign, tuition fee increases had significant implications for higher education access. At the increases of fees, students protested, at times violently, arguing that high fees make education inaccessible to poor students (Wangenge-Ouma, 2012). In light of these incidents, it is likely that university employees, their work and entrance into the institution were impacted. Based on the aforementioned literature, it was deemed necessary to measure unrest in HEIs in South Africa. In terms of the measurement, unrest pertains to any disruption that takes place at HEIs, as we live in a time fuelled with unrest.

Higher education unrest was measured with four items on a seven-response format, which ranged from 1 = never to 7 = always. An example of an item in this dimension is ‘Due to student protests, my work was negatively affected’.

The concept of change affects all facets of society, bringing about uncertainty and creating new possibilities of how to manage changes (Bowin, 2001). Similarly, organisations function in environments that are continuously changing, which presents both risks and opportunities. It could be argued that the introduction of certain changes is essential for the survival of an organisation. However, there are instances in which change, if not managed effectively, will result in the failure of an organisation (Govender & Rampersad, 2016).

South African HEIs were reconfigured in 2001, resulting in institutional restructuring to meet the needs of a developing democracy and paving the way for a new higher education landscape. In this regard, HEI systems were challenged by policy issues (Badat, 2009), mergers of technikons and universities, equity, access and quality (Le Grange, 2011).

According to Vandeyar (2010), HEIs are inherently resistant to change because of global pressures and international competitiveness. In light of the changes that occur within tertiary institutions, changes such as mergers can result in flagging staff morale, job insecurity and redeployment (Seijts & Farrell, 2003). In addition, Seijts and Farrell (2003) suggest that if change is not perceived as a means of creating positive outcomes, employees are likely to resist such initiatives. In terms of higher education literature, the work duties of an academic have been transformed in terms of work tasks and conditions of employment (Enders & Teichler, 1997; Schuster & Finkelstein, 2008).

Higher education institutions are prone to increased levels of uncertainty from societal change because of the production of knowledge (Barnett, 2000). In addition, academics have come to work under a ‘managerial paradigm’ which affects the way they work and in which they are managed (Barnett, 2012). Gill (2014) notes that essential changes in work at HEIs fail to sufficiently document the experiences of employees. Notably, changes that take place are likely to affect the way in which academics perform their work.

Change management was measured through seven items on a seven-response format, which ranged from 1 = never to 7 = always. An example of an item in this dimension is ‘There are continuous changes in processes and procedures which impact my job’.

A series of protest campaigns to ‘decolonise’ academia commenced around 2015 at universities throughout South Africa (Wamai, 2016). Decolonisation refers to addressing the colonial legacy (knowledge, violence and thought) (Pillay, 2015), which is still entrenched within education in South African HEIs (Heleta, 2016), and is perceived as a response to first- and second-generation colonialism (Le Grange, 2016).

Critical changes to the curricula in HEIs have not taken place for nearly a century, and they remain significantly unchanged despite the monumental shifts in the political landscape (Ramrathan, 2016). Such changes include moving away from colonial influences (Heleta, 2016). In terms of transformation, HEIs should include changes in the curriculum and epistemology as well as teaching and learning (Du Preez, Simmonds, & Verhoef, 2016). According to McKaiser (2016), HEIs in South Africa remain entrenched in a colonial outpost mentality and the few changes that have occurred have been superficial and were initiated through the response of policies, namely, professional bodies, qualification frameworks and curriculum frameworks (Ramrathan, 2016). Moreover, Mbembe (2016) indicates that syllabi are designed in a manner to meet the needs of colonialism; hence, it is important to bring about significant epistemological changes in HEIs. Although universities have developed policies and frameworks on change, transformation, equality and equity, they have still not progressively changed (Heleta, 2016).

The debate around the decolonisation of the curriculum is positioned within the context of universities that are strained from various quarters, notably, lack of government funding, increased running costs and increased student access, which makes time-intensive learning processes all the more expensive and unwelcome from a managerial perspective (Kamanzi, 2016). In order for the scholarship of academia and the curriculum to become decolonised, HEIs need to address matters by finding leaders, academics and administrators who possess knowledge of and passion for African content. Because of the past political landscape, white academics were out of touch with such knowledge (Mkhize, 2015). The recent initiation of decolonisation in HEIs may have an impact on academics and their work. With its continuation and universities prioritising decolonising imperatives, it is essential to see the way in which it is being perceived by academics.

Decolonisation was measured by five items on a seven-response format, which ranged from 1 = strongly disagree to 7 = strongly agree. An example of an item on this dimension is ‘Decolonisation of higher education will bring about changes in the way in which I approach my work’.

Fundamental changes in higher education have shifted the role of academics as they become increasingly involved in the changing nature of teaching, namely, online education (Bennett & Lockyer, 2004). The reform of teaching and learning has occurred through the use of technology which has highlighted change and transformation (Strobel & Tillberg-Webb, 2008). The proliferation of online technologies in higher education has resulted in transformation for students and teachers (Collis, 1996; Palloff & Pratt, 1999). The vast differences between traditional teaching and online teaching bring about advantages and disadvantages (Oliver & Herrington, 2001).

According to Allen and Seaman (2013), faculty staff members are increasingly prone to resisting online learning. They resist online teaching and learning as they perceive that educational values and culture are being threatened (Bennett & Lockyer, 2004). Concerns that arise from online teaching and learning include an increase in workload, lack of resources and technological skills and training as well as pedagogical concerns (Bennett & Lockyer, 2004; Hunt et al., 2014; Johnson, Wisniewski, Kuhlemeyer, Isaacs, & Krzykowski, 2012; Power & Gould-Morven, 2011). These concerns are also heightened amongst faculty members who have little or no experience with online teaching (Herman, 2012). Such academics cite fear and resist change to online learning. Some senior academics prefer to hold on to traditional face-to-face teaching (Goolnik, 2012). Academics may perceive that they are being challenged and are inadequately skilled to utilise technologically based teaching and learning tools. With online teaching and learning becoming a reality in HEIs, it is essential to establish whether it is considered a hindrance.

Online teaching and learning was measured by five items on a seven-response format, which ranged from 1 = strongly disagree to 7 = strongly agree. An example of an item on this dimension is ‘The drive towards online teaching and learning in my department is progressing well’.

Early research by Kahn (1990) refers to psychological safety as one’s ability to apply and show the fear of consequences towards one’s career, self-image and status. In more recent research, Edmondson and Lei (2014) explain that psychological safety refers to the feeling of safety whereby people are able to perform effectively in a rapidly changing world under the circumstances of growing and learning. Psychological safety assists people in overcoming barriers to change and learning and work environments that are interpersonally challenging (Edmondson, Higgins, Singer, & Weiner, 2016).

Edmondson, Bohmer, and Pisano (2001) assert that psychological safety is particularly important in environments that are characterised by complexity, high stakes and essential human interactions. There is a likelihood that individuals feel safe in environments where they are able to express themselves without fear of reprisal, being victimised or penalised. When conditions are perceived by individuals as unpredictable and ambiguous, a lack of psychological safety may be experienced (Chikoko, Buitendach, & Kanengoni, 2014).

Psychological safety influences the way in which an individual engages with his or her work. Factors that affect psychological safety are behavioural norms, supervisory relations and relationships between co-workers (Kahn, 1990). Edmondson (1999) notes that psychological safety varies amongst organisations and workgroups.

Academic staff in HEIs face challenges in their daily work that may lesson their sense of psychological safety. A climate may exist within the organisation, which may impact the psychological safety of staff members. The concept of psychological safety asserts itself with elements that allow for consistent social systems that are non-threatening and predictable to be created. This allows social situations to prevail and, hence, allows for engagement (Kahn, 1990). Therefore, it is imperative to assess the psychological safety of academics.

Psychological safety was drawn from Kahn (1990) and included three items on a seven-response format, where 7 = strongly agree, and 1 = never and 7 = always. An example of an item on this dimension is ‘I am afraid to be myself at work’. Olivier and Rothmann (2007) reported an acceptable Cronbach’s alpha coefficient of 0.71.

A non-probability sampling method was utilised and through a convenience sample data were collected from academics across universities in South Africa. Questionnaires were completed online; in total, 184 responses were received from academic staff. The size of a sample is deemed essential for factor analysis and varying guidelines exist in this regard (Tabachnick & Fidell, 2007; Williams, Onsman, & Brown, 2010). Hair, Anderson, Tatham, and Black (1995) and Comrey (1973) suggest that a sample size of 100 participants is considered sufficient and fair. It is noted that the sample size is small for factor analysis. This, however, does not invalidate the results and, hence, should be interpreted with caution.

The majority of the sample comprised women (56%) and men accounted for 44%. The participants comprised academic staff, with the majority of them employed as lecturers (40.8%), followed by senior lecturers (27.2%), associate professors and professors (19.6%), heads of departments/schools (5.4%), junior lecturers (2.2%), other (3.8%) and researchers (1.1%). The white race group (52.3%) was the most represented in the sample. The majority of the population had a master’s degree (31%). A high proportion of employees were permanently employed (90.8%) by their institutions in comparison to staff members who held temporary positions (2.2%). In terms of tenure, staff members had been employed at the HEIs for 0–5 years (28.8%), followed by 6–10 years (24.5%), 11–20 years (21.7%) and greater than 20 years (12.5%). The average age of employees was 43 years (SD = 10.48 years), and ranged between 23 and 67 years. It must be noted that 91.8% of the population was representative of the Faculty of Commerce.

Statistical analysis was carried out using R Studio 3.2.0 (R Development Core Team, 2015). Descriptive statistics were obtained for all factors on the hindrance demands scale and the data were analysed in terms of mean, standard deviation, skewness, kurtosis and Cronbach’s alpha coefficient of each item. The current study utilised exploratory factor analysis to examine construct validity of the HEHDS. To determine the suitability of the factor structure, the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity (Burton & Mazerolle, 2011; Pallant, 2011) were analysed. With this analysis, the bootstrap function was conducted with 1000 iterations. This function allows for the construction of confidence intervals (Kim, Kim, & Schmidt, 2007). It is valuable in estimating biases and variances and constructs confidence intervals in complex data sets (Ichikawa & Konishi, 1995).

In terms of KMO, the index ranged from 0 to 1, and a recommended value of 0.6 and above is considered suitable (Pallant, 2011). Items that are grouped into a set of interpretable factors can efficiently explain the constructs under investigation (Burton & Mazerolle, 2011). Bartlett’s test of sphericity provides a chi-square output which should be significant at <0.05 as recommended by Pallant (2011).

The study used unweighted least-squares method (minimum residual – MinRes) as the factor extraction method. To determine model fit, absolute and incremental fit indices were determined. The absolute fit indices for this study were the following: chi-square (χ²), fit statistic, the standardised root mean residual (SRMR) and the root mean square error of approximation (RMSEA). The assessment of incremental fit indices included the Tucker–Lewis Index (TLI) (Hair, Black, Babin, Andersen, & Tatham, 2010). The study utilised the rule of thumb whereby an RMSEA ≤ 0.05 indicates close approximate fit, values between 0.05 and 0.08 indicate reasonable error of approximation and an RMSEA ≥ 0.10 suggests a poor fit (Kline, 2005). In terms of the indices, if the TLI value is greater than roughly 0.90, this may suggest a reasonably good fit of the model, whereas values less than 0.10 are considered favourable. Consequently, the model with the largest TLI as well as the lowest RMSEA and SRMR is considered the best fit. The TLI can substantially improve models with an overall fit of less than 0.90 (Bentler & Bonett, 1980). The comparative fit index, namely, Bayesian information criterion (BIC), is indicative of model parsimony (Kline, 2011).

The initial examination of the HEHDS used Kaiser–Guttman’s criterion, while Cartell’s scree plot was used to determine the number of factors to be extracted. The KMO measure of sampling adequacy for the hindrance demands scale, which contained 29 items in the first iteration, was 0.8. This exceeded the recommended value of 0.6 (Tabachnick & Fidell, 2007). It is notable that normality and homoscedasticity preconditions for the study were fulfilled. The scree plot indicated that the five factors should be retained; however, the parallel analysis suggested six factors and six components. Hence, both models were tested to determine the number of factors to be extracted and to determine model fits. In this analysis, the bootstrap function was conducted with 100 iterations. The bootstrap function allows for the construction of confidence intervals (Kim et al., 2007).

The fits of the five- and six-factor models were then compared (see Table 1). Model 1 represents the five-factor solution with items loading freely on the five factors, while Model 2 is the six-factor solution with items loading freely on the six factors (as hypothesised). The fit indices of both models were satisfactory.

From the examination of the chi-square fit statistics for Model 1 (p < 0.000) and Model 2 (p < 0.000), it was inferred that Model 2 had a better fit. The TLI values were regarded as satisfactory for both the five-factor (0.755) and the six-factor (0.848) models. Reasonable error of approximation was indicated with regard to RMSEA for Model 1 (0.0790) and Model 2 (.062). Based on the squared mean square residual (SMSR), both models provided favourable results (Model 1 = 0.05 and Model 2 = 0.04), which were lower than the suggested cut-off point of < 0.10. The BIC value suggested that the five-factor model (-829.350) has a better fit than the six-factor model (-865.17) and, hence, allowed for differentiation between the competing models. In addition, 90% intervals were reported. Based on the aforementioned results, Model 2 has a suitable fit and factor structure.

The 29 items of the six-factor model of the HEHDS were subjected to a factor analysis using unweighted least square method (minimum residual – MinRes) and a direct oblimin rotation. Upon inspection of the analysis, the factor loadings that remained were undefined and unstructured. Hence, an uncorrelated factor solution was tested and the results indicated that the data best fitted an uncorrelated factor solution

Based on the statistical fit presented in Table 1, it was suggested that the six-factor analysis should be further inspected to facilitate the interpretation of the models and, hence, the respective factor solutions were examined. Solutions that presented two or more factors could have different factor structures and hence explain the data equally well. Factor loadings that were below the cut-off point of 0.30 were interpreted as non-significant (Diekhoff, 1992) and were not considered during the examination of the factor matrix. The standardised factor loadings for the six-factor solution indicated that the items DECO2, DECO1, OTL1 and OTL4 were to be removed as they did not load against any factors and were considered non-interpretative, which also scored low communalities between 0.14 and 0.23 (see Table 2). In addition, the items CHMAN 1, CHMAN5 and CHMAN4 loaded on more than one factor. These were considered problematic and were hence removed.

The remaining factor loadings for this solution ranged between 0.36 and 0.88, which is indicative of fairly strong to very strong factor loadings. The communalities ranged between 0.20 and 0.80. Evidently, the majority of the values fell above the cut-off point of >0.40. Furthermore, the solution provided six clearly defined sub-clusters, thus providing significant support for the six-factor model.

A second iteration of factor analysis was conducted using the unweighted least-squares method (MinRes). All items loaded accordingly on six factors and the factor loadings ranged from 0.31 to 0.89, with communalities ranging between 0.19 and 0.88. The item CHMAN1 loaded on more than one factor and hence it was decided to remove the factor and conduct the third iteration of the factor analysis.

The third iteration of the factor analysis was conducted (see Table 3) using the unweighted least-squares method (MinRes) and the respective solutions were examined. The factor solutions presented different factor structures and explained the data well. The second iteration data, CHMAN1, were removed. The standardised pattern matrix containing the factor loadings for the revised six-factor solution is indicated in Table 4. All of the factor loadings fell above the suggested cut-off point of 0.30 and ranged from 0.35 to 0.90, which is indicative of fairly strong to very high factor loadings. Communalities for the final structure ranged between 0.19 and 0.88, with the majority of the values falling above the cut-off point of >0.40. The solution indicated six clearly defined sub-clusters, thus providing significant support for the six-factor model. The proportion of variance explained by the six factors was examined. Factor 1 accounted for 10% of the variance, while Factor 2 accounted for 20%. Factors 3 and 4 of the six-factor model explained 33% and 42% of the variance, respectively. Fifty per cent of the variance was explained by Factor 5, while 57% was explained by Factor 6.

All the factors achieved acceptable Cronbach’s alpha coefficients, ranging from 0.77 (lowest) to 0.90 (highest). Hence, the scale satisfied the requirements of internal consistency.

Upon inspection of the parallel analysis and scree plot, it was suggested that a five-factor and a six-factor structure should be considered. Consequently, both structures were tested. However, based on the fit statistics, the six-factor structure fitted the data best and was also in line with the overall hypothesis of this study, namely, higher education unrest, workload, psychological safety, decolonisation, online teaching and learning and change management. It is not uncommon for academic work to be affected by numerous job demands, such as work overload, job insecurity, and lack of growth opportunities, which affect people in the workplace (Bakker & Demerouti, 2007), and more specifically, different types of job demands, such as emotional demands and physical demands (Bakker & Demerouti, 2007). Previous research further supports an expanded model of job demands and job resources which can affect stress at work (Barkhuizen, 2005; Bakker & Demerouti, 2007; Demerouti & Bakker, 2011).

The solutions of the exploratory factor analysis for a five-factor and a six-factor model were evaluated and compared. The goodness of fit indices and the model parameters were examined. The goodness of fit indices for both the five-factor model (TLI = 0.755; RMSEA = 0.079; report root-mean-square residuals [RMSR] = 0.05) and the six-factor model (TLI = 0.848; RMSR = 0.062; RMSR = 0.04) solutions were satisfactory. However, the five-factor model produced a lower BIC value (–829.35) than the six-factor model (–865.17). Hence, the six-factor model was deemed more suitable.

Following this, the proposed six-factor model containing the 29 items was subjected to a further iteration of exploratory factor analysis with an uncorrelated factor solution.

The inspection of the results from the first iteration analysis of the six-factor model suggested that problematic items were present. Such items loaded on more than one factor. CHMAN5 (Changes in the leadership in my work environment make me feel unsettled about the university’s future), CHMAN1 (I feel insecure about my job due to changes in my work environment) and CHMAN4 (Changes within my work environment make me feel anxious about my job) loaded on the factors ‘change management’ and ‘psychological safety’. It is likely that the way in which the item of change management was expressed is similar to that of psychological safety or bears resemblance to items measured under psychological safety. The overlapping factors are likely to suggest that the respective items do not provide adequate distinction or clarification for the factor upon which the double loading exists, except for CHMAN1, which loaded adequately on psychological safety. Hence, those items were considered problematic and discarded.

In addition, items DECO2 (Decolonisation of higher education makes me feel unsettled about my work) and DECO1 (Decolonisation of higher education will bring about changes in the way in which I approach my work) did not load on any factor. The concept of decolonisation is fairly new and refers to colonial legacy (knowledge, violence and thought) (Pillay, 2015) within academia in South African HEIs (Heleta, 2016). It is likely that the concept could be defined differently or perceived differently by academics and hence could possibly influence how it is measured.

OTL4 (My involvement in online teaching and learning initiatives has left me feeling anxious about my work) and OTL1 (The drive towards online teaching and learning in my department is progressing well) did not present any factor loadings and hence were discarded. It is possible that online teaching and learning may not have had a presence in the department or institution and may therefore not have been interpreted well.

The remainder of the 22 items, with the removal of items DECO2, CHMAN5, CHMAN4, DECO1, OTL5 and OTL1, attained factor loadings that ranged from 0.36 to 0.88, while communalities ranged between 0.20 and 0.80. This provided an insight that a further iteration of factor analysis was required and, hence, a third iteration was conducted to include 22 items to determine a finalised structure.

The fit statistics of the final 22-item scale indicated good fit and showed improvement from the original 29-item scale, which was done in the first iteration, and is indicative of a suitable structure. Hence, it can be confirmed as the final factor structure of the HEHDS which includes 22 items, still defined under the following six factors: (1) higher education unrest, (2) work load, (3) change management, (4) decolonisation, (5) online teaching and learning and (6) psychological safety.

It should be noted that scientific research is rarely without limitations. This study has several limitations which need to be addressed. Limitations included that the data were based on a convenience sample and did not utilise probability sampling. The study was conducted on a limited number of HEIs in South Africa because of accessibility and convenience. A number of other universities in South Africa could be considered for future research.

A subsequent limitation of the study was the number of participants. A larger sample size could yield more robust results. We acknowledge that the sample size was small and thus the results of the study should be interpreted with caution. A final limitation regarding the sample was that the majority of participants were from the Faculty of Commerce. It is possible that other disciplines within HEIs could experience hindrances differently. Future research could use the scale and consider validating it through the use of confirmatory factor analysis (CFA).

The items may seem narrow, and it is duly recommended that future studies build upon these items. Future studies can use the scale as a base and as development continues, add to it. It is acknowledged that more items can be added to the scale in the future.

There is a need for further research to be conducted on the challenges that exist within HEIs (Mouton, Louw, & Strijdom, 2012). Such studies could establish which of the demands in the present study could be differentiated as either a challenge or a hindrance demand, and could possibly add to the demands that are experienced by academic staff that are unique to the institution. The study contributed towards literature by investigating the perceived demands that challenge work within a specific occupational group, namely, academics.