This study aims to measure informal learning in situ and identify levers to foster it, as informal learning can offer a competitive advantage.
The authors used experience sampling to assess employees’ extent of informal learning from oneself (e.g. reflection), others (e.g. sharing knowledge) and media (e.g. searching the internet) and took into account work context factors (e.g. job autonomy). Previous retrospective studies have hardly considered the situational nature of informal learning processes and have yielded inconsistent results. Therefore, 364 German-speaking employees completed a survey about their work context, followed by twice-daily surveys for eight days about their current work task, informal learning activities used and perceived barriers to learning.
Multilevel structural equation models showed, at the within-person level, that learning from oneself, others and media increased in situations perceived as newer and more difficult. At the between-person level, individuals who perceived more difficult tasks learned more from oneself, and those who perceived newer tasks and a more positive learning culture learned more from oneself, others and media. No effects were found for task variety, job autonomy and leadership conducive to learning. Phone-related external interruptions were the main barrier to learning.
Levers for human resource development professionals and managers to foster informal learning can be the continuous integration of new/difficult tasks into jobs, the initiation of changes in learning culture through changes in formal structures and reducing interruptions.
The research highlights the importance of measuring situational constructs in situ to reduce bias and provide a robust basis for decisions aimed at fostering informal learning.
Introduction
In today’s rapidly changing work environments, employees must continuously adapt their knowledge and skills. Consequently, informal learning that occurs during work has become increasingly important, while traditional formal learning programs often lack the flexibility to meet individual and organizational needs (Cerasoli et al., 2018; Marsick and Watkins, 1990). When employees fail to solve work problems with their routine procedures, they engage in informal learning activities such as reflecting, trial and error, professional exchange with colleagues or searching the internet (Häußermann and Seufert, 2026; Marsick and Watkins, 1990). To foster the positive effects of informal learning, such as changes in employees’ knowledge, skills or attitudes and individuals’ and organizations’ professional achievement (see review of Smet et al., 2022), human resource development (HRD) professionals and managers must understand which factors influence the extent of informal learning.
Although numerous studies have examined antecedents of informal learning, their findings have been inconsistent. Poor comparability across studies partly explains these inconsistencies, as researchers have used different measurement instruments and examined diverse occupational groups. Methodological choices also contribute to the problem: most studies rely almost exclusively on single retrospective assessments (Cerasoli et al., 2018). This approach makes measurement quality highly dependent on participants’ ability to recall past experiences accurately (Beal, 2015). In addition, prior research has not adequately addressed the situational nature of informal learning. Some potential predictors, such as learning culture or self-efficacy, remain relatively stable over time, whereas others such as work tasks vary across situations and may require informal learning activities to differing degrees. Moreover, existing findings often provide limited insight because researchers typically report only a single aggregate measure of informal learning.
To address these methodological issues, we assessed employees’ use of informal learning activities and current work tasks up to 16 times close to the actual work situation using the experience sampling method (ESM; Beal, 2015). We also distinguished between three learning sources: informal learning from oneself (e.g. reflection), from others (e.g. professional exchange) and from media (e.g. searching the internet; Häußermann and Seufert, 2026) to obtain a more precise assessment. We focused on influencing factors from the work context, as prior research has produced particularly inconsistent results in this domain. Context factors can be subdivided into situationally varying work tasks and more stable job and organization factors (Watkins et al., 2021). Specifically, we examined task newness and task difficulty as situational work-task characteristics; task variety and job autonomy as stable job characteristics; and leadership conducive to learning and perceived learning culture as stable organizational characteristics. In addition, employees reported their situationally perceived barriers to learning.
The aim of this study was to overcome measurement issues from prior research and clarify which factors from the work context influence the extent of informal learning. Our primary contribution lies in the repeated in situ measurement of informal learning activities using the ESM.
How the work context relates to informal learning
Informal learning includes “all conscious learning behaviors that result from daily work, are learner driven, take place alone or in interaction with others, and aim to improve knowledge and skills” (Häußermann and Seufert, 2026, p. 2). In contrast to formal learning, informal learning is not guided by a teacher and largely unstructured (Tannenbaum et al., 2010). Informal learning can arise almost spontaneously from work (reactive) or can be planned with an explicit goal (deliberative; Eraut, 2004). Because of the heterogeneous compilations of informal learning activities in different studies, a large number of factor structures for informal learning have already been proposed (see e.g. Decius et al., 2023). We suggested an initial multilevel approach to the factor structure of informal learning that yielded the best model fit for a solution with three learning sources each at the within-person and between-person levels: learning from oneself was composed of the informal learning activities thinking/reflection, try out/trial and error and rethinking/thinking old problems in a new way; learning from others was composed of professional exchange/co-working with colleagues, searching for feedback, sharing knowledge, observing colleagues and exchange with supervisor; and learning from media was composed of watching videos/tutorials, searching the internet, asking questions digitally and reading books/journals/learning materials (Häußermann and Seufert, 2026).
To understand when employees use informal learning activities and how influencing factors come into play, we draw on three models that provide the theoretical frame for our study: the informal and incidental learning model (Cseh et al., 1999), which describes the process of informal learning, as well as the 3-P model (Tynjälä, 2013) and the job demand–control model (Karasek, 1979; Karasek and Theorell, 1990), which offer insights into different categories and functions of influencing factors. We combine these models because there is currently a lack of theoretical models that can fully explain why and how the work context shapes informal learning processes. Together with our in situ measurement approach, elements of all models may help us to explore underlying mechanisms linking work context and informal learning.
We first take a closer look at the process of informal learning. The informal and incidental learning model conceptualizes this process as cyclical (Cseh et al., 1999). Informal learning begins with an internal or external trigger that “signals dissatisfaction with current ways of thinking or being” (Marsick and Watkins, 2001, p. 29). Subsequently, employees assess the new problem or challenge in light of their prior experiences. Informal learning typically arises under non-routine conditions, when procedures and responses that individuals normally rely on fail. In these cases, employees question their tacit assumptions and recognize that a problem situation can be solved in many different ways (Marsick and Watkins, 1990). Employees then select among alternative courses of action; that is, they engage in one or more informal learning activities. At the end of the process, they evaluate the solution, and the concluding thoughts become part of a revised frame that shapes future actions (Cseh et al., 1999; Marsick and Watkins, 2001). The recently developed self-regulated informal learning cycle builds on this model and extends it by incorporating cognitive memory processes and metacognitive evaluation of the learning process, analogous to self-regulated learning (Decius et al., 2024).
While the cyclical process models only briefly acknowledge that access to appropriate resources affects the selection of informal learning activities, several models, theories and frameworks address this aspect more explicitly, such as the 3-P model (Tynjälä, 2013) or the model of informal workplace learning through problem-solving (Leiß et al., 2022). The models agree that the choice of learning activities depends on personal factors (e.g. motivation), the actual context (e.g. manager support) and learners’ interpretations of the situation. The exact mechanisms involved in the process are not covered. Empirical research has already established substantial evidence regarding the role of personal factors. For example, studies have shown clear positive effects, and qualitative evidence regarding the frequency of informal learning has been found for motivation (Choi and Jacobs, 2011; Schürmann and Beausaert, 2016), self-efficacy (Choi and Jacobs, 2011; Leiß and Rausch, 2023) or self-directed learning orientation (Decius et al., 2021; Gijbels et al., 2012). In contrast, findings on work context factors – which shape norms and determine access to learning resources – remain less consistent, as described in the next section. Therefore, our study focuses particularly on these context factors.
Most models in the field primarily list factors that influence the choice of informal learning activities and resulting learning outcomes. The added value of the job demand–control model lies in its distinction between two key functions of work characteristics, demands and control – in later models broadened to resources in general (Demerouti et al., 2001; Karasek, 1979; Karasek and Theorell, 1990). Even though the model primarily refers to work-related health, it proposes that learning is most likely to occur in active jobs characterized by high demands, combined with high control/resources. Demands represent sources of strain but also serve as necessary triggers for learning, as employees seek more effective work behaviors to achieve their work goals. This search involves the use of informal learning activities such as reflection and experimentation. Accordingly, demands can foster the use of informal learning activities, but they also increase stress (Cseh et al., 1999; Karasek, 1979; Karasek and Theorell, 1990; Marsick and Watkins, 1990; Van Ruysseveldt and van Dijke, 2011).
When demands are high, resources help reduce stress and increase learning (Karasek and Theorell, 1990). Typically, resources take a buffering role in the relationship between demands and negative outcomes such as stress (Demerouti et al., 2001). They support employees by enabling them to implement new work behaviors and cope successfully with challenging situations. The original model emphasized control as the central resource, defined as “the freedom to use all available skills” (Karasek and Theorell, 1990, p. 35). Control allows individuals to decide which response is most effective to address a given demand. Employees can test the efficacy of their actions and adjust them when necessary. Over time, researchers expanded the model and broadened the concept of resources to include aspects such as social support (job demand–control–support model, Johnson and Hall, 1988; job demand–resources model, Demerouti et al., 2001).
For the present study, we selected six demands and resources from the work context to establish a broad yet parsimonious research model. Incorporating all potential influencing factors into a single model would not be feasible given their multiplicity. We focused both on factors that are situational in nature but have not been sufficiently considered as such in previous research and on more stable factors for which prior findings have been inconsistent and therefore warrant further investigation. Task newness and task difficulty were considered as situationally varying task characteristics that function as demands and may trigger learning. Task variety and job autonomy were selected as relatively stable job characteristics referring to the overall job, with task variety conceptualized as a demand and job autonomy as a resource. Leadership conducive to learning, as well as learning culture, was included as a broader organizational resource. We restrict the analysis to the context factors as perceived by individuals and do not consider personal factors, although this restriction may reduce the observed effect sizes (Kwakman, 2003).
In the following, we provide an overview of the effects of context factors on the extent of informal learning in previous research. The hypotheses are presented collectively at the end of the section, after a description of the ESM for a better understanding of the research method.
Factors influencing informal learning
First, we take a closer look at the task characteristics, task newness and task difficulty. As outlined above, both factors function as demands and may trigger informal learning processes (Cseh et al., 1999; Karasek and Theorell, 1990). In his “two logics of learning,” Ellström (2006) argues that learners must master learning situations that require them to solve recurring routine problems efficiently as well as new situations that necessitate innovative approaches. New tasks, in particular, help build new cognitive structures and promote learning (Billett, 2004), while task complexity also determines the learning potential of a task (Ellström, 2006).
Empirical evidence remains limited. A small number of predominantly retrospective studies with only one measurement time have found new (e.g. Jeon and Kim, 2012) and complex tasks (as a related variable to task difficulty; Decius et al., 2021) as predictors of the effectiveness or frequency of informal learning. However, these studies did not account for the situational nature of work tasks and informal learning activities, which makes them susceptible to bias (Beal, 2015). Three diary studies conducted in vocational education and training with office workers represent exceptions. In all three studies, task newness predicted perceived learning potential, whereas only one small positive effect was found for task difficulty (Rausch, 2013). This pattern may be attributable to statistical reasons, as task newness and task difficulty were highly correlated in all three studies. Because Rausch (2013) assessed only the general learning potential of the task, we examine informal learning in greater detail by measuring 12 distinct informal learning activities and clustering them by learning source (Häußermann and Seufert, 2026). The ESM is very well suited to capturing task characteristics in situational contexts (Rausch et al., 2022) and can therefore extend prior research.
Second, we examine the job characteristics task variety and job autonomy. These constructs are less dynamic than task characteristics because they refer to the entirety of work tasks, yet they remain closely connected to employees’ daily work. Both play a central role in motivational work design (Hackman and Oldham, 1975). Task variety represents a demand and may trigger informal learning processes, whereas job autonomy constitutes a resource that enables employees to select among different problem-solving paths, even under high demands (Van Ruysseveldt and van Dijke, 2011). Task variety captures the totality of the different tasks of a job. Job autonomy (also called job control, scope of action and decision latitude) was defined as: “The degree to which the job provides substantial freedom, independence, and discretion to the employee in scheduling the work and in determining the procedures to be used in carrying it out” (Hackman and Oldham, 1975, p. 162). Contemporary measures distinguish three facets of autonomy: planning the timing and sequence of sub-tasks, making decisions independently and choosing working methods (Stegmann et al., 2010).
Empirical findings on task variety and the extent of informal learning remain inconclusive. While one study with teachers found a positive effect on collaborative informal learning activities (Kwakman, 2003), studies with police officers and employees from various industries reported no significant effects (Doornbos et al., 2008; Leiß et al., 2022). The studies did not explain the effects. Findings on job autonomy are even more heterogeneous. In their meta-analysis, Cerasoli et al. (2018) identified job autonomy as a strong predictor of informal learning outcomes. However, individual studies have reported positive effects (e.g. Ouweneel et al., 2009), non-significant effects (e.g. Doornbos et al., 2008; Gijbels et al., 2012; Kwakman, 2003; Leiß et al., 2022) and negative effects (Rausch, 2013). These studies included diverse occupational groups, such as home care managers, dual students of vocational education and training, police officers and teachers. Some evidence suggests that job autonomy may create more opportunities for informal learning among experienced professionals (Ouweneel et al., 2009). In contrast, newcomers may receive closer supervision and limited autonomy may restrict their learning opportunities, while social learning is more prevalent, at least among less experienced dual students (Rausch, 2013). In mixed-experience samples, such opposing effects may have offset each other. Overall, the studies are difficult to compare because they use different operationalizations of informal learning and contextual factors and focus on diverse occupational groups. Given that most rely on retrospective assessments, the in situ measurement of informal learning activities represents a promising avenue for our research.
Third, we considered leadership conducive to learning and perceived learning culture as organizational resources for informal learning processes. Supervisors are in direct contact with their employees and can facilitate informal learning processes, for example, by acting as multipliers, role models, learning facilitators, or feedback providers (Sonntag et al., 2004). Conversely, when supervisors are not explicitly aware of their role in learning or do not value it, they may become barriers to learning (Ellinger, 2005). Similarly, learning culture influences informal learning processes by embedding norms and values in the corporate strategy, making it prioritizable for supervisors (Foelsing and Schmitz, 2021). We define learning culture as the value attributed to learning in companies and follow Sonntag et al.'s (2004) broad understanding, which includes formal corporate structures. This perspective aligns with our aim to identify levers to foster informal learning and with the assumption that informal structures such as learning culture can only be changed indirectly via formal structures (Kühl, 2018). Without the organizational resources of leadership conducive to learning and a positively perceived learning culture, employees may revert to routine behaviors that are likely ineffective (Van Ruysseveldt and van Dijke, 2011).
The distinction between the effects of leadership and learning culture on the extent of informal learning is not always clear-cut. For example, the meta-analysis by Cerasoli et al. (2018) distinguished between formal support (e.g. rewards), informal support (e.g. culture and norms) and people support (e.g. supervisors and peers) and found significant effects on informal learning outcomes in each case. Contrary to the theoretically positively described relations between informal learning and leadership as well as learning culture, empirical evidence is ambiguous. Besides positive effects for social support (which includes supervisors as well as learning climate, e.g. Decius et al., 2021), supervisor support (e.g. Chan and Auster, 2003; Ouweneel et al., 2009), learning culture (e.g. Leslie et al., 1998) and the related learning climate (for social learning, e.g. Crans, 2023), findings are mixed. There were studies that did not find effects for social/management support (e.g. Gijbels et al., 2012; Kwakman, 2003), learning culture (e.g. Berg and Chyung, 2008), or supportive work environment, which predicted the frequency of using informal learning activities only indirectly via formal learning (Choi and Jacobs, 2011). Moreover, Enos et al. (2003) found that employees engaged more in informal learning in the absence of organizational support to fill the resulting gap but suggested that this effect would only occur when learners have sufficient metacognitive skills. Once again, the available studies are difficult to compare and susceptible to recall bias. For example, librarians in Canada were asked to estimate the number of hours they had engaged in informal learning during the previous month (Chan and Auster, 2003). In contrast, our approach uses ESM and assesses informal learning activities clustered by learning source, thereby providing a more precise and robust data basis for identifying work context factors that influence the extent of informal learning.
Finally, in contrast to the factors that can foster the extent of informal learning, we also considered barriers to learning. Barriers are defined as “simply those factors that prevent learning from starting, impede or interrupt learning or result in learning being terminated earlier than it might have been ordinarily” (Hicks et al., 2007, p. 64). A broad overview of barriers to learning, such as lack of time, can be found in Crouse et al. (2011) and Anselmann (2022). However, until now, mainly general barriers to learning were investigated retrospectively. The ESM, however, makes it possible to identify events that block learning directly in the work situation. We asked about these qualitatively in open-text fields. To the best of the authors’ knowledge, barriers to learning have not yet been examined in situational surveys.
This study – an experience sampling setting
Despite the advantages of ESM studies, the factors influencing the extent of informal learning in the workplace have rarely been investigated in an ESM study. The ESM involves “a representative sampling of immediate experiences in one’s natural environment” (Beal, 2015, p. 384). This means that many short surveys are conducted over periods of usually one to two weeks (Beal, 2015) with random measurement times, where people self-report their behaviors as close as possible to their occurrence (Hektner et al., 2007). The added value compared to retrospective studies is that the likelihood of biased results is minimized. Participants who cannot remember details and use heuristics are unlikely because data are collected directly in the work situation (Beal, 2015; Stone and Shiffman, 2002). Measuring the outcome variable in situ and in a more differentiated manner can also increase the accuracy of effects for stable predictor variables. Researchers typically analyze ESM data using multilevel models, which allow them to examine multiple scenarios. These models go beyond traditional between-person comparisons (Level 2) and make it possible to compare individuals across different situations within-person (Level 1; Hektner et al., 2007). Ignoring level-specific variance can lead to incorrect conclusions (Curran and Bauer, 2011). In our research, for example, we can find out that situations with difficult tasks (Level 1) increase informal learning from others in particular, while employees with more difficult tasks (Level 2) generally learn more from oneself, others and media than employees with less difficult tasks.
We measured informal learning from oneself, from others and from media as three latent factors on the basis of 12 informal learning activities (e.g. trial and error, sharing knowledge and searching the internet), each loading onto one of the factors (Häußermann and Seufert, 2026). We assessed the extent of informal learning and considered factors from the work context. The situationally varying factors task newness, task difficulty, use of informal learning activities and barriers to learning were measured in situ. All other factors – task variety, job autonomy, leadership conducive to learning and perceived learning culture – were more stable and only measured once. We set up hypotheses at the within-person and between-person levels, but we did not attempt to examine the data of individuals over time because systematic changes in the informal learning activities were not anticipated over the course of the eight-day survey period. Despite contradictory empirical findings, all hypotheses were formulated positively in accordance with theoretical argumentation. A research question was formulated for the barriers to learning:
In work situations that are perceived as new/difficult, individuals use informal learning activities from learning from oneself/others/media more (within-person, Level 1).
Individuals who experience new/difficult work situations more often use informal learning activities from learning from oneself/others/media more (between-person, Level 2).
Individuals with (a) higher task variety, (b) higher job autonomy, (c) leadership they perceive as conducive to learning and (d) a learning culture they perceive as positive use informal learning activities from learning from oneself/others/media more (between-person, Level 2).
Which barriers to learning arise in the daily working routine?
Figure 1 shows all hypotheses for the latent factor learning from others. For the hypotheses about learning from oneself and learning from media, the informal learning activities on the right side would have to be changed.
The model is divided into two sections labelled within and between. In the within section, task newness and task difficulty are connected to each other and both connect to learning from others within. Learning from others within connects to five learning activities: professional exchange co-working with colleagues, sharing knowledge, observing colleagues, searching for feedback, and exchange with supervisor. The five learning activities are interconnected. In the between section, task newness, task difficulty, task variety, job autonomy, leadership conducive to learning, and perceived learning culture all connect to learning from others between. Task variety is linked to 4 items, job autonomy to 9 items, leadership conducive to learning to 13 items, and perceived learning culture to 20 items. The six influencing factors are interconnected. Learning from others between connects to the same five learning activities: professional exchange co-working with colleagues, sharing knowledge, observing colleagues, searching for feedback, and exchange with supervisor. These five learning activities are also interconnected.Exemplary depiction of the hypotheses for learning from others
Note(s): The upper part of the figure shows the within-person level; the lower part shows the between-person level. On the right side are the informal learning activities that make up the factor of learning from others (identical for both within-person and between-person levels). Correlations are depicted with double-headed arrows. Manifest indicators are shown in square boxes, and latent factors are in oval shapes. Measurement errors have been omitted for reasons of clarity
Source: Authors’ own work
The model is divided into two sections labelled within and between. In the within section, task newness and task difficulty are connected to each other and both connect to learning from others within. Learning from others within connects to five learning activities: professional exchange co-working with colleagues, sharing knowledge, observing colleagues, searching for feedback, and exchange with supervisor. The five learning activities are interconnected. In the between section, task newness, task difficulty, task variety, job autonomy, leadership conducive to learning, and perceived learning culture all connect to learning from others between. Task variety is linked to 4 items, job autonomy to 9 items, leadership conducive to learning to 13 items, and perceived learning culture to 20 items. The six influencing factors are interconnected. Learning from others between connects to the same five learning activities: professional exchange co-working with colleagues, sharing knowledge, observing colleagues, searching for feedback, and exchange with supervisor. These five learning activities are also interconnected.Exemplary depiction of the hypotheses for learning from others
Note(s): The upper part of the figure shows the within-person level; the lower part shows the between-person level. On the right side are the informal learning activities that make up the factor of learning from others (identical for both within-person and between-person levels). Correlations are depicted with double-headed arrows. Manifest indicators are shown in square boxes, and latent factors are in oval shapes. Measurement errors have been omitted for reasons of clarity
Source: Authors’ own work
Method
The hypotheses presented are evaluated quantitatively, whereas the research question is examined qualitatively.
Data collection and participants
The data collection took place in June 2023 via the panel of Bilendi and Respondi. Part of the data, focusing on the multilevel factor structure of informal learning, is published elsewhere (Häußermann and Seufert, 2026). The informal learning activities and their factor structure, as well as the variables of task newness and task difficulty, are used again in this paper to analyze how context factors influence the extent of informal learning.
Participation in the study was open to German-speaking full-time employees who predominantly work at a desk and have regular working hours, to which the sending of the short surveys was tied. 52.2% of the sample were female, and the average age was 42.2 years (SD = 10.9). The majority of the sample (85%) were employees and 13% were supervisors. The professions were diverse but dominated by civil servants, clerical workers and assistants. The most frequently mentioned sector was the service industry (17.3%). Most participants had been working in their current position for 4–10 years (37.6%), followed by 23.3% for 1–3 years. More than half of the participants worked in companies with over 250 employees.
Procedure and experience sampling method
First, a pre-survey was conducted to collect information on demographics, participants’ workplace, task variety, job autonomy, leadership conducive to learning and perceived learning culture. In this survey, informed consent was given for the entire online study. Afterwards, the experience sampling began. As the approximate time of daily work start was queried in the pre-survey, each person could be assigned to one of three time slots in the morning and afternoon. Within each time slot, a survey was randomly sent out during a 1.5-h window (stratified random sampling; Stone and Shiffman, 2002), and participants had 4.5 h to complete it. On average, 50% of the questionnaires were completed within 1 h and 80% within 2 h.
Employees were surveyed twice daily over eight working days, resulting in 5,824 potential measurement situations at 16 measurement times. On average, the first survey took 2.8 min to complete, while the last (identical) survey required 1.2 min. In each of the 16 surveys, respondents were asked to select a task they had worked on in the previous 2 h, assign it to a rough category (e.g. telephoning or creating a written document) and rate task newness and task difficulty. They were then asked to indicate, for each of the 12 informal learning activities, whether and to what extent they had used them for the task at hand. Finally, participants could describe perceived barriers to learning in open-text fields and provide general comments.
In a post-survey, we asked questions about the study. Participants indicated that the surveys were easy to understand. Some participants reported that they had paid more conscious attention to their own learning processes (M = 3.10, SD = 1.10, scale 1–5), whereas many stated that the survey had not influenced their work behavior (M = 1.54, SD = 0.91, scale 1–5). Nine percent found the survey annoying, mainly because of short response intervals and redundancy.
Power, sample and missing data
Monte Carlo simulations indicated that, for multilevel analyses with a power of at least 0.80, 200 level-2 units (persons), 16 level-1 units (measurement times) and a medium or large intraclass correlation coefficient (ICC) typical of within-person measurements yielded a minimum detectable effect size of 0.08 for a direct level-1 effect and 0.21 for a direct level-2 effect (Arend and Schäfer, 2019). With only 10 measurement times, the minimal detectable effect sizes remained the same; however, we excluded participants who completed fewer than 10 short surveys from the analyses. A total of 20–30% missing data were also considered to be likely in comparable studies (Beal, 2015). In general, power can only be roughly approximated, as an exact determination would require parameters such as the variance components at both levels, which are unknown (De Jong et al., 2010). Because of overestimated dropout rates, we ended up with n = 364 participants across n = 4,977 measurement situations. 21.7% of the sample completed all 16 short surveys, and 58.8% completed at least 14 surveys. The number of completed surveys per measurement time varied between 229 and 343. Missed surveys were mostly because of forgetting or time conflicts, while technical issues were rare. Missing data were thus treated as missing completely at random (MCAR; Little and Rubin, 2019).
Measures
In the short surveys, each informal learning activity was measured on a Likert scale from 0 (not used) to 4 (intensive use over several minutes). Because higher values indicated a more extensive use of a learning activity, the measure captured both the frequency and intensity of informal learning. Task newness and task difficulty were each measured with one self-developed item, asking how new (from 1 routine to 5 new) and how difficult (from 1 easy to 5 difficult) the task described in the survey was. For all subsequent variables, a Likert scale from 1 (strongly disagree) to 5 (strongly agree) was used. Task variety and job autonomy were each measured with the German subscales of the Work Design Questionnaire (Stegmann et al., 2010). Task variety was measured with four items (e.g. In my job I do a lot of different things) and showed good internal consistency (α/ω = 0.86). Job autonomy was measured with three items each on planning, decisions and methods (e.g. I am free to schedule my work). The unidimensional scale was preferable to the three-factor structure according to confirmatory factor analyses. Cronbach’s alpha and McDonald’s omega were α/ω = 0.92. Leadership conducive to learning was assessed with the 13 items of the learning-oriented leadership scale from the Learning Culture Inventory (LKI; Sonntag et al., 2005; e.g. My manager supports me in learning). Internal consistency was excellent (α/ω = 0.96). To measure perceived learning culture, we used the “overall items” of the LKI subcategories, which were 20 items after some minor adjustments (Friebe, 2005; Sonntag et al., 2005; e.g. Learning and development opportunities: learning in groups is well supported at our company). Because of the complexity of the learning culture items, we added the option “I cannot judge,” which was used frequently (about 14%) for the items pertaining to the quality review of HRD measures, the company’s external networks and the design of change processes. Internal consistency was excellent (α/ω = 0.98).
Data analyses
Data analyses were performed using R (version 4.2.2). Multilevel structural equation models were estimated for each dependent variable – informal learning from oneself, others and media. Task newness and task difficulty were predictors at Level 1 and Level 2. They were included centered within cluster (cwc) at Level 1 as well as centered at the grand mean (cgm) at Level 2 to answer different questions at each level (Enders and Tofighi, 2007). Task variety, job autonomy, leadership conducive to learning and perceived learning culture were level-2 predictors that do not vary within persons. They were centered at the grand mean. First, null models and ICCs were estimated, followed by stepwise models with only level-1 predictors and models with all predictors. For reasons of parsimony, no interaction terms were included in the models. Analyses were conducted using the maximum likelihood estimator. We reported effects that are standardized for the latent factors and in Likert-scale units for the predictors. The loadings of the first item for each factor were set to 1, following the reference indicator approach.
To assess model quality, we examined the χ2 goodness-of-fit statistics and several local fit criteria: Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA) and the Standardized Root Mean Square at both levels (SRMR-W, SRMR-B). In the multilevel setting, the SRMR is the most sensitive index to model misspecification, as its output is separated by level (Hsu et al., 2015). Models are considered to fit the data well when the CFI and TLI exceed 0.90, the RMSEA is below 0.06 and the SRMR is below 0.08 (Hu and Bentler, 1999). The Akaike and Bayesian Information Criteria (AIC and BIC) were used for model comparison, with smaller values indicating better model fit.
Not all clusters (persons) showed variance for all level-1 variables (informal learning activities, task newness and task difficulty), but these clusters were still included in the analyses. A total of 4,250 measurement situations from 311 people were included in the analyses. Missing values occurred for participants without a supervisor, those with missing data on all learning culture items and those who missed measurement times.
For the qualitative evaluation of the perceived barriers to learning, responses were grouped into categories.
Results
Descriptive statistics
Descriptive statistics and bivariate correlations of all predictor variables are presented in Table 1. Regarding participants’ tasks, only 4.96% of the tasks of all persons at all measurement times were completely new, and 2.45% were rated as very difficult (Likert score 5), whereas 28.71% of all tasks were routine and 23.73% were very easy (Likert score 1). This pattern was consistent across measurement times. Moreover, across all measurement times, tasks were more likely to be performed alone (42–74%) than with others (12–30%). Creating/editing a written document was reported most frequently (668), followed by writing an e-mail (627) and internal or external meetings (612).
Descriptive statistics and bivariate within and between correlations of all predictors
| No. | Variable | n1 | n2 | M (SD) | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Task newness | 4,977 | 364 | 2.36 (0.73)a | 0.77 | 0.06 | 0.16 | 0.14 | 0.12 | |
| 2 | Task difficulty | 4,977 | 364 | 2.43 (0.70)a | 0.51 | 0.00 | 0.11 | 0.05 | 0.06 | |
| 3 | Task variety | 364 | 3.91 (0.76) | – | – | 0.48 | 0.38 | 0.32 | ||
| 4 | Job autonomy | 364 | 3.76 (0.77) | – | – | – | 0.41 | 0.40 | ||
| 5 | Leadership conducive to learning | 315 | 3.33 (0.98) | – | – | – | – | 0.79 | ||
| 6 | Perceived learning culture | 359 | 3.19 (1.05) | – | – | – | – | – |
On the left side, descriptive statistics are listed; on the right side, bivariate correlations are shown. Lower triangular: within-person correlations. Upper triangular: between-person correlations. n1 = Level 1, n2 = Level 2. A multilevel correlation was calculated between task newness and task difficulty. The person’s means of task newness and task difficulty were used for the between-person correlations with the other predictors. aThis mean value and standard deviation refer to the person-mean variable. For the standard deviation of person-mean variables, each person is weighted equally, regardless of the number of measurement times
For all stable predictors (task variety, job autonomy, leadership conducive to learning and perceived learning culture), the mean values were above the scale midpoint. A total of 13.5% of the sample had no supervisor; therefore, the corresponding analyses were based on n = 315 individuals. Five participants were excluded from the learning culture analyses because of missing data on all learning culture items (n = 359).
Factor loadings for all informal learning activities are presented in Table 2. Additional descriptive information on the informal learning activities and the multilevel factor structure of informal learning is omitted, as these data are published elsewhere (Häußermann and Seufert, 2026).
Factor loadings for the null models of learning from oneself, learning from others and learning from media
| Learning from oneself | Learning from others | Learning from media | ||||
|---|---|---|---|---|---|---|
| ILA | L1 | L2 | L1 | L2 | L1 | L2 |
| refl | 0.462 | 0.399 | ||||
| reth | 0.549 | 0.878 | ||||
| trial | 0.454 | 0.578 | ||||
| exch | 0.948 | 0.678 | ||||
| feedb | 0.705 | 0.786 | ||||
| shar | 0.831 | 0.746 | ||||
| obs | 0.551 | 0.694 | ||||
| sup | 0.477 | 0.749 | ||||
| tut | 0.329 | 0.775 | ||||
| ask | 0.213 | 0.832 | ||||
| read | 0.560 | 0.801 | ||||
| int | 0.546 | 0.713 | ||||
The presented factor loadings are standardized using latent variable standardization (Std.lv). L1 = Level 1; L2 = Level 2; ILA = informal learning activity; refl = thinking/reflection; reth = rethinking/thinking old problems in a new way; trial = try out/trial and error; exch = professional exchange/co-working with colleagues; feedb = searching for feedback; shar = sharing knowledge; obs = observing colleagues; sup = exchange with supervisor; tut = watching videos/tutorials; ask = asking questions digitally; read = reading books/journals/learning material; int = searching the internet. All factor loadings were significant (p < 0.001)
Multilevel structural equation models
The results of all hypotheses are summarized in Table 3. Manually calculated ICCs at the factor level (between-factor variance divided by the sum of between- and within-factor variance) based on the null models revealed that 42.7% of the variance in learning from oneself was attributable to the between-person level, compared to 33.7% for learning from others and 84.8% for learning from media. At the informal learning activity level, ICCs ranged from 29.9% to 61.5% across activities. Therefore, all informal learning activities showed significant variance at both levels.
Results of multilevel structural equation models
| Learning from oneself | Learning from others | Learning from media | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Parameters | Null modelb | L1b | L1 + L2 | Null model | L1 | L1 + L2 | Null model | L1 | L1 + L2 |
| Observations | 4,977 | 4,977 | 4,250 | 4,977 | 4,977 | 4,250 | 4977 | 4977 | 4250 |
| Clusters | 364 | 364 | 311 | 364 | 364 | 311 | 364 | 364 | 311 |
| Chi2 | 0.032 | 39.257 | 71.708 | 213.155 | 230.428 | 247.651 | 134.027 | 187.09 | 151.221 |
| df | 1 | 5 | 16 | 10 | 18 | 42 | 4 | 10 | 28 |
| CFI | 1.000 | 0.986 | 0.975 | 0.972 | 0.972 | 0.968 | 0.959 | 0.949 | 0.959 |
| TLI | 1.005 | 0.967 | 0.954 | 0.944 | 0.953 | 0.955 | 0.876 | 0.897 | 0.936 |
| AIC | 42,273.702 | 40,971.244 | 34,960.630 | 71,779.443 | 71,527.126 | 61,250.255 | 48543.088 | 48281.51 | 41616.216 |
| BIC | 42,364.878 | 41,075.446 | 35,106.788 | 71,942.257 | 71,702.966 | 61,459.959 | 48673.339 | 48424.78 | 41794.147 |
| RMSEA | 0.000 (0.000–0.021) | 0.037 (0.027−0.048) | 0.029 (0.022−0.036) | 0.064 (0.057−0.071) | 0.049 (0.043−0.054) | 0.034 (0.030−0.038) | 0.081 (0.069-0.093) | 0.060 (0.052-0.067) | 0.032 (0.027-0.037) |
| SRMR_W | 0.000 | 0.017 | 0.015 | 0.025 | 0.021 | 0.021 | 0.041 | 0.035 | 0.032 |
| SRMR_B | 0.003 | 0.003 | 0.035 | 0.041 | 0.041 | 0.030 | 0.012 | 0.012 | 0.018 |
| Task newness | 0.300*** | 0.309*** | 0.195*** | 0.192*** | 0.184*** | 0.186*** | |||
| Task difficulty | 0.611*** | 0.603*** | 0.133*** | 0.137*** | 0.207*** | 0.188*** | |||
| Task newness | 0.519*** | 0.315** | 0.570*** | ||||||
| Task difficulty | 0.251* | 0.206 | −0.109 | ||||||
| Task variety | 0.057 | 0.079 | 0.016 | ||||||
| Job autonomy | 0.016 | −0.026 | 0.033 | ||||||
| Leadership | −0.010 | 0.045 | −0.042 | ||||||
| Learning culture | 0.344*** | 0.393*** | 0.420*** | ||||||
| L1 R2 | 45.9% | 45.6% | 6.3% | 6.2% | 8.5% | 7.8% | |||
| L2 R2 | 47.0% | 38.4% | 36.3% | ||||||
| L1 Var (SE) | 0.213 (0.019) | 0.899 (0.033) | 0.108 (0.009) | ||||||
| L2 Var (SE) | 0.159 (0.031) | 0.459 (0.053) | 0.601 (0.051) | ||||||
| ICCa | 42.7% | 33.7% | 84.8% | ||||||
All reported regression coefficients are standardized using latent variable standardization (Std.lv). L1 = Level 1; L2 = Level 2; CFI =Comparative Fit Index; TLI = Tucker–Lewis Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; RMSEA = Root Mean Square Error of Approximation; SRMR-W = Standardized Root Mean Square at the within-person level; SRMR-B = Standardized Root Mean Square at the between-person level; leadership = leadership conducive to learning; learning culture = perceived learning culture; Var (SE) = variance (standard error); ICC = Intraclass Correlation Coefficient; L1 variables were centered within cluster; L2 variables were centered at the grand mean. aICCs do not contain an error term. bThe estimated residual variance of rethinking/thinking old problems in a new way was negative at Level 2 and was therefore fixed to 0 at Level 2. *p < 0.05, **p < 0.01, ***p < 0.001
For the null model of learning from oneself, the residual variance estimate for the informal learning activity rethinking/thinking old problems in a new way at Level 2 was negative. Possible reasons for so-called Heywood cases include, for example, small sample sizes with many indicators, factor loadings below 0.05, missing values or outliers (Farooq, 2022). As we did not find an obvious reason in our case, we fixed the variance of the item to zero at Level 2, which is the most widely used solution to Heywood cases (Farooq, 2022). We obtained an acceptable SRMR, which is considered the most reliable fit index in the multilevel setting with one value per level (Hsu et al., 2015).
In H1a, we proposed that in work situations that are perceived as new/difficult, individuals use learning activities from oneself/others/media more (within-person). To test this hypothesis, task newness and task difficulty were each centered within cluster. This procedure eliminates between-person variance and provides a more accurate estimate (Enders and Tofighi, 2007). The results were in line with our hypothesis. For learning from oneself, others and media, level-1 effects of task newness and task difficulty were significant. For example, an increase of one Likert point in task newness in one situation corresponded to 0.195 standard deviations more learning from others. The predictors explained 45.9% of the variance in learning from oneself, 6.3% in learning from others and 8.5% in learning from media. Analogous to the null model for learning from oneself, the variance of the item rethinking/thinking old problems in a new way was fixed to 0 at Level 2 because of a negative residual variance estimate warning in R. The model fit was very good, and results were interpreted as usual.
In H1b, we examined whether individuals who experience new/difficult work situations more often use learning activities from oneself/others/media more (between-person). To test this hypothesis, task newness and task difficulty were each centered at the grand mean. Under grand mean centering, a level-2 predictor indicates the additional explanatory value of the cluster means while controlling for level-1 predictors, thus representing the between-person effect (Enders and Tofighi, 2007). The results are partly in line with our hypothesis. For all three informal learning factors, the level-2 effects of task newness were significant. Task difficulty had only a significant effect for learning from oneself. For example, an increase of one Likert point in task newness for a person corresponded to 0.315 standard deviations more learning from others. Increasing task difficulty for a person by one Likert point did not significantly increase learning from others.
In H2, we investigated whether individuals with (a) higher task variety, (b) higher job autonomy, (c) leadership they perceive as conducive to learning and (d) a learning culture they perceive as positive use learning activities from oneself/others/media more. All four level-2 variables were centered at the grand mean. The results are not in line with our hypothesis. For all three informal learning factors, there was a strong effect of perceived learning culture (β = 0.344–0.420). For example, the effect of learning culture on learning from others indicates that an increase of one Likert point in perceived learning culture in one person corresponded to 0.393 standard deviations more learning from others. However, none of the other factors had a significant effect on the extent of informal learning. The overall models explained between 36.3% and 47.0% of the variance at Level 2.
Qualitative results on barriers to learning
The results addressing our research question of which barriers to learning arise in the daily working routine are shown in Figure 2. Participants could mention their currently perceived barriers to learning in open-text fields at each measurement time. During the survey, a total of 798 barriers to learning were reported across all participants, corresponding to an average of 2.19 reported barriers per person. The qualitative analysis was based on the absolute frequencies of mentions across all participants and measurement times. External learning barriers such as the phone were mentioned more frequently than internal barriers such as fatigue. Some participants also noted that it was difficult to turn off external sources of interruption such as the phone because they had to be available, for instance, because of on-call duties.
The vertical axis ranges from 0 to 350. The horizontal axis lists distraction categories. Phone has the highest frequency at 329, followed by colleagues supervisor at 247 and mails chats at 129. The remaining categories are noise at 37, meetings at 30, other tasks at 27, time work pressure at 26, care assistance for pets, kids, family, dependants, neighbours, vendors at 26, routine at 26, concentration tired at 25, customers at 19, general distraction at 17, technology at 8, and heat at 6.Results from the open-text fields for barriers to learning
Note(s): The absolute number of mentions for each barrier to learning, aggregated across all 364 study participants and covering 798 measurement situations, is shown on the y-axis. Barriers to learning could be entered voluntarily in the open-text fields at each measurement time, and multiple mentions per measurement time were possible
Source: Authors’ own work
The vertical axis ranges from 0 to 350. The horizontal axis lists distraction categories. Phone has the highest frequency at 329, followed by colleagues supervisor at 247 and mails chats at 129. The remaining categories are noise at 37, meetings at 30, other tasks at 27, time work pressure at 26, care assistance for pets, kids, family, dependants, neighbours, vendors at 26, routine at 26, concentration tired at 25, customers at 19, general distraction at 17, technology at 8, and heat at 6.Results from the open-text fields for barriers to learning
Note(s): The absolute number of mentions for each barrier to learning, aggregated across all 364 study participants and covering 798 measurement situations, is shown on the y-axis. Barriers to learning could be entered voluntarily in the open-text fields at each measurement time, and multiple mentions per measurement time were possible
Source: Authors’ own work
Discussion
The aim of this study was to assess the extent of informal learning from oneself, others and media in situ and to clarify which factors from the work context influence it to provide HRD professionals and managers with levers to foster informal learning. To achieve this aim and obtain a more precise measurement than previous retrospective research, we chose an experience sampling approach, in which employees reported their informal learning behavior close to the actual work situation twice a day for eight working days.
Theoretical implications
The combination of differentiating informal learning activities by source, applying situational measures and multilevel analyses, and drawing on established theoretical models allowed us to gain deeper insights into work context factors relevant to informal learning. We integrated the process model by Cseh et al. (1999), the 3-P model (Tynjälä, 2013) and the job demand–control model (Karasek, 1979) and took initial steps toward a more comprehensive and process-oriented understanding of how context factors operate within informal learning processes.
We hypothesized that informal learning from oneself, others and media would increase both in work situations perceived as new or difficult (H1a) and for individuals who experience such situations more often (H1b). Using the ESM and multilevel modeling allowed us to disentangle situational and person-level effects. We confirmed that the effects were positive at both levels, thereby avoiding potential misinterpretations caused by conflating within- and between-person variance (Curran and Bauer, 2011). In situations characterized by greater task newness and task difficulty, employees engaged more frequently in informal learning activities, regardless of learning source. At the person level, task newness remained a significant predictor of all three learning sources, whereas task difficulty predicted only learning from oneself. This pattern may reflect the high correlation between task difficulty and task newness, consistent with Rausch (2013). These positive effects support the theoretical view that work tasks function both as learning triggers (Cseh et al., 1999) and as demands that initiate informal learning processes (Karasek, 1979).
When examining stable job-level antecedents of learning from oneself, others and media (H2a + b), we found no effects for task variety or job autonomy. Neither construct has an inherent conceptual link to informal learning. Prior research has typically drawn on the job demand–control model (Karasek, 1979), particularly the active learning hypothesis (Karasek and Theorell, 1990), to justify such relationships. However, this hypothesis describes only broadly how demands and control may give rise to learning. One prerequisite is the “conversion of energy into action” (p. 36), and employees must be able to evaluate the efficacy of their responses to stressors and adjust their behavior accordingly (Karasek and Theorell, 1990). Task variety and job autonomy may therefore represent necessary but not sufficient conditions for learning (Ellström, 2001). As suggested by the 3-P model, interactions with personal characteristics likely play a crucial role (Tynjälä, 2013). Previous research indicates that engagement in learning activities depends on the value employees attribute to them (Doornbos et al., 2008). Studies on self-regulated learning (Kittel et al., 2021) and critical reflection (Van Woerkom, 2003) further indicate that self-efficacy moderates the relationship between job characteristics and learning activities. A review on why and how job characteristics influence learning identified motivational and metacognitive processes, such as goal setting, as key mediators (Wielenga-Meijer et al., 2010). Taken together, these findings underscore the importance of person-context interactions and highlight the need to examine interaction effects systematically in future research.
Our findings also raise questions about the functional assumptions of the job demand–control model. Karasek and Theorell (1990) note that demands are difficult to conceptualize because they may involve workload, interpersonal conflict or physical strain, although their model focuses primarily on task requirements. In the context of learning, our results suggest that further differentiating task requirements may be useful. The lack of an effect for task variety may reflect the “high strain” quadrant of the model, where high demands are not matched by sufficient resources (Karasek, 1979; Karasek and Theorell, 1990). Participants in our study may have experienced task variety as demanding rather than enriching and may have needed additional resources to benefit from it. Consistent with this interpretation, Hicks et al. (2007) found that task variety promoted learning in higher level positions, whereas trainees perceived multiple tasks primarily as stressful. This suggests that task variety functions differently depending on available resources and individual characteristics. An alternative interpretation would be that the tasks themselves may not have offered genuine learning potential but instead consisted of multiple routine activities. This is supported by the relatively low mean values for task newness and task difficulty in our study. In that case, task variety would not represent a learning-triggering demand but rather a barrier to learning. Distinguishing between learning-triggering demands and learning barriers could extend the model and improve its explanatory power for informal learning processes.
Another open question concerns whether job autonomy always functions as a resource that reduces stress and increases learning (Karasek and Theorell, 1990) or whether it can also operate as a demand (Van Ruysseveldt and van Dijke, 2011; Zolg and Herbig, 2023). Job autonomy may become demanding when cognitive requirements are too high, consuming additional capacity and potentially leading to overload (Zolg and Herbig, 2023). An inverse U-shaped relationship between job autonomy and learning was assumed by some scholars in this context. The corresponding empirical findings on learning but also on well-being remain inconclusive (Van Ruysseveldt and van Dijke, 2011; Zolg and Herbig, 2023). Because our study focused exclusively on learning, future research should also consider stress and its interaction with demands to clarify the function of job autonomy. Our null finding may also have methodological reasons: situations with high and low autonomy may have offset one another, as we measured job autonomy once as a stable job characteristic rather than in situ.
Regarding organizational factors (H2c + d), we found positive effects of learning culture on informal learning from oneself, others and media. This finding highlights that informal learning requires not only situational triggers but also a sense of legitimacy and organizational support. In contrast, we did not observe an effect of leadership conducive to learning, despite strong theoretical arguments emphasizing leadership’s importance for workplace learning. One explanation may be that supervisory support in formal learning contexts does not necessarily transfer to informal situations. Employees may initiate self-directed learning when needed, even without explicit encouragement. In addition, leadership functions increasingly appear to be distributed within teams, with experienced colleagues assuming supportive roles and reducing the centrality of supervisors (Foelsing and Schmitz, 2021). Qualitative research could further explore how employees perceive supervisory support in informal learning and whom they consider most influential for their informal learning processes.
The absence of a leadership effect may also reflect statistical issues. Leadership conducive to learning and perceived learning culture were highly correlated (r = .79). Because only the variance in leadership independent of learning culture was used to predict the extent of informal learning, multicollinearity cannot be ruled out (Alin, 2010). Kwakman (2003), for example, reported suppression effects caused by high predictor correlations, which produced a negative effect of management support on individual learning activities. This suggests that leadership conducive to learning and learning culture may exert similar influences on informal learning, which seems plausible given that both shape the overall conditions for learning (Foelsing and Schmitz, 2021).
Across all hypotheses, the pattern of results was consistent for the three learning sources (oneself, others, media). For all context factors, effects appeared either across all learning sources or not at all, with the exception of the Level 2 effect of task difficulty. This pattern suggests that the examined context factors do not selectively activate specific learning sources but rather shape general learning opportunity structures. Future in situ research could investigate whether employees prefer particular learning sources for specific types of tasks.
With regard to perceived barriers to learning (RQ), we primarily identified external barriers, consistent with Anselmann’s (2022) interview study. However, employees may have reported mainly external barriers even when these triggered internal barriers – for example, concentration difficulties following interruptions. Our findings extend existing classifications of stable structural barriers by adding a situational perspective. Informal learning processes appear highly vulnerable to situational interruptions, particularly from phones, colleagues, supervisors or incoming messages. Such interruptions may prevent learning from starting, interrupt ongoing processes or force employees to abandon them. Notably, our results revealed a paradoxical role of social embeddedness: social interaction is both a central source of learning and the most frequent barrier when employees do not consciously take measures to protect their learning. These micro-level dynamics, which hinder learning, are not addressed in the model by Cseh et al. (1999) but are essential for understanding realistic learning processes. They counteract work task characteristics that trigger learning. If (situational) barriers to learning were incorporated into the job demand–control model, they would represent demands not necessarily related to learning but inherent in the work environment. Employees could then only take advantage of learning-triggering demands and learn successfully if they had resources, such as control over their availability, that protect them from interruptions.
In sum, our study highlights the importance of situational measures to reduce bias when examining antecedents of informal learning activity use. Situational data also support a stronger process perspective by clarifying how and why context factors influence the selection of informal learning activities. Although the informal and incidental learning model (Cseh et al., 1999) emphasizes situational influences, prior empirical research and theoretical frameworks such as the 3-P model (Tynjälä, 2013) and the job demand–control model (Karasek, 1979) have largely overlooked this dimension. Future research should focus more strongly on person-context interactions and on the interplay among supportive resources, learning-triggering demands and learning barriers using situational measurement. Over time, insights from such research can inform and refine existing theoretical models.
Practical implications
Based on the results, we identified three immediate starting points for HRD professionals and managers to foster the extent of informal learning from an organizational perspective.
First, employees should work on new/difficult tasks on a regular basis. Since Taylorism, it has been known that variety in work tasks has a motivating effect. Job enrichment, as a job design technique that challenges employees by adding new and more difficult tasks to their jobs, could be one effective approach. When thinking further, this leads to self-managing teams, that is, autonomous working groups that independently allocate work tasks among themselves (Armstrong, 2006). This approach may offer a practical solution to provide employees with regular access to new tasks. However, maintaining a balance between sufficient resources and new challenges is essential to prevent the positive effects from being reversed (Van Ruysseveldt and van Dijke, 2011).
Second, the learning culture should be perceived as positively as possible by employees. Clearly, changing the learning culture is not a panacea and cannot be implemented as a short-term project. However, according to Kühl (2018), a change in learning culture can only be achieved indirectly through targeted adjustments to formal structures so that attitudes and norms toward learning can subsequently change as well. For instance, management can adjust learning time regulations to initiate subsequent changes in the learning culture. Given the strong leverage effect of learning culture on the use of informal learning activitites that we observed in this study, this approach represents a promising starting point, although such measures always entail the risk of unintended side effects.
Third, interruptions should be minimized. From an organizational perspective, employees are often expected to remain highly available, yet they also require opportunities to develop new skills without constant interruptions. Establishing fixed weekly learning times, during which employees are allowed to mute their phones and engage individually or in groups in self-directed informal learning – in the sense of Eraut’s (2004) deliberative informal learning – could be a valuable approach for many professions. The specific time window can be set individually. Alternatively, some companies have introduced concepts such as Learning Fridays, which vary considerably in implementation but share the common goal of encouraging learning and knowledge sharing (Eggmann, 2020). In doing so, companies send signals that can enhance the acceptance of learning throughout the organization.
If companies succeed in fostering informal learning, entire societies will benefit from adaptable employees and competitive organizations. The formal education system should lay the foundation for this by teaching students how to learn informally in a conscious and self-directed manner while also providing them with space to pursue their individual learning processes.
Limitations, future directions and conclusion
Despite the added value of this study, which reduced recall bias through the ESM, our analyses are based on self-report data, and common method bias cannot be ruled out (Podsakoff et al., 2003). Moreover, our measurement approach captures whether informal learning activities are applied but not whether actual learning occurs. We also cannot exclude the possibility that the ESM itself acted as an intervention and influenced learning behavior. In addition, ESM data depend heavily on the single measurement situation.
Future research should place greater emphasis on the process level of informal learning. ESM studies, for example, could explicitly ask why employees did or did not choose a particular learning activity. Qualitative approaches may also help uncover these underlying decision processes. In addition, researchers should examine interactions between personal factors, such as motivation and context factors, as prior studies have produced promising findings in this area (e.g. Kittel et al., 2021; Zia et al., 2022). When collecting data, scholars should also assess situationally varying constructs in situ. For example, job autonomy could also be measured at the task level, in line with the approach taken by Rausch (2013). Additionally, our study could be replicated in an occupational field with more new and difficult tasks or with blue-collar workers to examine the generalizability of our findings across diverse work contexts.
Finally, the ESM proved to be a valuable method for investigating situationally varying constructs such as informal learning. We identified task newness, task difficulty and a positively perceived learning culture as levers to foster informal learning. Conversely, external interruptions were found to be particularly detrimental. With these insights, HRD professionals and managers can specifically foster employees’ self-directed informal learning and thus enhance their companies’ competitiveness. Such efforts benefit not only organizations but also employees and society as a whole.
Thanks to the market research institute Bilendi and respondi, whose panel was used for the data collection.
Author contributions
Katja Häußermann designed the study, run the study, conducted the data analysis, interpreted the data and wrote the original draft. Ulrike Nett contributed to the data analysis and reviewed the draft. Tina Seufert contributed to the funding acquisition, the design of the study, supervised and reviewed the draft. All authors have approved the current version of the manuscript.
Ethics statement
This study was exempt from an ethic committee approval because of the recommendations of the German Research Association. All subjects were in no risk out of physical or emotional pressure; we fully informed all subjects about the goals and process of this study and none of the subjects were patients or minors. Participation was voluntary. All subjects signed a written informed consent and could stop participation in the study at any moment.
