Edinburgh Research Explorer Attitude and Peer Norm Predict How Students Use Lecture Recordings

The impact of lecture recordings on university students’ learning often depends on whether the recordings are used as a substitute for attending live lectures (watching instead of attending; usually undesirable) or as a supplement to attending (watching in addition to attending; usually desirable). However, little is known about the predictors of how students use the recordings. This study examined the demographic and psychological variables associated with attending live lectures, and with the use of lecture recordings as a substitute or supplement. In particular, we characterised the students’ own opinions about attending live lectures and using lecture recordings (attitude), students’ beliefs about what their classmates thought about these behaviours (peer norm), and students’ beliefs about what their lecturers thought about these behaviours (lecturer norm). Using data gathered in a large introductory psychology course (n=212), we found that attending live lectures, using recordings as a supplement and using recordings as a substitute were all viewed favourably and perceived to be accepted by peers. The perception of lecturer norm was more nuanced, with substitution perceived as the least acceptable to lecturers. Generally, the more positive the students’ own attitude and the perceived peer norm towards engaging with lectures in a particular way (attending live, using recordings as a supplement, or using recordings as a substitute), the more likely students were to engage with lectures in that way. These findings suggest that attitudes and peer norms may be valuable targets for educational interventions in this area.


Introduction
Over the past two decades, it has become commonplace in many countries, including the United Kingdom, for live university lectures to be recorded, and for the recordings to be provided to students as a learning aid (Ibrahim et al., 2021;Morris et al., 2019).This is received enthusiastically by the students, who see lecture recordings as a helpful tool for aiding their revision and allowing them to catch up on lectures missed due to illness or schedule clashes (Lokuge Dona et al., 2016).University staff remain divided in their attitudes, with some seeing lecture recording as a welcome technological advancement and others concerned about its impact on attendance and retention of information (Dommett et al., 2020).The existing evidence shows that any links between the use of lecture recordings and academic performance remain tenuous at best (Baillie et al., 2022), and one reason for this may be that students interact with lecture recordings in different ways, but empirical studies do not always capture these differences (Banerjee, 2021;Witchel et al., 2022).
One key distinction (Williams et al., 2012) lies in whether the recordings are being used as a substitute for live lectures (i.e., students watch recordings instead of attending live lectures) or as a supplement to live lectures (i.e., students attend live lectures, and subsequently use the recordings to complement their learning).Observational studies show that students use lecture recordings in both ways (Chinnery et al., 2021;MacKay et al., 2022).However, we have limited insight into why exactly students engage with lecture recordings in one way or the other.Thus, the first aim of this paper is descriptive: we set out to describe how students use lecture recordings, with a focus on supplementation and substitution.We also explore what attitudes students hold towards lecture recordings and attending live lectures, and what norms pertaining to lecture attendance and lecture recording use they perceive among their peers and lecturers.The second aim is to put lecture recordings in a social context: learning is often conceptualised as a social process (Salomon & Perkins, 1998), and therefore it is key to include social factors in our model of lecture recording use.The third aim is predictive: we examine the demographic and psychological variables that can help us predict whether students will use lecture recordings as a supplement or a substitute, and whether students will attend in-person lectures.

Lecture Recordings and Lecture Attendance
Much of the scholarship on lecture recordings has focused on their effect on attendance, largely because attendance is consistently associated with academic achievement (Credé et al., 2010), and so lowering attendance may also lead to a decrease in achievement.However, there seems to be no evidence of a consistent detrimental effect of the introduction or use of lecture recording on students' attendance (Banerjee, 2021).On one hand, Edwards and Clinton (2019) used a longitudinal approach to compare attendance on a similar course before and after lecture recording was introduced, and found that the introduction of recording was followed by a significant drop in attendance.In contrast, studies examining the correlation between attendance and the use of lecture recordings using student-level data available within a particular course (e.g.Nordmann et al., 2019) have found only a very weak association between the amount of recorded material students viewed and attendance at live lectures.
Reflecting on these inconsistent results, two recent chapters by Robson and Kauffman (2022) and Witchel and colleagues (2022) highlight that in the era of high-quality lecture recordings, it may be possible for students to perform excellently even without high levels of attendance.Some types of lectures, as Witchel et al. (2022) point out, may be replaced with recordings quite effectively, and some students may be particularly skilled at using the recordings to their advantage.Thus, to investigate how lecture recordings may be facilitating or hindering student achievement, we first need a more nuanced understanding of how students use the recordings in specific contexts.Williams et al. (2012) introduced an important distinction between using lecture recordings as a substitute for live lectures (watching recordings instead of attending live lectures) versus using them as a supplement to live lectures (watching recordings as well as attending live lectures).The authors went further to argue that the use of lecture recordings has a particularly positive association with achievement where recordings are used as a supplement rather than a substitute.

Substitution and Supplementation
However, as their study only included data on total attendance and recordings use, and not lecture-level data, the analysis assumes that students with high lecture attendance use recordings as a supplement and students with low lecture attendance use them as a substitute.In a more recent study, Bos and colleagues (2016) used individual lecture-level data to examine differences in academic achievement between four groups of students: those who used lecture recordings as a supplement, those who used them as a substitute, students who did not use the recordings at all but attended live lectures, and students who neither attended the lectures nor watched the recordings.The results showed that students who used lecture recordings as a supplement performed better on a course assessment than those who attended the lectures in person but did not watch any recordings afterwards.In contrast, students who just watched the recording did no worse than students who attended the lecture in person.
One way of gaining insight into how students use lecture recordings is to use learning analytics in conjunction with data mining techniques to find distinct patterns in how students engage with various learning materials.This approach has been implemented by O'Brien and Verma (2019), who found four distinct clusters of student behaviours regarding lecture resources: traditional (students who attended lectures and consulted lecture notes but did not utilise the lecture recordings), digital (students who did not attend lectures but relied on lecture notes and recordings), minimal (students who relied on lecture notes, with low attendance at live lectures and minimal use of recordings), and phantom (low attendance, low use of lecture recordings and notes).Surprisingly, this study did not identify a cluster of students whose behaviour would fit the supplementation pattern -it seemed that students either attended live lectures or relied on recordings as a substitute.Similar to Williams et al. (2012), a limitation of this study lay in the fact that it did not look at attendance and recording use on a lecture-by-lecture basis.Instead, what was analysed was the average number of lectures attended and recordings accessed.

Predictors of Lecture Recording Use
A few key studies (e.g., Nordmann et al., 2019;O'Brien & Verma, 2019;Sarsfield & Conway, 2018) have sought to understand what student-level characteristics are associated with specific study behaviours, including the use of lecture recordings.The variables used in these studies include demographic characteristics, such as gender, being a native speaker of English, or having learning difficulties that may affect engagement with lectures.While demographic variables are a common starting point for predictive models of behaviour, psychological models of learning behaviour often go beyond this by incorporating socio-cognitive predictors, such as attitudes or norms (Cheon et al., 2012).
The theoretical framing of this paper is based on the premise that there are both individual and social aspects to learning (Salomon & Perkins, 1998): while university instruction has an element of individual knowledge acquisition, classroom interactions with lecturers, tutors and peers are also a key part of learning.Crucially, other people are not only the transmitters of knowledge, but they also act as models and advisors for how learning should be done -lecturers often give advice on how to learn, and other students provide a reference point on study practices.
Consequently, we apply the socio-cognitive framework of Theory of Planned Behaviour (TPB; Ajzen, 1991) to investigate the predictors of how lecture recordings are used.The key constructs in TPB are attitudes, social norm and perceived behavioural control -these three are predictors of behavioural intention, and intention, in turn, predicts behaviour.While TPB has been subject to critique, especially in the health literature (see Sniehotta et al., 2014), its relative simplicity and high predictive power mean that it remains a key framework for studying predictors of behaviour, including in education (e.g.Allen et al., 2019;Pownall, 2012;White et al., 2008).Some voices have called for improving TPB by adding other known predictors of intention and behaviour (Conner, 2015;Conner & Armitage, 1998).Our work adopted exactly this approach, focusing on the TPB predictors of attitude and social norm, and adding social identification -a construct originating from the Social Identity Approach (Hornsey, 2008).

Attitude
Attitude is typically defined as the person's overall evaluation of the target behaviour.It is often found to be a strong predictor of the intention to engage in the behaviour, and intention tends to be predictive of behaviour (Sheeran et al., 2016).In the domain of education, attitude was found to be strongly associated with intentions to attend peer-assisted study sessions, and was an indirect predictor of attendance, mediated by intention (Allen et al., 2019;White et al., 2008).
The literature consistently reports that students' attitudes towards lecture recordings are overwhelmingly positive (MacKay et al., 2022;Mcgowan & Hanna, 2015;Morris et al., 2019;Nkomo & Daniel, 2021), and this holds for most levels and subjects of study, and also in most demographic groups.The positive attitude is related to the fact that providing lecture recordings helps students remove some of the anxiety around missed lectures or difficulties with understanding the lecturer (Nkomo & Daniel, 2021).Students also report that when they know a recording will be available, they worry less about being able to keep up with the lecture content in their notes (MacKay, 2020).

Social Norms
Social norms refer to whether other people approve of the target behaviour and whether they themselves engage in that behaviour (Cialdini et al., 1991).Smyth and colleagues have studied the influence of norms in university education (2015,2017,2018), showing that perceived learning norms of peers and educators affect student learning behaviours: if students believe that their peers and educators favour learning for understanding rather than just memorising, they are more likely to engage in learning for understanding themselves.Similarly, sending students a normative message where staff recommended spending more time on lesson activities (versus a control message) led to an increase in the amount of time spent on lesson activities and to a higher quiz score (Eyink et al., 2020).To our knowledge, there is no literature about the social norms surrounding the use of lecture recordings, and bridging this gap is one of the key aims of this study.

Social Identification
Social identification refers to "defining part of the self in terms of the social group" (Smyth et al., 2019, p. 409), and it is a key tenet of the Social Identity Approach (Hornsey, 2008).The key relevance of this construct to our theorising lies in the fact that when people highly identify with a social group, they derive significant meaning from that group membership, resulting in a motivation to follow the group norms and engage in behaviours prototypical for that group.Thus, a higher level of identification as a student would be expected to correlate with behaviours that are expected of students in that discipline.Indeed, students who report higher identification tend to engage in more thoughtful learning practices (Smyth et al., 2015), have stronger intentions to continue with academic study (Platow et al., 2013) and achieve higher marks (Bliuc et al., 2011a(Bliuc et al., , 2011b)).
Apart from the direct association with behaviour, identification has also been shown to moderate the effect of social norms (Turner et al., 1987), with highly identified students more likely to follow the peer and educator norm in favour of deep learning (Smyth et al., 2015(Smyth et al., , 2018)).Our study will investigate both the direct and the moderation effect of student identification on attending live lectures and using lecture recordings as a supplement or substitute.

The Current Study
Our study focused on two key research questions.First, what is the uptake of lecture recordings, and are they used primarily as a substitute or a supplement?Second, how do attitudes, norms and identification shape how students engage with lecture recordings?The answers will help educators and researchers understand what drives students' enthusiasm about lecture recordings, and also to what extent being positive about a particular way of using the recordings (e.g.substituting or supplementing) is then translated into behaviour.The results will also provide preliminary insights into whether social norm messaging would be useful for promoting effective lecture recording use, alongside information and practical guidance such as that provided by Nordmann et al. (2020).
To answer these questions, we gathered lecture-and student-level data on attendance and use of lecture recordings.We focused on the role of attitudes and perceived peer and educator norms in shaping three key outcomes: attendance at live lectures, using lecture recordings as a supplement to live lectures, and using lecture recordings as a substitute for live lectures.In line with Theory of Planned Behaviour (Ajzen, 1991), we predicted that attitudes and both types of norms would be positively associated with the corresponding outcomes: for example, that students whose attitudes towards attending live lectures are positive, and who perceive the norm among their peers and lecturers to be favourable towards attending in-person lectures, would attend live lectures more often.Further, in line with social psychological theorising (Turner et al., 1987), we predicted that the influence of both types of norms would be stronger among those students who highly identify with their course cohort.

Participants and Procedure
We recruited students enrolled in a semester-long introductory psychology course at a large research-intensive Scottish university.The data were collected in Autumn 2019, before the covid-19 pandemic led to the cancellation of in-person lectures.There were 327 students enrolled in this course, approximately 90% female, and 60% were enrolled in Psychology degrees (the Scottish system allows students from other degree programmes to take outside courses in the first two years).
All 327 students were invited to participate in the study, 263 consented to participation and completed the study questionnaire, and a subset of 212 students consented to having their questionnaire data linked with administrative records.The data presented below pertain to the 212 participants who consented to having their questionnaire data linked with the administrative data, and whose data we were able to retrieve.Demographic characteristics of these participants are presented in Table 1.Ethical approval was granted by the Psychology Research Ethics Committee (ref.1-1920/1).

Course in Study
The course where this study was embedded was delivered over 11 weeks -10 teaching weeks with a break in the middle.The teaching included three 50-minute-long didactic lectures each week, as well as small-group interactive tutorials with discussion of published articles (four in total) and larger practical labs involving data collection and simple analysis (four in total).There were 327 students enrolled.The course was embedded in the wider university context, where introductory courses are typically large (several hundred students) and follow a similar structure, with a combination of lectures and practical or discussion-based sessions and assessment consisting of both coursework and exams.During the data collection period, most courses (including this one) followed a traditional model, where attendance at live lectures was expected, although not mandatory.
Additional materials provided to the students included (1) a course handbook outlining the learning outcomes and course structure, (2) a virtual learning environment where lecture slides, recordings and other course information was posted, and (3) readings (assigned journal articles and a recommended textbook).The students typically had access to lecture slides at least 24 hours before the lecture but did not have access to any detailed lecture notes.The textbook was suggested as an optional resource, but lectures often went beyond the content covered by the textbook.
The assessment consisted of two substantive elements: an exam comprised of 100 multiple choice questions (45% of the final mark), testing knowledge of the course topics and based primarily on the lecture content, and a 1200-word essay (23% of the final mark), designed to test the students' ability to think critically about the published literature, and related loosely to the tutorial content.The remaining 32% of the course mark consisted of attendance at tutorials and labs, research participation and completion of study skills activities -each of these components accounted for 8% of the mark.
Students with additional support needs had access to the University's Disability Office and might have received learning adjustments, such as additional time at exams, extensions for submitting coursework, or support from a scribe.All students also had access to generic study skills support delivered at the university level.Students who spoke English as a second language had access to university-level support (such as guidance on academic English), but were not given any additional support within the course.

Administrative Data
Attendance was recorded during most lectures, using the attendance-checking functionality within the active learning system Top Hat (2019).During the lecture, attending students were shown a code on the projector screen and were asked to input that code on their devices to confirm attendance.We gathered attendance data from 24 out of 30 lectures delivered during the semester; for four lectures this was not organised in time, and in two cases the lecturer used questions on Top Hat rather than simple attendance checking 1 .These had a lower response rate and therefore could not be compared with the other data, which is why they were subsequently excluded from the analysis.
We had access to recording use data from all 30 lectures -these data were collected routinely by the system used for lecture recordings, Echo360 (2019).We were able to inspect data on the number of times that each student accessed each video and the percentage of the video that had been watched.When we considered each case of a student watching a video at least once (n = 3077), we found that in 71% of cases (n = 2189), if a student accessed a video, they watched half or more of the recording's duration (over one or more watching sessions)2 .Therefore, the data we use here is the number of videos accessed.
Given that we had access to both attendance and recording watching data for 24 lectures, for each participant we calculated two key indices: the number of lectures that the participant attended and then watched (broadly corresponding to supplementing), and the number of lectures that the participant did not attend but watched (broadly corresponding to substituting).Each of these indices ranged from 0 to 24, and they were the key outcome variables in our study.

Questionnaire Data
At the start of the semester, we distributed a questionnaire measuring students' attitudes towards attending live lectures and using lecture recordings, relevant social norms, and identification with the psychology student cohort.Participants responded on 7-point Likert-type scales.

Attitude
We asked students about their attitudes towards three behaviours: (1) attending live lectures in person, (2) attending lectures first and then watching the recordings (i.e.supplementation), and (3) watching recorded lectures online instead of attending them in person (i.e.substitution).For each of these behaviours, we asked five questions on 7-point semantic differential scales anchored at bad-good, harmful-beneficial, unpleasant-pleasant, foolish-wise, and unnecessary-necessary (adapted from Armitage & Conner, 1999).The responses were coded so that a higher score indicated a more positive attitude, and were averaged to form a single attitude score for each behaviour.The Cronbach's alpha for this scale was satisfactory, with alpha = 0.86 for attitude towards lecture attendance, alpha = 0.85 for attitude towards supplementation, and alpha = 0.86 for attitude towards substitution3 .

Peer Norm
For each of the target behaviours, we asked participants a single question to estimate the perceived peer norm (adapted from Bassili, 2008): "How useful do you believe [other students] feel the *live lectures* are in this course?"(attendance), "How useful do you believe [other students] feel the *recorded lectures* are in this course, when they are watched after attending the lecture?" (supplementation), and "How useful do you believe [other students] feel the *recorded lectures* are in this course, when watched instead of attending lectures in person?"(substitution).

Lecturer Norm
We also asked participants to estimate the relevant norm among course lecturers.The questions used were adapted from Bassili (2008): "How useful do you think your lecturers in [this course] consider the *live lectures* to be as a teaching tool?" (attendance), "How useful do you think your lecturers in [this course] consider the *recorded lectures* to be as a teaching tool, when they are watched after attending the lecture?" (supplementation), and "How useful do you think your lecturers in [this course] consider the *recorded lectures* to be as a teaching tool, when watched instead of attending lectures in person?"(substitution).

Student Identification
Identification with other students on the course was measured using a 4-item scale adapted from Cameron (2004): "I have a lot in common with other [course] students," "I feel strong ties to other [course] students," "I find it difficult to form a bond with other [course] students (reversescored)," "I don't feel a sense of being "connected" with other [course] students" (reverse-coded).
The scale was internally consistent (Cronbach's alpha = .84),and so the responses were averaged, with a higher score indicating a higher level of identification.

Analyses Plan
To test our hypotheses, we performed binomial regression, a suitable method for analysing count data with a large number of 0s (Theobald et al., 2019).The coefficient estimates are reported on a natural logarithmic scale and must be exponentiated before they can be interpreted.
The exponentiated coefficient (Odds Ratio, later abbreviated as OR) expresses the likelihood of the outcome, with 1 being a neutral value (i.e. a predictor with a coefficient of 1 does not affect the outcome at all).Predictors with coefficients larger than 1 increase the likelihood of the outcome, and predictors with coefficients below 1 decrease the likelihood.
We estimated three predictive models for each outcome of interest (lecture attendance, substitution, and supplementation).The first model only included demographic predictors: gender, age, being a native speaker of English, being enrolled in a Psychology degree, being a carer, being employed, and identifying as someone with a disability.The second model also included psychological predictors: attitude towards the behaviour, peer norm about the behaviour, educator norm about the behaviour, and group identification.In the third model, we added the predicted twoway interactions between norms and identification.All analyses were conducted using R (v4.3.0,R Core Team, 2023) and R Studio (v2022.07.1, RStudio Team, 2022).The data and analysis scripts, including a list of all R packages used, can be found online at https://osf.io/tgp8j/?view_only=7697958bb8df455fba205a94f0e1a9bd

Lecture Attendance
We had access to attendance data from 24 out of 30 lectures delivered in September -December 2019.Lecture attendance decreased over time (see Figure 1a), with each lecture having on average five fewer students attending than the previous one (linear regression analysis indicated a significantly negative slope, with B = -4.88,p < .001).This pattern aligns with anecdotal evidence from previous years.On average, each student attended 55% of lectures.

Lecture Recording Use
On average, each lecture recording was accessed by 31% of students enrolled in the course.The use of lecture recordings also decreased over time, albeit at a slower rate than attendance: each lecture had on average one fewer student watching the recording than the previous one (linear regression analysis indicated a significantly negative slope, with B = -1.10,p = .002).Students who did not attend a given lecture were more likely to watch its recording than those who did (see Figure 1b).Most students (80%) accessed at least one recording via the system, while 20% of students accessed no recordings at all.In most cases, students accessed a recording only once and watched it in its entirety.

Attitudes and Norms
The distributions of student attitude, peer norm and lecturer norm towards attending lectures in person were left-skewed, indicating that most students had a positive attitude towards this behaviour and reported that both their peers and their lecturers believed that attending live lectures was useful (see Figure 2a and Table 2).The number of lectures attended was significantly and positively correlated with the student's attitude towards attending live lectures (r = .29,p < .01)and their level of student identification (r = .16,p < .05),but not the peer or lecturer norm (ps > .05).
The distributions of attitudes and norms towards supplementation (i.e., attending the lecture in person and then watching the recording) were also left-skewed, though responses were distributed more evenly across the positive response options (see Figure 2b and Table 3).The number of times a recording was used to supplement a live lecture was significantly and positively correlated with the student's attitude towards supplementation (r = .20,p < .01)and the peer norm (r = .20,p < .01).
The distribution of attitude towards substitution (i.e., not attending a lecture and then watching its recording) was close to normal, with the median response indicating a neutral attitude (see Figure 2c and Table 4).The distribution of perceived lecturer norm was also approximately normal, with most responses falling in the middle of the scale.Perceived student norm showed a strong left skew, with most students indicating that their peers felt that substitution was helpful.The number of times a recording was used to substitute a live lecture was significantly and positively correlated only with the student's attitude towards substitution (r = .27,p < .01).

Predicting Lecture Attendance
The demographic model (Model 1a; see Table 5 for full model coefficients) indicated that male students were significantly less likely to attend lectures than female students (OR = 0.73, p = .001)and native speakers were significantly less likely to attend lectures than students for whom English was a second language (OR = 0.67, p < .001).Students enrolled in a psychology degree were significantly more likely to attend lectures than students enrolled in other degrees (OR = 2.22, p < .001),and students with a disability were significantly less likely to attend lectures than students without a disability (OR = 0.61, p < .001).
We then added psychological predictors to the model (Model 1b).The demographic predictors that were significant in Model 1a remained significant, except for gender.Looking at psychological variables, we found that students attended more lectures if they had a positive attitude towards lecture attendance (OR = 1.32, p < .001)and if they more strongly identified with other psychology students (OR = 1.07, p = .046).Somewhat surprisingly, we also found that a positive peer norm about lecture attendance was associated with lower odds of attending lectures (OR = 0.86, p < .001).
In Model 1c, we also included two-way interactions between norms and identification.We found a significant two-way interaction between the student norm and group identification (OR = 1.09, p = .019),in a direction opposite to what we expected: while there was a negative association between student norm and lecture attendance, the slope of this relationship was the steepest among participants with a low level of student identification.In other words, high levels of student identification buffered the effect of the student norm (see Figure 3).
We also found a significant interaction between the lecturer norm and student identification (OR = 1.16, p < .001) in the predicted direction: among students reporting high levels of identification, the odds of attending lectures increased with a more positive perceived lecturer norm.Among students who reported low levels of identification, the odds of attending lectures decreased with a more positive lecturer norm.In this model, we also found a significant effect of student employment, with students in paid employment having lower odds of attending live lectures (OR = 0.82, p = .022).

Predicting Supplementation Behaviour
The demographic model (Model 2a) indicated that male students were significantly less likely to use the recordings as a supplement to live lectures than female students (OR = 0.69, p = .02).Students enrolled in a psychology degree were significantly more likely to use recordings as a supplement (OR = 1.45, p < .001)and so were students with caring responsibilities (OR = 3.37, p < .001).In contrast, students with a disability were significantly less likely to use recordings as a supplement (OR = 0.36, p < .001).Other demographic predictors did not have a significant effect (see Table 6 for full model coefficients).
We then added psychological predictors to the model (Model 2b).The demographic predictors that were significant in Model 2a remained significant, except for gender.Native speaker status in this model became significant, with native speakers more likely to engage in supplementation (OR = 1.33, p = .004).Among the psychological variables, having a positive attitude towards supplementation was associated with higher odds of supplementing (OR = 1.27, p < .001),perceiving a positive peer norm towards supplementation was also associated with higher odds of supplementing (OR = 1.46, p < .001),but perceiving a positive lecturer norm towards supplementation was associated with lower odds of supplementing (OR = 0.81, p < .001).
In the third step, we added 2-way interactions to the model (Model 2c, Figure 4).The main effects from Model 2b remained significant.However, we also found a significant effect of student identification (OR = 1.12, p = .036)and two significant interactions: an interaction between peer norms and identification (OR = 0.64, p < .001),and an interaction between lecturer norms and identification (OR = 1.18, p < .001).Overall, a more favourable peer norm towards supplementation was associated with higher odds of using lecture recordings as a supplement; however, contrary to what we predicted, this relationship was strongest among students who did not strongly identify with their student cohort.Looking at the interaction between identification and lecturer norm, overall, a more favourable lecturer norm towards supplementation was associated with lower odds of using lecture recordings as a supplement.This relationship was strongest among students who did not strongly identify with their student cohort.

Predicting Substitution Behaviour
The demographic model (Model 3a) indicated that native speakers were significantly more likely to use the recordings as substitutes than students for whom English was a second language (OR = 1.45, p < .001).Students enrolled in a psychology degree were significantly less likely to use recordings as substitutes than students taking psychology as an outside course (OR = 0.44, p < .001),and students with a disability were significantly less likely to use recordings as substitutes (OR = 0.65, p = .01).Other demographic predictors did not have a significant effect (see Table 7 for full model coefficients).
We then added psychological predictors to the model (Model 3b).The demographic predictors from Model 3a remained significant.Both attitude and peer norm were significant predictors of substitution behaviour: having a positive attitude towards substitution was associated with higher odds of substituting (OR = 1.47, p < .001),and perceiving a positive peer norm towards substitution was also associated with higher odds of substituting (OR = 1.16, p < .003).
In the third step, we added 2-way interactions to the model (Model 3c, Figure 5).The main effects from Model 3b remained significant, and we also found two significant two-way interactions: an interaction between peer norms and identification (OR = 0.79, p < .001)and an interaction between lecturer norms and identification (OR = 1.09, p = .042).The interaction between student norm and identification ran contrary to what we predicted: among students who did not strongly identify with the course cohort, a more positive norm towards substitution was associated with higher odds of substitution behaviour.Among students who strongly identified with the course cohort, a more positive norm towards substitution was associated with lower odds of substituting.
In other words, identification seemed to act as a buffer against the permissive student norm.
Looking at student identification and lecturer norm, we found an interaction in the predicted direction: students who highly identified with the course cohort were more likely to follow the norm, engaging in more substitution if they perceived the educator norm to be permissive.Students who did not highly identify with the course cohort went against the lecturer norm and engaged in more substitution behaviour if they perceived the norm to be less positive.

Discussion
We found that overall, students had a positive attitude towards all three target behaviours: attending live lectures, using lecture recordings as a substitute and using recordings as a supplement.The peer norm was also largely positive -fellow students were perceived to feel positive all three behaviours.The perceived lecturer norm was more differentiated -very positive towards attending live lectures and using lecture recordings as a supplement, but more mixed towards using lecture recordings as a substitute.Notably, we found that most students accessed lecture recordings at least once in the semester, and that accessing a recording often meant watching that recording in full.This pattern could be interpreted as consistent with substitution.
However, an alternative interpretation might be that students found it difficult to search for specific sections due to how the user interface of the recordings was set up or the structure of the lectures themselves, and therefore resorted to watching the entire video.
Including demographic variables in the models allowed us to corroborate and extend previous findings from the literature.We found that students in psychology degrees were more likely than external students to attend live lectures and to use lecture recordings as a supplement, but they were less likely to use recordings as a substitute.This adds nuance to Nordmann et al.'s (2019) finding that there were no significant differences in how much psychology students and external students used lecture recordings -while we also did not find a difference in how much these two groups used lecture recordings, we did find a difference in how they used them, with external students being more likely to skip live lectures and watch the recordings instead.Another notable result pertains to native speakers -we found them to be more likely than non-native speakers to skip live lectures and use the recordings as both supplementation and substitution.This differs from the findings of Nordmann et al. (2019), who found no differences in attendance, but also found that non-native speakers used lecture recordings more.
Including psychological theorising and variables novel to the lecture recording literature allowed us to draw new conclusions about the factors that influence how students engage with lectures and lecture recordings.Across all three behaviours, we found a significant effect of student attitude and peer norm.Generally, the more positive students' attitudes were towards the focal behaviour, and the more positive the perceived peer norm, the more likely they were to engage with lectures in a particular way.The alignment between attitude and behaviour fits in with the TPB (Ajzen, 1991) and previous research on the predictors of student attendance at university (Ajzen & Madden, 1986;White et al., 2008).However, we found a somewhat surprising pattern for lecture attendance, as a less positive peer norm was associated with higher odds of attending lectures.One potential explanation for this is that the students who do attend lectures might notice the empty seats and develop a perception that the overall peer norm regarding attendance is quite low.An alternative explanation is that social norm is the weakest of TPB predictors, with a significant proportion of studies showing either no relationship between norms and behaviour, or a relationship in a direction opposite to predicted (Armitage & Conner, 2001;Niemiec et al., 2020).
As measurement issues may cause these inconsistencies, and our study included only a singleitem measure of norm, future research may look to use a longer measure of social norms to capture any effect more robustly.Finally, as our study was conducted relatively early in the students' university journey (for most students, this was the beginning of their first semester at the university), the attitudes and norms may not have fully developed yet; future research may look to investigate these relationships among students in later years, where the effects may be more robust.
In each of the predictive models, we found significant interactions between both types of norms and student identification.While the pattern of interactions was complex and not always aligned with our predictions, the overall findings suggest that (1) the level of identification early in the semester matters for how students engage with lectures later, and (2) the level of identification moderates how students respond to the norms present in their environment.Higher identification has been shown to correlate with deep learning practices, such as learning for understanding, looking up suggested readings, and self-testing (Bliuc et al., 2011b).Thus, it is no surprise that higher identification would generally be associated with using lectures in the way recommended by lecturers (e.g., attending live lectures).The premise that the level of identification would moderate the effect of social norms is at the heart of social psychological theorising (Turner et al., 1987), and we did indeed find those moderation effects.The fact that the exact direction of the interaction was not always as expected will require further investigation from future studies in this area.

Strengths
Our study contributes to the literature in two key ways.First, it addresses a common methodological gap in both lecture recording (Banerjee, 2021) and educational technology research (Hew et al., 2019) by applying a well-validated theoretical framework to its conception and undertaking.Having a solid grounding in theory gives us insight into the mechanisms behind empirical findings (Mueller & Urbach, 2017) and increases generalisability of the findings to other contexts (Jones & Czerniewicz, 2011).We found some support for the Theory of Planned Behaviour, with attitude and peer norm identified as key predictors of how students used lecture recordings.
Second, including socio-cognitive variables sheds light on the students' attitudes and perceived norms about attending live lectures and using lecture recordings.Student attitudes towards different learning behaviours are often assumed rather than measured explicitly.In our study, asking about attitudes highlighted that the assumptions educators hold are not necessarily correct.For example, it was a novel finding that the attitude towards supplementation was less positive than the attitude towards attending live lectures.As lecture recording researchers, we perceive supplementation to be the best strategy, as it combines the benefits of attending live lectures and the additional revision of difficult content from the recording.However, it is unclear whether the students have a similar view.Finally, responses to the norm questions were somewhat surprising, particularly in the case of lecturer norms about substitution -we expected the students to hold a strong perception that lecturers do not consider substitution to be a useful learning strategy; instead, we found that the responses to this question were distributed relatively evenly, with no positive skew, indicating that there was no clarity among the students about their lecturers' views on substitution.

Limitations
Our study focused on a cohort of students enrolled in a single course at a university, and therefore, the generalisability of our findings to other contexts could be questioned.Our response to this issue is twofold.First, we see our descriptive findings as another useful data point in the growing literature on lecture recordings.Second, and perhaps more importantly, our findings on the role of attitudes and social norms, since they are rooted in theory, are likely to be more widely generalisable, not only to other contexts but also to other behaviours.Thus, we would expect the use of lecture recordings at any university to be associated with student attitudes towards supplementation and substitution, and also that the student uptake of any new technology would depend on their attitudes towards that technology and social norms pertaining to it.Related to the above point, the course in question did have an assessment structure that rewarded attendance at small group teaching sessions (16% of the mark) and memorising factual information (45% of the mark).It is possible that this incentive structure encouraged students to prioritise attendance at small group sessions over attendance at lectures, and thus affected how students engaged with lectures and lecture recordings.Any research findings need to be interpreted within their context, and ours may not replicate in environments where the incentive structure is significantly different.Future studies investigating the predictors of lecture recording use in other courses will be able to paint a more complete picture.
While the focus on first-year psychology students in our study was driven by both theoretical (Nordmann et al., 2019)  We opted for a single questionnaire at the start of the semester and a one-item measure of student and lecturer norms, which may have been noisy and contributed to the unusual pattern of interactions between norms and identification.In setting up the study, we prioritised participant experience, ensuring that we minimised survey fatigue with a single short questionnaire and clear, succinct questions.This was particularly important given that we asked the same questions about the three target behaviours.We chose to distinguish between the student and lecturer norms, as this distinction was of interest theoretically, but the trade-off was including only one question about each type of norm.In future studies, we would like to see a more nuanced measurement of normative constructs and multiple measurements of attitude and norm over time.This is particularly important because introducing a clearer educator norm may be one promising avenue for intervention in this area.
Finally, it is a limitation that our study did not include academic performance as an outcome.Future research may seek to extend the findings of Williams et al. (2012) and Bos et al. (2016), and investigate how lecture attendance, and using the recordings as a supplement or substitute, are related to academic outcomes.In line with the arguments made by Robson and Kaufman (2022), the effect of lecture attendance on student achievement may have changed since the pre-pandemic times, and may also differ significantly between academic subjects.These relationships would thus best be studied across multiple university courses, or even multiple universities.

Conclusion
In summary, in line with previous work, we found that the uptake of lecture recordings was high, and that recordings were more likely to be watched by those students who did not attend the live lecture.Our key novel finding was that the "how" of lecture recording use (i.e.whether they were used as a supplement or a substitute to live lectures) depended not only on the students' attitude towards the way of using the recordings, but also on their peers' norm around it.Thus, there was robust evidence of lecture recording use being socially determined -students were likely to follow the norm set by their peers.Although quite complex, the pattern of interactions suggested that the level of identification moderates the effect of norms, and that highly identified students tend to engage in the more desirable behaviours (attending live lectures or supplementing with recordings), even if the perceived norms towards those behaviours are negative.These findings suggest potential avenues for further study and educational intervention -both student identification and social norms have previously been studied as targets of intervention and could be targeted in this context to improve effective engagement with lectures.If students feel that they are part of a larger community of learners, and if the peer norms are more positive towards using lecture recordings as a supplement than using them as a substitute, we can expect students to follow those norms and supplement attending live lectures with the recordings, which may in turn lead to more effective learning.Note: The solid black line corresponds to the proportion of students who did not attend the lecture, but watched the recording afterwards (i.e., proportion of students who substituted the recording for a live lecture on this occasion).The dashed grey line corresponds to the proportion of students who did attend the lecture and also watched the recording afterwards (i.e.proportion of students who used the recording as a supplement on this occasion).

Fig. 2
Distribution of scores for self-reported attitude and perceived norms regarding lecture attendance, using lecture recordings as a supplement and as a substitute Note.The dashed line indicates the median score.

Fig. 3
Interactive effect of norms and identification on lecture attendance Fig. 4 Interactive effect of norms and identification on supplementation behaviour Fig. 5 Interactive effect of norms and identification on substitution behaviour and practical factors, including students from other years and other subjects would have allowed us to compare whether the norms and attitudes towards different uses of lecture recordings are the same both within and across degrees.It is possible that as students progress in their degree programmes, they become more socialised in the university culture and develop a clearer view of their lecturers' expectations regarding different study behaviours.It is also possible that some programmes are better than others at communicating what they expect of their students and what learning behaviours they recommend.Future studies could test this hypothesis by measuring attitudes and norms towards lecture recording use in different cohorts, comparing both the year of study and the degree subject.

Table 1
Demographic Characteristics of Participants in the Sample

Table 2
Attending Live Lectures: Means, Standard Deviations, and Spearman's Correlations

Table 5
Demographic and Psychological Predictors of Live Lecture Attendance (Binomial Regression Coefficients)

Table 6
Demographic and Psychological Predictors of Using Lecture Recordings as a Supplement (Binomial Regression Coefficients)

Table 7
Demographic and Psychological Predictors of Using Lecture Recordings as a Substitute (Binomial Regression Coefficients)