Introduction

Instructional videos are widely used in instructional settings. Due to the COVID-19 pandemic, these media have been used more frequently as face-to-face instructional settings have had to be transferred to online environments. This raises the question of how to design these instructional videos to be enjoyable and effective for learning. In formal educational settings, such as in schools and higher education, a commonly used design of instructional videos presents PowerPoint slides accompanied by an instructor’s video or audio explanation only (e.g., Homer et al., 2008; Pi et al., 2020). However, whether or not to show the instructor in video-based instruction has strong practical relevance and has led to increased research interest aimed at evidence-based design recommendations (e.g., Alemdag, 2022; Henderson & Schroeder, 2021). Unfortunately, authors use different terms to refer to visible instructors in instructional videos, such as image effect (e.g., Fiorella & Mayer, 2021; Kizilcec et al., 2015), teacher video (e.g., Colliot & Jamet, 2018), and instructor presence (e.g., van Wermeskerken et al., 2018; Wang & Antonenko, 2017; Wilson et al., 2018). The image effect (Fiorella & Mayer, 2021) shows either a human instructor or a pedagogical agent, which can be an animated human or nonhuman virtual character. Pedagogical agents have been extensively discussed and meta-analytically reviewed (e.g., Castro-Alonso et al., 2021; Heidig & Clarebout, 2011; Schroeder et al., 2013). Because “instructor presence” refers exclusively to human instructors, we suggest using this term to investigate visible human instructors (as opposed to pedagogical agents) in instructional videos.

From a theoretical point of view, two contradictory hypotheses are discussed in the context of instructor presence: Benefits of the perceived social presence of the instructor for affective outcomes and learning (known as the Social Cues Hypothesis), and detrimental effects on learning due to an increase in extraneous processing caused by the attention drawn by the instructor’s video (known as the Interference Hypothesis). A recent review (Henderson & Schroeder, 2021) and meta-analysis (Alemdag, 2022) on instructor presence in instructional videos do not show general effects of a visible instructor on affective and cognitive outcomes, but rather mixed results. Most studies used relatively short instructional videos (up to 10 min) in laboratory or online studies that were not integrated into formal learning settings (e.g., Wang et al., 2020a; Wang et al., 2020b; van Wermesmerken et al., 2018; Zhang et al., 2021). In laboratory studies, the instructional videos are presented in a well-controlled environment where participants are physically present in a specific room at a specific time. In online studies, participants receive a link to the instructional video, which they can watch at any time and from any location. The instructional videos presented were stand-alone videos that were not part of an exam-relevant course. Therefore, we investigated instructor presence in instructional videos embedded in university courses to enhance external validity and to address the question of how to design instructional videos for lectures in higher education.

The instructional videos were presented as part of a series of lectures with exam-relevant content, and were presented by an instructor known to the learners. The current paper presents three field studies investigating instructor presence embedded in university courses.

Theoretical background

Social cues hypothesis

The transition from face-to-face to online instruction often reduces social interaction processes. One way to induce social processes in multimedia learning environments, particularly instructional videos, is to provide social cues to activate social schemata (Fiorella & Mayer, 2021; Moreno & Mayer, 2000; Reeves & Nass, 1996). Social cues such as voice, image, facial expression, eye gaze, and gestures can be implemented in instructional videos by presenting the instructor’s video next to the slides. According to social agency theory (Mayer & DraPra, 2012; Mayer et al., 2003), social cues create a sense of social connectedness even in online studies and activate social responses in learners. The presence of a visible instructor creates a feeling of social presence (Gunawardena, 1995; Short et al., 1976) and a sense of interacting with the instructor. Perceived social presence may then lead to increased learner engagement and better learning outcomes (Gunawardena, 1995; Tu, 2002). Accordingly, social agency theory (Mayer & DraPra, 2012; Mayer et al., 2003) further states that the priming of the social interaction schema by presenting social cues motivates the learner to engage more deeply in generative processing (Fiorella & Mayer, 2021). Generative processing is one of three instructional goals (along with minimizing extraneous processing and managing essential processing) derived from the Cognitive Theory of Multimedia Learning (CTML: Mayer, 2001, 2021). It refers to cognitive processes for making sense of instructional material and depends on learner motivation (Fiorella & Mayer, 2021). The instructor’s social presence may motivate the learners to invest more effort in understanding and processing the material, leading to better learning outcomes. This is referred to as Social Cues Hypothesis.

Research on multimedia design principles has been primarily based on CTML (Mayer, 2001, 2021), focusing on cognitive aspects and aimed at minimizing extraneous processing and managing essential processing according to Cognitive Load Theory (CLT, Sweller, 1988, 2020). However, few models have been introduced that extend CTML by including affective and motivational factors, such as the Cognitive-Affective Theory of Learning with Media (CATLM, Moreno, 2006) and the Integrated Cognitive Affective Model of Learning with Multimedia (ICALM, Plass & Kaplan, 2015). Schneider and colleagues (2021) additionally consider social responses (Cognitive-Affective-Social Theory of Learning in digital Environments, CASTLE). Providing social cues to induce social responses in multimedia learning is one way to facilitate affective processes in learners.

In sum, the Social Cues Hypothesis assumes that presenting the instructor in instructional videos next to the slides may lead to increased social presence, higher affective and motivational ratings, and more generative processing of the learning material, resulting in better learning outcomes.

Interference hypothesis

Contrary to the Social Cues Hypothesis, the Interference Hypothesis suggests that instructor presence in instructional videos may harm learning (e.g., Colliot & Jamet, 2018). From a cognitive perspective, the visible instructor is additional visual information for the learner to process. According to the CTML (Mayer, 2001, 2021), which is based on the dual-channel assumption (Paivio, 1986) and Baddeley’s model of working memory (Baddeley, 1986; Baddeley & Logie, 1999), humans have two separate channels for processing information, one for visual/spatial information and one for auditory/verbal information. Both channels have a limited capacity. The visible instructor adds other visual information to the slides’ text (and graphics) and may overload the visual/spatial channel.

According to Cognitive Load Theory (CLT, Sweller, 1988, 2020), the addition of a visible instructor increases extraneous processing as the additional visual information must be processed. Further, learners must divide their attention between the slides and the instructor, causing split attention (Ayres & Sweller, 2021; Tarmizi & Sweller, 1988). As working memory capacity is limited (Baddeley, 1986; Cowan, 2001), an increase in extraneous processing leaves fewer resources for generative processing and may therefore interfere with learning.

In sum, the Interference Hypothesis suggests that presenting a visible instructor next to the slides increases extraneous processing and may harm learning.

Empirical evidence of instructor presence in instructional videos

The empirical evidence on instructor presence in instructional videos was recently synthesized by Henderson and Schroeder (2021) in a systematic review and by Alemdag (2022) in a meta-analysis.

Henderson and Schroeder (2021) reviewed 12 studies that met their inclusion criteria (comparing on-screen and no on-screen instructors, human instructors (not pedagogical agents), and measuring learning outcomes). They found no general positive effect of visible instructors on cognitive or affective outcomes. Regarding cognitive outcomes, visible instructors were advantageous for learning in 7 out of 27 learning outcomes, whereas 20 comparisons showed no significant differences. This calls into question the Social Cues Hypothesis. However, the authors also found no evidence of a detrimental effect of visible instructors on learning that would confirm the Interference Hypothesis. Contrary to the Interference Hypothesis, results on extraneous cognitive load were also mixed. While the Interference Hypothesis states that visible instructors increase extraneous processing, some studies even found that learners in the visible instructor condition reported less extraneous load than learners without the instructor present (e.g., Wang et al., 2020a, 2020b). In terms of affective outcomes, visible instructors did not lead to higher ratings of perceived social presence, indicating that the instructor’s voice only created the same sense of social presence as additionally showing the instructor’s image (Henderson & Schroeder, 2021). Consistent with the Social Cues Hypothesis, higher student satisfaction ratings were shown when the instructor was present (Wang & Antonenko, 2017; Wang et al., 2020a; Zhang et al., 2021). According to Zhang and colleagues (2021), this was only the case when the instructors’ image was positioned on the right side of the screen. The results on social presence and satisfaction are based on only three quantitative studies (Henderson & Schroeder, 2021).

The meta-analysis by Alemdag (2022) used largely the same inclusion criteria as Henderson and Schroeder (2021), such as comparing on-screen and no on-screen instructors and including only human instructors (not pedagogical agents). They included studies that measured learning outcomes, cognitive load, social presence, or motivation. Consistent with the results of Henderson and Schroeder’s (2021) review, Alemdag’s (2022) meta-analysis of 20 empirical studies also found no effects of visible instructors on learning (knowledge acquisition and transfer) or social presence. Furthermore, visible instructors lead to higher ratings of learner motivation (Hedges’ g = 0.431, based on eight comparisons). Alemdag (2022) subsumed engagement, interest, and satisfaction under motivational ratings. In contrast to Henderson and Schroeder’s (2021) results, Alemdag (2022) found a significant effect of instructor presence on the increase of cognitive load (Hedges’ g = 0.319, based on seven comparisons). However, Alemdag (2022) did not differentiate between cognitive load and extraneous processing. Moderator analyses were conducted for knowledge acquisition (based on 22 comparisons), as few studies were available for other outcomes (less than ten comparisons). Alemdag (2022) found no significant moderating effects for video length (< 6 min vs. > 6 min) or learning domain (STEM vs. humanities). A significant effect was reported for study setting, suggesting that learning gains due to instructor presence are more likely to be seen in laboratory studies than in online studies (Alemdag, 2022).

In sum, the recent review (Henderson & Schroeder, 2021) and meta-analysis (Alemdag, 2022) found no effects of instructor presence on learning. Regarding affective outcomes, they report no effects on social presence, but benefits in learner satisfaction (Henderson & Schroeder, 2021) and motivation (including satisfaction, Alemdag, 2022). Thus, the results only partially support the Social Cues Hypothesis. They report mixed results on extraneous processing/cognitive load and no effects on learning, at least partially contradicting the Interference Hypothesis. Thus, the empirical evidence provides some rationale for including a visible instructor, as positive effects on affective measures may occur, but cannot be shown consistently across studies and for different affective variables. The empirical evidence also provides little rationale for excluding the visible instructor, as his or her presence did not affect learning across studies. However, individual studies have shown adverse effects on learning (e.g., Wilson et al., 2018).

Additionally, the results of the meta-analysis point out that instructor presence may occur mainly in laboratory settings. We argue that whether the effect of instructor presence might occur may not depend on the study setting (laboratory vs. online study), but rather on the learning setting (embedded vs. not embedded in higher education courses). We use the term “embeddedness” to refer to instructional videos that are (a) part of a course in which the learner is enrolled, (b) whose content is relevant to the learner’s exams, (c) is more than 30 min in length, and (d) is presented by an instructor known to the learner. This is in contrast to “no embeddedness,” where learning content that is not relevant to the learner’s exams is presented by instructors unknown to the learner and in very short instructional videos.

Instructor presence in instructional videos embedded in courses in higher education

The practical relevance of instructor presence arises from the question of whether instructors in authentic learning settings such as schools, higher education, or further education should include their videos in instructional videos to facilitate affective outcomes and learning results. Therefore, we focus on increasing external validity by presenting studies set in university courses.

Current studies, the review (Henderson & Schroeder, 2021), and the meta-analysis (Alemdag, 2022) do not explicitly consider the relevance of the learning materials to the learner, the relevance to the exams, or the personal relevance of the instructor. Regarding video length, Alemdag (2022) compared instructional videos of more than 6 min to less than 6 min, a video length that is not appropriate for instructional videos in higher education lectures. The complex topics in university lectures require longer instructional videos to replace face-to-face instruction. While Guo and colleagues (2014) recommend videos of 6 min or less to increase engagement in massive open online courses (MOOCs), it has been argued that this may not apply to on-campus university students because the learning setting is quite different from MOOCs (e.g., Ahn & Bir, 2018; Nielsen, 2020). Longer instructional videos that may replace face-to-face lectures should be examined to increase external validity. Most studies on instructor presence include videos up to 10 min (e.g., Kokoç et al., 2020; Wang et al., 2020a, 2020b; van Wermesmerken et al., 2018; Zhang et al., 2021). Only a few studies include videos longer than 20 min (Homer et al., 2008; Pi & Hong, 2016). We are not aware of any studies that include videos longer than 30 min.

Presenting instructional content that is relevant to learners and the exams they must pass in higher education may facilitate learner engagement in the learning task (e.g., Keller, 2007; Zander & Heidig, 2020). Further, the personal relevance of the instructor to the learner needs to be considered. Is he/she a real-life instructor of the learners or an unknown person? This is important in terms of the strength and valence of the social cues provided. The Cognitive-Affective-Social Theory on (digital) Learning Environments (CASTLE, Schneider et al., 2021) postulates that the number and strength (e.g., salience) of the social cues provided are essential for the degree of activation of social schemata. Domagk (2010) further argued that it is not the mere inclusion of social cues that is critical for learning, but the valence of those social cues. She showed that a likeable virtual instructor facilitated transfer performance, whereas the presentation of two unappealing social cues (appearance and voice) hindered transfer. In contrast to animated pedagogical agents, human instructors in instructional videos can be the actual instructors of the learners and therefore be known to the learners to varying degrees. Thus, they differ in their personal relevance to the learner. In a qualitative analysis of 68 surveys and a focus group interview, Reupert and colleagues (2009) showed that most distance students prefer instructors with personal presence created by self-disclosure, relationship building, humor, and individualized feedback. Instructors who are known to the learner from other learning settings transfer their personal presence and relationship with the learner to the instructional video. Therefore, presenting an instructor known to the learner may increase the salience and facilitate the valence of the social cues provided. This may lead to higher ratings of the instructor’s social presence and, consequently, more impact on affective and cognitive outcomes than an unknown instructor. Most studies investigating instructor presence in instructional videos include instructors who are not personally known to the learner (e.g., Colliot & Jamet, 2018; Fiorella et al., 2019; Kokoç et al., 2020; Wang & Antonenko, 2017; van Wermesmerken et al., 2018; Wang et al., 2020a, 2020b). Only a few studies present personally known instructors to the learners (e.g., Alfasor, 2021; Schmidt-Borcherding & Drendel, 2021). Both studies were set in courses with exam-relevant learning material. Alfasor’s (2021) study was set in a culturally specific context, presenting male instructors to female students at a Saudi Arabian university in a gender-segregated educational system. Schmidt-Borcherding and Drendel (2021) showed higher retention with coherent learning material and the presence of the instructor, but no differences in transfer performance. They did not report affective outcomes and presented a 10-min video.

Overview of the experiments

Research on instructor presence has shown no effects on learning across studies, mixed effects on extraneous processing/cognitive load, and some advantages for satisfaction, but no effect on social presence (Alemdag, 2022; Henderson & Schroeder, 2021). Further, the effects of a visible instructor are more likely to occur in laboratory studies than in online studies (Alemdag, 2022). This raises the question of whether the effect occurs primarily in controlled settings. However, this finding is based on studies that mostly use relatively short instructional videos (up to 10 min), irrelevant material for exams, and instructors who are unknown to the learner. Therefore, we investigate instructor presence in instructional videos embedded in courses in higher education to increase external validity. The three studies presented here seek to provide essential extensions to the existing research on instructor presence in instructional videos by (1) presenting exam-relevant learning material, (2) having an instructor known to the learners, (3) having a video length of more than 30 min, (4) being embedded in actual study courses, and by considering (5) a variety of affective outcomes, (6) cognitive outcomes, and (7) learners’ preference for whether or not the instructor’s video is shown. So far, it has not been investigated whether the effect of a visible instructor is more likely to occur when online studies are embedded in formal educational settings (as implemented in points (1), (2), (3), and (4)). Due to the mixed results in the previous evidence (Alemdag, 2022; Henderson & Schroeder, 2021), we collect several affective and cognitive outcome variables in order to be able to map differentiated effects.

In all three studies, an instructional video, including audio explanations to PowerPoint slides, was presented either with the video of the instructor delivering the lecture (visible instructor condition) or without the instructor’s video (no visible instructor condition). The design, content and embedding of the instructional videos used in the three studies are summarized in Table 1. Both instructional videos were designed in line with design recommendations from multimedia learning research (Mayer & Fiorella, 2021), such as coherence, spatial contiguity, and avoiding redundancy and seductive details in the slides (e.g., Fiorella & Mayer, 2021; Rey, 2012; Schroeder et al., 2018). Information about the instructor video provided in the visible instructor condition of Studies 1 to 3 is presented in Table 2.

Table 1 Design, content and embedding of the instructional videos used in the presented studies
Full size table
Table 2 Design of the instructor’s video as provided in the visible instructor condition
Full size table

The three studies were carried out as online studies. First, an exploratory field study (Study 1) was conducted, followed by two synchronous field studies (Study 2 & 3) in different lectures at two different universities. Study 2 was designed to replicate the exploratory Study 1 with a larger sample. It also investigated the learning setting as a separate factor, comparing instructional videos that were part of a lecture series and exam-relevant (embedded in a course) to stand-alone instructional videos that were not exam-relevant (not embedded in a course). Study 3 aimed to replicate Study 1 using a different instructional video embedded in a course at a different university. Dependent affective measures were social presence, well-being, and motivation. Dependent cognitive measures were extraneous processing and learning outcomes. In addition, we assessed the learner’s preference to show or not show the instructor. The research design of the three studies is summarized in Table 3.

Table 3 Research design of the presented studies
Full size table

Hypotheses

The studies presented in this paper, therefore, seek to answer the following research questions:

RQ1

Does a visible instructor in an instructional video embedded in higher education courses facilitate affective processes in learners (social presence, motivation, well-being) compared to a video without a visible instructor?

In line with the Social Cues Hypothesis, we expected that adding social cues by including a visible instructor would increase the sense of social presence, as well as learners’ motivation and well-being. Although previous studies do not consistently show positive effects of a visible instructor on affective ratings and often find no effect (cf. Alemdag, 2022; Henderson & Schroeder, 2021), we assumed a positive effect due to the embedding in courses (personal relevance of the instructor, exam-relevant topic, longer videos).

H1

Learners who watch an instructional video with a visible instructor embedded in higher education courses will report higher social presence than learners who watch a video without a visible instructor.

H2

Learners who watch an instructional video with a visible instructor embedded in higher education courses will report higher motivation than learners who watch a video without a visible instructor.

H3

Learners who watch an instructional video with a visible instructor embedded in higher education courses report higher levels of well-being than learners who watch a video without a visible instructor.

RQ2

Does a visible instructor in an instructional video embedded in higher education courses affect learners’ cognitive processes (extraneous processing and learning outcomes) compared to a video without a visible instructor?

In line with the Interference Hypothesis, we expected that showing a visible instructor might increase extraneous processing due to the additional visual information to be processed. Almedag’s (2022) meta-analysis showed an increase in cognitive load (not precisely extraneous processing), while Henderson and Schroeder (2021) reported mixed results on extraneous cognitive load.

H4

Learners who watch an instructional video with a visible instructor embedded in higher education courses will report higher extraneous processing than learners who watch a video without a visible instructor.

With respect to learning outcomes, the Social Cues Hypothesis and the Interference Hypothesis lead to contrary predictions: The Social Cues Hypothesis suggests that the presence of a visible instructor enhances affective processes, leading to more generative processing of the learning material and hence to better learning outcomes. The Interference Hypothesis suggests that the presence of a visible instructor interferes with learning by increasing extraneous processing. For this reason, and due to mixed results in previous studies (cf. Alemdag, 2022; Henderson & Schroeder, 2021), we formulated an undirected hypothesis.

H5

Learners who watch an instructional video with a visible instructor embedded in higher education courses will have different learning outcomes than learners who watch a video without a visible instructor.

Study 1

Method

Materials

The instructional video used in the study was part of a 14-week social psychology lecture. The video comprised a 34-min presentation on group conflicts. Two versions of the video were produced. The no visible instructor condition showed a PowerPoint presentation synchronized with the instructor’s verbal explanations. The visible instructor condition additionally included the synchronized video of the instructor giving the lecture in the lower right corner of the screen (Fig. 1). The verbal explanations and PowerPoint slides did not differ between the visible instructor and no visible instructor conditions. The instructional video was system-paced.

Fig. 1
figure 1

Instructional Video used in Studies 1 and 2 without Visible Instructor (left) and with Visible Instructor (right)

Full size image

Participants and design

Twenty-eight students majoring in psychology at a German university of applied sciences were enrolled in the social psychology course that we used as the setting for this first exploratory study. Due to the field setting, participation in the 14 parts of the lecture was voluntary. Eighteen college students watched the instructional video provided during the two weeks of the survey for this study, which was set in weeks eight and nine of the 14-week lecture. Therefore, the participants were 18 college students (15 female, three male, 0 diverse). The mean age was 22.4 years (SD = 2.91). They were randomly assigned to either the visible instructor condition (n = 10) or the no visible instructor condition (n = 8).

Measures

Affective measures

The learner’s initial motivation was measured at the beginning of the study to control for motivational differences between the experimental groups. Therefore, participants rated four items (e.g., “I find the course topic interesting.” and “I will probably understand the course content.”) on a 7-point Likert scale (1-low to 7-high, Cronbach’s alpha = 0.74). The learner’s current motivation after watching the instructional video was assessed by six items (e.g., “During the instructional video, I was in a kind of “flow”.”, “The lecture aroused my interest.”, “I didn’t even notice how the time passed.”) on a 7-point Likert scale (1-low to 7-high, Cronbach’s alpha = 0.83).

The learner’s well-being was measured by one item before watching the instructional video to control for differences between the experimental groups (“How comfortable do you feel today?”) and one item after watching the instructional video (“How comfortable did you feel during the lecture?”). It was also rated on a 7-point Likert scale (1-low to 7-high). After viewing the instructional video, participants were asked to complete a social presence questionnaire consisting of eight contrasting adjectives rated on a 7-point semantic differential (“personal/impersonal,” “cold/warm,” “interesting/boring,” “appealing/unappealing,” “activating/tiring,” “close/distant,” “comprehensible/incomprehensible,” “efficient/inefficient,” Cronbach’s alpha = 0.88). The social presence questionnaire was adopted from Gunawardena and Zittle (1997).

Cognitive measures

Three items measuring extraneous cognitive load, introduced by Klepsch and colleagues (2017), assessed extraneous processing during learning from the instructional videos. Participants rated their experienced extraneous cognitive load on a 7-point Likert scale (1-low to 7-high, Cronbach’s alpha = 0.80).

Two questions assessed prior knowledge. Participants were asked to indicate on a 7-point Likert scale (1-does not apply at all to 7-does fully apply) whether they were familiar with scientific knowledge about (a) intra- and intergroup conflicts and (b) social dilemmas.

Immediately after watching the instructional video and answering the social presence questionnaire, participants completed a learning test to assess what they had learned from the instructional video. The learning test consisted of eight single-select and five multiple-select multiple-choice retention questions (e.g., “How can social dilemmas be counteracted?,” maximum score = 20 points).

Preference

At the end of the study, participants were asked to indicate their preference for instructional videos: Do they prefer to watch instructional videos with the instructor’s video or instructional videos without a visible instructor? They could also indicate that they had no preference.

Procedure

The presented study was conducted as a field study. Students enrolled in the course on social psychology were emailed a link to the online survey. All participants completed a demographic questionnaire as well as items assessing their initial motivation, well-being, and prior knowledge. They were then randomly assigned to either the no visible instructor condition or the visible instructor condition. Immediately after watching the 34-min instructional video, participants completed the social presence questionnaire and the retention test. They also reported their experienced extraneous cognitive load, current motivation, and well-being during the lecture. Finally, they indicated their preference for instructional videos with or without a visible instructor.

Results

Total scores were calculated for the affective measures (initial/current motivation, social presence) and the cognitive measures (extraneous cognitive load, retention test). Mean scores and standard deviations for the dependent measures for the visible instructor and no visible instructor group are reported in Table 4.

Table 4 Mean scores and standard deviations of the dependent affective and cognitive measures in Study 1
Full size table

Pretest measures

The study participants reported moderate to high levels of initial motivation and well-being at the beginning of the study. A t-test yielded no significant differences between the experimental conditions (initial motivation: t (16) = 1.30, p = .21; well-being: t (16) = − 0.54, p = .60). Participants also reported low levels of prior knowledge about the content of the instructional video (M = 2.78, out of a possible 7, SD = 1.42). Subjective ratings of prior knowledge did not differ between experimental conditions (t (16) = 0.57, p = .58).

Affective post-test measures

T-tests were conducted to determine if the condition affected the learners’ experienced social presence, current motivation, and well-being during the lecture. The t-tests revealed no significant difference between the experimental conditions for current motivation (t (16) = 1.05, p = .16, d = 0.50) and well-being (t (16) = 1.67, p = .057, d = .79). They showed a significant difference for social presence (t (16) = 1.80, p = .045) with a large effect size (d = 0.85) indicating that learners who watched an instructional video with a visible instructor experienced higher social presence than learners who only heard the instructor’s voice.

Cognitive post-test measures

Participants reported low levels of extraneous cognitive load in both the visible instructor and the no visible instructor condition (MVI = 1.97, MNVI = 2.00 out of 7). According to the results of a t-test, the subjective ratings of extraneous processing did not differ significantly between the experimental groups (t (16) = − 0.07, p = .94). In addition, participants scored 15.5 (no visible instructor) and 16.5 (visible instructor) out of 20 on the retention test. The t-test conducted revealed no significant difference between the experimental conditions (t (16) = 1.36, p = .10, d = 0.65), indicating that learners who were presented with a visible instructor did not show higher retention than learners who were not presented with a visible instructor.

Preference

Two-thirds of the participants in this study indicated a preference for instructional videos with a visible instructor (n = 12, 66.67%). One person indicated a preference for instructional videos without a visible instructor (5.6%), and 5 participants indicated no preference for either option (27.8%). A chi-square test revealed no significant differences between the two experimental groups (X2 (2, N = 18) = 2.61, p = .27). Learners in the visible instructor and no visible instructor conditions did not differ in their preferences.

Summary Study 1

In line with Hypothesis 1, Study 1 indicates that a visible instructor in an instructional video led to higher ratings of social presence than an instructional video without a visible instructor. Hence, showing an instructor who is known to the learners in instructional videos embedded in university courses may increase the perceived social presence. However, contrary to Hypotheses 2 and 3, current motivation and well-being ratings did not differ between the experimental groups. Interestingly, participants in both experimental conditions reported very low ratings of extraneous processing, contradicting the Interference Hypothesis, which suggests that the visible instructor imposes higher levels of extraneous cognitive load (e.g., Colliot & Jamet, 2018). Contrary to Hypothesis 5, learning outcomes did not differ between experimental conditions. Thus, Study 1 provided initial exploratory results in higher education courses, but did not explicitly compare embedded and non-embedded instructional videos. In addition, the sample size was very small due to the field setting. Therefore, we replicated Study 1 in two further studies conducted in the same semester but in different courses at two different universities. Study 2 also explicitly examined the role of the learning setting by comparing instructional videos embedded and not embedded in higher education courses. In the following section, we present these two studies one after the other.

Study 2 replicated Study 1 with a larger sample size and further extended the experimental design in the following ways: (1) We additionally examined the role of the relevance of the learning task, resulting in a 2 × 2 design with the factors (a) instructor visible vs. not visible and (b) embedded vs. not embedded in a course. Embedded in a course refers to an instructional video that is part of a course in which students are enrolled and relevant to their exams, presented by an instructor known to the students. The term “not embedded in a course” refers to an instructional video whose content is not relevant to students’ exams and is not part of a course in which they are enrolled. In addition, learners in this condition did not know the instructor from other courses in which they were enrolled. (2) To assess motivation, we used Isen and Reeve’s (2005) questionnaire on intrinsic motivation rather than the items used in Study 1, which are more related to the flow concept (Csikszentmihalyi, 1975, 2010). (3) We used multiple-choice questions rather than subjective ratings to assess prior knowledge. (4) We extended the learning test, which included only retention items in Study 1, to include transfer items in Study 2.

Study 2

Hypotheses

Supplementing the hypotheses in Sect. "Hypotheses," we expected the learning setting (embedded vs. not embedded in a course) to have a main effect on affective outcomes (social presence, motivation, well-being) due to the personal relevance of the instructor.

H6

Learners who watch an instructional video embedded in a course in higher education report higher social presence than learners who watch a video that is not embedded in a course.

H7

Learners who watch an instructional video embedded in a course in higher education report higher motivation than learners who watch a video that is not embedded in a course.

H8

Learners who watch an instructional video embedded in a higher education course will report higher levels of well-being than learners who watch a video that is not embedded in a course.

Furthermore, we did not expect a main effect of learning setting on extraneous processing because the learning material itself and its cognitive demands did not differ between conditions (embedded vs. not embedded in a course).

H9

Learners who watch an instructional video embedded in a higher education course report the same level of extraneous processing as learners who watch a video that is not embedded in a course.

Finally, we expected a main effect of learning setting on learning outcomes due to higher motivation resulting from the exam-relevant topic of the instructional video.

H10

Learners who watch an instructional video embedded in a higher education course will have better learning outcomes than learners who watch a video that is not embedded in a course.

Method

Materials

Study 2 used the same materials as Study 1 (see Fig. 1): a 34-min instructional video on group conflicts with either a visible instructor or no visible instructor. The auditory explanations and the PowerPoint slides presented did not differ between the experimental conditions.

Participants and design

Participants in Study 2 were 53 students at a German university of applied sciences (47 female, four male, and two diverse). Their mean age was 21.8 years (SD = 3.81). Twenty-nine participants were psychology majors who watched the instructional video as part of a 14-week social psychology course relevant to their exams. The study was carried out in weeks eight and nine of the 14-week lecture. The sample consisted of all 29 students enrolled in the social psychology course. These participants knew the instructor from various courses in the previous semester. Thus, they form the “embedded in a course” group. The other 24 participants were students in other social science programs in the same faculty. The video was not part of a course they were enrolled in and was irrelevant to their exams. They did not know the lecturer from previous courses. Therefore, they form the “not embedded in a course” group. Both groups (embedded vs. not embedded in a course) were randomly assigned to either the visible instructor condition or the no visible instructor condition, resulting in four experimental groups in a 2 × 2-design: embedded in a course/visible instructor (EC/VI, n = 12), embedded in a course/no visible instructor (EC/NVI, n = 17), not embedded in a course/instructor visible (NEC/VI, n = 12), and not embedded in a course/no instructor visible (NEC/NVI, n = 12).

Measures

Affective measures

Study 2 used the same measures of initial motivation, well-being, and social presence as Study 1. We introduced another measure of learners’ current motivation after watching the instructional video by using Isen and Reeve’s (2005) intrinsic motivation questionnaire, which consists of 7 items rated on a 7-point Likert scale (1-low to 7-high, Cronbach’s alpha = 0.91). Two of the seven items were identical to the items used in Study 1. We chose to use Isen and Reeve’s (2005) intrinsic motivation questionnaire in Studies 2 and 3 because it focuses more on enjoyment and interest rather than flow (Csikszentmihalyi, 1975, 2010) - whether or not the learner is absorbed in the learning task, as assessed in Study 1. Furthermore, Isen and Reeve’s questionnaire on intrinsic motivation has been used previously in multimedia learning research (e.g., Liew et al., 2017; Um et al., 2012).

Cognitive measures

In Study 2, we used the same measures for extraneous processing as in Study 1. Prior knowledge was assessed by two single-select and two multiple-select multiple-choice questions (e.g., “What is meant by “positive interdependence?,” maximum score attainable = 8 points) rather than subjective rating as in Exploratory Study 1. We revised the retention test based on the task difficulties of the retention questions obtained in Study 1 by omitting items with a difficulty index > 0.85. Thus, the retention test consisted of five single-select and three multiple-select multiple-choice questions (e.g., “What causes cooperation between groups to decrease?,” maximum score = 12 points). We also added transfer questions to assess learning outcomes more differentiated than in Exploratory Study 1. The transfer test consisted of four single-select and three multiple-select multiple-choice questions (e.g., “What is an example of a commons dilemma?,” maximum score = 11 points).

Procedure

Study 2 was again carried out as a field study. Participants were emailed a link to the online survey. They completed a demographic questionnaire and measures of initial motivation, well-being, and prior knowledge. They were then randomly assigned to watch the instructional video with or without the visible instructor. They watched the 34-min instructional video and completed measures of well-being, intrinsic motivation, and social presence. They then completed the learning test, which included retention and transfer tasks. Finally, they reported their experienced extraneous cognitive load and indicated their preference for instructional videos with or without a visible instructor.

Results

Total scores were calculated for the affective measures (initial/intrinsic motivation, social presence) and the cognitive measures (extraneous cognitive load, retention, and transfer test). Mean scores and standard deviations for the dependent measures for the four experimental conditions are reported in Table 5.

Table 5 Mean scores and standard deviations of the dependent affective and cognitive measures in Study 2
Full size table

Pretest measures

Study participants reported moderate to high levels of initial motivation and well-being. They had low prior knowledge (M = 1.32, out of eight possible points, SD = 1.54) about the topic of the instructional video. Three separate ANOVAs yielded no significant differences between the four treatment groups on initial motivation, well-being, or prior knowledge (initial motivation: F(3, 49) = 0.61, p = .61; well-being: F(3, 49) = 1.52, p = .24; prior knowledge: F(3, 49) = 1.81, p = .16).

Affective post-test measures

Separate two-factor ANOVAs with learning setting (embedded vs. not embedded in a course) and instructor presence (visible vs. no visible instructor) as factors were conducted to determine if learning setting and instructor presence affected the learners’ experienced social presence, intrinsic motivation, or well-being during the lecture. The two-factor ANOVA with social presence as the dependent measure revealed no main effect of learning setting (F(1, 49) = 0.03, p = .87), no main effect of instructor presence (F(1, 49) = 1.04, p = .31), and no interaction between learning setting and instructor presence (F(1, 49) = .59, p = .45). The two-factor ANOVA for intrinsic motivation also revealed no main effect of learning setting (F(1, 49) = 1.98, p = .17, ηp2 = 0.04) or instructor presence (F(1, 49) = 0.45, p = .51), and no interaction between learning setting and instructor presence (F(1, 49) = 0.16, p = .69). Finally, the two-factor ANOVA for well-being also revealed no main effects of learning setting (F(1, 49) = 0.75, p = .39) or instructor presence (F(1, 49) = 0.59, p = .45) and no interaction effect (F(1, 49) = 0.75, p = .39). Thus, learners did not report different levels of experienced social presence, intrinsic motivation, or well-being as a function of learning setting or instructor presence.

Cognitive post-test measures

A two-factor ANOVA with learning setting (embedded vs. not embedded in a course) and instructor presence (visible vs. no visible instructor) as factors and subjective ratings of extraneous processing as the dependent measure yielded no main effect for either learning setting (F(1,49) = 0.08, p = .78) or instructor presence (F(1,49) = 0.33, p = .57), and also no interaction between learning setting and instructor presence (F(1,49) = 2.05, p = .16, ηp2 = 0.04).

Considering the learning outcomes, separate two-factor ANOVAs with learning setting (embedded vs. not embedded in a course) and instructor presence (visible vs. no visible instructor) as factors were conducted to investigate their effects on retention and transfer performance. The ANOVA with retention as the dependent measure yielded no main effect of learning setting (F(1,49) = 1.92, p = .17, ηp2 = 0.04), no main effect of instructor presence (F(1,49) = 1.96, p = .17, ηp2 = 0.04) and no interaction effect (F(1,49) = 0.98, p = .33). The ANOVA with transfer as the dependent measure showed no main effect of learning setting (F(1,49) = 1.22, p = .28), no main effect of instructor presence (F(1,49) = 0.68, p = .41) and no interaction effect (F(1,49) = 0.30, p = .58). Therefore, learners in the four experimental conditions did not report different levels of extraneous processing. They also showed no differences in retention or transfer performance.

Preference

Three-quarters of participants preferred instructional videos with a visible instructor (n = 36, 76.6%). Ten participants had no preference whether the instructor was visible or not (21.3%), and only one person indicated a preference for instructional videos with no visible instructor (2.1%). A chi-squared test revealed no significant differences in learner preference between the visible instructor and no visible instructor conditions (X2 (6, N = 47) = 5.65, p = .46).

Summary Study 2

Study 2 was designed to replicate Study 1 with a larger sample and to examine the learning setting (embedded vs. not embedded in a course) as a separate factor. No significant effects on affective or cognitive dependent measures were found in this study. Contrary to Study 1, we did not find an effect of a visible instructor on experienced social presence with exam-relevant learning material presented by an instructor of personal relevance to the learners (Hypothesis 1). Hypotheses 2 and 3, expecting differences in motivation and well-being, and Hypotheses 4 and 5, expecting differences in extraneous processing and learning outcomes, were also not confirmed. The comparison of embedded and non-embedded instructional videos in higher education courses did not yield significant results for affective ratings (Hypotheses 6–8) or learning outcomes (Hypothesis 10). In line with our expectations, extraneous processing did not differ for videos embedded in courses compared to videos not embedded in courses (Hypothesis 9). Therefore, this study adds evidence that instructor presence may be an effect that is more likely to occur in controlled laboratory studies than in online studies. This seems to hold even if the online study is embedded in courses. A limitation is that the sample size of this study is still relatively small due to the field setting, which limits the statistical power of the analyses. The positive effect of a visible instructor on social presence was found in Study 1 and not in Study 2. However, both studies used the same instructional video, the same instructor, and the same setting. This may be due to the context of the studies. Study 1 was conducted during distance learning due to the COVID-19 pandemic lack of face-to-face instruction, whereas Study 2 was conducted in the following standard predominantly face-to-face semester.

Parallel to Study 2, we conducted Study 3, which was also set in a predominantly face-to-face semester. Study 3 aimed to replicate Study 1 with a different instructional video embedded in a course at a different university. Therefore, this study (1) was conducted at a different German university with (2) another instructional video that (3) covered a different topic and was presented by (4) a different instructor who was known to the learners from a few lessons at the beginning of the semester. In addition, (5) the instructor’s video was integrated via a green screen (Fig. 2). As Study 3 aims to replicate the results of the exploratory Study 1, it relies on the same hypotheses as stated in Sect. “Hypotheses”.

Fig. 2
figure 2

Instructional video used in study 3 without visible instructor (left) and with visible instructor (right)

Full size image

Study 3

Method

Materials

The instructional video used in this study was also part of a 14-week lecture, but different from Studies 1 and 2, on the topic of “Design Recommendations for Multimedia Learning”. The video consisted of a 48-min presentation of design recommendations based on multimedia learning research relevant to the participants’ exams. Two versions of the video were produced. Both showed the same PowerPoint presentation synchronized with the instructor’s auditory narration. The auditory explanations did not differ between the experimental groups. In the visible instructor condition, the instructor’s video was integrated into the slides via a green screen (Fig. 2). Integrating the instructor’s video into the PowerPoint slides would sometimes cover relevant text on the slides. Therefore, we produced the video, including the instructor’s video, in a 16:9 format (Fig. 2). The instructional videos presented were system-paced.

Participants and design

This study involved 38 college students majoring in media communication at a German university (29 female, eight male, one diverse). The mean age was 21.53 years (SD = 3.51). Students were randomly assigned to either the visible instructor condition (n = 21) or the no visible instructor condition (n = 17). The study was conducted during the second week of the 14-week lecture.

Measures

Affective measures

Study 3 used the same measures of initial motivation, intrinsic motivation, well-being, and social presence as Study 2.

Cognitive measures

We also used the same measure of extraneous processing as in Studies 1 and 2. Other measures of prior knowledge, retention, and transfer performance were developed to match the content of the instructional video. As in Study 2, we applied a prior knowledge test rather than subjective ratings. It consisted of 4 open-ended questions (e.g., “What are independent and dependent variables in an experimental context?,” maximum score = 16 points). The learning test consisted of retention and transfer questions. Retention was assessed by 13 multiple-choice questions (e.g., “Which statements about dependent variables are correct?”) and one open-ended question (maximum score = 64 points). Transfer was assessed by four multiple-choice questions (e.g., “For which of these hypotheses is the assignment of dependent and independent variables correct?”). In addition, there were five open-ended questions (e.g., “How might we reduce the split-attention effects outlined in real learning situations?,” maximum score = 28 points). Participants’ responses to the open-ended questions in the prior knowledge and learning test were scored by two independent raters. The interrater reliability was ICC = 0.89 with 95% confidence interval = 0.85–0.92, based on a mean rating (k = 2), absolute agreement, 2-way random effects model. Thus, the level of interrater reliability can be considered good to excellent (cf. Koo & Li, 2016).

Procedure

Study 3 followed the same procedure as Study 2.

Results

Total scores were calculated for the affective measures (initial/intrinsic motivation, social presence) and the cognitive measures (extraneous cognitive load, retention, and transfer test). Means and standard deviations for the dependent measures in the visible instructor and no visible instructor condition are reported in Table 6.

Table 6 Mean scores and standard deviations of the dependent affective and cognitive measures in Study 3
Full size table

Pretest measures

Participants in the study had very little prior knowledge (M = 2.05, out of a possible 16, SD = 1.48) on the topic of the learning environment. One participant with exceptionally high prior knowledge was excluded from the analysis (M = 11.5). Further, the participants reported moderate to high levels of well-being and initial motivation. The t-tests conducted revealed no significant differences in well-being, initial motivation, or prior knowledge between the visible instructor and the no visible instructor group (well-being: t (36) = 0.80, p = .43, initial motivation: t (36) = 0.58, p = .57, prior knowledge: t (35) = − 0.93, p = .36).

Affective post-test measures

T-tests were conducted to investigate whether instructor visibility affected learners’ perceived social presence, intrinsic motivation, or well-being during the lecture. The t-tests revealed no significant difference between the experimental conditions for experienced social presence (t (36) = 1.10, p = .14, d = 0.36) and intrinsic motivation (t (36) = 0.15, p = .44). For well-being, the t-test showed a significant difference between the experimental groups (t (36) = 3.11, p = .002) with a large effect size (d = 1.01). Learners who watched an instructional video with a visible instructor reported higher levels of well-being than those who heard only the instructor’s voice.

Cognitive post-test measures

Participants in both the visible instructor and the no visible instructor condition reported relatively low levels of extraneous cognitive load (MVI = 2.35, MNVI = 3.04 out of 7). The conducted t-test showed no significant differences between the experimental conditions (t (36) = − 1.76, p = .088, d = 0.57). Contrary to the Interference Hypothesis, learners in the instructor visible condition even tend to report lower levels of extraneous cognitive load than learners in the no instructor visible condition. In terms of learning outcomes, participants scored 46.48 points (instructor visible condition) and 47.00 points (no instructor visible condition) out of 64 points on the retention test. In the transfer test, they scored 16.14 (instructor visible condition) and 16.60 points (no instructor visible condition) out of 28 points. The t-tests conducted revealed no significant difference between the experimental conditions in retention (t (35) = − 0.29, p = .39) or transfer performance (t (35) = − 0.36, p = .36). Therefore, learners who were shown an instructional video with a visible instructor did not show better retention or transfer performance than learners who were shown an instructional video without a visible instructor.

Preference

More than half of the participants in this study indicated a preference for instructional videos with a visible instructor (n = 20, 52.6%). Nearly 40% reported no preference for a visible or no visible instructor (n = 15, 39.5%). Only 3 participants indicated a preference for instructional videos with no visible instructor. (7.9%). A chi-square test revealed no significant differences in the preferences of learners in the visible instructor condition compared to learners in the no visible instructor condition (X2 (2, N = 38) = 0.51, p = .77).

Summary Study 3

The results of Study 3 show that having a visible instructor in a course-embedded instructional video led to higher well-being ratings than not having a visible instructor (confirming Hypothesis 3). However, Hypotheses 1 and 2, which expected positive effects of the visible instructor on social presence and motivation, could not be confirmed. Thus, the beneficial effects of a visible instructor on affective outcomes could be demonstrated in Study 1 for social presence (and in a trend for well-being) and in Study 3 for well-being in a predominantly face-to-face semester. Hence, they may occur in laboratory settings and online studies when embedded in higher education courses. However, the studies did not show consistent effects on the different affective measures. Therefore, the results only partially support the Social Cues Hypothesis. In Study 3, participants again reported relatively low levels of extraneous cognitive load and even a trend toward lower levels of extraneous load in the instructor visible condition, challenging the Interference Hypothesis. In terms of learning outcomes, contrary to Hypothesis 5, neither retention nor transfer performance differed between experimental conditions.

General discussion

Theoretical implications

The primary objective of the presented field studies was to determine whether an effect of instructor presence in instructional videos is more likely to be obtained when the instructional video is embedded in higher education courses. Therefore, we added an important distinction that is more specific than the distinction between laboratory and online studies only, as done in the meta-analyses of Alemdag (2022). In addition, we used instructional videos of more than 30 min in length, which are more similar to a lecture in higher education than concise instructional videos of less than ten minutes in length, which have been investigated in most studies of instructor presence so far (e.g., Kokoç et al., 2020; Wang et al., 2020a, 2020b; van Wermesmerken et al., 2018; Zhang et al., 2021).

The theoretical framework for addressing this question was social agency theory (Mayer & DraPra, 2012; Mayer et al., 2003), as well as CLT (Sweller, 1988, 2020) and CTML (Mayer, 2001, 2021). This results in two contradictory hypotheses that are often used in instructor presence: Social Cues versus Interference Hypotheses. For the affective measures, the Social Cues Hypothesis was only partially supported by the results of Studies 1 and 3. We found a positive effect of a visible instructor compared to showing no instructor on social presence with a large effect size (d = 0.85) and a trend on well-being in Study 1 (p = .057, d = 0.79) and a positive effect on well-being in Study 3 (d = 1.01). However, the other affective measures did not differ between the experimental conditions, and we found no significant differences in Study 2. By using instructional videos embedded in courses, we not only presented the visible instructor as an additional social cue, but also increased the valence of this social cue (Domagk, 2010) by presenting instructors who were known to the learners. These instructors are personally relevant to learners through self-disclosure and relationship building, which is preferred by distant students, as shown by Reupert and colleagues (2009). Thus, in contrast to the current review (Henderson & Schroeder, 2021), we found effects on social presence (Study 1) and well-being (Study 3, trend in Study 1), but still not on other affective measures or consistently across all three studies. Contrary to the results of the meta-analysis (Alemdag, 2022), we found no effects on learner motivation. However, Alemdag (2022) subsumed satisfaction under motivational ratings, which was not assessed in the present studies and is conceptually closer to well-being than to motivation.

Considering the cognitive measures, the results for extraneous processing of all three studies do not support the Interference Hypothesis, as no significant differences were found between the experimental conditions. Study 3 showed the opposite trend of less extraneous processing in the visible instructor condition. This result was also found by Wang and colleagues (2020a, 2020b), but is contradicted by the significant effect of instructor presence on cognitive load (Hedges’ g = 0.319) found in the meta-analysis (Alemdag, 2022). The latter might be due to a publication bias (Fanelli, 2012). These mixed results could be due to the use of differently visually rich learning materials in the studies. The PowerPoint slides in the presented studies did not impose high demands on visual processing as they consisted mainly of short texts. This may indicate that the social cues used in the studies (primarily the face) do not strongly influence the measured mental strain. This is supported by other research, which states that reading nonverbal information could be categorized as biological communication (Kirschner et al., 2018), which does not induce a high level of cognitive load.

Both hypotheses predict an effect of the visible instructor on learning outcomes, but in opposite directions. According to the Social Cues Hypothesis, the visible instructor may lead to more generative processing of the learning material, resulting in better learning outcomes. The Interference Hypothesis suggests that the visible instructor interferes with learning through increased extraneous processing and split attention (Ayres & Sweller, 2021; Tarmizi & Sweller, 1988). We found no evidence for either one, as we did not find effects on learning outcomes in any of the three studies. This is in line with the results of the current review (Henderson & Schroeder, 2021) and meta-analysis (Alemdag, 2022), which also report no significant differences in most studies. In our studies, even when embedded in courses, the visible instructor neither facilitated nor detracted from learning. Thus, the possible imposed changes in learners’ affective and cognitive processes due to instructor presence may be too small to be reflected in learning outcomes.

Limitations and future directions

The studies presented examined instructor presence in course-embedded instructional videos, transferring them from laboratory and online studies with mostly arbitrary content to exam-relevant learning settings in higher education. However, due to the field setting, the study samples were relatively small and predominantly female, limiting the statistical power of the analyses and the generalizability of the results. Future studies are needed to further explore instructor presence in instructional videos embedded in courses with larger samples and diverse learners to determine whether instructor presence can enhance affective and cognitive processes for specific learners. Previous research has shown that the positive effects of a visible instructor can be demonstrated in some studies, but not in others. Therefore, research interest should shift from investigating a general effect to investigating under what conditions visible instructors in instructional videos are beneficial for affective processes and learning outcomes. Our investigation of instructor presence in course-embedded instructional videos focused primarily on the characteristics of the learning environment. There are many other relevant conditions for the use of instructional videos in educational settings that can be considered in future research, such as the topic to be learned, the length of the video, the characteristics of the learner (e.g., prior knowledge, interest), the characteristics of the instructor’s video (e.g., size, position), and the characteristics of the instructor himself or herself, including likability, competence, and whether or not he or she is known to the learner. Analogous to the development in the field of pedagogical agents, these conditions should be systematically considered. For example, Zhang and colleagues (2021) investigated the position of the instructor’s video, and Yuan and colleagues (2021) considered the technical adjustment of the instructor’s image and voice. Since pedagogical agents are on-screen characters presented to introduce social cues that may increase learner motivation and learning outcomes, this line of research follows a similar logic to research on instructor presence in instructional videos. Henderson and Schroeder (2021) also pointed this out. Within the research on pedagogical agents, Heidig and Clarebout (2011) proposed two models that systematically consider these variables: the Pedagogical Agents—Conditions of Use model (PACU) and the Pedagogical Agents—Levels of Design model (PALD). Adjusted models for instructor presence in instructional videos are not yet available and would help to systematize previous results and future studies.

Practical implications

Given these limitations of the presented studies, some practical implications can be derived. Consistent with previous empirical evidence (e.g., Alemdag, 2022; Henderson & Schroeder, 2021), the studies presented did not show positive effects of instructor presence on learning outcomes and mixed results for affective measures. The three studies also showed no detrimental effects on learning outcomes or affective measures. Therefore, instructor presence in instructional videos (not only in the laboratory, but also in online studies and embedded in courses) may facilitate affective processes in some cases, but at least is unlikely to harm learning. From a practical point of view, instructors in formal educational settings such as higher education may choose to show themselves in instructional videos as negative effects are unlikely and positive affective effects may occur. This recommendation is further supported by the finding that most learners prefer instructional videos with a visible instructor (53–77%), 20 to 40% have no preference, and only a few prefer videos without a visible instructor (2–8%). This is consistent with the findings of Wilson and colleagues (2018), who found that learners prefer instructional videos with visible instructors and believe they learn more effectively when the instructor is present. However, including the visible instructor requires additional production effort (time, technical and software equipment, post-production expertise) that must be weighed against the potential benefits.

Conclusion

Research on instructor presence in instructional videos has yielded mixed results on affective outcomes and mostly no effects on learning outcomes (e.g., Alemdag, 2022; Henderson & Schroeder, 2021). The studies presented aimed to increase external validity by examining instructor presence in instructional videos embedded in higher education courses. The results are largely consistent with previous findings. Future research should therefore investigate the conditions under which visible instructors in instructional videos may be beneficial for affective processes and learning outcomes. Analogous to research on pedagogical agents that follows a similar logic, systematic models of the conditions of use and design of visible instructors in instructional videos are needed to clarify previous findings and inform future research,  teaching and design practice.