A zoom teaching experiment using CTML principles of multimedia design

ABSTRACT The purpose of this study was to examine how online synchronous teaching using Zoom might be enhanced by incorporating multimedia principles from the cognitive theory of multimedia learning (CTML). A live lecture teaching experiment was conducted where students were randomly assigned to attend a standard lesson on Zoom (control condition) or the same lesson but with the multimedia principles (signaling, embodiment, and generative activity principles) applied throughout lesson (treatment condition). Results revealed a direct effect of teaching with CTML principles on students’ learning, but there was no evidence for an indirect effect through working memory overload. Additionally, students had more affect toward the (same) instructor who taught with added CTML principles. Pedagogical implications and advice are offered for instructors who teach using Zoom.

platform Zoom (Stafford, 2020). Using Zoom, individuals interact digitally by joining online meeting rooms with unique multi-digit identification numbers. Users connect to meeting rooms using various devices (e.g., desktop computer, laptop computer, mobile phone, tablet), and once connected, can utilize numerous program features to communicate such as high-definition audio and video, screen-share, white-boarding, annotation, breakout rooms, virtual backgrounds, in-meeting chat, local recording, and nonverbal feedback (Zoom, n.d.). The flexibility Zoom provides its users has made it a popular interface for online instruction -expanding Zoom's daily user base from 10 million users in 2019 to 300 million users in 2020, largely due to its implementation across colleges and universities (Dean, 2022).
However, despite Zoom's comprehensive functionality and relatively intuitive user interface -as well as the litany of positive reviews it has received from educators -video-communication software cannot compensate for a teacher's lack of effective multimedia instructional design. Indeed, while the integration of video and audio has the potential to substantially enhance student learningparticularly in remote instruction -instructors must be deliberate to employ multimedia technologies in a manner aligned with how students learn. Given this, the purpose of this study was to examine if causal effects of longstanding principles of multimedia design might be strategically incorporated into online instruction via Zoom in a manner conducive to fostering students' learning. To achieve this goal, we conducted a live lecture teaching experiment based on the cognitive theory of multimedia learning (CTML, Mayer, 2001) to identify specific ways college instructors might enhance their online pedagogy using the technological features of Zoom.

The cognitive theory of multimedia learning (CTML)
CTML (Mayer, 2001), succinctly, is concerned with how individuals learn from words and pictures. CTML is based on the premise that individuals generally process information more deeply when words and pictures are presented together than when they are presented in isolation, an idea which Mayer (2001) described as the multimedia principle. However, Mayer (2005a) also cautioned that "simply adding pictures to words does not guarantee an improvement in learning -that is, all multimedia presentations are not equally effective" (p. 31). In fact, presenting words and pictures in certain ways (e.g., simultaneously presenting words and pictures conveying distinct pieces of information, presenting words and pictures in a disorganized manner) may ultimately impede learning rather than enhance it -undermining students' ability to efficiently process information (e.g., Harp & Mayer, 1998;Mayer et al., 2001). One of the foremost goals of CTML is to guide the development of effective multimedia messages based on how the human brain functions and the ways in which individuals cognitively process information (Mayer, 2001), facilitating the design of instructional multimedia messages based on the optimal ways that people learn. CTML's focus on how individuals learn is reflected in Mayer's (2001) conceptualization of a multimedia instructional message as "a presentation involving words and pictures that is intended to foster learning" (p. 3). From a perspective grounded in CTML, learning is an ongoing activity in which individuals engage in knowledge construction; an active process whereby learners create mental representations of new instructional material presented to them (Mayer, 2001). Mayer (2001Mayer ( , 2005a emphasized the importance of instructors being strategic in their implementation of effective multimedia instructional designs. Effective multimedia instructional designs are those which enable instructors to achieve three key goals: (1) assist learners in reducing extraneous cognitive processing, (2) support learners in managing essential cognitive processing, and (3) help facilitate learners' generative cognitive processing (Mayer, 2021). To that end, Mayer (2021) presents best instructional practices in multimedia contexts and identified a series of 15 instructional principles conducive to achieving those goals. We focus on three of these principles that are particularly applicable to online instruction using Zoom: the signaling principle, embodiment principle, and generative activity principle.

The signaling principle
The signaling principle suggests instructors can assist learners in reducing extraneous processing by employing effective organizational structure in their lessons -or, that "people learn more deeply from a multimedia message when cues are added that highlight the organization of essential material" (Mayer, 2005b, p. 184). In scenarios where extraneous material cannot be outright removed from lessons, the signaling principle suggests a viable solution is to integrate cues within those lessons which draw learners' attention away from extraneous details and toward essential material (Mayer, 2021). Instructors can apply the signaling principle in various ways while teaching in multimedia environments, such as through verbal signaling (i.e., directing attention to specific verbal information) or visual signaling (i.e., directing attention to specific pictures or graphics). Examples of verbal signaling include classic signaling (i.e., directing attention to specific information through the use of traditional organizational cues such as outlines, headings, and pointer words; Harp & Mayer, 1998;Mautone & Mayer, 2001), spatial outlines (i.e., visual arrangements of important words into organizational structures such as matrices, charts, or hierarchies; Ponce & Mayer, 2014a, 2014bStull & Mayer, 2007), and highlighting (i.e., manipulating the font or color of particular words or emphasizing particular words when spoken aloud; Mayer, 2021;Ponce & Mayer, 2014a). Examples of visual signaling include instructors' use of distinctive colors to emphasize particular parts of a graphic (Mautone & Mayer, 2001;Wang et al., 2018;Xie et al., 2019), coordinating visual and verbal cues so that graphical changes coincide with the presentation of verbal information (Xie et al., 2019), and specific cues such as pointing gestures, arrows, flashing, and graying out text (Li et al., 2019;Mayer, 2021;Wang et al., 2018). While signals do not add new information to lessons, they add organization and structure to guide learners' attention (Mayer, 2005b). Through signaling, instructors may serve as guides for learners' cognitive processing by drawing attention to important instructional material and assisting learners in selecting relevant information and organizing that information into coherent mental representations. Mayer (2021) points out that "without guidance in how to carry out appropriate cognitive processing, the learner is more likely to engage in extraneous cognitive processing -such as processing extraneous material and trying to organize it with the rest of the material" (p. 171).

The embodiment principle
The embodiment principle, originally encapsulated within the image principle (Mayer, 2005c), asserts learners understand multimedia lessons better when they can see an instructor or on-screen instructional agent who engages in body movement, gestures, eye contact, and facial expressions (Mayer, 2014a). As articulated in the revised image principle (Mayer, 2014a), simply being able to see an image of an instructor is not necessarily enough to cultivate a sense of social presence among learners in multimedia lessons. Rather, social presence is aroused when learners can observe their instructors engaging in behaviors which Mayer (2014a) described as highembodiment (i.e., behaviors similar to those used during in-person interactions in the real world). Elaborating further, Mayer (2021) wrote that examples of high-embodiment "include using hand gestures while talking . . . maintaining eye contact while talking . . . drawing graphics by hand while talking . . . or manipulating objects from a first-person perspective," ultimately characterizing high-embodiment behaviors as "the ways that onscreen instructors can use their bodies to enhance the act of instructional communication" (p. 344). When instructors in multimedia lessons engage in high-embodiment behaviors, Mayer (2021) explained that they provide "a positive social cue that primes a sense of social partnership in the learner, causing the learner to try harder to understand the instructional message and thereby learn more deeply" (p. 341). That is, when instructors are nonverbally animated and employ movement during multimedia lessons, they can motivate learners to engage in generative processing conducive to meaningful learning.
Unlike Mayer's (2005c) original articulation of the image principle, experimental research has consistently supported the embodiment principle (Mayer, 2014a). Studies have found that the extent to which instructors engage in nonverbal gestures (e.g., Li et al., 2019;Wang et al., 2018), maintain eye contact with learners (e.g., Fiorella et al., 2019Fiorella et al., , 2020, and physically draw graphics (e.g., Fiorella & Mayer, 2016a) during multimedia lessons can significantly enhance learners' performance on post-lesson assessments of both retention and transfer. Altogether, the available literature generally suggests that learners' social presence is enhanced when their instructors are nonverbally animated, catalyzing social responses conducive to motivating learners to engage in generative processing (Mayer, 2021).
Mayer's (2014a) notion of embodiment resembles Andersen's (1979) definition of nonverbal immediacy (i.e., "nonverbal behaviors that reduce physical or psychological distance between teachers and students" [p. 543]). Instructional communication research has suggested that instructors who engage in nonverbal immediacy behaviors (e.g., smiling, making eye contact, moving around the classroom) enhance their students' motivation (Frymier, 1994), participation (Rocca, 2009), and positive affective experiences during learning (Witt et al., 2004). In particular, nonverbal immediacy has been suggested to enhance teacher-student relationships (Frymier & Houser, 2000) by satisfying students' innate relational needs, motivating students to engage more deeply with instructional content (Frymier et al., 2019). Mayer's (2014a) suggestion that instructors' high-embodiment behaviors can arouse learners' sense of social presence, increase motivation, and facilitate generative information processing during multimedia lessons, aligns with instructional communication literature concerning the influence of instructor nonverbal immediacy. More recent research exploring instructors' use of nonverbal immediacy behaviors in online teaching (e.g., Dixson et al., 2017) lends further credence to Mayer's (2014a) arguments, demonstrating that the implications of instructors' nonverbal behaviors are not exclusive to face-to-face instruction.

The generative activity principle
The generative activity principle suggests that learners understand instructional content better when their instructors take the time to guide them through activities which stimulate generative processing (Mayer, 2021). Given that CTML (Mayer, 2001) is based on the assumption that meaningful learning is an active process, Mayer (2021) characterized activities which prompt learners to engage in active learning as inherently conducive to generative processing. Specifically, when instructors provide learners with opportunities to summarize, map, draw, imagine, self-test, selfexplain, teach, or enact instructional content, they stimulate learners' selection of important information, organization of information into coherent mental representations, and integration of mental representations with prior knowledge already stored within the long-term memory (Mayer, 2021). Generative activities thus refer to specific tasks that learners engage in during multimedia lessons with the intention of promoting deeper, more meaningful learning. Summarizing instructional content, for example, is a generative activity in that it requires learners to "select information to put into [their] summary, organize it into a coherent set of sentences, and integrate it with prior knowledge by putting it in [their] own words" (Mayer, 2021, p. 372). Similarly, mapping, drawing, imagining, selftesting, self-explaining, teaching, and enacting encourage learners to reflect more deeply on information presented in multimedia lessons in ways which stimulate generative processing (Mayer, 2021). Generative activities are thus a mechanism through which instructors can "clearly guide and scaffold the learner's generative activity" (Mayer, 2021, p. 376), assuming that they have already employed an instructional design conducive to reducing extraneous processing and assisting learners in effectively managing essential processing. Mayer's (2021) generative principle is well-supported by empirical research. Indeed, a number of experimental studies have consistently demonstrated that learners' generative processing is enhanced in multimedia lessons where they are directly prompted to engage in generative activities -whether by summarizing (e.g., Parong & Mayer, 2018), mapping (e.g., Ponce & Mayer, 2014a;Stull & Mayer, 2007), drawing (e.g., Fiorella et al., 2020), self-testing (e.g., Johnson & Mayer, 2009), self-explaining (e.g., Fiorella et al., 2020;Johnson & Mayer, 2010), teaching (e.g., Fiorella & Mayer, 2013), or enacting (e.g., Fiorella et al., 2017). Although Mayer (2021) urged instructors in multimedia contexts to be cautious in prompting learners to engage in generative activities before they are ready (i.e., when information is highly complex or learners lack sufficient foundational understanding or skill) or in a manner which learners might perceive as tedious (e.g., Johnson & Mayer, 2010;Stull & Mayer, 2007), he concluded that generative activities may represent the most direct pedagogical strategy through which instructors can stimulate learners' generative processing in multimedia lessons.

Theoretical rationale
Given the widespread transition to online instruction in the wake of the COVID-19 pandemic (Carillo & Flores, 2020;Stafford, 2020), the purpose of this study was to conduct a live lecture teaching experiment testing how CTML principles of effective instructional design might be applied to specific features of Zoom given its popularity as a remote teaching platform. The signaling principle, for example, seems particularly applicable to online lessons in which instructors utilize Zoom's annotation function to draw students' attention to specific instructional content. As instructors use Zoom's screen-share feature to present content displayed on their own computer screens to students, the annotation function allows them to draw various graphics (e.g., arrows, circles), place "stamps" (e.g., stars, exes), and employ "spotlights" to point at, highlight, or otherwise emphasize particular information (Zoom, n.d.). In these ways, Zoom's annotation feature affords instructors a potential means by which they might utilize to assist students in selecting information relevant to instructional goals.
Similarly, the embodiment principle can be implemented into Zoom teaching as simply as plugging in and turning on one's camera (Zoom, n.d.). Upon activating their cameras, instructors become visible to any students connected to a given Zoom chatroom, remaining visible even when using Zoom's screenshare feature (displayed in a smaller window in the corner of their students' screens). Zoom's inherent design thus enables students to see their instructors throughout online lessons, providing instructors with the opportunity to engage in the high-embodiment behaviors which Mayer (2014a) characterized as conducive to students' sense of social presence and generative processing.
Lastly, the generative activity principle may also be directly incorporated into online teaching using specific Zoom features. Polling, in particular, seems especially conducive to prompting students in online lessons to engage in the generative activities which Mayer (2021) suggested prime active learning and generative processing. Polling provides instructors with a means of soliciting student input or posing content-related questions to a class. Given this, instructors teaching via Zoom might use the program's polling feature to provide students with opportunities to self-test their understanding of recently presented information, scaffolding students' generative activity to facilitate appropriate selection, organization, and integration of instructional content.
In these ways, instructors teaching via Zoom might use specific program features strategically to incorporate the signaling, embodiment, and generative activity principles into their synchronous (i.e., live) online teaching -enhancing instructional clarity by using Zoom features to assist learners in selecting, understanding, and retaining information (Titsworth & Mazer, 2016). In contrast, Zoom features seem less conducive to implementing CTML principles into asynchronous (i.e., prerecorded) online teaching. Polling, for example, is designed to prompt feedback and participation from users during ongoing presentations (Zoom, n.d.). Inherently, then, the use of this feature is only applicable in live lessons where students are presented with instructional content simultaneously. Given the extent to which Zoom's features are designed to stimulate student activity in real-time (Zoom, n.d.), they seem more aligned with synchronous, rather than asynchronous, online teaching.
In light of the abundant empirical research already suggesting that implementation of the signaling (e.g., Bolkan, 2017b;Li et al., 2019;Mautone & Mayer, 2001), embodiment (e.g., Li et al., 2019;Wang et al., 2018), and generative activity principles (e.g., Fiorella & Mayer, 2013;Parong & Mayer, 2018) can significantly enhance students' learning in multimedia lessons by freeing up working memory constraints (working memory overload), we anticipated that instructors' efforts to incorporate these principles using specific Zoom features would similarly enhance student learning. Given that one of the fundamental purposes for which CTML (Mayer, 2001) was developed was to guide instructors in designing their multimedia presentations in ways which do not overwhelm learners' finite information processing capacity, we expected that students in an online lesson where an instructor enacts CTMLbased principles should experience less working memory overload than in an online lesson where an instructor does not. That is, our first hypothesis predicted that students participating in a Zoom lesson with an instructor who strategically uses Zoom features to enact the signaling, embodiment, and generative activity principles would experience less working memory overload than students in a Zoom lesson with an instructor who does not.
H: Compared to a standard synchronous lesson without Zoom features, students will experience less working memory overload and, in turn, exhibit greater learning in a synchronous lesson when their instructor uses Zoom features to enact the signaling, embodiment, and generative activity principles. Our second hypothesis predicted that students participating in a synchronous lesson with an instructor who strategically uses Zoom features to enact CTML principles would also exhibit greater affect for their instructor compared to students in a lesson with an instructor who does not. Recall that the theoretical rationale for the embodiment principle is that students in multimedia lessons who can see their instructors engaging in body movements are more likely to positively identify with those instructors, thereby stimulating students' sense of social presence (Mayer, 2014a). As reviewed previously, there is ample empirical evidence to suggest this is the case, such that instructors' use of gestures (e.g., Li et al., 2019;Wang et al., 2018) and eye contact (e.g., Fiorella et al., 2019Fiorella et al., , 2020 have been found to facilitate students' sense of social presence in a manner conducive to enhanced learning during multimedia lessons. Similarly, instructional communication research also suggests that instructors' nonverbal behaviors can significantly influence students' learning experiences, with instructors who are nonverbally immediate enhancing students' motivation (Frymier et al., 2019;Frymier, 1994), participation (Rocca, 2009), and positive affective experiences (Witt et al., 2004). It is feasible that students participating in multimedia lessons where they can see an instructor engaging in nonverbally immediate behaviors will experience greater positive affect for that instructor than students in a lesson where they cannot. Further, studies suggest that clear teaching, in and of itself, is positively related to students' affect for their instructors (see Titsworth et al., 2015), such that students experience greater positive affect for teachers who effectively assist them in selecting, understanding, and retaining material (Titsworth & Mazer, 2016). Given this, it was anticipated that students in an online lesson with an instructor who uses Zoom features to enact CTML principles would report greater positive affect for their instructor than students in a lesson with an instructor who does not.
H2: Compared to a standard synchronous lesson without Zoom features, students will report greater positive affect for their instructor in a synchronous lesson where their instructor uses Zoom features to enact the signaling, embodiment, and generative activity principles.

Participants
Following a pilot test of experimental materials (see Online Supplemental File for manipulation check, presentation slides, teaching scripts, and posttest quiz), 160 undergraduate students were randomly assigned to participate in a live lecture teaching experiment using Zoom. Participants' ages ranged from 18 to 54 years (M = 21.05, SD = 5.23), with 60 participants identifying as men, 99 participants as women, and one participant indicating that they "prefer not to answer." Fifty-nine participants identified as first-years, 36 as sophomores, 31 as juniors, and 33 as seniors, with one participant responding "Other." One hundred eighteen participants identified as White, 21 as Middle Eastern, eight as Black, five as Asian, three as Latinx, and three as "Other" (e.g., biracial). Two respondents did not self-identify. Participants' GPAs ranged from 1.90 to 4.00 (M = 3.30, SD = .55).
Of the 160 total participants recruited for this study, data from 18 were omitted from analyses due to participants having had previous student experiences in courses taught by the instructor of the online lesson (the same instructor who taught the pilot study lesson). Data provided by an additional two participants were also omitted because they had already participated in previous pilot test activities. Following the omission of these data, the final sample size for the experiment was 140 participants. Of the final 140 participants, 72 participated in an online lesson where the instructor used Zoom features to enact CTML principles (i.e., the treatment lesson) and 68 participated in a standard online lesson in which the instructor did not (i.e., the control lesson).

Procedures
After receiving IRB approval, consenting student participants were randomly assigned to attend an online lesson using Zoom at a scheduled date and time. The online lesson topic was on Arthur Chickering's Theory of College Student Development. We selected this lesson because of its relevance to college students' lives. The full teaching materials of the lesson including PowerPoint slides, Treatment and Control lecture scripts, and lecture quiz questions are available in the online supplemental materials file.
Participants accessed their online lesson using the hyperlink and password provided to them upon sign-up. Identical content was taught in both online lessons, and the same PowerPoint slides were used in identical ways during each lesson. The only difference between the two online lessons was whether the instructor utilized Zoom features to enact CTML principles of instructional design during the lesson.

Control Condition.
The hyperlink provided to participants assigned to the standard Control lesson directed them to a Zoom meeting room. Throughout the lesson, participants viewed a PowerPoint presentation of lesson information accompanied by instructor narration. The instructor was not visible to participants in this lesson. All participants in this online lesson had their microphones muted, were unable to activate their cameras or display images, and were unable to type or send messages using Zoom chat. Treatment Condition. Throughout the treatment lesson, participants viewed a PowerPoint presentation of lesson information accompanied by instructor narration, as well as by a live video of the instructor in a small box on the right side of their screen. Participants were able to see the instructor teaching for the duration of the lesson. The instructor was an early-thirties man dressed in business-casual attire, and was the same instructor who taught the online lesson for participants assigned to the control condition. All participants in the treatment condition had their microphones muted, were unable to activate their cameras or display images, and were unable to type or send messages using Zoom chat. The CTML manipulations in the treatment condition included the live video of the instructor displaying immediacy (embodiment principle) and three Zoom polls on the content being taught (generative activity principle), as well as 32 instances of the instructor using Zoom annotations to draw shapes and symbols (16) or create labels (16) drawing attention to visually presented information (signaling principle).

Posttests.
After attending the lesson, participants completed a post-lesson survey assessing working memory overload, affect toward instructor, and cognitive learning (quiz score %), along with demographic information.
Participants' cognitive learning was measured using a 10-item test assessing retention of lesson content which pilot study participants completed (see Supplemental Material for test questions and answers). Test questions were designed to measure information retention by employing a series of multiplechoice questions assessing basic recall of lesson content. Each multiple-choice question was developed based on Suskie's (2018) guidebook for assessing student learning and employed four possible answers (a, b, c, d). Responses to multiple-choice questions were coded as (1) for correct responses and (0) for incorrect responses, then scored to reflect the percentage of total questions answered correctly.

Results
Hypothesis one predicted that participants attending a CTML-based Zoom lesson would experience less working memory overload and, in turn, exhibit greater learning on a post-lesson test. This hypothesis was tested using PROCESS (Hayes, 2022), specifying a simple mediation model using ordinary least squares path analysis. The indirect effect was estimated using a 95% confidence interval based on 10,000 percentile bootstrap samples. We found no evidence of an indirect effect for the influence of CTML-based instruction on participants' test performance through working memory overload (ab = .703, CI: -.530, 2.520; ab ps = .034, CI: -.026, .122), but instead, found evidence for a direct effect of CTML-based principles on students' test performance (c′ = 7.367, c′ ps = .350, CI: .470, 14.265). That is, Zoom teaching using CTML-based principles of instructional design directly caused an average increase of seven percent across participants' test scores, holding constant working memory overload. Although CTML-based teaching did directly cause increases in students' learning, hypothesis one was not supported. Unstandardized model estimates for the analysis are provided in Table 1. Hypothesis two predicted that participants who attended the CTMLbased Zoom lesson would report greater affect for their instructor. An ANOVA revealed significant differences in participants' reported levels of affect toward their instructor across conditions, Welch's F (1, 103) = 7.836, p = .006, η 2 = .058, suggesting that participants assigned to the CTML-based lesson reported greater affect toward their instructor (M = 6.74, SD = .48) than participants assigned to the lesson which was not taught in line with CTML-based best practices (M = 6.42, SD = .80). Another ANOVA also revealed significant differences in participants' reported levels of affect toward enrolling in future courses with their instructor, Welch's F (1, 117) = 5.036, p = .027, η 2 = .037, suggesting that participants assigned to the CTML-based lesson reported greater affect toward enrolling in future courses with the same instructor (M = 6.41, SD = .90) than participants assigned to the lesson which was not taught in line with CTML-based best practices (M = 5.99, SD = 1.23). Given these results, hypothesis two was supported.

Discussion
The study tested CTML principles (Mayer, 2001) to identify specific ways that college and university educators can enhance their online pedagogy using the technological features of Zoom. Contrary to expectations, CTML-based instruction did not indirectly influence students' performance on a post-lesson test through reduced working memory overload. Rather, an instructor's implementation of CTML principles directly increased students' average performance on a post-lesson test from roughly 61% to 70%. An instructor's implementation of CTML principles also created students' affect toward their instructor, such that students who attended a CTML-based lesson exhibited significantly greater affect toward both their instructor and the possibility of enrolling in future courses with their instructor than students who attended a standard online lesson.
Students who attended a CTML-based online lesson did not perceive a reduction in their working memory overload compared to students who attended a standard online lesson; yet nonetheless learned more, which is perplexing and -at first glance -seems to contradict the central propositions of CTML. However, such conclusions may not be necessarily accurate. From the theoretical perspective of CTML, working memory is effectively synonymous with active consciousness (Mayer, 2021); it is limited both temporally and in terms of capacity, lasting only seconds and capable of processing an extremely finite number of pictorial or verbal representations at a single point in time (Mayer, 2001). Despite these limitations, Mayer (2021) contended that the working memory is where most of the sense-making processes that ultimately drive learning occurs. More specifically, it is within the working memory that learner's selection of sensory information, organization of sensory information into coherent verbal and pictorial models, and integration of models with preexisting knowledge structures take place. In other words, CTML assumes that the step-by-step processes whereby learning occurs take place almost entirely within the finite scope of the working memory, with many cognitively intensive processes are taking place in a limited space and a brief span of time. Given this, it is feasible that this process may seem almost instantaneous considering that selection, organization, and integration -as well as working memory overload -all coalesce to influence students' learning within the span of just a few seconds. That is, the fundamental processes through which CTML argues learning occurs are likely still applicable when interpreting the direct effect of CTML-based instruction on students' learning observed in this study, despite the lack of an observed indirect effect through working memory overload.
Indeed, the limited capacity, fleeting nature, and complexity of the working memory also make it plausible that students who participated in this study may have been unable to accurately self-report the extent to which they experienced working memory overload. The instrument used to measure working memory overload (Bolkan, 2017a) prompted students to respond to a series of statements related to their information processing following each lesson, asking students to indicate the extent to which they agreed or disagreed with each statement (e.g., "I felt flustered trying to keep up with the amount of information presented in this lesson") after their assigned online lesson had concluded. Asking students to report on their working memory overload after-the-fact and in aggregate may have ultimately been inappropriate for collecting data which would allow the researchers to appropriately characterize the moment-to-moment cognitive processes which occur within the working memory and mediate learning. Further, asking students to subjectively report on their own information processing, generally, may not be appropriate for examining the ways in which students learn. A meta-analysis by Sitzmann et al. (2010) concluded that self-assessments of learning are likely better suited for measuring students' affect regarding their learning experiences rather than the extent to which they truly learn. Echoing these sentiments, Zheng and Greenberg (2018) expressed concern related to the influence of response biases when employing self-report measurements of cognitive load -chiefly, that students' perceptions regarding their information processing and learning may not necessarily reflect their actual information processing or the extent to which they actually learn. While the test scores used to operationalize learning served as a more objective indicator of students' information processing, the subjectivity of the instrument used to measure working memory overload may have impeded our ability to capture the extent to which working memory overload truly occurred from moment-to-moment.
Despite the lack of the hypothesized indirect effect for CTML-based instruction, students learned more when the instructor strategically used Zoom features to highlight the organization of essential lesson content, made themself visible to students throughout the online lesson, and stimulated students' generative processing of important material. Synthesizing relevant research exploring these three principles -the majority of which have analyzed the total effects of CTML-based instruction on learning- Mayer (2021) reported median effect sizes of d = .51 for the signaling principle, d = .58 for the embodiment principle, and d = .71 for the generative activity principle on student learning during multimedia instruction. In our experiment, when all three principles were manipulated together, the effect size was d = .41, consistent with the range of effect sizes reported by Mayer (2021) in his review as our findings support CTML.
How, specifically, might an instructor's use of Zoom features have contributed to improving learning outcomes for the students who attended the CTML-based online lesson? First, an instructor's use of Zoom's annotation and screen-sharing features is a direct implementation of the signaling principle, which asserts that students will engage in more effective cognitive processing in multimedia lessons "when cues are added that highlight the organization of essential material" (Mayer, 2005b, p. 184). An instructor emphasizing specific information presented during a lesson by employing annotations such as circles, lines, arrows, squares, numbers, and text constitutes an act of signaling in that each annotation serves to direct students' attention to specific verbal or pictorial information, rather than placing the burden of deciding which information to pay attention to, or determining how information is interrelated, upon students (Mayer, 2021). The Zoom annotations employed during the CTML-based online lesson thus likely enhanced the instructor's organizational clarity (i.e., "the methods with which teachers use verbal, nonverbal, and visual resources to organize information for students," Titsworth & Mazer, 2016, p. 119) by assisting students in selecting relevant information to attend to and organizing that information into coherent mental representations.
Second, the instructor remaining visible to students on camera for the duration of the CTML-based online lesson, and engaging in body movements while on camera, likely enhanced student learning by stimulating positive affective responses from students (Mayer, 2014a). When students can see their instructors engaging in high-embodiment behaviors (i.e., behaviors similar those used during in-person interactions in the real world; Mayer, 2014a), it can "help establish a stronger social bond between the teacher and learner, causing the learner to try harder to make sense of the instructional message and thereby build a deeper learning outcome" (Mayer, 2021, p. 345). Third, the instructor's use of Zoom's polling feature during the CTML-based online lesson served to further guide students' selection, organization, and integration of important instructional material in a manner which was unavailable to the instructor of the standard online lesson. More specifically, each of the three polls provided a mechanism through which the instructor of the CTML-based online lesson could directly stimulate students' generative processing by enacting activities which prompted learners to actively engage with material. Each poll posed a question which required students to reflect upon information presented earlier in the lesson, consider the ways in which that information related to other lesson material, and apply that information in responding to each individual poll. That is, each poll provided students with opportunities to self-test their understanding of information presented during the online lesson, as well as highlighted specific pieces of information which were particularly important for students to understand. Like previous research exploring the influence of self-testing on student learning (e.g., Johnson & Mayer, 2009), our results similarly position it as a viable generative activity which can prime more effective selection and organization of instructional content, as well as demonstrate that Zoom's polling feature is an effective tool for implementing this generative activity during synchronous online instruction.
As predicted, students who participated in a CTML-based online lesson reported significantly greater affect toward their instructor than students who participated in a standard online lesson. Moreover, students who attended a CTML-based online lesson exhibited significantly greater affect toward the possibility of enrolling in future courses with their instructor. These results are consistent with findings from previous CTML research, particularly those suggesting that instructors can stimulate a sense of social presence among students in multimedia classrooms by making themselves visible to students (e.g., Fiorella et al., 2019Fiorella et al., , 2020Li et al., 2019;Wang et al., 2018), as well as with instructional communication research suggesting that instructors who engage in nonverbally immediate behaviors (Andersen, 1979) can facilitate positive affective experiences for their students (e.g., Frymier & Houser, 2000;Frymier et al., 2019;Witt et al., 2004). Our findings thus complement previous research in suggesting that an instructor's nonverbal behaviors may have significant implications for students' affective responses during instruction, even in remote learning environments.
Although CTML (Mayer, 2001) is primarily concerned with providing instructors guidance in creating multimedia lessons based upon the cognitive processes through which humans learn, extant CTML research nevertheless highlights the salient role that students' affective experiences can play in their information processing. Moreno (2005) asserted that affective elements of students' instructional experiences can have a profound effect on the extent to which students engage in the generative information processing which is fundamentally necessary for students to achieve meaningful learning outcomes. This sentiment was echoed by Mayer (2014b), who wrote that student affect is integral for learning insofar as it "initiates, maintains, and energizes the learner's effort to engage in learning processes" (p. 171). From a CTML-based perspective, learning is an active process (Mayer, 2001) -one in which students must put forth cognitive effort as they engage in making sense of novel information. Elements of instruction (e.g., embodiment, signaling, generative activity) which stimulate positive affective experiences for students thus facilitate generative processing by motivating students to pay attention to important instructional content (i.e., selecting) and exert the cognitive resources necessary to comprehend it (i.e., organizing and integrating; Mayer & Estrella, 2014).
Given that literature suggests many instructors unexpectedly taught online following the COVID-19 pandemic and struggled to adapt to online teaching (e.g., Huber & Helm, 2020), it is also important to consider potential practical implications for instructors who use Zoom. First, instructors teaching via Zoom (or similar remote conferencing software) should make strategic efforts to avail themselves of program features that can be used to enact signaling during online instruction. While the instructor of the CTML-based Zoom incorporated signaling by creating shapes (e.g., circles, arrows), employing highlighting using the "draw" feature, and including additional numbers and text to provide organizational cues while sharing their screen, this does not represent the full array of ways in which instructors might use Zoom's annotation function to assist students in selecting and organizing information. Zoom's annotation feature also allows instructors to "stamp" specific pieces of information with check marks, stars, arrows, exes, hearts, and question marks, as well as "spotlight" specific information using colorful icons which can be substituted for the instructor's mouse cursor on-screen (Zoom, n.d.). Further, instructors can modify the formatting of annotations in a variety of ways, such as by adjusting the thickness of lines, bolding or italicizing numbers and text, increasing or decreasing font size, or employing different colors for specific annotations to assist students in differentiating between them (Zoom, n.d.). In doing so, however, instructors should be mindful that they do not employ so many annotations that they inadvertently violate CTML's coherence principle (Mayer, 2001) by drawing students' attention away from important lesson material, thereby facilitating extraneous processing. As evidenced by previous research (e.g., Sung & Mayer, 2012), instructors' signaling attempts must be relevant to instructional goals in order to be effective -that is, instructors should be deliberate in ensuring that the annotations they use while teaching via Zoom direct students' attention toward, rather than away from, important core content.
Second, instructors should make themselves visible to their students when teaching online by turning and keeping their cameras on during instruction. Given that Zoom's default settings allow on-camera instructors to remain visible to students even while sharing their screens, allowing students to see their instructor should be as straightforward as activating one's camera at the beginning of an online lesson (Zoom, n.d.). While making oneself visible to students may be relatively intuitive, instructors should also be mindful of their nonverbal behaviors while on camera. As originally emphasized by Mayer (2014a), students simply being able to see their instructor may not necessarily be enough to stimulate positive affective responses in-and-of itself. Rather, those affective responses are largely influenced by the extent to which students perceive their instructors' nonverbal behaviors as high-embodiment (Mayer, 2014a) or immediate (Andersen, 1979). Instructors can facilitate these perceptions by making a conscious effort to look into their cameras while speaking to students as a substitute for in-person eyecontact, using hand gestures and physical movement to emphasize instructional content as if they were conveying information to students face-to-face, and employing appropriate facial expressions when communicating with students (e.g., smiling). While doing so, instructors should carefully consider the timing and purpose of their nonverbal behaviors.
Third, instructors teaching via Zoom should leverage program features strategically to incorporate activities into their online lessons which stimulate students' generative processing of information. That is, instructors teaching online should use Zoom features to prompt their students to engage in active, as opposed to passive, learning. In the CTML-based lesson of this study, the instructor utilized Zoom's polling feature to prompt students to engage in selftesting (i.e., studying information and completing practice assessments; Mayer, 2021) -assisting students in recognizing particularly important pieces of information presented during the lesson (i.e., selecting), understanding how that information was interrelated (i.e., organizing), and remembering that information later (i.e., integrating; Mayer, 2021). However, this is just one way in which instructors might use the various features of Zoom to implement generative activities in their online classrooms. For example, instructors teaching via Zoom could provide students with opportunities for self-explaining or teaching by utilizing Zoom's breakout room feature to split students into smaller discussion groups (Zoom, n.d.), facilitating students' active engagement with both lesson content and other students attending an online lesson. Instructors might also consider ways in which they can allow their students to create their own annotations during lessons by adjusting Zoom's screensharing settings (Zoom, n.d.), stimulating students' generative processing by providing opportunities to engage in mapping or drawing during instruction, or prompt their students to guide their own use of Zoom annotations to create visual representations of instructional content via microphone or chat. Utilizing Zoom features in any of these ways may stimulate action from otherwise passive students during online instruction, thereby promoting greater engagement with material and facilitating deeper learning (Fiorella & Mayer, 2016b).

Limitations and future directions
The main limitation of this study involves the types of questions used to assess students' learning on the post-lesson test. While the multiple-choice questions which comprised the post-lesson test provide a means of assessing students' basic information recall, they do not measure the types of meaningful learning outcomes which constitute the penultimate goal of CTML-based instructional design (Mayer, 2001). Meaningful learning, according to Mayer (2001), takes place when learners exhibit both retention and understanding of instructional material. Mayer (2021) wrote that deeper understanding of content is best assessed using transfer questions which prompt students to apply newly learned information to solve novel problems; such as by posing open-ended troubleshooting, redesign, prediction, or conceptual questions. While multiplechoice questions are appropriate for assessing students' recall and recognition of lesson content, "tests of retention such as recall or recognition tests provide a limited view of what someone knows" (Mayer, 2021, p. 98). Future researchers interested in exploring how learning occurs during Zoom-based instruction should thus carefully consider ways in which they can employ both retention and transfer questions to measure students' deeper learning outcomes. Another limitation is that we manipulated three features of Zoom together, instead of individually manipulating the signaling, embodiment, and generative activity principles in isolation. By doing this, we were able to compare the same lecture with and without Zoom features enabled in a way that might mirror real teaching environments (camera on, highlighting aspects of slides, polling). However, this limitation does not allow us to determine the effects of each individual Zoom feature (e.g., just the camera on with instructor immediacy). Future researchers might test other principles of CTML using Zoom in isolation as we found evidence that the three principles tested here appear to be effective instructional communication behaviors.
As teaching with Zoom persists beyond the pandemic, including the use of Zoom recordings for asynchronous instruction, CTML principles should continue to be tested. For instance, instructors who record asynchronous lecture videos may still employ CTML principles by displaying and recording their faces instead of narrating without showing themselves on camera, highlighting main points of their videos after post processing and editing, and perhaps conducting asynchronous polls for students to engage in the elements of recorded lectures (and posting asynchronous poll results on an online course message board). Similarly, although we disabled the Zoom chat features to ensure experimental control of the teaching conditions, the chat feature and question/answer feature might offer other avenues to test CTLM as instructors respond to students' questions or comments and students communicate with each other in Zoom.

Conclusion
This live teaching experiment examined the causal impact of CTML-based principles of instructional design to Zoom-based online teaching environments. An instructor's strategic use of Zoom features to enact CTML's signaling, embodiment, and generative activity principles directly enhanced students' performance on a post-lesson test, as well as stimulated greater affect among students toward the instructor of their online lesson. Despite its drawbacks, Zoom may be well-suited to create online learning environments grounded in CTML principles of instructional design that are conducive to students' learning outcomes. In particular, it appears that merely "showing up" and lecturing with Zoom is not enough. Rather, instructors' active utilization of Zoom's annotation, screen-share, and polling features provide significant improvements for students' online synchronous learning experiences.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Notes on contributors
Kevin C. Knoster (Ph.D., West Virginia University, 2021) works in industry conducting data analysis.