Active learning includes any meaningful instructional activity that requires students to actively engage and respond during class, and to think about what they are doing (Prince, 2004; Felder and Brent, 2009). Rather than passively viewing a lecture, active learning engages students in constructing or applying knowledge for themselves. Commonly-described active learning activities range from collaborative or cooperative learning, personal response systems (e.g., clicker questions), answering questions in a think-pair-share format, and problem-solving activities (Felder and Brent, 2009; Freeman et al., 2014).

Active learning can help students engage in constructive or interactive levels of engagement known to be associated with greater learning (Chi and Wiley, 2014). Active learning methods can also support social-psychological needs, such as self-efficacy and perceptions of belonging (Ballen et al., 2017). Courses using these methods have been shown to reduce or eliminate gaps in learning outcomes between minoritized and majority students otherwise found in traditional lecture-based courses (Haak et al., 2011; Ballen et al., 2017; Theobald et al., 2020). Despite these important benefits, active learning methods are less frequently used in online instruction. Moreover, there are few studies comparing learning outcomes between online course sessions that use active learning methods vs. more traditional lecture-only instruction (cf., McClellan et al., 2023).

Instructors often resist using active learning methods, because they assume use of these methods will require a great deal of time and effort (Mogavi et al., 2021). However, certain active learning methods can be implemented online without an entire course overhaul. Even interspersing well-designed learning activities among shorter lectures can be beneficial (Sandrone et al., 2021). Exploratory learning activities provide one such approach. In exploratory learning, students complete a novel learning activity (e.g., solve problems) before receiving instruction on the procedure and concepts (DeCaro and Rittle-Johnson, 2012). This overarching term describes the two-phase sequence used across studies from the literatures on productive failure (e.g., Kapur, 2008), preparation for future learning (e.g., Schwartz et al., 2009), and problem-solving-instruction methods (PS-I; e.g., Loibl and Rummel, 2014).

Multiple studies across these literatures, using exploratory learning during in-person class or lab settings, have shown that exploring before instruction can improve learning (Loibl et al., 2017; Darabi et al., 2018; Sinha and Kapur, 2021). However, relatively few studies have implemented exploratory learning in undergraduate STEM courses (e.g., Weaver et al., 2018; Chowrira et al., 2019; Bego et al., 2022; DeCaro et al., 2022). Even fewer studies have compared the use of exploratory learning before instruction in an online format to a more traditional online lecture-then-practice approach (e.g., Hieb et al., 2021). Given the importance of active learning methods for learning and persistence in STEM disciplines, exploratory learning provides a promising, relatively simple, method to accompany lecture in an online course.

In traditional lecture-based instruction, students often experience cognitive fluency, in which they process the information superficially and think they understand the information better than they actually do (Bjork, 1994; Felder and Brent, 2009; DeCaro and Rittle-Johnson, 2012). When engaging in exploratory learning before lecture, students are given a novel problem or activity that is relevant to the target topic. Students can begin to attempt this activity using prior knowledge, but they have not necessarily encountered the problem before (Kapur, 2016). This process better enables students to integrate the new information with their existing knowledge, creating more connected schemas in long-term memory (Schwartz et al., 2009; Kapur, 2016; Chen and Kalyuga, 2020).

However, the activity is novel, and students often solve the problem incorrectly. Through this process, students can become more aware of the gaps in their knowledge (Loibl and Rummel, 2014; Glogger-Frey et al., 2015). Students might become more curious (Glogger-Frey et al., 2015; Lamnina and Chase, 2019), and desire to make sense of the content, leading them to pay greater attention during the subsequent lecture (Wise and O’Neill, 2009). This exploratory process has therefore been described as productive failure (Kapur, 2016).

During the exploration process, students also test hypotheses, and begin to discern which features of the problem are relevant or useful for a solution, and which elements are less important (DeCaro and Rittle-Johnson, 2012). Some of these elements may include the relationships between individual features in a more complex, to-be-learned formula (Alfieri et al., 2013; Chin et al., 2016). These cognitive, metacognitive, and motivational processes can help to deepen students’ understanding (Loibl et al., 2017).

The benefits of exploring before instruction are generally found on measures of conceptual understanding (Loibl et al., 2017), although some studies find benefits for procedural knowledge as well (e.g., Kapur, 2010, 2011; Kapur and Bielaczyc, 2012). Conceptual knowledge consists of underlying principles and relationships between connected ideas, whereas procedural knowledge includes the memorization or use of a series of actions to solve problems (Rittle-Johnson et al., 2001).

For example, Bego et al. (2022, Experiment 1) examined the causal benefits of exploratory learning during an in-person undergraduate physics lesson on gravitational field. Bego et al. randomly assigned half of the class to a traditional lecture-then-practice condition (instruct-first condition). The other half were assigned to an exploratory learning condition (explore-first condition), in which they instead completed the same learning activity prior to instruction. On a subsequent posttest, students in the explore-first condition showed equal procedural knowledge as those in the instruct-first condition, but higher conceptual knowledge. Thus, students learned and applied the formulas they were taught equally well between conditions, but understood the relational principles better after having explored the topic for themselves first.

The current experiment examined whether the benefits of exploratory learning extend to a synchronous, online learning environment, during which the class meets live via videoconference. We adapted Bego et al.’s (2022) materials and procedures for an online physics lesson on gravitational field, and students participated during their regular class time. The course instructor was the same as in Bego et al.’s study. Students were randomly assigned to either an explore-first or instruct-first condition, and participated on separate class days. Students in the instruct-first condition completed the traditional instructional order (lecture then activity). Students in the explore-first condition completed the same materials in reverse order (activity then lecture). The activity included a series of contrasting cases, examples that vary in ways that help students encounter important features of the problems (e.g., Schwartz et al., 2011; Roll et al., 2012; Roelle and Berthold, 2015). Students completed a posttest assessing their procedural and conceptual knowledge of gravitational field. By randomly assigning students to condition, and providing the exact same materials in reverse order, this experimental research design allowed us to examine the causal effect of exploring before instruction in an online setting.

Although we used the same basic materials as Bego et al. (2022), we made several general modifications to the procedure, due to time constraints. The class period in our study was shorter (50 min vs. 75 min in Bego et al.). To account for this decreased time, first we focused only on the gravitational field lesson, and did not include other measures that Bego et al. used (i.e., learning transfer or survey). Second, students completed the posttest on their own, outside of class time. Also, as is commonly needed with online courses, we allowed more time during class for transitions (Mogavi et al., 2021; Venton and Pompano, 2021). We allowed for more class time for students to locate and download the learning materials and transition in and out of breakout groups.

Our specific research questions included the following:

We expected that the modality of teaching would not change the cognitive, metacognitive, or motivational learning benefits of exploring, even in an online setting. We hypothesized that students in the explore-first condition would score higher on the posttest, especially on the conceptual knowledge subscale. Thus, we also hypothesized that our results would resemble those of Bego et al. (2022), extending those findings to an online, synchronous learning setting.

Participants (N = 78) included all students who attended an online, synchronous, introductory undergraduate physics course (Fundamentals of Physics I) on the dates of the experiment and submitted all learning materials. Participants were primarily upper-level undergraduate students working toward degrees with a pre-professional health science focus. Additional participants were excluded from analyses for experiencing an internet connection issue during class that lasted more than 1 min during a critical time period (n = 3) or for scoring a 0 on the procedural knowledge subscale of the posttest (n = 1), indicating lack of attention or effort.

Students from one Fundamentals of Physics I course section were randomly assigned to explore-first (n = 37) or instruct-first (n = 41) conditions, and attended class on the day of their assigned condition. Students participated online during one 50-min class period through Blackboard Collaborate, the learning management system (LMS) used all semester.

The session included four phases: instruction, activity, activity review, and posttest. The only difference between conditions was the order of the phases. In the explore-first condition, students completed the learning activity first. The instructor provided “tips for effective exploration” on a slide, explaining that correct answers were less important than the process of working through the activity and trying to explain ideas. Students accessed the activity through a link on the LMS page and worked in breakout groups (13 min). Then, students rejoined the whole class and were given the lecture-based instruction (24 min), followed by a brief review of the learning activity and the answers (5 min). Students had the opportunity to ask questions throughout. Students in the instruct-first condition followed these same procedures, except they began with the lecture-based instruction, followed by the learning activity, then activity review.

At the end of class, all students were directed to the posttest link on their LMS, referred to as a “Gravitational Field Practice Assignment.” Students were advised that the assignment should take around 15 min to complete, to work individually, and that they would be graded for effort, not accuracy. Students were asked to submit their learning activities and posttests before the next class period.

All procedures were approved by the university Institutional Review Board. Students were debriefed about the study in a letter emailed at the end of the semester and given the opportunity to withdraw their data.

Learning activity scores were analyzed using a between-subjects ANOVA as a function of condition (explore-first, instruct-first). Posttest scores for each subscale were transformed into percentages and analyzed using a 2 (condition: instruct-first, explore-first) × 2 (subscale: procedural, conceptual) mixed-factorial ANOVA, with condition between-subjects and subscale within-subjects.

As expected, students in the explore-first condition had lower scores on the learning activity than students in the instruct-first condition, F(1, 76) = 11.12, p = 0.001, η² = 0.13 (Table 1).

As shown in Figure 2, there was a main effect of condition, in which students in the explore-first condition (M = 76.49%, SE = 3.05, 95% CI[70.41, 82.56]) had higher overall posttest scores than students in the instruct-first condition (M = 66.36%, SE = 2.90, 95% CI [60.59, 72.13]), F(1, 76) = 5.79, p = 0.019, η_p² = 0.07. There was also a main effect of subscale, in which students scored higher on the procedural items (M = 76.91%, SE = 3.08, 95% CI [70.79, 83.03]) than the conceptual items (M = 65.94%, SE = 1.65, 95% CI [62.65, 69.22]), F(1, 76) = 18.15, p < 0.001, η_p² = 0.19. No interaction was found, F(1, 76) = 1.98, p = 0.164, η_p² = 0.03, suggesting that the effects of condition occurred similarly across both procedural and conceptual subscales. This observation was confirmed by examining 95% confidence intervals. As shown in Table 1, students in the explore-first condition scored higher than those in the instruct-first condition on both the procedural subscale and conceptual subscale.

Although these findings provide promising support for the use of exploratory learning activities in online courses, there are potential limitations both in using online active-learning methods and in experimentally studying their use. First, because the study took place online, we cannot be certain that students worked on the learning activity when they were told to (e.g., before, rather than after, instruction). Given that students who were asked to explore before instruction on average scored lower on the learning activity, it seems likely that they followed these procedures in our study. However, instructors are less able to monitor what students do while working online. For similar reasons, controlling how students approach assessments is more difficult when done online vs. in person. Although we gave students ample time to work on the posttest, results might have looked different if students were asked to complete the posttest immediately and given time constraints (e.g., using an exam function in their LMS). However, our results generally mirror those of Bego et al. (2022), who used these same materials during an in-person class session. If control during assessment was an issue, it would more likely have diminished differences between conditions. That we found similar effects as Bego et al. suggests that control did not seem to be a factor in our study.

We also focused on only one outcome measure—posttest results for procedural and conceptual knowledge. Our results would be strengthened by additional survey and outcome measures, such as the transfer scale used in Bego et al.’s (2022) study.

Online learning in this study occurred in a live, synchronous, virtual classroom. More research is needed to determine if exploratory learning can be implemented, and benefit learning outcomes, in an online course conducted asynchronously. LMSs provide the tools to implement this method. For example, students can be required to submit an exploration activity (graded for effort, not accuracy) prior to being granted access to the video lecture on the topic. Similar methods can also be used in hybrid courses such as flipped classrooms. In such courses, instructors typically ask students to view a lecture outside of class, and spend class time using active learning methods such as cooperative problem solving (Lage et al., 2010). Rather than (or in addition to) asking students to solve problems after viewing the lecture, students could be given exploratory learning activities in class, prior to watching the assigned video lecture on the topic outside of class (Kapur et al., 2022). Like the current study, such methods have the potential to increase learning outcomes, and potentially other aspects of engagement as well.