Studying with Friends in Introductory Biology Is Associated with Low Impact Study Strategies and No Improvement of Exam Scores

Abstract

Students often choose to study for exams with friends. Since active learning in class boosts success, instructors might expect studying with peers to also help. However, research offers little support for this. We investigated whether students study with peers because of low metacognitive knowledge about study strategies. At the start and end of a first-term introductory biology course, students reported their study strategies and their knowledge about their effectiveness. These data were combined with student demographic information and grades. We found that students entered university demonstrating only modest metacognitive knowledge, and this was associated with course performance. Study group use was popular, with students valuing the support and collaboration, but it had no significant effect on exam scores. Students chose encoding over retrieval strategies regardless of whether they studied alone or with others, and students who more often used encoding strategies while studying scored lower on exams. We conclude that studying with friends is not harmful, but it is based on incomplete metacognitive knowledge, and encoding strategies in general are used too close to exams. We recommend that instructors encourage peer study groups to meet during nonexam weeks when students are learning rather than studying.

INTRODUCTION

Many students prepare for a difficult exam by studying with friends outside of class time. They might ask and answer questions, quiz each other, or independently study in the same room. Although there is significant evidence supporting peer learning within class activities, the effect of informal out-of-class study groups on exam performance remains largely untested. This paper explores the effectiveness of student-organized study groups in a first-year, first-term introductory biology course with difficult exams.

Learning with Peers in STEM Classes

There is significant support for peer interaction during class time in biology and other STEM disciplines (Michael, 2006; Prince and Felder, 2006; Freeman et al., 2007; Bishop and Verleger, 2013; Theobald et al., 2020). Examples of in-class peer learning techniques include peer instruction in physics (Crouch and Mazur, 2001), two-stage exams in earth science (Bruno et al., 2017) and biology (Cooke et al., 2019), and clicker questions with discussion in biology (Smith et al., 2011). Using cooperative teams in the classroom is also an effective strategy. POGIL (process-oriented guided inquiry learning) is a team-learning program often used in chemistry (Bailey et al., 2012), and PBL (problem-based learning) is common in medical schools and biology classrooms (Allen and Tanner, 2003). All of these forms of active learning demonstrate the benefits of organized discussion between students during class.

Theories of Peer Learning

Peer learning in the classroom, exemplified by the active learning techniques discussed above, is supported by considerable research on cooperative learning (summarized by Johnson et al., 1998; Springer et al., 1999; Johnson et al., 2014; Nokes-Malach et al., 2015). The primary cognitive theory associated with cooperative learning is social interdependence theory (Johnson and Johnson, 1994). This theory states that students benefit from collaborative work because they can divide labor, keep each other accountable, and solve problems together. Multiple studies have also examined cooperative learning during class time (see reviews by Johnson and Johnson, 1994; Slavin, 1996; Johnson et al., 1998).

The research measuring the effectiveness of voluntary out-of-class study groups is more limited. Unlike peer learning in the classroom, study outside the classroom is organized by the students themselves. The cognitive theory of collaborative inhibition argues that learning with peers may have a negative effect. Students encode information into long-term memory in different ways, so trying to collaborate while storing and retrieving information may tangle up the process rather than streamline it (Basden et al., 1997; Weldon and Bellinger, 1997). It is therefore important to consider the context of collaborative learning before assuming that all types are equally effective.

The Strategy of “Studying with Friends”

Although student learning can occur during organized activities during class time, most of it likely occurs out of class while students study the course content. Significant research has been done on the strategies that students use and the strategies most associated with demonstrated learning. As part of these projects, researchers often survey students on their study strategies. An option often chosen by students is to “study with friends.” Hartwig and Dunlosky (2012) found that 50% of their surveyed introductory psychology students say they regularly study with friends. Morehead et al. (2016) report a similar 48% of introductory psychology students do so. Rodriguez et al. (2018) report 21% of students in a second-year biology class studied with friends, and Williams et al. (2021) found that study groups were used by 30% of first-year university students across multiple courses. In the context of this study, we define “study with friends” as 1) an optional meeting outside of class time 2) with students invited by someone whose name they know 3) to prepare for an exam that all are taking 4) using their own choice of strategies within the study group. Students were not asked to describe friend depth in this study, so the term “friends” includes friend depths from simple acquaintances to close confidants.

Metacognition and Effective Study Strategies

Consider a student who is invited to attend a study group for an upcoming exam. Unlike an in-class peer learning activity, this study group is optional. The student must decide whether the time in the group will be effective compared with studying alone, and then decide to attend. They then must choose what activities to use while with the group. Should they plan to re-read their notes, asking questions of the group when they have them? Should they encourage the group to take turns explaining difficult concepts? Should the group take turns quizzing each other? Although time interacting with peers in class is organized using the metacognitive skill of an experienced instructor, time spent studying outside of class is controlled by the students and is therefore based on their own metacognitive skills.

Metacognition in a student can be broadly defined as the student being aware of and being in control of how they work to learn (Stanton et al., 2021; Halmo et al., 2024). We will focus here on two aspects of metacognition: 1) metacognitive knowledge (here, specifically knowing what study strategies are effective, Pintrich et al., 2000; Biwer et al., 2020); and 2) self-regulation (here, specifically choosing and implementing appropriate study strategies, Schraw and Moshman, 1995). There is considerable research available to guide students on the most effective study strategies, but only rarely is student knowledge of effective strategies, or metacognitive knowledge, measured in the context of an actual college class. This type of knowledge is very likely to be important, as students vary in general metacognitive knowledge (Zhong et al., 2025), and those who demonstrate higher general metacognitive skills perform better in biology courses (Raković et al., 2022; Bernacki et al., 2023). There is also evidence that co-curricular academic skills courses that include study strategy knowledge can improve biology course grades for students (Wienhold and Branchaw, 2018; Hawkins et al., 2021).

We define a student with high metacognitive knowledge as one who can describe which study strategies have been shown to be most effective for learning according to research. The experimental research on study strategies consistently supports three strategies as effective for long-term learning. Retrieval practice, or self-testing, is most highly recommended (Dunlosky et al., 2013). Most research on retrieval practice has participants carefully read a passage and then either attempt to repeat the information from memory (retrieval) or reread the passage (Roediger and Karpicke, 2006; Carpenter et al., 2008). Participants who used retrieval strategies performed better in later tests of understanding. Studies that measure self-testing as a study strategy in a real course generally find that students who use self-testing in their course outperform students who do not (see meta-analyses by Moreira et al., 2019; Agarwal et al., 2021; Carpenter et al., 2022). The strategy of retrieval practice is assumed to also incorporate other studying strategies that similarly require students to retrieve information from memory, such as rewriting notes from memory.

A second effective study strategy is explaining concepts out loud, particularly without notes. In addition to incorporating the retrieval benefits of “from memory,” the act of organizing knowledge and reframing it to explain or “learn by teaching” has been shown to improve test scores—particularly if students teach earlier on in the learning process and speak aloud (Lachner et al., 2022).

Third, there is evidence that drawing both representational drawings and concept maps requires students to select and reorganize information in a way that aids learning (Bobek and Tversky, 2016; Fiorella and Zhang, 2018; Wong et al., 2021). This strategy is more likely to be specific to disciplines like biology that have complex pathways or flow charts to understand and memorize (Van Meter and Garner, 2005).

Study Groups as a Study Strategy

How does studying with peers outside of class rank as a study strategy? Evidence from actual college courses is limited. Tang (1993) reports that students in a physiotherapy major that chose to work together on a research project did not score higher than students who worked alone. Rybczynski and Schussler (2011) surveyed students on the use of study groups in large introductory biology courses. They found no significant relationship between exam scores and the use of group study. A 2015 survey of undergraduate medical students in Saudi Arabia found that high-grade point average (GPA) students prefer to study alone (Al Shawwa et al., 2015). Chi and Kadandale (2022) helped organize optional study groups in an active-learning biochemistry course and found that students who chose to work in groups on an activity outperformed students who did not. Outside of biology and medicine, Stockdale and Williams (2004) placed students in cooperative learning groups for one module of an educational psychology course and encouraged group study outside of class. They found low achievers improved somewhat on the next exam, while high achievers showed no improvement. Finally, a meta-analysis of the effectiveness of the Motivated Strategies for Learning Questionnaire (MSLQ) for predicting grades by Credé and Phillips (2011) found that the relationship between peer learning behavior and GPA was actually negative. These scattered studies indicate that peer study groups outside of the classroom are unlikely to have the same benefits as cooperative learning does inside the classroom.

Self-Regulation and Differences in Study Strategies among Students

In addition to metacognitive knowledge, students must also effectively use self-regulation. If a student knows that explaining from memory is effective but finds that strategy to be difficult or exhausting, they might instead choose to use a more approachable or more immediately rewarding strategy, like summarizing notes. When students spend time studying, either alone or with friends, what study strategies do they actually use? Students who say they use more effective strategies usually have higher test scores (Ewell et al., 2023; Hazlett et al., 2025), so it is important to provide support to students who have not had access to instruction regarding effective strategies. There have been regular calls to improve study strategy skills for students from under-resourced communities (Ford and Thomas, 1997; Engle and Tinto, 2008), even though there is only modest evidence that students who identify as members of historically under-resourced groups use fewer effective study strategies (Rodriguez et al., 2018; Williams et al., 2021). In this study, we seek to quantify student knowledge of effective study strategies, plus measure what strategies they chose to actually use in their total strategy time and in the time they spent in study groups. We combined institutional data (gender, ethnicity, high school GPA, etc.) with student survey responses to determine whether students from any demographic group actually need more support than others, or whether all students need support.

Research Questions

We organize our study as follows. When a student arrives in a difficult introductory biology class during their first term at college, are they able to describe what study strategies are likely to be most effective? Which students then study with friends? Are they able to spend time in study groups in an effective manner? If studying with friends is not effective, why do students continue to study with others? We break this down into the following research questions:

1. What metacognitive knowledge about study strategies do students start college with? Does this affect course performance?

2. Do students who identify as part of different demographic groups vary in metacognitive knowledge about study strategies?

3. Who studies with friends, and what are their primary study strategies during group study?

4. Is studying with friends associated with exam performance?

5. Why do students value studying in groups?

MATERIALS AND METHODS

Participants

The participants in this study are from a large, public R1 institution in the Western United States. The students were in two sections of a large introductory biology course, a course required for biological science majors. The first author (A.E.W.) co-taught the course during the Fall quarter of 2022. Students in this course are 92% first-term college students. The most common majors are biological science (46%), undeclared (22%), and public health science (15%), and most students consider themselves pre-med or pre-health. There were 643 students enrolled, 590 were first-year students, and 502 students took all the midterms and completed at least one survey. In Table 1, we have indicated the student-identified demographics of biology majors at the institution, in the course as a whole, plus provided the demographic identities of the participants for each research question. In all cases, the percentages of students who identify with each characteristic (gender, ethnicity (PEER: persons excluded because of ethnicity or race), first generation, low income status) and course grade were similar to those of the first-year students in the course as a whole. In the course, 51% were PEER students, 73% were women, 42% were first-generation students, and 39% were low-income students.

TABLE 1.

Student demographic characteristics and performance in course for each cohort

			Research Questions
			RQ1, RQ2	RQ3, RQ4	RQ5
	First-year biology students at the institution	First-year students in the course	First-year students who took all midterms and the week 0 survey	First-year students who took all midterms and the week 10 survey	First-year students who took all midterms and said in the week 10 survey that they studied with friends
Cohort:	%	%	%	%	%
Black, Latinx, or Indigenous students	39	51	48	48	53
Women	70	73	73	79	78
First-generation	38	42	40	40	38
Low income	45	39	37	39	40
		Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Grade in course	—	3.23 (0.81)	3.31 (0.76)	3.30 (0.72)	3.21 (0.72)
Total	n = 707	n = 590	n = 502	n = 375	n = 199

Survey Development and Administration.

The survey design was based on the list of strategies used in Hartwig and Dunlosky (2012) and Morehead et al. (2016). It was modified in two major ways. We added a metacognitive knowledge component: asking students which strategies researchers had determined to be most effective. We also changed the study strategy that used questions to be more precise—instead of asking what study strategies students use regularly, we asked students to estimate the percentage of their study time spent using each strategy. The majority of the survey questions were quantitative, with students asked to indicate a score, a frequency (regularly, occasionally, rarely), or a percentage. A single free-response question asked students who said they studied with friends to explain why they valued that study strategy. The survey questions are summarized in Table 2.

TABLE 2.

Survey questions on metacognition

The following questions are part of longer surveys administered to students. The association with the relevant research questions is provided, along with the question scales. The study strategy questions provided a list of sample study strategies as listed. An abbreviation for each strategy is shown here in parentheses for use in the paper, but was not given to students.

Week	RQ	Question	Scale
Week 0 and Week 10	1, 2, 3, 4	Researchers have tried to measure the effectiveness of different strategies for students studying for an exam. Predict the effectiveness of each strategy, where 0 = Least Effective and 100 = Most Effective. Give each strategy a different grade. Explaining concepts out loud from memory (Explain) Drawing diagrams, charts, or pictures (Draw) Re-watching course videos (Rewatch) Condensing or summarizing notes (Summarize) Answering practice problems from memory (Practice Problems) Having someone explain concepts to me while studying (Listen) Using flashcards without looking at notes (Flashcards) Re-reading notes or the textbook (Re-read) Studying with friends (Friends) Watching non-course videos on the topic (Videos) Re-writing notes from memory (Re-write)	0–100
Week 10	3, 4	Which of these study strategies listed did you use while studying for the last exam in this class? Indicate the percent of the time you used for each until your total is 100%. Answering practice exam problems from memory Answering study guide questions from memory Condensing or summarizing my notes Drawing diagrams, charts, or pictures Explaining concepts out loud from memory Having someone explain concepts to me Re-reading my notes Re-reading the textbook Re-watching course videos Re-writing notes from memory Studying with friends Using flashcards without looking at notes Watching noncourse videos on the topic Other	0–100
Week 0	3	Did you study with friends for science exams in the past?	Regularly, Occasionally, Rarely, or never
Week 10	3, 4, 5	Did you study with friends in this class?	Regularly, Occasionally, Rarely, or never
Week 10 (for students who reported studying with friends)	3	What percentage of time did you spend on each of these study activities while studying with friends? Indicate the percent of the time you used for each until your total is 100%. I am listening to someone else explain a topic I am explaining a topic to someone I am answering a practice problem out loud from memory I am answering a practice problem out loud from my notes or the book I am quizzing someone else I am reading on my own I am quietly quizzing myself I am chatting with/discussing test strategies with/encouraging my friends	0–100
Week 10 (for students who reported studying with friends)	5	Why do you value studying with friends?	Open-ended response

The surveys for the beginning and end of the term were administered in Qualtrics (Qualtrics, Provo, UT). The Week 0 survey link was emailed to enrolled students 2 wk before the quarter began, and was open until the third day of the term. The Week 10 survey was emailed to students at the start of the last week of class and was open for two weeks until the end of finals week. Students were given a small amount of extra credit for completing the surveys. Participants occasionally skipped individual questions on surveys, and thus the cohort size varies across research questions from 199 to 502 students.

Course Exams and Student Performance Measures.

The 10-wk course contained midterms at Weeks 3, 7, and 10. A final exam was offered for students who wished to raise their grade, but it was not included in the analysis since many students did not take it. The midterm exams combined were worth 70% of the course grade. The format of the midterm exams was open-note but timed to provide ∼2 min per question. Roughly three-quarters of the exam questions were application problems or research graph interpretation questions, and the raw mean scores were generally 60 to 65% before additional points were awarded, which raised the mean to over 70%. About 96% of students passed the class with an A, B, C, or P (pass). We used the sum of the three midterm exam scores (each with 27 questions) as a proxy for student performance.

Course Design.

The introductory biology course in this study was designed to foster a sense of belonging and engagement in students. Pre-class videos provided the biology content, and class time contained minimal lecture and significant peer instruction with clicker questions (Caldwell, 2007). A digital backchannel was provided to allow more students to ask questions anonymously during class (Baron et al., 2016). Seating arrangements matched discussion groups to encourage social support (Payne et al., 2025). We utilized undergraduate learning assistants in lecture and discussion to support group work (Clements et al., 2022). There were group practice exams five times during the 10-wk quarter to encourage students to discuss old exam questions and gain experience with application-type problems (Parsell et al., 2022).

Course Discussion of Study Strategies.

The course instructors discussed effective study strategies in a 10-min lecture at the end of the first week of the term. Students were encouraged to research confusing concepts during nonexam weeks and focus on high-impact strategies during exam weeks (practice problems from memory, spacing of studying, explaining from memory, drawing diagrams from memory). The instructor specifically indicated that rereading notes, asking questions, listening to others, and watching videos should occur before exam weeks. Students were offered extra credit to complete short “debrief” survey assignments every weekend, where they described their study strategies for that week. The instructor did not discuss the strategy of “studying with friends” during the lecture, and group study was not directly encouraged or discouraged during the weekly debrief activities.

Institutional Data and IRB.

The institutional review board characterized this research as exempt (IRB #2022-1337). Each survey contained an option that students could check that would remove their results from the study while still allowing them to receive the extra credit. The study information sheet provided the contact information of a researcher not associated with the class to contact to opt out of the study. After the course had ended, demographic identities of all participants were collected from the institutional data: matriculation date, major, sex, race/ethnicity, first-generation status, and low-income status (associated with Pell Grant eligibility). Demographic identity is self-reported by the students to the institution at the time of enrollment.

Statistical Analysis

We have organized the statistical analyses by each research question below. For many analyses, we made comparisons as follows: women versus non-women (reference group [RG]: men, nonbinary, and students who did not state their gender), PEER students (i.e., students who are Black, Latino/a/e/x, or indigenous to the US and its territories) versus non-PEER students (i.e., students who are White or Asian) (RG: non-PEER students), first-generation students versus non–first-generation students (RG: non-first-generation students), low-income students versus non–low-income students (RG: non–low-income students), Biological Sciences majors versus students who were non-Biological Sciences majors (RG: non-Biological Sciences majors), and a categorical variable for how students studied with friends during high school which included three levels (RG [i.e., group 1]: students who rarely or never studied with friends during high school, group 2: students who occasionally studied with friends during high school, group 3: students who regularly studied with friends during high school).

RQ1: What metacognitive knowledge about study strategies do students start college with? Does this affect course performance?

We used a linear regression model to understand the relationship between academic performance in the class (total midterm exam scores) and metacognitive knowledge of the most effective study strategies while also taking into consideration student demographic identities (gender, race/ethnicity, first-generation status, low-income status, and major) and previous academic performance (high school GPA). To indicate metacognitive knowledge about study strategies, students ranked, from 0 to 100, how much they predicted that each strategy (explain, draw, rewatch, summarize, practice problems, listen, flashcards, reread, friends, videos, and rewrite) was supported by research. We used an ordinary least squares (OLS) regression because the course exam scores are a quantitative variable that has an approximately normal distribution. The assumption of the OLS model is that there is a linear relationship between the predictor variables (knowledge of effective study strategies and demographic identities of students) and the response variable (total midterm exam scores). Assuming we have a sample of η independent observations, , we estimated . Let y_i be the total midterm exam score, be the associated predictor vector, and be the random error for the ith student, which is given by , that is:

We calculated the 95% confidence intervals for the average students’ perception of the effectiveness of 11 different study strategies: 1) explaining concepts out loud from memory 2) drawing diagrams, charts or pictures, 3) rewatching course videos, 4) condensing or summarizing notes, 5) answering practice problems from memory, 6) having someone explain concepts to me while studying, 7) using flashcards without looking at notes, 8) re-reading notes or the textbook, 9) studying with friends, 10) watching noncourse videos on the topic, and 11) re-writing notes from memory.

RQ2: Do students who identify as part of different demographic groups vary in metacognitive knowledge about study strategies?

Separate linear regression models for five common study strategies (solving practice problems, drawing diagrams, explaining from memory, listening to others, rewatching course videos, and studying with friends) were used to understand the relationships between student demographic groups (gender, race/ethnicity, first-generation status, low-income status, and whether or not the student was a biological sciences major) as well as previous academic performance (high school GPA) and metacognitive knowledge of study strategies. To measure metacognitive knowledge, students ranked from 0 to 100 how much they predicted that each strategy (solving practice problems, drawing diagrams, explaining from memory, listening to others, rewatching course videos, and studying with friends) was supported by research from 0 to 100. Assuming we have a sample of η independent observations, , we estimated . Let y_i be the study strategy, be the associated predictor vector, and be the random error for the ith student, which is given by . We fit an OLS model for each study strategy given by:

RQ3: Who studies with friends, and what are their primary study strategies?

There were several analyses in this research question.

We first compared the general popularity of studying with friends before the course and at the end of the course. We used the χ² goodness of fit test to compare proportions from Week 0 to 10 for those who rarely or never study with friends, occasionally study with friends, or regularly study with friends. The test compares the observed frequency distribution with the frequency distribution under the null hypothesis (where the null hypothesis was that the counts would be evenly distributed across the rows and columns).

We then examined whether subgroups of students changed their group study behavior. This analysis compares the frequency of group study at Week 0 within student groups of all major demographic identities (gender, race/ethnicity, first-generation status, low-income status, and whether or not the student was a biological sciences major) and repeated that analysis for the frequency of group study at Week 10 using a χ² goodness of fit test.

We were also curious what factors might predict group study by the end of the course, so we modeled the log odds of a student regularly studying with friends to address our specific research question. We used a logistic regression model since the response variable has only two outcomes (1 = the student regularly studies with friends at the end of the term, 0 = the student does not regularly study with friends at the end of the term). More specifically, we wanted to know if there is an increase in the odds of a student regularly studying with friends at the end of the term for students of different demographic groups (gender, race/ethnicity, first-generation status, low-income status, and whether or not the student was a biological sciences major), different previous academic performance (high school GPA), and whether or not the student studied with friends during high school (studies with friends in the Week 0 survey). To accomplish this, we fit a logistic regression model utilizing the stats package in R (R Core Team, 2019). Assuming we had a sample of n independent observations, , we estimated . Let Y be whether or not the student regularly studies with friends at the end of the term, and the probability that the student regularly studies with friends be , and let be the associated predictor vector. We assumed there is a linear relationship between the predictor variables and the log-odds of the event that the student regularly studies with friends. The logistic regression model was given by , that is:

Finally, we explored how students behaved while studying in groups. We examined the difference in the average percentage of total study time spent using different study strategies (explaining concepts out loud from memory; drawing diagrams, charts or pictures; re-watching course videos; condensing or summarizing notes; answering practice problems from memory; having someone explain concepts to me while studying; using flashcards without looking at notes; re-reading notes or the textbook; studying with friends; watching noncourse videos on the topic; re-writing notes from memory) in students who rarely or never studied with friends, those who occasionally studied with friends, and those who regularly studied with friends. We conducted an overall F test for the difference across the three groups of students in the Week 10 survey (studying with friends: rarely or never, occasionally, and regularly) for each of the study strategies. To adjust for multiple testing, we used a Bonferroni correction with For the study strategies that had significant differences across the three groups of students (Week 10 students studying with friends: rarely or never, occasionally, and regularly), we examined the pairwise differences between each of the pairs of students (Week 10: rarely or never vs. occasionally, regularly vs. occasionally, and regularly vs. rarely or never) using Tukey's Honest Significant Difference testing (Miller, 1981; Yandell, 1997).

RQ4: Is studying with friends associated with exam performance?

We used a linear regression model to examine the relationship between academic performance in the class (total midterm exam scores) and which study strategies the students used (Week 10: explain, draw, rewatch, listen, flashcards, reread, friends, videos, rewrite) while also taking into consideration the demographic identities of students (gender, race/ethnicity, first-generation status, low-income status, and major), previous academic performance (high school GPA), and whether or not the student regularly studied with friends (RG: occasionally, rarely, or never studied with friends at the end of the term). We assume that there is a linear relationship between the predictor variables and the response variable (total midterm exam scores). Assuming we have a sample of η independent observations, , we estimated . Let y_i be the total midterm exam score, be the associated predictor vector, and be the random error for the ith student, which is given by , that is:

RQ5: Why do students study in groups? (Qualitative Analysis)

One question from the Week 10 survey asked students for a short response if they indicated they studied with friends at least occasionally (“Why do you value studying with friends?”). We performed a qualitative data analysis on the responses. The authors (K.D. and A.E.W.) conducted an inductive content analysis (Mayring, 2000) of student responses to this question. The responses were short, such as, “I value the input that they provided on subjects that were otherwise confusing or unclear to me.” Interrater reliability was determined on the basis of 42% of the data coded by both researchers using joint probability of agreement (Cohen, 1960; Uebersax, 1987), with 92% agreement. The code definitions were refined to meet 100% consensus. The two authors individually coded the remaining responses and again achieved 92% agreement. After individually coding the remaining responses, the authors discussed any disagreements and reached 100% consensus after discussion of any original disagreements. The researchers counted codes as primary if they occurred first in the response and secondary if they occurred second. Generic responses that were not codable (e.g., “I value the learning.”) were coded in the “Other” category. The results section presents summaries of the qualitative responses. The codebook (codes plus multiple example responses) is in the Supplemental Table S32.

We then used a linear regression model to examine the relationship between academic performance in the class (total midterm exam scores) and why students value studying with friends (coded qualitative responses: collaborative, explaining, learning content, social, metacognition, other) while also taking into consideration student demographic identities (gender, race/ethnicity, first-generation status, low-income status, and major) and previous academic performance (high school GPA). We assume that there is a linear relationship between the predictor variables (why students study with friends, gender, race/ethnicity, first-generation status, low-income status, and major) and the response variable (total midterm exam scores). Assuming we have a sample of η independent observations, , we estimated . Let y_i be the total midterm exam score, be the associated predictor vector, and be the random error for the ith student, which is given by , that is:

Model Selection

To address the research questions, we used backward elimination to identify a more parsimonious model. Starting with the full model, we sequentially removed the predictor that provided the greatest reduction in the Bayesian Information Criterion (BIC) until the BIC could no longer be reduced. Specifically, BIC assesses the fit of a model to the data, while including a penalty for each additional term included in the model. We chose to use BIC since it applies a stronger penalty for model complexity compared with the Akaike Information Criterion (AIC), leading to the selection of smaller, more parsimonious models (Chakrabarti and Ghosh, 2011). Furthermore, “the BIC is more useful in selecting a correct model, while the AIC is more appropriate in finding the best model for predicting future observations” (Chakrabarti and Ghosh, 2011). Any terms included in the final (reduced) model are considered important to explaining the data, even if they don't achieve statistical significance (e.g., p < 0.05) (Anderson and Burnham, 2002). This sequential model selection process reduced complexity without sacrificing explanatory power and helped ensure that the retained predictors would generalize beyond our sample. The supplemental material contains the tables of the variables that were removed, the resulting BIC after each elimination, and the change in BIC (i.e., ΔBIC) to ensure reproducibility and transparency. In addition, we conducted a likelihood ratio test (LRT) to compare the goodness-of-fit of the complex model (i.e., the initial or full model) with a simpler restricted model (i.e., the final or reduced model) by calculating the ratio of their maximum likelihood values. The null hypothesis is that the reduced model is sufficient for describing the relationship between the remaining terms in the model and the outcome variable. The alternative hypothesis is that the complex model significantly improves the fit (i.e., the full model is superior to the reduced model) (Chambers and Hastie, 2017). Thus, we are using the LRT as a safeguard in our model selection process to ensure that we can safely remove the variables and favor the model with fewer parameters. These Supplementary Tables are provided for all models.

RESULTS

This study attempts to determine why students choose to study with friends, how they study with friends, and whether this study strategy is helpful. Our analysis below is organized by research question; Supplemental Table S1 summarizes the research questions and the tables and figures associated with each.

RQ1: What metacognitive knowledge about study strategies do students start college with? Does this affect course performance?

When students were asked to rank, from 0 to 100, which strategies researchers consider to be the most effective, they gave all study strategies very similar mean scores, ranging from 56 (SD = 28) to 77 (SD = 20). They rated two strategies that are supported by research to improve metacognition highly: explaining concepts out loud and drawing diagrams, charts, or pictures. However, they ranked answering practice problems (a high-impact study strategy) lower than two strategies (re-watching course videos and summarizing notes) that have low research support. And while rewriting notes from memory is supported as a high-impact study strategy that requires retrieval, it was ranked lower than all other strategies, including listening to someone explain the concepts and studying with friends (Figure 1; all data in Supplemental Table S2).

FIGURE 1.

The average metacognitive knowledge about study strategies by students is modest. Symbols show the student responses at Week 0 to the metacognitive knowledge question: “Researchers have tried to measure the effectiveness of different strategies for students studying for an exam. Predict the effectiveness of each strategy, where 0 = least effective and 100 = most effective. Give each strategy a different grade.” Answers are mean responses +/1 SEM. The four strategies most supported by evidence are indicated in gold. Only two of the four are ranked highly by students. Additionally, all strategies have similar scores with averages between 56 and 77.

We next explored whether knowledge of the most effective study strategies is associated with higher performance in this class (Table 3). Examining the impact of variables that affect exam performance, there was a positive association between high school GPA and exam performance (b_{HS GPA} = 18.76, p < 0.001), and a negative association for women compared with men (b_women = −4.81, p < 0.001), PEERs compared with non-PEERs (b_PEER = −2.61, p = 0.012), first-generation students compared with non–first-generation students (b_FG = −0.37, p = 0.782), and low-income students compared with non–low-income students (b_low-income = −2.86, p = 0.004). Furthermore, students who were both first-generation students and PEERs had an additional negative impact of exam scores compared with non–first-generation and non–low-income students (b_{linteraction: FG*PEER} = −4.69, p = 0.005). Metacognitive knowledge of study strategies had small effect sizes while being important predictors of exam performance. Students who ranked low-impact strategies higher did slightly worse than students who ranked low-impact strategies lower (having someone explain concepts to me while studying: b_listen = −0.06, p < 0.001; watching noncourse videos on the topic: b_{watch other videos} = −0.04, p = 0.008, re-watching course videos: b_{rewatch course videos} = −0.06, p = 0.001).

TABLE 3.

RQ1. Metacognitive knowledge is associated with exam performance—final linear regression model based on backward elimination

	Estimate	Standard Error	t value	p value
(Intercept)	−4.98	9.11	−0.55	0.585
Women	−4.81	0.87	−5.54	<0.001
PEER	−2.61	1.04	−2.52	0.012
First Generation	−0.37	1.35	−0.28	0.782
Low Income	−2.86	0.99	−2.89	0.004
High School GPA	18.76	2.14	8.75	<0.001
Week 0 Research: Rewatch	−0.06	0.02	−3.25	0.001
Week 0 Research: Listen	−0.06	0.02	−3.82	<0.001
Week 0 Research: Video	0.04	0.02	2.66	0.008
PEER x First Generation	−4.69	1.66	−2.83	0.005
Adjusted R2: 0.415
F-statistic: 40.48 on 9 and 492 DF, p-value: < 0.001

Note: The response variable is the total score on the midterm exams.

RQ2: Do students who identify as part of different demographic groups vary in metacognitive knowledge about study strategies?

It is possible that metacognitive knowledge of study strategies might vary significantly by gender, ethnicity, or other demographic identities that are often associated with past educational experiences or access. We looked for significant differences in study strategy knowledge in students who identify with different demographic groups for the six most commonly used strategies. The independent linear models showed few significant differences in metacognitive knowledge of study strategies among students associated with different demographic groups. Furthermore, the final models themselves had very low predictive power (i.e., adjusted R₂ < 0.10), and thus we only include the analysis tables in the supplement (Supplemental Tables S5–S22) and suggest that instructors do not assume that metacognitive knowledge of study strategies can be predicted based on demographic groups and prior academic performance.

RQ3: Who studies with friends, and what are their primary study strategies?

In this project, we were particularly interested in students who study in peer study groups (“studying with friends”), a strategy that has only modest research support. In this research question, we examined metacognitive self-regulation as indicated by student use of study strategies.

We first compared how many students changed their amount of group study during their first quarter at university. Students were surveyed at the start of the quarter (Week 0) on how they studied with friends during high school. At the end of the course (Week 10), we then asked students how they studied for the last midterm exam in this course. In high school, students studied with friends less often, with only ∼10% saying they regularly study with friends, 45% saying they occasionally studied with friends, and 45% saying they rarely or never studied with friends (descriptive statistics of the frequencies of all the study strategies used are provided in Supplemental Table S28). There was a significant increase in studying with friends by the end of the term (Figure 2, full contingency table in Supplemental Table S23, χ² = 27.88, p < 0.001), with 19% saying they regularly study with friends, 45% saying they occasionally studied with friends, and 37% saying they rarely or never studied with friends at Week 10.

FIGURE 2.

Students increase their studying with friends by the end of the term. The Sankey diagram shows that students who rarely and occasionally studied with friends before college tended to increase their studying with friends by the end of this introductory biology class. The arrow width is proportional to student counts; colors map to Week 10 categories (navy, regularly; gold, occasionally; gray, rarely/never). The number of students who rarely or never studied with friends decreased. χ² = 27.88, p < 0.001. Constructed in Python with Matplotlib.

We next examined whether the frequency of group study varied among students who identify as members from different demographic groups. At Week 0, students who identify as first-generation and low income were less likely to have regularly studied with friends compared with continuing generation (χ² = 11.34, p = 0.003) and middle and high income students (χ² = 10.43, p = 0.005). By Week 10, students from all groups had increased their studying with friends, and there was no significant difference between the demographic groups and their reference controls. Figure 3 shows the percentages of students who regularly study with friends; the full contingency tables are in Supplemental Table S24.

FIGURE 3.

Differences in the frequency of group study by demographic group at the start of the course are not present at the end of the course. For clarity, only the percentage of students who said they regularly studied with friends is shown. There was a significant difference in frequency of group study at Week 0 in low-income students compared with non–low-income students (χ² = 10.43, p = 0.005) and first-generation students compared with non–first-generation students (χ² = 11.34, p = 0.003). By the end of the term, all groups had increased studying in groups, and there was no longer any statistically significant difference between demographic groups (i.e., women and men, first-generation and non–first-generation, PEER and non-PEER, and BioSci Major and non-BioSci Major). The full contingency tables are in Supplemental Table S24.

Our logistic regression analysis for the log odds of the event that the student regularly studies with friends at the end of the term provided consistent results with our goodness of fit results (Supplemental Table S25–S27). Demographic characteristics, previous academic performance, and how students studied with friends during high school had weak predictive power for the log odds of the event that the student studied with friends during college (Supplemental Table S25). After model selection procedures, the only predictor remaining was how students studied with friends during high school (Supplemental Table S26). Although students who regularly study with friends in high school are 4.8 times more likely to use peer study groups by the end of their first quarter in college compared with students who rarely or never studied with friends during high school, the overall predictive power of the model is weak (Supplemental Table S27).

Effective metacognition requires students to not only understand what study strategies are most effective but to implement those strategies correctly and to accurately judge when the strategies are working. We were next interested in this middle step: the study behaviors of students who use peer study groups (regular group studiers and occasional group studiers) and students who study alone (solo studiers). For this analysis, we surveyed all students on how they divided up their total study time among different study strategies for their last midterm. We used an overall F test to test whether there was a difference between the three types of group studiers (solo studiers [i.e., rarely or never], occasional group studiers, and regular group studiers) for each of the study strategies (Supplemental Table S29). There was no significant difference between regular group studiers, occasional group studiers, and solo studiers in the percentage of time spent on high-impact study strategies like explaining from memory or answering practice problems. Regular group studiers, occasional group studiers, and solo studiers are, unsurprisingly, different in the amount of time they spend studying with friends as a study strategy (F = 86.3, p < 0.001). Three other significant differences between the three groups (regular group studiers, occasional group studiers, and solo studiers) were found: having someone explain concepts to me while studying, re-reading notes or the textbook, and watching noncourse videos on the topic. We followed by conducting pairwise comparisons for each study strategy that had a significant difference across the three types of group studiers (Supplemental Table S30). Both occasional and regular group studiers spend, on average, more time having someone explain concepts to them (7.4 and 10.9% of study time for group studiers compared with 2.7% of study time for solo studiers). Solo studiers, compared with both occasional group studiers and regular group studiers, spend more time re-reading notes (15.2% of time for solo studiers vs. 10.4 and 8.3% for occasional and regular group studiers, respectively) and watching noncourse-videos on the topics (6.3% of study time for solo studiers vs. 5.1 and 2.9% of studying time for occasional and regular group studiers, respectively) (Supplemental Table S30; Figure 4).

FIGURE 4.

Students who study in groups use different low-impact study strategies than students who study alone. The figure shows the average percentage of time spent using different study strategies at the end of the term in students who reported they studied with friends regularly or occasionally compared with those who reported they did not. Students who reported that they use peer study groups (unsurprisingly) also report that they spend a significant percentage of time using “study with friends” as a study strategy. In addition, students who use peer study groups are more likely to spend study time listening to explanations and less likely to spend study time rereading notes or the textbook, and less likely to spend time watching noncourse videos. * ≤ 0.05, ** ≤ 0.01, *** ≤ 0.001.

For students who regularly or occasionally studied in groups, we also asked how those students spent their time during their group study sessions. The most common activity during group study was listening to others explain the material (on average, 19% of their time), followed by explaining (14.8%) and reading (13.2%). Answering practice problems took 12.1% of their time, and they quizzed each other 5.4% of the time.

To summarize, low-income and first-generation students are less likely to arrive at college having studied with friends in high school. By the end of the quarter, all students, regardless of identity, increased their frequency of studying with friends, and there is no longer a significant difference in the percentage of group study by identity. Students who studied with friends in high school are more likely to study with friends in college. A comparison of how students spent their study time showed that students who regularly study with friends are more likely to listen to explanations from fellow students about the topics, while students who study alone are more likely to reread their notes or watch noncourse videos on the topics. Group study time is focused on listening to others, explaining, and re-reading notes.

RQ4: Is studying with friends associated with exam performance?

Above, we have shown that students vary in their metacognitive knowledge and cannot, on average, identify all the study strategies most supported by research. They tend to rate all study strategies similarly. We then explored student self-regulated learning, particularly what study strategies students chose to actually use while studying for exams. By the end of the course, more students chose to spend their study time studying with friends, and those who do often listened to their friends explain the content.

We then looked to see which of the study strategies students used were associated with exam performance. When we controlled for high school GPA and demographic identity variables in our linear model, we found that increased use of four study strategies were associated with slightly lower exam scores: summarizing notes (b_summarizing = −0.13, p = 0.001), listening to friends explain the content (b_listening = −0.18, p = 0.001), using flashcards (b_flashcards = −0.55, p < 0.001), and re-writing notes from memory (b_{re-writing notes} = −0.18, p = 0.032). Studying with friends at the end of the term was in the initial (full) model (Supplemental Table S31), but was not included in the final (reduced) model (Table 4), implying that studying with friends during the course did not add additional predictive value for exam performance (Supplemental Table S32).

TABLE 4.

RQ4. Low-Impact study strategy use is associated with lower exam Scores—Final linear regression model based on backward elimination

	Estimate	SE	t value	p value
(Intercept)	−1.34	9.72	−0.14	0.891
Women	−2.48	1.02	−2.43	0.015
PEER	−4.20	0.91	−4.65	<0.001
First-Generation	−3.74	1.07	−3.51	<0.001
Low-Income	−2.29	1.04	−2.20	0.029
High School GPA	17.12	2.31	7.41	<0.001
Week 10: Summarize	−0.13	0.03	−3.95	<0.001
Week 10: Listen	−0.18	0.05	−3.51	<0.001
Week 10: Flashcards	−0.55	0.13	−4.21	<0.001
Week 10: Rewrite	−0.18	0.08	−2.27	0.024
Adjusted R2: 0.423
F-statistic: 31.44 on 9 and 365 DF, p-value: < 0.001

Note: The response variable is the total score on the midterm exams.

RQ5: Why do students study in groups?

The effect of peer study groups will likely be influenced by the reason students choose to attend. We ended our survey by asking students who occasionally or regularly studied with friends why they valued this strategy. The coded short answer responses fell into the following categories: 1) learning (student values receiving help or explanations from others, 38%), 2) collaboration (student values teaching others and having others teach them, 32%), 3) metacognition (studying with others keeps the student on task, improves studying, 12%), 4) social (student values the presence of others to feel comfortable or happy, 10%), 5) explaining (student values learning by explaining the material to others, 2%), and 6) other (6%). Examples of statements associated with each code are in Supplemental Table S33. The full code counts and percentages are listed in Supplemental Table S34.

Learning and collaboration were the most frequent reasons given by students for why they valued studying with friends. We found that these two codes came from similar statements from students. Statements were coded as “learning” if the student only mentioned that they learned from the group. Statements were coded as “collaboration” if the student said both they learned and they helped others learn. The social, metacognition, and explaining codes were rare. When we conducted model selection, none of the reasons why students valued studying with friends (learning, collaboration, metacognition, social, explaining, and other) remained in the final model, suggesting that the expressed reason for studying with friends provides an insignificant increase in predictive power on exam scores when we already take into account previous high school performance and student demographic characteristics (High school GPA, gender, first-generation status, PEER status) (Supplemental Table S37).

To summarize, students who study with friends most often say the value is from listening to their friends explain or from taking turns listening and explaining. None of the codes associated with the student responses was associated with exam performance in the class.

DISCUSSION

This study seeks to determine why some students choose to study with friends and whether it is effective for them. We gathered information at both the start and end of a first-term introductory biology class that contained significant in-class peer learning. We found that most students entered introductory biology demonstrating an incomplete understanding of effective study strategies (Figure 1), and students with a better understanding had higher exam scores (Table 3). First-generation and lower-income students studied with friends less often in high school, but in this course, students from all demographic groups spent more of their total study time in peer study groups by the end of the term (Figure 3). Students who regularly studied with friends had exam scores that were not significantly different from those of students who studied alone. But we also found that students who spent more of their total study time summarizing their notes, listening to friends explain content, or using flashcards did worse on exams (Table 4). Students who studied in peer groups spent more time listening to others explain the material, while students who studied alone spent more time rereading notes and rewatching course videos (Figure 4). Students seem to begin exam weeks with an incomplete understanding of the material and choose either solo study or peer group study to better encode the information. Both groups spend too much time using encoding strategies near exam time, and therefore, group study generates no comparative increase in exam score.

Metacognitive Knowledge of Study Strategies Is Rarely Measured

The framework of memory described by Bjork et al. (2013) describes the first step of learning as the selection of the study strategy to use in knowledge acquisition. The selection students make is based on their knowledge of what strategies are effective, and correct knowledge is needed for effective selection. This ability to describe what study strategies are associated with effective exam performance is a part of metacognition called metacognitive knowledge (Stanton et al. 2021). But metacognitive knowledge is rarely measured directly. Instead, much of the existing education research quantifies student metacognition by measuring student behavior (e.g., in biology, see van Vliet et al., 2015; Walck-Shannon et al., 2021; Hazlett et al., 2025). In this study, we directly measured students’ metacognitive knowledge about study strategies. We asked first-term college students in introductory biology to grade study strategies according to their perceived research support. This is one of the first studies we are aware of to gather this information. These undergraduates are only moderately able to complete this task, and we found that demonstrated metacognitive knowledge of study strategies was associated with a small but significant effect on performance in the class (Table 3).

In addition, researchers rarely compare how students from different demographic groups differ in their knowledge of effective study strategies. We did not find higher metacognitive knowledge of study strategies in students who identify as any of the measured demographic groups, which indicates that all students would benefit equally from additional training. Undergraduates regularly report that they are not taught effective study strategies (Kornell and Bjork, 2007). When asked which strategy would be most effective in a study scenario, students often choose incorrectly (Figure 1, McCabe, 2011; Blasiman et al., 2017). Many instructors also regularly give advice not supported by research (Morehead et al., 2016; Ewell et al., 2025). Although there is evidence that co-curricular programs and workshops can effectively increase the metacognitive knowledge (Hattie et al., 1996), students are more likely to be able to immediately apply what they learn if the support is in the context of a course (Bernacki et al., 2023). We therefore encourage instructors to incorporate the most evidence-based information about effective strategies into their courses.

The Use of Peer Study Groups Increased During the Course

Multiple studies have measured how common peer study is in undergraduate learning. But little data have been gathered on which students choose to study with friends. Chi and Kadandale (2022) offered students the option to join learning communities and found no significant difference in participation among students of different demographic groups, but transfer students were slightly more likely to choose to be in a group. Williams et al. (2021) surveyed first-year college students across multiple courses and found men were slightly more likely to say they studied with friends, but found no difference associated with income, first-generation status, or ethnicity. In this study, both low-income and first-generation students reported less studying with friends in high school. By the end of their first college term, students of all demographic groups had increased their frequency of studying with friends, and there was no longer any difference by identity (Figure 3). There are several factors that might have encouraged all students to use peer study groups in this course. Students were encouraged to sit with peers from their discussion groups in class, to collaborate during clicker questions, and to use explaining out loud as a study strategy. We did not directly encourage or discourage students to study with friends, and specifically said that explaining does not require an audience. We do not have a control class to judge whether this course design caused the increase in student use of peer study groups. Sokolove and Marbach-Ad (1999) showed higher peer study groups in an active learning course compared with a more traditional course, but students in the active learning section were specifically encouraged to use study groups (1999). We do find it promising that the use of peer study groups by low-income and first-generation students matched that of other students by the end of the course, indicating that all students who wanted a connection with others in the class seemed to be equally able to find it. Students who have access to a community of peers show reduced anxiety and have a higher sense of belonging and higher motivation (Wu et al., 2024). Undergraduate students in learning communities also show higher STEM retention, though the effects on GPA seem minimal (Johnson et al., 2023).

Study Group Strategies Are Collaborative and Context-Specific

Our quantitative analysis showed students in study groups spent most of their group study time “having someone explain concepts to me” (19% of the time) or “explaining concepts out loud from memory” (14.9% of the time). This focus on having peers explain course concepts during group study is also seen in the research of others, although other studies used less precise measurements. Wei et al. (2024) found that psychology undergraduates who studied with others most often choose “teaching each other” (44% of students), “quizzing each other” (40% of students), and “asking for help” (22% of students) as strategies they used while studying with friends. Wissman and Rawson (2016) sampled students across many different majors. About 68% of students who studied with friends discussed notes, 78% went over a study guide, and 86% quizzed each other. McCabe and Lummis (2018) surveyed students in multiple psychology, mathematics, and biology courses. They found the most common strategies during group study were asking questions, discussing course materials, and quizzing each other (reported as average Likert scores). These studies showed similarities to ours, with various versions of teaching each other, asking questions, and going over notes together. Course design likely influences study strategy choice, particularly for strategies like quizzing each other. These other studies surveyed students from many different courses or focused on psychology undergraduates. These other studies also only characterized whether students used the strategy at all, or used it “often.” Our students were asked to explicitly report the percentage of time spent on each study strategy while studying with friends, and ours reported spending only 5% of their time quizzing each other. We hypothesize that this difference is because of exam design. The exams in this course were primarily application-level problems rather than memorization-level problems, and students who reported a higher use of flash cards did significantly worse in the course. Most of our students likely found that quizzing of key terms and vocabulary was not likely an effective use of study time. This difference between our results and those of other studies indicates that the strategies students use while studying with peers is expected to vary considerably depending on the course. But if instructors are using assessments with a significant amount of application, we predict students will spend most of their peer study time asking each other questions about the course material and less time using flashcard-level quizzing. It would be interesting to compare our quantitative results with those in another introductory biology course with similar students but more memorization-focused exams.

Our qualitative analysis showed that group study enabled students to learn from others or see other perspectives. This matches our quantitative results above—that students primarily spend group time listening or discussing. It also corresponds to the qualitative research on academic help-seeking behavior, where researchers found a theme of “mutual aid” to be a common one in students who were comfortable seeking help (Ruihua et al., 2025). Research also suggests that students form friendship groups based on student similarity (McCabe and Lummis, 2018) and similar academic achievement (Brouwer and Engels, 2022). This likely means that most study groups are students taking turns trying to understand and explain difficult concepts, and students who choose peer study groups feel that it is helpful for learning. Our focus as instructors should therefore not be to discourage peer study but to make sure it is only a small part of a student's study strategy plan.

Comparing Study Strategy Use During Peer Study and Solo Study

Some students only study alone, and even students who regularly study with friends also spend time studying alone. We compared students who regularly studied in groups with students who rarely or never studied in groups to see how students spent their total time studying for the last midterm. Both groups spent a similar percentage of time explaining concepts out loud, even though not all studied with friends. Although it might seem like solo studiers could not utilize explaining as a study strategy, the instructors in this class explicitly encouraged all students to spend time explaining out loud regardless of whether anyone was listening, and presented research that demonstrated the effectiveness of this strategy (e.g., Koh et al., 2018; Fiorella and Kuhlmann, 2020). The largest difference between the two groups was not in the high-impact strategies, but in the low-impact strategies. Students who regularly studied with friends spent a higher percentage of time listening to explanations. Students who regularly studied alone spent a higher percentage of time rereading their notes and book and watching noncourse videos. We consider this one of the primary findings of this study—all students spend a significant amount of study time on study strategies that are not considered high impact, but group studiers focus on listening to their friends explain content to them, and solo studiers focus on rereading their notes and watching resource videos (Figure 4). This likely explains why, when controlling for high school GPA and student demographic identities, we found no significant relationship between studying with friends and exam score (Table 4).

Do these results match those of other researchers? Few studies have attempted to determine if using peer study groups improves exam performance, and fewer still collect student-identified information about study group use and exam scores and control for preparation factors like high school GPA. Studies may depend on students to provide GPA or exam scores (Rybczynski and Schussler, 2011; McCabe and Lummis, 2018) or use instructor-supported groups (Zhou et al, 2023; Chi and Kadandale, 2022). Our study differs from these. We gathered precise information on student study strategies both inside and outside of peer group study, and have more complete statistical control of student preparation levels. We also allowed students to choose on their own whether to study with friends. This level of precision gives us more confidence in our conclusion that students who independently choose to use peer study groups do not use more effective study strategies or have higher exam grades.

Application: Teach Students Metacognition Instead of Study Strategies

Our findings show that many students spend too much time the week before exams listening to friends explain concepts, re-reading notes, and watching online videos. But we also find that students find value in peer study groups, and study group use increases during the term. Because of this information, we use a modified approach when we talk about study strategies in our class. Rather than telling students whether they should or should not study with friends or re-read notes, we instead focus on teaching students how to determine the best time to study with friends or re-read notes.

The research on memory and learning has generated a conceptual framework that focuses on three aspects of memory: encoding, storage, and retrieval (Roediger and Guynn, 1996). Encoding is the term used to describe initially learning information—understanding the topic and connecting it to existing knowledge. Storage is the term for the neural synapse building that occurs when information is moved to long-term memory. And retrieval is the process of producing the stored information upon demand (Owens and Tanner, 2017). Here, we asked students to describe the research support for several study strategies (metacognitive knowledge), and students produced average ratings very similar to each other (Figure 1). We conclude that many students judge all strategies as helpful because they do not recognize that some strategies are good for encoding and others for retrieval, and they do not understand that what researchers mean by “effective” study strategies are specifically retrieval strategies that are effective in a testing environment. Because of this, we are developing an intervention to teach students that none of the listed study strategies are bad, but some are encoding strategies and should be done as part of daily studying, and others are retrieval strategies and should be the focus during exam preparation. We are deliberately naming strategies and timings, providing training on their use, and regularly providing feedback on how well students are implementing the strategies (Dignath and Veenman, 2021). In our next project, we will report on whether this training is effective at improving both the demonstrated metacognitive knowledge about study strategies and demonstrated self-regulated learning.

Limitations

Although this study provides new information about student use of peer study groups, it does have several limitations. Like most research on study strategies, we asked students about their study strategies after studying was completed, so we are measuring how students think they studied (or wish they studied). It is possible that students answered the behavior surveys to give the answers that the instructors wanted, although our first survey collected information before the class was underway, and we did not direct students to include or avoid peer study groups. Surveys at the end of the class had many students who said they used encoding strategies like listening to others and rereading notes during the exam week, so many clearly felt comfortable being honest. A contextual limitation of this study is that course design and exam type likely affect the strategies used during study groups. For example, other studies indicate that quizzing each other is often used in psychology study groups, but was not used by our students. In addition, students study differently for open-note exams such as ours (Driessen et al., 2022). We recognize that students will adjust their studying to fit the exams of the course (Stanger-Hall, 2012; Ewell et al., 2023). But because most students rely on encoding strategies more than they should, encouraging earlier use of encoding strategies will be appropriate regardless of the course. Lastly, we did not collect information on how students actually used a study strategy. For example, both students who studied alone and studied with friends said that they used “explaining” as a strategy. It would be interesting to explore how much learning occurs when explaining to an inanimate object compared with explaining to a real friend who gives feedback but may also interrupt.

CONCLUSION

Metacognition can be summarized as knowing what study strategies are most effective and using those strategies in the appropriate way and at the appropriate time. Students will naturally use encoding study strategies first (gathering information, organizing it, clearing up confusions), as this is the first part of memory building. Some students will do this within a group, others will do it on their own. The most effective students will then switch to retrieval strategies and have sufficient time to use them before an exam.

Our work provides several new findings. First, students arrive in college demonstrating an incomplete understanding of the research on study strategies. If their first biology course has difficult application problems on the exams, students who rate encoding strategies as high impact do worse in the course. There is no student demographic group that indicates a more reduced understanding of study strategy effectiveness, so all students should receive support. Second, although low-income and first-generation students are less likely to arrive at college having studied with friends, all students will increase their studying with friends when they are encouraged to form a community with their classmates. Third, group study itself is not a predictor of exam scores, but using encoding strategies during exam weeks is predictive of lower exam scores. We therefore encourage instructors not to focus on the context of studying (alone or with friends). Instead, focus on the metacognitive knowledge needed to describe the high-impact strategies and the self-regulation needed to use encoding strategies earlier and high-impact retrieval strategies later, as the exam approaches.