Introduction

Biology is a graphic science, wherein the learners study living organisms and their interactions with one another and their environments (Goodwin and Dawkins, 1995, p. 47). This is a broad definition since the scope of biology is vast. High school biology students often feel difficulty in mastering the subject. The potential reason for this is due to complex materials, invisible/intangible objects, and complex terms (Cimer, 2012). This difficulty in learning could influence students’ variables such as their performance, attitudes, behaviors, knowledge, skills, etc. Furthermore, traditional education has not been efficacious in solving these issues (Ebrahim and Naji, 2021; Yapici and Akbayin, 2012). Thus, many non-traditional pedagogical models/approaches are being developed, tested, and implemented for the efficient teaching of biology concepts. The reported non-traditional, students centered pedagogical models in high school biology classrooms include inquiry-based, problem-based, project-based, virtual simulation-based, game-based, argument-based, etc. (Klisch et al., 2013; Nunaki et al., 2019; Ping et al., 2020; Sivia et al., 2019a; Thisgaard and Makransky, 2017; Thurrodliyah et al., 2020; Yapici and Akbayin, 2012). Furthermore, a combination of these models has also been reported (Anazifa and Djukri, 2017; Lui and Slotta, 2014; Thompson et al., 2020).

Despite the various pedagogies employed, it is often tricky to conclude the best pedagogical approach in high school biology education. Even the literature shows disproportionate results with varying levels of effectiveness. Such as the reported impact of problem-based learning is [d = 0.89 (Xu et al., 2021)], project-based learning are [d = 1.36 (Balemen and Keskin, 2018; d = 0.95 (Ayaz and Söylemez, 2015)], inquiry-based are [d = 1.26 (Funa and Prudente, 2021); d = 0.35 (Wang et al., 2011)] and web-based is [d = 0.668 (Bayraktar, 2001)] in biology education. This difference in effect sizes might be due to the diversity in students (low/average/high/mixed achievers, gifted, at-risk students, etc.) and their pedagogical requirements (cognitive, affective, behavioral learning gains). Different students may benefit disproportionately from different learning gains (Korkor Sam et al., 2018; Steenbergen-Hu et al., 2020).

Therefore, this article seeks to compute the impact of various non-traditional educational models, particularly for a mixed-ability (low/average/high achievers) biology classroom in terms of students’ learning gains (cognitive, affective, and, behavioral). This meta-analysis intends to compute the impact of non-traditional (modern) pedagogical models and offer pedagogical justifications for the requirement of efficacious and appropriate practices in high school biology. For this, the study has first investigated the diverse pedagogical approaches that are being employed in high school biology classrooms. The overall effectiveness of the non-traditional pedagogies compared with the traditional lecture model has also been inspected. Subsequently, a comparative examination of different teaching approaches has been conducted, taking into account the different learning gains. Therefore, the research questions addressed in the paper are:

  1. 1.

    What are the various non-traditional pedagogical models/ approaches employed in high school biology education?

  2. 2.

    What is the overall impact of the non-traditional pedagogical models employed for mixed-ability high school biology classrooms (when compared to the traditional lecture model)?

  3. 3.

    What is the comparative effectiveness of each pedagogical model, concerning the students’ gains (cognitive, affective, behavior)?

Review of literature and conceptual framework

non-traditional pedagogies in biology education

Literature has witnessed the implication of diverse teaching approaches that are often employed explicitly (i.e., targeting a single pedagogical model) or in conjugation (targeting two/more models). Dichotomizing these, project-based pedagogy in biology, a student-centered and multidisciplinary approach, is often employed, where students work on projects to investigate and find answers to complex questions/problems. Prior literature reports that project-based models are often employed for teaching genetics (Sivia et al., 2019a) and animal physiology such as digestive/circulatory/respiratory systems (Sukmawati et al., 2019; Anazifa and Djukri, 2017; Blacer-Bacolod, 2022). This approach in high school biology has reported improved knowledge, critical thinking skills, civic engagement, conceptual understanding, and application skills (Sivia et al., 2019a; Sukmawati et al., 2019; Anazifa and Djukri, 2017; Blacer-Bacolod, 2022; Sukmawati et al., 2019; Sari et al., 2019). In contrast, the problem-based model is often single subject/problem-based, employing case studies/fictitious scenarios as the problem, which helps students acquire 21st-century quintessential skills of problem-solving. Problem-based model is often used to instruct and educate students on environmental issues (Özalemdar, 2021; Thurrodliyah et al., 2020; Thinkhamchoet et al., 2021). Many studies have already shown enhanced students’ skills of reflection, knowledge, creativity, critical thinking, positive attitudes and behavior, and psychomotor learning outcomes (Anazifa and Djukri, 2017; Thurrodliyah et al., 2020; Özalemdar, 2021; Hugerat et al., 2021; Kolarova et al., 2014). The primary similarity between the project-based and problem-based models is their focus on open-ended questions/problems, driving the inquiry process. Another model is the inquiry-based pedagogical model that supports students to acquire knowledge independently via the inquiry process (Hadjichambis et al., 2022; Nunaki et al., 2019; Wilson et al., 2010). Studies have reported the successful execution of inquiry-based models for environmental biology and biodiversity (Ristanto et al., 2022; Lui and Slotta, 2014; Hadjichambis et al., 2022), and cell biology (Saputri et al., 2019; Thompson et al., 2020; Ping et al., 2020). Many studies have reported better students’ achievements, metacognitive skills, perception, and, the conception of biology learning via the inquiry approach (Hadjichambis et al., 2022; Kagnici and Sadi, 2021; Nunaki et al., 2019; Ristanto et al., 2022; Saputri et al, 2019; Wilson et al., 2010). Inquiry-based teaching has also been executed in collaboration with argument-based models (Ping et al., 2020; Ristanto et al., 2022), and game-based models (Lui and Slotta, 2014; Thompson et al., 2020) for high school biology. And, it is reported to promote critical thinking, argumentation skills, and science process skills, (Lui and Slotta, 2014; Ping et al., 2020; Ristanto et al., 2022; Thompson et al., 2020). Therefore, other than typical pedagogical alternatives (e.g., project-based, problem-based, inquiry-based, etc.); the gamification approach has gained much traction. A possible justification is the contribution of the game-based model toward students’ engagement and interactivity, which are supposed to be the most challenging variables that should be considered (Cai et al., 2022). Similarly, modeling and virtual simulations have also been reported to be impactful in promoting the knowledge and understanding of complicated biological concepts such as cell biology and genetics (Marbach-Ad et al., 2008; Mulder et al., 2016; Thisgaard and Makransky, 2017; Li and Ma, 2010). Also, a careful blend of pedagogical approaches (termed as the blended model) in a wisely framed online and offline setting is often recommended to improve students’ cognitive skills, achievements, and attitudes toward the course contents and the internet (Yapici and Akbayin, 2012; Ebrahim and Naji, 2021; Kazu and Demirkol, 2014).

Students’ learning gains

A systematic review of diverse high school pedagogies in association with learning gains seems to be an underexplored area (in the context of biology). The notion of students’ learning gain is defined as growth or change in knowledge, skills, and attitudes over time (Cronbach and Furby, 1970; Roohr et al., 2017). The affective learning gains account for attitudes, confidence, motivation, satisfaction, and well-being. While behavioral learning gains account for students’ behavioral skills such as engagement, leadership skills, and teamwork. Cognitive learning gains pertain to skills associated with cognitive growth, including comprehension, information retention, critical thinking, creative thinking, logical reasoning, analytical thinking, and scientific reasoning. (Bloom et al., 1956). In psychology and education, the interlinked affective, behavioral, and cognitive learning gains were used to understand and unravel the multidimensional notions of learning gains (Ostrom, 1969). However, the range of learning gains evaluated in the literature seems to be extensively diverse (in the context of biology education). For example, a longitudinal study on nearly 17000 students across 50 US colleges was conducted to study a range of learning gains. Their study reported positive learning gains in engagement, critical thinking, moral reasoning, and leadership; while negligible learning gains in literacy, and political and social engagement (Pascarella and Blaich, 2013). In contrast, some studies evaluated only particular kinds of learning gains. For instance, some studies only concentrated on the cognitive learning gains of biology students (Nunaki et al., 2019; Özalemdar, 2021; Ping et al., 2020; Ristanto et al., 2022; Thompson et al., 2020). While some others focused only on affective gains (Anazifa and Djukri, 2017; Brom et al., 2011; Venville and Dawson, 2010). This diversity in evaluating the learning gains might be predisposed by the diversity in student needs. This was well explained by a meta-analysis and systematic review by Steenbergen-Hu et al., 2020). Their study has reported that gifted underachievers benefit more from pedagogies focusing on affective and behavioral gains (such as self-efficacy, goal evaluation, positive perceptions, motivation, and psychosocial functioning), rather than cognitive gains such as course grades (Steenbergen-Hu et al., 2020). In this context, Korkor Sam et al. (2018) insist on the 3E learning model (exploration, explanations, and expansion) for improving the performance of low achievers in biology (Korkor Sam et al., 2018). Recommendations by Chaplin (2007), include modeling and coaching learning in classrooms to augment the critical thinking skills of students at risk in introductory biology courses. Interestingly a study by Yaduvanshi and Singh, 2019 has recommended structured cooperative learning for mixed achievers (low, average, and high achievers) in secondary biology classrooms (Yaduvanshi and Singh, 2019). Therefore, concluding the best pedagogical model in a mixed-ability high school biology must contemplate the wide-ranging students’ needs (cognitive, affective, and behavioral gains in collaboration) (Wei et al., 2021).

Methods

The research employed a systematic literature review approach in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. (Page et al., 2021). The screening for studies as per PRISMA methodology was employed in three stages (Fig. 1). The three stages include (1) Identifying the articles employing a specific search strategy; (2) Initial screening of the articles based on title, abstract, and content relatedness; and (3) Final screening of the studies based on predefined inclusion/exclusion criteria.

Fig. 1: PRISMA 2020 flow diagram.
figure 1

The figure reveals the inclusion-exclusion criteria for studies included in the meta-analysis.

Full size image

Search strategy

The relevant research articles were gathered using widely used web search engines like “Web of Science,” “Journals for Educational Research Information Center (ERIC),” and “Scopus”. Specific keywords were included for searching the articles: (“STEM” OR “science” OR “biology” OR “genetics” OR “anatomy” OR “botany” OR “zoology”) AND (“teach*“ OR “learn*“ OR “pedagog*“ “education” OR “*school” OR “grade 10” OR “grade 11” OR “grade 12”) AND (“online” OR “digital” OR “laboratory” OR “project” OR “discovery” OR “blended” OR “flip*” OR “inquiry” OR “discovery” OR “problem-*“ OR “game” OR “virtual-*” OR “immersive”). All the papers included in the study were empirical research studies.

A total of 799 papers were located through keyword searches: Scopus (n = 330), ERIC (n = 212), and Web of Science (n = 257). The identified records were exported to Excel to find the duplicates. Wherein, 327 articles were discarded due to repetition. The articles were screened for apt title, gist, abstract, appropriate context, inclusion/exclusion criteria, retrieval, and relevancy (Fig. 1). The initial selection of the studies was based on skim-reading the title and abstract. Followingly articles have been thoroughly investigated for context-relatedness. And finally, the whole article has been reviewed for relevance. The context-relatedness and relevancy were checked via inclusion and exclusion criteria, set under the scope of the review paper.

Inclusion/exclusion criteria

The inclusion criteria for article selection were: Biology education model at senior high school (e.g., Grades 10–12); Articles extracted from 2008–2022 March; Peer-reviewed articles; Articles in English. The exclusion criteria were: Education level other than high school (e.g., junior high school, middle school, UG, PG, etc.); The teaching models for non-biology subjects (chemistry, physics, language, mathematics, etc); and qualitative research and review papers. The final exclusion criteria were as follows: non-empirical studies; studies lacking a pre-/post-test or control/experiment design; studies with data that did not conform to the Comprehensive Meta-Analysis (CMA) format (e.g., mean, standard deviation, Cohen’s d, t-value, p value, etc.); studies presenting results based on non-student variables (e.g., teachers’ perception, author’s opinions, etc.); insufficient data for calculating effect size (i.e., p value/t-value = 0). Ultimately, 32 peer-reviewed empirical articles met all the criteria and were included in the meta-analysis Table 1 displays the descriptive characteristics of the 32 selected studies considered for the meta-analysis review: (a) studies, (b) pedagogical model (c) sub-pedagogical approach, (d) study design, (e) participants’ grade, (f) topic of biology, (g) sample size, (h) study results, and, (i) country of publication.

Table 1 Descriptive features of the 32 shortlisted studies for the meta-analysis.
Full size table

Figure 2 illustrates the distribution of included studies categorized by continent of publication, year, and grade. It can be noted in Fig. 2a, that more studies were from Asia (n = 13), followed by the Middle East (n = 7), America (n = 6), Europe (n = 5), and Australia (n = 1). Thus, research studies from almost all continents were included in the study. When analyzing the grades (in Fig. 2b), studies incorporating the grade 10 participants only were the most (n = 16), followed by grade 11 (n = 7). Figure 2c depicts the studies by the year of publication. The studies were extracted from the past 15 years (2008–March 2022), and most of the studies were extracted from the years 2019 and 2021 (n = 5 each).

Fig. 2: Features of shortlisted studies.
figure 2

The bar graphs demonstrate the number of studies included in the meta-analysis by a continent of publication, b by grades, and c by year of publications.

Full size image

Coding procedure

The coding of the studies was conducted by two authors, who have more than 15 years of educational research experience. They read each paper individually and employed the content analysis technique to extract the data, (Hsu et al., 2013). Consecutively, an inter-rater-reliability test has been computed to analyze how often two coders agree with each other. Inter-rater reliability using Cohen’s kappa statistic was employed and this value was found to be 0.93, corresponding to “almost perfect agreement” (Warrens, 2015). Occasional disagreements between the coders were resolved by discussions and consensus. The quality of the shortlisted studies was evaluated by using the Medical Education Research Study Quality Instrument (MERSQI) (Reed et al., 2007). The instrument comprised 10 constructs, encompassing six domains related to study quality, including study design, sampling, type of data, validity of evaluation instrument, data analysis, and outcomes. The potential total score ranged from 5 to 18, with this research attaining a mean score of 15.61 (SD = 0.88). This suggests that the studies incorporated into our meta-analysis demonstrated commendable overall methodological quality (Smith and Learman, 2017).

For the categorization of studies, in terms of pedagogical approach and learning gains, terms and keywords used by the respective authors of the shortlisted studies were considered.

The meta-analysis and interpretation

The meta-analysis was conducted using the CMA software package, specifically version 3.3.070. The Der Simonian and Laird methods were employed to calculate both individual and average effect sizes, along with 95% confidence intervals (DerSimonian and Laird, 2015). This analysis aimed to elucidate the influence of diverse pedagogical approaches on high school students. The raw empirical data were extracted from the finalized studies in the form of pre-/post- or control/treatment means, SDs, t-values, p values, etc., to calculate the effect size (Cohen’s d index) using CMA software. A forest plot diagram revealed the distribution of the effect sizes (at 95 percent confidence intervals) of all the shortlisted studies. As per Cohen’s classification, Cohen’s d values of ≤0.2, approximately 0.5, and ≥0.8 are considered to indicate low, moderate, and high levels of significance, respectively (Cohen, 2013). Furthermore, according to Arnold, any effect size (Cohen’s d) index=0.4 was considered to be educationally impactful (Arnold, 2011). Most of the studies included in the review were large sample-sized, therefore Cohen’s d value was preferred over Hedge’s g value for measuring the effect sizes. A random-effects model was used to compute the mean effect sizes. This model assumes that the variation in the effect sizes of the individual studies is due to sampling error or study design differences. In random-effects analyses, it assumes that each study tends to have a different “true” effect. Thus, the model has been used to account for the heterogeneity between the studies. The impact of each pedagogical model is further investigated by a sub-group meta-analysis and has been explained in the results and discussion section. Before analysis, tests for heterogeneity, test for publication bias, and sensitivity tests were performed.

Test for heterogeneity

The heterogeneity test in meta-analysis determines the variation in study outcomes between studies. To check the heterogeneity of the studies, Cochran’s Q statistic and I2 statistic were used (Cochran, 1954). A significant Q statistic indicates that the effect sizes are derived from diverse populations, indicating heterogeneity. Conversely, a non-significant Q statistic suggests that all studies are presumed to share the same population effect, which implies the application of fixed-effect models.

In our study, a random-effects model was employed because of the assumption that effect sizes vary across studies due to true variance and also due to the fact of reporting the means for the universe of all comparable studies (Borenstein et al., 2009). The difference in effect sizes of the studies could be due to interference of variables (such as sampling error or research design); Therefore, to identify the interfering variable, the heterogeneity analysis was performed using Q and I2 statistics. In our study, the Q statistic is also employed as a test for the null hypothesis. When the Q value deviates from the degrees of freedom (Df), we reject the null hypothesis, which posits that the true effect size is consistent across all studies. In this instance, the Q-value is 401.194 with 31 degrees of freedom. Consequently, we reject the null hypothesis, signifying that true effect size values differ among the studies.

On the other hand, the I2 statistic represents the proportion of observed variance in effect sizes compared to the actual variance. I2 values falling within the ranges of <20%, 20–50%, 50%–75%, and ≥75% are interpreted as indicating low, moderate, high, and very high levels of heterogeneity, respectively. Increased I2 values reveal low dispersion. The study estimates an average I2 value of 92.273%, showing lower dispersion (Cochran, 1954).

Test for publication bias

Publication bias exists when the results of an experiment influence the decision of its publication, which is often not recommended in meta-analysis research. To assess the potential presence of publication bias in this study, several methods were employed. These included a funnel plot (Borenstein et al., 2009), a trim-and-fill model (Duval and Tweedie, 2000), and a classic fail-safe N (Rosenthal, 1979). The funnel plot visually represents the association between effect sizes and their corresponding standard errors (SE). A symmetrical funnel plot suggests an absence of publication bias. This study’s findings reveal an asymmetrical funnel plot (Fig. 3) corresponding to publication bias due to small-size studies. Therefore, other methods (classic fail-safe N, Trim and fill, and Eggers’ linear regression tests) to compute the publication bias were employed. Examination of publication bias employing Egger’s test suggested that publication bias would not impact the study analysis and results (p > 0.05). Also, the difference between the observed and adjusted estimates was found to be 39.1%, which fell into the “moderate” cut-off value (20% to 40%), indicating the publication bias to be negligible. The classic fail-safe N was also computed to examine the number of missing studies to prove the non-significance of the meta-analysis conducted due to publication bias (Rosenthal, 1979). The presence of publication bias is not deemed significant if the number of missing studies, as indicated by the classic fail-safe N, exceeds the tolerance level of 5n + 10. Here, “n” represents the original number of studies extracted without duplication (n = 472). Accordingly, according to the classic fail-safe N method applied to our study, an additional 3554 new results would need to be included to render the overall effect size statistically insignificant. This suggests a considerable degree of robustness in the findings against potential publication bias. This number is greater than the tolerance level value of 2370 [i.e., 5(472) + 10], indicating the publication bias to be negligible and acceptable to run this meta-analysis.

Fig. 3: A funnel plot.
figure 3

The diagram represents the standard error by effect sizes to assess the potential publication bias.

Full size image

Test for sensitivity

In a meta-analysis, sensitivity tests are frequently utilized to assess the resilience of the overall conclusions. Sensitivity analysis aims to identify any effect size that might exert an undue impact on the central tendency (overall effect size) and variability of the data. This often arises from highly influential effect sizes (studies) located at the extremes of the distribution. Hence, the effect sizes of the studies were assessed using the “one study removed procedure” within the CMA software (Borenstein et al., 2009). The findings indicated that, with each study removed, the highest mean in the random model was d = 0.740 (n = 32, SE = 0.089), while the lowest mean was d = 0.689 (n = 32, SE = 0.086). Remarkably, both of these new average effect sizes fell within the confidence interval of the complete dataset, which was [n = 32, d = 0.718, 95% CI {0.547; 0.890}, p < 0.001]. This suggests that no anomalies were observed to exert a substantial influence on the calculated average ESs.

Results

All the finalized studies included in this study were reviewed and confirmed for pedagogies employed in a mixed-ability high school biology classroom. Using the meta-analysis, the effect sizes based on Cohen’s d (standardized difference in means) were estimated, to assess the effectiveness of each type of pedagogical approach, considering the learning gains.

What are the various non-traditional pedagogical models/ approaches employed in high school biology education?

To investigate the pedagogical approach employed in high school biology education, the methodology section of the studies included was thoroughly screened. The studies (n = 32) included in the paper were categorized based on the type of pedagogical model/approach employed. The following categories were identified initially: (a) game-based (n = 6), (b) inquiry-based (n = 5), (c) problem-based (n = 5), (d) project-based (n = 4), (e) virtual laboratory & simulation-based (n = 3), (f) blended model (n = 3). Some combinations were also classified (g) game & inquiry-based (n = 2), (h) argumentation & inquiry-based (n = 2), (i) project & problem-based (n = 1), (j) problem & argument-based (n = 1). Figure 4 represents a pie chart showing the pedagogical approach by studies.

Fig. 4: Types of pedagogical models in high school biology.
figure 4

The pie chart reveals the pedagogical model types in high school biology by the number of studies included.

Full size image

What is the overall impact of the non-traditional pedagogical models employed for mixed-ability high school biology classrooms (when compared to the traditional lecture model)?

The measure of effectiveness (effect size) has been determined by Cohen’s d value. Indeed, Cohen’s d value represents the standardized difference between means. Cohen’s d is commonly used in scenarios where the independent variable is binary, and the dependent variable is continuous. A positive effect size would favor the non-traditional intervention, while a negative effect size would favor the conventional teaching model. Additionally, Cohen’s d values of ≤0.2, 0.2–0.5, 0.5–0.7, and ≥0.8 are classified as indicating low, medium, high, and very high effects, respectively. The larger the Cohen’s d values, the greater the mean difference compared to the variability, indicating greater reliability of the study findings. A random-effect model has been employed (refer to the “Methods” section). The overall effect size of non-traditional pedagogical approaches for mixed-ability high school biology classrooms, when compared to the traditional model has shown a high impact with an effect size value of 0.718. Figure 5 depicts a forest plot that has been used to reveal the distribution of effect sizes [n = 32, d = 0.718, 95% CI {0.547; 0.890}, p < 0.001]. In other words, non-traditional pedagogical models for mixed-ability high school biology classrooms are more highly effective than traditional methods. In the forest plot, the squares on the right represent the effect size of individual studies, while the diamond at the bottom illustrates the overall effect. The lines extending from the squares and diamonds indicate the confidence intervals. The range of effect sizes varied from low (d = 0.012) to high (d = 3.15) effects. Importantly, all the studies demonstrated a positive effect size, indicating a beneficial impact of the interventions. While only 5 among the 32 studies revealed a low effect size i.e., d < 0.2 (Mulder et al., 2016; Klisch et al., 2012; Su et al., 2014; Sivia et al., 2019a; Ebrahim and Naji, 2021). The heterogeneity analysis has estimated a Q value of 297.831 and an I2 index of 89.59%. The distribution of true effects was also determined using CMA prediction interval software. This analysis indicates that the true effect sizes in 95% of all comparable populations are expected to fall within the prediction interval range of −0.21 to 1.69. This provides a range within which the true effects are likely to lie across different populations.

Fig. 5: Forest plot diagram depicting the distribution of effect sizes for the included studies.
figure 5

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image

What is the comparative effectiveness of each pedagogical approach in improving students’ gains (cognitive, affective, behavior) in mixed-ability high school biology classrooms?

To investigate the effectiveness of the various pedagogical models, a subgroup meta-analysis was performed. The studies were classified into six pedagogical categories; (1) project-based; (2) problem-based; (3) inquiry-based; (4) blended model; (5) game-based; (6) virtual simulation-based (Table 2). Studies employing a combination of pedagogical models/ approaches were classified, as per Table 2 for the sub-group meta-analysis. The findings from the table reveal that the problem-based, inquiry-based, game-based, and argumentation-based pedagogies sought to improve students’ cognitive, affective, and behavioral gains. While the studies incorporating the project-based models reported only cognitive and behavioral gains. Similarly, the virtual simulation-based educational models contributed only toward cognitive and affective gains. These findings postulate the need for more research on project-based and virtual simulation-based educational models for high school biology, investigating the students’ affective and behavioral gains respectively.

Table 2 Types of pedagogical models employed in high school biology by studies.
Full size table

Figures 6–12 show the forest plot of the subgroup meta-analysis (comparative analysis). Wherein, the average effect size of project-based learning pedagogies has shown moderate effects on students’ cognitive, and, behavioral gains (n = 5, d = 0.374, 95% CI [0.137; 0.612], p = 0.002, refer Fig. 6). The effect sizes ranged from non-significant small effect size (d = 0.155; Sivia et al., 2019a) to significant large effect size (d = 0.77; Sari et al., 2019).

Fig. 6: Forest plot depicting the distribution of effect size values for studies employing the project-based pedagogical approach.
figure 6

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 7: Forest plot depicting the distribution of effect size values for studies employing the problem-based pedagogical approach.
figure 7

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 8: Forest plot depicting the distribution of effect size values for studies employing the inquiry-based pedagogical approach.
figure 8

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 9: Forest plot depicting the distribution of effect size values for studies employing the blended pedagogical approach.
figure 9

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 10: Forest plot depicting the distribution of effect size values for studies employing the game-based pedagogical approach.
figure 10

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 11: Forest plot depicting the distribution of effect size values for studies employing the virtual simulation-based pedagogical approach.
figure 11

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image
Fig. 12: Forest plot depicting the distribution of effect size values for studies employing the argumentation-based pedagogical approach.
figure 12

The squares depict effect sizes from individual studies, with the diamond representing the overall effect; lines extending from both indicate confidence intervals.

Full size image

The average effect size of problem-based learning pedagogies has also revealed a highly significant impact on students’ cognitive, affective, and, behavioral gains (n = 7, d = 0.913, 95% CI [0.511; 1.315], p < 0.001, refer to Fig. 7). The effect sizes spanned from non-significant small effect size (d = 0.250; (Özalemdar, 2021)) to significant large effect size (d = 2.919; Thurrodliyah et al., 2020). As the I2 value was reported to be greater than 75% (I2-value = 91.3%); thus, it indicated that the large proportion of variability appears to be of true variance.

The random average effect size of the inquiry-based model has revealed a highly significant impact on students’ cognitive, affective, and, behavioral gains (n = 9, d = 0.882, 95% CI [0.536; 1.227], p < 0.001, refer Fig. 8). The effect sizes varied from moderate effect size (d = 0.500; Ristanto et al., 2022) to significant large-sized effect (d = 2.463; Nunaki et al., 2019).

The overall effect size of blended learning approaches has shown high effectiveness on students’ cognitive, affective, and, behavioral gains (n = 3, d = 0.720, 95% CI [−0.178; 1.619], p = 0.116, refer Fig. 9). The effect sizes varied from non-significant small effect size (d = 0.099; (Ebrahim and Naji, 2021)) to significant large effect size (d = 1.542; (Yapici and Akbayin, 2012). Finally, a comparative analysis have been performed and project-based learning is found to be most effective (d = 0.913) in augmenting learning gains in mixed-ability high school biology classrooms (Fig. 13).

Fig. 13: Comparative analysis of effect sizes.
figure 13

The bar graph represents the effect sizes (Cohen’s d) of various pedagogical approaches in high school biology education.

Full size image

The average effect size of game-based strategies has revealed a high impact on improving students’ cognitive, affective, and, behavioral gains (n = 8, d = 0.662, 95% CI [0.335; 0.988], p < 0.001, refer to Fig. 10). The effect sizes ranged from low effect size (d = 0.018; Su et al., 2014) to significantly large effect size (d = 3.157; (Lham and Sriwattanarothai, 2018)). Among the 8 shortlisted studies employing a game-based model, 2 studies showed a non-significant p value (Su et al., 2014; Lokayut and Srisawasdi, 2014). However, the results are reliable as the average p < 0.001 with an average effect size of 0.662.

The random average effect size of virtual simulation-based models in high school biology education has shown moderate impacts in terms of students’ cognitive, and, affective gains (n = 3, d = 0.407, 95% CI [0.089; 0.724], p = 0.012, refer Fig. 11). The effect sizes varied from non-significant small effect size (d = 0.012; Mulder et al., 2016) to significant large effect size (d = 0.688; Thisgaard and Makransky, 2017).

The average effect size of argumentation-based pedagogies in high school biology education has revealed high impacts in terms of students’ gains (n = 3, d = 0.815, 95% CI [0.444; 1.185], p = 0.010, refer to Fig. 12). The effect sizes ranged from moderate effect size (d = 0.500; Ristanto et al., 2022) to significant large effect size (d = 1.090; (Venville and Dawson, 2010)).

Perspective

With the growing demand for personalization and improving students’ learning gains, diverse non-traditional pedagogical practices are developed, tested, and employed by researchers and academicians. We believe that for a satisfactory, and, suitable approach, a dynamic model/ approach must be utilized for varied student populations/communities. Therefore, our first objective of the study was to investigate the diverse pedagogical models/ approaches employed in high school biology education. Some of the non-traditional models employed for biology education have been reported to be virtual laboratory simulation-based, inquiry-based, argumentation-based, problem-based, project-based, game-based, blended models, etc. (Klisch et al., 2013; Nunaki et al., 2019; Ping et al., 2020; Sivia et al., 2019a; Thisgaard and Makransky, 2017; Thurrodliyah et al., 2020; Yapici and Akbayin, 2012). Furthermore, a meta-analysis review of the existing literature has been performed to investigate the impact of the non-traditional pedagogical approaches on mixed-ability high school biology classrooms. The effectiveness was revealed by the standardized difference in means (Cohen’s d). The overall impact of the non-traditional intervention was found to be significantly effective [n = 32, d = 0.809, 95% CI {0.604; 1.015}, p < 0.001] when compared to traditional lectures. A subgroup meta-analysis of various non-traditional pedagogical approaches has also been performed to reveal the pedagogical comparative assessment. Figure 5 reports the overall random average effect sizes of the various non-traditional pedagogies. Furthermore, results from moderator analysis reveal that there is no influence of “biology topics” on the effect size of the studies (due to in-significant QB statistics: between-class variance component statistics).

The problem-based pedagogical model was found to be highly effective in augmenting students’ cognitive, and, behavioral gains [n = 7, d = 0.913, 95% CI {0.511; 1.315}, p = 0.00]. These results are in line with the study findings by Xu et al. (2021), who also reported high effectiveness of the problem-based model with Cohen’s d value ≈ 0.9 for biology education (Xu et al., 2021). It is important to note that no studies in this category reported significant affective gains. This low effectiveness could be due to the limitations of the problem-based model. Which includes devoting greater time and effort to problem-based learning might cause low performance in standardized assessments, as the students might have been involved actively and acquired sufficient knowledge but might not have the depth of knowledge required to score well. Further analysis showcased that students’ affective outcomes are often ignored in literature while reporting/assessing the problem-based models in biology education.

On investigating the inquiry-based learning pedagogies, it has been reported to be highly impactful on students’ cognitive, affective, and, behavioral learning gains (n = 9, d = 0.882, 95% CI [0.536; 1.227], p < 0.001, refer Fig. 8). All the studies employing an inquiry-based model were statistically significant and impactful. These findings are partially in agreement with the study by Funa and Prudente (2021), which reported the impact of inquiry-based pedagogies for biology education to be Cohen’s d value = 1.26 (Funa and Prudente, 2021). The biological science inquiry model is one of the progressing approaches that aid students in processing information using various techniques used by biologists (Furtak et al., 2012). In this area, the students/researchers try to identify different problems and use scientific methodology to solve the problems (Funa and Prudente, 2021). The reported I2 value was greater than 75% (I2-value = 87.2%); indicating a larger proportion of variability to be true variance.

Similarly, the average effect size of argumentation-based pedagogies in high school biology education has been reported to be highly impactful in terms of students’ cognitive, affective, and, behavioral learning gains (n = 3, d = 0.815, 95% CI [0.444; 1.185], p = 0.010, refer Fig. 12). Argumentation in biological science teaching is demonstrated as a process of scientific investigation that involves the justification of claims based on evidence. Thus, by acquiring argumentation skills the student could develop science process skills (Ping et al., 2020; Ristanto et al., 2022). A possible strategy to acquire argumentation skills in the science classroom is by performing laboratory investigations. Wherein, the students will be involved in investigating, tabulating, and analyzing data systematically, thus producing evidence to defend the claims (Ping et al., 2020). Further investigation revealed that students’ affective outcomes are often ignored in literature while evaluating the argumentation-based models in biology education (Ping et al., 2020, Ristanto et al., 2022; Venville and Dawson, 2010).

Followingly, the blended and game-based models have revealed a high effect size of 0.720 and 0.662 respectively. Even though the average effect size of blended learning pedagogies has revealed high effectiveness on students learning (n = 3, d = 0.720, 95% CI [−0.178; 1.619], p = 0.116, refer Fig. 9). There were only 3 studies employing a blended model and the overall p-value is not significant. Therefore, more studies might be included to conclude and justify its impact on students. Moreover, among these three studies, behavioral gains were neglected and not evaluated. Followingly, in the context of game-based pedagogies, findings have illustrated a high impact in terms of students learning (n = 8, d = 0.662, 95% CI [0.335; 0.988], p < 0.001, refer to Fig. 10).

Finally, on investigating the effect of project-based and virtual simulation-based models, their effect sizes are found to be moderately impactful (effect sizes 0.374 and 0.0.407 respectively). Wherein, the average effect size of project-based learning pedagogies has shown moderate effects on students’ cognitive, and, behavioral gains (n = 5, d = 0.374, 95% CI [0.137; 0.612], p = 0.002, refer Fig. 6). These results do not align with the previous meta-analytical research where project-based learning in biology education has reported greater effect sizes [i.e., d = 1.36 (Balemen and Keskin, 2018); d = 0.95 (Ayaz and Söylemez, 2015)]. Despite the project-based model showing its effectiveness, the studies have only reported/assessed cognitive and behavioral gains, giving meager importance to affective gains. Followingly, in the context of virtual simulation-based pedagogies, findings revealed moderate impacts in terms of students’ cognitive learning gains (n = 3, d = 0.407, 95% CI [0.089; 0.724], p = 0.012, refer to Fig. 11). These results are inconsistent with prior meta-analyses reporting a greater effect size of virtual simulation-based models for biology education [d = 0.668; Bayraktar, 2001]. It is also noteworthy that only cognitive gains were reported/assessed in the shortlisted studies, neglecting the affective and behavioral gains.

Furthermore, Table 3 briefs the comparative aspect along with the heterogeneity analysis of the sub-groups. Cohen’s d value determining the effect sizes of the sub-group pedagogical models has revealed very high effects for argumentation-based (d = 0.815), problem-based (d = 0.913), and inquiry-based (d = 0.882). And high effects for blended (d = 0.720) and game-based models (d = 0.662). And moderate impacts for virtual simulation (d = 0.407) and project-based (d = 0.374)] on students‘ gains. However, among all the sub-group analyses, only the p value for the blended model is non-significant (p value = 0.116). The I2 value is highly significant for problem-based, inquiry-based, game-based, and blended models (I2-value = 91.3%, 87.2%, 92.21%, and 88.21% respectively). The higher I2 value has revealed that a large proportion of variability is due to true variance. Likewise, the argumentation-based, virtual simulation-based, and project-based have also shown significant I2-value = 58.23%, 57.83%, and 50.71% respectively.

Table 3 Comparative analysis of the pedagogical subgroups: heterogeneity statistics.
Full size table

In conclusion, it is important to interpret the findings of this study with an awareness of certain limitations. These encompass the absence of a discussion on the feasibility of non-traditional pedagogical approaches, as these practices can often be constrained by factors such as implementation costs, and the time and effort involved. Additionally, it’s worth noting that our study exclusively incorporated peer-reviewed articles, with data from theses, books, and proceedings excluded from consideration. Including such gray literature could have given a holistic view of the available evidence. Another limitation is the inability of study findings to be expressed in terms of different types of learning gains (i.e., creativity, scientific thinking skills, etc.) under a specific pedagogical category due to the lack of studies. Rather authors could infer if cognitive, affective, and behavioral gains are targeted in each pedagogical model.

Conclusion

The main contribution of this study is the analysis of the impact of non-traditional pedagogical models/approaches employed for mixed-ability high school biology classrooms. Considering the diversity in students’ needs, and with the growing demand for personalization and improving students’ learning gains (cognitive, affective, and behavioral), diverse teaching practices are developed, tested, and employed by researchers and academicians. Some of the non-traditional models employed for biology education are virtual laboratory simulation-based, inquiry-based, argumentation-based, problem-based, project-based, game-based, blended model, etc. This meta-analysis review has reported a high impact of the non-traditional pedagogical approach in comparison to the traditional lecture method in mixed-ability high school biology classrooms [n = 32, d = 0.809]. In addition, a subgroup meta-analysis of various non-traditional pedagogical approaches revealing the comparative aspect reported a very high impact of the problem-based model for augmenting students’ learning gains [n = 7, d = 0.913]. Further investigation showcases that it is cognitive gains that are often explored, followed by behavioral, and least by affective gains in mixed-ability high school biology classrooms. Therefore, this study proposes the necessity for future studies evaluating affective gains during project-based, problem-based, and argumentation-based models and behavioral gains during blended models. Thus, these findings propose the necessity for further studies investigating affective and behavioral gains during project-based and virtual simulation-based models respectively.

Acknowledging the limitations discussed in the previous section, our study highlights several promising avenues for future research. As a result, it offers valuable insights for researchers engaged in the study of non-traditional pedagogies. Subsequent studies could delve into the necessity for longitudinal pedagogical interventions, recognizing that certain student variables may require time to manifest and subsequently be assessed. This can contribute to a deeper understanding of the long-term impacts of such teaching methods. Studies on how low-competency students perform differently than high-competency students during/after the pedagogical intervention could be researched. The future scope of the study also includes various moderators’ analyses, such as grades of the participants, duration of the intervention, gender of participants, knowledge type targeted, etc. The moderator analysis will aid in understanding the effect of pedagogy with respect to specific characteristics of the intervention. Thus, finally, we believe that this research would pave the way for academicians to design, customize, and implement novel pedagogies for a dynamic education system in high school biology classrooms (considering the learning requirements of the classrooms).