State-of-the-Art of STEAM Education in Science Classrooms: A Systematic Literature Review

1.1 Existing Review on STEAM Education

Existing reviews have reported both general and specific aspects of STEAM education pedagogy and research without an explicit focus on science classroom settings. This includes the general analysis by Belbase et al. (2022) reporting the problems, processes, priorities, and prospects in STEAM education. Based on their review, Belbase et al. (2022) proposed a new alternative thinking by extending STEAM to the ecological and humanistic well-being, defined as STEAM-h, where “h” refers to humanity for future sustainable environmental balance. The work of Belbase et al. (2022) is an example that illustrates the effort to understand and define STEAM education. Such efforts create different views and varying definitions about STEAM education.

The inconsistencies in defining STEAM terminology, including the matter of defining “science,” “technology,” “engineering,” “arts,” and “mathematics” within the STEAM acronym, cause difficulties for STEAM pedagogy and research (Colucci-Gray et al., 2017; Perignat & Katz-Buonincontro, 2019). For example, very recently, Leavy, Dick, Meletiou‐Mavrotheris, Paparistodemou, and Stylianou (2023) engaged in a systematic literature review identifying the use and prevalence of up-and-coming technologies within the landscape of STEAM education. They found that although various kinds of emerging technologies, e.g., extended realities, maker spaces, and robotics, have been used in the design of STEAM learning, there are discrepancies in the way how they are defined and implemented (Leavy et al., 2023). In Leavy et al.’s (2023) review, the research methodologies and study design characteristics of STEAM education were not properly captured because they were not available in many of the extracted studies. This is probably because Leavy et al. (2023) also included grey literature in their review, which leads to selective publication bias. Perignat and Katz-Buonincontro (2019) reviewed how reported studies define the “arts” within the context of STEAM education. Perignat and Katz-Buonincontro (2019) discovered that, from 44 reviewed articles, not only did researchers fail to be consistent in defining “arts,” but also many of the studies did not support the claim that STEAM education is to enhance problem-solving, creativity, and other twenty-first-century skills. Nearly 55% of the reviewed articles mentioned such learning outcomes in their introductions with lack of explanation. This indicates that many practitioners are struggling to integrate “arts” in STEAM education and measuring the relevant learning outcomes (Herro & Quigley 2017; LaJevic 2013).

From the bibliographic analysis results reported by Marín-Marín, Moreno-Guerrero, Dúo-Terrón, and López-Belmonte (2021), one of their highlights was a trend in relating STEAM education and computational thinking. In response to their findings, Zhang, Ng, and Leung (2023) reported a descriptive review on investigating computational thinking and STEAM education in the context of early childhood education. In the other specific context, Li and Wong (2023) conducted a comprehensive review of STEAM education for personalized learning. They categorized personalization in STEAM education as applied in different modes of education, subject disciplines, and across various countries (Li & Wong, 2023). They found that the trend of research in this particular topic has been increasing over the last 10 years, yet many of the included studies only focused on one STEAM discipline (Li & Wong, 2023). Li and Wong (2023) therefore suggested that more integrative STEAM practices should be implemented in the context of personalized learning. The limitation of Li and Wong (2023) work relies on the use of only one database, namely Scopus. A single database searching may generate a biased representation in a literature review (Kugley et al., 2017).

Despite the growing body of research on STEAM education, there is a notable gap in reviews focusing specifically on its implementation within science classroom settings. General reviews of STEAM education (e.g., Belbase et al., 2022; Perignat & Katz-Buonincontro, 2019) and specific contexts, such as computational thinking (Marín-Marín et al., 2021) and personalized learning (Li & Wong, 2023), have been published. However, a comprehensive review of STEAM integration in science classrooms remains absent. It is crucial to review the implementation of STEAM education in science classrooms not only because science education is fundamental for global educational (Martins & Veiga, 2001), but also for addressing global challenges and fostering sustainable development (Li, Sjöström, Ding, & Eilks, 2022; Won et al., 2021). The attempt to develop effective approaches for science learning based on STEAM education can be enhanced by evaluating its current state-of-the-art. In addition, our search in Web of Science (WOS) and Scopus databases, using keywords “STEAM” AND “Education” within the years 2014–2023, reveals that there are no reported reviews on the STEAM education in the context of the science classroom. Therefore, this article is the first to capture the integration of STEAM learning in science classroom settings. The article provides insights into how STEAM learning enhances science education.

Since resolving real-world problems requires scientifically informed decisions, science education is therefore critical to prepare students to understand and generate creative solutions to real-world problems (Won et al., 2021). In line with that STEAM learning promotes a scientific understanding of real-world phenomena and various possible approaches to solving problems at individual and social levels (Quigley, Herro, Shekell, Cian, & Jacques, 2020a; Wilson et al., 2021). Implementing STEAM learning in science classrooms has become a substantial response toward education for sustainable development (Sachs et al., 2019). However, studies have suggested that teachers had difficulty in implementing STEAM education, particularly in science classes. This is reported that teachers who want to adopt STEAM in their classrooms have difficulties in selecting integrated topics and having a tendency to reduce the science content (Son & Jung, 2019). Therefore, the article holds significant insight into adopting STEAM learning in science classroom settings.

1.2 Research Questions

The research questions that guide this review are as follows:

1. What education levels, subject areas, and research methodologies dominate in science classroom STEAM learning studies?

2. What are the definitions and integration levels of STEAM adopted in science classroom research?

3. What are the pedagogical approaches applied in science classroom STEAM learning?

2.1 Minimizing Bias

A systematic literature review is done because researchers’ concern about the efforts of minimizing bias to obtain reproducible, objective, and thorough search results (Kugley et al., 2017). In this article, three common biases were taken into consideration, i.e., representation, judgment, and publication. First, Kugley et al. (2017) highlighted that the results may not represent all the studies if the search is only done in one database. Rice et al. (2016) recommended considering multi-database searches, moreover when the topic of interest is cross-disciplinary, like STEAM education. To minimize the result bias, WOS and Scopus databases were chosen because they are one of the biggest bibliographic databases. Second, Lasserson, Thomas, and Higgins (2019) addressed that too many judgments may arise during systematic reviews. Therefore, a good review protocol must be developed prior to the document search, wherein PRISMA 2020 protocol is applied to minimize judgment bias. Lastly, publication bias can be a big concern for literature review. Although Hopewell, Loudon, Clarke, Oxman, and Dickersin (2009) discovered that unpublished works were not necessarily of poorer quality than published studies, it is argued that the published articles have generally undergone a review process to ensure their quality. As highlighted by Bakker and Traniello (2019), peer-reviewed publications ensure that the published works have been scientifically validated by experts and therefore can be trusted. Hence, to minimize publication bias and ensure reproducibility, only research papers published in peer-reviewed international journals were included in this systematic literature review.

2.2 Search Strategy

In this systematic review, search terms “STEAM education” AND “science,” “STEAM” AND “science education,” “STEAM education” AND “physics,” “STEAM education” AND “biology,” and “STEAM education” AND “chemistry” were used. The searches were executed in WOS and Scopus databases in the timespan from 1 January 2012 to 31 May 2023. The strings used in the literature review search are provided in Table 1. The initial search from WOS and Scopus resulted in 773 and 465 articles, respectively. However, documents from article reviews, conference papers, conference reviews, books, book chapters, editorial materials, meeting abstracts, short surveys, and notes were automatically excluded from the initial search. Therefore, 289 and 350 documents were respectively recorded from WOS and Scopus for further screening and inclusion stages.

2.3 Selection Criteria

These records were identified and screened based on the specific selection criteria to justify the eligibility for inclusion. In this study, PRISMA 2020 (Page et al., 2021) was applied. The PRISMA 2020 statement is the updated version of the earlier version, PRISMA 2009 (Moher, Liberati, Tetzlaff, Altman, & PRISMA Group, 2009). The PRISMA 2020 consists of 27 checklist items, the expanded version of the previous version, to ensure a transparent and accurate systematic literature review. It also includes a three-phase flow diagram, i.e., identification, screening, and inclusion, as the systematic procedure for review (Figure 1). This systematic process ensures the accountability and consistency of systematic literature analysis documentation. Herein, the search focuses on mapping existing empirical studies on the implementation of STEAM learning in science classrooms context.

First, the documents underwent general selection criteria, as depicted in Table 2. Only English-written original research articles published in the ranges of January 2012 and May 2023 in the scopes of education, social sciences, and humanities were included. Next, more specific criteria for inclusion, as given in Table 3, were applied to the identified records (350 and 289 documents, respectively, from Scopus and WOS).

2.4 Quality Assessment

The specific criteria for inclusion (Table 3) allow us to maintain the good quality of the review. The identification, screening, and inclusion process are shown in Figure 1. This protocol was conducted by two independent reviewers (authors). In cases of discrepancies between the reviewers during this process, a consensus was reached through discussion between the reviewers. The 639 total identified records were carefully checked for specific inclusion criteria. The filtration of duplicate documents resulted in the exclusion of 97 records. Further exclusion of documents that did not meet criteria 1 (document type) and 2 (language) yielded 231 saved articles for further screening. The title, abstract, and keywords screening were then executed to test the sample on criteria 3 (keyword appearance) and 4 (original research article identification). This resulted in the rejection of 119 records that were not relevant to the implementation of STEAM education in classroom settings. During this screening process, one article’s abstract was not found, and another article was retracted. Therefore, 110 articles were further assessed for eligibility, i.e., tested on criteria 5 (focus on science-related subjects) and 6 (published in peer-reviewed international journals). This eligibility test led to the final inclusion of 22 published studies that were further analyzed to answer the research questions.

3.1 Education Levels, Subject Areas, and Research Methodologies that Dominate in Science Classroom STEAM Learning Studies

Table 4 shows a list of the articles included in this study, highlighting the country, subject area, and educational level. Studies on STEAM education in a science classroom context have been conducted in many countries. The international scope of the studies is evident, with contributions from countries such as Greece (Conradty & Bogner, 2019), Spain (Bassachs et al., 2020), Thailand (Khamhaengpol, Sriprom, & Chuamchaitrakool, 2021), China (Jia, Zhou, & Zheng, 2021), Indonesia (Rahmawati, Taylor, Taylor, Ridwan, & Mardiah, 2022), and Saudi Arabia (Alkhabra, Ibrahem, & Alkhabra, 2023). This geographical diversity represents the global significance of effective STEAM education practices in the science classroom context.

Out of the 22 studies, 12 focused on middle school education, such as those conducted by Quigley and Herro (2016) in the USA and Park and Park (2020) in Korea. It emphasizes the importance of tailoring interventions to address the unique challenges and learning needs of students in this crucial stage of academic development. Additionally, investigations at the elementary school level were conducted by researchers like Bassachs et al. (2020), Cheng et al. (2022), and Wilson et al. (2021). Only two studies were conducted at high school levels (Conradty & Bogner, 2019; Khamhaengpol et al., 2021). The distribution of articles based on the school level is provided in Figure 2.

Figure 2

Distribution of articles based on school level.

The 22 articles included in this review spanned various science subjects, ranging from Geoscience to Nanotechnology. Out of the 22 articles, 13 focused on Science in general, 5 emphasized Physics, 3 focused on chemistry, and the remaining articles addressed Geoscience and Nanotechnology. The distribution of articles based on subject area is shown in Figure 3. Furthermore, research methodologies can be classified into three different types, i.e., quantitative research, qualitative research, and mixed methods (Creswell & Creswell, 2017; Snyder, 2019). In this review, 14 studies employed a quantitative approach, 7 studies were based on qualitative data, and the remaining two studies used mixed methods. The distribution of articles based on research methodology is shown in Figure 4. Meanwhile, the details of the data collection techniques are shown in Table 5.

Figure 3

Distribution of articles based on subject area.

Figure 4

Distribution of articles based on research methodology.

From Table 5, the data collection techniques in the reviewed study can be described as follows:

1. Test, which is designed to measure specific variables or constructs in a systematic and standardized manner. It includes pre- and post-tests (Alkhabra et al., 2023; Arpaci et al., 2023; Cheng et al., 2022; Conradty & Bogner, 2019; Hsiao et al., 2022; Hughes et al., 2022; Jia et al., 2021; Ozkan & Umdu Topsakal, 2021b; Tran et al., 2021a,b), worksheets (Khamhaengpol et al., 2021), and questions (Bassachs et al., 2020).

2. Survey, which involves collecting data from a sample of individuals via the administration of a set of standardized questions. It includes pre- and post-survey (Gates, 2017), digital survey (Jiang et al., 2020), and survey open-ended questions (Wilson et al., 2021).

3. Case study, which involves an in-depth, detailed examination of a specific instance or case. The articles that employed the case study were authored by Park and Park (2020).

4. Observation, which involves carefully and systematically watching and recording behaviors, events, or interactions. It can be done using observational rubrics (Quigley et al., 2020a,b; Rahmawati et al., 2022).

5. Interview, which aims to explore participants’ perspectives, experiences, attitudes, and opinions on a particular topic. Interviews were conducted by Jiang et al. (2020), Ozkan and Topsakal (2020), Quigley et al. (2020a), and Rahmawati et al. (2022).

Based on the findings, it is discovered that STEAM education in science classrooms spans across multiple countries, including Greece, Spain, Thailand, China, Indonesia, and Saudi Arabia. It implied that there is a global significance of effective STEAM practices. Furthermore, many of the included studies focused on middle school education, covering diverse science subjects. Research methodologies in STEAM education research varied from quantitative approaches to mixed methods. Mixed methods combined quantitative and qualitative techniques to provide comprehensive findings into STEAM learning practices in science classrooms.

3.2 Definitions and Integration Levels of STEAM Adopted in Science Classroom Research

Table 6 shows how the researchers define STEAM education in their research in the context of science classrooms. The table depicts the STEAM definition and the corresponding level of integration. In terms of integration level, as can be seen from Figure 5, STEAM education can be classified in four levels, i.e., disciplinary, multidisciplinary, interdisciplinary, and transdisciplinary integration (Vasquez, Sneider, & Comer, 2013).

Figure 5

Levels for integrated learning (adapted from Vasquez et al., 2013).

The disciplinary STEAM approach is defined by the separation of concepts and skills associated with each STEAM component. The term “multidisciplinary” refers to the study of multiple disciplines within a common theme. Interdisciplinary STEAM requires a strong connection between two or more STEAM disciplines to develop deeper knowledge and abilities. Transdisciplinary STEAM is considered the highest level of integration. The transdisciplinary approach to problem-solving requires the application of knowledge and abilities from two or more subjects. In transdisciplinary STEAM, students engage in problem-solving activities that demand the synthesis of concepts, methodologies, and techniques from different disciplines (Aguilera & Ortiz-Revilla, 2021). This approach encourages collaboration, communication, and the development of holistic thinking skills. Transdisciplinary STEAM projects often tackle complex, real-world problems that do not fit neatly within the confines of a single subject (Vasquez et al., 2013).

The distribution of articles based on level of integration is given in Figure 6. The analysis of the provided data on STEAM integration levels suggests a predominant emphasis on Interdisciplinary approaches, cited by 11 of the authors. This indicates a shared perspective among scholars that STEAM education thrives on collaborative and cross-disciplinary interactions. Multidisciplinary approaches, mentioned by six authors, underscore the importance of enriching STEM subjects with Arts or combining specific disciplines to achieve learning objectives. The less frequent but notable presence of Transdisciplinary approaches was mentioned by five authors. It suggests a holistic problem-solving orientation, where collective expertise from various disciplines is harnessed to address real-world problems.

Figure 6

Distribution of articles based on level of integration.

From the definitions of STEAM adopted in the literature depicted in Table 6, the characteristics of STEAM integration levels and their relevancies for education are inferred in Table 7. Researchers have delineated distinct perspectives on the definitions and integration levels of STEAM education in the science classroom context. Transdisciplinary STEAM, as expounded by Quigley and Herro (2016), emphasizes a holistic nature within the formal education context in addressing real-world issues. This approach involves the integration of various disciplines to address problems and focuses on the issue itself rather than a specific discipline (Quigley et al., 2020a; Wilson et al., 2021). Thus, transdisciplinary STEAM fosters a comprehensive real-world problem-solving framework. Meanwhile, multidisciplinary STEAM enriches STEM subjects with the inclusion of arts (Conradty & Bogner, 2019; Hughes et al., 2022). This approach aims to facilitate students’ disciplinary identity development (Jiang et al., 2020) and combines arts with STEM subjects (Cheng et al., 2022; Khamhaengpol et al., 2021; Rahmawati et al., 2022). It offers a balanced integration of disciplines that make it suitable for contexts where the Arts incorporation into specific STEM disciplines becomes the primary goal. Although transdisciplinary STEAM is less frequently advocated by the researchers, it promotes a holistic approach to real-world problem-solving through the integration of multiple disciplines. Alternatively, interdisciplinary STEAM emphasizes the integration of science, technology, engineering, arts, and mathematics (STEM), which highlights core competencies and emphasizes on integration of various disciplines (Bassachs et al., 2020; Gates, 2017; Park & Park 2020). This perspective signifies an interdisciplinary education strategy that encourages experiential learning and shifts from traditional teaching paradigms (Arpaci et al., 2023; Hsiao et al., 2022; Jia et al., 2021; López-Banet et al., 2021; Ozkan & Topsakal, 2021; Tran et al., 2021a,b).

3.3 Pedagogical Approaches Applied in Science Classroom STEAM Learning

Table 8 summarizes the learning approaches for STEAM education in the context of science classrooms. Various pedagogical approaches have been applied in science classroom STEAM learning. Project-based learning emerges as the most frequently cited approach, appearing six times in the data. Arts/Technology-based learning follows closely with four instances. Inquiry-based learning is referenced three times. Contextual learning was cited in two articles. Problem-based learning, engineering design-based learning, role-taking-based learning, explicit learning, and reflective learning each appear once in the reviewed articles. The distribution of articles based on the pedagogical approach is shown in Figure 7.

Figure 7

Distribution of articles based on a pedagogical approach.

The STEAM teaching strategies in the science classrooms can be elaborated as follows:

1. Arts/technology-based learning, applied by Alkhabra et al. (2023), Gates (2017), Jia et al. (2021), and Ozkan and Topsakal (2021), recognizes the value of combining creativity, artistic expression, and technological skills to enhance the overall educational experience.

2. Project-based learning is an instructional approach where students actively engage in complex and real-world problems, resulting in a presentation or product for them to acquire knowledge and skills (Hypolite & Rogers, 2023). It was subscribed by Cheng et al. (2022), Hasio et al. (2022), Rahmawati et al. (2022), Tran et al. (2021a,b), and Wilson et al. (2021).

3. Problem-based learning, adopted by adopted by Quigley et al. (2020a), allows students to apply knowledge and skills to generate solutions to a defined problem (Savery, 2015).

4. Reflective learning, adopted by Bassachs et al. (2020), involves the activation of experiential knowledge through ongoing inquiry and self-assessment and ensures adaptability and responsiveness to the dynamic challenges of the environment (Sachs et al., 2019).

5. Explicit learning, adopted by Park and Park (2020), is defined as carefully planned lessons presented to students with the aim of assisting them in gaining a comprehensive understanding of a particular subject (Park, 2008).

6. Role-taking-based learning, followed by Jiang et al. (2020), incorporates activities or methods where students engage in taking on different roles or perspectives to develop science identities (Jiang et al., 2020).

7. Engineering design-based learning, applied by Khamhaengpol et al. (2021), emphasizes active engagement, problem-solving, and the application of engineering principles in a practical context.

8. Inquiry-based learning, adopted by Conradty and Bogner (2019), Hughes et al. (2022), López-Banet et al. (2021), refers to actively involved students by establishing connections to the real world through exploration and use of advanced questioning techniques (NRC, 2013).

9. Contextual learning, adopted by Arpaci et al. (2023) and Ozkan and Topsakal (2020), emphasizes the importance of placing learning within a meaningful context or setting (Hwang, Hariyanti, Chen, & Purba, 2023).

In the exploration of pedagogical approaches within the context of STEAM education in science classroom settings, project-based learning emerges as the predominant method. Many of studies showed its effectiveness in engaging students with real-world problems and fostering creativity. Since the “Arts” distinguishes STEAM from STEM, Arts/technology-based learning also plays a significant role. This learning approach emphasizes the integration of artistic and technological skills to enhance educational experiences. However, no studies investigate students’ creativity when they apply arts/technology-based learning. Other student-centered learning strategies, for example, inquiry-based learning, contextual learning, and role-taking-based learning, have been studied for their roles in implementing STEAM education in science classrooms. Each approach brings unique strengths to STEAM education, collectively offering a diverse range of strategies to enhance science learning outcomes.