The effects of the medium of instruction in certificate-level physics on achievement and motivation to learn
Abstract
A 3-year study was launched in a Hong Kong secondary school to investigate the effects of the medium of instruction (MOI), specifically English and Chinese, on the learning of certificate-level physics. A total of 199 Secondary Four (S4 or tenth-grade) students, divided into three major ability groups, participated in a teaching intervention designed to determine the effects of MOI on their learning achievement and motivation. The results of conceptual assessments and physics examinations revealed Chinese to be a superior MOI in enabling low-ability students to attain a higher level of achievement, whereas English was more suitable for their high-achieving counterparts. However, little conclusive evidence regarding the role of MOI for the medium-ability groups was found. A questionnaire-based survey indicated that students were more motivated to learn physics through Chinese as the MOI (CMI) rather than English (EMI), although significant limitations to its use were identified for the topic of “Heat.” Deficiencies in the vocabulary needed for abstract scientific concepts in Chinese may account for these limitations (for instance, Chinese uses the same word, “re” (
), for both “heat” and “hot”). Finally, follow-up interviews at the end of the study revealed a sharp contrast between the learning prospects of EMI and CMI students. © 2014 Wiley Periodicals, Inc. J Res Sci Teach 51: 1219–1245, 2014.
The English language has enjoyed a superior status to Chinese in Hong Kong (HK) ever since the territory became a colony of Great Britain in 1842. The colonial government adopted English as the colony's official language, which gave a competitive edge to students with a comparatively high degree of English proficiency (Ho, Morris, & Chung, 2005). English proficiency also conferred distinct advantages in securing well-paid jobs in the government, public organizations, and the commercial arena until Chinese was granted equal status in 1974 (Tollefson & Tsui, 2004; Yip & Tsang, 2007). Given the advantages that proficiency in English conferred, it is no surprise that a large majority of secondary schools (S1–S7, seventh to thirteenth-grade levels) in HK chose English as their medium of instruction (MOI) in the past century, particularly for science subjects (Tao, 1996, 2003). However, after the signing of the Sino-British Joint Declaration in 1984, which confirmed the People's Republic of China's resumption of sovereignty over HK in 1997 (Fung & Yip, 2010), concerns began to be expressed over whether Chinese would be more appropriate as the MOI in secondary schools. As a result of HK's remarkable political transition, its MOI policy has been one of the most controversial issues on the education agenda since the 1980s (Lin & Morrison, 2010).
At the beginning of the post-colonial era, and after HK's first Chief Executive presented his blueprint for educational reform, emphasizing the need for school leavers to be bi-literate (i.e., to have mastery of written Chinese and English) and trilingual (i.e., fluent in Cantonese, English, and Putonghua), the education authority implemented a mandatory scheme governing the MOI in all government, grant, and aided secondary schools. Under the new scheme, individual schools no longer had the autonomy to choose the MOI for junior-level students (S1–S3). Instead, they were required to determine their MOI by the quality of their prospective S1 students and teachers' English proficiency (Yip & Tsang, 2007). According to the Medium of Instruction Guidance for Secondary Schools (referred to as the Guidance hereafter), only those secondary schools that could admit no less than 85% of S1 students who belonged to the top 50% of the population in terms of language proficiency were allowed to opt for English (EMI) in the junior years.1 All other schools were constrained to the use of Chinese (CMI) alone. In its entirety, the Guidance not only rigidly assigned the MOI of individual schools, but also banned the mixed mode of instruction. In the government's view, junior students would learn non-language subjects better through adherence to a single-language environment (Education Commission, 1995). As a result of the new language policy, only around 100 secondary schools in HK retained EMI at the junior level, with approximately 300 required to switch from EMI to CMI starting from the 1998–99 academic year.2 Although the effectiveness of CMI had not been sufficiently proven, and many secondary school teachers were somewhat dubious about it (So, 2000), the government's visible manifestation of commitment to the mother tongue was applauded by a considerable number of education stakeholders, including the Association of Chinese Middle Schools and Hong Kong Federation of Education Workers.
At the same time, however, the Guidance imposed no restrictions on schools' choice of MOI at the senior secondary level. Therefore, most schools that were required by the government to use CMI for their junior-year classes continued to employ EMI to teach students in the senior grades so as to remain popular choices with parents, who continued to favor English. Indeed, a labeling effect was inevitably created by the new language policy, whereby elite students were eventually streamed into schools that were permitted to use EMI in the junior years. This practice arguably jeopardized the MOI policy itself, as its initial aim was to promote the mother tongue as an effective teaching medium in the expectation that students so taught would eventually achieve better cognitive and academic development in practical subjects. Regrettably, however, the labeling effect resulted in CMI students being perceived as inferior and discriminated against. Shifting some classes of junior students from CMI to EMI when they reached the higher forms (i.e., S4–S7) thus became common practice in most HK secondary schools (Ho, 2008).
Literature Review
The Role of Language in Science Education
There has been growing interest in the role played by language in science learning in recent years (i.e., Airey & Linder, 2006; Haagen-Schuetzenhoefer, 2010; Liebscher & Dailey-O'Caine, 2005). On the one hand, language is seen as a crucial part of scientific literacy (Yore, Bisanzb, & Hand, 2003), which can be further distinguished by its fundamental and derived sense. Norris and Phillips (2003) believe that emphasis on the fundamental sense of scientific literacy, that is, teaching students to read and write scientific texts, is important for acquisition of its derived sense, that is, the state of “being knowledgeable, learned, and educated in science” (p. 56). To achieve scientific literacy, scholars such as Huang (2006) believe that students must be proficient in both the language and concept of science inquiry. On the other hand, language is also seen as a symbolic representation of culture in the science context (Brown & Ryoo, 2008). Scientific terms written in a local language sometimes imply meanings that are difficult for students to understand. For instance, the Chinese characters for “melt” (
), which is related to heat or fire, and “dissolve” (
), which involves the use of water or a solvent, look quite similar, and are thus easily confused even by native Chinese speakers. Hence, to counter the effects of local languages on science learning, it has been proposed that student diversity (i.e., students' home languages) be taken into account in science education (e.g., Lee & Luykx, 2006; Luykx & Lee, 2007).
In addition, a review of the literature (e.g., Gee, 2008) indicates that the academic language of science usually differs from the everyday language used by students. For example, certain scientific terms, such as energy, temperature, heat, power, and speed, have precise implications in science, but often have very different meanings in daily usage. Learning science has thus been likened to learning a new language, with Wellington and Osborne (2001) noting that “[s]cience education involves dealing with familiar words … and giving them new meanings in new contexts” (p. 5). The inconsistency and distinction between daily language and academic language may even constitute an obstacle to the acquisition of scientific knowledge, and hence to construction of the derived sense of scientific literacy. Language is “living” in terms of its meanings and uses over time (Roth, 2014). It may be understood by one group of speakers (e.g., scientifically literate students) but not by others. It is no wonder that Wellington and Osborne (2001) consider language to be a major barrier to science learning for most students, even native speakers. In a similar vein, Maskill (1988) claimed that “abstract ideas have to be correctly imagined in order to gain meaning. Classroom language is what guides the imagination” (p. 486).
English as a Second Language in Science Education
A review of research undertaken in Kenya, South Africa, Japan, and Israel, where students use a second language—that is, English—to learn science (see Cleghorn, Merritt, & Abagi, 1989; Kawasaki, 2002; Rollnick, 2000; Strevens, 1971), indicates the importance accorded to language in science education. Notably, there is strong consensus that English is indispensable for scientific development and communication in the international context (Rollnick, 2000). Research in the past few years (i.e., Kawasaki, 2002; Lee, 2005) reveals that many languages, including Japanese and Chinese, lack the vocabulary required for the precise identification of certain abstract scientific terms and concepts. Such deficiencies no doubt account for the increasing dominance of English as the world's leading MOI in science education, although it is still acknowledged that learning science through the medium of a second language constitutes a formidable task, one that involves the simultaneous mastery of science content and a new language (Isa & Maskill, 1982; Strevens, 1980). Okebukola, Owolabi, and Okebukola (2013) also recently found that parental preferences play a major role in schools' choice of English as the MOI for science learning, with those preferring English strongly resistant to a change to home-language instruction. Those preferences conflict with the findings of a recent review of language policies in African countries, which called into question the prevailing belief that using English to study science promotes a better understanding of the subject, thus urging examination of the roots of such belief (Brock-Utne, 2012).
Although a lack of English-language proficiency is a negative predictor of non-native English learners' performance in science assessments (e.g., Maerten-Rivera, Myers, Lee, & Penfield, 2010), there is a considerable body of evidence to support the practice of home language instruction in secondary schools (Cummins, 2000). For example, Heugh (1999) and Setati, Adler, Reed, and Bapoo (2002) showed that when black secondary-school students in South Africa were forced to learn through the medium of Afrikaans in half their subjects and English in the other half (neither of which was their native language), pass rates dropped significantly (cited in Rollnick, 2000). South Africa is not an isolated example. Bunyi (1999) analyzed Kenyan data and found that when science is taught through the medium of English (the second language in Kenya), students find it more difficult to associate science learning with their daily lives. Moreover, Cleghorn et al. (1989) discovered that restricting native language use during instruction in Kenyan primary schools ultimately hampers student understanding of important concepts. Although the positive effects of native language use in science learning have not been conclusively proved—for instance, a study on the performance of Malaysian eighth graders in the Trends in International Mathematics and Science Study (TIMSS) in 1999, 2003, and 2007 indicated no significant link between that performance and the language used at home (Mohammadpour, 2012)—mainstream theories (e.g., Lee, 2005; Maerten-Rivera et al., 2010; Okebukola et al., 2013) on the effects of MOI still favor the use of the home language in science teaching at the lower grade levels.
The Effects of MOI on Learning Achievement and Motivation in HK
Research on the effects of MOI on academic achievement and the motivation to study science in HK dates back to the late 1970s. In a study of 19 secondary schools (9,095 students), Siu (1979) revealed that medium- and low-ability students (in terms of academic performance) achieved the poorest academic results in science when EMI was adopted. Further analysis confirmed that such underperformance was the result of these students' lack of thinking and communication skills in English. A few studies conducted in the 1980s reported similar findings. For instance, the Education Commission (1986), Ip and Chan (1985), and Lo, Chan, and Ip (1985) all argued that CMI helps to boost S1–S3 students' learning motivation and achievement, particularly those of low achievers. They also found EMI to have a negative effect on the learning of the latter group. Marsh, Hau, and Kong (2000) followed approximately 12,000 secondary-school students for three successive years, from S1 to S3, to determine the effect of MOI on the study of content subjects. Their results favored the use of the native language (CMI in this case) to study science.
A more recent study (Yip, Tsang, & Cheung, 2003) involving 100 secondary schools reported EMI junior science students to perform much worse than their CMI counterparts, despite superior initial academic performance and abilities. Their analysis showed the EMI students to struggle with questions related to an understanding of abstract concepts, discrimination between scientific terms, and the application of scientific knowledge in novel situations. In an in-depth follow-up study, Yip, Coyle, and Tsang (2007) reported EMI students to display a lower self-concept in science than their CMI peers. However, in contrast to previous studies, the EMI students in their later research were more academically oriented and demonstrated a deeper interest in science.
Although there is general consensus in the HK literature that the use of students' native language is beneficial to science learning, particularly for underachievers, most studies have arguably focused too heavily on junior secondary students. Although we agree that an achievement gap and poor motivation in the early years can impede science learning at a later stage, a few questions remain unanswered, as very few studies have tackled the relationship between MOI and science learning at the senior secondary level. What happens when the language of instruction changes from students' native language to English as they progress from junior secondary (S3) to senior secondary (S4)? Do these students achieve poorer academic results and display less motivation to learn science than those who continue learning in their native tongue? Or, alternatively, do they become more interested in learning science and thus achieve better results? And, finally, is students' learning in S4 driven by the prospect of learning through a specific language? The aforementioned MOI shift at the senior secondary level in HK provides an ideal setting in which to answer these questions. The study reported herein set out to do so using both quantitative and qualitative research methods.
Methodology
Research Design
The study, which adopted a quasi-experimental research design (Robson, 2002), was conducted in a secondary school in HK that was among the approximately 300 required by the government to use Chinese as the MOI in the junior years. Hence, all of the students had been instructed in Chinese prior to research commencement. However, as they progressed from the junior to senior secondary years, some of them chose to begin learning physics through English (the EMI class), whereas others continued to learn through Chinese (the CMI class). The research involved a tenth-grade physics teaching intervention implemented in three cohorts of students. The intervention ran as a cycle designed to inquire into students' first year of study in certificate-level physics. The topics of “Heat” and “Mechanics” were chosen and taught according to the syllabus provided by the Hong Kong Examination and Assessment Authority (HKEAA). The teaching intervention lessons employed two instructional languages, Chinese and English. In line with the MOI, the textbook and assessments were in English in the EMI physics class, whereas all materials were in Chinese in the CMI class, with teachers and students alike required to use the specified language to the greatest extent possible. For instance, the students were regularly encouraged to interact and participate in the learning activities (such as take-home assignments and project work) using Chinese or English alone (for further details, refer to The Teaching Intervention section below).
Participating School and Students
The participating school is a typical CMI school that receives public funding and belongs to HK's government-aided secondary school sector. Similar to many other CMI schools, most of its students are classified as Band 2, indicating a medium level of academic ability (Yip et al., 2003). Moreover, as is the case with most CMI schools, the vast majority of the participating school's students come from working-class or poor families. About one-fourth of the participating ninth-graders' families were Comprehensive Social Security Assistance recipients, and approximately 60% of them had at least half their school fees remitted by the government (see Table 1).
| Total Sample Size | CMI Group (N = 92) | EMI Group (N = 107) |
|---|---|---|
| Sex | ||
| Male | 59.8% | 60.7% |
| Female | 40.2% | 39.3% |
| Means age (SD) | 16.35 (0.62) | 16.30 (0.60) |
| Ethnicity | ||
| Chinese (Hong Kong) | 94.6% | 90.7% |
| Chinese (Mainland China) | 4.3% | 8.4% |
| Other | 1.1% | 0.9% |
| Home language | ||
| Chinese | 100% | 100% |
| Family education level (highest level attained by either father or mother) | ||
| Primary school | 2.2% | 2.8% |
| Junior secondary school | 48.9% | 45.8% |
| Senior secondary school | 38.0% | 40.2% |
| Tertiary education | 9.8% | 9.3% |
| Missing data | 1.1% | 1.9% |
| Socioeconomic status | ||
| Comprehensive Social Security Assistance recipient | 26.0% | 25.2% |
Sixty-five, 66, and 68 students participated in the study in three separate academic years. Boys (N = 119) constituted a small majority (approximately 60%) of the total of 199 participating students. In addition to a low socioeconomic status, the demographic data of the participating school reveal another common feature of CMI schools in HK: the home language of the students is Chinese, and their parents have a relatively low level of education (compared to the parents of EMI school attendees) and are thus unlikely to provide sufficient support, guidance, or counseling to facilitate their children's academic development (Catsambis, 1995). As previously noted, when the MOI policy was implemented in 1998, the participating school was required by the government to use Chinese in almost all subjects in the junior years. Accordingly, in the years since, its younger students' exposure to English has been almost exclusively limited to English lessons. Like many other typical CMI schools, however, it offers various out-of-class activities, including an English festival, musical dramas, English corners, and supplementary classes during the holidays, to boost students' English proficiency. Moreover, every student attends a 3-week English bridging course in the summer they progress from ninth to tenth grade to strengthen their English reading and writing skills. The school administrators hope that such holistic support strategies will enable students to catch up to the English standards of their peers in EMI schools.
The Teaching Intervention
Both the experimental (CMI class) and control group (EMI class) students participated in the teaching intervention, which was conducted during physics lessons scheduled in the school's normal timetable for an entire academic year. There were five 40-minute teaching sessions per week. Although a combination of didactic and inquiry-based teaching approaches was adopted, the distribution of teaching time was about 20 minutes of lecturing, 10 minutes of activities and discussion, a 5-minute assessment, and a 5-minute conclusion. In addition to the instructional, textbook, and assessment language, there were two other major differences between the CMI and EMI classes. First, because the EMI students had just began to learn physics through English, the course instructor provided them with vocabulary lists containing key terms and phrases extracted from the textbook. These lists also offered brief explanations of certain signaling words (e.g., assume, predict, and speculate) and connectives (e.g., as a result of, due to, and consequently) frequently used in science texts to facilitate students' reading and understanding of the textbook content. The CMI students also received vocabulary lists, but these included more detailed descriptions of the formal application of scientific terms in the Chinese language. For example, the words “force” and “power” are sometimes used interchangeably and informally carry the same meaning in everyday usage. Hence, the vocabulary lists for the CMI students emphasized the clarification and illustration of their precise and formal application in the science context.
Second, in the EMI class, the instructor read aloud-selected keywords and sentences from the textbook with the students, and highlighted the main points of various chapters for their reference. The aim was to improve students' pronunciation of the English words and further facilitate their comprehension of the text. In the CMI class, in contrast, the same instructor spent more time explaining the abbreviated symbols of measurement used in physics (e.g., “J,” “N,” “s,” and “Kg”) to the CMI students. This was necessary because although Chinese textbooks use these English symbols to represent various units, for example, “J” for the unit of energy, they provide little information on their origins relative to English textbooks (i.e., “J” is derived from the name of physicist James Joule [1818–1889]).
With the exception of these differences, both the CMI and EMI students engaged in a series of learning activities, including project work, group discussions, and presentations, in addition to receiving traditional physics tuition. The learning activities, which accounted for about a quarter of the total lesson time, offered students ample opportunities to reflect upon their conceptual understanding of physics. To practice the concepts they learned on the course, they also entered an inter-class competition in small groups of five to six. Students were required to complete the project work by applying and elucidating principles related to the topics of “Heat” and “Mechanics” in certificate-level physics. Furthermore, during the first month of the research, both the CMI and EMI students took part in training sessions on the use of the Internet for data collection and project planning, and extra laboratory lessons and self-reflection sessions were also organized to help them to prepare for the inter-class competition. The training sessions started and finished with the administration of pre- and post-conceptual assessments. In addition, a questionnaire-based survey and follow-up interviews were administered after the teaching intervention. Finally, at the end of the physics course, all students took the same 4-hour examination containing multiple-choice questions and structured problems in accordance with their MOI.
Assignment of MOI Classes and Ability Groups
In this research, the allocation of the students to the CMI and EMI classes was based on their own choice rather than random assignment by the school or researchers. The school briefed students on the advantages and disadvantages of each at the end of their final junior year, and gave them two months to consider which class would best suit their physics learning needs. This arrangement is common practice in most HK secondary schools, with students asked to select their preferred MOI and elective subjects for the certificate-level examination at the tenth- and eleventh-grade levels at the end of the ninth grade. It allows students to learn their chosen subjects through the medium of their preferred language. To avoid any deviation from school practice, no attempt was made in this research to impose the rigid random allocation of students to the two MOI classes, which would have created a purely experimental, but ecologically invalid, research setting in the HK educational context.
Prior to research commencement, the principal researcher gathered and collated students' results on the ninth-grade physics examination. In each MOI group within a year cohort, the students were then ranked according to prior achievement based on their exam results, and an ordered list (on a yearly basis) was constructed to classify them into different physics ability groups. In alignment with the Secondary School Places Allocation (SSPA) mechanism used by the Hong Kong Education Bureau (Au & Watkins, 1997), which ranks and evenly distributes students into Bands 1, 2, and 3 (from highest to lowest ability), the participating students were evenly divided into high-, medium-, and low-ability groups (in terms of relative ability within each MOI class) based on the ordered list. Although data from all three-study years were pooled for analysis, the students were ranked into groups using this procedure on a yearly basis rather than re-grouping them on the basis of a new ordered list ranking them across the 3-year research period. We believed that this method would better reflect students' physics ability in the HK educational context, in which the classification of student ability (using the SSPA exercise) takes place annually.
Research Hypotheses and Factorial Design
Drawing on the empirical evidence outlined in the literature review, this study set out to test two research hypotheses. Hypothesis 1 is “Students display the same level of achievement in learning certificate-level physics regardless of whether Chinese or English is the MOI.”3 Accordingly, the alternative hypothesis is “Students who study using Chinese and English as the MOI display different levels of achievement in learning certificate-level physics.” Hypothesis 2 is “Students have the same level of motivation to learn certificate-level physics regardless of whether Chinese or English is the MOI.” The alternative hypothesis is “Students who study using Chinese and English have different levels of motivation to learn certificate-level physics.” The two research hypotheses were subjected to quantitative analysis in the current study.
With regard to the study's analytical framework, in addition to the factors of MOI and ability group, students' demographic characteristics, such as sex, ethnicity, and home language, could also be considered to potentially influence student learning in science (Maerten-Rivera et al., 2010). However, chi-square non-parametric tests revealed no significant differences between the CMI and EMI students in the distributions of sex (χ2(1,199) = .019, p > .05), ethnicity (χ2(2,199) = 1.34, p > .05), family education level (χ2(4,199) = .45, p > .05), or socioeconomic status (χ2(1,199) = .019, p > .05), indicating that these factors had been statistically controlled prior to study commencement. In addition, because both the CMI and EMI classes were taught by the same course instructor and held in an identical, and authentic, classroom setting, that is, the typical HK secondary school seating arrangement of desks and chairs arranged in neat rows facing the chalkboard, the effects of possible differences between the classes in terms of instructional strategies and learning environment were assumed to be adequately minimized. The research thus constituted a 2 × 3 factorial design. The independent variables were MOI (EMI and CMI) and Ability Group (low-, medium-, and high-ability), and the corresponding dependent variables were Achievement and Motivation.
Data Collection
Data were collected from three main sources—(a) the ninth- and tenth-grade physics examinations, (b) conceptual assessments, and (c) a questionnaire-based survey and follow-up interviews. Each source is now described in turn (also see Supporting Information Appendix I).
Ninth- and Tenth-Grade Physics Examinations
As noted, prior to research commencement, the student participants had taken a 3-hour ninth-grade physics examination comprising 30 multiple-choice questions and seven structured problems. As they had all studied physics through Chinese in their junior secondary years (prior to their division into the CMI and EMI groups at the tenth-grade level), this examination was administered in Chinese only. The exam covers the physics topics for ninth-graders in the junior secondary science curriculum (Curriculum Development Council, 1998), as stipulated by the Education Bureau, including the reflection of light, electromagnetic spectrum, color dispersion, total internal reflection, and very basic concepts concerning forces in the context of space travel. The exam's multiple-choice questions are designed to assess students' ability to recall physics principles and apply them to real-life situations, for example, to determine the angle of reflection at a plane mirror by a given incident light angle, whereas the structured questions require them to demonstrate their knowledge of physics by explaining a number of everyday phenomena and applying certain principles to novel scientific problems (e.g., to explain the optical phenomenon of a mirage). Finally, students are asked to read a short Chinese newspaper article on a recent development in physics and answer questions on how it relates to science, technology, and society.
At the end of each research year, all of the CMI and EMI students took the tenth-grade physics examination in which they were administered an identical set of questions in Chinese and English, respectively. The format of this 4-hour examination is similar to that of the ninth-grade exam, and is aligned with the HKEAA's public physics examination (HKEAA, 2002). It comprises two papers. In the first, students have 90 minutes to answer 45 multiple-choice questions assessing their conceptual understanding of the topics of “Heat” and “Mechanics.” These questions also assess students' understanding of units of measurement, terminology, definitions, laws, models, and investigative methods in both theoretical and experimental physics. In the second paper, they have 150 minutes to answer 10 structured questions containing two newspaper items related to science, environment, and health issues (i.e., the CMI and EMI students responded to the Chinese and English versions of the newspaper items, respectively). They evaluate students' content knowledge of physics and their scientific awareness. The second paper also requires students to answer several tailor-made questions related to the principles and applications demonstrated in their project work. Both papers evaluate to the greatest extent possible the competencies that students acquired in their tenth-grade physics study.
Pre- and Post-Conceptual Assessments
The CMI and EMI students were also administered identical (apart from language) pre- and post-tests on “Heat” and “Mechanics.” These tests were adapted from two instruments, the Thermal Concept Evaluation (TCE; Yeo & Zadnik, 2001) and Force Concept Inventory (FCI; Hestenes, Wells, & Swackhamer, 1992), as normed assessments of the concepts covered in the two units. Examples of the questions in each are provided in Supporting Information Appendix II. The TCE comprises 26 multiple-choice questions on the concepts of temperature, heat, and heat transfer, and the FCI 30 multiple-choice items associated with the theories of Newton's laws of motion, kinematics, and action/ reaction force. The validity of the original (English-language) version of the TCE was assessed in a study of more than 450 Western Australian high-school students in nine institutions who undertook formal instruction in thermal physics (Yeo & Zadnik, 2001). The test reliability was 0.81 (using a split half-correlation with Spearman–Brown correction), and the difference between the means of the pre- and post-test scores was statistically significant (p < .05). Similarly, in a study of approximately 1,000 college students in the U.S., the FCI attained internal consistency coefficients (Croker & Algina, 1986) of α = .86 (pre-test) and 0.89 (post-test), and the discrimination index of 85% of its questions reached 0.80 (Savinainen & Scott, 2002). We argue that this reliability and validity evidence holds in our research for two reasons. First, we investigated a similar age group of students (aged 15–17) to those in the Australian and U.S. studies, and all three-student groups (in HK, Australia, and the U.S.) studied a comparable entry-level physics course (i.e., tenth-grade introductory thermodynamics). Second, the students in our study answered an identical set of questions in the same amount of time as their Australian and U.S. counterparts, and the length of time between administration of the pre- and post-tests was comparable.
The Chinese versions of the TCE and FCI were translated by a physicist who is fluent in both Chinese and English, and then cross-checked by a native Chinese-speaking translator. The principal researcher made minor revisions to the wording to improve the concordance between the original and translated versions, and the translated conceptual assessments were then piloted in a small group of students (N ∼ 25) to remove any potential language issues. The TCE had an internal consistency coefficient of α = .80 when the mean item difficulty was found to be 0.61 and the discriminating power was equal to 0.52. Similarly, the levels of internal consistency, average item difficulty and discrimination for the FCI were 0.82, 0.59, and 0.56, respectively.
Questionnaire-Based Survey and Follow-Up Interviews
A questionnaire-based survey was carried out after the teaching intervention. Adapted from Singh, Granville, and Dika (2002), this self-report survey contained eight questions assessing students' motivation to learn physics and learning attitudes. Students were asked to rate themselves on scales ranging from “never to daily” and “strongly disagree to strongly agree,” which permitted us to ascertain whether the CMI and EMI students had different levels of motivation to learn physics. The Cronbach's alpha for the survey was 0.83, indicating an acceptable level of internal consistency. The original (English) version was translated into Chinese and then back-translated to ensure that the Chinese version captured the original intent. Its content validity was then assessed and safeguarded by the research team, and any ambiguity and/or discrepancies between the Chinese and English versions were corrected and eliminated.
The final, qualitative, stage of data collection involved follow-up interviews with six focus groups comprising three students in each academic year (N = 6 groups × 3 students × 3 years). Each group represented one ability group from each MOI (low-, medium-, and high-ability × CMI and EMI). The interview participants were randomly sampled by the principal researcher, and all agreed to participate and express their opinions. The 54 interviewees constituted about a quarter of the total student sample (199) across the three study years. The first part of the interview consisted of a formalized set of questions focusing on how the MOI had affected physics learning with respect to classroom atmosphere, learning motivation, school assessment, and project work completion. Sample questions included “Were you motivated to study physics by using EMI (CMI) in class?” and “Did you find it difficult to comprehend the physics concepts taught in the lessons?” The second part comprised specific follow-up questions probing the interviewees' comments and opinions in greater depth. A few questions were asked in all of the interviews to elicit students' opinions on the learning materials, public examinations, and further studies in science. To maintain uniformity and ensure that the same topics were addressed in all focus groups, an interview protocol was developed, although its questions were under constant review for clarity and relevance to the study's key concerns. The interview protocol is given in Supporting Information Appendix III. The interviews lasted approximately 15 minutes and were audiotaped. The resulting 4 hours of interview data were subjected to qualitative analysis, providing evidence of the impact of MOI on students' daily lessons, experiments, project work, assignments, and academic aspirations.
Data Analysis
The tenth-grade physics examination was marked by the course instructor. Inter-rater reliability was calculated by randomly extracting 40 responses from the structured question component of the examination and comparing the scores assigned by the instructor with those of a second rater. The level of agreement between markers was above 80%. Similarly, the internal consistency of the exam's multiple-choice questions was also deemed satisfactory, with a Cronbach's alpha (α) of 0.83. The physics examination results were also subjected to analysis of covariance (ANCOVA), which tests statistically whether certain factors are related to the outcome variable after controlling the variance of the covariates (Lindquist, 1940; Pedhazur, 1997). ANCOVA was deemed appropriate in this study, as its aim was to investigate how MOI and Ability Group (i.e., factors) were associated with the Tenth-Grade Examination Results (i.e., outcome variable) after statistically controlling the Ninth-Grade Examination Results (i.e., covariate). Moreover, because the pre- and post-conceptual assessments were identical, repeated-measures analysis of variance (ANOVA) was adopted for statistical analysis of the FCI and TCE results. More specifically, it was performed to determine the relative improvement exhibited by the CMI and EMI groups following the teaching intervention.
There were several reasons for including Ability Group as an intermediating factor and Ninth-Grade Physics Examination Results as a covariate in ANCOVA and repeated-measures ANOVA. First, empirical findings in HK (e.g., Ip & Chan, 1985; Lo et al., 1985; Marsh et al., 2000; Siu, 1979) show that students of different ability groups (in terms of relative ability in the SSPA context) attain different levels of achievement and motivation in both MOI groups and that student ability is a crucial factor (correlated with MOI) in the academic performance of students in junior science. Therefore, the significance of Ability Group to student learning, particularly of a senior secondary science subject such as physics, was deemed worthy of investigation in the current research. Second, both Ability Group and Ninth-Grade Examination Results are useful variables in increasing the effect size4 of the analytical tests. Although they are correlated with the dependent variable (i.e., Tenth-Grade Physics Examination Results), they help to strengthen the statistical power of the tests. Finally, Ability Group and Ninth-Grade Examination Results constitute significant categorical and continuous variables, respectively.5 The latter helps to account for the variances in the tenth-grade results, as they reflect students' absolute physics ability, whereas the former factor focuses on the relative ability of students within the same MOI group. In sum, the two independent variables played different but important roles in the statistical tests.
Moreover, the administration of the ninth-grade physics examination in Chinese, which was used to rank students into ability groups, can be explained by the design of the study. The Ability Group variable was created to represent the students' physics ability prior to research commencement. To accurately capture physics competencies in isolation from language difficulties (or cultural barriers), it is ideal that students take the examination in the language in which they are most proficient. Furthermore, the aim of analyzing the questionnaire-based survey results was to reveal any significant differences between the CMI and EMI students' responses. Accordingly, we applied chi-square non-parametric tests to compare the two sets of responses, and drew conclusions from these categorical data. Finally, the audiotapes of the follow-up interviews were reviewed several times before transcription to “digest” the dialogue content. The process of qualitative analysis involved transcribing the tapes verbatim, entering the transcripts into NVivo, and then comparing the content from the various groups. We adopted the respondent validation method (Denzin, 1970), meaning that we crosschecked all of our notes with the interviewees. The interview data served to triangulate the data collected in the questionnaire-based survey.
Results
Tenth-Grade Examination Results
As illustrated above, to investigate the effects of MOI and student ability on the results of the tenth-grade examination, we performed two-way factorial ANCOVA with MOI (CMI and EMI) and Ability Group (low-, medium-, and high-ability) as the factors, Ninth-Grade Examination Results as the covariate, and Tenth-Grade Examination Results as the dependent variable. The ANCOVA assumptions were first checked to ensure that they were met. For instance, prior to the ANCOVA test, Levene's test for equality of variances was performed. It was found to be not significant, F(5, 193) = 1.94, p > .05, indicating no violation of the homogeneity of variance assumption.
The results of factorial ANCOVA revealed significant main effects for both MOI (F(1, 192) = 4.51, p < .05) and Ability Group (F(2, 192) = 6.42, p < .005), as well as a significant interaction effect between the two (F(2, 192) = 78.42, p < .001). To explore the nature of this interaction, post-hoc tests were conducted to evaluate the pair-wise differences among the adjusted means (or estimated marginal means) for the tenth-grade examination results. More specifically, Bonferroni-corrected post-hoc follow-up tests6 were carried out to control for Type I error across the pair-wise comparisons. The results showed that, although the differences of the pairs “Low (CMI vs. EMI)” (p < .001) and “High (CMI vs. EMI)” (p < .005; effect size estimated by Cohen's d measure = 1.7) were significant, that of “Medium (CMI vs. EMI)” (p > .05) was not. Results of the adjusted mean comparisons among the low-, medium-, and high-ability groups of CMI and EMI students are presented in Figure 1.
In summary, ANCOVA tests on tenth-grade physics performance showed the low-ability CMI students to have a statistically higher adjusted mean than their low-ability EMI peers in their first year of certificate-level physics study. The implication is that students with poorer mastery of physics can benefit from the use of their native language for classroom instruction and interaction. However, no such language advantage was found for their counterparts with superior mastery of the subject, and the medium-ability group displayed a mixed and inconclusive pattern with regard to classroom language. Accordingly, the alternative Hypothesis 1 is statistically supported by the results of the tenth-grade examination.
Results of Conceptual Assessments
We next performed repeated-measures ANOVA on the conceptual assessment (FCI and TCE) results. MOI and Ability Group were entered as the between-subjects factors and Time as a within-subjects factor for the scores of these assessments (the dependent variables). The assumptions of repeated-measures ANOVA were first checked to ensure that they were met. In the following sections, we present the results of the conceptual assessments individually.
Force Concept Inventory Results
Levene's test on the FCI pre-test results showed the variance in the scores of the low-, medium-, and high-ability groups to be homogeneous, F(pre-test) (5,193) = 1.23, ns, and showed a similar degree of variance in the groups' post-test scores: F(post-test) (5,193) = 1.87, ns. Both sets of results satisfied the condition of variance homogeneity. Repeated-measures ANOVA on the FCI results revealed significant main effects for Ability Group (F(2, 193) = 146.10, p < .001) and Time (F(1, 193) = 287.80, p < .001), but an insignificant main effect for MOI (F(1, 193) = 2.33, p > .05). At the same time, a significant interaction effect was found: F(2, 193) = 8.95, p < .001. Post-hoc analyses were performed to explore the nature of this interaction.
Bonferroni-corrected post-hoc follow-up tests7 were conducted to control for Type I error across the pair-wise comparisons. The results showed the differences of all six pairs of “Post-test versus Pre-test” in the different ability groups of CMI and EMI students to be significant (p < .001). As shown in Figure 2, all six groups performed significantly better in the post-test than in the pre-test. The difference in means between the post- and pre-test results in each ability group was subsequently analyzed (see Figure 3), with the results revealing the CMI students in the low-ability group to have improved the most. In addition, the EMI students in the high-ability group realized significantly greater improvement than their CMI peers, but no significant difference was found between the medium-ability CMI and EMI groups.
Overall, these standardized test results on the topic of “Mechanics” displayed a similar pattern to that of the internal tenth-grade examination results. The low-ability CMI group again appears to have benefited most from the use of the mother tongue in classroom instruction, and the high-ability EMI students displayed a higher adjusted mean than their CMI peers, whereas the language of instruction made little difference to the medium-ability group.
Thermal Concept Evaluation Results
Similar to the FCI analysis, Levene's test on the TCE pre- and post-test results showed the variance in the scores of the low-, medium-, and high-ability groups to be homogeneous, F(pre-test) (5,193) = 1.56, ns and F(post-test) (5,193) = 1.29, ns, indicating that the homogeneity of variance assumption was satisfied. Repeated-measures ANOVA on the TCE results revealed significant main effects for Ability Group (F(2, 193) = 19.23, p < .001), MOI (F(1, 193) = 74.75, p < .001), and Time (F(1, 193) = 476.34, p < .001), as well as a significant interaction effect: F(2, 193) = 10.30, p < .001. Post-hoc analyses were performed to explore the nature of that interaction.
Bonferroni-corrected post-hoc follow-up tests were again employed to control for Type I error across the pair-wise comparisons. The results showed that the differences between all “Post-test versus Pre-test” pairs in the different ability groups of CMI and EMI students were significant (“Post-test versus Pre-test”(CMI Medium Ability): p < .005; all other pairs: p < .001). In other words, all groups of students performed significantly better on the post-test than on the pre-test (see Figure 4).
The difference in means between the post- and pre-test results for each ability group was subsequently analyzed, with the results shown in Figure 5. It can be seen that, whereas the EMI students in the medium- and high-ability groups achieved significantly greater improvement than their CMI peers in the same groups, no significant difference was found between the improvement levels of the low-ability CMI and EMI groups.
Questionnaire-Based Survey
We next examined the CMI and EMI students' responses to the questionnaire-based survey, which was carried out after the teaching intervention to assess students' motivation to learn physics (see Table 2). To compare the differences between their responses statistically, we ran chi-square non-parametric tests on the survey results, revealing significant differences between the two groups' responses to questions QS002, QS003, QS005, QS006, QS007, and QS008 (asterisked in Table 2). No significant differences were found for the responses to the two other survey questions.
| Item | Question | Group | Percentage (%) | Chi-square tests | ||||
|---|---|---|---|---|---|---|---|---|
| Never | Seldom | Often | Sometimes | Daily | ||||
| QS001 (missday) | Number of days of school missed during the intervention | CMI | 87 | 8 | 5 | 0 | 0 | χ2(2,199) = 32, p = .85 > .05 |
| EMI | 84 | 9 | 7 | 0 | 0 | |||
| QS002 (lateschl)* | Number of times late for physics lesson during the intervention | CMI | 76 | 16 | 2 | 2 | 3 | χ2(4,199) = 15.63, p = .004 < .01 |
| Low | 25 | 5 | 1 | 1 | 0 | |||
| Med | 25 | 7 | 0 | 1 | 1 | |||
| High | 26 | 4 | 1 | 0 | 2 | |||
| EMI | 52 | 22 | 13 | 4 | 9 | |||
| Low | 13 | 2 | 7 | 3 | 7 | |||
| Med | 17 | 15 | 4 | 1 | 2 | |||
| High | 22 | 5 | 2 | 0 | 0 | |||
| QS003 (nonotes)* | Frequency of coming to class without notes | CMI | 91 | 4 | 4 | 0 | 0 | χ2(2,199) = 9.39, p = .009 < .01 |
| Low | 29 | 1 | 2 | 0 | 0 | |||
| Med | 30 | 2 | 1 | 0 | 0 | |||
| High | 32 | 1 | 1 | 0 | 0 | |||
| EMI | 75 | 11 | 14 | 0 | 0 | |||
| Low | 17 | 6 | 9 | 0 | 0 | |||
| Med | 25 | 4 | 5 | 0 | 0 | |||
| High | 33 | 1 | 0 | 0 | 0 | |||
| QS004 (nobooks) | Frequency of coming to class without books | CMI | 93 | 4 | 2 | 0 | 0 | χ2(2,199) = 4.33, p = .12 > .05 |
| EMI | 84 | 9 | 7 | 0 | 0 | |||
| QS005 (nohw)* | Frequency of coming to class without assignment completed | CMI | 43 | 33 | 11 | 13 | 0 | χ2(4,199) = 41.85, p < .001 |
| Low | 12 | 13 | 2 | 6 | 0 | |||
| Med | 15 | 11 | 3 | 4 | 0 | |||
| High | 16 | 9 | 6 | 3 | 0 | |||
| EMI | 44 | 7 | 2 | 42 | 5 | |||
| Low | 2 | 2 | 1 | 24 | 4 | |||
| Med | 18 | 2 | 1 | 12 | 1 | |||
| High | 24 | 3 | 0 | 6 | 0 | |||
| Item | Question | Group | Percentage (%) | Chi-square tests | ||||
|---|---|---|---|---|---|---|---|---|
| Strongly disagree | Disagree | Neutral | Agree | Strongly agree | ||||
| QS006 (phyuse)* | Ever bored in physics class? | CMI | 43 | 31 | 16 | 8 | 2 | χ2(4,199) = 30.87, p < .001 |
| Low | 14 | 9 | 5 | 4 | 0 | |||
| Med | 11 | 13 | 7 | 3 | 1 | |||
| High | 18 | 9 | 4 | 1 | 1 | |||
| EMI | 22 | 34 | 5 | 24 | 15 | |||
| Low | 1 | 6 | 2 | 15 | 9 | |||
| Med | 7 | 14 | 2 | 6 | 5 | |||
| High | 14 | 14 | 1 | 3 | 1 | |||
| QS007 (phyfor)* | Look forward to physics class? | CMI | 1 | 2 | 11 | 34 | 52 | χ2(4,199) = 51.39, p < .001 |
| Low | 1 | 1 | 3 | 12 | 15 | |||
| Med | 0 | 1 | 5 | 11 | 17 | |||
| High | 0 | 0 | 3 | 11 | 20 | |||
| EMI | 18 | 19 | 16 | 34 | 13 | |||
| Low | 12 | 12 | 1 | 5 | 2 | |||
| Med | 3 | 4 | 12 | 11 | 4 | |||
| High | 3 | 3 | 3 | 18 | 7 | |||
| QS008 (bored)* | Physics is useful for my future | CMI | 22 | 25 | 5 | 25 | 23 | χ2(4,199) = 24.89, p < .001 |
| Low | 3 | 4 | 1 | 12 | 12 | |||
| Med | 7 | 11 | 1 | 9 | 7 | |||
| High | 12 | 10 | 3 | 4 | 4 | |||
| EMI | 7 | 7 | 9 | 33 | 43 | |||
| Low | 3 | 2 | 4 | 10 | 14 | |||
| Med | 1 | 3 | 3 | 13 | 14 | |||
| High | 3 | 3 | 2 | 10 | 15 | |||
- * Items with significant differences between the CMI and EMI students' responses.
The first two survey items (QS001 and QS002) explored the number of days that students had missed school and been late for their physics classes, respectively, during the intervention. The vast majority of both CMI and EMI students were on time for these lessons, although slightly fewer CMI students were late. These results are consistent with the outcomes of QS003 and QS004, which showed that more than three-quarters of students in both language groups always brought their notes and textbooks to their physics lessons, although slightly more EMI students arrived without their notes. However, there was a sharp contrast between the CMI and EMI groups for the fifth question (QS005). Approximately 76% of the CMI students reported always coming to class with their assignments completed compared to only about 51% of the EMI students. The latter's responses were stratified on this item, with just over half agreeing that they regularly finished their assignments before lessons and just under half disagreeing. Indeed, the results for this item also reveal stratification within the EMI ability groups. Of the 47% of EMI students who came to class without completed assignments, just 6% belonged to the high-ability group, whereas approximately 28% were low-ability students. Notably, a small proportion (around 4%) of EMI students (i.e., those in the low-ability group) encountered persistent difficulties in completing their assignments.
Divergent figures appeared again in the EMI students' responses to item QS006, with 39% agreeing that they felt bored during physics lessons and 56% disagreeing. In comparison, only one-tenth of the CMI students expressed agreement with this item. This stratified response on the EMI students' part was further underpinned by the pattern that emerged for item QS007: 37% of these students did not look forward to physics lessons, whereas 47% did. Interestingly, there was also a relatively large proportion (about 16%) of neutral opinions among the EMI students for this item. The percentage of low-ability EMI students who felt interested in studying physics (7% vs. 28%) and looked forward to physics lessons (7% vs. 25%) was around one-fourth that among the high-ability EMI groups.
The findings for item QS008 are striking. Slightly more than three-quarters (about 76%) of EMI students expressed the belief that physics would be useful to them in the future, whereas just under half (about 48%) of their CMI counterparts expressed that belief, with a similar percentage (47%) seeing little prospect of learning physics beyond the certificate level. The difference becomes even more obvious when we compare the high-ability groups of CMI and EMI students. The percentage of high-ability EMI students who said they believed that physics would be useful in the future was slightly more than three times (25% vs. 8%) higher than that of their CMI counterparts.
In general, the results of the questionnaire-based survey provide statistical support for the alternative Hypothesis 2. Although synthesis of the survey results revealed a number of interesting findings, and a variety of implications regarding the role of MOI in students' motivation to learn physics, we decided to carry out follow-up interviews to see whether further, more conclusive evidence would emerge.
Follow-Up Interviews
It was easier for me to feel bored when I read the textbook in English. Sometimes I even gave up reading the textbooks and just read the teacher's notes for quizzes, tests, and examinations (Steven, EMI, Low-Ability Group).
It was hard for me to memorize the vocabulary and understand the abstract concepts at the same time (Amy, EMI, Low-Ability Group).
Although there were some abstract terms in the topic of “Heat” that were difficult for me to express in Chinese, for instance, the key term “re” (
) is always confusing, as I am unsure whether it means “heat” or “hot,” in general, I understood and felt interested in learning physics (Ching, CMI, Low-Ability Group).
Some textbooks translated and described “heat” as “re liáng” (
) [the capacity of heat], but others stated that “heat” is “re” (
) [a kind of energy]. Because we sometimes use this word, “re” (
), to describe “hot” in daily-life situations, we are confused by it. However, I think in general I learned physics better by using Chinese, although some other students might experience difficulty in learning the “Heat” topic in Chinese (Jason, CMI, Low-Ability Group).
I was quite satisfied with my learning [through English]. However, I think using Chinese would give an absolute advantage in the HKCEE [HK public examination]. This is because there are some daily-life questions that require students to describe and explain some scientific principles. [The ability to use your] mother tongue always gives you an edge (Billy, EMI, Medium-Ability Group).
I insisted on learning physics through English because if I never use English, I will never become familiar with it. Indeed, I think my English standard has gradually improved (Kitty, EMI, Medium-Ability Group).
I think I did gradually improve my general English standard. However, this was not the case for all students. I noted that some of my low-achieving classmates failed to cope with the difficulty of learning physics through English, and they were adversely affected (John, EMI, High-Ability Group).
I felt good studying [physics] using English. I think science should be studied in English because the world is using English for knowledge communication, especially in science. Western countries are dominating advances of science, so we can only learn the latest developments in science by using English. In fact, I am confident about using English in science communication globally in the future (Benson, EMI, High-Ability Group).
I was doing well [studying physics] using Chinese. My marks were over 70 on many quizzes and tests. However, all of the universities in HK teach science through EMI. Thus, if we want to study science in the future using CMI, our prospects are very limited. So, many classmates decided to start using EMI at the certificate level, instead of at the advanced or university level, if they wanted to proceed to higher education (Ivy, CMI, Medium-Ability Group).
Compared to the EMI students, we found it more difficult to explore resources extensively for our projects. Moreover, extra learning materials in Chinese are very scarce. I think that is the drawback of using Chinese to study physics (Jack, CMI, High-Ability).
In general, it was more difficult for us to use Chinese keywords to search for information on the Internet regarding the concepts and principles of our project work. We needed to use English keywords to search for resources and then translate certain websites into Chinese for reference (Jenny, CMI, Low-Ability).
Findings and Discussion
Situated in the distinctive educational context of HK, where most secondary schools shift their MOI from Chinese to English at the senior secondary level, this research was designed to examine the effect of MOI on student learning in certificate-level physics. The results of the ninth- and tenth-grade physics examinations, two standardized conceptual assessments (TCE and FCI), a questionnaire-based survey, and follow-up interviews were analyzed in terms of students' academic achievement and motivation to test the two proposed research hypotheses. Because Chinese and English alone were used in teaching the topics of “Mechanics” and “Heat” in a tenth-grade physics intervention, that is, the bilingual and mixed-code approaches were prohibited in the school, we were able to evaluate whether the benefits of learning in the mother tongue identified in the previous literature would persist as students progressed from junior to senior secondary study.
In line with existing research on junior science students in HK (e.g., Education Commission, 1986; Ip & Chan, 1985; Lo et al., 1985; Marsh et al., 2000), the results of the tenth-grade examination, conceptual assessments, and follow-up interviews reveal conclusive evidence that learning in their native language, that is, Chinese, helps low-ability students at the senior secondary level to make greater improvement in physics than learning in a second language. The findings of the questionnaire-based survey also indicated that, overall, the CMI students in this study were more motivated to learn physics than their EMI counterparts. Underpinned by the literature on language in science learning (e.g., Maerten-Rivera et al., 2010; Okebukola et al., 2013), we reason that native language use is advantageous in the science context, as science requires students to simultaneously learn new reading, writing, listening, and speaking skills in acquiring both the academic language of science and the subject content (Wellington & Osborne, 2001). At the same time, however, the data revealed a different orientation among the high-ability students. In contrast to the results of the research targeting junior science students conducted by Yip et al. (2003), ours suggest that English is a more effective MOI for high-achieving senior students. The most likely reason is that the high-ability EMI group in our research became more competent in mastering both science content and academic language in English as they progressed from the junior to senior level of study, which is borne out by the interview data showing that the students in this group (e.g., Billy and Benson) felt capable of learning physics in English and were confident in using the language in science communication globally in the future.
In contrast to their EMI counterparts, the CMI students in both the high- and medium-ability groups faced a greater dilemma in being instructed in Chinese alone, which not only affected their learning of physics, but also their future studies and even their career prospects. The interview data reveal that they struggled with the issue of whether they should switch to EMI if they progressed to advanced-level study or university education. Moreover, the lack of relevant science materials in Chinese also made them hesitant about learning in their mother tongue. Their dissatisfaction in this regard was heightened by their attempts to find suitably extensive reference materials for their project work.
Based on these results, we argue that the MOI in the teaching intervention made a significant difference to the outcomes of the CMI and EMI groups. First, bearing in mind that the EMI students had just begun to learn physics through English (i.e., their second language), whereas their CMI peers continued learning in Chinese, the teaching medium in the lessons is considered to be the crucial factor influencing their performance in both the assessments (i.e., FCI, TCE, and tenth-grade examination) and learning activities (e.g., classroom discussion and project work). Indeed, as stated above, science learning involves the simultaneous mastery of both subject content and what is effectively a new language (Isa & Maskill, 1982; Strevens, 1980), that is, the language of science. When this must be accomplished in a second language, which inevitably presents a double challenge for students whose scientific competence is limited (i.e., low-ability students), the added cognitive load affects students' performance (in science) negatively. Although language support was provided in the lessons in this research (e.g., read-aloud activities and vocabulary lists), it is no surprise that the use of a non-native language was difficult for low-ability students beginning their senior secondary career already lagging behind the average in science proficiency.
Second, in addition to the MOI in the lessons, the language of the textbooks and learning resources was another significant factor affecting students' learning. Both the CMI and EMI students expressed concerns about this issue in the interviews. The low-ability EMI students reported that it was difficult for them to read the textbooks in English and memorize the vocabulary, whereas their CMI counterparts in all ability groups complained about the difficulty of finding adequate learning resources in Chinese. Rollnick's (2000) research sheds light on the former finding. She noted that to communicate well in a “scientific way” through a second language, students need to learn the required vocabulary, sentence patterns, and texts gradually. Therefore, students (especially low-ability students) have to take time to acquire these skills to read and “digest” textbook content. Further, as Gee (2008) pointed out, because the academic language of science is quite different from daily language, it is virtually impossible for students to comprehend its meaning simply by translating scientific texts. With regard to the CMI students, given that the availability of extensive learning materials (e.g., eBooks, videos, and software) encourages a deeper interest in the subject and enhances study independence, limited such availability may hinder students' capacity to hunt for more in-depth explanations of novel scientific concepts (Petitto & Dunbar, 2004). It is reasonable to believe that high-ability students would be more eager to explore supplementary materials, which may explain why the high-ability CMI students, who suffered from a lack of such materials, were outperformed by their EMI peers.
Implications and Future Research Directions
Several of the observations made herein have long-term implications for the local education community, and can likely be extended to scholarship in the international context. First, some researchers may have doubts about our unanticipated finding that the CMI students in the low-ability group demonstrated approximately the same level of improvement on the TCE conceptual assessment as their EMI counterparts, whereas the medium- and high-ability CMI students achieved less improvement than their EMI peers. With the exception of the high-ability group, these results present a different picture from the tenth-grade examination and FCI results. However, this deviation is explicable when we turn to the follow-up interviews, in which some of the CMI students expressed difficulties in understanding the concepts behind the keywords in the topic of “Heat.” These findings supplement previous studies (i.e., Isa & Maskill, 1982; Kawasaki, 2002) showing that many languages, including Chinese, do not possess the precise vocabulary required to describe abstract scientific concepts. We thus believe the CMI students' underperformance on the TCE to be understandable, and argue for the integrity of our conclusions.
More broadly, this finding has significant implications for the cultural dimension of scientific language (Brown & Ryoo, 2008) and highlights the drawbacks of teaching and learning science in certain home languages, including Chinese, Vietnamese (Jenkins, 1979), Serbian (Trenkic, 2002), and Japanese (Ogawa, 1998). The students in the current research were somewhat confused about the meaning and concept of “heat” because Chinese uses the same word, “re” (
), for both “heat” (noun) and “hot” (adjective). It is thus rather difficult to express this abstract concept using Chinese alone. Similarly, although Japanese students are generally able to think in abstract terms, their language uses “concrete and intuitive signs” that are ill-suited to semantic modification and the presentation of abstract concepts (Tudjman, 1991). We thus suggest that science teachers and educators, particularly those who teach students through the aforementioned languages, place greater emphasis on elucidating the abstract concepts behind key terms in various scientific contexts.
Second, special attention should be paid to our results concerning CMI students' learning prospects. In the questionnaire-based survey, roughly half of these students claimed that they had no prospect of learning physics beyond the senior secondary level. The widespread use of English in science teaching in HK higher education (owing to public expectations) and severe lack of supplementary learning materials in Chinese affects CMI students' aspirations for the future, with the difficulty of exploring advanced science resources using their native tongue deterring them from pursuing further studies. Given the limitations of CMI in science study, we recommend that English be used as a supplementary teaching medium in secondary science classrooms to serve as a bridging device for CMI students. In fact, in the interviews in this research, both the EMI and CMI students expressed the need for home language and second language support. The latter reported that learning key terms in English was essential for their future studies, whereas the former said the use of Chinese for enrichment proposes could improve their learning. As the practice of helping second-language students to comprehend meanings in their native language is widely accepted (Liebscher & Dailey-O'Caine, 2005; Webb, 2010), we believe that home language instruction with supplementary vocabulary given in the second language would benefit students in HK, particularly low-ability students, and may also be appropriate for other cultures in which two languages are widely used in education. At the same time, we also recommend that the second language (i.e., English) be adopted as the major MOI for high-ability students, with home language support.
Third, although we investigated a single school in this study, which may not be a strong basis for statistical generalization, we argue that it allows for “naturalistic generalization” (Stake, 1978). In this view, generalization is seen as a product of human experiences, with researchers identifying “the similarities of objects and issues” (Stake, 1978, p. 6). In this research, we offer empirically grounded results addressing a real-life issue (the change in MOI from Chinese to English at the tenth-grade level) that affects most secondary schools in HK. Our research thus provides a template for exploratory study (Yin, 1993). Future investigations can replicate and build upon it to extend the knowledge it offers (on the effects of MOI on students' physics learning) to a broader intellectual context (i.e., secondary schools in other countries). However, it must be acknowledged as a limitation that we were unable to engage in in-depth analysis of student performance on the examinations and conceptual assessments and/or the ways in which they interacted during CMI and EMI classes. Moreover, although the factors of instructional strategies and learning environment between the two MOI classes were controlled, the issue of classroom culture is yet to be addressed. As Morton (2012) pointed out, a more dialogic approach in class would possibly allow learners to use scientific language more meaningfully. A closer look at classroom talk and interactions (e.g., Lehesvuori, Viiri, Rasku-Puttonen, Moate, & Helaakoski, 2013; Zhai & Dillon, 2014) would constitute a methodological response to Roth's (2014) call for better analysis of how a superior understanding of science language can be constructed among students. Therefore, we believe careful examination of classroom episodes would provide more information on how the views expressed in the student interviews were constructed and whether those perspectives affected the questionnaire responses. Future studies could explore these issues through qualitative analysis.
Fourth, with the recent endorsement of the government's fine-tuning policy, secondary schools in HK may now decide to introduce different MOI arrangements for S1 students in accordance with the language proficiency of teachers and students' academic performance. It can be foreseen that schools that are currently using CMI will be pressurized into switching to EMI in response to parental demand and intense competition for better student intakes. On the surface, such a measure would seem to afford greater flexibility to schools, allowing them to switch back to the use of English alone or adopt a mixed code. At the same time, however, it will create new challenges for a large group of students, low-ability students in particular, in acquiring language and science literacy at the same time. From the teachers' perspective, after years of teaching through the medium of Chinese, CMI teachers will now need to undertake professional training to improve their English-language skills. It would be of interest in future studies to examine how the re-introduction of mixed code or a purely English or Chinese MOI affects science students and teachers alike. As there is no conclusive evidence showing which language of instruction best supports medium-ability students, or what types of support should be provided to these students, further investigation is necessary in this regard.
We thank Mr. Raymond Lam for his consultation on and involvement in this research project.
Notes
1 For further details, refer to Yip and Tsang (2007, pp. 394–396).
2 These numbers include government, grant, and aided secondary schools, but exclude private schools.
3 The expression is a null hypothesis, and, because an ANCOVA model was adopted in the analysis, adjusted means (i.e., means adjusted for the covariate) were considered for the level of achievement. Further details are provided in the following sections.
4 The effect size in ANCOVA can be estimated by Cohen's d measure (Cohen, 1988; Lipsey, 1990): 
, where µ1 and µ2 are the respective adjusted means of the tenth-grade results of the two groups under comparison, σ is the standard deviation, and ρ is the correlation between the tenth-grade and ninth-grade results.
5 The coefficients of both Ability Group and Ninth-Grade Examination Results were found to be statistically significant (p < .05 and p < .01) in the ANCOVA model.
6 Bonferroni correction is an adjustment on the p-values of multiple pair-wise statistical tests. It can reduce the chances of Type I errors (i.e., false positives) when several tests are performed simultaneously on the same dataset. In the current case, as there were three pairs for comparison, that is, Low (CMI vs. EMI), Medium (CMI vs. EMI), and High (CMI vs. EMI), Bonferroni correction was adopted to adjust the p-value for each pair to 0.05/3 = 0.016 to account for the increased possibility of Type I errors.
7 Refer to the previous footnote. As there were six pairs for comparison in this case, Bonferroni correction was adopted to adjust the p-value for each pair to 0.05/6 = 0.0083.
