Jeannette Wing (2006, 2011) opened the field of CT by inaugurating the term on a conceptual level and defining it as the “thought processes involved in formulating problems and their solutions so that the solutions are represented in a form that can effectively be carried out by an information-processing agent.” Since then, there have been multiple efforts attempting to encapsulate the theoretical and operational underpinnings of the ill-established term. Currently, CT is widely understood as a competence that builds on the disciplinary notions in computing and the power of modern digital computer devices while solving problems in the real world (Denning and Tedre, 2019). Although such conceptual ideas have been recently framed specifically in CT, they have been taught, learned, and studied in schools and in the field of computing education for decades (see e.g., Pea et al., 1987) and, in a broader sense, long before modern digital devices had been invented (Tedre and Denning, 2016).

In the light of this framing for CT, there have been various scholarly efforts shaping the core CT skills and areas of knowledge that teachers should teach and students should learn in basic education today. Recent studies have begun distinguishing a “multiliteracy” dimension in CT by expanding it with sociocultural approach (Kafai et al., 2019; Mertala, 2021) and encompassing such pedagogical notions as relating to designing with technology on a personal level and critically reflecting the societal impact of contemporary computational technologies (Høholt et al., 2021). Despite this nascent but educationally meaningful viewpoint, the “problem-solving” dimension in CT has been generally more prevalent, portraying how CT can provide skills to understand and solve concrete real-life problems with computational methods. This dimension has been characterized with such core concepts and practices as abstraction, decomposition, algorithms, evaluation, and generalization that students are expected to learn (Barr and Stephenson, 2011; Grover and Pea, 2013; Shute et al., 2017). Establishing on this dimension, the theoretical definition of CT in ICILS 2018 (Fraillon et al., 2019b, pp. 27–28) involves two “strands,” that is, conceptual categories framing its core skills and knowledge, and five more specific “aspects,” that is, content categories for the two strands (see Figure 1).

The core conceptual and practical principles of CT are generally introduced to students through computer programming in teaching and learning practice in basic education (Grover and Pea, 2013). Nonetheless, CT and programming are not synonyms: programming is a concrete hands-on activity that can foster the more generic CT skills that go beyond mere programming knowledge and transfer across computational problem-solving domains (e.g., tools used to design computational solutions) (Tang et al., 2020). Cognitive underpinnings involved with CT are thus relevant beyond mere programming contexts (Wing, 2006) and have been portrayed in the context of various school subjects as well (see e.g., Settle and Perković, 2010; Moreno-León et al., 2017).

Educational programming tasks and activities for CT typically involve a real-world computational problem that students need to understand and analyze and subsequently solve or meet by designing a computational solution, such as an algorithm implemented with computer code. Although a CT methodology focused on data-driven machine learning is currently rising (see e.g., Vartiainen et al., 2020), the solutions—and thus views of CT—have commonly followed the traditional “step-by-step” algorithmic computing methodology. Several educational programming environments that build upon this methodology are easy to access and use for both teachers and students: they emphasize learner-centered learning that is expected to capture the interest of students and engage them in learning, which lowers the threshold to include them in teaching and learning activities (Lye and Koh, 2014). Popular approaches include a variety of block-based programming or “coding” environments on the web, such as Scratch, purchasable programming tools and kits, such as Lego robotics and Micro:bits, and text-based programming languages, such as Python and JavaScript (Garneli et al., 2015). The manifold programming environments, tools, and languages are used in more drill-like exercises, such as those presented in (Lambić et al., 2021), and more learner-centered, creative craft-like projects in which students can more broadly apply their creativity and interest areas (Brennan and Resnick, 2012).

The scarce prior empirical studies on teachers’ programming motivation have focused mainly on pre-service teachers and computer science (CS) or information technology (IT) teachers likely because programming and CT are relatively new cross-curricular contents in basic education. As expected, pre-service CS teachers have had moderately positive attitudes toward programming and moderately high self-efficacy in programming, which has also correlated with their achievement in programming-themed training courses (Gurer et al., 2019). Otherwise, there have been preliminary findings showing that basic education teachers on average consider training programming to be difficult and may lack understanding regarding the purpose of programming education (Hartell et al., 2019). Concurrently, some teachers have been noted to disapprove programming curricula (Mühling et al., 2010). Teachers’ comfort levels to teach programming, create CT teaching materials, and integrate CT in their lessons have also varied widely, suggesting that some teachers may not be well prepared to integrate CT in their teaching (Garvin et al., 2019).

Several studies have investigated teachers’ motivation-related aspects in specific educational programming contexts. Notably, gaining hands-on experience of programming has been shown to increase positive attitudes and negate initial gender differences in attitudes between pre-service IT teachers (Erol and Kurt, 2017). Gaining programming experience has also improved pre-service teachers’ attitudes toward programming and utilizing ICT generally in teaching (Fesakis and Serafeim, 2009). Similarly, in Yukselturk and Altiok’s (2017) study, pre-service teachers’ enjoyment of programming (an aspect of intrinsic programming motivation) and self-efficacy increased whereas their fear of programming decreased after attending a Scratch programming course. However, their attitudes toward the value and importance of learning programming (an aspect of extrinsic programming motivation) did not increase. Choi’s (2013) study demonstrated a similar change: pre-service teachers’ initial thoughts on programming being, for instance, difficult, scary, and perplexing changed to a level of enjoyment, accomplishment, and confidence after gaining programming experience. However, teachers have been shown to have positive attitudes especially toward specific pedagogical solutions (e.g., lesson structure plans) in programming education (Sentance et al., 2019), suggesting that the nature of the manifold CT contents being adopted by teachers are important to consider. Similarly, CT has been shown to include specific substances that can be stronger causes of uncertainty for teachers (Rich et al., 2021).

Compared to scarce previous research on teachers’ programming motivation, several previous studies have examined motivation and attitudes toward programming-related topics among students of different ages and with different backgrounds. Notably, Sun et al. (2022) recently found that Chinese Grade 7 students’ high programming interest (an aspect of intrinsic motivation) was related to better CT learning outcomes. The study also found that girls had on average higher CT proficiency than boys but more negative programming attitudes, including interest. Other studies have also provided evidence that boys have more positive attitudes toward programming-related topics than girls (Mahoney, 2010; Kier et al., 2014; Gunbatar and Karalar, 2018; Kong et al., 2018), corresponding with the common view of computing-related studies and work fields being male-dominant. However, other studies (Mason and Rich, 2020; Gul et al., 2021) have not found gender differences in students’ programming attitudes, and gender differences have also been shown to disappear in some studies after gaining programming experiences (Gunbatar and Karalar, 2018). On another note, results regarding the role of gender in the level of students’ CT proficiency have also varied: in ICILS 2018 (Fraillon et al., 2019a), the average CT proficiency in all participating countries was in favor of boys. However, statistically significant differences were found only in two countries: in favor of boys in Portugal and of girls in Finland. Then again, another Finnish study (Vainikainen et al., 2022) found that boys had higher programming proficiency than girls.

Previous studies have discovered that students’ programming motivation associates with such factors as ethnicity, grade level, coding frequency, and math interest (Mason and Rich, 2020; Gul et al., 2021). Although social factors, such as home socioeconomic background, have been positively related to students’ better performance in CT tasks (Fraillon et al., 2019a), their role in programming attitudes has been so far inconclusive (see Mannila et al., 2020; Mason and Rich, 2020). However, as with teachers, gaining actual programming experiences has been shown to promote students’ positive programming attitudes, including their interest (Sun et al., 2022). The main reason is likely that programming with contemporary tools is generally designed to be motivating for students in different educational levels and in different contexts (Garneli et al., 2015). Several context-specific studies (e.g., Ruf et al., 2014; Asad et al., 2016; Jiang and Wong, 2017; Lambić et al., 2021; Tisza and Markopoulos, 2021) have illustrated in practice how students’ attitudes toward programming can increase after programming with a variety of tools and while participating in collaborative, engaging, and appropriately challenging and interesting coding tasks (see Lye and Koh, 2014; Dohn, 2019; Sharma et al., 2019). Connectedly, the number of years of programming learning has been shown to positively relate to students’ programming attitudes (Sun et al., 2022), demonstrating the importance of sustained exposure to programming rather than gaining singular introductions to only specific ways of programming, such as the popular “Hour of Code” (see also Mason and Rich, 2020).

International Computer and Information Literacy Study is an international study organized by International Association for the Evaluation of Educational Achievement (IEA) in partnership with Australian Council for Educational Research and national research centers of participating countries in 2013 and 2018. ICILS 2018 (Fraillon et al., 2019a, pp. 9–10) gathered data of Grade 8 students’ computer and information literacy (CIL) proficiency from 12 countries and two benchmarking areas as well as their CT proficiency as an optional assessment from eight countries and one benchmarking area. In addition to assessing students’ skills and knowledge, contextual questionnaires were presented for students, teachers, school ICT coordinators, principals, and national research centers. The tests and surveys were conducted on computers.

In the present study, we utilize student and teacher data collected in ICILS 2018 in Finland. The sampling design (see Fraillon et al., 2020) involved multi-stage sampling, stratification, and cluster sampling. The sampling of students and teachers specifically was a two-stage cluster sampling (see also Fraillon et al., 2019a, p. 11). In Finland, first, 150 schools with the target grade students were randomly selected with a probability proportional to size and utilizing the NUTS classification (see Statistics Finland, 2022) for regions. Second, 20 Grade 8 students (or all students if less than 25 students were enrolled in the grade) were randomly sampled in each sampled school. Additionally, 15 teachers that taught Grade 8 students at the time of testing (or all teachers if less than 20 Grade 8 teachers worked in the school) were randomly sampled in each sampled school. The sampled teachers taught different subjects, and the sampling did not consider if a teacher had taught any of the sampled students or not. The final national data in Finland concluded 2,546 students and 1,853 teachers from 145 schools.

Teachers in ICILS 2018 responded to a questionnaire enquiring about their perceptions and use of ICT and various background variables. In the present study, the examined background variables were gender (binary), age, subjects taught, and previous programming teaching experience (yes/no). Teachers’ CT instruction emphasis was assessed in the questionnaire with a question enquiring how much in their teaching of the reference class¹ during the current school year they emphasized the given CT skills (see items in Table 1). The response options were presented on a 4-point Likert scale (1 = strong emphasis, 2 = some emphasis, 3 = little emphasis, 4 = no emphasis). Items A to I were international items and items J to M nationally added items that were more contextualized to programming.

Teachers’ programming motivation was assessed in a separate question using two subscales utilized from the Intrinsic Motivation Inventory (IMI²; for validity, see McAuley et al., 1987): (1) interest/enjoyment (items A to C in Table 2, Cronbach’s α = 0.81) and (2) value/usefulness (items D to F, Cronbach’s α = 0.82). The interest/enjoyment subscale measures teachers’ intrinsic motivation, whereas the value/usefulness subscale measures teachers’ identified motivation (see also McAuley et al., 1987; Deci et al., 1994), that is, one form of extrinsic motivation. Teachers were asked, “When thinking about your relationship toward programming and [teaching/considering teaching] it, to what extent do you agree or disagree with the following statements?” As demonstrated with the square brackets, the subscales had two slightly different but construct-wise not dissimilar formats for teachers who responded “yes” or “no” to a previous question regarding whether they had prior experience in teaching programming. The response options were presented on a 4-point Likert scale (1 = completely disagree, 2 = disagree, 3 = agree, 4 = completely agree). The two subscales correlated moderately positively both for teachers with prior experience in teaching programming (r = 0.37, SE 0.07) and those without it (r = 0.44, SE 0.02). Teachers who answered the questionnaire also selected different subjects (one or more) that they taught for at least four lessons each week in the school in the school year (see Table 3).

In Finland, each sampled student in ICILS 2018 completed two of a total of five randomly selected 30-min CIL modules and two 25-min CT modules that were the same for all students (Fraillon et al., 2019a, pp. 8–9). Both CT modules had several problem-solving tasks with a unifying theme, and they were designed to measure the aspects presented in Figure 1. We used students’ CT task scores, which are based on five different plausible values. The analyses are performed for the each plausible value separately after which their average is used (see Fraillon et al., 2020, pp. 152–153, 224). The task scores from all the CT tasks were aggregated to establish a total CT test score with a standardized international mean score of 500 and a standard deviation of 100 (Fraillon et al., 2019a, p. 92).

Between the CIL and CT modules, the students completed a questionnaire enquiring their learning about ICT, CIL, and CT and various background variables. In the present study, the examined background variables were gender (binary), prior programming experience (yes/no), and home socioeconomic background, which was combined from the parents’ highest level of education, parents’ highest occupation based on the International Standard Classification of Occupations, and the amount of books at home (see Fraillon et al., 2019b, p. 40).

Students’ programming motivation was assessed in the questionnaire using the same two subscales utilized from the IMI as with the teachers. The students were asked, “When thinking about your relationship toward programming and [studying/considering studying] it, to what extent do you agree or disagree with the following statements?” The subscales were similarly those of (1) interest/enjoyment (items A to C in Table 4, Cronbach’s α = 0.88) and (2) value/usefulness (items D and E, Cronbach’s α = 0.76). The interest/enjoyment subscale measured students’ intrinsic motivation, whereas the value/usefulness subscale measured their identified motivation (see also McAuley et al., 1987; Deci et al., 1994), that is, one form of extrinsic motivation. Both aspects of programming motivation were measured using a 4-point Likert scale (1 = completely disagree, 2 = disagree, 3 = agree, 4 = completely agree), and, as demonstrated by the square brackets, the questions had two slightly different but construct-wise not dissimilar formats for students with and without prior experience in programming. The two examined types of motivations correlated moderately positively both for students with prior experience in programming (r = 0.53, SE 0.03) and those without it (r = 0.58, SE 0.02).

As the ICILS 2018 sample is not based on simple random sampling, the generalization of the results to the target population is not straightforward. The complex sampling design and the non-participation of schools, students, or teachers could lead to biased results if the data was treated as if it was drawn from a simple random sample. To achieve unbiased estimates of the corresponding population, sampling weights and non-response adjustments for each school were used when analyzing the data (Fraillon et al., 2020, p. 79). When estimating the variances and standard errors for the population statistics, jackknife repeated replication technique was used (see also Fraillon et al., 2020, p. 221).

For RQ1, descriptive statistics of the students’ and teachers’ responses to the programming motivation questions were explored. For RQs 2.1 and 2.2, teachers’ gender age, and subjects taught were set as explanatory variables, and the effect of those variables on the subscales of intrinsic and extrinsic motivation were analyzed by examining (weighted) mean differences and by linear regression. Similar analyses to explore the effects on the two motivation subscales were performed for students (RQs 3.1 and 3.2) using gender and home socioeconomic background as explanatory variables. The analyses for gender, age, and home socioeconomic background were performed separately for teachers and students with and without prior programming teaching/learning experience to examine differences between these two experience groups. This approach was chosen instead on just one regression analysis in which some connections were not visible and also because the motivation questions for the two experience groups were slightly different (see Tables 2, 4). Correlation coefficients between the subscales were examined using Pearson correlation coefficient. For RQs 2.3 and 3.3, regression analyses were performed for the scales of interest, that is, CT instruction emphasis for teachers (see also Fraillon et al., 2020, p. 191) and CT test scores for students separately for teachers and students with and without prior programming teaching/learning experience.

Statistical significances of the mean differences and regression coefficients were also obtained. The significances are based on observed t-values, which were compared to the critical values of the standard normal distribution. This is a standard procedure in large scale assessments and is based on large sample size and the asymptotic normality of estimators. In the regression analyses, no multicollinearity issues were observed. All analyses were performed with the IEA IDB Analyzer data analysis software.

Our next RQs concerned how teachers’ prior programming teaching experience, gender, subject taught, and age are related to their programming motivation. These RQs were examined by exploring (weighted) mean differences and linear regression analysis.

Teachers with prior experience in teaching programming had substantially higher intrinsic motivation than teachers without said experience (Mean = 2.63 cf. 1.79; difference 0.83, SE 0.04, p < 0.001, Cohen’s d = 1.34). In turn, extrinsic programming motivation was only slightly higher among teachers with prior experience in teaching programming (Mean = 2.82 cf. 2.68; difference 0.14, SE 0.04, p < 0.001, Cohen’s d = 0.24). Statistically significant gender differences did not emerge in teachers’ extrinsic programming motivation in either experience group, but such differences were found in intrinsic motivation among both groups (see Table 6). Male teachers generally reporting a higher level of intrinsic motivation than female teachers. Intrinsic motivation among male teachers with prior programming teaching experience situated on the positive side of the 4-point Likert scale (2.5) whereas among female teachers the average was in the middle (difference −0.15, SE 0.06, p < 0.05, Cohen’s d = 0.26). Gender differences among teachers without prior programming teaching experience followed a similar pattern (difference −0.22, SE 0.04, p < 0.001, Cohen’s d = 0.35), but the average values were comparably much lower and on the negative side of the scale. In contrast, both experience groups’ average extrinsic programming motivation leaned toward the positive side of the scale.

The subjects that the teachers taught were related to the teachers’ intrinsic and extrinsic motivation statistically significantly (see Table 7). Specifically, teachers’ intrinsic programming motivation was higher if the teacher taught IT-related subjects, mathematics, practical and vocational subjects, and creative arts. In turn, teachers’ intrinsic motivation was lower if they taught national or foreign languages, subjects in the aggregated category of “other subjects,” and human sciences or social studies. Higher extrinsic motivation related only to teaching IT-related subjects, and, for lower extrinsic motivation, teaching foreign languages.

Teacher’s age was weakly but statistically significantly related only to intrinsic motivation (but not extrinsic motivation) at the p < 0.05 level. Among teachers with prior experience in teaching programming, the regression coefficient was 0.01, denoting that intrinsic motivation was slightly higher among teachers in this experience group that were older than the average age of the respondents. Among teachers without prior experience in teaching programming, the coefficient was −0.01, SE = 0.00, β = −0.09, SE (β) = 0.03, p < 0.01, showing an inverted effect.

The broader adoption of CT in basic education could be aided by guiding all teachers to integrate CT in teaching through clear curricular guidelines. Effective implementation of such guidelines would naturally necessitate systematic training, which has been previously lacking in Finland. Because teachers have much autonomy, it is important to promote especially their intrinsic programming motivation, which appears meaningful for adopting CT in instruction and currently varied in the population. Teachers’ awareness of the possibilities of CT and programming could be expanded by providing them with further understanding of different computational tools and CT-related substance areas to pique their interest. It seems especially important to pay particular attention to in-service teachers especially with low programming motivation as well as to conduct further investigations among pre-service teachers to further understand the current situation in initial teacher training. It would also be interesting to examine potential changes in teachers’ and students’ motivation, for instance, due the continued presence of programming in the newest national core curriculum as well as examine how other types of motivation may relate to teachers’ CT instruction.

Alongside motivational incentives in spreading CT awareness among teachers, it is important to ensure that teachers are equipped with pedagogically effective ways to integrate CT that are also meaningful for the disciplinary nature of their subject. While there has been discussion about whether all teachers should be expected to include programming in their teaching, CT, the expected learning outcome of programming, has been expressed as being cross-contextual (Wing, 2006), and there are many concrete examples of integrating these topics across the curriculum (see e.g., Settle and Perković, 2010; Moreno-León et al., 2017). Training should thus also consider solutions on how the manifold aspects of CT could be content-wise reflected in the lessons of different subjects. It appears crucial to spread awareness especially regarding such more cross-contextual CT activities as understanding diagrams and using real-world data to solve problems with computational methods (see e.g., Barr and Stephenson, 2011; Grover and Pea, 2013). Thus, also the extents and ways in which teachers of different subjects emphasize different aspects of CT (see Figure 1 and Table 1) could also be studied more thoroughly in the future. Additionally, expanding pedagogical possibilities in the more non-problem-solving aspects of CT, that is, the dimension of computational multiliteracy (see Kafai et al., 2019; Høholt et al., 2021; Mertala, 2021) concretely for different school subjects could be highly impactful.

To support teachers’ overall teaching success in CT instruction, it can be vital to also consider how comfortably teachers can adopt different kinds of CT-related pedagogical contents (Sentance et al., 2019; Rich et al., 2021) and what role teachers’ CT self-efficacy can play in professional learning in CT (see also Mühling et al., 2010; Yukselturk and Altiok, 2017; Garvin et al., 2019). Future teacher training and research should thus also critically and meticulously consider teachers’ subject and pedagogical content knowledge in CT (see Mäkitalo et al., 2019; Kong et al., 2020). Additionally, the availability of teaching materials and assessment practices (Weintrop et al., 2019) and even such mundane factors as time for lesson planning (Waite et al., 2020), time in the syllabus, and time altogether for professional development are important, turning focus also into school resources and school leaders’ decision-making. Illustratively, in Finland, only 58% of school principals rated the educational outcome of developing students’ skills to write apps or programs as quite important or very important (Strietholt et al., 2021, p. 28), showing that nearly a half of school principals do not perceive students’ programming skills as important educational outcomes.

Teachers’ broader adoption of CT would likely also improve students’ exposure to CT while studying different school subjects. However, in terms of students’ programming motivation specifically, it is altogether vital to provide learning experiences that are interesting and enjoyable and that promote a sense of value and usefulness. The latter could be targeted by clearly highlighting the significant role of CT and programming in daily life and the world of professional work. Regarding the first, prior studies have already illustrated that programming in a variety of ways can be intrinsically motivating. Specifically, learners’ positive attitudes have been shown to increase after, to name a few, designing games and animations in the contexts of creative computing with Scratch (Ruf et al., 2014), maker-like activities (Tisza and Markopoulos, 2021), “unplugged” activities (Jiang and Wong, 2017), (Lambić et al., 2021), and text-based programming (Asad et al., 2016). Students’ motivational dimensions could thus be increased by providing personally interesting and enjoyable ways of programming from the variety of existing ways to do programming rather than through singular isolated introductions. Especially girls are in desperate need of such opportunities to diminish the prevalent views of programming being not for their gender. On the previous note regarding teachers’ pedagogical content knowledge in CT, it is important to note that students’ interest toward programming can wane when faced with difficult and tedious tasks (Dohn, 2019). Proper pedagogical planning in terms of appropriate instructional guidance (Lye and Koh, 2014) and collaborative, student-centered, and craft-oriented learning (Taub et al., 2012; Dohn, 2019; Sharma et al., 2019) can thus be crucial specifically from the viewpoint of student motivation and achievement.

This study is not without limitations. The first limitation is the cross-sectional study design. Despite the large and representative sample and accounting for the effects of meaningful covariates, the present study was correlational, which inhibits assertions on causality. For instance, we were only able to hypothesize whether motivation may lead to increased instructional emphasis and learning or vice versa. Another limitation is that the sampling design in the study prevented certain potentially valuable analyses, such as cross-examining the responses of teachers and students in the same schools or classrooms. In addition, even though we measured intrinsic and extrinsic motivation for both students and teachers, the measures were not identical since the student measures focused on motivation for learning programming whereas teachers’ measures focused on motivation in teaching programming. Otherwise, the items were identical apart from that the extrinsic motivation questionnaire consisted of three items for teachers and two items for students.

The international questions in the large-scale international comparative study were developed and the data was collected following a common approach agreed by several partners. The student questionnaire especially was broad, and therefore only few additional questions on motivation could be added to investigate select theoretical dimensions. Moreover, the data was self-reported by the teachers and students, and this setting can involve factors related to, for instance, self-esteem and self-presentation styles that can influence responses. The data was also collected in 2018, and the results thus indicate the situation on the eve of the new curriculum rather than its impact.