Introduction

How students approach a learning task is expected to determine the quality of learning outcomes. For instance, Marton and Saljö (1976) were the first researchers to make the distinction between approaches to learning by distinguishing between deep and surface learners. Deep learners are students who have the intention to look for meaning in the study materials by closely examining the content to sieve out the underlying concepts and relating these concepts to everyday life and one’s prior knowledge. Surface learners, on the other hand, are students who have the intention to only meet task requirements such as fulfilling course requirements or passing the examination. The most common strategy adopted by surface learners is rote learning or memorising the study materials. The tendency is to remember the concepts or symbols that represent the knowledge without understanding the meaning of these concepts and symbols, and hence they fail to internalise information. Generally it is assumed that the deep approach to learning results in “higher quality learning outcomes” and the surface approach to “lower quality learning outcomes” (Gijbels et al. 2005). In short, students that are surface learners are expected to perform less well in school as compared to deep learners. Besides the distinction between the surface approach and deep approach to learning, Biggs (1976) and Ramsden (1981) introduced a third approach, the achieving (or strategic) approach to learning. Achieving learners (or strategic learners) refers to learners who aim to get high marks by optimising their efforts and by organising their time and study strategies to earn a good grade. Given that achieving students’ focus is on doing well on the test, it is expected that they will generally perform better than surface learners (Biggs 1987a).

The concept of the three approaches to learning has been operationalized in many studies across different disciplines, educational contexts, and countries to measure how students generally approach learning and to make predictions about their academic achievement. One of the instruments that has extensively been used is the Study Process Questionnaire (SPQ) developed by Biggs (1987b). The SPQ is a 42-item self-report instrument that measures students’ deep, surface, and achieving approach to learning. Many studies were conducted with the SPQ, which generally revealed relatively weak correlations between surface, deep and achieving approaches and students’ academic achievement.

There are several studies that found no relationship at all between student approaches to learning and academic achievement. For instance, Groves (2005) conducted a study with second year students from a medical school and found no significant correlation between SPQ scores and academic achievement. In another study conducted by Jones and Jones (1996) with first-year Chinese health-care students in Hong Kong, no significant correlation was found as well. Likewise, Gijbels et al. (2005) conducted a study with second-year law students to gain more insight into the relationships between students’ approaches to learning and students’ academic achievement. The results replicated the findings from Groves; there were no significant correlations between the SPQ scales and students’ academic achievement. In a longitudinal study conducted by Wilding and Andrews (2006), students responded to the SPQ 1 month before enrolment into a university college and again mid way through their second year of study. Similarly, no significant correlations between the SPQ scales and academic achievement were found. Each of these four studies concluded that students’ approaches to learning do not predict students’ academic achievement.

However, in a large-scale study based on data collected from first-year accounting students, Ramburuth and Mladenovic (2004) found a negative and weak, yet significant, correlation between the surface scale of the SPQ and academic achievement (r = −.09, p = .01). Also, Snelgrove and Slater (2003) found a weak but significant correlation between the surface scale of the SPQ and nursing students’ academic achievement on an examination (r = −.22, p < .05). In yet another study with accounting students, Booth et al. (1999) found a significant correlation between the surface approach to learning and academic achievement (r = −.24, p = .01). Unlike the first three studies, the results of these three studies suggest that there is a relationship—albeit weak—between approaches to learning and academic achievement.

Contrary to the studies mentioned above, a longitudinal study conducted by Zeegers (1999) found a relatively strong correlation of .41 (p < .05) between deep approach to learning and academic achievement. The results of their study suggest that a deep approach to learning could be used to explain almost 17% of the variance in academic achievement. This is the first study that was able to explain more that 5% of the variance in academic achievement based on the SPQ scales. However, the caveat to this outcome is that the relatively stronger correlation of .41 (p < .05) was observed at the end of a three-year longitudinal study. The longitudinal study surveyed 200 science students in five different stages of their academic life at a university to evaluate the predictive validity of the approaches to learning on learning outcomes. Over the years, the number of students that responded to the study decreased to 60 students at the final stage of the study (for which the higher correlation coefficient was found). The correlation coefficient between deep approach and Grade Point Average (GPA) was initially weak r = .11 (p < .05), but gradually increased over the 3 years to r = .41 (p < .05). This increase in the magnitude of the correlation coefficient could imply (1) that students become indeed more deep learners over the 3 years of study or (2) that only the better students responded to the survey. Astin (1970) and Neilson et al. (1978), observed the latter in their studies where they found that students who responded to follow-up surveys are more motivated and tend to do better academically as compared to those who do not respond. The same may have happened in the Zeeger study.

Finally, Watkins (2001) conducted a meta-analysis with 55 independent samples including 27,078 respondents from 15 countries. The studies that were included in the meta-analysis used various self-report instruments that measure students’ approaches to learning with the SPQ used in one-third of the reported studies. Since the present study is concerned specifically with the SPQ, only the correlations between the SPQ scales and academic achievement are reported here. The meta-analysis revealed the following average correlations between the approaches to learning scales and academic achievement: surface approach (r = −.14), deep approach (r = .16), achieving (r = .16). Overall, this meta-analysis confirms the findings from the earlier mentioned studies that the relationship between students’ approaches to learning and academic achievement is generally weak, explaining only about 1–3% of the variance in academic achievement.

The results of the above studies suggest that student approaches to learning, as measured by the SPQ, is a relatively weak predictor of academic achievement. The question is why is this the case? An answer to this question may lie in what has been referred to by various researchers as constructive alignment of learning objectives, teaching and learning activities, and assessment task (Biggs 1992; Biggs et al. 2001; Biggs and Tang 2007). For instance, if the intended learning goals (and corresponding learning task) demand a deep conceptual understanding of a scientific principle but the assessment constitutes a mere reproduction of isolated facts about the scientific principle, then one could speak of a constructive misalignment because the assessment is inadequate in determining students’ thorough understanding of the topic as set out by the intended learning goals. Theoretically, the reverse could also be the case; the assessment task is designed to measure students’ deep understanding of a topic, but the intended learning goals and the learning task demand minimal in-depth understanding of the topic. Although the latter scenario is possible, the assessment method has been found to be the most salient factor influencing student approaches to learning (Biggs 1973, 1989; Marton and Saljo 1976; Ramsden 1992; Scouller and Prosser 1994). As such, the low correlations in the above mentioned studies may be due to inadequate assessments and assessment standards. Watkins (2001), for instance, argued that the low correlation is mainly due to teachers having different criteria in assessing students’ performance. He warns that it is possible that the assessments do not necessarily reflect the application of deep learning strategies. This implies that in order to have an adequate measure of students’ approaches to learning, the assessment system needs to measure students’ corresponding learning behaviours that reflect surface, deep and achieving approaches to learning. It may be that most assessments, however, do not measure students’ learning behaviours, but mainly their knowledge retention (Choppin 1990).

Because of this, we suggest that it may be more appropriate to include assessment measures, which provide a truer picture of students’ approaches to learning in the classroom. It may be possible that self-report measures of students’ approaches to learning need to manifest themselves in an actual classroom first before they can be considered as an adequate predictor of academic achievement. In other words, students may report that they generally perceive themselves as deep learners, but that does not necessarily mean that they apply these deep learning strategies in the actual classroom. Only if they demonstrate the appropriate learning behaviours in the classroom (e.g., being actively involved in searching for meaning, relating and applying concepts to real life examples and engaging in discussions in identifying the main ideas with others), they will perform well on the assessment because they have translated their self-reported approaches to learning into actual behaviour.

Based on the above, we propose that the relationship between approaches to learning and academic achievement is mediated by students’ actual learning behaviours in the classroom, which we will refer to as achievement-related classroom behaviours. This observational measure of classroom behaviours we used in the present study captured (1) the level of students’ engagement in the learning process, (2) the extent to which they engaged and persisted in self-directed learning, (3) the degree to which they participated in group discussions and work in teams, and (4) their understanding of what they had learned. Teachers observed and subsequently generated a performance judgement based on these four attributes (see Appendix for a detailed description of the rubrics used). Students’ engagement in the learning process captured their willingness and active participation during the lesson. For instance, teachers observed how actively students engaged in trying to identify and formulate the key concepts to be learned, be able to provide a coherent description of the learning objectives and their commitment to understand the learning task. Students’ self-directed learning entailed their ability to independently search for information and evaluate its usefulness in relation to the learning task. In short, it is a student’s ability to find, organize, analyse and evaluate information. The degree to which students engaged in collaborative learning was determined by their observable interactions with other peers during the lesson. This attribute requires the student to play an active role in the team and to remain committed in helping the team to become successful, both in terms of completing the assigned task as well as encouraging healthy relationship among team members. The quality of students’ understaning was observed in the way they verbalised their thought processes and their ability to constructively comment as well as critiques other team members’ ideas. For instance, the students were expected to demonstrate an inquisitive attitude, generate many useful questions in the course of the lesson, actively listen to other team members and provide constructive feedback to their ideas.

Gijbels et al. (2008) observed that students change their approach to learning based on the way they perceived the learning environment. A perceived heavy workload, perceived good teaching, clear goals and more freedom in learning were related to deep approach to learning. However, the study did not address to what extent these elements mediate the relation between assessment and students’ approaches to learning. With the present study, we propose to use an observational measure of students’ actual learning behaviours instead rather than a self-report measure, which may be more adequate as a mediator. Although the present study is not the first to suggest a mediator variable to improve and explain the relationship between approaches to learning and academic achievement, it is the first to our best knowledge that used an observational variable of students’ actual behaviours in the classroom.

In summary, the present study examined whether achievement-related classroom behaviours is an adequate mediator between student’s approaches to learning and academic achievement. In addition, this study examined which approach to learning has a greater influence on how well students perform academically. To that end, the SPQ was administered to 1,608 students at a polytechnic in Singapore. Correlational analysis, as well as structural equation modelling was used to examine the relationships. By doing so we expected to cast more light on the seemingly problematic relationship between the SPQ scales and academic achievement.

Method

Participants

A total of 1,608 students (56% female) with an average age of 18 (SD = 1.4) years participated in the study. The students were enrolled in a polytechnic in Singapore.

Measures

Study Process Questionnaire (SPQ)

The SPQ (Biggs 1987c) was administered to assess the extent to which students used the deep, surface and achieving approaches to learning. The SPQ is a 42-item self-report instrument consisting of three main scales: (1) deep approach to learning, (2) surface approach to learning and (3) achieving approach to learning. All items were scored on a 5-point Likert scale, ranging from 1 (never or only rarely true of me) to 5 (always or almost true me). Further, the two subscales that determine each main scale are determined by measuring students’ attitudes towards studies (motives) and their usual ways of studying (strategy). Thus, the SPQ is designed to provide six subscales scores: Surface Motive and Strategy, Deep Motive and Strategy and Achieving Motive and Strategy. The validity of the SPQ was established by means of confirmatory factor analysis (CFA). Items of the SPQ were clustered into groups representing the six subscales. This technique is called “item parceling” (Bandalos and Finney 2001; Little et al. 2002). Item Parceling is a measurement practice that is commonly used in latent variable analyses. According to Little et al. (2002), a parcel can be defined as an aggregate-level indicator, comprised of the average of two or more items. For further analysis we parcelled all items belonging to subscales (i.e., six parcels were formed). This procedure has been applied to the SPQ in previous studies (Biggs et al. 2001; Snelgrove and Slater 2003; Zeegers 2002). The data fitted the model very well: Chi-square/df ratio = .12, CFI = 1.00, RMSEA = .00. The reliability of the SPQ was determined by means of the Cronbach’s α, which was .74 for surface learning approach .87 for deep learning approach and .86 for achieving approach (average SPQ.82). Overall the construct validity and reliability was deemed adequate.

Achievement-related classroom behaviours

This measure was based on teacher observations which captured (1) the level of students’ engagement in the learning process, (2) the extent to which they engaged and persisted in self-directed learning, (3) the degree to which they participated in group discussions and work in teams, and (4) their understanding of what they had learned. A more detailed description of the observational assessment criteria can be found in the rubrics in the Appendix. A grade was assigned to each student after every lesson for all subjects over one semester (i.e., 16 observations per subject). Since some students were absent for some classes we encountered missing values (missing cells were 3.2% of cases). Since structural equation analysis does not allow having missing values, we used EM imputation to estimate and replace the missing values. This approach is admissible as long as the number of the missing cells is less than 5% (Graham and Hofer 2000). The grade was reflected on a 5-point performance scale: 0 (fail), 1 (conditional pass), 2 (acceptable), 3 (good), and 4 (excellent). The construct validity of this observational measure was established by means of CFA. For each student there were 16 observational measures (or indicator variables), which determined the latent variable achievement-related classroom behaviours. The data fitted the model well: Chi-square/df ratio = 5.33, CFI = .96, RMSEA = .05. The reliability of this measure was established by means of Cronbach’s alpha, which was .89. These values are indicative of adequate construct validity and reliability.

Academic achievement

As an academic achievement measure, written tests of 30 min duration were conducted every 4 weeks (totalling four test grades) over the semester for all subjects to measure students’ understanding of the concepts learned. Most of the tests were a combination of open-ended questions and multiple-choice questions. A grade was assigned to each student for each test. Scores were distributed on a scale ranging from 0 to 4 with .5 increments: 0 (full fail), .5 (fail), 1.0 (conditional pass I), 1.5 (conditional pass II), 2.0 (acceptable), 2.5 (satisfactory), 3.0 (good), 3.5 (very good), and 4.0 (excellent). The construct validity of this observational measure was established by means of CFA. For each student there were four indicator variables, which determined the latent variable academic achievement. The data fitted the model very well: Chi-square/df ratio = .02, CFI = 1.00, RMSEA = .00. The reliability of this measure was established by means of Cronbach’s alpha, which was .72. These values are indicative of adequate construct validity and reliability.

Procedure

The SPQ was administered online to all students 2 weeks before the end of the semester. Both achievement-related classroom behaviours and academic achievement measures were obtained from the institution’s registry database at the end of the semester. Overall mean scores were calculated for the written achievement tests and the achievement-related classroom behaviours as well as the mean scores for each subscale of the SPQ.

Analyses

In the studies mentioned earlier, almost all of them conducted correlation analysis to examine the strength and direction of the linear relationship between students’ approaches to learning and academic achievement. We followed this approach with the present study by first calculating Pearson’s correlation coefficients to examine whether we could replicate the findings from previous studies.

To determine if achievement-related classroom behaviours is indeed an adequate mediator between student approaches to learning and academic achievement, we tested a full and partial mediation model by means of structural equation modelling (SEM). We evaluated the assumptions of multivariate normality and observed 31 multivariate outliers. We removed the outliers from the subsequent analyses. We chose maximum likelihood parameter estimation over other estimation methods because the data were distributed normally. Chi-square accompanied by degrees of freedom, sample size, p value and the root mean square error of approximation (RMSEA) were used as indices of absolute fit between the models and the data.

The Chi-square is a statistical measure to test the closeness of fit between the achievement-related and predicted covariance matrix. A small Chi-square value, relative to the degrees of freedom, indicates a good fit (Byrne 2001). A Chi-square/df ratio of less than 3 is considered to be indicative of a good fit. RMSEA is sensitive to model specification and is minimally influenced by sample size and not overly affected by estimation method (Fan et al. 1999). The lower the RMSEA value, the better the fit. A commonly reported cut-off value is .06 (Hu and Bentler 1999). In addition to these absolute fit indices, the comparative fit index (CFI) was calculated. The CFI value ranges from zero to one and a value greater than .95 is conventionally considered a good model fit (Byrne 2001).

Results and discussion

As a first step in the analysis, we generated the correlation coefficients between the three approaches to learning and academic achievement. See Table 1 for the correlation matrix. Results revealed that there is a negative relationship between the surface approach to learning and academic achievement (r = −.07, p = .01) and a positive relationship between the deep approach to learning and achievement (r = .10, p = .00) as well as the achieving approach to learning and academic achievement (r = .14, p = .00).

Table 1 Correlation matrix and descriptive statistics of the SPQ and academic achievement (N = 1,639)
Full size table

Similar to the findings of the previous studies mentioned, as well as the meta-analysis by Watkins (2001), our data confirmed that the relationships between students’ approaches to learning and academic achievement are generally weak, explaining less than 3% of the variance in academic achievement. As such, the results of our study replicate the findings of the existing studies that used the SPQ to predict students’ academic achievement.

We argued earlier that the weak correlations could be due to a misalignment between students’ approaches to learning and how these approaches are assessed by means of achievement tests. We proposed that the self-report measures of approaches to learning need to translate themselves into actual classroom behaviours first in order to be an adequate predictor of students’ academic achievement.

The results of the causal model in which we included a measure of achievement-related classroom behaviours as a mediator, are depicted in Fig. 1.

Fig. 1
figure 1

Path model depicting relationships between approach to learning, achievement-related classroom behaviours and academic achievement. Note: numbers above the arrows represent standardised regression weights. All standardized regression weights are statistically significant at the 1% level

Full size image

The results of the analysis revealed that the data fitted the hypothesized model well, Chi-square/df ratio = 2.26, p < .01, CFI = .98, RMSEA = .03. All factor loadings were statistically significant at the 1% level. In this model, about 32% of the variance in achievement-related classroom behaviours could be explained by approaches to learning. This is considerably more than in the reported studies in the literature.

In line with previous studies, the achieving approach was the strongest predictor of achievement-related classroom behaviours (β = .61, p < .01). Achieving learners are versatile learners and they tend to adapt the way they approach learning tasks to what the teachers or the assessment scheme demands in order to get good grades. As such, they get cues from what they think they will be tested on. This finding is consistent with other studies, whereby Biggs’ achieving approach to learning has consistently been shown to have a higher positive relationship with academic performance. The high orientation towards achieving approach to learning is not necessary a bad thing as studies have indicated, particularly in an Asian context, it has been found that students adopt a wide variety of strategies to understand the subject matter (Kember 1996, 2000).

Surface approach had a weak to medium strong inverse relationship with the achievement-related classroom behaviours (β = −.37, p < .01). This inverse relationship was expected and indicates that students who follow a surface approach to learning do generally demonstrate less engagement in the learning activities. This result is similar to that obtained by Zeegers (1999), Watkins and Hattie (1981) and Ramburuth and Mladenovic (2004). It seems that the surface approach to learning has consistently shown to be a very strong indicator on how well students will perform academically, albeit negatively.

An interesting finding is that the deep approach to learning also had a weak inverse relationship with the achievement-related classroom behaviours (β = −.08, p < .01). A possible factor that might explain this negative relationship is how deep approach to learning is being measured in SPQ.

Achievement-related classroom behaviours in turn was a strong predictor of academic achievement (β = .42, ρ < .01). In addition, surface approach to learning has a direct inverse relation with academic achievement (β = .08, ρ < .01). Together they explain 18% of the variance in academic achievement. It seems, however, that achievement-related classroom behaviours has a higher impact on academic achievement. Thus, the result of the path model lent support for our hypothesis that achievement-related classroom behaviours, as observed by a teacher, are indeed an adequate mediator for the relationship between students’ approaches to learning and academic achievement.

General discussion

The objective of this study was to investigate how approaches to learning, as measured by the SPQ, are related to students’ academic achievement. Previous findings suggest that this relationship is relatively weak. It was hypothesised that the relationship between approaches to learning and achievement is mediated by achievement-related classroom behaviours. In addition, it was examined which approach to learning has the strongest influence on how well students perform academically. To test the above hypothesis, the SPQ was administered to a large cohort of 1,608 students that were enrolled in six different three-year diploma programmes at a polytechnic in Singapore.

Our first analysis replicated the results of existing studies; the three approaches to learning are relatively weak predictors of students’ academic achievement, explaining less than 4% of the variance in academic achievement. The subsequent analysis revealed that including achievement-related classroom behaviours turned out to be an adequate mediator, significantly increasing the explained variance in academic achievement to about 18%. A small part of this variance is contributed by surface approach to learning. This indicates that students that adopt a surface approach to learning, with the intention to only meet task requirements and involving rote learning of the study materials, will not do well academically. Overall, the results of the study suggest that the students’ approaches to learning need to be translated into actual classroom behaviours before they can be used to predict academic achievement.

An additional finding was that the deep approach to learning has a significant inverse relationship, albeit weak, with achievement-related classroom behaviours. Our model also shows that behaviours exhibited by students’ having strong tendency towards achieving and surface approach to learning can be observed fairly well by the teachers. The positive relationship between achieving approach to learning and surface approach to learning found in this study are comparable to results found in other studies.

A rather unexpected finding was that the relationship between deep approach to learning and achievement-related classroom behaviours is negative. Does it mean that the teachers do not reward students who exhibited characteristics of a deep learner? To answer this question, we examined the statements in the questionnaire further and noted some differences in the manner the statements were being phrased to measure deep approach to learning as compared to the other two approaches to learning. Examining the statements that are used to measure deep behaviours in the SPQ, we realised that they tended to be more cognitive and philosophical in nature whereby it may not be that apparent in classroom behaviour. This is in particular apparent to statements that are used to measure deep motives. An example of such a statement is “I believe strongly that my main aim in life is to discover my own philosophy and belief system and to act strictly in accordance with it” and “My studies have changed my views about such things as politics, my religion, and my philosophy of life”. In comparison with deep approach to learning, statements that are used to measure surface approach to learning tends to be more visible and easier for both students and teachers to identify with. An example is “Lecturers shouldn’t expect students to spend significant amounts of time studying material everyone knows won’t be examined” and “Even when I have studied hard for a test, I worry that I may not be able to do well in it”. This is the same for the statements that are used to measure achieving approach to learning. For instance, “I try to complete my given tasks as soon as possible after they are given out” and “I would see myself basically as an ambitious person and want to achieve top grades/position, whatever I do”. What has been able to produce a consistent outcome in predicting how well students will perform academically is surface approach to learning. As mentioned earlier, statements that are used to measure surface and achieving approach to learning are more behavioural in nature. As such, it is easier for students to identify with as they go through the SPQ.

The relationship between approach to learning and academic achievement is always of great interest to educational practitioners and researchers. In this research, we were able to develop a causal model to show that using the achievement-related classroom behaviours as a mediator has effectively improved the variance explained in academic achievements by more than 6 times (from 4 to 18%). Based on the findings of the study, it is recommended that if one wish to make predictions of academic achievement it would be more fruitful to used observational measures of students’ actual behaviours in the classroom. Other determinants of academic success such as factors relating to cognitive and non-cognitive predictors of academic achievements can be used as well. Ackerman and Heggestad (1997) established that cognitive ability is one important determinant of academic achievement. However, cognitive ability might lose their predictive power at higher level of education (Ackerman et al. 2001) and is not able to account sufficiently for individual differences in academic success. This leads to examining non-cognitive predictors, such as personality traits, to predict academic achievements. To understand personality traits, the Big Five model of personality is widely accepted and used model by many researchers. A meta-analysis on the Big Five showed that out of the five independent dimensions of personality, the dimension “Conscientiousness” correlates strongly and consistently with academic success (O’Connor and Paunonen 2007).

The present study provides empirical evidence that teachers’ observation of how students behave and learn in the classroom can effectively predict academic achievements. Future studies may examine the role cognitive ability and personality traits play (in combination with the observational measure) to further enhance the prediction of students’ academic achievement in a variety of educational settings.