Refining questionnaire-based assessment of STEM students’ learning strategies
- Birgit Griese^{1}Email author,
- Malte Lehmann^{2} and
- Bettina Roesken-Winter^{2}
https://doi.org/10.1186/s40594-015-0025-9
© Griese et al. 2015
Received: 14 January 2015
Accepted: 8 July 2015
Published: 12 August 2015
Abstract
Background
Learning strategies are considered a key aspect for academic performance not only for science, technology, engineering and mathematics (STEM) students. Refining their assessment thus constitutes a worthwhile research purpose. The aim of this study is to examine the 69-item LIST questionnaire (ZDDP 15:185-200, 1994) with the aim of shortening it while keeping its factor structure and thus its potential for describing learning behaviour and for identifying significant changes therein. This includes exploring if reduced scales remain internally reliable, both in terms of reliability measures and content, and to examine if they stay sensitive enough to capture developments in a pre-post design.
Results
Our cohorts consist of STEM students (N = 2374) from different engineering courses at Ruhr-Universität Bochum in Germany, typically predominantly males, some with insufficient background in mathematics or non-native speakers of German. The data was analysed using various statistic methods, e.g. reliability measurement and confirmatory factor analysis. Our findings show that about half of the original items (36 out of 69) are sufficient, reliability holds (Cronbach’s α > 0.7) and more variance is explained (56.17 % as compared to 45.65 %). Most content-related changes occurring when eliminating so many items survive critical scrutiny.
Conclusions
The study shows that it is possible to refine and considerably shorten the LIST questionnaire to the effect that all but one factor are kept intact. This will simplify future research enormously. Furthermore, the refined factor structure suggests reconsidering the postulate of metacognition as an easily accessed facet of learning behaviour—thus implying promising research perspectives.
Keywords
Background
In science, technology, engineering and mathematics (STEM) education, the passage from secondary to tertiary education is considered problematic, which is especially true regarding the challenges students encounter when being confronted with university mathematics (cf. Gueudet 2008; Henn et al. 2010). Students’ failure rates in STEM fields and particularly in engineering subjects are alarmingly high in many countries. In Germany, for instance, almost 48 % of engineering students fail in their first year of university studies (Heublein et al. 2012). In relation to other subjects, the gap between school and university mathematics seems to be extremely high and causes difficulties for students taking mathematics courses. The obstacles can be categorized in three categories, thus stressing the wide range of aspects: epistemological and cognitive difficulties, sociological and cultural difficulties and didactical difficulties (de Guzmán et al. 1998). Their dramatic character is depicted in the word “abstraction shock”, referring to the fact that university mathematics is adding a formal world to the mathematics encountered at school (e.g. Artigue et al. 2007; Tall 2004). The challenge of understanding (and influencing) how the learning of mathematics at university works is often addressed by the use of cognitive development theories, as mathematics is commonly regarded as a rational and cognitively demanding subject (cf. Dreyfus 1995). Studies therefore often elaborate on cognitive difficulties and conceptual obstacles experienced by students in how mathematics is communicated to them (Artigue et al. 2007), particularly referring to the formal level of university mathematics and the prevalent role of proofs (Selden and Selden 2005).
Students’ success is linked to a great extent to their capacity of developing “meta-level” skills allowing them, for instance, to self-diagnose their difficulties and to overcome them, to ask proper questions to their tutors, to optimize their personal resources, to organize their knowledge, to learn to use it in a better way in various modes and not only at a technical level. (p. 760)
Thus, a promising perspective is provided by exploring general and meta-level skills in terms of learning strategies (e.g. Wild 2005; Rach and Heinze 2011), whose investigation allows for revealing both the cognitive dispositions as well as affective barriers and pathways—and finally the interrelations between them.
Numerous research and instruction projects that attend to these issues can be found in many countries (cf. Dunn et al. 2012). The causality between adequate learning strategies and successful learning seems well established (for example, cf. Erdem Keklik and Keklik 2013). Depending on the specific research interest, previous studies have different emphases, for example, affective aspects or problem-solving skills. Most of these studies include data collection via questionnaires. To meet the challenge of assessing learning strategies appropriately, in the past, questionnaires were developed aiming at capturing different facets of these skills. In this paper, we report on refining and adapting one commonly used German questionnaire for assessing students’ learning strategies. Our research is dedicated to gaining a reliable and valid instrument that simultaneously allows for making good economic sense. Often, research is not restricted to merely capturing students’ learning strategies but also includes interventions, leading to a pre-post design with repeated surveys, so one essential goal lies in keeping the load for students within reasonable limits.
All data was collected within the scope of the design-based research study MP^{2}-Math/Plus/Practice, which focuses on supporting engineering students in mathematics (cf. Griese et al. 2013; Dehling et al. 2014). Among other issues, the project aims at supporting those with insufficient learning strategies and motivation by attempting to remedy these obstacles. To reveal the influence of the project interventions, the study draws on capturing students’ learning strategies and their development in the first semester at university.
Learning strategies
Learning strategies as indications for learning behaviour
Thus, we suggest that trait approaches recognize the necessity to measure behaviorally, and that behavioral approaches to competence recognize the role of cognitive, affective and conative resources. At this time, we encourage research on competence in higher education emanating from either perspective and paying attention particularly to the steps in between. (Blömeke et al. 2015, p. 7)
Accordingly, in the paper at hand, learning strategies are understood as all kinds of planned and conscious learning behaviour and the attitudes behind it, involving observable actions (e.g. solving tasks, asking questions, taking notes) as well as thought processes (e.g. planning, reflecting) on the basis of both cognitive and affective-motivational dispositions. This extends to the lack of planning and conscious actions as it also presents a characteristic of an individual’s learning strategy.
Research on the significance of learning strategies in mathematics education has its roots in contributions highlighting the role of affect, motivation and beliefs (cf. McLeod 1992, as a starting point), as all cognitive processes involve affective stances that moderate the tension between modes of intuitive and analytical thinking (e.g. Fischbein 1987; Stavy and Tirosh 2000). In particular, the theory of dual processes in cognitive psychology has been adapted to mathematics education, and the role of affective variables has been pointed out in this context (e.g. Evans 2007). These perspectives provide fresh views on learning processes and have done much to reach a deeper understanding of the obstacles involved. Findings reveal that students’ cognitive reflection, as a metacognitive variable, their beliefs about mathematics, and their self-efficacy, are all correlated positively and significantly with mathematical achievement (Gómez-Chacón et al. 2014). There is also evidence that metacognition impacts positively on learning strategies which in turn influences achievement (Griese et al. 2011).
In summary, it can be said, therefore, that this has led to a fortified interest in certain kinds of learning strategies. In the context of mathematics, overcoming motivational and affective barriers with the help of meta-skills, e.g. self-regulation, has become an important issue. What is more, mathematics demands the use of effective planning as well as organized and consistent work (cf. Rach and Heinze 2011). More than many other subjects, mathematics is cognitively challenging and needs motivational perseverance, thus representing an ideal research area for the influence of interventions addressing learning strategies, both on a general and a meta-level. However, the goals when assessing students’ learning behaviour in science are various: taking an inventory, describing the development, comparing or improving learning behaviour (cf. Lovelace and Brickman 2013), and so are the research interests, in mathematics, popularly performance prediction or the identification of at-risk students. Apart from more time-consuming methods, this has led to a great variety of questionnaires in many languages.
Questionnaires for assessing learning strategies
Questionnaires with different focal points (according to the background of the authors and their research interests) originate from this variety (cf. Pintrich et al. 1993; Weinstein and Palmer 2002). Schellings (2011) gives a comprehensive overview from an international and a Dutch perspective. Though her work is based on the text-heavy learning of history, the general categories of learning behaviour can be applied to other subjects as well. Differentiating between motivational and cognitive aspects when dealing with learning strategies is a widely accepted concept (cf. Nenniger 1999) and is in keeping with the understanding of affective aspects as a key issue.
In the following, approaches to capture learning strategies which have influenced subsequent research fundamentally will be outlined. The selection includes only those which reflect the importance of affective and motivational issues. Pintrich et al. (1993) developed a questionnaire “to measure college undergraduates’ motivation and self-regulated learning” (Artino 2005, p. 3), the Motivated Strategies for Learning Questionnaire (MSLQ). The MSLQ measures motivation and self-regulated learning in general and for a particular course by means of six motivation and nine learning strategies subscales. Initially, Pintrich and De Groot (1990) started by postulating a five latent factor structure comprising expectancy, value, affect, learning strategies and self-regulation. The items that were developed for operationalizing these constructs later formed the basis for the 15 subscales mentioned above. MSLQ has been applied in many research studies (Duncan and McKeachie 2005), partly aiming at developing a new conceptualization with respect to the significance of the single sub-scales (Dunn et al. 2012; Hilpert et al. 2013). MSLQ’s reliability has proved “robust”, and its predictive validity to actual course performance is considered “reasonable” (Pintrich et al. 1993, p. 801).
The Approaches to Studying Inventory (ASI) by Entwistle and Ramsden (1983) and its refinements (ASSIST by Tait et al. 1998, ALSI by Entwistle and McCune 2004) feature the main distinction of categorizing learning behaviour as being of either strategic (deep) or of apathetic (surface) approach. The dichotomy forms the inventory’s two main factors which in turn contain up to 16 subscales, depending on the version of the questionnaire. Although the authors do not group their items into motivational and (meta-)cognitive scales, the object of research is nearly identical to that of MSLQ users. A specific feature of ASI and its variations is the idea to measure not only the desired learning behaviour (strategic approach) but also what is hypothesized as less success-oriented (apathetic approach). This produces a multifarious picture of learners’ behaviour.
Another well-known instrument to capture students learning strategies is the Learning and Study Strategies Inventory (LASSI) by Weinstein and Palmer (2002). LASSI covers thoughts, behaviours, attitudes and beliefs in relation to successful learning that can also be fostered by interventions. Its ten scales are classed into affective strategies, goal strategies and comprehension monitoring strategies, thus covering cognitive, metacognitive (particularly, self-regulative), affective and motivational aspects. LASSI is not only used for research purposes but is also recommended to students to use for themselves in order to get feedback on their strengths and weaknesses. LASSI’s reliability coefficients (Cronbach’s α) for its different scales are reported to score between 0.86 and 0.68, the lowest often being considered insufficient (Weinstein et al. 1987). Its validity to academic performance depends on the specific scale, e.g. Cano (2006) found, using multiple regression, that two scales (namely Affective Strategies and Goal Strategies) contributed to academic performance, whereas one (Comprehension Monitoring Strategies) did not.
All questionnaires described so far resort to self-assessment of student behaviour. It must be conceded that this entails the weakness that “the learner’s perceptions of his or her strategies are measured” (Schellings 2011, p. 94), which need not coincide precisely with the strategies themselves. In this context, it is interesting to compare self-reported learning behaviour (especially concerning metacognition) to the results gained with other methods, e.g. Thinking Aloud (David 2013). When it comes to affective aspects, however, the learner’s subjective perspective is what counts. Other problems, like assessing the sufficiency or efficiency of study time or effort, might be harder to overcome. In our context, beginner students might initially judge their efforts as sufficient (in relation to what they have been used to at school), but later rate them as inadequate (when their frame of reference has shifted after some months at university). This can result in a seeming decrease, although objectively, they have actually increased their efforts. Particularly in pre-post designs, this must be taken into account when interpreting results. Some authors thus favour using a retrospective pretest data design to measure programme effectiveness (Nimon et al. 2011; Lam and Bengo 2003) which is easily applicable to inventories measuring learning strategies.
LIST inventory
For our research at Ruhr-Universität Bochum in Germany, the decision fell for the German LIST questionnaire (Learning Strategies at University, Wild and Schiefele 1994), which is based on the same classification as MSLQ and takes up aspects from LASSI as well. LIST was invented for measuring learning strategies of medium generality, between learning styles and learning tactics (Wild 2000). The instrument distinguishes between cognitive, metacognitive and resource-related learning strategies and comprises dimensions of learning strategies grouped accordingly. This mirrors the acceptance of this taxonomy in the German-speaking community (Wild 2000). Just like its English predecessors, this approach originates from educational research and thus is not subject-specific. However, in STEM education, the instruments are frequently used to assess students’ learning behaviour on a general level while combining the results with subject-related measures (Lovelace and Brickman 2013). The LIST questionnaire for measuring learning strategies in academic studies was first compiled in the 1990s (Wild and Schiefele 1994) and has since been modified and tested several times. LIST has been applied in the context of many subjects, mathematics among them (cf. Liebendörfer et al. 2014, for an overview), with overall satisfying results with regard to reliability (Wild and Schiefele 1994 found Cronbach’s α between 0.64 for Metacognition and 0.90 for Attention) and validity (Wild 2000, 2005). In the following, we explore in detail how the LIST takes up scales and items from MSLQ and LASSI, in order to understand its origins and to illustrate its structure.
Apart from Motivation, the scales from LIST are derived directly from MSLQ, although the number of items varies. Some items in LIST are translations of MSLQ items. In addition to LIST, MSLQ seems very differentiated in terms of motivation, it sports six Motivation scales (Intrinsic Goal Orientation, Extrinsic Goal Orientation, Task Value, Control of Learning Beliefs, Self-Efficacy for Learning and Performance, Test Anxiety) comprising 31 items. LIST does not have items with the label Motivation as such, but LIST’s six items on Attention (which are all reverse coded) and eight items on Effort more or less cover this aspect, for example, “I work late at night or at the weekends if necessary”. And other LIST scales, in particular, the resource-related ones, are meant to measure the degree of motivation a student possesses when preparing for an important exam, with items like “I fix the hours I spend daily on learning in a schedule”. The main difference between the two questionnaires is that MSLQ puts more emphasis on including different aspects of motivation as Goal Orientation or Control of Learning Beliefs. For LIST, on the other hand, the aim was to clearly keep apart cognitive and motivational aspects.
Synoptic table for LIST’s roots in MSLQ and LASSI (item numbers)
LIST | MSLQ | LASSI |
---|---|---|
Critical Checks (8) | Critical Thinking (5) | |
Elaborating (8) | Elaboration (6) | Information Processing (8) |
Organizing (8) | Organization (4) | Selecting Main Ideas (8) |
Repeating (7) | Rehearsal (4) | |
Attention (6) | Motivation | Attitude (8) |
Effort (8) | Effort Regulation (4) | Concentration (8) |
Learning Environment (6) | Time / Study Environment Management (8) | |
Time Management (4) | Time Management (8) | |
Peer Learning (7) | Peer Learning (3) | |
Reference (4) | Help Seeking (4) | Study Aids (8) |
Planning (4) | Metacognitive Strategies (12) | Self-Testing (8) |
Monitoring (4) | ||
Regulating (3) | ||
… | … |
As the comparison of sub-scales and items of the three instruments reveals, there are ample reasons for aiming at compactifying and balancing LIST, particularly when planning to use it with large samples. One side of the coin is receiving a reliable and valid instrument to access learning strategies. However, the next thing to consider is that the distribution, filling in and collection of the LIST questionnaire in a lecture hall holding several hundred students lasts almost 30 min, precious time that both lecturers and students would rather spend on mathematics. In addition, many studies aim at assessing learning strategies in a pre-post design at least twice, meaning that the drop-out rate is a serious issue to consider in any study design. That is, if there were a possibility to reduce the time to 15 min, participants’ cooperation would be easier to gain.
Study perspectives and research questions
The purpose of the project MP^{2}-Math/Plus/Practice is to support engineering students in their first year at the university by enhancing their learning strategies and motivation. The length of MP^{2}-Math/Plus/Practice, which has completed its fifth year, permits to change the project interventions according to the principles of educational design research (cf. McKenney and Reeves 2012). The fact that MP^{2}-Math/Plus/Practice employs the LIST questionnaire in a pre-post design allows for conducting meta-analyses and prompted the study at hand which aims at exploring if the LIST questionnaire can be shortened while keeping the factor structure comparable to the original, and therefore its potential for describing learning strategies and significant changes therein:
(RQ 1): Do reduced LIST scales remain internally reliable both in terms of reliability measures and content?
(RQ 2): Are the reduced scales still sensitive enough to capture developments in a pre-post design?
The evaluations of the MP^{2}-Math/Plus/Practice interventions, in which the results gained from LIST play an important role, are not the focus of this paper. They will address more content-related and less technical issues, e.g. in what respect students modify their learning strategies and to what extent this can be influenced by the project interventions (for more details, see Griese et al. 2011). We restrict ourselves to refer to MP^{2}-Math/Plus/Practice only so far as information is needed to pursue the final goal of gaining a more workable version of LIST, making future research easier.
Methods
Participants of the study
LIST has been used for different groups of students from various backgrounds reading all sorts of subjects. Our research focuses on engineering students in their first semester at the university. Out of the students questioned, 77.70 % are males, 22.30 % females. The average age is M_{age} = 20.34 years (SD_{age} = 2.22 years). Almost one quarter (24.85 %) of the students have a first language different from German. Only 60.86 % of them attended an advanced course in mathematics at school, and no more than 57.55 % went to the preparation course in mathematics at the university prior to their engineering course. Moreover, 20.33 % of the students neither took part in an advanced mathematics course at school nor did they visit the preparation course at university. These last characteristics classify the academic qualifications of a substantial number of students for the chosen subject of engineering as insufficient.
Study design and instruments
In order to study the development of learning strategies, a pre-post design was decided upon, the post questionnaire consisting of identical items in retrospect, presented in the past tense. The questionnaires were distributed during the break or at the end of the mathematics lecture and collected on-site after completion. This lecture covers what is considered basic mathematics: linear algebra, differential and integral calculus. The approach is axiomatic, with an emphasis on calculation. On the whole, proofs are presented in the lectures but not tested in the examination. The pre-test took place around the second week of the regular lectures; the post-test around the second last week.
The modification of the original LIST used in the research at hand is minimal: the scale Critical Checks was eliminated as it did not seem appropriate for mathematics at the beginning of the university. The items referring to the use of reference works were reworded in order to relate to digital sources as well. A new starting item (I study for my courses) was introduced but not used for evaluation—and consequently not included in the following countings of items, which starts at 69 for the consequently named LIST69. Whereas MSLQ and LIST usually prefer five-point Likert scales, we opted for a four-point Likert scale (with poles very rarely and very often) in order to get more concise results.
Number of LIST69 questionnaires (60 or more items answered)
Pre | Post | |
---|---|---|
2011 / 2012 | 345 (231) | 103 (99) |
2012 / 2013 | 421 (376) | 254 (233) |
2013 / 2014 | 955 (780) | 296 (253) |
Sum | 1721 (1387) | 653 (585) |
Total | 2374 (1972) |
Process of data analysis
The objective of shortening LIST69 while keeping as much of the original scales as possible was conducted as follows, balancing consideration of content with statistical calculations: in a first step, the complete sample of data from 2011/2012 until 2013/2014 (N = 2374) was randomly split in half. The first half (N = 1187) was used to reduce the questionnaire by calculating Cronbach’s α values for each original scale, including how they would develop when single items were left out. Thus, a scale was shortened by eliminating the item whose deletion yielded the best alpha. This process was repeated as long as the alphas stayed good (>0.7), preferably until there were only half as many items as before, in order to produce a perceptible pruning. The shortened scales were tested on the second half of the sample (N = 1187), meaning their Cronbach’s α values were calculated as well. This complete process was repeated three times for three different random splits of the sample, thus enabling an assessment for the stability of the process, in order to minimize the risk of coincidental results. Three rounds also mean that in case of incongruent results, a tendency might be detected.
As a second step, the shortened version of LIST69 (named LIST45 as it contained 45 items) then underwent tests on a second half of the sample, i.e. a principal component analysis (PCA) to explore its structure, followed by some minor changes in item deletion and further PCA as well as confirmatory factor analysis (CFA). The results were compared to the results of PCA and CFA of LIST69 from before, including model fit. Slight modifications of factor descriptions and changes in the loadings of single items onto factors were expected. As the pre and post scores would be explored separately, Cronbach’s α values for single questionnaires were calculated from the complete sample as well as a further test for scale reliability.
The aim of researching the development of learning strategies demands further investigation into whether significant developments can be identified with the help of the shortened questionnaire. For this purpose, the item scores of each scale were combined using the formula 100/3 · [1/n · (x _{1} + … + x _{ n }) − 1] = 100/3 · (\( \overline{x} \) − 1), which renders values between 0 and 100 for a four-point Likert scale (n = number of items in the scale; scale scores under 25 describing rare use, between 25 and under 50 infrequent use, between 50 and under 75 regular use, 75 or more continual use of the learning strategies). Using LIST69 and the reduced data, the differences of means of these combined scores were calculated in dependent t-tests (listwise exclusion of cases, level of significance 0.05).
Results
The exploration of LIST69 identified the subscales of Metacognition (Planning, Monitoring and Regulating) as problematic, as all three had Cronbach’s α < 0.7 from the beginning. Thus, it was only possible to treat it as a complete scale where the eleven items rendered α = 0.728. PCA of LIST69 was conducted with orthogonal rotation (varix) and pairwise exclusion of cases on the complete sample. The chosen rotation takes into account that from the construction of LIST, the scales are expected to be unrelated. The Kaiser-Meyer-Olin (KMO) measures proved good, KMO_{LIST69} = 0.913, attesting the sample adequacy. All KMO values for individual items were greater than 0.76, which meets the accepted limit of 0.5. Bartlett’s test of sphericity χ ^{2} _{LIST69} (2346) = 37,916.098, p_{LIST69} = 0.000 indicated that correlations between items were sufficiently large for PCA. For LIST69, 13 components had eigenvalues over Kaiser’s criterion of 1 and in combination explained 50.59 % of variance. Traditionally, the LIST questionnaire is expected to render 10 scales (12 if Metacognition is regarded as having subscales Planning, Monitoring and Regulating), whereas in our case, Kaiser’s criterion of extracting as many components as have eigenvalues >1 renders more. Extraction of ten factors in LIST69 resulted in explaining 45.65 % of variance. All this hints that the intended factor structure should be reconsidered.
Cronbach’s α for LIST69 and shortened LIST45, N = # used data sets
Scale | LIST69 | LIST45 | ||||
---|---|---|---|---|---|---|
#items | α | N | #items | α | N | |
Organizing | 8 | .81 | 1605 | 3 | .74 | 1969 |
Elaborating | 8 | .77 | 1606 | 5 | .73 | 1816 |
Repeating | 7 | .73 | 1646 | 6 | .72 | 1702 |
Metacognition | 11 | .73 | 1513 | 9 | .71 | 1603 |
Effort | 8 | .76 | 1556 | 5 | .70 | 1751 |
Attention | 6 | .75 | 1663 | 3 | .82 | 1899 |
Time Management | 4 | .76 | 1901 | 3 | .71 | 1956 |
Learning Environment | 6 | .70 | 1721 | 5 | .72 | 1775 |
Peer Learning | 7 | .78 | 1664 | 3 | .77 | 1915 |
Using Reference | 4 | .77 | 1920 | 3 | .73 | 1971 |
Total | 69 | 45 |
However, some items remained problematic in both versions of the questionnaire. The Metacognition items did not form a separate scale. This is in keeping with the results of other research (cf. Wild 2000), and the items were therefore candidates for deletion. The items from the Metacognition subscale, Planning, consistently loaded on the same component as Time Management, prompting a fresh inspection of this new scale. This lead to further deletion of items following the procedure described above, which quickly revealed all four Planning items as obsolete. Repeating items did not satisfy either; the items loaded on different scales. Not surprisingly, the Monitoring item that mentions fellow students (monitoring 4, see Additional file 1) loads on Peer Learning.
Summary of PCA for LIST36, item loadings <0.4 suppressed, N = 1187
LIST items | Elaborating | Repeating | Effort | Learning Environment | Attention | Peer Learning | Time Management | Using Reference | Organizing |
---|---|---|---|---|---|---|---|---|---|
ela2 | 0.74 | ||||||||
ela4 | 0.71 | ||||||||
ela6 | 0.68 | ||||||||
ela3 | 0.63 | ||||||||
ela5 | 0.56 | ||||||||
repeating7 | 0.70 | ||||||||
repeating6 | 0.65 | ||||||||
repeating3 | 0.60 | ||||||||
repeating4 | 0.56 | ||||||||
repeating2 | 0.48 | ||||||||
repeating1 | 0.47 | ||||||||
effort3 | 0.66 | ||||||||
effort2 | 0.64 | ||||||||
effort8 | 0.61 | ||||||||
effort6 | 0.56 | ||||||||
effort1 | 0.53 | ||||||||
environ5 | 0.75 | ||||||||
environ6 | 0.75 | ||||||||
environ1 | 0.57 | ||||||||
environ4 | 0.56 | ||||||||
environ2 | 0.55 | ||||||||
attention6 | 0.87 | ||||||||
attention5 | 0.85 | ||||||||
attention4 | 0.84 | ||||||||
peer1 | 0.84 | ||||||||
peer2 | 0.83 | ||||||||
peer6 | 0.76 | ||||||||
time3 | 0.79 | ||||||||
time4 | 0.77 | ||||||||
time1 | 0.72 | ||||||||
reference4 | 0.75 | ||||||||
reference1 | 0.73 | ||||||||
reference3 | 0.72 | ||||||||
orga2 | 0.76 | ||||||||
orga5 | 0.74 | ||||||||
orga3 | 0.69 | ||||||||
Percentage of variance | 16.52 | 7.20 | 6.38 | 5.75 | 5.38 | 4.53 | 3.86 | 3.46 | 3.08 |
Cronbach’s α | 0.72 | 0.72 | 0.71 | 0.73 | 0.82 | 0.77 | 0.72 | 0.73 | 0.72 |
All in all, the deletions show slight variations to our first attempt at refining the LIST questionnaire (cf. Griese et al. 2014), when we achieved a version with 32 items and slightly worse alphas, using data from earlier project years. As a side project, oblique rotation (direct oblimin) was also tried: the results stay the same, meaning that the same items load on the same factors as strongly as in orthogonal rotation, the same amount of variance is explained; and the structure matrix does not reveal any disturbing interconnections between the components.
Fit indices for CFA for original and shortened questionnaires
CFI | RMSEA | SRMR | χ ^{2}(df) | p | |
---|---|---|---|---|---|
LIST69 | 0.8001 | 0.0420 | 0.0523 | 11464.87(2211) | 0.00 |
LIST45 | 0.8419 | 0.0448 | 0.0466 | 5191.93(900) | 0.00 |
LIST36 | 0.8364 | 0.0556 | 0.0495 | 4644.44(558) | 0.00 |
Cronbach’s α of LIST36 for the individual surveys
Pre 2011/12 | Post 2011/12 | Pre 2012/13 | Post 2012/13 | Pre 2013/14 | Post 2013/14 | |
---|---|---|---|---|---|---|
Organizing | 0.72 | 0.77 | 0.71 | 0.77 | 0.70 | 0.79 |
Elaborating | 0.67 | 0.74 | 0.75 | 0.72 | 0.67 | 0.74 |
Repeating | 0.69 | 0.78 | 0.65 | 0.72 | 0.71 | 0.69 |
Effort | 0.68 | 0.76 | 0.70 | 0.70 | 0.70 | 0.62 |
Attention | 0.76 | 0.89 | 0.80 | 0.75 | 0.81 | 0.85 |
Time Management | 0.63 | 0.66 | 0.73 | 0.70 | 0.69 | 0.77 |
Learning Environment | 0.71 | 0.72 | 0.74 | 0.77 | 0.71 | 0.72 |
Peer Learning | 0.77 | 0.78 | 0.79 | 0.72 | 0.74 | 0.82 |
Using Reference | 0.66 | 0.73 | 0.76 | 0.70 | 0.71 | 0.78 |
Overview of content-related aspects of item deletions
Scale | Summary of content-related aspects of item deletion |
---|---|
Organizing | The deletions are acceptable (e.g. the aspect of structuring subject matter to make it easier to remember is contained in orga2 and orga3). However, compiling one’s own lists is not included anymore—but writing your own summaries is |
Elaborating | The deletions seem particularly appropriate for engineering students, who are either explicitly taught the connections to their other subjects and courses, or they do not stand a chance to discover them on their own |
Repeating | The deletion only concerns one item rather referring to texts (reciting paragraphs) which does not fit to mathematics very well |
Metacognition | The deletions can be regarded as critical, as here we find aspects that are neither covered by other scales nor is their depth reflected anywhere else, see text for details |
Effort | This scale does not lose its focus on perseverance but a more concrete manifestation of this attitude, namely working late and long hours. Comparing one’s learning time with others’, though, measures not only objective effort but also subjective self-confidence and seems problematic in terms of multidimensionality |
Attention | The six original items are paraphrases of each other, so that deleting half of them does not result in any loss of information, as the high Cronbach’s α for this scale shows (cf. Table 6) |
Time Management, Using Reference | The deleted items are mirrored in those kept |
Learning Environment, Peer Learning | The deletions seem unexciting. The deleted item about always sitting at the same place when studying does not match students’ reality well, as most have places of study at home and meet fellow students on campus. The same applies to comparing notes with fellow students in times of digital uploads |
For the metacognitive scales Monitoring and Regulating, at a closer look, the aspects inquired for are mostly mirrored in Effort, though more superficially. One item in Monitoring (monitoring 4, see Additional file 1) can well be associated with Peer Learning, as it contains interacting with fellow students as much as it contains testing oneself. The metacognitive scale Planning, whose close connection to Time Management was obvious from the data analysis, nevertheless stresses substantial planning rather than simple timing. Its deletion means losing a more effective and more reflected learning strategy. It is doubtful, though, if techniques of this depth can be recorded well in a quick and self-reporting questionnaire. All in all it can be said that deleting the metacognitive scales must be regarded critically.
Discussion
Investigations into an approved and validated questionnaire originally designed to explore learning strategies in general have led to a shorter version consisting of only 36 of the original 69 items, which will make it easier to ensure students’ cooperation for further research. LIST36 was developed using the data from research into STEM students’ learning strategies and mirrors nine out of the ten scales in LIST69 and has proven its reliability and its factors explain more variance. In addition, the shortened questionnaire appears as suitable for use in pre-post designs as its predecessor. If the differences in pre and post scores are a valid measurement for competence, development at all has to be viewed critically, though, as individual reference frames may have shifted. Furthermore, the validity in connection with academic performance will have to be further explored in our future research. Another point is that the generalizability of the new instrument will need to be tested on other populations than engineering students, as some of the content-related issues might be weighted differently for different groups of students.
Leaving aside the numerous calculations, it is nonetheless important to take a close look at which items have suffered deletion, if they contain any specific content that might get lost in the scale or if they involve a shift in content focus. For our purposes, we can keep in mind that we are dealing with first year engineering students and their learning of mathematics, which on the one hand makes some deletions acceptable but which, on the other hand, also represents a limitation. It still needs to be tested if the findings from our sample of first year engineering students can be generalized.
To avoid problems stemming from inter-language or inter-cultural differences, it seemed wise to use a German questionnaire in our research, although a transfer to another country and language can present a test for a questionnaire’s validity. International cooperation has been effected with colleagues from Spain (cf. Gómez-Chacón et al. 2015, accepted for CERME9), showing that LIST keeps its qualities when being used in another country: After being translated into English and then into Spanish, the cognitive and metacognitive scales from LIST kept their reliability, an indication for the questionnaire’s universal applicability.
In addition to this, it has become apparent that some of our initial assumptions about the impact and measurability of metacognitive behaviour of STEM students might be incorrect, thus guiding further research towards other ways of examining this concept. This aspect represents a point that might interest researchers from other areas, as metacognition and its measurability often are popular research interests.
Conclusions
Depending on the purpose of the individual study, it might still be advisable to keep the Metacognition items in the questionnaire for further tests as the decision to consider deleting all of them was a hard one. Our research into the interventions of the project MP^{2}-Math/Plus/Practice itself had been planned to keep a close eye on metacognitive learning strategies, hypothesizing that they make the difference when it comes to successful learning behaviour. However, the results of our analysis show that the characteristics formerly believed to be separate from more technical learning strategies actually load on already existing factors, and there they do not even contribute relevantly. One logical conclusion might be that metacognitive dispositions are latent factors which work in the background and become apparent only in other learning strategies—where they can then be measured accordingly. Another explanation could lie in the nature of the data acquisition used: metacognitive reflections might be inappropriate for being measured in self-assessing printed questionnaires. Other forms of procuring data, for example, from interviews, videographed peer discussions, classroom observations, flash interviews, reflective essays or learning diaries, might prove more adequate. Further research is required, and our data collected from interviews, learning diaries and other sources will be used for this purpose.
Declarations
Authors’ Affiliations
References
- Artigue, M, Batanero, C, & Kent, P. (2007). Mathematics thinking and learning at post-secondary level. In F Lester (Ed.), Second handbook of research on mathematics teaching and learning (pp. 1011–1049). Greenwich, Connecticut: Information Age Publishing.Google Scholar
- Artino, AR (2005). A review of the motivated strategies for learning questionnaire. http://eric.ed.gov/?id=ED499083. Accessed 1 December 2014.
- Blömeke, S, Gustafsson, JE, & Shavelson, RJ. (2015). Beyond dichotomies. Competence viewed as a continuum. Zeitschrift für Psychologie, 223(1), 3–13.View ArticleGoogle Scholar
- Cano, F. (2006). An in-depth analysis of the learning and study strategies inventory (LASSI). Educational and Psychological Measurement, 66(6), 1023–1038.View ArticleGoogle Scholar
- David, A (2013). Aufgabenspezifische Messung metakognitiver Aktivitäten im Rahmen von Lernaufgaben (PhD thesis). Technische Universität, Chemnitz. http://nbn-resolving.de/urn:nbn:de:bsz:ch1-qucosa-133054. Accessed 7 April 2015.
- de Guzmán, M, Hodgson, BR, Robert, A, & Villani, V. (1998). Difficulties in the passage from secondary to tertiary education. In G Fischer & U Rehmann (Eds.), Proceedings of the international congress of mathematicians—documenta mathematica extra volume III (pp. 747–762). Rosenheim: Geronimo GmbH.Google Scholar
- Dehling, H, Glasmachers, E, Griese, B, Härterich, J, & Kallweit, M. (2014). MP^{2}-Mathe/Plus/Praxis, strategien zur vorbeugung gegen studienabbruch. Zeitschrift für Hochschulentwicklung, 9(4), 39–56.Google Scholar
- Dreyfus, T (Guest Editor) (1995). Advanced mathematical thinking [special issue]. Educational Studies in Mathematics, 29(2).Google Scholar
- Duncan, TG, & McKeachie, WJ. (2005). The making of the motivated strategies for learning questionnaire. Educational Psychologist, 40, 117–128.View ArticleGoogle Scholar
- Dunn, KE, Lo, WJ, Mulvenon, SW, & Sutcliffe, R. (2012). Revisiting the motivated strategies for learning questionnaire: a theoretical and statistical reevaluation of the metacognitive self-regulation and effort regulation subscales. Educational and Psychological Measurement, 72, 312–331.View ArticleGoogle Scholar
- Entwistle, N, & McCune, V. (2004). The conceptual bases of study strategies inventories. Educational Psychology Review, 16(4), 325–346.View ArticleGoogle Scholar
- Entwistle, N, & Ramsden, P. (1983). Understanding student learning. London: Croom Helm.Google Scholar
- Erdem Keklik, E, & Keklik, I. (2013). Motivation and learning strategies as predictors of high school students’ math achievement. Cukurova University Faculty of Education Journal, 42(1), 96–109.Google Scholar
- Evans, JSBT. (2007). Hypothetical thinking. Dual processes in reasoning and judgement. Hove: Psychology Press.Google Scholar
- Field, A. (2009). Discovering statistics using SPSS (3rd ed.). Thousand Oaks: Sage Publications.Google Scholar
- Fischbein, E. (1987). Intuition in science and mathematics. An educational approach. Dordrecht: Kluwer Academic Publishers.Google Scholar
- Gómez-Chacón, IM, García-Madruga, JA, Vila, JO, Elosúa, MR, & Rodríguez, R. (2014). The dual processes hypothesis in mathematics performance: beliefs, cognitive reflection, reasoning and working memory. Learning and Individual Differences, 29, 67–73.View ArticleGoogle Scholar
- Gómez-Chacón, IM, Griese, B, Rösken-Winter, B, & Gonzàlez-Guillén, C. (2015). Engineering students in Spain and Germany—varying and uniform learning strategies. In N Vondrova & K Krainer (Eds.), Proceedings of the 9th conference of european researchers in mathematics education. Prague, Czech Republic: Charles University. in press.Google Scholar
- Griese, B, Glasmachers, E, Kallweit, M, & Roesken, B. (2011). Engineering students and their learning of mathematics. In B Roesken & M Casper (Eds.), Current state of research on mathematical beliefs XVII. Proceedings of the MAVI-17 conference (pp. 85–96). Bochum: Professional School of Education, RUB.Google Scholar
- Griese, B, Lehmann, M, & Roesken-Winter, B. (2014). Exploring university students’ learning strategies in mathematics: refining the LIST questionnaire. In S Oesterle, C Nicol, & P Liljedahl (Eds.), Proceedings of the joint meeting of PME 38 and PME-NA 36, Vol. 6 (p. 131). Vancouver, Canada: PME.Google Scholar
- Griese, B, Roesken-Winter, B, Kallweit, M, & Glasmachers, E. (2013). Redesigning interventions for engineering students: learning from practice. In AM Lindmeier & A Heinze (Eds.), Proceedings of the 37th conference of the international group for the psychology of mathematics education vol. 5 (p. 65). Kiel, Germany: PME.Google Scholar
- Gueudet, G. (2008). Investigating the secondary–tertiary transition. Educational Studies in Mathematics, 67(3), 237–254.View ArticleGoogle Scholar
- Henn, HW, Bruder, R, Elschenbroich, J, Greefrath, G, Kramer, J, & Pinkernell, G. (2010). Schnittstelle schule - hochschule. In A Lindmeier & S Ufer (Eds.), Beiträge zum mathematikunterricht 2010 (pp. 75–82). Münster: WTM.Google Scholar
- Heublein, U, Richter, J, Schmelzer, R, & Sommer, D. (2012). Die entwicklung der schwund- und studienabbruchquoten an den deutschen hochschulen: statistische berechnungen auf der basis des absolventenjahrgangs 2010. Forum hochschule: vol. 2012,3. Hannover: HIS.Google Scholar
- Hilpert, J, Stempien, J, van der Hoeven Kraft, KJ, & Husman, J. (2013). Evidence for the latent factor structure of the MSLQ: a new conceptualization of an established questionnaire. SAGE-Open, 3, 1–10. doi:10.1177/2158244013510305.View ArticleGoogle Scholar
- Lam, TC, & Bengo, P. (2003). A comparison of three retrospective self-reporting methods of measuring change in instructional practice. American Journal of Evaluation, 223(1), 3–13.Google Scholar
- Liebendörfer, M, Hochmuth, R, Schreiber, S, Göller, R, Kolter, J, Biehler, R, Kortemeyer, J, & Ostsieker, L. (2014). Vorstellung eines fragebogens zur erfassung von lernstrategien in mathematikhaltigen studiengängen. In J Roth & J Ames (Eds.), Beiträge zum mathematikunterricht 2014 (pp. 739–742). Münster: WTM.Google Scholar
- Lovelace, M, & Brickman, P. (2013). Best practices for measuring students’ attitudes toward learning science. CBE – Life Sciences Education, 12, 606–617.Google Scholar
- McKenney, SE, & Reeves, TC. (2012). Conducting educational design research. New York: Routledge.Google Scholar
- McLeod, DB. (1992). Research on affect in mathematics education: a reconceptualization. In DA Grouws (Ed.), Handbook of research on mathematics teaching and learning (pp. 575–596). New York: Macmillan.Google Scholar
- Nenniger, P. (1999). On the role of motivation in self-directed learning. The ‘two-shells-model of motivated self-directed learning’ as a structural explanatory concept. European Journal of Psychology of Education, 14(1), 71–86.View ArticleGoogle Scholar
- Nimon, K, Zigarni, D, & Allen, J. (2011). Measures of program effectiveness based on retrospective pretest data: are all created equal? American Journal of Evaluation, 32(1), 8–28.View ArticleGoogle Scholar
- Pintrich, P, & de Groot, E. (1990). Motivational and self-regulated learning components of classroom academic performance. Journal of Educational Psychology, 81(1), 33–40.View ArticleGoogle Scholar
- Pintrich, P, Smith, D, Garcia, T, & McKeachie, W. (1993). Reliability and predictive validity of the motivated strategies for learning questionnaire (MSLQ). Educational and Psychological Measurement, 53(3), 801–813.View ArticleGoogle Scholar
- Rach, S, & Heinze, A. (2011). Studying mathematics at the university: the influence of learning strategies. In B Ubuz (Ed.), Proceedings of the 35th conference of the international group for the psychology of mathematics education (Vol. 4, pp. 9–16). Ankara, Turkey: PME.Google Scholar
- Schellings, C. (2011). Applying learning strategy questionnaires: problems and possibilities. Metacognition and Learning, 6(2), 91–109.View ArticleGoogle Scholar
- Schreiber, JB, Nora, A, Stage, FK, Barlow, EA, & King, J. (2006). Reporting structural equation modeling and confirmatory factor analysis results: a review. The Journal of Educational Research, 99(6), 323–337.View ArticleGoogle Scholar
- Selden, A, & Selden, J. (2005). Perspectives on advanced mathematical thinking. Mathematical Thinking and Learning, 7(1), 1–13.View ArticleGoogle Scholar
- Stavy, R, & Tirosh, D. (2000). How students (mis-)understand science and mathematics: intuitive rules. New York: Teachers College Press.Google Scholar
- Tait, H, Entwistle, N, & McCune, V. (1998). ASSIST: a reconceptualisation of the approaches to studying inventory. In C Rust (Ed.), Improving student learning: improving students as learners (pp. 262–271). Oxford: Oxford Centre for Staff and Learning Development.Google Scholar
- Tall, DO. (2004). Building theories: the three worlds of mathematics. For the Learning of Mathematics, 24(1), 29–33.Google Scholar
- Weinstein, CE, & Palmer, DR. (2002). Learning and study strategies inventory (LASSI): user’s manual (2nd ed.). Clearwater: H & H Publishing.Google Scholar
- Weinstein, CE, Schulte, AC, & Palmer, DR. (1987). Learning and study strategies inventory (LASSI). Clearwater: H & H Publishing.Google Scholar
- Wild, KP. (2000). Lernstrategien im studium: strukturen und bedingungen. Münster: Waxmann.Google Scholar
- Wild, KP. (2005). Individuelle lernstrategien von studierenden. Konsequenzen für die hochschuldidaktik und die hochschullehre. Beiträge zur Lehrerbildung, 23(2), 191–206.Google Scholar
- Wild, KP, & Schiefele, U. (1994). Lernstrategien im studium. Ergebnisse zur Faktorenstruktur und Reliabilität eines neuen Fragebogens. Zeitschrift für Differentielle und Diagnostische Psychologie, 15, 185–200.Google Scholar
Copyright
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.