thesis learning styles

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Recommended Articles
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

The relationship between learning styles and academic performance: consistency among multiple assessment methods in psychology and education students.

1. Introduction

1.1. learning styles: experiential learning theory, 1.2. assessment method, academic performance and learning dimensions, 1.3. present study, 2. materials and methods, 2.1. participants, 2.2. instruments.

• Multiple choice questions (MCQ). The contents were evaluated through multiple choice tests. Using this closed question method, students had to identify a single valid response among four alternatives. The following is an example of an MCQ used in the assessment: “The increase in explicit memory in children is associated with: (a) the autonomy of the child when learning to walk and handle objects; (b) an increase in the density of synapses at four months; (c) an increase in the density of synapses at eight months (correct answer); (d) the importance of holophrases in children to store memory”. A total of 30 MCQ were used with a final score of 0 to 10.
• Short questions (SQ). This method used short and closed questions. In this test, the student is given a statement in order to identify a concept. An example is given below: “Identify the developmental theory that states that the zone of proximal development refers to the distance between the actual level of development and the level of potential development”. In this case, the correct answer would be: Vygotsky’s Sociocultural Theory. A total of 10 SQ were used with a final score ranging from 0 to 10.
• Creation-elaboration questions (CEQ). In this open-ended question, students must create a practical activity based on the contents studied in the developmental psychology course. This method involves mainly a practical question. The CEQ used in the tests was: “Create an activity to work on the understanding other people’s emotions in a class of 5-year-olds”. With a score between 0 and 10 points, the structure of the activity, creativity, and suitability with the contents of the course were used as criteria to evaluate performance on the CEQ.
• Elaboration Questions on the Relationship between Theory and Practice (EQRTP). In this open-ended assessment question, students must link theoretical concepts to practical application. A video is used to illustrate a teaching-learning process between children and an expert, and various interactions between children. After watching the video, students must analyze its content using theoretical concepts. Below is the examination question and a link to the video used ( https://www.dropbox.com/s/xux6p9di5pj885m/Sustainability.mp4?dl=0 ) (accessed on 29 January 2021) [ 48 ] “Associate the following video with theoretical contents of the Developmental Psychology course”. The number of associations between practical aspects and theoretical contents, coherence between concepts, presentation of theory and clarity in the written composition were used as criteria to assess performance from 0 to 10.

2.3. Procedure

2.4. data analysis, 3.1. learning styles according to personal and educational dimensions, 3.2. assessment methods: academic performance and consistency and their relation to learning dimensions and styles, 3.3. influence of learning dimensions on performance consistency in the different assessment methods, 4. discussion, 5. conclusions, author contributions, institutional review board statement, informed consent statement, acknowledgments, conflicts of interest.

• Cardoso, S.; Rosa, M.J.; Stensaker, B. Why is quality in higher education not achieved? The view of academics. Assess. Eval. High. Educ. 2016 , 41 , 950–965. [ Google Scholar ] [ CrossRef ]
• Kolb, A.Y.; Kolb, D.A. The Kolb Learning Style Inventory 4.0: A Comprehensive Guide to the Theory, Psychometrics, Research on Validity and Educational Applications ; Hay Resources Direct: Boston, MA, USA, 2013. [ Google Scholar ]
• Tormo-Carbó, G.; Seguí-Mas, E.; Oltra, V. Business ethics as a sustainability challenge: Higher education implications. Sustainability 2018 , 10 , 2717. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Ghazivakili, Z.; Norouzi Nia, R.; Panahi, F.; Karimi, M.; Gholsorkhi, H.; Ahmadi, Z. The role of critical thinking skills and learning styles of university students in their academic performance. J. Adv. Med. Educ. Prof. 2014 , 2 , 95–102. [ Google Scholar ]
• Hendry, G.D.; Heinrich, P.; Lyon, P.M.; Barratt, A.L.; Simpson, J.M.; Hyde, S.J.; Gonsalkorale, S.; Hyde, M.; Mgaieth, S. Helping students understand their learning styles: Effects on study self-efficacy, preference for group work, and group climate. Educ. Psychol. 2005 , 25 , 395–407. [ Google Scholar ] [ CrossRef ]
• Deem, R.; Baird, J.A. The English Teaching Excellence (and Student Outcomes) Framework: Intelligent accountability in higher education? J. Educ. Chang. 2020 , 21 , 215–243. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• United Nations. Transforming Our World: The 2030 Agenda for Sustainable Development ; United Nations Population Fund: New York, NY, USA, 2015. [ Google Scholar ]
• Watson, M.K.; Pelkey, J.; Noyes, C.; Rodgers, M.O. Using Kolb’s Learning Cycle to Improve Student Sustainability Knowledge. Sustainability 2019 , 11 , 4602. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Alonso, C.; Gallego, D.; Honey, P. Los Estilos de Aprendizaje. Procedimientos de Diagnóstico y Mejora ; Mensajero: Bilbao, Spain, 1997. [ Google Scholar ]
• Kalleberg, A.L.; Dunn, M. Institutional Determinants of Labor Market Outcomes for Community College Students in North Carolina. Community Coll. Rev. 2015 , 43 , 224–244. [ Google Scholar ] [ CrossRef ]
• Kolb, D.A. Experiential Learning: Experience as The Source of Learning and Development. Prentice Hall Inc. 1984 , 20–38. [ Google Scholar ] [ CrossRef ]
• White, R.K.; Jantrania, A. Computer program design for land treatment systems. Biocycle 1989 , 30 , 66–67. [ Google Scholar ]
• Keefe, J.W. Profiling and Utilizing Learning Style ; NASSP Learning Style Series; NASSP: Reston, VA, USA, 1988. [ Google Scholar ]
• Myers, I.B. MBTI Manual: A Guide to the Development and Use of the Myers-Briggs Type Indicator ; Consulting Psychologists Press: Palo Alto, CA, USA, 1998. [ Google Scholar ]
• Kolb, A.Y.; Kolb, D.A. Learning styles and learning spaces: Enhancing experiential learning in higher education. Acad. Manag. Learn. Educ. 2005 , 4 , 193–212. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Sadler-Smith, E. Does the Learning Styles Questionnaire Measure Style or Process? a Reply to Swailes and Senior (1999). Int. J. Sel. Assess. 2001 , 9 , 207–214. [ Google Scholar ] [ CrossRef ]
• Kolb, A.Y.; Kolb, D.A. The Kolb Learning Style Inventory—Version 3. 1 2005 Technical Specifications ; Hay Resources Direct: Boston, MA, USA, 2005. [ Google Scholar ]
• Romero, L.N.; Salinas, V.; Mortera, F.J. Estilos de aprendizaje basados en el modelo de Kolb en la educación virtual. Apert. Rev. Innov. Educ. 2010 , 2 , 72–85. [ Google Scholar ]
• Sabariego, M.; Regos, R.A.; Villaverde, V.A.; José, M.; Tarazona, C.; Benito, V.D.; Serrano, L.R.; Dolores, M.; González, S. El Pensamiento Reflexivo a Través De Las Metodologías Narrativas: Experiencias de Innovación en Educación Superior ; Octaedro: Barcelona, Spain, 2018. [ Google Scholar ]
• Rodríguez, H.; Pirul, J.; Robles, J.; Pérez, L.; Vásquez, E.; Galaz, I.; Cuellar, C.; Díaz, H.; Arriaza, C. Analysis of learning styles in Medical students in the University of Chile. Educ. Med. 2018 , 19 , 2–8. [ Google Scholar ] [ CrossRef ]
• Bahamón, M.; Viancha, M.; Alarcón, L.; Bohorquez Olaya, C.I. Estilos y estrategias de aprendizaje: Una revisión empírica y conceptual de los últimos diez años. Pensam. Psicol. 2012 , 10 , 129–144. [ Google Scholar ]
• Gargallo, B.; Almerich, G.; Suárez, J.M.; García, E.; Garfella, P.R. Learning styles and approaches to learning in excellent and average first-year university students. Eur. J. Psychol. Educ. 2013 , 28 , 1361–1379. [ Google Scholar ] [ CrossRef ]
• Gravini, M.; Cabrera, E.; Ávila, V.; Vargás, I. Estrategias de enseñanza en docentes y estilos de aprendizaje en estudiantes del programa de psicología de la Universidad Simón Bolívar, Barranquilla. Rev. Estilos Aprendiz 2009 , 3 , 124–140. [ Google Scholar ]
• İlçin, M.; Tomruk, N.; Yeşilyaprak, S.S.; Karadibak, D.; Savcı, S. The relationship between learning styles and academic performance in Turkish physiotherapy students. Physiotherapy 2016 , 102 , e84–e85. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Brown, T.; Cosgriff, T.; French, G. Learning style preferences of occupational therapy, physiotherapy and speech pathology students: A comparative study. Internet J. Allied Health Sci. Pract. 2008 , 6 , 7. [ Google Scholar ]
• D’Amore, A.; James, S.; Mitchell, E.K.L. Learning styles of first-year undergraduate nursing and midwifery students: A cross-sectional survey utilising the Kolb Learning Style Inventory. Nurse Educ. Today 2012 , 32 , 506–515. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Shein, P.P.; Chiou, W. Bin Teachers as role models for students’ learning styles. Soc. Behav. Pers. 2011 , 39 , 1097–1104. [ Google Scholar ] [ CrossRef ]
• Steinmayr, R.; Meißner, A.; Weidinger, A.F.; Wirthwein, L. Academic Achievement. Oxford Bibliographie ; Oxford University Press: New York, NY, USA, 2015. [ Google Scholar ]
• Ackerman, D.S.; Hu, J. Effect of type of curriculum on educational outcomes and motivation among marketing students with different learning styles. J. Mark. Educ. 2011 , 33 , 273–284. [ Google Scholar ] [ CrossRef ]
• Urda, J.; Ramocki, S.P. Assessing students’ performance by measured patterns of perceived strengths: Does preference make a difference? Assess. Eval. High. Educ. 2015 , 40 , 33–44. [ Google Scholar ] [ CrossRef ]
• Pham, H.; Trigg, M.; Wu, S.; O’Connell, A.; Harry, C.; Barnard, J.; Devitt, P. Choosing medical assessments: Does the multiple-choice question make the grade? Educ. Health Chang. Learn. Pract. 2018 , 31 , 65–71. [ Google Scholar ] [ CrossRef ]
• Schuwirth, L.W.T.; Van Der Vleuten, C.P.M.; Donkers, H.H.L.M. A closer look at cueing effects in multiple-choice questions. Med. Educ. 1996 , 30 , 44–49. [ Google Scholar ] [ CrossRef ]
• Scouller, K. The influence of assessment method on students’ learning approaches: Multiple choice question examination versus assignment essay. High. Educ. 1998 , 35 , 453–472. [ Google Scholar ] [ CrossRef ]
• Palmer, E.J.; Duggan, P.; Devitt, P.G.; Russell, R. The modified essay question: Its exit from the exit examination. Med. Teach. 2010 , 32 . [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Struyven, K.; Dochy, F.; Janssens, S. Students’ perceptions about evaluation and assessment in higher education: A review. Assess. Eval. High. Educ. 2005 , 30 , 325–341. [ Google Scholar ] [ CrossRef ]
• Bleske-Recheka, A.; Zeuga, N.; Webbb, R.M. Discrepant performance on multiplechoice and short answer assessments and the relation of performance to general scholastic aptitude. Assess. Eval. High. Educ. 2007 , 32 , 89–105. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Hogan, T. Constructed-response approaches for classroom assessment. In Sage Handbook of Research on Classroom Assessment ; McMillan, J., Ed.; Sage: Thousand Oaks, CA, USA, 2013; pp. 275–293. [ Google Scholar ]
• Bridgeman, B.; Morgan, R. Success in College for Students with Discrepancies between Performance on Multiple-Choice and Essay Tests. J. Educ. Psychol. 1996 , 88 , 333–340. [ Google Scholar ] [ CrossRef ]
• Gilar-Corbi, R.; Pozo-Rico, T.; Castejón, J.L.; Sánchez, T.; Sandoval-Palis, I.; Vidal, J. Academic achievement and failure in university studies: Motivational and emotional factors. Sustainability 2020 , 12 , 9798. [ Google Scholar ] [ CrossRef ]
• Schneider, M.; Preckel, F. Variables associated with achievement in higher education: A systematic review of meta-analyses. Psychol. Bull. 2017 , 143 , 565–600. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Yazici, H.J. Role of learning style preferences and interactive response systems on student learning outcomes. Int. J. Inf. Oper. Manag. Educ. 2016 , 6 , 109. [ Google Scholar ] [ CrossRef ]
• Lynch, T.G.; Woelfl, N.N.; Steele, D.J.; Hanssen, C.S. Learning style influences student examination performance. Am. J. Surg. 1998 , 176 , 62–66. [ Google Scholar ] [ CrossRef ]
• Rutz, E. Learning Styles and Educational Performance: Implications for Professional Development Programs. In Proceedings of the CIEC Conference, Tucson, AZ, USA, 28–31 January 2003. [ Google Scholar ]
• Camarero, F.J.; de Martín, F.A.; Herrero, F.J. Estilos y estrategias de aprendizaje en estudiantes universitarios. Psicothema 2000 , 12 , 615–622. [ Google Scholar ]
• Oughton, J.M.; Reed, W.M. The effect of hypermedia knowledge and learning style on student—centered concept maps about hypermedia. J. Res. Comput. Educ. 2000 , 32 , 366–384. [ Google Scholar ] [ CrossRef ]
• Cornwell, J.M.; Manfredo, P.A. Kolb’S Learning Style Theory Revisited. Educ. Psychol. Meas. 1994 , 54 , 317–327. [ Google Scholar ] [ CrossRef ]
• Socarrás, V.; Donat, R.; Fornons, D.; Vaque, C.; Milà, R. Estilos de aprendizaje identificados según el modelo VARK y el cuestionario de Kolb: Implicación en la Educación para la Salud. Cult. Educ. Soc. 2015 , 6 , 63–76. [ Google Scholar ]
• La Vida Secreta de los Niños: ¡Es el cumpleaños de Eneko! [Video]. Available online: https://www.youtube.com/watch?v=Y3NSTCosyqA (accessed on 11 October 2018).
• Maya, J.; Maraver, J. Teaching-learning processes: Application of educational psychodrama in the university setting. Int. J. Environ. Res. Public Health 2020 , 17 , 3922. [ Google Scholar ] [ CrossRef ]
• IBM Corp. IBM SPSS Statistics for Windows, Version 26.0 ; IBM Corp.: Armonk, NY, USA, 2019. [ Google Scholar ]
• Bhattacharyya, E. Learning Style and Its Impact in Higher Education and Human Capital Needs. Procedia Soc. Behav. Sci. 2014 , 123 , 485–494. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Newton, P.M.; Miah, M. Evidence-based higher education—Is the learning styles «myth» important? Front. Psychol. 2017 , 8 , 444. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Dryjanska, L.; Giua, M. Diversified path to the psychological career: Europe vs. USA. Int. Psychol. Bull. 2016 , 20 , 51–56. [ Google Scholar ]
• Yonker, J.E. The relationship of deep and surface study approaches on factual and applied test-bank multiple-choice question performance. Assess. Eval. High. Educ. 2011 , 36 , 673–686. [ Google Scholar ] [ CrossRef ]
• Bataineh, Z.M. Comparison of Students’ Performance in Multiple-Choice vs. Short Answer Question Formats in Anatomy Spotter Exams. FASEB J. 2019 , 33 , 439.3. [ Google Scholar ] [ CrossRef ]
• Li, C.; Zhou, H. Enhancing the Efficiency of Massive Online Learning by Integrating Intelligent Analysis into MOOCs with an Application to Education of Sustainability. Sustainability 2018 , 10 , 468. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Roediger, H.L.; Karpicke, J.D. The Power of Testing Memory: Basic Research and Implications for Educational Practice. Perspect. Psychol. Sci. J. Assoc. Psychol. Sci. 2006 , 1 , 181–210. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Sáinz, M.; Martínez-Cantos, J.L.; Rodó-de-Zárate, M.; Romano, M.J.; Arroyo, L.; Fàbregues, S. Young Spanish People’s Gendered Representations of People Working in STEM. A Qualitative Study. Front. Psychol. 2019 , 10 , 996. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• González-Zamar, M.D.; Abad-Segura, E. Implications of Virtual Reality in Arts Education: Research Analysis in the Context of Higher Education. Educ. Sci. 2020 , 10 , 225. [ Google Scholar ] [ CrossRef ]
• Magulod, G. Learning styles, study habits and academic performance of Filipino University students in applied science courses: Implications for instruction. J. Tech. Sci. Educ. 2019 , 9 , 2. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Wang, R.; Lowe, R.; Newton, S.; Kocaturk, T. Task complexity and learning styles in situated virtual learning environments for construction higher education. Autom. Constr. 2020 , 113 , 103148. [ Google Scholar ] [ CrossRef ]
• Buldu, M.; Buldu, N. Concept mapping as a formative assessment in college classrooms: Measuring usefulness and student satisfaction. Procedia Soc. Behav. Sci. 2010 , 2 , 2099–2104. [ Google Scholar ] [ CrossRef ] [ Green Version ]

Click here to enlarge figure

Share and Cite

Maya, J.; Luesia, J.F.; Pérez-Padilla, J. The Relationship between Learning Styles and Academic Performance: Consistency among Multiple Assessment Methods in Psychology and Education Students. Sustainability 2021 , 13 , 3341. https://doi.org/10.3390/su13063341

Maya J, Luesia JF, Pérez-Padilla J. The Relationship between Learning Styles and Academic Performance: Consistency among Multiple Assessment Methods in Psychology and Education Students. Sustainability . 2021; 13(6):3341. https://doi.org/10.3390/su13063341

Maya, Jesús, Juan F. Luesia, and Javier Pérez-Padilla. 2021. "The Relationship between Learning Styles and Academic Performance: Consistency among Multiple Assessment Methods in Psychology and Education Students" Sustainability 13, no. 6: 3341. https://doi.org/10.3390/su13063341

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Differentiating the learning styles of college students in different disciplines in a college English blended learning setting

Roles Conceptualization, Formal analysis, Funding acquisition, Methodology, Supervision, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Affiliations Department of Linguistics, School of International Studies, Zhejiang University, Hangzhou City, Zhejiang Province, China, Center for College Foreign Language Teaching, Zhejiang University, Hangzhou City, Zhejiang Province, China, Institute of Asian Civilizations, Zhejiang University, Hangzhou City, Zhejiang Province, China

Roles Formal analysis, Project administration, Writing – review & editing

Affiliation Department of Linguistics, School of International Studies, Zhejiang University, Hangzhou City, Zhejiang Province, China

Roles Formal analysis, Writing – original draft

Roles Writing – review & editing

• Jie Hu, 
• Yi Peng, 
• Xueliang Chen, 

• Published: May 20, 2021
• https://doi.org/10.1371/journal.pone.0251545
• Peer Review
• Reader Comments

Learning styles are critical to educational psychology, especially when investigating various contextual factors that interact with individual learning styles. Drawing upon Biglan’s taxonomy of academic tribes, this study systematically analyzed the learning styles of 790 sophomores in a blended learning course with 46 specializations using a novel machine learning algorithm called the support vector machine (SVM). Moreover, an SVM-based recursive feature elimination (SVM-RFE) technique was integrated to identify the differential features among distinct disciplines. The findings of this study shed light on the optimal feature sets that collectively determined students’ discipline-specific learning styles in a college blended learning setting.

Citation: Hu J, Peng Y, Chen X, Yu H (2021) Differentiating the learning styles of college students in different disciplines in a college English blended learning setting. PLoS ONE 16(5): e0251545. https://doi.org/10.1371/journal.pone.0251545

Editor: Haoran Xie, Lingnan University, HONG KONG

Received: May 15, 2020; Accepted: April 29, 2021; Published: May 20, 2021

Copyright: © 2021 Hu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and its Supporting Information files.

Funding: This research was supported by the Philosophical and Social Sciences Planning Project of Zhejiang Province in 2020 [grant number 20NDJC01Z] with the recipient Jie Hu, Second Batch of 2019 Industry-University Collaborative Education Project of Chinese Ministry of Education [grant number 201902016038] with the recipient Jie Hu, SUPERB College English Action Plan with the recipient Jie Hu, and the Fundamental Research Funds for the Central Universities of Zhejiang University with the recipient Jie Hu.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Research background.

Learning style, as an integral and vital part of a student’s learning process, has been constantly discussed in the field of education and pedagogy. Originally developed from the field of psychology, psychological classification, and cognitive research several decades ago [ 1 ], the term “learning style” is generally defined as the learner’s innate and individualized preference for ways of participation in learning practice [ 2 ]. Theoretically, learning style provides a window into students’ learning processes [ 3 , 4 ], predicts students’ learning outcomes [ 5 , 6 ], and plays a critical role in designing individualized instruction [ 7 ]. Knowing a student’s learning style and personalizing instruction to students’ learning style could enhance their satisfaction [ 8 ], improve their academic performance [ 9 ], and even reduce the time necessary to learn [ 10 ].

Researchers in recent years have explored students’ learning styles from various perspectives [ 11 – 13 ]. However, knowledge of the learning styles of students from different disciplines in blended learning environments is limited. In an effort to address this gap, this study aims to achieve two major objectives. First, it investigates how disciplinary background impacts students’ learning styles in a blended learning environment based on data collected in a compulsory college English course. Students across 46 disciplines were enrolled in this course, providing numerous disciplinary factor resources for investigating learning styles. Second, it introduces a novel machine learning method named the SVM to the field of education to identify an optimal set of factors that can simultaneously differentiate students of different academic disciplines. Based on data for students from 46 disciplines, this research delves into the effects of a massive quantity of variables related to students’ learning styles with the help of a powerful machine learning algorithm. Considering the convergence of a wide range of academic disciplines and the detection of latent interactions between a large number of variables, this study aims to provide a clear picture of the relationship between disciplinary factors and students’ learning styles in a blended learning setting.

Literature review

Theories of learning styles..

Learning style is broadly defined as the inherent preferences of individuals as to how they engage in the learning process [ 2 ], and the “cognitive, affective and physiological traits” of students have received special attention [ 14 ]. To date, there has been a proliferation of learning style definitions proposed to explain people’s learning preferences, each focusing on different aspects. Efforts to dissect learning style have been contested, with some highlighting the dynamic process of the learner’s interaction with the learning environment [ 14 ] and others underlining the individualized ways of information processing [ 15 ]. One vivid explication involved the metaphor of an onion, pointing out the multilayer nature of learning styles. It was proposed that the outermost layer of the learning style could change in accordance with the external environment, while the inner layer is relatively stable [ 16 , 17 ]. In addition, a strong concern in this field during the last three decades has led to a proliferation of models that are germane to learning styles, including the Kolb model [ 18 ], the Myers-Briggs Type Indicator model [ 19 ] and the Felder-Silverman learning style model (FSLSM) [ 20 ]. These learning style models have provided useful analytical lenses for analyzing students’ learning styles. The Kolb model focuses on learners’ thinking processes and identifies four types of learning, namely, diverging, assimilating, converging, and accommodating [ 18 ]. The Myers-Briggs Type Indicator model classifies learners into extraversion and introversion types, with the former preferring to learn from interpersonal communication and the latter inclining to benefit from personal experience [ 19 ]. As the most popular available model, the FSLSM identifies eight categories of learners according to the four dimensions of perception, input, processing and understanding [ 20 ]. In contrast to other learning style models that divided students into only a few groups, the FSLSM describes students’ learning styles in a more detailed manner. The four paired dimensions delicately distinguish students’ engagement in the learning process, providing a solid basis for a steady and reliable learning style analysis [ 21 ]. In addition, it has been argued that the FSLSM is the most appropriate model for a technology-enhanced learning environment because it involves important theories of cognitive learning behaviors [ 22 , 23 ]. Therefore, a large number of scholars have based their investigations of students’ learning styles in the e-learning/computer-aided learning environment on FSLSM [ 24 – 28 ].

Learning styles and FSLSM.

Different students receive, process, and respond to information with different learning styles. A theoretical model of learning style can be used to categorize people according to their idiosyncratic learning styles. In this study, the FSLSM was adopted as a theoretical framework to address the collective impacts of differences in students’ learning styles across different disciplines (see Fig 1 ).

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

This model specifies the four dimensions of the construct of learning style: visual/verbal, sensing/intuitive, active/reflective, and sequential/global. These four dimensions correspond to four psychological processes: input, perception, processing, and understanding.

https://doi.org/10.1371/journal.pone.0251545.g001

The FSLSM includes learning styles scattered among four dimensions.

Visual learners process information best when it is presented as graphs, pictures, etc., while verbal learners prefer spoken cues and remember best what they hear. Sensory learners like working with facts, data, and experimentation, while intuitive learners prefer abstract principles and theories. Active learners like to try things and learn through experimentation, while reflective learners prefer to think things through before taking action. Sequential learners absorb knowledge in a linear fashion and make progress step by step, while global learners tend to grasp the big picture before filling in all the details.

Learning styles and academic disciplines.

Learning styles vary depending on a series of factors, including but not limited to age [ 29 ], gender [ 30 ], personality [ 2 , 31 ], learning environment [ 32 ] and learning experience [ 33 ]. In the higher education context, the academic discipline seems to be an important variable that influences students’ distinctive learning styles, which echoes a multitude of investigations [ 29 , 34 – 41 ]. One notable study explored the learning styles of students from 4 clusters of disciplines in an academic English language course and proposed that the academic discipline is a significant predictor of students’ learning styles, with students from the soft-pure, soft-applied, hard-pure and hard-applied disciplines each favoring different learning modes [ 42 ]. In particular, researchers used the Inventory of Learning Styles (ILS) questionnaire and found prominent disparities in learning styles between students from four different disciplinary backgrounds in the special educational field of vocational training [ 43 ]. These studies have found significant differences between the learning styles of students from different academic disciplines, thus supporting the concept that learning style could be domain dependent.

Learning styles in an online/blended learning environment.

Individuals’ learning styles reflect their adaptive orientation to learning and are not fixed personality traits. Consequently, learning styles can vary among diverse contexts, and related research in different contexts is vital to understanding learning styles in greater depth. Web-based technologies eliminate barriers of space and time and have become integrated in individuals’ daily lives and learning habits. Online and blended learning have begun to pervade virtually every aspect of the education landscape [ 40 ], and this warrants close attention. In addition to a series of studies that reflected upon the application of information and communication technology in the learning process [ 44 , 45 ], recent studies have found a mixed picture of whether students in a web-based/blended learning environment have a typical preference for learning.

Online learning makes it possible for students to set their goals and develop an individualized study plan, equipping them with more learning autonomy [ 46 ]. Generally, students with a more independent learning style, greater self-regulating behavior and stronger self-efficacy are found to be more successful in an online environment [ 47 ]. For now, researchers have made substantial contributions to the identification and prediction of learning styles in an online learning environment [ 27 , 48 – 51 ]. For instance, an inspiring study focused on the manifestation of college students’ learning styles in a purely computer-based learning environment to evaluate the different learning styles of web-learners in the online courses, indicating that students’ learning styles were significantly related to online participation [ 49 ]. Students’ learning styles in interactive E-learning have also been meticulously investigated, from which online tutorials have been found to be contributive to students’ academic performance regardless of their learning styles [ 51 ].

As a flexible learning method, blended courses have combined the advantages of both online learning and traditional teaching methods [ 52 ]. Researchers have investigated students’ learning styles within this context and have identified a series of prominent factors, including perceived satisfaction and technology acceptance [ 53 ], the dynamics of the online/face-to-face environment [ 54 ], and curriculum design [ 55 ]. Based on the Visual, Aural, Reading or Write and Kinesthetic model, a comprehensive study scrutinized the learning styles of K12 students in a blended learning environment, elucidating the effect of the relationship between personality, learning style and satisfaction on educational outcomes [ 56 ]. A recent study underscored the negative effects of kinesthetic learning style, whereas the positive effects of visual or auditory learning styles on students’ academic performance, were also marked in the context of blended learning [ 57 ].

Considering that academic disciplines and learning environment are generally regarded as essential predictors of students’ learning styles, some studies have also concentrated on the effects of academic discipline in a blended learning environment. Focusing on college students’ learning styles in a computer-based learning environment, an inspiring study evaluated the different learning styles of web learners, namely, visual, sensing, global and sequential learners, in online courses. According to the analysis, compared with students from other colleges, liberal arts students, are more susceptible to the uneasiness that may result from remote teaching because of their learning styles [ 11 ]. A similar effort was made with the help of the CMS tool usage logs and course evaluations to explore the learning styles of disciplinary quadrants in the online learning environment. The results indicated that there were noticeable differences in tool preferences between students from different domains [ 12 ]. In comparison, within the context of blended learning, a comprehensive study employed chi-square statistics on the basis of the Community of Inquiry (CoI) presences framework, arguing that soft-applied discipline learners in the blended learning environment prefer the kinesthetic learning style, while no correlations between the learning style of soft-pure and hard-pure discipline students and the CoI presences were identified. However, it is noted that students’ blended learning experience depends heavily on academic discipline, especially for students in hard-pure disciplines [ 13 ].

Research gaps and research questions

Overall, the research seems to be gaining traction, and new perspectives are continually introduced. The recent literature on learning styles mostly focuses on the exploration of the disciplinary effects on the variation in learning styles, and some of these studies were conducted within the blended environment. However, most of the studies focused only on several discrete disciplines or included only a small group of student samples [ 34 – 41 ]. Data in these studies were gathered through specialized courses such as academic English language [ 42 ] rather than the compulsory courses available to students from all disciplines. Even though certain investigations indeed boasted a large number of samples [ 49 ], the role of teaching was emphasized rather than students’ learning style. In addition, what is often overlooked is that a large number of variables related to learning styles could distinguish students from different academic disciplines in a blended learning environment, whereas a more comprehensive analysis that takes into consideration the effects of a great quantity of variables related to learning styles has remained absent. Therefore, one goal of the present study is to fill this gap and shed light on this topic.

Another issue addressed in this study is the selection of an optimal measurement that can effectively identify and differentiate individual learning styles [ 58 ]. The effective identification and differentiation of individual learning styles can not only help students develop greater awareness of their learning but also provide teachers with the necessary input to design tailor-made instructions in pedagogical practice. Currently, there are two general approaches to identify learning styles: a literature-based approach and a data-driven approach. The literature-based approach tends to borrow established rules from the existing literature, while the data-driven approach tends to construct statistical models using algorithms from fields such as machine learning, artificial intelligence, and data mining [ 59 ]. Research related to learning styles has been performed using predominantly traditional instruments, such as descriptive statistics, Spearman’s rank correlation, coefficient R [ 39 ], multivariate analysis of variance [ 56 ] and analysis of variance (ANOVA) [ 38 , 43 , 49 , 57 ]. Admittedly, these instruments have been applied and validated in numerous studies, in different disciplines, and across multiple timescales. Nevertheless, some of the studies using these statistical tools did not identify significant results [ 36 , 53 , 54 ] or reached only loose conclusions [ 60 ]; this might be because of the inability of these methods to probe into the synergistic effects of variables. However, the limited functions of comparison, correlation, prediction, etc. are being complemented by a new generation of technological innovations that promise more varied approaches to addressing social and scientific issues. Machine learning is one such approach that has received much attention both in academia and beyond. As a subset of artificial intelligence, machine learning deals with algorithms and statistical models on computer systems, performing tasks based on patterns and inference instead of explicit instruction. As such, it can deal with high volumes of data at the same time, perform tasks automatically and independently, and continuously improve its performance based on past experience [ 54 ]. Similar machine learning approaches have been proposed and tested by different scholars to identify students’ learning styles, with varying results regarding the classification of learning styles. For instance, a study that examined the precision levels of four computational intelligence approaches, i.e., artificial neural network, genetic algorithm, ant colony system and particle swarm optimization, found that the average precision of learning style differentiation ranged between 66% and 77% [ 61 ]. Another study that classified learning styles through SVM reported accuracy levels ranging from 53% to 84% [ 62 ]. A comparison of the prediction performance of SVM and artificial neural networks found that SVM has higher prediction accuracy than the latter [ 63 ]. This was further supported by another study, which yielded a similar result between SVM and the particle swarm optimization algorithm [ 64 ]. Moreover, when complemented by a genetic algorithm [ 65 ] and ant colony system [ 66 ], SVM has also shown improved results. These findings across different fields point to the reliability of SVM as an effective statistical tool for identification and differentiation analysis.

Therefore, a comprehensive investigation across the four general disciplines in Biglan’s taxonomy using a strong machine learning approach is needed. Given the existence of the research gaps discussed above, this exploratory study seeks to address the following questions:

• Can students’ learning styles be applied to differentiate various academic disciplines in the blended learning setting? If so, what are the differentiability levels among different academic disciplines based on students’ learning styles?
• What are the key features that can be selected to determine the collective impact on differentiation by a machine learning algorithm?
• What are the collective impacts of optimal feature sets?

Materials and methods

This study adopted a quantitative approach for the analysis. First, a modified and translated version of the original ILS questionnaire was administered to collect scores for students’ learning styles. Then, two alternate data analyses were performed separately. One analysis involved a traditional ANOVA, which tested the main effect of discipline on students’ learning styles in each ILS dimension. The other analysis involved the support vector machine (SVM) technique to test its performance in classifying students’ learning styles in the blended learning course among 46 specializations. Then, SVM-based recursive feature elimination (SVM-RFE) was employed to specify the impact of students’ disciplinary backgrounds on their learning styles in blended learning. By referencing the 44 questions (operationalized as features in this study) in the ILS questionnaire, SVM-RFE could rank these features based on their relative importance in differentiating different disciplines and identify the key features that collectively differentiate the students’ learning style. These steps are intended to not only identify students’ learning style differences but also explain such differences in relation to their academic disciplinary backgrounds.

Participants

The participants included 790 sophomores taking the blended English language course from 46 majors at Z University. Sophomore students were selected for this study for two reasons. First, sophomores are one of the only two groups of students (the other group being college freshmen) who take a compulsory English language course, namely, the College English language course. Second, of these two groups of students, sophomores have received academic discipline-related education, while their freshmen counterparts have not had disciplinary training during the first year of college. In the College English language course, online activities, representing 55% of the whole course, include e-course teaching designed by qualified course teachers or professors, courseware usage for online tutorials, forum discussion and essay writing, and two online quizzes. Offline activities, which represent 45% of the whole course, include role-playing, ice-breaker activities, group presentations, an oral examination, and a final examination. Therefore, the effects of the academic discipline on sophomores’ learning styles might be sufficiently salient to warrant a comparison in a blended learning setting [ 67 ]. Among the participants, 420 were male, and 370 were female. Most participants were aged 18 to 19 years and had taken English language courses for at least 6 years. Based on Biglan’s typology of disciplinary fields, the students’ specializations were classified into the four broad disciplines of hard-applied (HA, 289/37.00%), hard-pure (HP, 150/19.00%), soft-applied (SA, 162/20.00%), and soft-pure (SP, 189/24.00%).

Biglan’s classification scheme of academic disciplines (hard (H) vs. soft (S) disciplines and pure (P) vs. applied (A) disciplines) has been credited as the most cited organizational system of academic disciplines in tertiary education [ 68 – 70 ]. Many studies have also provided evidence supporting the validity of this classification [ 69 ]. Over the years, research has indicated that Biglan’s typology is correlated with differences in many other properties and serves as an appropriate mechanism to organize discipline-specific knowledge or epistemologies [ 38 ] and design and deliver courses for students with different learning style preferences [ 41 ]. Therefore, this classification provides a convenient framework to explore differences across disciplinary boundaries. In general, HA disciplines include engineering, HP disciplines include the so-called natural sciences, SA disciplines include the social sciences, and SP disciplines include the humanities [ 41 , 68 , 71 ].

In learning style research, it is difficult to select an instrument to measure the subjects’ learning styles [ 72 ]. The criteria used for the selection of a learning style instrument in this study include the following: 1) successful use of the instrument in previous studies, 2) demonstrated validity and reliability, 3) a match between the purpose of the instrument and the aim of this study and 4) open access to the questionnaire.

The Felder and Soloman’s ILS questionnaire, which was built based on the FSLSM, was adopted in the present study to investigate students’ learning styles across different disciplines. First, the FSLSM is recognized as the most commonly used model for measuring individual learning styles on a general scale [ 73 ] in higher education [ 74 ] and has remained popular for many years across different disciplines in university settings and beyond. In the age of personalized instruction, this model has breathed new life into areas such as blended learning [ 75 ], online distance learning [ 76 ], courseware design [ 56 ], and intelligent tutoring systems [ 77 , 78 ]. Second, the FSLSM is based on previous learning style models; the FSLSM integrates all their advantages and is, thus, more comprehensive in delineating students’ learning styles [ 79 , 80 ]. Third, the FSLSM has a good predictive ability with independent testing sets (i.e., unknown learning style objects) [ 17 ], which has been repeatedly proven to be a more accurate, reliable, and valid model than most other models for predicting students’ learning performance [ 10 , 80 ]. Fourth, the ILS is a free instrument that can be openly accessed online (URL: https://www.webtools.ncsu.edu/learningstyles/ ) and has been widely used in the research context [ 81 , 82 ].

The modified and translated version of the original ILS questionnaire includes 44 questions in total, and 11 questions correspond to each dimension of the Felder-Silverman model as follows: questions 1–11 correspond to dimension 1 (active vs. reflective), questions 12–22 correspond to dimension 2 (sensing vs. intuitive), questions 23–33 correspond to dimension 3 (visual vs. verbal), and questions correspond 34–44 to dimension 4 (sequential vs. global). Each question is followed by five choices on a five-point Likert scale ranging from “strongly agree with A (1)”, “agree with A (2)”, “neutral (3)”, “agree with B (4)” and “strongly agree with B (5)”. Option A and option B represent the two choices offered in the original ILS questionnaire.

Ethics statements

The free questionnaires were administered in a single session by specialized staff who collaborated on the investigation. The participants completed all questionnaires individually. The study procedures were in accordance with the ethical standards of the Helsinki Declaration and were approved by the Ethics Committee of the School of International Studies, Zhejiang University. All participants signed written informed consent to authorize their participation in this research. After completion of the informed consent form, each participant was provided a gift (a pen) in gratitude for their contribution and participation.

Data collection procedure

Before the questionnaires were distributed, the researchers involved in this study contacted faculty members from various departments and requested their help. After permission was given, the printed questionnaires were administered to students under the supervision of their teachers at the end of their English language course. The students were informed of the purpose and importance of the study and asked to carefully complete the questionnaires. The students were also assured that their personal information would be used for research purposes only. All students provided written informed consent (see S2 File ). After the questionnaires were completed and returned, they were thoroughly examined by the researchers such that problematic questionnaires could be identified and excluded from further analysis. All questionnaires eligible for the data analysis had to meet the following two standards: first, all questions must be answered, and second, the answered questions must reflect a reasonable logic. Regarding the few missing values, the median number of a given individual’s responses on 11 questions per dimension included in the ILS questionnaire was used to fill the void in each case. In statistics, using the median number to impute missing values is common and acceptable because missing values represent only a small minority of the entire dataset and are assumed to not have a large impact on the final results [ 83 , 84 ].

In total, 850 questionnaires were administered to the students, and 823 of these questionnaires were retrieved. Of the retrieved questionnaires, the remaining 790 questionnaires were identified as appropriate for further use. After data screening, these questionnaires were organized, and their respective results were translated into an Excel format.

Data analysis method

During the data analysis, as a library of the SVM, the free package LIBSVM ( https://www.csie.ntu.edu.tw/~cjlin/libsvm/ ) was first applied as an alternative method of data analysis. Then, a traditional ANOVA was performed to examine whether there was a main effect of academic discipline on Chinese students’ learning styles. ANOVA could be performed using SPSS, a strong data analysis software that supports a series of statistical analyses. In regard to the examination of the effect of a single or few independent variables, SPSS ANOVA can produce satisfactory results. However, SVM, a classic data mining algorithm, outperforms ANOVA for dataset in which a large number of variables with multidimensions are intertwined and their combined/collective effects influence the classification results. In this study, the research objective was to efficiently differentiate and detect the key features among the 44 factors. Alone, a single factor or few factors might not be significant enough to discriminate the learning styles among the different disciplines. Selected by the SVM, the effects of multiple features may collectively enhance the classification performance. Therefore, the reason for selecting SVM over ANOVA is that in the latter case, the responses on all questions in a single dimension are summed instead of treated as individual scores; thus, the by-item variation is concealed. In addition, the SVM is especially suitable for statistical analysis with high-dimensional factors (usually > 10; 44-dimensional factors were included in this study) and can detect the effects collectively imposed by a feature set [ 85 ].

Originally proposed in 1992 [ 86 ], the SVM is a supervised learning model related to machine learning algorithms that can be used for classification, data analysis, pattern recognition, and regression analysis. The SVM is an efficient classification model that optimally divides data into two categories and is ranked among the top methods in statistical theory due to its originality and practicality [ 85 ]. Due to its robustness, accurate classification, and prediction performance [ 87 – 89 ], the SVM has high reproducibility [ 90 , 91 ]. Due to the lack of visualization of the computing process of the SVM, the SVM has been described as a “black box” method [ 92 ]; however, future studies in the emerging field of explainable artificial intelligence can help solve this problem and convert this approach to a “glass box” method [ 67 ]. This algorithm has proven to have a solid theoretical foundation and excellent empirical application in the social sciences, including education [ 93 ] and natural language processing [ 94 ]. The mechanism underlying the SVM is also presented in Fig 2 .

Hyperplanes 1 and 2 are two regression lines that divide the data into two groups. Hyperplane 1 is considered the best fitting line because it maximizes the distance between the two groups.

https://doi.org/10.1371/journal.pone.0251545.g002

The SVM contains the following two modules: one module is a general-purpose machine learning method, and the other module is a domain-specific kernel function. The SVM training algorithm is used to build a training model that is then used to predict the category to which a new sample instance belongs [ 95 ]. When a set of training samples is given, each sample is given the label of one of two categories. To evaluate the performance of SVM models, a confusion matrix, which is a table describing the performance of a classifier on a set of test data for which the true values are known, is used (see Table 1 ).

https://doi.org/10.1371/journal.pone.0251545.t001

ACC represents the proportion of true results, including both positive and negative results, in the selected population;

SPE represents the proportion of actual negatives that are correctly identified as such;

SEN represents the proportion of actual positives that are correctly identified as such;

AUC is a ranking-based measure of classification performance that can distinguish a randomly chosen positive example from a randomly chosen negative example; and

F-measure is the harmonic mean of precision (another performance indicator) and recall.

The ACC is a good metric frequently applied to indicate the measurement of classification performance, but the combination of the SPE, SEN, AUC, F-measure and ACC may be a measure of enhanced performance assessment and was frequently applied in current studies [ 96 ]. In particular, the AUC is a good metric frequently applied to validate the measurement of the general performance of models [ 97 ]. The advantage of this measure is that it is invariant to relative class distributions and class-specific error costs [ 98 , 99 ]. Moreover, to some extent, the AUC is statistically consistent and more discriminating than the ACC with balanced and imbalanced real-world data sets [ 100 ], which is especially suitable for unequal samples, such as the HA-HP model in this study. After all data preparations were completed, the data used for the comparisons were extracted separately. First, the processed data of the training set were run by using optimized parameters. Second, the constructed model was used to predict the test set, and the five indicators of the fivefold cross-validation and fivefold average were obtained. Cross-validation is a general validation procedure used to assess how well the results of a statistical analysis generalize to an independent data set, which is used to evaluate the stability of the statistical model. K-fold cross-validation is commonly used to search for the best hyperparameters of SVM to achieve the highest accuracy performance [ 101 ]. In particular, fivefold, tenfold, and leave-one-out cross-validation are typically used versions of k-fold cross-validation [ 102 , 103 ]. Fivefold cross-validation was selected because fivefold validation can generally achieve a good prediction performance [ 103 , 104 ] and has been commonly used as a popular rule of thumb supported by empirical evidence [ 105 ]. In this study, five folds (groups) of subsets were randomly divided from the entire set by the SVM, and four folds (training sample) of these subsets were randomly selected to develop a prediction model, while the remaining one fold (test sample) was used for validation. The above functions were all implemented with Python Programming Language version 3.7.0 (URL: https://www.python.org/ ).

Then, SVM-RFE, which is an embedded feature selection strategy that was first applied to identify differentially expressed genes between patients and healthy individuals [ 106 ], was adopted. SVM-RFE has proven to be more robust to data overfitting than other feature selection techniques and has shown its power in many fields [ 107 ]. This approach works by removing one feature each time with the smallest weight iteratively to a feature rank until a group of highly weighted features were selected. After this feature selection procedure, several SVM models were again constructed based on these selected features. The performance of the new models is compared to that of the original models with all features included. The experimental process is provided in Fig 3 for the ease of reference.

https://doi.org/10.1371/journal.pone.0251545.g003

The classification results produced by SVM and the ranking of the top 20 features produced by SVM-RFE were listed in Table 2 . Twenty variables have been selected in this study for two reasons: a data-based reason and a literature-based reason. First, it is clear that models composed of 20 features generally have a better performance than the original models. The performance of models with more than 20 is negatively influenced. Second, SVM-based studies in the social sciences have identified 20 to 30 features as a good number for an optimal feature set [ 108 ], and 20 features were selected for inclusion in the optimal feature set [ 95 ]. Therefore, in this study, the top 20 features were selected for subsequent analysis, as proposed in previous analyses that yielded accepted measurement rates. These 20 features retained most of the useful information from all 44 factors but with fewer feature numbers, which showed satisfactory representation [ 96 ].

https://doi.org/10.1371/journal.pone.0251545.t002

Results of RQ (1) What are the differentiability levels among different academic disciplines based on students’ learning styles?

To further measure the performance of the differentiability among students’ disciplines, the collected data were examined with the SVM algorithm. As shown in Table 2 , the five performance indicators, namely, the ACC, SPE, SEN, AUC and F-measure, were utilized to measure the SVM models. Regarding the two general performance indicators, i.e., the ACC value and AUC value, the HA-HP, HA-SA, and HA-SP-based models yielded a classification capacity of approximately 70.00%, indicating that the students in these disciplines showed a relatively large difference. In contrast, the models based on the H-S, A-P, HP-SA, HP-SP, and SA-SP disciplines only showed a moderate classification capacity (above 55.00%). This finding suggests that these five SVM models were not as effective as the other three models in differentiating students among these disciplines based on their learning styles. The highest ACC and AUC values were obtained in the model based on the HA-HP disciplines, while the lowest values were obtained in the model based on the HP-SA disciplines. As shown in Table 2 , the AUCs of the different models ranged from 57.76% (HP-SA) to 73.97% (HA-HP).

To compare the results of the SVM model with another statistical analysis, an ANOVA was applied. Prior to the main analysis, the students’ responses in each ILS dimension were summed to obtain a composite score. All assumptions of ANOVA were checked, and no serious violations were observed. Then, an ANOVA was performed with academic discipline as the independent variable and the students’ learning styles as the dependent variable. The results of the ANOVA showed that there was no statistically significant difference in the group means of the students’ learning styles in Dimension 1, F(3, 786) = 2.56, p = .054, Dimension 2, F(3, 786) = 0.422, p = .74, or Dimension 3, F(3, 786) = 0.90, p = .443. However, in Dimension 4, a statistically significant difference was found in the group means of the students’ learning styles, F (3, 786) = 0.90, p = .005. As the samples in the four groups were unbalanced, post hoc comparisons using Scheffé’s method were performed, demonstrating that the means of the students’ learning styles significantly differed only between the HA (M = 31.04, SD = 4.986) and SP (M = 29.55, SD = 5.492) disciplines, 95.00% CI for MD [0.19, 2.78], p = .016, whereas the other disciplinary models showed no significant differences. When compared with the results obtained from the SVM models, the three models (HA-HP, HA-SA, and HA-SP models) presented satisfactory differentiability capability of approximately 70.00% based on the five indicators.

In the case of a significant result, it was difficult to determine which questions were representative of the significant difference. With a nonsignificant result, it was possible that certain questions might be relevant in differentiating the participants. However, this problem was circumvented in the SVM, where each individual question was treated as a variable and a value was assigned to indicate its relative importance in the questionnaire. Using SVM also circumvented the inherent problems with traditional significance testing, especially the reliance on p-values, which might become biased in the case of multiple comparisons [ 109 ].

Results of RQ (2) What are the key features that can be selected to determine the collective impact on differentiation by a machine learning algorithm?

To examine whether the model performance improved as a result of this feature selection procedure, the 20 selected features were submitted to another round of SVM analysis. The same five performance indicators were used to measure the model performance (see Table 2 ). By comparing the performance of the SVM model and that of the SVM-RFE model presented in Table 2 , except for the HA-SP model, all other models presented a similar or improved performance after the feature selection process. In particular, the improvement in the HA-HP and HP-SA models was quite remarkable. For instance, in the HA-HP model, the ACC value increased from 69.32% in the SVM model to 82.59% in the SVM-RFE model, and the AUC score substantially increased from 73.97% in the SVM model to 89.13% in the SVM-RFE model. This finding suggests that the feature selection process refined the model’s classification accuracy and that the 20 features selected, out of all 44 factors, carry substantive information that might be informative for exploring disciplinary differences. Although results for the indicators of the 20 selected features were not very high, all five indicators above 65.00% showed that the model was still representative because only 20 of 44 factors could present the classification capability. Considering that there was a significant reduction in the number of questions used for the model construction in SVM-RFE (compared with those used for the SVM model), the newly identified top 20 features by SVM-RFE were effective enough to preserve the differential ability of all 44 questions. Thus, these newly identified top 20 factors could be recognized as key differential features for distinguishing two distinct disciplines.

To identify these top 20 features in eight models (see Table 2 ), SVM-RFE was applied to rank order all 44 features contained in the ILS questionnaire. To facilitate a detailed understanding of what these features represent, the questions related to the top 20 features in the HA-HP model are listed in Table 3 for ease of reference.

https://doi.org/10.1371/journal.pone.0251545.t003

Results of RQ (3) What are the collective impacts of optimal feature sets?

The collective impacts of optimal feature sets could be interpreted from four aspects, namely, the complexities of students’ learning styles, the appropriate choice of SVM, the ranking of SVM-RFE and multiple detailed comparisons between students from different disciplines. First, the FSLSM considers the fact that students’ learning styles are shaped by a series of factors during the growth process, which intertwine and interact with each other. Considering the complex dynamics of the learning style, selecting an approach that could detect the combined effects of a group of variables is needed. Second, recent years have witnessed the emergence of data mining approaches to explore students learning styles [ 28 , 48 – 50 , 110 ]. Specifically, as one of the top machine learning algorithms, the SVM excels in identifying the combined effects of high-order factors [ 87 ]. In this study, the SVM has proven to perform well in classifying students’ learning styles across different disciplines, with every indicator being acceptable. Third, the combination of SVM with RFE could enable the simultaneous discovery of multiple features that collectively determine classification. Notably, although SVM-FRE could rank the importance of the features, they should be regarded as an entire optimal feature set. In other words, the combination of these 20 features, rather than a single factor, could differentiate students’ learning styles across different academic disciplines. Last but not least, the multiple comparisons between different SVM models of discipline provide the most effective learning style factors, giving researchers clues to the nuanced differences between students’ learning styles. It can be seen that students from different academic disciplines understand, see and reflect things from individualized perspectives. The 20 most effective factors for all models scattered within 1 to 44, verifying students’ different learning styles in 4 dimensions. Therefore, the FSLSM provides a useful and effective tool for evaluating students’ learning styles from a rather comprehensive point of view.

The following discussions address the three research questions explored in the current study.

Levels of differentiability among various academic disciplines based on students’ learning styles with SVM

The results suggest that SVM is an effective approach for classification in the blended learning context in which students with diverse disciplinary backgrounds can be distinguished from each other according to their learning styles. All performance indicators presented in Tables 2 and 3 remain above the baseline of 50.00%, suggesting that between each two disciplines, students’ learning style differences can be identified. To some extent, these differences can be identified with a relatively satisfactory classification capability (e.g., 69.32% of the ACC and 73.97% of the AUC in the HA-HP model shown in Table 2 ). Further support for the SVM algorithm is obtained from the SVM-RFE constructed to assess the rank of the factors’ classification capacity, and all values also remained above the baseline value, while some values reached a relatively high classification capability (e.g., 82.59% of the ACC and 89.13% of the AUC in the HA-HP model shown in Table 2 ). While the results obtained mostly show a moderate ACC and AUC, they still provide some validity evidence supporting the role of SVM as an effective binary classifier in the educational context. However, while these differences are noteworthy, the similarities among students in different disciplines also deserve attention. The results reported above indicate that in some disciplines, the classification capacity is not relatively high; this was the case for the model based on the SA-SP disciplines.

Regarding low differentiability, one explanation might be the indistinct classification of some emerging “soft disciplines.” It was noted that psychology, for example, could be identified as “a discipline that can be considered predominantly ‘soft’ and slightly ‘purer’ than ‘applied’ in nature” [ 111 ] (p. 43–53), which could have blurred the line between the SA and SP disciplines. As there is now no impassable gulf separating the SA and SP disciplines, their disciplinary differences may have diminished in the common practice of lecturing in classrooms. Another reason comes from the different cultivation models of “soft disciplines” and “hard disciplines” for sample students. In their high school, sample students are generally divided into liberal art students and science students and are then trained in different environments of knowledge impartation. The two-year unrelenting and intensive training makes it possible for liberal art students to develop a similar thinking and cognitive pattern that is persistent. After the college entrance examination, most liberal art students select SA or SP majors. However, a year or more of study in university does not exert strong effects on their learning styles, which explains why a multitude of researchers have traditionally investigated the SA and SP disciplines together, calling them simply “social science” or “soft disciplines” compared with “natural science” or “hard disciplines”. There have been numerous contributions pointing out similarities in the learning styles of students from “soft disciplines” [ 37 , 112 – 114 ]. However, students majoring in natural science exhibit considerable differences in learning styles, demonstrating that the talent cultivation model of “hard disciplines” in universities is to some extent more influential on students’ learning styles than that of the “soft disciplines”. Further compelling interpretations of this phenomenon await only the development of a sufficient level of accumulated knowledge among scholars in this area.

In general, these results are consistent with those reported in many previous studies based on the Felder-Silverman model. These studies tested the precision of different computational approaches in identifying and differentiating the learning styles of students. For example, by means of a Bayesian network (BN), an investigation obtained an overall precision of 58.00% in the active/reflective dimension, 77.00% in the sensing/intuitive dimension and 63.00% in the sequential/global dimension (the visual/verbal dimension was not considered) [ 81 ]. With the help of the keyword attributes of learning objects selected by students, a precision of 70.00% in the active/reflective dimension, 73.30% in the sensing/intuitive dimension, 73.30% in the sequential/global dimension and 53.30% in the visual/verbal dimension was obtained [ 115 ].

These results add to a growing body of evidence expanding the scope of the application of the SVM algorithm. Currently, the applications of the SVM algorithm still reside largely in engineering or other hard disciplines despite some tentative trials in the humanities and social sciences [ 26 ]. In addition, as cross-disciplines increase in current higher education, it is essential to match the tailored learning styles of students and researchers studying interdisciplinary subjects, such as the HA, HP, SA and SP disciplines. Therefore, the current study is the first to incorporate such a machine learning algorithm into interdisciplinary blended learning and has broader relevance to further learning style-related theoretical or empirical investigations.

Verification of the features included in the optimal feature sets

Features included in the optimal feature sets provided mixed findings compared with previous studies. Some of the 20 identified features are verified and consistent with previous studies. A close examination of the individual questions included in the feature sets can offer some useful insights into the underlying psychological processes. For example, in six of the eight models constructed, Question 1 (“I understand something better after I try it out/think it through”) appears as the feature with the number 1 ranking, highlighting the great importance attached to this question. This question mainly reflects the dichotomy between experimentation and introspection. A possible revelation is that students across disciplines dramatically differ in how they process tasks, with the possible exception of the SA-SP disciplines. This difference has been supported by many previous studies. For example, it was found that technical students tended to be more tactile than those in the social sciences [ 116 ], and engineering students (known as HA in this study) were more inclined toward concrete and pragmatic learning styles [ 117 ]. Similarly, it was explored that engineering students prefer “a logical learning style over visual, verbal, aural, physical or solitary learning styles” [ 37 ] (p. 122), while social sciences (known as SA in this study) students prefer a social learning style to a logical learning style. Although these studies differ in their focus to a certain degree, they provide an approximate idea of the potential differences among students in their relative disciplines. In general, students in the applied disciplines show a tendency to experiment with tasks, while those in the pure disciplines are more inclined towards introspective practices, such as an obsession with theories. For instance, in Biglan’s taxonomy of academic disciplines, students in HP disciplines prefer abstract rules and theories, while students in SA disciplines favor application [ 67 ]. Additionally, Question 10 (“I find it easier to learn facts/to learn concepts”) is similar to Question 1, as both questions indicate a certain level of abstraction or concreteness. The difference between facts and concepts is closely related to the classification difference between declarative knowledge and procedural knowledge in cognitive psychology [ 35 , 38 ]. Declarative knowledge is static and similar to facts, while procedural knowledge is more dynamic and primarily concerned with operational steps. Students’ preferences for facts or concepts closely correspond to this psychological distinction.

In addition, Questions 2, 4, 7, and 9 also occur frequently in the 20 features selected for the different models. Question 2 (“I would rather be considered realistic/innovative”) concerns taking chances. This question reflects a difference in perspective, i.e., whether the focus should be on obtaining pragmatic results or seeking original solutions. This difference cannot be easily connected to the disciplinary factor. Instead, there are numerous factors, e.g., genetic, social and psychological factors, that may play a strong role in defining this trait. The academic discipline only serves to strengthen or diminish this difference. For instance, decades of research in psychology have shown that males are more inclined towards risk taking than females [ 118 – 121 ]. A careful examination of the current academic landscape reveals a gender difference; more females choose soft disciplines than males, and more males choose hard disciplines than females. This situation builds a disciplinary wall classifying students into specific categories, potentially strengthening the disciplinary effect. For example, Question 9 (“In a study group working on difficult material, I am more likely to jump in and contribute ideas/sit back and listen”) emphasizes the distinction between active participation and introspective thinking, reflecting an underlying psychological propensity in blended learning. Within this context, the significance of this question could also be explained by the psychological evaluation of “loss and gain”, as students’ different learning styles are associated with expected reward values and their internal motivational drives, which are determined by their personality traits [ 122 ]. When faced with the risk of “losing face”, whether students will express their ideas in front of a group of people depends largely on their risk and stress management capabilities and the presence of an appropriate motivation system.

The other two questions also convey similar messages regarding personality differences. Question 4 concerns how individuals perceive the world, while Question 7 concerns the preferred modality of information processing. Evidence of disciplinary differences in these respects was also reported [ 35 , 123 – 125 ]. The other questions, such as Questions 21, 27, and 39, show different aspects of potential personality differences and are mostly consistent with the previous discussion. This might also be a vivid reflection of the multi-faceted effects of blended learning, which may differ in their consonance with the features of each discipline. First, teachers from different domains use technology in different ways, and student from different disciplines may view blended learning differently. For instance, the characteristics of soft-applied fields entail specialized customization in blended courses, further broadening the gulf between different subjects [ 126 ]. Second, although blended learning is generally recognized as a stimulus to students’ innovation [ 127 ], some students who are used to an instructivist approach in which the educator acts as a ‘sage on the stage’ will find it difficult to adapt to a social constructivist approach in which the educator serves as a ‘guide on the side’ [ 128 ]. This difficulty might not only negatively affect students’ academic performance but also latently magnify the effects of different academic disciplines.

Interpretation of the collective impact of optimal feature sets

In each SVM model based on a two-discipline model, the 20 key features (collectively known as an optimal feature set) selected exert a concerted effect on students’ learning styles across different disciplines (see Table 2 ). A broad examination of the distribution of collective impact of each feature set with 20 features in the eight discipline models suggests that it is especially imperative considering the emerging cross-disciplines in academia. Current higher education often involves courses with crossed disciplines and students with diverse disciplinary backgrounds. In addition, with the rise of technology-enhanced learning, the design of personalized tutoring systems requires more nuanced information related to student attributes to provide greater adaptability [ 59 ]. By identifying these optimal feature sets, such information becomes accessible. Therefore, understanding such interdisciplinary factors and designing tailor-made instructions are essential for promoting learning success [ 9 ]. For example, in an English language classroom in which the students are a blend of HP and SP disciplines, instructors might consider integrating a guiding framework at the beginning of the course and stepwise guidelines during the process such that the needs of both groups are met. With the knowledge that visual style is dominant across disciplines, instructors might include more graphic presentations (e.g., Question 11) in language classrooms rather than continue to use slides or boards filled with words. Furthermore, to achieve effective communication with students and deliver effective teaching, instructors may target these students’ combined learning styles. While some methods are already practiced in real life, this study acts as a further reminder of the rationale underlying these practices and thus increases the confidence of both learners and teachers regarding these practices. Therefore, the practical implications of this study mainly concern classroom teachers and educational researchers, who may draw some inspiration for interdisciplinary curriculum design and the tailored application of learning styles to the instructional process.

Conclusions

This study investigated learning style differences among students with diverse disciplinary backgrounds in a blended English language course based on the Felder-Silverman model. By introducing a novel machine learning algorithm, namely, SVM, for the data analysis, the following conclusions can be reached. First, the multiple performance indicators used in this study confirm that it is feasible to apply learning styles to differentiate various disciplines in students’ blended learning processes. These disciplinary differences impact how students engage in their blended learning activities and affect students’ ultimate blended learning success. Second, some questions in the ILS questionnaire carry more substantive information about students’ learning styles than other questions, and certain underlying psychological processes can be derived. These psychological processes reflect students’ discipline-specific epistemologies and represent the possible interaction between the disciplinary background and learning style. In addition, the introduction of SVM in this study can provide inspiration for future studies of a similar type along with the theoretical significance of the above findings.

Despite the notable findings of this study, it is subject to some limitations that may be perfected in further research. First, the current analysis examined the learning styles without allowing for the effects of other personal or contextual factors. The educational productivity model proposed by Walberg underlines the significance of the collected influence of contextual factors on individuals’ learning [ 129 ]. For example, teachers from different backgrounds and academic disciplines are inclined to select various teaching methods and to create divergent learning environments [ 130 ], which should also be investigated thoroughly. The next step is therefore to take into account the effects of educational background, experience, personality and learning experience to gain a more comprehensive understanding of students’ learning process in the blended setting.

In conclusion, the findings of this research validate previous findings and offer new perspectives on students’ learning styles in a blended learning environment, which provides future implications for educational researchers, policy makers and educational practitioners (i.e., teachers and students). For educational researchers, this study not only highlights the merits of using machine learning algorithms to explore students’ learning styles but also provides valuable information on the delicate interactions between blended learning, academic disciplines and learning styles. For policy makers, this analysis provides evidence for a more inclusive but personalized educational policy. For instance, in addition to learning styles, the linkage among students’ education in different phases should be considered. For educational practitioners, this study plays a positive role in promoting student-centered and tailor-made teaching. The findings of this study can help learners of different disciplines develop a more profound understanding of their blended learning tendencies and assist teachers in determining how to bring students’ learning styles into full play pedagogically, especially in interdisciplinary courses [ 131 – 134 ].

Supporting information

https://doi.org/10.1371/journal.pone.0251545.s001

S2 File. Informed consent for participants.

https://doi.org/10.1371/journal.pone.0251545.s002

S1 Dataset.

https://doi.org/10.1371/journal.pone.0251545.s003

Acknowledgments

The authors would like to thank the anonymous reviewers for their constructive comments on this paper and Miss Ying Zhou for her suggestions during the revision on this paper.

• View Article
• Google Scholar
• 15. Dunn R, Dunn K, Perrin J. Teaching young children through their individual learning styles. Boston, MA: Allyn & Bacon, Inc;1994.
• 17. Curry L. Integrating concepts of cognitive or learning style: A review with attention to psychometric standards. Ottawa, ON: Canadian College of Health Service Executives;1987.
• 18. Kolb DA. Experiential learning: experience as the source of learning and development. Englewood Cliffs, NJ: Prentice-Hall;1984.
• 21. Anitha D, Deisy C, Lakshmi SB, Meenakshi MK. Proposing a Classification Methodology to Reduce Learning Style Combinations for Better Teaching and Learning. 2014 IEEE Sixth International Conference on Technology for Education, Amritapuri, India, 2014; 208–211. https://doi.org/10.1109/T4E.2014.5
• 23. Kuljis J, Liu F. A comparison of learning style theories on the suitability for e-learning. Web Technologies, Applications, and Services, 2005; 191–197. Retrieved from: https://www.mendeley.com/catalogue/da014340-bfb5-32d1-b144-73545a86d440/
• 30. Richardson JTE. Researching student learning. Buckingham: SRHE and Open University Press; 2000.
• 33. Marton F, Säljö R. Approaches to learning. Edinburgh: Scottish Academic Press; 1984.
• PubMed/NCBI
• 52. Thorne K. Blended learning: How to integrate online and traditional learning, London: Kogan Page; 2003.
• 59. Graf S. Adaptivity in learning management systems focusing on learning styles. Vienna, Austria: Vienna University of Technology;2007.
• 79. Crockett K, Latham A, Mclean D, Bandar Z, O’Shea J. On predicting learning styles in conversational intelligent tutoring systems using fuzzy classification trees. IEEE International Conference on Fuzzy Systems. 2011; 2481–2488. https://doi.org/10.1109/FUZZY.2011.6007514
• 82. Coffield, Ecclestone K, Moseley, Hall E. Learning styles and pedagogy in post 16 education: A critical and systematic review. London, UK: Learning and Skills Research Centre;2004.
• 86. Acuna E, Rodriguez C. The Treatment of Missing Values and its Effect on Classifier Accuracy. In Banks D, House L, McMorris FR, Arabie P, Gaul W, editors. Classification, Clustering, and Data Mining Applications. Springer Berlin Heidelberg;2004. p. 639–647.
• 88. Boser BE, Guyon IM, Vapnik VN. A training algorithm for optimal margin classifiers. A training algorithm for optimal margin classifiers. Proceedings of The Fifth Annual Workshop on Computational Learning Theory. New York: ACM Press, 1992: 144–152. https://doi.org/10.1145/130385.130401
• 100. Maloof MA. Learning when data sets are imbalanced and when costs are unequal and unknown. Proceedings of the 20th International Conference on Machine Learning (ICML-2003). 2003. Retrieved from: http://www.site.uottawa.ca/~nat/Workshop2003/maloof-icml03-wids.pdf
• 101. Yan L, Rodier R, Mozer M, Wolniewicz R. Optimizing classifier performance via the Wilcoxon-Mann-Withney statistics. Proceedings of the 20th International Conference on Machine Learning (ICML-2003). 2003. Retrieved from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.324.1091&rep=rep1&type=pdf
• 105. James G, Witten D, Hastie T, Tibshirani R. An introduction to statistical learning: with applications in R (Springer Texts in Statistics). 1st ed. New York, NY: Springer Verlag;2013.
• 118. Anderson JR. Language, memory, and thought. Mahwah, NJ: Lawrence Erlbaum Associates;1976.
• 119. Anderson JR. The architecture of cognition. Cambridge, Massachusetts: Harvard University Press;1983.

• Research article
• Open access
• Published: 04 December 2018

The relationship between learning styles and academic performance in TURKISH physiotherapy students

• Nursen İlçin   ORCID: orcid.org/0000-0003-0174-8224 1 ,
• Murat Tomruk 1 ,
• Sevgi Sevi Yeşilyaprak 1 ,
• Didem Karadibak 1 &
• Sema Savcı 1  

BMC Medical Education volume  18 , Article number:  291 ( 2018 ) Cite this article

379k Accesses

50 Citations

4 Altmetric

Metrics details

Learning style refers to the unique ways an individual processes and retains new information and skills. In this study, we aimed to identify the learning styles of Turkish physiotherapy students and investigate the relationship between academic performance and learning style subscale scores in order to determine whether the learning styles of physiotherapy students could influence academic performance.

The learning styles of 184 physiotherapy students were determined using the Grasha-Riechmann Student Learning Style Scales. Cumulative grade point average was accepted as a measure of academic performance. The Kruskal-Wallis test was conducted to compare academic performance among the six learning style groups (Independent, Dependent, Competitive, Collaborative, Avoidant, and Participant).

The most common learning style was Collaborative (34.8%). Academic performance was negatively correlated with Avoidant score ( p  < 0.001, r  = − 0.317) and positively correlated with Participant score ( p  < 0.001, r  = 0.400). The academic performance of the Participant learning style group was significantly higher than that of all the other groups ( p  < 0.003).

Conclusions

Although Turkish physiotherapy students most commonly exhibited a Collaborative learning style, the Participant learning style was associated with significantly higher academic performance. Teaching strategies that encourage more participant-style learning may be effective in increasing academic performance among Turkish physiotherapy students.

Peer Review reports

Learning can be defined as permanent changes in behavior induced by life [ 1 ]. According to experiential learning theory, learning is “the process whereby knowledge is created through the transformation of experience” [ 2 , 3 ].

Facilitating the learning process is the primary aim of teaching [ 4 ]. Understanding the learning behavior of students is considered to be a part of this process [ 5 ]. Therefore, the concept of learning styles has become a popular topic in recent literature, with many theories about learning styles put forward to better understand the dynamic process of learning [ 2 , 3 ].

Learning style refers to an individual’s preferred way of processing new information for efficient learning [ 6 ]. Rita Dunn described the concept of learning style as “a unique way developed by students when he/she was learning new and difficult knowledge” [ 7 ]. Learning style is about how students learn rather than what they learn [ 1 ]. The learning process is different for each individual; even in the same educational environment, learning does not occur in all students at the same level and quality [ 8 ]. Research has shown that individuals exhibit different approaches in the learning process and a single strategy or approach was unable to provide optimal learning conditions for all individuals [ 9 ]. This may be related to students’ different backgrounds, strengths, weaknesses, interests, ambitions, levels of motivation, and approaches to studying [ 10 ]. To improve undergraduate education, educators should become more aware of these diverse approaches [ 11 ]. Learning styles may be useful to help students and educators understand how to improve the way they learn and teach, respectively.

Determining students’ learning styles provides information about their specific preferences. Understanding learning styles can make it easier to create, modify, and develop more efficient curriculum and educational programs. It can also encourage students’ participation in these programs and motivate them to gain professional knowledge [ 9 ]. Therefore, determining learning style is quite valuable in order to achieve more effective learning. Researching learning styles provides data on how students learn and find answers to questions [ 5 ].

Considering the potential problems encountered in the undergraduate education of physiotherapists, determining the learning style of physiotherapy students may enable the development of strategies to improve the learning process [ 12 ]. Studies on learning styles in the field of physiotherapy have mostly been conducted in developed countries such as Canada and Australia [ 13 , 14 ]. A study conducted in Australia examined the learning styles of physiotherapy, occupational therapy, and speech pathology students. The results of this study suggest that optimal learning environment should also be taken into consideration while researching how students learn. The authors also stated that future research was needed to investigate correlations between learning styles, instructional methods, and the academic performance of students in the health professions [ 14 ].

To the best of our knowledge, there are no prior publications in the literature that report Turkish physiotherapy students’ learning styles. Furthermore, previous studies mostly used Kolb’s Learning Style Inventory (LSI), Marshall & Merritts’ LSI, or Honey & Mumford’s Learning Style Questionnaire (LSQ) to assess learning styles [ 5 , 13 , 15 , 16 , 17 , 18 ]. Some of these studies also suggested that learning behavior and styles should be investigated using different inventories [ 5 ]. Moreover, a scale that was indicated as valid and reliable for Turkish population was needed to accurately determine the learning styles of Turkish physiotherapy students. Therefore, we opted to use the Grascha-Riechmann Learning Style Scales (GRLSS) to assess the learning styles of physiotherapy students, which will be a first in the literature.

Learning style preferences are influential in learning and academic achievement, and may explain how students learn [ 19 ]. Previous studies have demonstrated a close association between learning style and academic performance [ 20 , 21 ]. Learning styles have been identified as predictors of academic performance and guides for curriculum design. The aim of this study was to determine whether learning style preferences of physiotherapy students could affect academic performance by identifying the learning styles of Turkish physiotherapy students and assessing the relationship between these learning styles and the students’ academic performance. Since physiotherapy education mainly consists of practice lessons and clinical practice and mostly requires active student participation, we hypothesized that physiotherapy students with a Collaborative learning style according to the GRLSS would have higher academic performance.

A cross-sectional survey design using a convenience sample was used. The study population consisted of 488 physiotherapy students who were officially registered for the 2013–2014 academic year in Dokuz Eylul University (DEU) School of Physical Therapy and Rehabilitation. A minimum sample size of 68 participants was calculated with 95% confidence interval and 80% power by using “Epi Info Statcalc Version 6”. Inclusion criteria were (i) age ≥ 17 years, (ii) official registration in DEU School of Physical Therapy and Rehabilitation for the 2013–2014 academic year, (iii) being a first-, second-, third-, or fourth-year undergraduate student of physiotherapy, (iv) ability to read, write, and understand Turkish, and (v) being willing and able to participate in the study. Exclusion criteria were (i) unwilling to participate in the study, (ii) inability to read, write, and understand Turkish. The questionnaire was distributed to the physiotherapy students in a classroom setting during the final exam week of the academic year. Due to the absence of participants who did not attend final exams and were not actively attending classes (non-attendance students), questionnaires were distributed to 217 students in total.

184 physiotherapy students with a mean ± SD age of 21.52 ± 1.75 years participated in the study. Participants were informed verbally and in writing about the purpose of the study and the survey that would be implemented. A research assistant was available in the classroom to provide assistance if required. Demographic characteristics (age, gender, undergraduate year) comprised the first section of the questionnaire, followed by the GRLSS to assess learning style.

Cumulative grade point average (CGPA) shown on the students’ transcripts was used as the measure of academic performance. The students’ CGPAs at the end of the 2013–2014 academic year were obtained from the records held in the student affairs unit of the DEU School of Physical Therapy and Rehabilitation. CGPA was derived by multiplying the grade point (out of 100) with the credit units for each module or course and then dividing the total sum by the total credit units taken in the program.

The local university ethics committee provided ethical approval and informed consent was obtained from the participants before inclusion. Ethical protocol number was 1432-GOA.

Data collection

Grasha-riechmann student learning style scales.

The GRLSS is a five-point Likert-type scale ( response format: strongly disagree, moderately disagree, undecided, moderately agree, strongly agree ) consisting of 60 items which was designed based on student interviews and survey data [ 22 , 23 ]. In accordance with the response to student attitudes toward learning, classroom activities, teachers and peers, six learning styles were defined [ 24 ]. Learning styles that form subscales are the Independent, Avoidant, Collaborative, Dependent, Competitive, and Participant learning styles [ 24 , 25 ]. The six main styles in the GRLSS are described in Table  1 and the scoring of the GRLSS is shown in Table  2 [ 23 , 24 ]. The GRLSS was adapted to Turkish in 2003 and found to have good reliability [ 25 ] (Table  3 ).

The learning styles of the physiotherapy students in the current study were identified according to GRLSS and the students were grouped based on their predominant (highest scoring) style. The mean and median academic performance values of each group were calculated and the significance of the differences between groups was statistically analyzed.

Statistical analysis

Statistical analyses were performed to compare academic performances among the learning style groups and test the significance of pairwise differences. All data were analyzed using Statistical Package for Social Science software (IBM Corporation, version 20.0 for Windows). Descriptive statistics were summarized as frequencies and percentages for categorical variables. Continuous variables were presented as mean and standard deviation when normally distributed and as median and interquartile range when not normally distributed. Mann-Whitney U test was used for between-group analyses of abnormally distributed variables. The variables were investigated using visual (histograms, probability plots) and analytical methods (Kolmogorov-Smirnov/Shapiro-Wilk test) to determine whether they showed normal distribution. As parameters were not normally distributed, the correlation coefficients and their significance were calculated using Spearman test. Strength of correlation was defined as very weak for r values between 0.00–0.19, weak for r values between 0.20–0.39, moderate for r values between 0.40–0.69, strong for r values between 0.70–0.89, and very strong for r values over 0.90 [ 26 ]. As the academic performance was not normally distributed, the Kruskal-Wallis test was conducted to compare this parameter among the six learning style groups. The Mann-Whitney U Test was performed to test the significance of pairwise differences using Bonferroni correction to adjust for multiple comparisons. An overall 5% type-I error level was used to infer statistical significance ( p  < 0.05).

A total of 217 physiotherapy students were invited to participate in the study. Eighteen students refused to participate. Fifteen surveys were discarded due to missing item responses. As a result, data obtained from 184 students were used for the analyses. Overall response rate was 84.8%.

Demographic characteristics (gender, year) and learning style preferences are presented in Table  4 . The most common learning styles among the physiotherapy students according to the GRLSS were Collaborative (34.8%) and Independent (22.3%). The results of GRLSS subscale scores were given in Table  5 . The highest subscale score was Collaborative (Mean ± SD = 3.57 ± 0.62), while Competitive score was the lowest (Mean ± SD = 2.81 ± 0.69).

A moderate positive correlation between academic performance and Participant score was found (p < 0.001, r = 0.400) . A weak negative correlation was also found between academic performance and Avoidant score (p < 0.001, r = − 0.317) . No other significant correlation between academic performance and subscale scores was found (Table  6 ) .

When students were grouped according to learning styles, between-group (Kruskal-Wallis) analysis showed a significant difference in the academic performance of the groups (p < 0.001). Post-hoc (Mann-Whitney U) analysis revealed significantly higher academic performance in the Participant learning style group compared to all of the other learning style groups (Independent, Avoidant, Collaborative, Dependent, and Competitive) (Table  7 ).

The current study assessed the learning styles of Turkish physiotherapy students, and investigated the relationship between their learning styles and academic performance. The results revealed that the Collaborative learning style was most common among the Turkish physiotherapy students. However, students with Participant learning style had statistically higher academic performance when compared to the others. In addition, we found a positive correlation between Participant score and academic performance of the students, which supports the previous finding, while a negative correlation was found between Avoidant score and academic performance. In the case of physiotherapy students in this study, the emphasis should be on developing Participant and Collaborative learning skills. This might involve providing more class activities, discussions, and group projects.

The physiotherapy program at DEU has a combined case study-based and traditional style curriculum including lectures, tutorials, seminars, case study presentations, and supervised small group clinical practice in the hospital and at other health centers. Learning tasks and assessment methods include individual written examinations, practical examinations, homework and assignments as well as collaborative oral presentation and research projects. In the physiotherapy discipline, clinical practice improves students’ occupational skills and is seen as a crucial part of the teaching process [ 12 , 27 ]. Similarly, the teaching and learning approach at DEU is heavily based on practical training and requires active participation and group work. This could be a reason for the greater preference for Collaborative learning style.

Previous studies have indicated that physiotherapy students prefer abstract learning styles [ 28 ] and have desirable approaches to studying [ 29 ]. Canadian and American physiotherapy students preferred Converger (40 and 37% respectively) or Assimilator (35 and 28% respectively) learning styles [ 13 ]. According to descriptions of the learning style categories in the Kolb LSI, Convergers enjoy learning through activities like homework problems, computer simulations, field trips, and reports and demonstrations presented by others. On the other hand, Assimilators prefer attending lectures, reading textbooks, doing independent research and watching demonstrations by instructors when learning. In our study, Turkish physiotherapy students preferred Collaborative (34.8%) or Independent (22.3%) learning styles. According to GRLSS, Collaboratives prefer lectures with small group discussions and group projects (similar to Assimilators), while Independents prefer self-pace instruction and studying alone (similar to Convergers). Therefore, it can be concluded that learning styles of Canadian, American, and Turkish physiotherapy students are similar to each other.

Katz and Heimann used the Kolb LSI in their study and reported average learning style scores instead of the number of students in each of the four learning styles. They reported Converger as the “average” learning style for physiotherapy students [ 30 ]. In our study, the largest proportion of the physiotherapy students had a Collaborative learning style. Moreover, the average learning style was also Collaborative, with the highest average score.

Competitive learning style was the least frequently preferred (5.4%) by Turkish physiotherapy students in our study. The low preference for Competitive learning style indicates that students were less likely to compete with other students in the class to get a grade. Mountford et al. assessed learning styles of Australian physiotherapy students using Honey & Mumford’s LSQ and found that the Pragmatic learning style was the least preferred. According to LSQ, Pragmatists tend to see problem solving as a chance to rise to a challenge [ 31 ]. Considering that both Competitives and Pragmatists like challenges, the least frequently preferred styles of Australian and Turkish physiotherapy students seem to be similar to each other.

Alsop and Ryan pointed out that “personal awareness of learning styles and confidence in communicating this are first steps to achieve an optimal learning environment” [ 32 ]. According to Kolb’s theory, a preferred learning style affects a person’s problem solving ability [ 13 ]. Wessel et al. also stated that in order to provide students the best learning opportunity, educators must be aware of the learning styles and students’ ability to solve problems [ 13 ]. Indeed, evidence supporting these views can be found in the literature. Previous studies showed that students who were aware of their learning style had improved academic performance [ 33 , 34 ]. Nelson et al. found that college students who were tested on their learning style and were given appropriate education according to their learning style profile achieved higher academic performance than other students [ 33 ]. Linares also investigated learning styles in different health care professions (physiotherapy, occupational therapy, physician assistants, nursing and medical technology) and found a significant relationship between learning style and students’ readiness to undertake self-directed learning [ 15 ]. However, Hess et al. found no association between learning style and problem-solving ability in their study [ 35 ].

While planning this study, we hypothesized that students with a Collaborative learning style would have higher academic performance. Although the Collaborative learning style was the most common, these students did not show significantly higher academic performance. However, students with Participant learning style had statistically higher academic performance when compared to the other learning style groups. Characteristics specific to the Participant learning style are enjoyment from attending and participating in class and interest in class activities and discussions. These students enjoy opportunities to discuss class materials and readings. This may suggest that increasing in-class activities and discussions, which encourage participant-style learning, is needed to increase academic performance. Another approach would be to adapt teaching strategies according to the characteristics of Collaboratives, as they represented the largest body of students. Creating a convenient environment in which students could spend more time sharing and cooperating with their teacher and peers may facilitate collaborative learning, thus enhancing academic performance. Organizing the curriculum to include small group discussions within lectures and incorporate group projects may also be beneficial. As Ford et al. stated, “ Identification teaching profiles could be used to tailor the collaborative structure and content delivery ” [ 36 ].

The most important reason for determining learning style is to create a proper teaching strategy [ 37 , 38 , 39 , 40 ]. However, there seems to be no exact relationship between students’ learning style and the curriculum of a program described in the literature [ 13 ]. Learning style alone is not the only factor that may influence a learning situation. Many factors (educational and cultural context of university, individual awareness, life experience, other learning skills, effect of educator, motivation, etc.) may influence the learning process [ 31 ]. Therefore, expecting a simple relationship between learning style and teaching strategy may not be realistic. Moreover, the review of Pashler et al. showed that there was virtually no evidence that people learn better when teaching style is tailored to match students’ preferred learning style [ 41 ]. Nevertheless, future studies investigating physiotherapy educators’ teaching styles and their association with learning styles and academic performance may elucidate this complex issue.

The major strength of this study is that, to the best of our knowledge, ours is the first study investigating the learning styles of Turkish physiotherapy students with relation to academic performance.

There were some limitations to this study. It should be noted that learning style is a self-reported measure that can change based on experience and the demands of a situation. Therefore, it is subjective and able to provide adaptive behavior [ 42 ]. It should also be kept in mind that the conclusions of this study could be limited due to the cross-sectional design, and respondent bias may be an issue because convenience sampling was used to recruit participants. One possible limitation of the study could be the fact that the three of the scale reliabilities reported for GRLSS was poor.

This study investigated the learning styles of physiotherapy students in only one university (DEU) and this could preclude the generalization of our results. Subsequent studies should include students enrolled in the physiotherapy departments of multiple universities in Turkey to achieve an accurate geographical representation. Moreover, future studies on this topic should be conducted in collaboration with universities in Europe, with which we share a cultural connection.

The results of this study showed that the Collaborative learning style was most common among Turkish physiotherapy students. On the other hand, the physiotherapy students with Participant learning style had significantly higher academic performance than students with other learning styles. Teaching strategies consistent with the unique characteristics of the Participant learning style may be an effective way to increase academic performance of Turkish physiotherapy students. Incorporating more in-class activities and discussions about class material and readings may facilitate Participant learning, thus impacting academic performance positively. Another approach may be to adopt teaching strategies that target the predominant Collaborative learning style. Creating a convenient environment for students to share and cooperate with their teacher and peers and organizing the curriculum to include more small group discussions and group projects may also be supportive. Future studies should investigate physiotherapy educators’ teaching styles and their relations with learning styles and academic performance.

Abbreviations

Cumulative Grade Point Average

Dokuz Eylul University

Grascha-Riechmann Learning Style Scales

Learning Style Inventory

Learning Style Questionnaire

Hunt DE. Learning style and student needs: An introduction to conceptual level. In: NASSP, editor. Student learning styles: diagnosing and prescribing programs. Reston: National Association of Secondary School Principals; 1979. p. 27–38.

Google Scholar  

Arthurs JB. A juggling act in the classroom: managing different learning styles. Teach Learn Nurs. 2007;2(1):2–7.

Article   Google Scholar  

Coffield F, Moseley D, Hall E, Ecclestone K. Learning styles and pedagogy in post 16 learning: a systematic and critical review. London: The Learning and Skills Research Centre; 2004.

Ramsden P. Learning to teach in higher education. Oxon: Routledge; 2003.

Book   Google Scholar  

Mountford H, Jones S, Tucker B. Learning styles of entry-level physiotherapy students. Adv Physiother. 2006;8(Suppl 3):128–36.

Huston JL, Huston TL. How learning style and personality type can affect performance. The Health Care Supervisor. 1995;13(Suppl 4):38–45.

Dunn RS, Dunn KJ. Teaching secondary students through their individual learning styles: practical approaches for grades 7–12. New Jersey: Prentice Hall; 1993.

Saban A. Ögrenme-Ögretme Süreci Yeni Teori ve Yaklasimlar. Ankara: Nobel; 2005.

Brown T, Zoghi M, Williams B, Jaberzadeh S, Roller L, Palermo C, McKenna L, Wright C, Baird M, Schneider-Kolsky M. Are learning style preferences of health science students predictive of their attitudes towards e-learning? AJET. 2009;25:4.

Felder RM, Brent R. Understanding student differences. J Eng Educ. 2005;94(Suppl 1):57–72.

Carmo L, Gomes A, Pereira F, Mendes A. Learning styles and problem solving strategies. Proceedings of 3rd E-Learning Conference. 2006:7–8.

Hobbs C, Henley E, Higgs J, Williams V. Clinical education program strategies for challenging times. Focus on Health Professional Education: A Multidisciplinary Journal. 2000;2:7.

Wessel J, Loomis J, Rennie S, Brook P, Hoddinott J, Aherne M. Learning styles and perceived problem-solving ability of students in a baccalaureate physiotherapy programme. Physiother Theor Pr. 1999;15(Suppl 1):17–24.

Brown T, Cosgriff T, French G. Learning style preferences of occupational therapy, physiotherapy and speech pathology students: a comparative study. Int J Allied Health Sci Prac. 2008;6:1–12.

Linares AZ. Learning styles of students and faculty in selected health care professions. J Nurs Educ. 1999;38(Suppl 9):407–14.

Zoghi M, Brown T, Williams B, Roller L, Jaberzadeh S, Palermo C, McKenna L, Wright C, Baird M, Schneider-Kolsky M. Learning style preferences of Australian health science students. J Allied Health. 2010;39(Suppl 2):95–103.

Hauer P, Straub C, Wolf S. Learning styles of allied health students using Kolb's LSI-IIa. J Allied Health. 2005;34(Suppl 3):177–82.

Severiens S, Dam GT. Gender and gender identity differences in learning styles. Educ Psychol-Uk. 1997;17(Suppl 1):79–93.

Yazici HJ. Role of learning style preferences and interactive response systems on student learning outcomes. Int J Inf Oper Manag Educ. 2016;6(Suppl 2):109–34.

Ghazivakili Z, Nia RN, Panahi F, Karimi M, Gholsorkhi H, Ahmadi Z. The role of critical thinking skills and learning styles of university students in their academic performance. J Adv Med Educ Prof. 2014;2(Suppl 3):95–102.

Good JP, Ramos D, D'Amore DC. Learning style preferences and academic success of preclinical allied health students. J Allied Health. 2013;42(Suppl 4):81–90.

Riechmann SW, Grasha AF. A rational approach to developing and assessing the construct validity of a student learning style scales instrument. J Psychol. 1974;87(Suppl 2):213–23.

Grasha AF. Teaching with style: a practical guide to enhancing learning by understanding teaching and learning style. Pittsburgh: Alliance publishers; 1996.

Kumar P, Kumar A, Smart K. Assessing the impact of instructional methods and information technology on student learning styles. Issues in Informing Science and Information Technology. 2004;1:533–44.

Uzuntiryaki E, Bilgin I, Geban O. The effect of learning styles on high school students’ achievement and attitudes in Chemistry 2003. https://eric.ed.gov/?id=ED475483 . Accessed 15 May 2016.

Streiner DL, Norman GR, Cairney J. Health measurement scales: a practical guide to their development and use. Oxford: Oxford University Press; 2014.

Forbes R, Nolan D. Factors associated with patient satisfaction in student-led physiotherapy clinics: a qualitative study. Physiotherapy Theory and Practice. 2018;34(9):705–13.

Barris R, Kielhofner G, Bauer D. Learning preferences, values. and student satisfaction J Allied Health. 1985;14(Suppl 1):13–23.

Van Langenberghe HV. Evaluation of students’ approaches to studying in a problem-based physical therapy curriculum. Phys Ther. 1988;68(Suppl 4):522–7.

Katz N, Heimann N. Learning style of students and practitioners in five health professions. Occup Ther J Res. 1991;11(Suppl 4):238–44.

Honey P, Mumford A. The manual of learning styles. 3rd ed. Maidenhead: Peter Honey; 1992.

Alsop A, Ryan S. Making the most of fieldwork education: a practical approach. California: Springer; 2013.

Nelson B, Dunn R, Griggs SA, Primavera L. Effects of learning style intervention on college students retention and achievement. J Coll Stud Dev. 1993;34:364–9.

Sandmire DA, Boyce PF. Pairing of opposite learning styles among allied health students: effects on collaborative performance. J Allied Health. 2004;33(Suppl 2):156–63.

Hess D, Frantz JM. Understanding the learning styles of undergraduate physiotherapy students. Afr J Health Prof Educ. 2014;6(Suppl 1):45–7.

Ford JH, Robinson JM, Wise ME. Adaptation of the Grasha Riechman student learning style survey and teaching style inventory to assess individual teaching and learning styles in a quality improvement collaborative. BMC Medical Education. 2016;16:252.

Kell C, Rv D. Student learning preferences reflect curricular change. Med Teach. 2002;24(Suppl 1):32–40.

Rappolt S, Tassone M. How rehabilitation therapists gather, evaluate, and implement new knowledge. J Contin. Educ Health. 2002;22(Suppl 3):170–80.

Al Maghraby MA, Alshami AM. Learning style and teaching method preferences of Saudi students of physical therapy. Journal of family & community medicine. 2013;20(Suppl 3):192.

Celik Y, Ceylantekin Y, Kilic I. The evaluation of simulation maket in nursing education and the determination of learning style of students. Int J Health Sci. 2017;11(3):74–9.

Pashler H, McDaniel M, Rohrer D, Bjork R. Learning styles: concepts and evidence. Psychol Sci Public Interest. 2008;9(3):105–1.

Cassidy S. Learning styles: an overview of theories, models, and measures. Educ Psychol-Uk. 2004;24(Suppl 4):419–44.

Download references

Acknowledgements

The authors like to thank all physiotherapy students who participated in this study.

No funding was obtained for this study.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Author information

Authors and affiliations.

School of Physical Therapy and Rehabilitation, Dokuz Eylul University, 35340, Inciralti, Izmir, Turkey

Nursen İlçin, Murat Tomruk, Sevgi Sevi Yeşilyaprak, Didem Karadibak & Sema Savcı

You can also search for this author in PubMed   Google Scholar

Contributions

Nİ conducted the literature search for the background of the study, analyzed and interpreted statistical data, and wrote the majority of the article. MT conducted the literature search, collected data for the study, analyzed statistical data, and contributed to writing the article. SSY and DK were involved in study planning, data processing, and revising the article. SS contributed to study design and oversaw the study. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Nursen İlçin .

Ethics declarations

Author’s information.

Nursen İlçin, PT, PhD.

İlçin graduated from Dokuz Eylul University, School of Physical Therapy and Rehabilitation in 1998. She received her Master’s degree in 2002 and PhD in 2009 from Dokuz Eylül University. She is currently a associate professor in Geriatric Physiotherapy Department.

Murat Tomruk, PT, PhD.

Tomruk graduated from the School of Physical Therapy and Rehabilitation at Dokuz Eylul University in 2009. He received his MSci degree in Musculoskeletal Physiotherapy in 2013 and his PhD degree in 2018. His doctorate thesis was about manual therapy. He works as a research assistant at Dokuz Eylul University since 2011.

Sevgi Sevi Yeşilyaprak, PT, PhD.

Sevgi Sevi Yeşilyaprak’s speciality is shoulder rehabilitation. Her primary research interests are orthopaedic and sports injuries of the shoulder, shoulder biomechanics, proprioception, and exercise. She has one active and two completed grants. Yeşilyaprak teaches courses on musculoskeletal physiotherapy including sports physiotherapy, musculoskeletal disorders, therapeutic exercises, exercise prescription, and manual physiotherapy techniques.

Didem Karadibak, PT, PhD.

Karadibak obtained her BS degree in Physiotherapy from Hacettepe University in 1992 and her MS and PhD degrees from the Physical Therapy Program of the Institute of Health and Sciences, Dokuz Eylul University in 1998 and 2003, respectively. She is currently a professor of Cardiopulmonary Rehabilitation in the Dokuz Eylul University School of Physical Therapy and Rehabilitation.

Sema Savcı, PT, PhD.

Savcı obtained her BS degree in Physiotherapy from Hacettepe University in 1988 and her MS and PhD degrees from the Physical Therapy Program of the Institute of Health and Sciences, Hacettepe University in 1990 and 1995, respectively. She is currently a professor and serving as the Head of Cardiopulmonary Rehabilitation in the Dokuz Eylul University School of Physical Therapy and Rehabilitation.

Ethics approval and consent to participate

Written ethical approval was taken from the Dokuz Eylül University’s local ethics committee (approval number 1432-GOA) and written informed consent obtained from all the participants.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

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.

Reprints and permissions

About this article

Cite this article.

İlçin, N., Tomruk, M., Yeşilyaprak, S.S. et al. The relationship between learning styles and academic performance in TURKISH physiotherapy students. BMC Med Educ 18 , 291 (2018). https://doi.org/10.1186/s12909-018-1400-2

Download citation

Received : 19 June 2018

Accepted : 22 November 2018

Published : 04 December 2018

DOI : https://doi.org/10.1186/s12909-018-1400-2

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Learning style
• Academic performance
• Physiotherapy

BMC Medical Education

ISSN: 1472-6920

• Submission enquiries: [email protected]
• General enquiries: [email protected]

SYSTEMATIC REVIEW article

Is it really a neuromyth a meta-analysis of the learning styles matching hypothesis.

• Department of Education, Health, and Behavior Studies, University of North Dakota, Grand Forks, ND, United States

Learning styles have been a contentious topic in education for years. The purpose of this study was to conduct a meta-analysis of the effects of matching instruction to modality learning styles compared to unmatched instruction on learning outcomes. A systematic search of the research findings yielded 21 eligible studies with 101 effect sizes and 1,712 participants for the meta-analysis. Based on robust variance estimation, there was an overall benefit of matching instruction to learning styles, g  = 0.31, SE = 0.12, 95% CI = [0.05, 0.57], p  = 0.02. However, only 26% of learning outcome measures indicated matched instruction benefits for at least two styles, indicating a crossover interaction supportive of the matching hypothesis. In total, 12 studies without sufficient statistical details for the meta-analysis were also examined for an indication of a crossover effect; 25% of these studies had findings indicative of a crossover interaction. Given the time and financial expenses of implementation coupled with low study quality, the benefits of matching instruction to learning styles are interpreted as too small and too infrequent to warrant widespread adoption.

Introduction

Learning styles have been the topic of ongoing debate in education. Teacher education textbooks often state matching instruction to students’ preferred style will optimize learning outcomes (i.e., the matching hypothesis; Cuevas, 2015 ; Wininger et al., 2019 ). In contrast, cognitive scientists have argued there is a lack of empirical evidence to support the claims of the matching hypothesis ( Kirschner, 2017 ; Willingham, 2018 ). Because of the lack of known empirical evidence supporting the matching hypothesis, there is understandable concern that perpetuating the concept of learning styles could lead to wasting resources (namely, educator time and effort) to match instruction as well as stereotyping students into restrictive categories ( Newton and Miah, 2017 ). However, a meta-analysis aggregating findings compiled from an exhaustive search for studies on matching instruction to learning styles has not been conducted. Such a meta-analysis could be very helpful in informing this ongoing debate between educational practitioners and researchers. The purpose of this study was to conduct a meta-analysis of learning outcomes comparing conditions in which instruction is matched to students’ preferred learning styles to when instruction is unmatched to students’ preferred learning styles.

Literature review

It is not controversial that there are substantial individual differences in student learning—teacher education and cognitive science scholars agree on this concept. There is substantial empirical evidence that students’ academic performance and learning vary due to background knowledge, motivation, and study strategies ( Fong et al., 2021 ; Smith et al., 2021 ), just to name a few examples. However, the concept in learning styles theories that is controversial is the meshing or matching hypothesis in which students learn better when their instruction matches their preferred learning style ( Pashler et al., 2008 ; Cuevas, 2015 ; Lyle et al., 2023 ). A key aspect of the matching hypothesis is that there is a crossover interaction ( Kirschner, 2017 ), also known as a qualitative interaction, in which a particular treatment (in the case of learning styles, a particular modality of instruction) is effective for at least one subgroup but a different treatment is effective for another subgroup ( Qiu and Wang, 2019 ). Generally speaking, these crossover treatment interactions are rare ( Petticrew et al., 2012 ; Preacher and Sterba, 2019 ), but important to determining optimal treatments for individuals ( Olsen et al., 2019 ; Qiu and Wang, 2019 ).

There are numerous learning styles ( Dunn, 1990 ) as well as cognitive styles in which the preferred order of processing information varies ( Calcaterra et al., 2005 ; Fiorina et al., 2007 ). The most prevalent are preferred modalities for learning information ( Dekker et al., 2012 ; Brown, 2023 ). Learners are generally categorized through self-reports of preferred modalities ( An and Carr, 2017 ), such as the VAK typology (visual, auditory, and kinesthetic; Fallace, 2023a , b ). An example of accommodating these styles would be to provide information for learners categorized as “visual” in pictures, learners categorized as “auditory” would process the same information best aurally, and learners categorized as kinesthetic would have a hands-on activity ( Dunn and Dunn, 1975 ). Then, a read/write category was added making it the VARK typology for learners who were thought to best process information through reading verbal information (as opposed to visual learners who better processed pictures; Fleming and Mills, 1992 ).

A typology similar to the VARK for categorizing learning styles is the verbalizer/visualizer approach ( Riding and Rayner, 1998 ). According to this framework, verbalizers tend to mentally represent information in words whereas visualizers (also called imagers) tend to mentally represent information in mental pictures or diagrams ( Riding and Sadler-Smith, 1992 ; Knoll et al., 2017 ). Subsequently, the developers of this framework argue that verbalizers better learn the material presented in text and images better learn the material presented in images ( Riding and Sadler-Smith, 1992 ). This is analogous to the visual and read/write learners in the VARK model. Importantly, both the VARK and the verbalizer/visualizer approach advocate matching the modality of the instruction to the students’ learning style.

Adapting instruction based on modality learning styles may be conflated with multimodal instruction. Multimodal instruction is providing information to students in more than one modality, such as a text with relevant pictures or diagrams ( Bouchey et al., 2021 ). The rationale for providing students with multiple modalities is grounded in dual coding in which visual and verbal information are processed in separate channels or pathways in the architecture of human cognition ( Paivio, 1991 ; Reed, 2006 ). Having information presented in two modalities (and subsequently two channels) allows for more information to be processed at a given time ( Mayer and Anderson, 1992 ; Mayer, 2011 ). Multimodal instruction has been found to benefit learning for students ( Mayer, 2017 ; Noetel et al., 2022 ). However, it should be noted there are individual differences in the degree of benefit, such as students with lower levels of background knowledge tend to have more benefit from adding visuals to verbal information compared to their peers with higher levels of background knowledge ( Mayer, 2017 ). This is distinct from learning styles in that certain students learn better than others in multimodal instruction, but there is not a crossover in which students receive harm or benefit from multimodal instruction. Individuals who support learning styles have been found to also support multimodal instruction ( Nancekivell et al., 2021 ). However, matching instruction to learning styles is more time-consuming as it involves assessing for styles and purposefully assigning modalities, rather than providing multiple options available for all students.

There are concerns that matching instruction to learning styles relates to psychological essentialism, which is the belief that categories of people are innate and biologically based ( Gelman, 2003 ; Nancekivell et al., 2020 ). An essentialist view of learning styles would be that, for example, visual learners are born with a predisposition to learning visually and that this limits what they can learn through other modalities. Indeed, essentialist and non-essentialist believers in learning styles have been identified ( Nancekivell et al., 2020 ). Essentialist belief in learning styles may explain why visual learners are perceived as more intelligent and better performing academically than kinesthetic, “hands on,” learners ( Sun et al., 2023 ). Relatedly, learners who are told they have a particular style may have a self-fulfilling prophecy in which they believe they can only learn in a particular modality and subsequently do not develop necessary skills in modalities outside of their style ( Vasquez, 2009 ).

Given the resources involved and potential consequences relevant to psychological essentialism, learning styles would logically need to demonstrate remarkable efficacy to justify their use in education. In a review of learning styles efficacy, a team of cognitive scientists focused on student learning explained the criteria for validating the matching or meshing hypothesis ( Pashler et al., 2008 ). One is to categorize learners based on a measure of learning style into at least two groups. A second is that participants need to be randomly assigned to receive instruction in a minimum of two methods (e.g., visual compared to auditory information). Learners need to be assessed in the same manner across styles and conditions. Finally, there needs to be a crossover in which there is an interaction between the learning style group and instruction in which matched instruction has higher learning gains than unmatched instruction for each of the learning style groups. This avoids the possibility that the instruction intended to be matched for a particular style is simply better across style groups. For example, college students who were prompted to visualize statements (visual matching) remembered more statements than their peers who were prompted to consider the sounds in the statements (auditory matching) across learning style categories ( Cuevas and Dawson, 2018 ).

The review by Pashler et al. (2008) concluded that there was a lack of empirical evidence to support matching instruction to student’s learning styles that met their criteria for validating the matching hypothesis. Since this time, there have been other reviews similarly concluding that there is a lack of empirical support for matching instruction to students’ learning styles ( Cuevas, 2015 ; Klitmøller, 2015 ; Aslaksen and Lorås, 2018 ). However, there has not been a meta-analysis aggregating effects across studies to provide an estimate of magnitude. Such an approach provides more precision that can be deduced from individual studies and more power to detect effects that may be provided by a single study sample ( Deeks et al., 2023 ). Moreover, meta-analyses may help resolve controversies based on conflicting study findings ( Deeks et al., 2023 ).

Potential moderators

The modality of instruction for matching to learning styles should be considered when considering effects. For example, verbalizer or read/write learners may have their matched instruction involve reading and auditory learners would receive the same information aurally (e.g., Rogowsky et al., 2015 , 2020 ; Lehmann and Seufert, 2020 ). However, reading comprehension is somewhat better than listening comprehension for inferential understanding in which readers need to connect ideas from the text ( Clinton-Lisell, 2022 ). However, listening may be more effective than reading when accompanied by relevant visual representations, such as pictures or diagrams ( Noetel et al., 2022 ). In addition, non-verbal images (pictures) tend to be remembered better than the same information presented in words ( Paivio and Csapo, 1973 ).

The modality of the assessment should be considered as a potential moderator. Pashler et al.’s (2008) criteria understandably require the learning assessment to be the same modality in order to make comparisons between matched and unmatched instructions based on learning styles. However, this typically involves one method of instruction being in the same modality as the assessment and the comparison method of instruction being in a modality different than the assessment. For example, a listening task would be considered matched for auditory learners and a reading task would be considered matched for read/write or verbal learners and the assessment would be in writing, which is the same modality as the matched instruction for read/write or verbal learners. It is possible that there is a modality-match effect in which having the same modality at learning and assessment would affect results ( Mulligan and Osborn, 2009 ). Encoding and producing information in the same modality may be easier than in different modalities ( Staudigl and Hanslmayr, 2019 ), and subsequently, whether the instruction modality and assessment modality were the same should be considered.

Experimental studies comparing instructions matching and unmatched to learning styles have been conducted with between-subjects and within-subjects designs. With a between-subjects design, participants are in separate groups and only experience one condition. In the case of learning styles, participants would be placed in a group to either receive instruction matched or unmatched to their categorized learning style. An advantage to between-subject designs is that participants are unaware of conditions they were not assigned to thereby preventing carryover effects from other conditions as well as practice effects ( Charness et al., 2012 ). However, different individuals are compared by condition, and subsequently, prior group differences could confound effects thought to be due to condition ( Gray et al., 2003 ; Adesope et al., 2017 ). In contrast, a within-subjects design involves participants experiencing both instruction matched and instruction unmatched to their learning style with different materials and counterbalanced to prevent order effects. With a within-subjects design, each participant serves as their own control, which prevents prior group differences at baseline to confound results ( Charness et al., 2012 ). Because these research designs are comparable, but not identical, it is recommended that the study design be tested as a moderator in meta-analyses ( Borenstein et al., 2009 ).

Study quality is an important consideration in meta-analyses as it is possible for treatment effects to vary as a function of study quality ( Sterne et al., 2001 ; Feeley, 2020 ). However, removing low-quality studies from analyses may lead to missing valuable data and meta-analyses should strive to be as inclusive as possible to have an accurate understanding of the accumulated evidence ( Weaver, 2011 ; Feeley, 2020 ). Narrow inclusion criteria themselves may bias meta-analytic findings. Moreover, studies in social sciences and education (such as the ones for the current meta-analysis) tend to receive low-quality ratings due to methodological details (particularly internal consistency) not being reported ( Singer and Alexander, 2017 ; Feeley, 2020 ). However, the potential influence of study quality should be considered by coding the quality of each study using predetermined quality criteria and including study quality as a moderator to assess its potential contribution to varying effects ( Pigott and Polanin, 2020 ; see Austin et al., 2019 ; Zhu et al., 2021 ; Lam and Zhou, 2022 ).

The current study

The purpose of this study was to conduct a meta-analysis of matching instruction to modality learning styles. In doing so, the criteria from Pashler et al. (2008) are generally followed. One exception is that non-randomized quasi-experiments are included given the valuable information they provide due to their external validity in education research ( Waddington et al., 2022 ). Three research questions guide this inquiry:

1. What is the aggregated effect of matching instruction to learning styles compared to unmatched instruction on learning outcomes?

2. How frequent is the crossover of matching instruction by style? In other words, is there an interaction indicating benefits to matched instruction over unmatched instruction for at least two of the styles examined?

3. Does the study design (between or within subjects), type of styles, modality of instruction, or study quality moderate the effects of matching instruction to learning styles?

The data extracted from the included studies and R code used for analyses are available on the Open Science Framework ( Clinton-Lisell, 2023b ).

Author positionality

Following guidance from Castillo and Babb (2024) , information about the authors’ backgrounds and identities is shared in this section.

The first author learned a cognitive approach to educational psychology during her doctoral and postdoctoral studies. During these times, she was taught that there was a lack of empirical evidence to support the concept of learning styles and that it was a common myth of education. Furthermore, she is aware that learning styles have origins rooted in ethnocentrism in which white scholars developed the concept based on condescending attitudes toward children of color ( Fallace, 2019 ). As a white woman, this is a history she works to be mindful of not repeating.

The second author has a master’s degree in school counseling and was influenced by behavioral and school counseling theories. The career aspects of school counseling education supported the use of learning style inventories during the time she received her training. As a researcher, the second author became aware of learning styles research that did not support the career education practices being utilized in the educational setting. The second author shifted practices in her work away from using learning style inventories as part of career education because of the current research on the topic. As a white woman and first-generation college student, she works to be mindful of the social/cultural underpinnings that could influence the understanding of research in learning styles.

Inclusion criteria

Following Pashler et al.’ (2008) guidelines, studies for the learning styles meta-analysis were included if they met the following criteria: (1) participants were categorized in at least two types of learning styles (e.g., visual and auditory), (2) there was at least one condition with instruction and/or learning materials matching to the participants’ learning styles and at least one condition with instruction and/or learning materials not matching to the participants’ learning styles, (3) the unmatched condition for one type of learning style was considered a matched condition for another learning style (so that a crossover interaction could be examined), (4) there was a measurement of learning that was identical across conditions and styles, (5) the study was disseminated in English because of the linguistic skills of the research team, and (6) descriptive statistics were reported to calculate effect sizes or the author of the study provided these upon request.

Systematic search

The first step in the systematic search for relevant articles included a broad search of the databases Web of Science, Scopus, PsycInfo/EBSCOhost, ERIC, and ProQuest Dissertations and Theses using the search terms such as “learning style*” and “learning preference*.” Dissertations and theses were important to include in the search as they are less likely to be influenced by publication bias in which journal articles are more likely to get published when reporting significant results ( Paez, 2017 ). A total of 6,299 citations were found (see Figure 1 for a flow chart of the systematic search process). After duplicates were removed, 1,810 remained. These citations were screened based on titles and abstracts by at least two researchers working independently using Abstrackr ( Wallace et al., 2012 ). Based on this screening, 40 reports were selected for full-text screening. Of these studies, 12 were selected for inclusion (see Figure 1 for reasons for exclusion). A backwards search of the citations in these 12 reports was conducted but did not yield additional studies. A forwards search of the 12 reports yielded an additional 8 reports. The citations of previous reviews were examined ( Pashler et al., 2008 ; Cuevas, 2015 ; Aslaksen and Lorås, 2018 ; Dinsmore et al., 2022 ), which yielded one more report. This led to a total of 21 reports of 21 independent studies in this meta-analysis.

Figure 1 . PRISMA flow chart of the systematic review process.

To prepare the studies for analyses, two researchers coded the methodological and bibliographic information about each study (see Table 1 ; κ  = 0.89). Specifically, the sample, study design, learning styles examined, measures of learning, content of instruction/materials, and assessment were recorded to describe studies (see Appendix A for codebook). Study quality was determined based on What Works Clearinghouse (2022) criteria and categorized as meeting standards, meeting standards with reservations, or not meeting standards (see Appendix A for details). Based on these standards, a study must be a randomized experiment to meet standards (although not all randomized experiments meet standards). In randomized experiments, the chance of students being the control or intervention should be equal. In contrast, quasi-experiments involve naturally occurring groups, typically classes in educational research, or controls matched through propensity score matching or regression discontinuity design. Whether a study had randomization (for between subjects) or counterbalancing (for within subjects) is noted in the summary of studies in Table 1 . Other study quality criteria such as the face validity for each outcome measure, reliability standards for each outcome measure, and whether there was consistent data collection across conditions are reported in Appendix Table B1 . As can be seen in Appendix Table B2 , five studies were determined to WWC standards and the remainder did not meet WWC standards.

Table 1 . Summary of studies.

Statistical procedures

The effect sizes for each learning outcome comparing matched and unmatched instruction were calculated. Hedges’ g was used as an effect size calculated using Meta-Essential tools ( Suurmond et al., 2017 ). A positive Hedges’ g indicates better learning outcomes for matched than unmatched instruction. To account for multiple effect sizes within each study, a robust variance estimation (RVE) was used. An RVE is a statistical technique that accounts for dependencies within studies while still allowing for the unique contribution of each effect size to be considered ( Tanner-Smith et al., 2016 ). Each of the study effect sizes is shown in Table 2 , and a forest plot is in Figure 2 . Learning outcomes indicating a crossover interaction as articulated in Pashler et al. (2008) in which at least two styles had higher learning outcomes with matched instruction are bolded in Table 2 .

Table 2 . Effect sizes with variances and number of participants.

Figure 2 . Forest plot of effect sizes.

The overall main effect of matching instruction to learning styles on learning outcomes was estimated using RVE based on 21 studies and 101 effect sizes and assumed dependency (intercorrelation of dependent effects within studies) of ρ = 0.8. The findings indicated an overall positive effect on learning outcomes for matching instruction to learning styles compared to unmatched instruction, g  = 0.32, SE = 0.12, 95% CI = [0.07, 0.57], p  = 0.01. There was substantial variability with a τ 2 of 0.77 and I 2 of 91.17. A sensitivity analysis was conducted with a range of dependent effect size correlations. As can be seen in Table 3 , the effect was consistent across assumed dependencies.

Table 3 . Sensitivity analyses for the assumed dependency of effect sizes.

Publication bias

Publication bias was examined to see whether there was overreporting of positive effects. A funnel plot was generated using the “metafor” package in R ( Viechtbauer, 2010 ; see Figure 3 ). Based on a visual inspection of the funnel plot, the distribution of effect sizes was approximately symmetrical with smaller and larger studies having similar distances away from the mean (indicated by the vertical line; Lin and Chu, 2018 ). Egger’s test of the intercept was not significant, b  = −0.058, 95% CI [−0.52, 0.41], p  = 0.11. Taken together, the funnel plot and Egger’s test indicate that publication bias does not appear to be the reason for the positive effect of matching instruction to learning styles.

Figure 3 . Funnel plot of effect sizes.

Crossover interactions

The number of crossover interactions (at least two styles benefited from matched instruction within a learning outcome measure) was calculated. Based on the tally of the bolded effect sizes in Table 2 , there are 11 learning outcome measures in which matched instruction benefited at least 2 learning styles as indicated by Hedges’ g greater than 0. This was out of a total of 42 learning outcome measures that were compared indicating that 26.19% had the type of crossover interaction necessary to support the meshing hypothesis as articulated by Pashler et al. (2008) .

As indicated in Figure 1 , five that had their full texts screened did not have sufficient statistics to calculate the effect sizes reported. In addition, seven reports identified through other searches did not have sufficient statistics to calculate effect sizes but otherwise met inclusion criteria. Based on the descriptions of the findings of these 12 studies in Tables 3 , 4 of these studies indicated a crossover interaction (25.00%). Therefore, the findings from the studies without sufficient statistics reported appear to be similar to the studies included in the meta-analysis in terms of crossover interactions.

Table 4 . Studies without sufficient descriptive statistics reported to calculate effect sizes.

Moderator analysis

To estimate whether these potential moderators varied the effect of matching instruction to learning style, the package “robumeta” in R was used ( Fisher and Tipton, 2015 ). The study design (between or within subjects), modality of instruction/materials (visual, verbal, or auditory), whether the assessment was in the same modality as the instruction, and study quality (does not meet WWC standards or meets WWC standards) were all coefficients estimated in the meta-regression model. For consistency across studies, “visual” matched instruction that was text-based was coded as “verbal or read/write.” Based on the output of the meta-regression model, none of the moderators were significant (see Table 5 ). Therefore, it is unclear what the source of variability in the aggregate findings is.

Table 5 . Meta-regression model.

Sensitivity analysis

A sensitivity analysis was conducted to examine whether altering the inclusion criteria changed the results. There was only one study in which instruction was adopted to a kinesthetic learning style ( Papanagnou et al., 2016 ). Removing this study from the RVE analyses indicated an overall aggregated benefit of matched instructions to learning styles compared to unmatched, g  = 0.34, SE = 0.13, 95% CI = [0.08, 0.60], p  = 0.01 with 20 studies and 98 effect sizes. This finding is similar to the findings when the kinesthetic learning intervention was included. There were two studies that were quasi-experiments without random assignment ( Moussa-Inaty et al., 2019 ; Mujtaba et al., 2022 ). This was a concern given that they did not demonstrate baseline equivalence in the study quality coding (see Appendix Table B1 ). Therefore, an RVE was conducted with the two quasi-experiments removed from the analyses. The results of the RVE were similar to the quasi-experiments removed in that there was an overall aggregated benefit of matched instruction to learning styles compared to unmatched, g  = 0.33, SE = 0.13, 95% CI = [0.05, 0.61], p  = 0.02 with 19 studies and 91 effect sizes.

Educational researchers consider the concept of better learning through matching instruction to learning styles to be a neuromyth that completely lacks empirical evidence ( Brown, 2023 ). However, the findings from this meta-analysis indicated a small, but statistically reliable benefit of matching instruction based on learning styles. This aligns with the majority of educators’ perspectives ( Dekker et al., 2012 ; Nancekivell et al., 2020 ; Eitel et al., 2021 ) but conflicts with the conclusions of previous reviews by educational researchers ( Cuevas, 2015 ; An and Carr, 2017 ; Aslaksen and Lorås, 2018 ; Dinsmore et al., 2022 ; Yan and Fralick, 2022 ). What distinguishes this meta-analysis from previous reviews is (1) its singular focus on studies comparing instruction matched and unmatched to modality learning styles and (2) its systematic approach to gathering relevant studies and aggregating findings. The lack of evidence noted in previous reviews may be due to a lack of power in individual studies. The cumulative evidence of aggregated effects appeared to have sufficient power to detect an effect. However, the majority of learning outcomes did not indicate a crossover interaction that would validate accommodation to learning styles. However, a non-trivial minority of learning outcomes did indicate the crossover interaction indicative of supporting the matching hypothesis based on Pashler et al. (2008) . An important caveat is that most of the studies indicating a crossover interaction did not meet quality standards as determined by the What Works Clearinghouse (2022) . Taken together, these findings may be interpreted that it is too much of an overreach to insist learning styles should be incorporated into instructional practices.

Given the time and resources required for matching instruction to learning styles coupled with the potential for harm through psychological essentialism ( Vasquez, 2009 ; Fallace, 2019 , 2023a , b ; Nancekivell et al., 2020 ; Sun et al., 2023 ), we stated in the literature review that accommodating instruction to learning styles would need to have substantial benefits to mitigate their potential for harm. To consider this issue, it may be helpful to compare the effect size noted in this meta-analysis (Hedges’ g  = 0.32) with those from other methods of adapting instruction. For example, the modality effect , in which listening to verbal information while viewing visual representations, rather than reading the same verbal information alongside visuals has Hedges’ g of 0.70 ( Noetel et al., 2022 ); that is, the benefit of listening, rather than reading verbal information that accompanies visual representations across all students, appears to have twice the effect than was noted in this meta-analysis matching instruction to learning styles and would be less time-consuming and expensive to implement. Removing interesting or irrelevant information included with the lesson has Hedges’ g of 0.33 ( Sundararajan and Adesope, 2020 ). Segmenting instruction into meaningful learner-paced units has a benefit of Hedges’ g of 0.32 compared to continuous information ( Rey et al., 2019 ). Finally, an overall application of multimedia principles to learning has Hedge’s g of 0.28 ( Noetel et al., 2022 ).

When examining an overview of meta-analysis on multimedia design for learning, accommodating instruction based on learning styles in the current meta-analysis is generally about the same size or smaller than various multimedia designs (e.g., signaling important information, animation, and pleasant colors/anthropomorphic; Noetel et al., 2022 ). However, all of the multimedia design principles reviewed involved having students each receive the same instructional changes, whereas accommodating instruction based on learning styles by definition involves at least two types of instruction ( Noetel et al., 2022 ). In addition, 85% of the studies in the current meta-analysis did not include all participants in the final sample because their learning styles scores did not allow for confident categorization and matching/unmatching to instructional modality. Therefore, we, the authors, deeply question whether the found benefits of learning styles in this meta-analysis warrant accommodating instruction, especially for widespread use. Based on previous studies, well-structured instructional design may be more effective across all students and would involve less time than accommodating to learning styles.

Participant expectations may be relevant to interpreting the findings from the studies in this meta-analysis ( Vasquez, 2009 ; Sun et al., 2023 ). Generally, participants were asked about their modality preferences and then engaged in a learning activity shortly thereafter (only three studies specifically stated participants were asked to complete learning style measures after the learning activity; see Table 1 ). If participants were aware of their learning styles prior to engaging in a task that matched or unmatched their style, they may have had different expectations for success and engagement that affected their learning ( Vasquez, 2009 ). For example, one study categorized students based on fake/induced learning styles ( Moser and Zumbach, 2015 ). In other words, students took a learning styles assessment and were told (incorrectly) that they scored as visualizers or verbalizers. Students scored higher when their instructional materials “matched” their fake/induced learning style compared to the unmatched conditions, but there were no benefits to matching based on their actual categorizations based on the learning styles assessment ( Moser and Zumbach, 2015 ). This is described in the learning styles genesis model in which appraisal and decision processes based on external feedback about learning styles along with previous experiences with modalities shape learning outcomes ( Moser and Zumbach, 2018 ). Moreover, participants may have had a situational interest in the content triggered by immediately receiving instruction in their stated preferences ( Bernacki and Walkington, 2018 ). This likely would not continue long term as maintained interest requires a personal connection ( Høgheim and Reber, 2017 ).

Learning styles may be conflated with modality-specific skills ( An and Carr, 2017 ). In other words, participants may simply be better at reading if they indicate a read/write style or listening if they indicate an auditory style. This results in a jangle fallacy in which two similar constructs (e.g., modality skill and learning styles) are considered different because they have different terminology ( Kelley, 1927 ; Beisley, 2023 ). It should be noted, however, that if this is the case, skills in a modality may have been developed because of preferences in that modality, which would, in turn, lead to more practice and more skill in a particular modality. It would be extremely difficult to disentangle the initiating factor in this (hypothetical) perpetual cycle of skill and preference. However, more inquiry into fake/induced learning styles such as that by Moser and Zumbach (2015) would be a useful means of testing whether the skill is confounded with style given that fake/induced styles would be randomly assigned and subsequently modality skills should be similar across “styles.”

A key feature of the matching hypothesis is a crossover interaction in which matched instruction benefits learning for only the group for which it is matched. This matched instruction differs depending on the learning style of the student. In the current meta-analysis, approximately one-fourth of the learning measures indicated a crossover interaction in which there were positive effect sizes for matched instruction for two different learning styles ( Kassaian, 2007 ; Chen and Sun, 2012 ; Hazra et al., 2013 ; Kam et al., 2020 ; Lehmann and Seufert, 2020 ; Chui et al., 2021 ; Tadayonifar et al., 2021 ). This raises the question of what characteristics of these studies and learning measures may be responsible for the crossover interaction. However, these studies are quite heterogeneous. Samples include young adult college students, elementary school students, and aircraft pilot trainees. The learning styles and their inventories varied and included the Styles of Processing Scale ( Childers et al., 1985 ), VARK questionnaire ( Fleming, 2001 ), Index of Learning Styles Scores ( Felder and Soloman, 1997 ), and Caption Reliance Test ( Leveridge and Yang, 2014 ). Content and learning measures were also varied such as memory and recall of history ( Hazra et al., 2013 ), multiple-choice questions about energy education ( Chen and Sun, 2012 ), and flight simulator performance ( Chui et al., 2021 ). Therefore, there does not seem to be any consistent feature across these studies based on the information coded for this meta-analysis that would elucidate the mechanism behind the crossover interaction. Subsequently, the lack of understanding of what circumstances could foster a crossover interaction is an additional reason for caution in implementing the matched instruction for learning styles. Without knowing what features are conducive to effective matched instruction, it is extremely difficult to have effectively matched instruction across identified learning styles.

Implications

We advise extreme caution if using the findings from this meta-analysis to justify matching instruction to learning styles. If choosing to incorporate learning styles, then learning styles should never be ascribed as a feature of a cultural group, especially by individuals outside of that group, as this leads to unwarranted and potentially harmful expectations based on group membership ( Gutiérrez and Rogoff, 2003 ; Fallace, 2023a , b ). Moreover, learning style interventions are costly in terms of both time and money ( Pashler et al., 2008 ). By definition, matching instruction based on learning styles requires multiple versions of instruction or materials to be developed.

If learning styles are incorporated into education, we strongly recommend that they be implemented in the context of multimodality for learning. By providing information in more than one modality, such as text with visuals, the same materials could arguably appeal to both verbal and visual learning styles while grounded in theories of human cognition such as dual coding ( Noetel et al., 2022 ). Engaging multiple senses is generally beneficial for learning ( Nguyen et al., 2022 ). In addition, providing students with audio-assisted text may also be beneficial, particularly for learning beyond one’s native language ( Clinton-Lisell, 2023a ), and logically appeal to auditory and verbal preferences. Not only is multimodality known to be effective for learning but even individuals with strong essentialist beliefs about learning styles support multimodal learning as effective ( Nancekivell et al., 2021 ). Moreover, offering multiple modalities for learning provides an inclusive education for students with perceptual disabilities to have access to the content ( Thomas et al., 2015 ; Griful-Freixenet et al., 2017 ).

Limitations and future directions

Limitations to the studies were included in the meta-analysis. As indicated in the study quality coding, the majority of the outcome measures did not have reliability metrics reported. The lack of information about reliability, as noted in the study quality scoring, leads to challenges in determining the validity of the findings. Indeed, the primary issue with study quality is due to an inability to assess reliability due to a lack of reporting across multiple studies. Unfortunately, a lack of reporting reliability statistics is a common issue across multiple social science and education disciplines ( Barry et al., 2014 ; Lovejoy et al., 2014 ; Han, 2016 ; Parsons et al., 2019 ; Flake, 2021 ). This illustrates the need to ensure that reliability is reported throughout the peer review and publication process. Indeed, publication reporting standards in psychology, through the American Psychological Association ( Appelbaum et al., 2018 ), state that the reliability of measures should be reported.

The studies were all single sessions in duration and subsequently claims about long-term effects cannot be determined from the meta-analysis. Moreover, there was substantial variation in the findings across outcomes that was not explained in the meta-regression. This could be due to insufficient power to identify moderators in the meta-regression analyses ( Schmidt, 2017 ). Furthermore, the studies were limited to those disseminated in English due to the linguistic limitations of the research team. It is possible the inclusion of more languages would have led to different outcomes. In addition, all but two of the reviewed studies were from journal articles. Although the publication analyses did not indicate publication bias, it is still an issue to consider given that only two studies were from the gray literature in which non-significant findings are more likely to be reported ( Cairo et al., 2020 ). There is also possible bias when considering studies as several authors were contacted with requests for data to calculate effect sizes, but only some of the authors provided this information. There may be response bias regarding the findings that were calculated based on author-provided data. However, it should be noted that authors frequently do not respond to requests for data ( Tedersoo et al., 2021 ).

The studies in this meta-analysis all categorized their participants based on learning styles, but the methods of categorization varied. There were a range of measures used and the cutoff for categorization of learning styles differed by study as well. This makes the generalizability of the findings challenging. Moreover, there was substantial variability in the outcome measures. Only 21 studies were identified that met the criteria for testing the matching hypothesis and reported sufficient statistics to conduct effect sizes. In particular, there were not enough studies to examine whether having the learning styles assessment before or after the learning activity varied the findings. This is unfortunate given the concerns about self-fulfilling prophecies and findings from Moser and Zumbach (2015) with fake, induced learning styles.

Learning styles are a controversial topic in education. In this meta-analysis, we sought to inform the controversy with aggregated findings based on a comprehensive search for studies. An overall small, positive effect was noted. However, this aggregated effect should be interpreted with caution given that most studies did not indicate a crossover interaction. Such a crossover interaction would have been necessary to support the claim that matching instruction to learning styles benefits students from different learning styles. Given the high amount of variability in the findings and infrequent crossover interactions, it is far from conclusive that there is actually a benefit to matching the modality of instruction to students’ learning styles. Teaching with multiple modalities may be preferable to the costly and labor-intensive practice of matching instruction to learning styles given the empirical evidence for benefits across students for multimodal instruction.

Data availability statement

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://osf.io/5heum/ .

Author contributions

VC-L: Conceptualization, Data curation, Funding acquisition, Supervision, Writing – original draft, Writing – review & editing. CL: Conceptualization, Data curation, Writing – original draft, Writing – review & editing.

The author(s) declare that financial support was received for the research, authorship, and/or publication of this article. We thank the University of North Dakota Alumni Foundation for supporting this research by VC-L through the Rose Isabella Kelly Fischer Professorship.

Acknowledgments

We thank Maylynn Riding In for her assistance screening titles and abstracts.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg.2024.1428732/full#supplementary-material

* included in meta-analysis

Google Scholar

Adesope, O. O., Trevisan, D. A., and Sundararajan, N. (2017). Rethinking the use of tests: a meta-analysis of practice testing. Rev. Educ. Res. 87, 659–701. doi: 10.3102/0034654316689306

Crossref Full Text | Google Scholar

An, D., and Carr, M. (2017). Learning styles theory fails to explain learning and achievement: recommendations for alternative approaches. Personal. Individ. Differ. 116, 410–416. doi: 10.1016/j.paid.2017.04.050

Appelbaum, M., Cooper, H., Kline, R. B., Mayo-Wilson, E., Nezu, A. M., and Rao, S. M. (2018). Journal article reporting standards for quantitative research in psychology: the APA publications and communications board task force report. Am. Psychol. 73, 3–25. doi: 10.1037/amp0000191

PubMed Abstract | Crossref Full Text | Google Scholar

Aslaksen, K., and Lorås, H. (2018). The modality-specific learning style hypothesis: a mini-review. Front. Psychol. 9:1538. doi: 10.3389/fpsyg.2018.01538

*Aslaksen, K., and Lorås, H. (2019). Matching instruction with modality-specific learning style: effects on immediate recall and working memory performance. Educ. Sci. , 9:32. doi: 10.3390/educsci9010032

Austin, C. R., Wanzek, J., Scammacca, N. K., Vaughn, S., Gesel, S. A., Donegan, R. E., et al. (2019). The relationship between study quality and the effects of supplemental reading interventions: a meta-analysis. Except. Child. 85, 347–366. doi: 10.1177/0014402918796164

Bareither, M. L., Arbel, V., Growe, M., Muszczynski, E., Rudd, A., and Marone, J. R. (2013). Clay modeling versus written modules as effective interventions in understanding human anatomy. Anat. Sci. Educ. 6, 170–176. doi: 10.1002/ase.1321

Barry, A. E., Chaney, B., Piazza-Gardner, A. K., and Chavarria, E. A. (2014). Validity and reliability reporting practices in the field of health education and behavior: a review of seven journals. Health Educ. Behav. 41, 12–18. doi: 10.1177/1090198113483139

Beisley, A. H. (2023). The jingle-jangle of approaches to learning in prekindergarten: a construct with too many names. Educ. Psychol. Rev. 35:79. doi: 10.1007/s10648-023-09796-4

Bernacki, M. L., and Walkington, C. (2018). The role of situational interest in personalized learning. J. Educ. Psychol. 110:864. doi: 10.1037/edu0000250

Borenstein, M., Hedges, L. V., Higgins, J. P. T., and Rothstein, H. R. (2009). Introduction to meta-analysis . Hoboken, New Jersey, USA: John Wiley & Sons, Ltd.

Bouchey, B., Castek, J., and Thygeson, J. (2021). “Multimodal learning” in Innovative learning environments in STEM higher education: Opportunities, challenges, and looking forward . eds. J. Ryoo and K. Winkelmann (Cham, Switzerland: Springer International Publishing), 35–54.

Brown, S. B. R. E. (2023). The persistence of matching teaching and learning styles: a review of the ubiquity of this neuromyth, predictors of its endorsement, and recommendations to end it. Front. Educ. 8:1147498. doi: 10.3389/feduc.2023.1147498

*Burns, L. (n.d.). Visual versus auditory processing preference and mode of presentation: Differences in confidence, attention, and recall performance in online learning. Undergraduate capstone project at the University of Missouri-Columbia. Available at: https://osf.io/5heum/?view_only=510f56642e974fbeb000dc8a8b214b68

Cairo, A. H., Green, J. D., Forsyth, D. R., Behler, A. M. C., and Raldiris, T. L. (2020). Gray (literature) matters: evidence of selective hypothesis reporting in social psychological research. Personal. Soc. Psychol. Bull. 46, 1344–1362. doi: 10.1177/0146167220903896

Calcaterra, A., Antonietti, A., and Underwood, J. (2005). Cognitive style, hypermedia navigation and learning. Comput. Educ. 44, 441–457. doi: 10.1016/j.compedu.2004.04.007

Castillo, W., and Babb, N. (2024). Transforming the future of quantitative educational research: a systematic review of enacting QuantCrit. Race Ethn. Educ. 27, 1–21. doi: 10.1080/13613324.2023.2248911

Charness, G., Gneezy, U., and Kuhn, M. A. (2012). Experimental methods: between-subject and within-subject design. J. Econ. Behav. Organ. 81, 1–8. doi: 10.1016/j.jebo.2011.08.009

*Chen, C.Y. (2020). The influence of representational formats and learner modality preferences on instructional efficiency using interactive video tutorials. J. Educ. Train. 7, 1–17. doi: 10.5296/jet.v7i2.17415

*Chen, C.M., and Sun, Y.C. (2012). Assessing the effects of different multimedia materials on emotions and learning performance for visual and verbal style learners. Comput. Educ. , 59, 1273–1285. doi: 10.1016/j.compedu.2012.05.006

Childers, T. L., Houston, M. J., and Heckler, S. E. (1985). Measurement of individual differences in visual versus verbal information processing. J. Consum. Res. 12, 125–134. doi: 10.1086/208501

Chislet, V., and Chapman, A. (2005). VAK learning styles self-test .

*Chui, T. K., Molesworth, B. R., and Bromfield, M. A. (2021). Feedback and student learning: matching learning and teaching style to improve student pilot performance. Int. J. Aerospace Psychol. , 31, 71–86. doi: 10.1080/24721840.2020.1847650

Clinton-Lisell, V. (2022). Listening ears or reading eyes: a meta-analysis of reading and listening comprehension comparisons. Rev. Educ. Res. 92, 543–582. doi: 10.3102/00346543211060871

Clinton-Lisell, V. (2023a). Does reading while listening to text improve comprehension compared to reading only? A systematic review and meta-analysis. Educ. Res. Theory Pract. 34, 133–155.

Clinton-Lisell, V. (2023b). Learning styles meta-analysis. Available at: https://osf.io/5heum/?view_only=510f56642e974fbeb000dc8a8b214b68

Cohen, A.D., Oxford, R.L., and Chi, J.C. (2019) Learning style survey: Assessing your own learning styles. Available at: http://carla.umn.edu/maxsa/documents/LearningStyleSurvey_MAXSA_IG.pdf

Cuevas, J. (2015). Is learning styles-based instruction effective? A comprehensive analysis of recent research on learning styles. Theory Res. Educ. 13, 308–333. doi: 10.1177/1477878515606621

*Cuevas, J., and Dawson, B. L. (2018). A test of two alternative cognitive processing models: learning styles and dual coding. Theory Res. Educ. , 16, 40–64. doi: 10.1177/1477878517731450

Deeks, J. J., Higgins, J. P. T., and Altman, D. G. (2023). “Chapter 10: analysing data and undertaking meta-analyses” in Cochrane handbook for systematic reviews of interventions version 6.4 . eds. J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, and M. J. Page (Cochrane Statistical Methods Group).

Dekker, S., Lee, N., Howard-Jones, P., and Jolles, J. (2012). Neuromyths in education: prevalence and predictors of misconceptions among teachers. Front. Psychol. 3:429. doi: 10.3389/fpsyg.2012.00429

Dinsmore, D. L., Fryer, L. K., and Parkinson, M. M. (2022). The learning styles hypothesis is false, but there are patterns of student characteristics that are useful. Theory Pract. 61, 418–428. doi: 10.1080/00405841.2022.2107333

Dunn, R. (1990). Understanding the Dunn and Dunn learning styles model and the need for individual diagnosis and prescription. J. Read. Writ. Learn. Disabil. Int. 6, 223–247. doi: 10.1080/0748763900060303

Dunn, R., and Dunn, K. (1975). Educator’s self-teaching guide to indiviudalizing instructional programs . Englewood Cliffs, New Jersey, USA: Parker Publishing Company.

Eitel, A., Prinz, A., Kollmer, J., Niessen, L., Russow, J., Ludäscher, M., et al. (2021). The misconceptions about multimedia learning questionnaire: An empirical evaluation study with teachers and student teachers. Psychol. Learn. Teach. 20, 420–444. doi: 10.1177/14757257211028723

Fajari, L. E., Sarwanto,, and Chumdari,. (2020). Improving elementary school’s critical thinking skills through three different learning media viewed from learning styles. J. E Learn. Knowledge Soc. 16, 55–65. doi: 10.20368/1971-8829/1135193

Fallace, T. (2019). The ethnocentric origins of the learning style idea. Educ. Res. 48, 349–355. doi: 10.3102/0013189X19858086

Fallace, T. (2023a). Herman Witkin and the rise and fall of the black learning style idea, 1960–2003. Teach. Coll. Rec. 125, 67–94. doi: 10.1177/01614681231178589

Fallace, T. (2023b). The long origins of the visual, auditory, and kinesthetic learning style typology, 1921–2001. Hist. Psychol. 26, 334–354. doi: 10.1037/hop0000240

Feeley, T. H. (2020). Assessing study quality in Meta-analysis. Hum. Commun. Res. 46, 334–342. doi: 10.1093/hcr/hqaa001

Felder, R. M., and Soloman, B. A. (1997). Index of learning styles questionnaire . Cochrane Statistical Methods Group.

Fiorina, L., Antonietti, A., Colombo, B., and Bartolomeo, A. (2007). Thinking style, browsing primes and hypermedia navigation. Comput. Educ. 49, 916–941. doi: 10.1016/j.compedu.2005.12.005

Fisher, Z., and Tipton, E. (2015). Robumeta: An R-package for robust variance estimation in meta-analysis. ArXiv Preprint ArXiv 1503:02220. doi: 10.48550/arXiv.1503.02220

Flake, J. K. (2021). Strengthening the foundation of educational psychology by integrating construct validation into open science reform. Educ. Psychol. 56, 132–141. doi: 10.1080/00461520.2021.1898962

Fleming, N. D. (2001). Teaching and learning styles. Christchurch, New Zealand: VARK strategies

Fleming, N. D., and Mills, C. (1992). Not another inventory, rather a catalyst for reflection. Improve Acad. 11, 137–155. doi: 10.1002/j.2334-4822.1992.tb00213.x

Fong, C. J., Krou, M. R., Johnston-Ashton, K., Hoff, M. A., Lin, S., and Gonzales, C. (2021). LASSI's great adventure: a meta-analysis of the learning and study strategies inventory and academic outcomes. Educ. Res. Rev. 34:100407. doi: 10.1016/j.edurev.2021.100407

*Ge, Z. G. (2021). Does mismatch between learning media preference and received learning media bring a negative impact on academic performance? An experiment with e-learners. Interact. Learn. Environ. , 29, 790–806. doi: 10.1080/10494820.2019.1612449

Gelman, S. A. (2003). The essential child: origins of essentialism in everyday thought . New York, NY, USA: Oxford University Press.

Gray, N. S., Snowden, R. J., Peoples, M., Hemsley, D. R., and Gray, J. A. (2003). A demonstration of within-subjects latent inhibition in the human: limitations and advantages. Behav. Brain Res. 138, 1–8. doi: 10.1016/s0166-4328(02)00181-x

Griful-Freixenet, J., Struyven, K., Verstichele, M., and Andries, C. (2017). Higher education students with disabilities speaking out: perceived barriers and opportunities of the universal Design for Learning framework. Disabil. Soc. 32, 1627–1649. doi: 10.1080/09687599.2017.1365695

Gutiérrez, K. D., and Rogoff, B. (2003). Cultural ways of learning: individual traits or repertoires of practice. Educ. Res. 32, 19–25. doi: 10.3102/0013189X032005019

Han, C. (2016). Reporting practices of rater reliability in interpreting research: a mixed-methods review of 14 journals (2004-2014). J. Res. Design Statistics Linguist. Commun. Sci. 3, 49–75. doi: 10.1558/jrds.29622

*Hazra, A.K., Patnaik, P., and Suar, D. (2013). Relation of modal preference with performance in adaptive hypermedia context: An exploration using visual, verbal and multimedia learning modules. 2013 IEEE fifth international conference on Technology for Education (T4e 2013), 163–166

Höffler, T. N., and Schwartz, R. N. (2011). Effects of pacing and cognitive style across dynamic and non-dynamic representations. Comput. Educ. 57, 1716–1726. doi: 10.1016/j.compedu.2011.03.012

Høgheim, S., and Reber, R. (2017). Eliciting mathematics interest: new directions for context personalization and example choice. J. Exp. Educ. 85, 597–613. doi: 10.1080/00220973.2016.1268085

Huang, T.-C. (2019). Do different learning styles make a difference when it comes to creativity? An empirical study. Comput. Hum. Behav. 100, 252–257. doi: 10.1016/j.chb.2018.10.003

*Kam, E. F., Liu, Y.-T., and Tseng, W.-T. (2020). Effects of modality preference and working memory capacity on captioned videos in enhancing L2 listening outcomes. ReCALL , 32, 213–230. doi: 10.1017/S0958344020000014

*Kassaian, Z. (2007). Learning styles and lexical presentation modes. Estud. Linguistica Inglesa Aplicada , 7, 53–78

Kelley, T. L. (1927). Interpretation of educational measurements . Chicago, Illinois, USA: World Book.

Kirschner, P. A. (2017). Stop propagating the learning styles myth. Comput. Educ. 106, 166–171. doi: 10.1016/j.compedu.2016.12.006

Klitmøller, J. (2015). Review of the methods and findings in the Dunn and Dunn learning styles model research on perceptual preferences. Nord. Psychol. 67, 2–26. doi: 10.1080/19012276.2014.997783

Knoll, A. R., Otani, H., Skeel, R. L., and Van Horn, K. R. (2017). Learning style, judgements of learning, and learning of verbal and visual information. Br. J. Psychol. 108, 544–563. doi: 10.1111/bjop.12214

Koć-Januchta, M. M., Höffler, T. N., Eckhardt, M., and Leutner, D. (2019). Does modality play a role? Visual-verbal cognitive style and multimedia learning. J. Comput. Assist. Learn. 35, 747–757. doi: 10.1111/jcal.12381

Koć-Januchta, M. M., Höffler, T. N., Prechtl, H., and Leutner, D. (2020). Is too much help an obstacle? Effects of interactivity and cognitive style on learning with dynamic versus non-dynamic visualizations with narrative explanations. Educ. Technol. Res. Dev. 68, 2971–2990. doi: 10.1007/s11423-020-09822-0

Kollöffel, B. (2012). Exploring the relation between visualizer–verbalizer cognitive styles and performance with visual or verbal learning material. Comput. Educ. 58, 697–706. doi: 10.1016/j.compedu.2011.09.016

Kraemer, D. J. M., Schinazi, V. R., Cawkwell, P. B., Tekriwal, A., Epstein, R. A., and Thompson-Schill, S. L. (2017). Verbalizing, visualizing, and navigating: the effect of strategies on encoding a large-scale virtual environment. J. Exp. Psychol. Learn. Mem. Cogn. 43, 611–621. doi: 10.1037/xlm0000314

Lam, K. K. L., and Zhou, M. (2022). Grit and academic achievement: a comparative cross-cultural meta-analysis. J. Educ. Psychol. 114, 597–621. doi: 10.1037/edu0000699

*Lehmann, J., and Seufert, T. (2020). The interaction between text modality and the learner’s modality preference influences comprehension and cognitive load. Front. Psychol. , 10:2820, doi: 10.3389/fpsyg.2019.02820

Leveridge, A. N., and Yang, J. C. (2014). Learner perceptions of reliance on captions in EFL multimedia listening comprehension. Comput. Assist. Lang. Learn. 27, 545–559. doi: 10.1080/09588221.2013.776968

Lin, L., and Chu, H. (2018). Quantifying publication bias in meta-analysis. Biometrics 74, 785–794. doi: 10.1111/biom.12817

Lovejoy, J., Watson, B. R., Lacy, S., and Riffe, D. (2014). Assessing the reporting of reliability in published content analyses: 1985–2010. Commun. Methods Meas. 8, 207–221. doi: 10.1080/19312458.2014.937528

Lyle, K. B., Young, A. S., Heyden, R. J., and McDaniel, M. A. (2023). Matching learning style to instructional format penalizes learning. Computers Edu. Open 5:100143. doi: 10.1016/j.caeo.2023.100143

Massa, L. J., and Mayer, R. E. (2006). Testing the ATI hypothesis: should multimedia instruction accommodate verbalizer-visualizer cognitive style? Learn. Individ. Differ. 16, 321–335. doi: 10.1016/j.lindif.2006.10.001

Mayer, R. E. (2011). “Chapter three—applying the science of learning to multimedia instruction” in Psychology of learning and motivation . eds. J. P. Mestre and B. H. Ross, vol. 55 (Cambridge, Massachusetts, USA: Academic Press), 77–108.

Mayer, R. E. (2017). Using multimedia for e-learning. J. Comput. Assist. Learn. 33, 403–423. doi: 10.1111/jcal.12197

Mayer, R. E., and Anderson, R. B. (1992). The instructive animation: helping students build connections between words and pictures in multimedia learning. J. Educ. Psychol. 84, 444–452. doi: 10.1037/0022-0663.84.4.444

Mayer, R. E., and Massa, L. J. (2003). Three facets of visual and verbal learners: cognitive ability, cognitive style, and learning preference. J. Educ. Psychol. 95, 833–846. doi: 10.1037/0022-0663.95.4.833

Moser, S., and Zumbach, J. (2015). Explicit and implicit measuring of visual and verbal learning styles. Paper presented at the European Association for Research in learning and instruction (EARLI) conference 2015, 25.-29.08.2015 in Limassol, Cyprus.

*Moser, S., and Zumbach, J. (2018). Exploring the development and impact of learning styles: An empirical investigation based on explicit and implicit measures. Comput. Educ. , 125, 146–157. doi: 10.1016/j.compedu.2018.05.003

*Moussa-Inaty, J., Atallah, F., and Causapin, M. (2019). Instructional mode: a better predictor of performance than student preferred learning styles. Int. J. Instr. , 12, 17–24. doi: 10.29333/iji.2019.1232a

*Mujtaba, M.S., Parkash, R., Kaur Mehar Singh, M., and Kamyabi Gol, A. (2022). The effect of computer-mediated feedback on L2 accuracy. Does the difference in learners’ perceptual style moderate the effectiveness of the feedback? Comput. Sch. , 39, 99–119. doi: 10.1080/07380569.2022.2041891

Mulligan, N. W., and Osborn, K. (2009). The modality-match effect in recognition memory. J. Exp. Psychol. Learn. Mem. Cogn. 35, 564–571. doi: 10.1037/a0014524

Nancekivell, S. E., Shah, P., and Gelman, S. A. (2020). Maybe they’re born with it, or maybe it’s experience: toward a deeper understanding of the learning style myth. J. Educ. Psychol. 112, 221–235. doi: 10.1037/edu0000366

Nancekivell, S. E., Sun, X., Gelman, S. A., and Shah, P. (2021). A slippery myth: how learning style beliefs shape reasoning about multimodal instruction and related scientific evidence. Cogn. Sci. 45:e13047. doi: 10.1111/cogs.13047

Newton, P. M., and Miah, M. (2017). Evidence-based higher education–is the learning styles ‘myth’ important? Front. Psychol. 8:444. doi: 10.3389/fpsyg.2017.00444

Nguyen, N. N., Mosier, W., Hines, J., and Garnett, W. (2022). Learning styles are out of style: shifting to multimodal learning experiences. Kappa Delta Pi Rec. 58, 70–75. doi: 10.1080/00228958.2022.2039520

Noetel, M., Griffith, S., Delaney, O., Harris, N. R., Sanders, T., Parker, P., et al. (2022). Multimedia design for learning: An overview of reviews with meta-meta-analysis. Rev. Educ. Res. 92, 413–454. doi: 10.3102/00346543211052329

Olsen, M. K., Stechuchak, K. M., and Steinhauser, K. E. (2019). Comparing internal and external validation in the discovery of qualitative treatment-subgroup effects using two small clinical trials. Contemp. Clin. Trials Commun. 15:100372. doi: 10.1016/j.conctc.2019.100372

Paez, A. (2017). Gray literature: An important resource in systematic reviews. J. Evid. Based Med. 10, 233–240. doi: 10.1111/jebm.12266

Paivio, A. (1991). Dual coding theory: retrospect and current status. Can. J. Psychol. 45, 255–287. doi: 10.1037/h0084295

Paivio, A., and Csapo, K. (1973). Picture superiority in free recall: imagery or dual coding? Cogn. Psychol. 5, 176–206. doi: 10.1016/0010-0285(73)90032-7

*Papanagnou, D., Serrano, A., Barkley, K., Chandra, S., Governatori, N., Piela, N., et al. (2016). Does tailoring instructional style to a medical student’s self-perceived learning style improve performance when teaching intravenous catheter placement? A randomized controlled study. BMC Med. Educ. , 16:205. doi: 10.1186/s12909-016-0720-3

Parsons, S., Kruijt, A. W., and Fox, E. (2019). Psychological science needs a standard practice of reporting the reliability of cognitive-behavioral measurements. Adv. Methods Pract. Psychol. Sci. 2, 378–395. doi: 10.1177/2515245919879695

Pashler, H., McDaniel, M., Rohrer, D., and Bjork, R. (2008). Learning styles: concepts and evidence. Psychol. Sci. Public Interest 9, 105–119. doi: 10.1111/j.1539-6053.2009.01038.x

Petticrew, M., Tugwell, P., Kristjansson, E., Oliver, S., Ueffing, E., and Welch, V. (2012). Damned if you do, damned if you don’t: subgroup analysis and equity. J. Epidemiol. Community Health 66, 95–98. doi: 10.1136/jech.2010.121095

Pigott, T. D., and Polanin, J. R. (2020). Methodological guidance paper: high-quality Meta-analysis in a systematic review. Rev. Educ. Res. 90, 24–46. doi: 10.3102/0034654319877153

Preacher, K. J., and Sterba, S. K. (2019). Aptitude-by-treatment interactions in research on educational interventions. Except. Child. 85, 248–264. doi: 10.1177/0014402918802803

Qiu, X., and Wang, Y. (2019). Composite interaction tree for simultaneous learning of optimal individualized treatment rules and subgroups. Stat. Med. 38, 2632–2651. doi: 10.1002/sim.8105

*Rassaei, E. (2018). Computer-mediated textual and audio glosses, perceptual style and L2 vocabulary learning. Lang. Teach. Res. , 22, 657–675. doi: 10.1177/1362168817690183

*Rassaei, E. (2019). Computer-mediated text-based and audio-based corrective feedback, perceptual style and L2 development. System , 82, 97–110. doi: 10.1016/j.system.2019.03.004

Reed, S. K. (2006). Cognitive architectures for multimedia learning. Educ. Psychol. 41, 87–98. doi: 10.1207/s15326985ep4102_2

Reid, J. M. (1987). The learning style preferences of ESL students. TESOL Q. 21, 87–111. doi: 10.2307/3586356

Rey, G. D., Beege, M., Nebel, S., Wirzberger, M., Schmitt, T. H., and Schneider, S. (2019). A Meta-analysis of the segmenting effect. Educ. Psychol. Rev. 31, 389–419. doi: 10.1007/s10648-018-9456-4

Riding, R. J. (1991). Cognitive styles analysis . Birmingham, United Kingdom: Learning and Training Technology.

Riding, R. J., and Ashmore, J. (1980). Verbaliser-imager learning style and children's recall of information presented in pictorial versus written form. Educ. Stud. 6, 141–145. doi: 10.1080/0305569800060204

*Riding, R., and Douglas, G. (1993). The effect of cognitive style and mode of presentation on learning performance. Br. J. Educ. Psychol. , 63, 297–307. doi: 10.1111/j.2044-8279.1993.tb01059.x

Riding, R., and Rayner, S. (1998). Cognitive styles and learning strategies: Understanding style differences in learning and behaviour . London, United Kingdom: David Fulton Publishing.

Riding, R., and Sadler-Smith, E. (1992). Type of instructional material, cognitive style and learning performance. Educ. Stud. 18, 323–340. doi: 10.1080/0305569920180306

*Rogowsky, B. A., Calhoun, B. M., and Tallal, P. (2015). Matching learning style to instructional method: effects on comprehension. J. Educ. Psychol. , 107, 64–78. doi: 10.1037/a0037478

*Rogowsky, B. A., Calhoun, B. M., and Tallal, P. (2020). Providing instruction based on students’ learning style preferences does not improve learning. Front. Psychol. , 11:164. doi: 10.3389/fpsyg.2020.00164

Rundle, S., and Dunn, R. (2010). Learning styles: Online learning style assessments and community . Available at: http://www.learningstyles.net

Sankey, M. D., Birch, D., and Gardiner, M. W. (2011). The impact of multiple representations of content using multimedia on learning outcomes across learning styles and modal preference. Int. J. Educ. Dev. Using Inf. Commun. Technol. 7, 18–35.

Schmidt, F. L. (2017). Statistical and measurement pitfalls in the use of meta-regression in meta-analysis. Career Dev. Int. 22, 469–476. doi: 10.1108/CDI-08-2017-0136

Singer, L. M., and Alexander, P. A. (2017). Reading on paper and digitally: what the past decades of empirical research reveal. Rev. Educ. Res. 87, 1007–1041. doi: 10.3102/0034654317722961

Slack, N., and Norwich, B. (2007). Evaluating the reliability and validity of a learning styles inventory: a classroom-based study. Educ. Res. 49, 51–63. doi: 10.1080/00131880701200765

Smith, R., Snow, P., Serry, T., and Hammond, L. (2021). The role of background knowledge in Reading comprehension: a critical review. Read. Psychol. 42, 214–240. doi: 10.1080/02702711.2021.1888348

Staudigl, T., and Hanslmayr, S. (2019). Reactivation of neural patterns during memory reinstatement supports encoding specificity. Cogn. Neurosci. 10, 175–185. doi: 10.1080/17588928.2019.1621825

Sterne, J. A. C., Egger, M., and Smith, G. D. (2001). Investigating and dealing with publication and other biases in meta-analysis. BMJ 323, 101–105. doi: 10.1136/bmj.323.7304.101

Sun, X., Norton, O., and Nancekivell, S. E. (2023). Beware the myth: learning styles affect parents’, children’s, and teachers’ thinking about children’s academic potential. NPJ Sci. Learn. 8:1. doi: 10.1038/s41539-023-00190-x

Sundararajan, N., and Adesope, O. (2020). Keep it coherent: a Meta-analysis of the seductive details effect. Educ. Psychol. Rev. 32, 707–734. doi: 10.1007/s10648-020-09522-4

Suurmond, R., van Rhee, H., and Hak, T. (2017). Introduction, comparison and validation of Meta-essentials: a free and simple tool for meta-analysis. Res. Synth. Methods 8, 537–553. doi: 10.1002/jrsm.1260

*Tadayonifar, M., Entezari, M., and Valizadeh, M. (2021). The effects of computer-assisted L1 and L2 textual and audio glosses on vocabulary learning and reading comprehension across different learning styles. J. Lang. Educ. , 7, 223–242. doi: 10.17323/jle.2021.11020

Tanner-Smith, E. E., Tipton, E., and Polanin, J. R. (2016). Handling complex meta-analytic data structures using robust variance estimates: a tutorial in R. J. Dev. Life-Course Criminol. 2, 85–112. doi: 10.1007/s40865-016-0026-5

Tedersoo, L., Küngas, R., Oras, E., Köster, K., Eenmaa, H., Leijen, Ä., et al. (2021). Data sharing practices and data availability upon request differ across scientific disciplines. Scientific Data 8:192. doi: 10.1038/s41597-021-00981-0

Thomas, P. R., and McKay, J. B. (2010). Cognitive styles and instructional design in university learning. Learn. Individ. Differ. 20, 197–202. doi: 10.1016/j.lindif.2010.01.002

Thomas, C. N., Van Garderen, D., Scheuermann, A., and Lee, E. J. (2015). Applying a universal design for learning framework to mediate the language demands of mathematics. Read. Writ. Q. 31, 207–234. doi: 10.1080/10573569.2015.1030988

Vasquez, K. (2009). “Learning styles as self-fulfilling prophecies” in Getting culture . eds. R. A. R. Gurung and L. R. Prieto (New York, NY, USA: Routledge), 55–63.

Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. J. Stat. Softw. 36, 1–48. doi: 10.18637/jss.v036.i03

Waddington, H. S., Wilson, D. B., Pigott, T., Stewart, G., Aloe, A. M., Tugwell, P., et al. (2022). Quasi-experiments are a valuable source of evidence about effects of interventions, programs and policies: commentary from the Campbell collaboration study design and Bias assessment working group. J. Clin. Epidemiol. 152, 311–313. doi: 10.1016/j.jclinepi.2022.11.005

Wallace, B. C., Small, K., Brodley, C. E., Lau, J., and Trikalinos, T. A. (2012). Deploying an interactive machine learning system in an evidence-based practice center: Abstrackr. Proceed. ACM Int. Health Inform. Symp. , 819–824. doi: 10.1145/2110363.2110464

Weaver, A. J. (2011). A Meta-analytical review of selective exposure to and the enjoyment of media violence. J. Broadcast. Electron. Media 55, 232–250. doi: 10.1080/08838151.2011.570826

What Works Clearinghouse (2022). WWC version 5.0 procedures and standards handbook. Available at: https://ies.ed.gov/ncee/wwc/handbooks#procedures

Willingham, D. T. (2018). Ask the cognitive scientist: does tailoring instruction to “learning styles” help students learn? Am. Educ. 42:28.

Wininger, S. R., Redifer, J. L., Norman, A. D., and Ryle, M. K. (2019). Prevalence of learning styles in educational psychology and introduction to education textbooks: a content analysis. Psychol. Learn. Teach. 18, 221–243. doi: 10.1177/1475725719830301

Yan, V. X., and Fralick, C. M. (2022). “Consequences of endorsing the individual learning styles myth: helpful, harmful, or harmless?” in Learning styles, classroom instruction, and student achievement . eds. D. H. Robinson, V. X. Yan, and J. A. Kim (Cham, Switzerland: Springer International Publishing), 59–74.

Zhu, J., Racine, N., Xie, E. B., Park, J., Watt, J., Eirich, R., et al. (2021). Post-secondary student mental health during COVID-19: a Meta-analysis. Front. Psych. 12:777251. doi: 10.3389/fpsyt.2021.777251

Keywords: learning styles, meta-analysis, modality, systematic review, crossover interaction

Citation: Clinton-Lisell V and Litzinger C (2024) Is it really a neuromyth? A meta-analysis of the learning styles matching hypothesis. Front. Psychol . 15:1428732. doi: 10.3389/fpsyg.2024.1428732

Received: 06 May 2024; Accepted: 17 June 2024; Published: 10 July 2024.

Reviewed by:

Copyright © 2024 Clinton-Lisell and Litzinger. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Virginia Clinton-Lisell, [email protected]

† ORCID: Virginia Clinton-Lisell, https://orcid.org/0000-0002-4705-2217

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.