Grounded on the literature on the related fields (mentioned above) and our preliminary studies, DinL is defined based on a set of learning dimensions, each dimension representing a psycho-social-educational way to process information. The learning profiles comprising the combination of different levels of such dimensions will allow us to determine the learning characteristics of the student from an integrative approach. The set of different types of DinL profiles among students will define the DinL in a specific classroom.

Five key learning dimensions composing the DinL construct are defined, according to the literature research in these fields:

(1) Coping with Difficulties, understood as the level of optimal regulation of emotional and psychosocial difficulties in learning such as anxiety, nervousness, bad mood, irritability, impulse control, attention deficits, apathy, discouragement, success expectations instudy, difficulties at home and university, the attribution of success or failure to circumstances, self-esteem, and the learning climate in the class (e.g., Alves et al., 2007; Bernard et al., 2024; Conley et al., 2023; Del Valle et al., 2020; Dumitrescu & De Caluwé, 2024; Herrera et al., 2020; Heritage et al., 2023; Khalil et al., 2020; Lew et al., 2019; MacCann, et al., 2020; Martos et al., 2023; Matalinares et al., 2016; Melo et al., 2021; Morales & Pérez, 2019; Naser et al., 2022; Pintrich, 2004; Pozar, 2002; Samaniego & Buenahora, 2016; Sanagavarapu et al., 2019; Santander et al., 2013; Tan et al., 2023; Tholen et al., 2022; Trigueros et al., 2020; Weinstein et al., 2002; Ya et al., 2024; Zandvliet et al., 2019).

(2) Effort, defined as the levels of perseverance, constancy, regular study habits, the capacity to delay reward, control of the time and situation for studying, and internal attribution of academic performance (Mondragón et al., 2017; Muñoz, 2022; Weinstein et al., 2002).

(3) Autonomy, defined as the extent to which the student employs self-regulation learning (active process to choose learning goals, and to control and regulate cognitive processes, motivation and behaviour), the active search for information sources, the integration of information from different sources, the generation of own theories, and the active search of empirical evidence of the theories and the applications of new knowledge (Al Mulhim, 2021; Allgood et al., 2000; Alonso et al., 2007; Beltrán et al., 2006; Jiménez et al., 2018; Kolb & Koln, 2005; Lilia et al., 2021; Maya et al., 2021; Merino-Soto et al., 2022; Pérez et al., 2011; Pintrich, 2004; Said et al., 2010).

(1) Understanding and Career Interest, understood as the levels to which the student focusses on deep learning, studying to attain a profound comprehension of the concepts and their interrelationships, and closely related to the interest in focusing learning on the knowledge and skills required for the chosen profession (Alonso et al., 2007; Álvarez & Fernández, 2005; Han et al., 2024; Herrera & Rodríguez, 2011; Jiménez et al., 2018; Kolb & Koln, 2005; Leite et al., 2010; Mondragón et al., 2017; Weinstein et al., 2002).

(5) Social and Physical Context, refers to the preferences to study alone vs. with other students, and preferences about the study place (Acevedo et al., 2009; Álvarez & Fernández, 2005; Mondragón et al., 2017; Park et al.2021; Yang & Kang, 2020).

The present study aims to obtain evidence of validity for a scale that measures the Diversity in Learning (DinL) construct. The validity evidence includes four key aspects: (1) internal structure, expecting to confirm a five-dimension model, (2) invariance across male and female participants, (3) concurrent validity with academic performance, and (4) internal consistency of the dimensions. The scale measures this construct individually, understood at individual level as the combination of main dimensions comprising the psychological processes, styles, habits, difficulties, internal and external resources, contextual and psychosocial factors defining the individual learning profile that distinguish it from other individual profiles within a diversity in learning in the academic classroom. It is hypothesized that the DinL scale will show optimal evidence of validity and reliability. Likewise, the five-dimensional structure of the scale defining the DinL integrative construct will show adequate internal consistency qualities, expecting this structure and measurement properties to hold across men and women.

Using a cross-sectional design, the initial sample was composed of 1,841 students enrolled at the Complutense University of Madrid (UCM) in Spain; 11% of cases were excluded from the study due to major errors in responses or incomplete questionnaires. The final sample was 1,644 students with a mean age of 24.37 (SD = 8.73), 1,167 (70.9%) of whom were women, 1,285 (78.2%) were undergraduate, 233 (14.2%) were studying at master’s level, and 126 (7.6%) were PhD students from different disciplines (social science, science, humanities and engineering); 1,331 of the students were born in Spain and other European countries (85.1%) and the rest were from South America (9.2%), Asia (2.3%), North America (2.1%), and Africa (1.3%). The sample size was determined by the recommendation that a sufficient sample for analysis with WLSMV estimator must be made up of more than 200-500 participants (Kyriazos, 2018). (See Table 3 for the sampling procedure).

Table 3

The Diversity in Learning Scale (DinL) is an individually self-administered scale that describes the psychosocial diversity of learning patterns observed in the classroom. It comprises of 28 items and represents the above-defined dimensions in 5 sub-scales: Coping with Difficulties, Effort, Autonomy, Understanding and Career Interest, and Social/Physical Context. The items were rated on a four-point Likert scale from 1,= nothing or very little, 2 = some, 3 = quite a lot, 4 = a lot (see Table 4 for a list of the items grouped in each of the sub-scale). The scale was constructed in three stages: (1) an extensive bibliographic research of the fundamental psychosocial and psycho-educational dimensions regarding learning styles, habits, strategies and difficulties; (2) a theoretical and empirical development of the DinL construct, including its constituent dimensions; and (3) a survey conducted with a large sample of university students for the purpose of scale validation.

Academic performance was measured by an item that asked for the average grade obtained in the previous academic year.

Participation in the study was voluntary, and all participants provided informed consent before completing the questionnaire. Data collection adhered to the ethical standards outlined in the Declaration of Helsinki (World Medical Association, 2013) and was conducted in compliance with national and European data protection regulations, ensuring anonymity and confidentiality. In addition to the approval of the Research Ethical Committee at the University of Madrid (Ref. No CE_20211118-15_SOC), the study was guided by principles of ethical and scientific rigor. These included:

Data analyses were performed using the statistics program R (RStudio, 4.2.2) to address the research objectives. First, descriptive analyses were conducted to characterize the sample. To evaluate the internal structure of the DinL scale, confirmatory factor analysis (CFA) was conducted to test the hypothesized five-factor model. Given the categorical nature of the data, analyses were performed using the weighted least squares means and variance adjusted (WLSMV) estimator, which is robust to violations of multivariate normality (DiStefano & Morgan, 2014). The polychoric correlation matrix served as the basis for the analysis. The scale was assessed using the goodness of fit indexes: comparative fix index > .90 (CFI; Bentler, 1990) and the Tucker Lewis index > .90 (TLI; Tucker & Lewis, 1973), the root mean square error of approximation < .80 (RMSEA; Steiger, 1990) and the standardized root mean residual < .80 (SRMR; Fan & Sivo, 2007).

To investigate invariance across sexes, a multigroup confirmatory factor analysis was performed, comparing nested models of configural, metric, scalar, and strict invariance. The fit was evaluated using χ², CFI, TLI, and RMSEA. Differences between the models were examined with a ΔCFI test and a ΔRMSEA test, each considering the less restrictive model. A non-optimal result in ΔCFI is less than -.01 and .015 in ΔRMSEA (Chen, 2007; Dimitrov, 2010). Internal consistencies of each of the five subscales of the DinL scale were assessed as reflected by the omega McDonald coefficient (ω) and confidence intervals (Viladrich et al., 2017). The score ≥ .70 is considered “acceptable”, ≥ .80 “good” and ≥ .90 “excellent” indicators (Taber, 2018). The model parameters were estimated under the same conditions as in the confirmatory factor analysis, using the polychoric correlation matrix and the WLSMV estimation method. Finally, the concurrent validity was evaluated through Pearson correlation coefficients using the five scales of DinL and perceived academic performance.

Table 4 provides the means, standard deviations, skewness, and kurtosis of the 28 items. The skewness and kurtosis values are below ± 1.96, indicating that the items conform to a normal distribution (Mardia, 1970; West et al., 1995). Additionally, the discrimination indices (DI) demonstrate that the items effectively differentiate between participants with varying levels of the measured constructs.

Table 4

Note. DI = discrimination index; M = mean; SD = standard deviation; Sk = skewness; Ku = kurtosis. Item numbers in bold are the inverse items.

The tested measurement of the goodness-of-fit showed CFI = .932, TLI = .925, RMSEA = .065 [.063, .068] and SRMR = .067 for the five subscales of the model. The model fit values were adequate for the threshold proposed. The reliability of the scales in terms of internal consistency score between ω = .62 to .80, with an acceptable but lower score for the subscale Understanding/Career Interest (Table 4). Figure 1 shows the path diagram of the model with the standardized weights.

Figure 1 Note. Con = social/physical context; Cop = coping with difficulties; Und = understanding/career interest; Eff = effort; Aut = autonomy.

The measurement invariance of the DinL scale was assessed to examine its applicability across male and female subgroups. Table 5 summarizes the fit indices for the configural, metric, scalar, and strict invariance models. The results indicated adequate fit for the configural, metric, and scalar models, demonstrating that the scale retains the same factor structure and metric properties across subgroups. However, the strict invariance model showed a slight decrease in fit indices (ΔCFI = -.0018), suggesting minor variations in item residuals between groups. Despite this, the results confirm that the scale is robust for comparing latent constructs across genders at the configural, metric, and scalar levels, supporting its use for examining gender-based differences in the DinL construct.

Table 5

*p < .05, **p < .01, ***p < .001.

The five DinL subscales were correlated with perceived academic performance. Table 6 indicates that all the subscales were statistically significant at the .01 level. The perceived academic performance showed a positive relationship with Coping with Difficulties, Effort, Autonomy and Understanding/Career Interest, but a negative one with Social/Physical Context.

Table 6

Note. AP = academic performance.

*p < .05, **p < .01.

This study has examined the concept of diversity in students’ study and learning processes, utilizing a holistic definition of the Diversity in Learning (DinL) construct. This definition encompasses not only behavioral and psycho-educational aspects but also contextual and psychosocial factors that influence learning. The findings underscore the relevance and potential of an integrative approach to understanding the increasing DinL in university classrooms. DinL extends beyond established fields of research such as learning styles and strategies, which have shown limited applicability in addressing the complexity and diversity of higher education environments.

Additionally, this study demonstrates the feasibility of operationalizing DinL using a concise scale that evaluates five meaningful dimensions, characterized by robust psychometric properties. This scale holds significant potential for research and practical application, enabling a more personalized, equitable, and inclusive approach to higher education. Such an approach may enhance academic performance and success, particularly for students who do not conform to the predominant learning patterns within their respective academic fields.

The present study validates the soundness and robustness of an innovative and integrative DinL framework within the university context. The developed scale offers practical applications for educators, allowing the categorization of students into DinL profiles based on their unique combinations of learning dimensions. This categorization could facilitate the alignment of teaching strategies with the diverse learning needs of students, moving away from the traditional “one-size-fits-all” approach (Khalil et al., 2020) that persists in many academic contexts.

This tool may be particularly beneficial in designing targeted strategies for groups of students whose learning patterns deviate significantly from the norm, thereby fostering inclusivity and equity. Additionally, it supports the organization of DinL-based student groups to enhance collaborative learning outcomes. Such an approach repositions DinL from being perceived as a challenge to being viewed as a resource for promoting learning and improving academic performance. Furthermore, the scale provides insights into the psychological and social factors influencing these learning dimensions, enabling interventions to enhance learning experiences.

This study acknowledges certain limitations, including lower omega reliability in one subscale and minor shortcomings in strict invariance. Regarding future studies, the analysis of the influence of digital technologies on learning patterns is set to be of fundamental importance. The present research project did not find significant evidence to suggest that the available digital tools have fundamentally altered students’ learning methods, strategies, or habits to date. Research in this area, including that of Zambrano et al. (2018), supports this observation, indicating that digital tools currently function primarily as supplementary aids rather than transformative influences on established learning patterns. Nevertheless, future studies should explore the potential impact of emerging digital learning methodologies on the development of diverse learning strategies. Finally, while this study was conducted at a large public university in Spain with substantial psychosocial diversity, the scale is currently being tested in universities across different countries to evaluate its psychometric properties in varied social contexts.