Motivational profiles for physical activity in Chinese university students based on the theory of planned behavior

Abstract

Physical activity (PA) is essential for health and well-being, yet college students often struggle to maintain consistent engagement. This study employed the Theory of Planned Behavior (TPB) and latent profile analysis (LPA) to investigate how TPB constructs attitude, subjective norm, perceived behavioral control, and intention combine within individuals to form distinct motivational configurations among Chinese college students, moving beyond variable-centered approaches that assume uniform relationships. We aimed to identify these subgroups, examine gender and age differences in subgroup membership, and compare physical activity levels across subgroups.We recruited 313 Chinese college students (M = 19.56, SD = 1.18) and used LPA to identify subgroups based on TPB constructs. A three-step method with BCH examined how gender and age related to profile membership and how profiles differed in PA.LPA identified four optimal profiles: Disengaged (9.1%), Moderately Driven (31.2%), Autonomously Active (34.2%), and Fully Integrated (25.6%). Gender significantly predicted profile membership, with female students less likely to belong to more active profiles. Age showed no significant effect. The BCH method showed that the Autonomously Active and Fully Integrated profiles had significantly higher PA than the Disengaged and Moderately Driven profiles.These four motivational configurations reveal heterogeneity in how TPB constructs combine within individuals to influence physical activity. The findings support tailoring interventions to specific profiles rather than using a one-size-fits-all approach, offering a refined foundation for both research and practice.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-50543-w.

Introduction

Regular participation in physical activity (PA) has been found to have several benefits for the healthy development of youths, such as improving physical fitness, and promoting mental health¹. Although moderate physical activity is beneficial for people’s health, many researchers revealed that many college students struggle to maintain regular physical activity in many countries^2,3.

Internal factors such as motivation, intention, and belief play a vital role in influencing students’ participation in physical activities⁴. The Theory of Planned Behavior (TPB) provides a framework for understanding these factors⁵. However, most prior TPB research relies on variable-centered approaches that assume the relationships between TPB constructs and behavior are uniform across individuals, potentially masking meaningful differences in how these psychological factors combine within different people⁶. In contrast, person-centered analyses such as latent profile analysis (LPA) identify distinct subgroups who share similar motivational configurations, thereby uncovering heterogeneity in psychological processes^7,8. This approach can reveal configurations that may range from profiles with uniformly low scores to those with elevated scores on specific combinations of constructs. This study addresses this gap by applying LPA within the TPB framework.

Accordingly, this study aims to identify distinct motivational profiles based on TPB constructs among Chinese college students using LPA, and to examine how these profiles differ in physical activity levels. This person-centered approach seeks to provide new insights for developing tailored interventions to promote student participation in PA.

Literature review

Physical activity patterns among college student

Despite the well-documented health benefits of regular PA, many college students fail to meet recommended activity guidelines^9,10. Understanding this inactivity requires looking beyond individual health beliefs to the social and developmental context of university life. For students, engagement in PA is not merely a health behavior but can be a critical arena for forming social connections and developing a sense of institutional belonging. Participating in sports or recreational activities can shape how students identify with their peers and the university itself^11,4.

Participation patterns are further characterized by notable demographic variations. There may be gender differences in participation in sports activities, with boys participating more frequently than girls¹². There are many influencing factors that may contribute to gender difference in sports, such as social norms and gender stereotypes¹³. Children and adolescents are influenced by their environment during their growth process, which may lead to gender bias in certain sports¹⁴. For example, some sports such as football and basketball are considered male dominated, resulting in fewer women participating in them¹⁵. These observed differences may reflect not merely innate preference but also differential access to resources, visibility, and socially valued roles within physical activity contexts, which shape institutional belonging and identity formation¹⁶. This perspective is particularly relevant for university students, whose engagement in sports may influence their sense of connection to the institution and their peers. Such established patterns, where boys are often found to be more active than girls from adolescence onward, can perpetuate health inequalities^13,17.

Moreover, the age difference among college students can also affect their participation in sports. When students transition from high school to college, their physical activities may decrease¹⁸. Specifically, young students, like freshmen and sophomores, usually have more energy and enthusiasm, so they are more likely to participate in competitive and high-intensity sports. On the other hand, older high school students may be more focused on academic and social activities, which leads to less participation in organized sports activities¹⁹. It is important to note, however, that the restricted age range and shared life stage of a university sample may limit observable age-related variability in some psychological or behavioral measures.

By systematically exploring these gender and age differences in physical activity among college students, universities can be better informed to develop more inclusive and effective sports programs that meet the diverse needs of all students, thereby working to improve overall student health and well-being²⁰.

Theory of planned behavior in physical activity research

TPB was proposed by Ajzen and has been widely used to study education, consumer, and health-related behaviors of different groups^5,18,²¹. At its core, the TPB provides a structured framework for understanding the cognitive antecedents of volitional behavior. It posits that an individual’s intention to perform a behavior is the most immediate and primary determinant of their actual behavior²². These behavioral intentions are themselves shaped by three foundational constructs: (1) attitude, which refers to an individual’s overall positive or negative evaluation of performing the behavior,(2) subjective norm, which captures the perceived social pressure from significant others to engage or not engage in the behavior,and (3) perceived behavioral control (PBC), which reflects one’s perception of the ease or difficulty of performing the behavior, encompassing factors like available resources, skills, and potential obstacles. Due to its clear and testable propositions, the TPB is considered a highly useful theoretical model and has been extensively applied in the study of youth health behaviors, including physical activity^23,24.

Early studies used TPB to observe PA in young people^25,26, these studies consistently explore how variations in TPB constructs explain differences in PA participation levels across groups. Research focusing specifically on college students has reinforced the theory’s utility. For example, Norman and Conner have a strong connection between intention and behavior. The three TPB factors have a positive predictive effect on the intentions and behaviors of British university students²⁷. Further supporting its relevance in non-Western contexts, a study in South Korea found that within the TPB framework, factors such as health awareness and self-efficacy played significant roles in strengthening the relationships between attitudes, intentions, and sports behaviors among college students³. These studies indicate that TPB is very useful for understanding and shaping college students’ participation in physical activity. They provide important insights into how psychological factors affect young people’s participation in sports activities.

However, the application of the TPB has also revealed a significant theoretical and practical challenge known as the "intention-behavior gap." This refers to the common finding that individuals’ positive intentions do not always translate into actual behavior²⁸. In the context of physical activity, many people express strong intentions to exercise but consistently fail to follow through. This gap suggests that the relationship between intention and behavior is not straightforward and may be moderated by other factors not fully captured by the standard TPB variables²⁹. The existence of this gap underscores the complexity of motivational processes and implies that there may be distinct subgroups of individuals for whom the intention-behavior link is stronger or weaker. This insight provides a key rationale for moving beyond models that assume a uniform motivational pathway and towards person-centered analyses that can uncover different motivational archetypes, some of which may be more or less susceptible to this intention-behavior gap.

Rationale for person-centered analysis

Previous research has identified various factors influencing the physical activity of university students within the TPB framework. For instance, studies have shown that psychological satisfaction can enhance attitudes and participation²⁵, and that enjoyment is a key predictor of TPB constructs and exercise behavior³⁰. These variable-centered studies, which examine relationships between variables across an entire sample, have provided valuable insights. However, they are characterized by a significant methodological and theoretical limitation that inherently assume population homogeneity. This means they operate on the premise that the relationships between TPB constructs and physical activity are essentially the same for every individual in the population⁶. This assumption of a uniform motivational process may obscure the existence of distinct subgroups with unique psychological profiles.

To address this gap and move beyond the homogeneity assumption, person-centered analytical methods are essential. Person-centered approaches, such as Latent Profile Analysis (LPA), do not focus on relationships between variables for the whole sample. Instead, they identify latent subgroups of individuals who share similar patterns of responses across multiple variables^7,8. Whereas variable-centered methods aim to understand how constructs are associated on average, LPA seeks to classify individuals based on their shared response profiles, thereby uncovering the common typologies of motivation that exist within the population³¹. This allows researchers to discover whether there are, for example, subgroups of students who are primarily driven by high personal attitude but low social concern, or those with high perceived control but ambivalent attitudes. LPA has proven effective in health behavior research for identifying such subgroups based on psychological profiles^32,33.

Thus, while prior variable-centered TPB research has established important average trends, it may have overlooked meaningful heterogeneity. By applying LPA to TPB constructs among Chinese college students, this study aims to uncover this hidden diversity. It seeks to identify distinct motivational archetypes and examine how these qualitatively or quantitatively different profiles are associated with varying levels of physical activity. This person-centered shift in perspective is crucial for developing interventions that are not based on a “one-size-fits-all” model of motivation but are instead tailored to the specific needs of different student subgroups.

The present study

This study used latent profile analysis to explore TPB factors and physical activity among Chinese college students, which can provide a new insight for further research and intervention to improve college student’s physical activity. Specially, this study aims to (a) explore the influence factors in physical activity based on TPB by LPA; (b) examine the gender and age difference; (c) explore the relationship between the relationship between subgroups and their physical activity.

Methods

Participants

In this study, a total of 501 college students were recruited using convenience sampling, and the final sample consisted of 313 participants (M age = 19.560, SD = 1.179, Male = 170, Female = 143). Blank questionnaires or those with patterned (e.g., straightlining), random, or contradictory responses will be excluded³⁴. This study satisfied the requirements of sample size. This research satisfied the minimum sample size requirement of latent profile analysis⁸.

Measures

Scale of the theory of planned behavior (TPB)

The TPB is a mature and practical theoretical framework. As²² has noted, there is no single standardized questionnaire for the TPB and measures must be tailored to the specific behavior and population under investigation. Therefore, consistent with³⁵ guidelines for constructing TPB questionnaires, this study developed a scale specifically to measure the core TPB constructs in the context of regular physical activity among Chinese university students.

The scale was developed for this research. Item generation was guided by the conceptual definitions of the four TPB constructs (Attitude, Subjective Norm, Perceived Behavioral Control, and Intention) applied to the target behavior of “exercising at least 6 times in the next two weeks.” Item phrasing was designed to be clear and direct for a student population, drawing upon the common semantic structure used in established TPB research. The final questionnaire comprised 19 items: 6 items for attitude (e.g., “For me, exercising at least 6 times in the next two weeks is…”, α = .919, ω = .920), 4 items for subjective norm (e.g., “Most people who are important to me think that I should exercise at least 6 times in the next two weeks”, α = .861, ω = .862), 5 items for perceived behavioral control (e.g., “If I wanted to, I could exercise at least 6 times in the next two weeks”, α = .870, ω = .874), and 4 items for behavioral intention (e.g., “I intend to exercise at least 6 times in the next two weeks”, α = .921, ω = .922). All items were rated on a 7-point Likert scale ranging from totally disagree (1) to totally agree (7). The complete set of scale items is provided in the Supplementary Materials.

The psychometric properties of the newly developed TPB scale were evaluated and found to be excellent. The scale demonstrated outstanding internal consistency, with both Cronbach’s alpha (α) and McDonald’s omega (ω) coefficients for all subscales ranging from .861 to .922. Confirmatory factor analysis (CFA) was conducted to test the hypothesized four-factor structure. The model fit indices were as follows: χ² = 519.357, df = 146, RMSEA = .090, CFI = .924, TLI = .911, SRMR = .039. These robust reliability and validity results confirm that the scale developed for this study is a psychometrically sound instrument for measuring TPB constructs within the present sample³⁶.

Physical activity

Physical activity was assessed by the metabolic equivalent of task (MET) of the student and measured by the International Physical Activity Questionnaire short form (IPAQ—SF), which consists of 7 items³⁷. Six items investigate the participants’ physical activity, and one item examines the duration of sedentary behavior. The weekly MET for each participant is calculated as the sum of the activity type value × weekly frequency × daily duration. Especially, MET in this research was calculated as 8.0 (vigorous-intensity activity) × weekly frequency × daily duration + 4.0 (moderate-intensity activity) × weekly frequency × daily duration + 3.3 (walking) × weekly frequency × daily duration^38,39. Following standardization, which converted the total MET-min/week scores to z-scores for subsequent profile comparisons, extreme outliers defined as values beyond ±3.29 standard deviations from the mean were excluded from analysis to ensure data robustness^40,41.

Demographic variables

Demographic variables included participants’ gender (coded as 1 = male, 2 = female), age, height, and weight.

Procedure

This research adopted a cross-sectional design and the university ethics committee has approved this research (IRB: No. 2025LCLL—009). The survey was conducted between February and March 2025. Participants came from two different colleges in the south of China and completed the online survey. Primary participants were adults and all participants provided digital written informed consent through mandatory checkbox confirmation on the Wenjuanxing survey platform. This online survey presented study details including purpose, voluntary participation, confidentiality protocols, withdrawal rights, and compensation (small gift upon completion). This study strictly adhered to standardized survey protocols during the data collection process.

Statistical analysis

Descriptive statistics, correlations, Harman’s single-factor test, and multivariate eanalysis of variance (MANOVA) were analyzed by SPSS 25 and reliability was conducted by JASP 0.95.3.0. We conducted the CFA, bifactor model analysis, and LPA using Mplus 8.3. As for LPA, this study evaluated 1-5 profiles to identify the optimal profile based on established model fit criteria⁷. The selection of the optimal profile was guided by several statistical indices, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), adjusted BIC (aBIC), Lo-Mendell-Rubin adjusted likelihood ratio test (LMR), bootstrap likelihood ratio test (BLRT), and entropy⁸. Lower AIC, BIC, and aBIC values indicate superior model fit. Furthermore, statistically significant BLRT and LMR values suggest that the k-class model provides a significantly better fit than the (k−1)—class model. Finally, entropy reflects classification accuracy, with values ≥ 0.8 indicating high precision in profile assignment⁴².

After identifying the optimal latent profile solution, a three-step method was used to examine how covariates relate to profile membership and how profiles differ in a distal outcome. This method accounts for the uncertainty in classifying individuals into profiles. Firstly, the R3STEP procedure was applied to test if demographic variables predicted profile membership. Secondly, the BCH method was used to compare the levels of physical activity across the different profiles⁴³.

Results

Common methods bias

To assess common method bias (CMB), Harman’s single-factor test was conducted⁴⁴, yielding a variance of 48.628%, below the 50% threshold, indicating no severe CMB concerns⁴⁵. Further, a bifactor model analysis showed no better fit than the trait-only model (Table 1), confirming that CMB⁴⁶ did not substantially affect the data⁴⁷.

Table 1.

Results of common method bias by Bifactor model.

Model	AIC	BIC	aBIC	χ2	df	RMSEA	CFI	TLI	SRMR
M1	18090.725	18326.736	18126.920	519.357	146	.090	.924	.911	.039
M2	18116.624	18382.605	18157.415	529.256	138	.095	.921	.902	.058
M3	18433.254	18650.534	18466.576	871.886	151	.124	.854	.834	.055
M4	18489.594	18703.496	18522.710	930.594	152	.128	.842	.822	.057

Notes. M1 = Baseline model, M2 = Bifactor model, M3 = Combined model, M4 = One factor model.

Descriptive statistics

As shown in (Table 2), the Pearson’s correlation analysis indicated that attitude was positively correlated with subjective norm (r = .671, p < .001), PBC (r = .775, p < .001), intention (r = .758, p < .001), and physical activity (r = .317, p = .001). Subjective norm was positively correlated with PBC (r = .730, p < .001), intention (r = .761, p < .001), and physical activity (r = .236, p < .001). PBC was positively correlated with intention (r = .870, p < .001), and physical activity (r = .320, p < .001). Intention was positively correlated physical activity (r = .310, p < .001).

Table 2.

Descriptive statistics and correlations of key variables.

	Variables	M	SD	Ske.	Kur.	1	2	3	4	5	6
1	Attitude	4.689	1.375	−0.234	−0.516
2	SN	4.501	1.340	−0.194	−0.283	.671***
3	PBC	4.768	1.307	−0.234	−0.356	.775***	.730***
4	Intention	4.725	1.455	−0.312	−0.356	.758***	.761***	.870***
5	PA	-0.118	0.837	1.158	1.446	.317***	.236***	.320***	.310***
6	Age	19.560	1.176	1.293	4.974	.174***	.139*	.141*	.167**	.252***
7	Gender	/	/	/	/	−.339***	−.285***	−.422***	−.380***	−.366***	−.213***

SN = Subjective norm, PBC = Perceived behavioral Control, PA = Physical activity.

Notes. * p < .05,** p < .01, *** p < .001. Ske = Skewness, Kur = Kurtosis.

Moreover, as for covariate, age was significantly correlated with attitude (r = .174, p = .002), subjective norm (r = .139, p = .014), PBC (r = .141, p = .012), intention (r = .167, p = .003), and physical activity (r = .252, p < .001). Gender was negatively correlated with attitude (r = -.339, p < .001), subjective norm (r = -.285, p < .001), PBC (r = -.422, p < .001), intention (r = -.380, p < .001), and physical activity (r = -.366, p < .001).

Latent profile analysis of TPB Factors

This study employed LPA to categorize TPB factors. (Table 3) presents the model fit indices for 1-5 profiles. The AIC, BIC, and adjusted BIC values decreased sharply from 1 to 3 profiles but plateaued at 4 to 6 profiles, suggesting improved fit with additional categories. However, the 5-profile model’s LMR p-value exceeded 0.05, indicating no significant improvement over the 4-profile model. Further, the 4-profile solution demonstrated high classification accuracy (entropy > 0.8). Based on these criteria, the 4-profile model was selected as optimal in this study.

Table 3.

Latent profile model information criteria, likelihood ratio test, and entropy (n = 313).

Model	K	Log(L)	AIC	BIC	aBIC	Entropy	BLRT	LMR	Group size
1	8	−2167.077	4350.154	4380.124	4354.750	/	/	/	1.000
2	13	−1864.088	3754.176	3802.876	3761.644	.869	<.001	<.001	0.443/0.577
3	18	−1730.184	3496.374	3563.806	3506.716	.888	<.001	.009	0.139/0.486/0.375
4	23	−1651.200	3348.400	3434.563	3361.614	.887	<.001	.004	0.091/0.312/0.342/0.256
5	28	−1632.718	3321.435	3426.329	3337.522	.877	<.001	.339	0.040/0.283/0.108/0.323/0.247

Additionally, the distribution of scores on the four dimensions of the 4-category model is shown in Fig. 1. Profile 1, the smallest subgroup (9.1%, n = 28), was characterized by uniformly low scores across all four TPB constructs. Within this pattern of global lows, perceived behavioral control was marginally elevated relative to the other dimensions. This configuration of uniformly low motivational cognitions underpins its label as the “Disengaged Profile.”

Fig. 1.

Line graph of the best fitting model Abbreviations. SN = Subjective norm, PBC = Perceived behavioral control.

Profile 2, the second largest subgroup (31.2%, n = 98), displayed moderately low scores with relatively balanced values across all four TPB dimensions. This flat configuration of mid-range scores supports its label as the “Moderately Driven Profile.”

Profile 3, the largest subgroup (34.2%, n = 108), exhibited the second highest scores across all TPB constructs, with particularly elevated intention and perceived behavioral control. This configuration of high self-determined motivational resources underpins its label as the “Autonomously Active Profile.”

Profile 4 (25.6%, n = 79) displayed the highest scores across all TPB dimensions compared to the other subgroups. A notable feature of this profile’s shape is that subjective norm, while still high in absolute terms, was relatively lower than attitude, perceived behavioral control, and intention. This configuration of globally high yet internally differentiated motivational resources supports its label as the “Fully Integrated Profile.”

These four profiles demonstrate that the underlying motivational structure differs both in magnitude and in shape. Profiles 3 and 4 exhibited higher overall levels of motivational resources compared to Profiles 1 and 2, reflecting differences in magnitude. Additionally, the profiles differed in shape: Profile 4 showed a distinct pattern with relatively lower subjective norm despite globally high scores, whereas Profile 2 displayed a flat configuration with balanced values across all dimensions. This distinction between magnitude and shape is central to understanding the heterogeneous nature of student motivation.

Latent profile analysis and auxiliary variable testing

The results of the three-step (R3STEP) multinomial logistic regression analysis, which accounts for classification uncertainty in profile assignment, are presented in (Table 4). The Disengaged Profile served as the reference group for comparisons.

Table 4.

Results of the three-step (R3STEP) multinomial logistic regression predicting latent profile membership.

Predictor	Profile comparison (vs. Profile 1)	Odds ratio (OR)	95% CI	p
Age	Profile 2 (Moderately driven)	1.16	[0.92, 1.45]	.207
	Profile 3 (Autonomously active)	1.00	[0.79, 1.27]	.996
	Profile 4 (Fully integrated)	1.10	[0.88, 1.37]	.409
Gender	Profile 2 (Moderately driven)	0.55	[0.31, 0.99]	.045
	Profile 3 (Autonomously active)	0.14	[0.02, 0.86]	.033
	Profile 4 (Fully integrated)	0.07	[0.02, 0.23]	< .001

Note. Gender was coded as 1 = Male, 2 = Female. An odds ratio (OR) less than 1 for the Gender predictor indicates that female students have lower odds of belonging to the comparison profile than male students, relative to the Disengaged Profile.

Age did not significantly predict profile membership when compared to the reference group. The odds ratios for age were 0.88 (95% CI [0.71, 1.08], p = .212) when comparing the Autonomously Active Profile to the reference group, and 1.02 (95% CI [0.83, 1.26], p = .841) for the Fully Integrated Profile comparison.

Gender significantly predicted profile membership. Using the Disengaged Profile as a reference, female students (coded as 2) were significantly less likely to belong to the Fully Integrated Profile (OR = 0.14, 95% CI [0.05, 0.42], p < .001). Female students were also significantly less likely to belong to the Moderately Driven Profile (OR = 0.55, 95% CI [0.31, 0.99], p = .045) and the Autonomously Active Profile (OR = 0.14, 95% CI [0.02, 0.86], p = .033) compared to the reference group.

Differences in physical activity across latent profiles

Differences in standardized PA levels across the four latent profiles were examined using the Bolck-Croon-Hagenaars (BCH) method, which adjusts for classification uncertainty. The BCH-adjusted mean PA levels for each profile are presented in (Table 4). A significant overall difference in PA across the four profiles was confirmed, χ² (3) = 41.49, p < .001.

Pairwise comparisons revealed that the Fully Integrated Profile (M = 0.22, SE = 0.11) and the Autonomously Active Profile (M = 0.02, SE = 0.09) reported significantly higher PA levels than the Disengaged Profile (M = -0.48, SE = 0.14; χ² (1) = 16.56, p < .001 and χ² (1) = 9.61, p = .002, respectively). The Fully Integrated Profile also reported significantly higher PA than the Moderately Driven Profile (M = -0.45, SE = 0.07; χ² (1) = 28.40, p < .001), as did the Autonomously Active Profile (χ² (1) = 16.23, p < .001). No significant differences in PA levels were found between the Disengaged and Moderately Driven profiles (χ² (1) = 0.05, p = .829), or between the Autonomously Active and Fully Integrated profiles (χ² (1) = 1.81, p = .178) (Table 5).

Table 5.

Tests of physical activity differences across latent profiles using the BCH method.

Latent profile	n	Adjusted mean (PA)	SE	Pairwise comparisons	p
1. Disengaged	28	-0.480	.135	/	/
2. Moderately driven	98	-0.447	.065	vs. Profile 1: 0.047	.829
3. Autonomously active	108	0.024	.091	vs. Profile 1: 9.613	.002
				vs. Profile 2: 16.234	< .001
4. Fully integrated	79	0.223	.108	vs. Profile 1: 16.562	< .001
				vs. Profile 2: 28.401	< .001
				vs. Profile 3: 1.810	.178

Note. PA = Physical Activity, presented as standardized scores. Means and standard errors (SE) are adjusted for classification uncertainty using the BCH method

Discussion

This study employed a person-centered approach to identify distinct motivational profiles among Chinese college students based on the Theory of Planned Behavior (TPB). By moving beyond variable-centered analyses that assume population homogeneity, this research offers a nuanced understanding of the diverse psychological pathways to physical activity (PA) engagement. The findings reveal four distinct motivational profiles characterized by varying configurations of TPB constructs: Disengaged (9.1%), Moderately Driven (31.2%), Autonomously Active (34.2%), and Fully Integrated (25.6%). These groups exhibited significant differences in PA levels. The findings offer a nuanced understanding of how TPB constructs combine within individuals and highlight the need for tailored interventions in the Chinese higher education context.

Characterization of latent profiles

The results reveal four distinct subgroups among Chinese college students, each demonstrating unique configurations of TPB constructs (attitude, subjective norm, PBC, and intention). The Inactive Group (9.1%) exhibits uniformly low scores across all dimensions, with PBC being marginally higher than other factors, suggesting that even students with minimal motivation recognize some degree of personal control over physical activity. The Moderate Group (31.2%) displays mid-range scores with balanced contributions from all TPB components, indicating moderate engagement but lacking strong motivational drivers. The Active Group (34.2%), the largest subgroup, demonstrates elevated scores, particularly in intentions, reinforcing the centrality of intentionality in physical activity engagement²². Finally, the Highly Active Group (25.6%) scores highest in all dimensions except subjective norm, which are comparatively lower-a finding that aligns with prior research suggesting that highly motivate individuals may rely less on social pressure²⁴.

These profiles extend previous variable-centered TPB studies by illustrating how psychological factors interact within individuals⁶. The dominance of the Active and Highly Active Groups suggests that many Chinese college students possess the cognitive foundations for physical activity, yet structural barriers (e.g., academic workload) may prevent translation into behavior⁹.

Covariate analysis: the role of gender and age

Our analysis revealed significant gender differences in profile membership. Female students were significantly less likely than males to belong to the more active and motivationally robust profiles (Autonomously Active and Fully Integrated) compared to the Disengaged Profile. This finding suggests that female students in this sample may face systemic or socio-cultural barriers that affect the formation of strong, self-determined motivational configurations for PA. These barriers could include differential access to sports facilities, fewer visible female role models in university athletics, or internalized gender stereotypes⁴⁸. The broader role of PA in shaping student identity and institutional belonging is an important consideration in higher education research.

Contrary to some previous research in broader populations, age did not predict profile membership. This null finding is likely attributable to the restricted age range and shared life stage of our college student sample, which creates a homogeneous context that limits age-related variability in motivational structures⁸.

Physical activity disparities across profiles

The BCH-adjusted comparisons confirmed significant PA disparities across profiles, validating the motivational typology. As hypothesized, profiles with stronger TPB configurations (Autonomously Active and Fully Integrated) reported significantly higher PA levels than the Disengaged and Moderately Driven profiles. Notably, the difference between the two highest profiles was non-significant, suggesting that once a threshold of self-determined motivation (high attitude and control) is reached, further increases in subjective norms or intention may yield diminishing returns for behavior. This aligns with the concept of the intention-behavior gap, where intention does not always translate to action²⁹.

The strong predictive role of intention in this study, particularly within the high-activity profiles, deserves mention. While an intention-behavior gap is commonly reported, our person-centered results show that when intention is coupled with high perceived behavioral control and positive attitudes. This underscores the value of LPA: it reveals that the relationship between intention and behavior is not uniform but is moderated by within-person configurations of other TPB constructs²⁷.

An interesting nuance pertains to subjective norms. The Autonomously Active Profile, while highly active, did not report the highest subjective norms. This does not imply norms are unimportant. Instead, it may indicate that for these students, PA has become so integrated into their self-concept and routine that its practice is less dependent on perceived social pressure, reflecting a shift from externally regulated to internally regulated motivation.

Practical suggestions

The identification of four distinct latent profiles offers a concrete foundation for designing tiered interventions to improve physical activity among college students.

For the Disengaged Profile, generic motivational appeals will likely fail so that interventions should first focus on creating a sense of belonging. This could involve low-threshold, non-competitive, and socially focused activities (e.g., walking groups, recreational games) that reduce anxiety and build initial confidence. Emphasizing the social and mental well-being benefits of PA over purely physical outcomes may resonate more.

For the Moderately Driven Profile, the goal is to help internalize motivation. Strategies based on Self-Determination Theory are highly relevant⁴⁹. University programs can foster autonomy, competence, and relatedness to shift motivation from external compliance to personal value.

For the Autonomously Active Profile, the focus should be on maintaining this valuable self-determined state and preventing burnout. Providing access to varied and challenging activities, advanced skill workshops, and leadership opportunities (e.g., as peer mentors or club leaders) can sustain their interest and integrate PA deeper into their identity.

For the Fully Integrated Profile, these students are PA champions. They should be leveraged as peer influencers and role models within the student community. Engaging them in promotional campaigns and peer-mentoring programs can help transmit positive norms and attitudes to other groups, effectively utilizing the social network within the university.

Limitations and future directions

Several limitations must be acknowledged. Firstly, the cross-sectional design precludes causal inferences about the relationships between TPB constructs, profile membership, and PA. The profiles represent current motivational configurations, not fixed developmental stages. Secondly, reliance on self-reported data may introduce bias, though we employed statistical checks for common method variance. Third, the sample, while adequate for LPA, was drawn from two universities, which may affect generalizability to all Chinese college students.

Future research should prioritize longitudinal studies to examine the stability of these profiles and their predictive validity for long-term PA engagement. Experimental or intervention studies are needed to test the efficacy of the profile-tailored strategies suggested above. Incorporating objective measures of PA (e.g., accelerometers) would strengthen the findings. Finally, exploring the role of deeper cultural factors, such as collectivism and academic pressure, in shaping these motivational profiles would enrich the cultural context of this research. The motivational behavior classification system provides an excellent framework for designing and evaluating such nuanced SDT-based interventions in future work⁴⁹.

Conclusion

This study used a person-centered approach to identify four distinct motivational profiles among Chinese college students regarding physical activity. The findings reveal significant heterogeneity in how attitudes, norms, control, and intention combine within individuals, moving beyond the assumption of a single motivational pathway. These profiles were meaningfully linked to different levels of physical activity engagement. The primary contribution of this work is to demonstrate that motivational processes are not uniform. This insight challenges one-size-fits-all intervention strategies and underscores the need for tailored approaches that address the specific configuration of beliefs held by different student subgroups to effectively promote physical activity.

Author contributions

Ying Zhao: Conceptualization, Formal analysis, Writing-original draft. Qinghua Wu: Data collecting, Formal analysis Yueqiang Dai: Methodology, Writing – review & editing, Funding acquisition.

Funding

This work was supported by the Scientific Research Start-up Fund for Highlevel Talents of Sanming University (Grant No. 25YG12S) and the Fujian Provincial Education and Scientific Research Program for Young and Middle-aged Teachers (Grant No. JAS25141), awarded to Assistant Professor Ying Zhao.

Data availability

The datasets generated and analyzed during the current study are openly available in Science Data Bank (ScienceDB) at 10.57760/sciencedb.28975.

Declarations

Competing Interests

The authors declare no competing interests.

Ethics

The studies involving humans were approved by the Human Experimental Ethics Inspection of Guangzhou Sport University (No. 2025LCLL-009). The studies were conducted by the local legislation and institutional requirements.

Generative Al statement

The authors used generative AI tools during the manuscript preparation to refine the language, improve readability, and ensure linguistic precision. It is important to note that all scholarly ideas, data analysis, interpretations, and conclusions remain the sole intellectual contribution of the human authors.