Last year I helped supervise a study of mindfulness among 344 university students in Pakistan, using the 15-item Mindful Attention Awareness Scale. The results were interesting, modest, and easy to misread, which is exactly why I want to describe them carefully. We found that about one in three students fell into a low-mindfulness group, that women scored lower on average even after adjustment, and that social media use was not an independent predictor once other factors were accounted for. Our model explained just 4.2% of the variation in scores. Here is what that does and does not mean.
Mindfulness in university students: the study design
We surveyed 344 undergraduate and graduate students aged 18 to 28 (mean age 21.76), recruited online and in person during 2025. Alongside the MAAS, we collected demographic, academic, health, activity, phone-use, and social media variables. The analysis combined reliability testing (Cronbach alpha), K-means clustering to form score groups, chi-square tests, one-way ANOVA, and multiple linear regression in SPSS 26.
My contribution ran through research planning, guidance of our student collaborators, analytical thinking, interpretation of the findings, and manuscript development. I spent a lot of that time making sure the discussion said only what the statistics could support. That discipline is the subject of this post.
What we found
- The MAAS showed strong internal consistency in our sample (alpha .894). The instrument measured reliably, whatever else you say about the results.
- K-means clustering produced three score groups: low (32.8%, 113 students, mean 41.84, SD 5.55), middle (47.4%, 163 students, mean 55.94, SD 4.57), and high (19.8%, 68 students, mean 75.50, SD 7.12).
- Women were overrepresented in the low-score group and averaged about five points lower after adjustment (coefficient -5.16, p = .001).
- Social media frequency was associated with group membership in the simple comparison, but it was not an independent predictor after gender, academic year, and socioeconomic status were considered (p = .228).
- The full regression model explained 4.2% of the variation in mindfulness scores (R squared 0.042).
What these results do not show
This is the section I care about most. The design was cross-sectional, so nothing here shows cause. The social media finding is the one people want to overstate: because the association vanished after adjustment, this study does not show that cutting social media use would raise anyone’s mindfulness. It also does not show the reverse.
The gender finding needs the same care. Women averaging five points lower is a real, adjusted result, but gender alone explains only a fraction of individual differences, and the groups themselves are data-driven clusters, not clinical cutoffs. Nobody should be sorted or labeled on this basis.
And the 4.2% deserves emphasis rather than embarrassment. It means nearly 96% of the variation in mindfulness scores had nothing to do with the variables we measured. Stress, sleep, rumination, mental health, and academic demands were not in the model. The honest reading is that we mapped a small corner of a large room.
Why I am telling you about a small R-squared
Researchers face constant pressure to make findings sound bigger than they are. Reporting that your model explains 4.2% of variance feels like confessing a failure. It is not. It is information. It tells the next researcher where not to look, and it tells universities that simple demographic targeting will not identify who struggles with attention. A small, honest number moves science further than a large, stretched one. If you supervise students, by the way, this is one of the hardest lessons to teach, and one of the most valuable.
What could come next
The study supports further work, not final answers. Universities could pilot broadly available attention and mindfulness support and evaluate the outcomes properly. Digital-wellness education is reasonable to explore, as long as nobody claims our data proves it works. And the gender pattern warrants further study with better measures of the things we did not model. Longitudinal or experimental designs are the natural next step.
Limitations, stated plainly
Cross-sectional data cannot determine cause. Self-report invites social-desirability bias. The Pakistani student sample limits generalizability. We did not name universities, cities, or the sampling frame. Low, medium, and high are statistical groups, not diagnoses. Keep all of that in mind before quoting any single number from this study, including mine.
The paper
Nisar S, Khurshid A, Sahar I, Shahid B, Asghar T, Saeed R. Mindfulness and Its Demographic and Behavioral Correlates Among University Students in Pakistan: A Cross-Sectional Study. Multidisciplinary Surgical Research Annals. 2025;3(3):404-415. Published August 3, 2025.
A plain-language summary of this study also lives on my research page. This article discusses published research findings. It is not medical advice.