Learning Disabilities (LD) can be categorized into logical, analytical, grammatical, Vocabularyulary, sequential, and inference disabilities. Analysis of such disabilities assists students to identify and strengthen their weak areas. A wide variety of analysis models are proposed by researchers to perform such tasks, but most of these models are highly complex, and cannot be scaled for multimodal parameter sets. To overcome these issues, this text proposes a model for correlative assessment of differential usage patterns in students with-or-without learning disabilities via multimodal analysis. The proposed model initially collects real-time inference sets for students with Learning Disabilities (LD), and without LDs. These sets consist of question-specific recorded responses for ‘Addition’, ‘Carry Propagation’, ‘Basic to Advanced Grammar’, ‘Direct, Inference and Vocabulary Comprehension’, ‘Finding odd-man-out’, ‘Sequencing’, and ‘Pseudo and Sight Spelling’ for different question sets. Answers to these questions and their metadata were processed via a correlative engine which assisted in evaluation of correctness, time needed per question per category, number of skips, number of revisits, and unanswered ratio for different students. This evaluation was combined with temporal analysis in order to identify per-category progress of students. Based on this progress, students were either upgraded to next level, or given lower-level questions, which assisted them to incrementally improve their grades. The model proved that the performance of LD students is 55% less than the non LD students and an average of 18 LD students have achieved an average of 33% of improvement after having multiple attempts of the adaptive lessons. The model uses a correlation function, which enables to identify answering patterns of LD and non-LD students with 98.4% accuracy, thus can be used for clinical scenarios.