Problem solving activities require students to apply mathematics rules on non-routine problems. Mathematical connection ability is required so that students can relate real problems to mathematical model nor relate the interrelationships between the concepts that required to solve problems. This study aims to examine the effectiveness of MEA in geometry and to analyze the mathematical connections ability of students in geometry based on upper, middle, and lower spatial intelligence categories on MEA. The research method is Mixed Methods and research design is Sequential Explanatory Design. The subjects of the study is the 9th grade students. The study started by first spatial intelligence test in experimental class, then followed by MEA learning in experimental class and expository learning in control class. The study ended by mathematical connection ability test in experimental and control classes and last spatial intelligence test in experimental class. The results showed that MEA learning in geometry were effective. The result of mathematical connection ability in geometry of upper spatial intelligence category students has reached five indicators of mathematical connection ability well. Students of middle spatial intelligence category on first, second and fourth indicators have been maximized, but third and fifth indicators have not been maximized. Students of lower spatial intelligence category on first indicator have been maximized, second indicator has not been maximized, while third, fourth and fifth indicators can not be achieved well. Teachers should know the spatial intelligence of students so that the learning process especially in geometry achieve maximum results.