Various studies have rendered self-directed learning disposition to be significant in the learning of mathematics, however several previous studies have pointed the level of self-directed learning disposition to be at a low point. This research is aimed to enhance self-directed learning through implementing a metacognitive strategy in learning mathematical statistics. This research incorporates both methods of quantitative and qualitative ones while employing a sequential explanatory strategy. Research design takes a pretest-posttest control group design (Creswell, 2009). The subjects of the research are students of Mathematics Education Study Program of the University of Kupang, East Nusa Tenggara Province currently taking Mathematical Statistics course. Research sample is filtered from two class levels of mathematical statistics course Research instruments include the Likert scale to measure self-directed learning disposition. Results point an increment of self-directed learning disposition of students who were being taught through either a Metacognitive Strategy Approach (MSA) or Conventional Approach (CA). The Self-directed learning disposition of students based on the MSA approach was found to be better than those of the CA approach. The lowest average increase of the CA class was based on the identification of the goals indicator. Furthermore, the lowest average increase of the MSA class was based on the evaluation of the advantages and the disadvantages of learning. The highest average increase for either MSA or CA classes was observed because of the presence of learning initiatives. The average increase of the presence of learning initiatives indicator for the MSA class was higher than those of the CA class.