The published research studies to date indicate that pair programming has a positive impact on some aspects of students' performance. In the normal practice of pairing programming in the academic field, the students were paired by assigning partners according to their level of programming skill. In other words, students were paired according to their programming compatibility that was perceived by their lecturers. However, research studies did not attempt to identify the main element that the students are looking into when they are given the freedom to select their partner in pair programming practice. An experiment with 76 students during a one-week programming workshop shows that 59.2% will choose their partner according to gender while 30.3% will choose their partner based on the ethnics group. The study shows that only 5.2% of the students focus on the skills of their choice of partner. At the end of the workshop, 96% of the students agree that pairing with a partner helps them in solving a programming problem. However, only 89.2% of the students prefer to work in pairs when solving programming while 5.4% prefer to work as an individual. This initial finding tallies with the other research whereby it shows that pair programming benefits the students in solving a programming problem. Despite the normal belief that the pairs are compatible if they are almost the same level in terms of technical competency in programming, students tend to choose according to gender when they are given a choice.