The present study aims to reveal the effects of test type, pronunciation and proficiency levels of the students on speaking test scores. A total of 147 Turkish EFL students consisting of 38 beginner, 36 elementary, 37 pre-intermediate and 36 intermediate levels participated in the study. Presentation as planned, and paired speaking test as unplanned one were used as instruments to figure out the effects of two different test types on the test scores. Being prepared for their performances, the students did their presentations lasting between 5 and 8 minutes in front of two raters. The same participants, at the end of each level, were invited to the paired-speaking tests designed for the students to show their spontaneous performances. The performances in both tests were scored through a scale with five criteria, and the criteria along with overall scores were examined through Paired Samples t-test to reveal the effects of test type, and through bivariate regression to see the proportion of pronunciation aspect on overall scores. The results showed that in beginner level, although no differences were found in the overall scores, the test type induced differences in pronunciation, vocabulary and relevance aspect of their speaking performances. Similar results were found in elementary level besides the difference found in accuracy aspect, which resulted in a significant difference in overall scores. However, in pre-intermediate level, the only significant difference was found in pronunciation. On the other hand, in intermediate level all the aspects along with the overall scores were found to be affected by the test type except for fluency and pronunciation. The bivariate regression revealed that the effect of pronunciation sub-score on overall scores is significant in each level and test type.