Emotion recognition from speech plays important role for natural human-computer interaction. This study investigates binary classification performances of 4 fundamental emotion classes in Turkish Emotional Speech (TurES) Database using acoustic features for various classifiers. Results shows that Angry emotion class has higher classification rate (70%-80%) than others; lowest classification rate is obtained as 64% for Happy-Neutral emotion pair. Best classification results are obtained with J48 (C4.5) classifier for all emotion pairs.