Gender recognition has been widely used in research areas like biometric security systems, human computer interaction and verbal/nonverbal communication in recent times. However, due to different appearances of human face and negative environmental effects on face images, gender recognition is still a hot research area in the field of computer vision. And, current gender recognition studies are carried out using geometric, texture and appearance based features of face images. In this study, different appearance based approaches' gender recognition performances on same datasets are investigated and a comparison is made between these methods. For this purpose, firstly, a new dataset which contains different facial expressions is created using a standard camera. Appearance based features are extracted using Principal Component Analysis (PCA) and Local Binary Pattern Histograms (LBPH). Subsequently, learning and classification is performed using Support Vector Machines (SVM). Finally, comparison of these methods' performances on gender recognition is made with our proposed and FERET datasets. The best classification accuracy of 94.53% is obtained when using LBPH+DVM method pair our proposed dataset and accuracy of 91.07% is obtained on FERET dataset when using PCA+SVM method pairs. Consequently, it can be understood that gender recognition problems can be solved under different conditions using appearance based methods.