Despite the rapid advancement of computer technology, computers remain incapable of understanding human emotion. As a result, users have often been unaided for their aversive emotion that may take place during their computer tasks. This may be detrimental to positive and productive interactions between users and computers. This paper reviews some empirical studies regarding the effect of emotion on computer work and conceptualizes what constitutes an emotional computer. It is proposed that the emotional computer be designed to understand human emotion and adapt its interface accordingly. This paper raises a number of research questions in relation to such issues as measurement (e.g., automatic detection of human emotion, time delay), signal processing (e.g., accuracy) and user interfaces (e.g., ways to alleviate the intensity of negative emotion). Considering that there has been very little research on the design and aftermath of emotional computers, further studies are urgently needed.