In this paper we study the feasibility of seismocardiography (SCG) for the detection of Atrial Fibrillation (AF). In this preclinical study, data acquired from one patient having paroxysmal AF (no other heart diseases) is used to introduce specific changes in SCG signal due to AF. Observed changes and phenomena create a foundation for the development of SCG-based AF detection algorithms. SCG data was recorded from the sternum of an AF patient in dorso-ventral direction while at rest in a supine position using a three-axis high precision MEMS accelerometer simultaneously with a one-lead ECG. In contrast to ECG, the magnitude of beats registered with SCG varies considerably from beat to beat during AF. We show that the magnitude of the beats is not random but is in relation to beat intervals. It is shown that extra indicators for detecting AF become available when SCG data is combined with electrocardiographical (ECG) data; there is a certain behavior in the electromechanical delay characteristic of the AF. It is discussed how all this information can be taken advantage of in the detection of AF. Today electrocardiography (ECG) is the primary method for diagnosing arrhythmias, but there is a growing need for simpler and more convenient method for detecting asymptomatic AF. Given the very small dimensions of modern MEMS accelerometers (2mm×2mm), a reliable MEMS based measurement may provide totally new venues for arrhythmia detection.