The purpose of this paper is to analyze and detect changes in body position (BPC) during electrocardiogram (ECG) recording. These changes are often manifested as shifts in the electrical axis and may be misclassified as ischemic changes during ambulatory monitoring. We investigate two ECG signal processing methods for detecting BPCs. Different schemes for feature extraction are used (spatial and scalar), while preprocessing, trend postprocessing and detection are identical. The spatial approach is based on VCG loop rotation angles and the scalar approach is based on the Karhunen-Loève transform (KLT) coefficients. The methods are evaluated on two different databases: a database with annotated BPCs and the STAFF III database with recordings from rest and during angioplasty-induced ischemia but not including BPCs. The angle-based detector results in performance values of detection probability PD = 95%, false alarm probability PF = 3% in the BPC database and false alarm rate in the STAFF III database in control ECGs during rest RF(c) = 2 h(-1) (episodes per hour) and in ischemia recordings during angioplasty RF(a) = 7 h(-1), whereas the KLT-based detector produces values of PD = 89%, PF = 3%, RF(c) = 4 h(-1), and RF(a) = 11 h(-1), respectively. Including information on noise level in the detection process to reduce the number of false alarms, performance values of PD approximately equal to 90%, PF approximately equal to 1%, RF(c) approximately equal to 1 h(-1) and RF(a) approximately equal to 2 h(-1) are obtained with both methods. It is concluded that reliable detection of BPCs may be achieved using the ECG signal and should work in parallel to ischemia detectors.