Myocardial infarction (MI), also known as a heart attack, is a leading cause of mortality in the world. Spatial vectorcardiogram (VCG) signals are recorded on the body surface to monitor the underlying cardiac electrical activities in three orthogonal directions of the body, namely, frontal, transverse, and sagittal planes. The 3-D VCG vector loops provide a new way to study the cardiac dynamical behaviors, as opposed to the conventional time-delay reconstructed phase space from a single ECG trace. However, few, if any, previous approaches studied the relationships between cardiac disorders and recurrence patterns in VCG signals. This paper presents the recurrence quantification analysis (RQA) of VCG signals in multiple wavelet scales for the identification of cardiac disorders. The linear classification models using multiscale RQA features were shown to detect MI with an average sensitivity of 96.5% and an average specificity of 75% in the randomized classification experiments of PhysioNet Physikalisch-Technische Bundesanstalt database, which is comparable to the performance of human experts. This study is strongly indicative of potential automated MI classification algorithms for diagnostic and therapeutic purposes.