The spatial coherence structure of 30 infrasound array detections, with source-to-receiver ranges of 25–6500km, has been measured within the 0.25–1Hz passband. The data were recorded at International Monitoring System (IMS) microbarograph arrays with apertures of between 1 and 4 km. Such array detections are of interest for Comprehensive Nuclear-TestBan Treaty monitoring. The majority of array detections (e.g. 80 percent of recordings in the third-octave passband centred on 0.63 Hz) exhibit spatial coherence loss anisotropy that is consistent with previous lower frequency atmospheric acoustic studies; coherence loss is more rapid perpendicular to the acoustic propagation direction than parallel to it. The thirty array detections display significant interdetection variation in the magnitude of spatial coherence loss. The measurements can be explained by the simultaneous arrival of wave fronts at the recording array with angular beamwidths of between 0.4 and 7◦ and velocity bandwidths of between 2 and 40ms−1. There is a statistically significant positive correlation between source to-receiver range and the magnitude of coherence loss. Acoustic multipathing generated by interactions with fine-scale wind and temperature gradients along stratospheric propagation paths is qualitatively consistent with the observations. In addition, the study indicates that to isolate coherence loss generated by propagation effects, analysis of signals exhibiting high signal-to-noise ratios (SNR) is required (SNR2 >11 in this study). The rapid temporal variations in infrasonic noise observed in recordings at IMS arrays indicates that correcting measured coherence values for the effect of noise, using pre-signal estimates of noise power, is ineffective.