Since the pioneering studies of Carpenter & Narin (1983), and Narin & Noma (1985), non-patent references (NPRs) in patent documents have been widely used as an indicator of science-technology links. Meyer (2000) reviewed previous work in the patent citation literature and found that citation links between patents and papers are, if not explicitly, at least implicitly viewed as an indication of the contribution of science to technology. Using a sample of 850 patents of New Zealand companies granted by the USPTO between 1976 and 2004, we find evidence of systematic noise in NPR data. We suggest that future research should pay close attention to heterogeneity among countries, and that one should demonstrate more caution in applying and interpreting results based on the NPR methodology.