The strength of Earth's magnetic field, on timescales greater than a million years, is known primarily from Thellier‐Thellier analyses of igneous rocks. In the Thellier‐Thellier method, the natural rémanent magnetization (NRM) is compared with a thermoremanent magnetization (TRM) imparted during heating in a known laboratory field. Unfortunately, the multiple heating steps required by the method often result in alteration of magnetic minerals in clay‐bearing groundmass of whole rock samples. This alteration may result in the creation of new magnetic minerals that could affect TRM acquisition. Analyses of single plagioclase crystals, which contain magnetic inclusions that retain paleofield values, may offer a way to retrieve paleointensity data that is less affected by alteration. We test this hypothesis through a study of whole rocks and single plagioclase crystals from a mid‐Cretaceous (113–115 Ma) basalt flow of the Rajmahal Traps (northeastern India). Magnetic hysteresis and thermal demagnetization data from plagioclase crystals and whole rock samples are similar, suggesting a common magnetic mineralogy. However, hysteresis data from heated whole rock samples indicate that a fine‐grained magnetic phase has formed. This effect was either absent, or much less pronounced in the single plagioclase crystals tested. Paleointensity data from these crystals suggest a paleofield value of 65.1±5.3 μT (n=15). Data from whole rock samples overlap these values but are on average lower. We attribute the difference in this study to the preferential growth of new magnetic minerals in the whole rock samples, resulting in an enhanced acquisition of TRM and shallowing of the slope of NRM‐lost versus TRM‐gained curves. These data suggest that subtle alteration effects may bias Thellier‐Thellier results from some whole rock samples toward low paleofield values. In these cases the single plagioclase crystal approach may yield more reliable estimates of geomagnetic paleointensity.