We present a new method of paleointensity determination based on comparing the thermal demagnetization of natural remanent magnetization (NRM) with that of an artificial total thermoremanent magnetization (TRM). Igneous rocks often contain pseudo-single domain (PSD), multidomain (MD), and/or single domain (SD) particles as magnetic remanence carriers under strong magnetic grain (domain) interactions. The magnetic grain interactions have particular disastrous effects on paleointensity experiments, which make determination of paleointensity unreliable. We have critically examined how magnetic grain interactions affect the Thellier experiment, and have developed a new technique for correcting grain-interaction effects in the experiment of paleointensity estimation. The essential point of our experimental method is that by comparing the thermal demagnetization of natural remanent magnetization (δ NRM_loss
) with that of an artificial total TRM (δ TRM_loss
) for estimating its paleointensity, rather than that by comparing the remaining of NRM during thermal demagnetization (NRM_remaining
) with a progressive TRM_gain
in the traditional Thellier-Coe method, which essentially requires the additivity of partial TRM and independence of pTRMs. Using our new method, a mild alternating field (AF) demagnetization pre-treatment is applied to destroy most of the low coercivity remanence, which makes the samples behave more suitebly for a paleointensity study. We also make an apparent paleointensity estimation with pTRM, which is acquired in the perpendicular direction of NRM in a narrow non-overlapping temperature interval and cooled slowly in air. In this way, the non-ideal behavior of samples is detected most sensitively by the discrepancy between NRM loss and pTRM gain. Finally, we employ an artificial total TRM test to elucidate the relation between TRM_loss
, and to correct interference caused by the non-ideal behavior. We have applied our new method to several representative suites of historical lava flows of known geomagnetic field intensity, and successfully extracted reliable paleointensity with a precision higher than 95% from samples even containing PSD and MD grains.