One necessary condition for successful determination of relative paleomagnetic field intensity using anhysteretic remanent magnetization (ARM) methods is reciprocity: a partial ARM, produced by a steady field H applied over a narrow interval ($widetilde{H}$2, $widetilde{H}$1) of alternating field (AF), must demagnetize over the same interval ($widetilde{H}$2, $widetilde{H}$1). Experimentally, we find that partial ARMs of single-domain (SD) and pseudosingle-domain (PSD) grains demagnetize mainly between $widetilde{H}$2 and $widetilde{H}$1, whereas >50% of partial ARMs of large PSD and multidomain (MD) grains are erased below $widetilde{H}$1, giving a low-field tail in the coercivity distribution. Natural pumices, granites, and oceanic basalts violated reciprocity, but lake sediments, gabbros, andesite, and red scoria had relatively small low-coercivity tails and are better candidates for paleointensity work. Using total ARM to simulate natural remanence, we carried out pseudo-Thellier paleointensity determinations for coarse PSD and MD grains. ARM demagnetization outweighed partial ARM acquisition at the same AF step, resulting in convex-down curves of ARM remaining versus partial ARM gained (pseudo-Arai plot). Pseudo-Arai plots predicted from experimentally determined distributions of blocking and unblocking fields agreed well with measured pseudo-Thellier results, in particular explaining convex-down MD curves.