A restudy of 13 bricks from a 19th-century kiln, using the Shaw AF-demagnetization method of paleointensity determination, revealed alteration of the magnetic carrying natural remanent magnetization (NRM) during heating to the Curie point. Anhysteretic remanences before heating (ARM1) and after heating (ARM2) had similar intensities, but the ARM2 coercivity spectrum was softer than the ARM1 spectrum. The coercivity spectrum of thermoremanence (TRM) softened in a similar way after heating. We therefore calculated a ''corrected TRM'', TRMc=(ARM1/ARM2)TRM at each AF demagnetization step. The NRM vs. TRMc plot was in most cases more nearly linear than the original NRM vs TRM plot. However, the paleointensities determined using uncorrected total NRM/TRM ratios (57 &mgr;T) and using our TRMc values (55.5 &mgr;T) agreed equally well with a recent Thellier-method determination of 56 &mgr;T by Aitken and others. Grain-size alteration of the magnetites has changed their fundamental TRM/ARM ratios as well as their coercivities, making our correction procedure untrustworthy. Lack of significant changes in TRM or ARM coercivity spectra after heating is evidence of minimal grain-size alteration, and remains the best criterion for reliable Shaw-method paleointensities. ¿ American Geophysical Union 1987