The furcation involvement of 100 molars in 25 patients suffering from moderate to advanced periodontitis was investigated. The horizontal probing attachment level (PAL-H) within the furcations was assessed 2x within 2 weeks using the pressure-calibrated (0.25 N) flexible plastic universal explorer version of the TPS probe (TPS). To determine the measurement error of PAL-H assessments, the standard deviation of single measurements was calculated. The measurements were repeated using a colour-coded Nabers probe and compared to the TPS assessments. 253 furcations were evaluated (100 buccal, 47 lingual, 53 mesiolingual and 53 distolingual, respectively). For buccal, lingual, mesiolingual and distolingual furcations, the standard deviations were 0.486 mm, 0.598 mm, 0.846 mm, 1.039 mm, respectively. Measurement error was less in buccal and lingual furcations than in mesiolingual and distolingual sites (p < 0.005). The agreement of replicate measurements of furcation degrees was excellent for buccal and lingual furcations (weighted kappa [standard error] 0.824 [0.076] and 0.779 [0.107], respectively), but only moderate for mesiolingual and distolingual furcations (weighted kappa 0.688 [0.096] and 0.544 [0.101], respectively). Only in distolingual sites there was a significant (p < 0.025; paired t-test) underestimation of PAL-H by the TPS as compared to the Nabers probe. At all locations the TPS underestimated furcation degrees significantly (p < 0.1; Stuart-Maxwell's chi 2) as compared to Nabers probe. Measurement error of mesiolingual and distolingual furcations was significantly higher than of buccal or lingual sites. The reproducibility of PAL-H measurements in furcations using TPS is comparable to data published for scorings with colour-coded Nabers probes. Through and through furcations are likely to be underestimated by using the flexible plastic version of the TPS probe. Hence, it seems to be unsuitable for a proper assessment of the degree of furcation involvement.