Discordance has been reported in human chorionic gonadotropin (hCG) concentrations measured by different immunoassay kits. We examined the results for 40 serum samples assayed with 10 different hCG immunoassay kits. Results varied considerably. Individual sample results varied by as much as 58-fold. Average results for different kits varied by as much as 1.4-fold for pregnancy (20 samples) and 2.2-fold for trophoblast disease (20 samples) serum. We investigated the causes of this discordance. hCG or hCG beta are general names for mixtures of hCG, hCG alpha, or hCG beta immunoreactive molecules in serum. These mixtures include regular hCG, nicked hCG (missing peptide linkages at beta 44-45 or beta 47-48), carbohydrate variants of hCG, hCG missing the beta-subunit C-terminal segment, free beta-subunit, beta-core fragment, and free alpha-subunit. We prepared standards for each of these major variants and measured their reactivities in the 10 hCG immunoassay kits. Free beta-subunit reactivity varied from nonrecognition (anti-beta:anti-alpha type kits; Hybritech Tandem-R and others) to overrecognition (one kit had five-fold greater affinity for free beta than for hCG). Kits with antibodies to beta-subunit C-terminal segment (Organon NML and others) failed to recognize hCG missing this segment, a component of serum hCG in trophoblast disease. Kits with anti-hCG antibodies (Serono MAIA-clone and others) had minimal recognition of nicked hCG (12%), a component of all serum hCG samples, and consistently gave the lowest values with all serum samples. We conclude that differences in recognition of nicked hCG, free beta, and these other hCG variants cause discordance in hCG immunoassay results.