Accurate measurement of testosterone is important for reproductive endocrinology research, but the validity of direct (nonextraction) testosterone immunoassays, developed and validated for human serum, has not been appraised for application to mouse serum or steroidogenic tissue extracts. Testosterone was measured in serum and extracts of testis or ovary from male and female wild-type mice by 2 commercial direct testosterone immunoassays, with and without preassay extraction, and by the liquid chromatography, tandem mass spectrometry reference method. Results were compared hierarchically by correlation (Kendall's τ), regression (Passing-Bablok), and deviance (Bland-Altman) analysis, under the null hypothesis of perfect agreement between assays (slope = 1, intercept and deviation = 0). For mouse serum, immunoassays displayed an upward bias with performance better for male vs female sera and, within gender, improved by preassay extraction relative to liquid chromatography, tandem mass spectrometry. Testosterone was detectable in all serum samples, but few (male 54%, female 9%) were accurate (within 20% of the reference measurement). For mouse testis extracts, immunoassays were biased upwards, and preassay extraction improved immunoassay performance. Although testosterone was detectable in all extracts, a minority (45%) was accurate. For mouse ovary extracts, all correlations were poor with severe, upward bias, and while testosterone was detectable in all samples, virtually none were accurate. We conclude that these direct testosterone immunoassay kits provide relatively, but not absolutely, accurate results with male mouse serum and testis extracts but not with female mouse serum and ovary extracts, with performance improved by preassay extraction. Whether relative accuracy is fit for purpose depends on the experimental aims, design, and interpretation.