The second to fourth digit ratio (2D:4D) indicates androgen exposure during early development. Numerous methods are used in the assessment of 2D:4D, however, some are not reliable enough and others are difficult to perform in large epidemiological studies. We assessed the reliability of 2D:4D using a direct method with a caliper, and compared it with a computer-assisted analysis on scanned images. Both methods were moderately correlated, but the scan method produced slightly lower 2D:4D. The main source of variation was differences between subjects (real variation). Reliability was higher among men and among younger participants. All reliability coefficients were higher than 0.8 when three repeated measurements were averaged. Our results suggest that reliability is influenced by participants’ characteristics. Digit ratios determined directly with calipers are reliable when repeated measurements are averaged.