limit of detection
/ LIM-it of dee-TEK-shun /
Listen for a friend whispering across a noisy café. There is a faintest whisper that you can just tell apart from the background chatter — quieter than that and you genuinely cannot say whether anyone spoke at all. The limit of detection is that threshold for a chemical measurement: the smallest amount of an analyte you can reliably say is present, rather than just noise.
Formally, the limit of detection is the lowest concentration that gives a signal you can confidently distinguish from a blank. A common rule sets it where the signal rises about three times the standard deviation of the blank's noise above the average blank reading. Below this point the analyte may well be there, but its signal is buried in the random scatter, so the only honest answer is 'not detected'.
It matters because reporting a real but tiny amount of a contaminant or drug can have big consequences, and the limit of detection draws the line between 'we can see it' and 'we cannot'. The caveat is that it is only a yes-or-no threshold: detecting something is not the same as measuring how much there is. For a trustworthy number you need the higher limit of quantitation.
An atomic absorption method for lead in drinking water has a limit of detection of 2 micrograms per litre: a sample reading 1 µg/L is reported as 'below detection limit', while one at 5 µg/L is confidently flagged as containing lead.
The smallest amount of lead that can be told apart from blank noise.
A signal-to-noise ratio of about three is the usual rule of thumb for the limit of detection; the higher ratio of about ten marks the limit of quantitation, where you can not only detect but also measure the analyte.