A calibrated mercury thermometer holds about ±0.1 °C (±0.18 °F) under ideal conditions, while a clinical digital thermometer typically holds ±0.1 °C to ±0.2 °C in its specified range—so on paper they are close peers, and the real accuracy gap comes from how each device is used, aged, and calibrated rather than from the sensing principle itself. Mercury wins on drift resistance and self-consistency. Digital wins on speed, readability, and safety. Neither category is uniformly more accurate than the other across every scenario.

The physics inside each device

Mercury thermometers exploit thermal expansion. Borosilicate glass and mercury expand at very different rates, and the narrow capillary amplifies the difference into a readable column. The relation is close to linear over the clinical range because the expansion coefficients stay nearly constant. That linearity is why a single calibration point plus a known ice-point reference can validate the whole scale.

The capillary bore geometry sets sensitivity. A finer bore gives more length per degree but raises surface-tension effects and breakage risk. Constriction above the bulb holds the column in place after removal—the "shake-down" design—so the reading reflects the peak temperature reached, not the current one.

Digital thermometers convert temperature to voltage. A thermistor's resistance falls predictably with heat, and the chip linearizes that curve against stored coefficients. The display value is a computed estimate, so accuracy depends on sensor tolerance, the analog-to-digital converter, and firmware calibration. A probe's thermal time constant, driven by mass and surface area, governs how fast it converges.

Plain language: mercury reads temperature mechanically; digital reads it electronically and does the math for you.

Quantified comparison

The table below reconciles typical manufacturer specifications with the tolerances published in standards. The ASTM E1112 standard for electronic clinical thermometers sets a maximum error of ±0.1 °C in the 37–39 °C band and ±0.2 °C outside it, which is the number to weigh against marketing claims.

Attribute Mercury (clinical glass) Digital (thermistor probe)
Typical accuracy ±0.1 °C (±0.18 °F) ±0.1 to ±0.2 °C (±0.18 to ±0.36 °F)
Response time 3–8 minutes 10–60 seconds
Drift over years Very low if undamaged Firmware and sensor drift possible
Readability 0.1 °C graduations, parallax error Numeric display, no interpolation
Breakage hazard Mercury exposure if broken None from mercury
Field recalibration Ice-point check only Ice bath or factory service
Typical cost Largely phased out of retail USD 8 to USD 60

A 0.2 °C difference is smaller than the natural variation in body temperature across a day, which is roughly 0.5 °C. That matters for interpreting any single reading.

The time-course problem

The biggest hidden error is reading too early. A glass thermometer pulled at two minutes may sit 0.5 °C below the true value. A digital unit that beeps "done" may be predicting final temperature from a rate-of-rise model rather than measuring equilibrium, and that prediction degrades if probe contact is poor.

Elapsed time Mercury column Digital probe
0 seconds At starting temperature At starting temperature
30 seconds Rising, well below final Often beeps; predicted value
60 seconds Still climbing Near equilibrium with good contact
3 minutes Approaching equilibrium Stable
5–8 minutes Equilibrium reached Stable, no further change
After removal Holds peak reading Holds last reading

The professional caveat: a predicted reading is a model output, not a measurement. For clinical decisions near a threshold, wait for the stable value rather than the first beep.

Edge cases that flip the ranking

Contact quality dominates everything. A digital probe in an axillary fold with poor skin contact can read 0.5 °C low regardless of sensor accuracy. Mercury in the same spot suffers the same physics but gives you time to notice the column stalling.

Age and damage matter more than type. A 20-year-old glass thermometer with a separated column is not accurate at all. A digital unit dropped on tile may have a cracked thermistor housing that shifts readings silently.

Site matters. Oral, rectal, axillary, and tympanic readings differ by design, often by 0.3 to 0.5 °C, so comparing a mercury oral reading to a digital axillary reading tests anatomy, not instrumentation.

Low batteries and cold storage distort digital output. A unit left in a freezing car may show implausible values until it warms. Mercury is immune to that failure mode but vulnerable to breakage and to meniscus reading errors.

Counterintuitive detail: the glass thermometer's much-criticized slow response is partly a feature. Slow convergence is harder to fake, and a stalling column is a visible signal that contact is poor.

Common misconceptions

"The mercury one is the gold standard." It was the reference in many labs, but only when calibrated and undamaged. Retail glass thermometers vary in bore quality, and parallax reading error can exceed the device's nominal accuracy.

"Digital means precise, so it must be accurate." Display resolution of 0.1 °C is not accuracy. A unit can show three digits and still be off by 0.3 °C.

"Both are interchangeable at any body site." They are not. Site-specific offsets mean the comparison is only valid at the same site with the same technique.

"Beeping means the reading is final." It often means the prediction model has converged, not that thermal equilibrium is reached.

"Mercury is banned everywhere, so the question is moot." Sales restrictions vary by jurisdiction, and legacy units remain in homes and clinics. The accuracy question still applies to what people actually own.

Practical actions and thresholds

Verify a glass unit with an ice-point check: a slurry of crushed ice and water should read 0.0 °C (32 °F) within the graduation tolerance. Discard or recalibrate if it is off by more than 0.2 °C. Test a digital unit the same way, then compare against a second unit. A 0.3 °C disagreement between two digitals suggests one is out of specification. Use the same site for repeat measurements. Oral readings need 15 minutes after hot or cold drinks. Wait for the stable value, not the first beep, when a decision hinges on a threshold near 38.0 °C (100.4 °F). Replace digital units after a hard drop or when readings look implausible against a reference. Retire any glass thermometer with a separated column, cracked bulb, or unreadable scale.

Red flags to watch:

A reading that does not change across repeated attempts. A digital display that flickers or shows a low-battery icon during measurement. Visible mercury beads or a broken bulb, which requires safe cleanup rather than normal disposal. A glass thermometer whose column will not shake down.

Limits of what can be concluded

Specifications describe new, calibrated units under controlled conditions. Real-world accuracy depends on technique, site, age, and contact quality, and those variables often outweigh the device category. Without a traceable calibration check, you cannot claim a specific accuracy figure for any thermometer in your drawer.

The honest summary: neither type is categorically more accurate. Choose digital for speed, safety, and readability, and choose a well-made glass unit for drift resistance and self-checking. Then verify whichever you own against a known reference, because the calibration check, not the sensing principle, determines whether the number on the device means anything.