There are exactly 1,000 milliseconds in one second. That figure is fixed by the International System of Units, which defines the milli- prefix as a factor of one thousandth, so the conversion never changes with region, year, or platform. The boundary worth knowing up front: that 1,000 figure is a pure decimal prefix conversion, while the length of the second itself is set separately by atomic physics. One is arithmetic; the other is metrology.
The Exact Conversion and Why the Prefix Decides It
Every millisecond count in a second traces back to one design choice in the metric system: the prefix milli- means one thousandth. The International Bureau of Weights and Measures, which publishes the SI brochure, lists milli as 10⁻³. So one millisecond equals 0.001 seconds, and multiplying by 1,000 reverses it.
This makes the conversion one of the few in measurement that requires no correction factor, no leap adjustment, and no rounding. You aren't converting between two independently defined units, like inches and centimeters, where both sides carry their own history. You're subdividing a single unit by a power of ten.
Run the arithmetic in either direction and it stays closed. Divide 1 second by 1,000 and you get 0.001 s. Multiply 0.001 s by 1,000 and you're back to 1 s. A stopwatch showing 250 ms is showing a quarter of a second; 500 ms is half; 1,500 ms is one and a half seconds, not "one second plus an error."
The same prefix family gives you the neighboring scales for free. Micro- means 10⁻⁶, so a second holds 1,000,000 microseconds. Nano- means 10⁻⁹, so it holds 1,000,000,000 nanoseconds. Each step is another factor of 1,000, which is why engineers can move between millisecond, microsecond, and nanosecond work without touching a conversion table.
How the Second Itself Is Defined Today
The second is no longer tied to the Earth's rotation, and that shift matters for anyone using millisecond precision. Since 1967, the SI second has been defined by the frequency of the cesium-133 atom: exactly 9,192,631,770 cycles of the radiation corresponding to the transition between two hyperfine levels of its ground state. The International Bureau of Weights and Measures controls that definition.
That fixed atomic count is what makes the millisecond dependable. Because the second is pinned to a quantum transition, it doesn't stretch or shrink with the planet's spin. A millisecond is always the same duration anywhere on Earth, which is the property that global timing systems depend on.
The older definition is worth knowing because it explains a common confusion. Before 1967, the second was a fraction of the mean solar day, roughly 1/86,400 of it. A day's length fluctuates with tides, earthquakes, and the redistribution of mass in the Earth's crust, so a solar-day second was never perfectly stable.
That instability is why your phone's clock and the atomic clocks in GPS satellites can diverge if left uncorrected. The fix is the leap second, inserted by international agreement when atomic time and Earth rotation drift apart by about 0.9 seconds. The last one was added on December 31, 2016, per the International Earth Rotation and Reference Systems Service. A single leap second equals 1,000 milliseconds, and the body that decides it is the same service.
Milliseconds at the Human Scale
Human perception sits well above the millisecond, which is why the unit feels abstract. Research on auditory and visual timing puts the threshold for noticing two separate events at roughly 20 to 30 milliseconds. Below that, clicks blur into a single sound rather than a fast pair.
Reaction time is the more concrete anchor. A typical adult responds to a simple visual stimulus in about 200 to 250 milliseconds, a figure repeated across decades of experimental psychology. Elite sprinters leave the starting blocks faster, but not arbitrarily so: World Athletics rules treat a reaction under 100 milliseconds as a false start, on the reasoning that no human can respond that quickly to the gun.
Blinking runs slower than most people guess. A full blink takes roughly 100 to 400 milliseconds, and the eyelid's closing phase alone is about 100 milliseconds. If you've ever missed a detail in a video because you blinked, you lost something in that 100 ms window.
The practical upshot is that everyday life is measured in hundreds of milliseconds. Anything you can consciously see as "fast" is already in the 100 ms range or higher. A millisecond is the scale of machinery, not of perception.
Milliseconds Per Second Across Common Hardware and Standards
The interesting variation isn't in the conversion, which is constant, but in how precisely different systems slice a second into milliseconds and whether they get it right over long runs. Here's how common platforms compare on granularity and drift.
| System or standard | Smallest typical tick | Millisecond granularity | Known drift behavior |
|---|---|---|---|
| JavaScript Date.now() | 1 ms (integer) | Whole milliseconds only | Sub-millisecond fraction discarded; can go backward if the system clock is adjusted |
| Windows default system timer | ~15.6 ms | Coarse unless raised | Default timer resolution rounds sleeps up |
| Linux clock_gettime | 1 ns | Full millisecond and finer | Reads a hardware clock; no rounding to 1 ms |
| GPS time signal | ~20–30 ns | Far below 1 ms | Atomic-referenced; no leap-second drift by design |
| NTP-synced computer clock | 1–10 ms typical | Millisecond | Corrected by network; may step or slew |
| 60 Hz display refresh | 16.67 ms per frame | Not 1 ms aligned | Frame boundaries quantize visible timing |
| World Athletics false-start rule | 100 ms | Millisecond measurement | Rule threshold, not a device limit |
The table exposes a mistake people make constantly: they assume a millisecond-resolution clock is a millisecond-accurate clock. Windows' default timer illustrates the gap. Until a program raises the timer resolution, its default granularity is about 15.6 milliseconds, so a request to sleep for 1 millisecond may round up to roughly 15 ms. That's a 15-fold error on a function that advertises millisecond units.
On the other side, Linux's clock_gettime can report nanoseconds, which means the millisecond value is a truncation rather than a limit. A browser's Date.now() truncates toward the millisecond, so two calls 0.4 ms apart can return the same number even though real time moved. The conversion stays 1,000; the measurement underneath does not always cooperate.
Where the Conversion Breaks: Performance Measurement Traps
The clean 1,000-to-1 ratio invites a trap in timing code: dividing by 1,000 assumes the underlying counter started at zero and never drifted. Most system clocks violate both assumptions. Measure a function with a wall clock and you inherit whatever drift the clock accumulated since boot or since the last network correction.
A common failure case is benchmarking with the system clock on a machine synced by NTP. If the clock steps backward during your test run, elapsed time can come out negative or absurdly small. The clamp that follows hides the problem rather than fixing it. Monotonic clocks exist for exactly this reason and should be the default choice for duration measurement.
Timer resolution causes a subtler error. On a system where the effective tick is 15.6 ms, a loop that computes "elapsed = end − start" in milliseconds has a granularity floor of about 16 ms per sample. Averaging thousands of runs hides that floor; timing a single run exposes it. Worse, consecutive samples can appear identical even when real durations differ.
The third trap is truncation compounding. If you measure 100 sub-millisecond operations individually and each reading truncates down by an average of 0.5 ms, your total can understate real time by 50 ms. That's not the conversion's fault, but it's the conversion that gets the blame when the numbers don't add up.
Converting Seconds and Milliseconds Without Losing Count
Once you know the ratio, the conversions are mechanical, but direction and decimals trip people. The table below covers the cases that show up in practice, with the arithmetic shown so you can verify any row by hand.
| Seconds | Milliseconds | Microseconds | Nanoseconds |
|---|---|---|---|
| 0.001 | 1 | 1,000 | 1,000,000 |
| 0.01 | 10 | 10,000 | 10,000,000 |
| 0.1 | 100 | 100,000 | 100,000,000 |
| 0.25 | 250 | 250,000 | 250,000,000 |
| 1 | 1,000 | 1,000,000 | 1,000,000,000 |
| 60 | 60,000 | 60,000,000 | 60,000,000,000 |
| 3,600 (one hour) | 3,600,000 | 3,600,000,000 | 3,600,000,000,000 |
| 86,400 (one day) | 86,400,000 | 86,400,000,000 | 86,400,000,000,000 |
The multiplication by 1,000 is the whole method, so the safe habit is to move the decimal three places and keep the unit attached. Convert 2.5 seconds and you get 2,500 milliseconds. Convert 0.08 seconds and you get 80 milliseconds. Going the other way, drop three decimal places: 450 milliseconds is 0.45 seconds.
Watch for decimal drift in spreadsheet and programming conversions. Entering 1.5 seconds and multiplying by 1,000 gives 1,500, but if you instead add 1 second (1,000 ms) plus 0.5 seconds (500 ms) in separate cells and one formula treats 0.5 as a percentage, you'll be off by 500. The unit, not the ratio, is usually where the error enters.
Milliseconds in Computing, Audio, and Networking
Millisecond budgets are how engineers decide whether a system feels instant, and the thresholds are surprisingly specific. A response within about 100 ms is generally perceived as immediate. The standard figure, used in interface guidance for decades, treats 0.1 second as the boundary of "real-time" sensation.
Audio makes the number concrete. At a 44,100 Hz sample rate, which is standard for CD-quality audio, one millisecond holds exactly 44.1 samples. That means a 10 ms latency in an audio interface is about 441 samples, and musicians notice it. Round-trip monitoring latency above roughly 10 ms starts to feel disconnected when playing.
Networking turns the millisecond into a distance measure. Google's public DNS service advertises typical response times in the 10 to 25 ms range for users near its resolvers, and each additional 100 ms of latency measurably affects how quickly users abandon a page. The conversion is trivial; the engineering pressure to shave milliseconds is not.
Distinguishing Milliseconds From Seconds and Similar Units
The unit most often confused with the millisecond is the microsecond, partly because both appear in performance logs and both get abbreviated. A microsecond is one thousandth of a millisecond, so a second contains 1,000,000 microseconds. A nanosecond is one thousandth of a microsecond, giving 1,000,000,000 per second.
A frequent misuse is labeling a microsecond-reading clock as millisecond-precision. If a profiler reports 3,000,000 units per second, it's counting microseconds, not milliseconds. Multiplying when you should divide, or the reverse, produces errors of a thousandfold, not a small percentage.
The other confusion is between the millisecond and the "tick" of a particular clock. Windows ticks, .NET ticks, and JavaScript timestamps all have their own scales, and a .NET tick, for reference, is 100 nanoseconds, not 1 millisecond. Never assume a field called "ticks" is milliseconds; check the platform documentation.
The Unambiguous Rule for Any Millisecond Question
The conversion's simplicity makes it easy to state as a rule and hard to apply when a tool reports something adjacent. When a device or API claims millisecond values, the question to ask is whether it's measuring milliseconds or just displaying them. A 1,000-to-1 ratio describes the units, not the accuracy.
Here's the decision path worth internalizing. If you're converting a duration, multiply seconds by 1,000 and you're done. If you're measuring one, use a monotonic clock, not the wall clock. If you're comparing across systems, check each system's effective tick before trusting the digits.
A worked example shows why this matters. Suppose you measure 12.5 seconds with a system clock that drifted backward by 4 ms during the run. The naive conversion reports 12,500 ms. The actual duration was 12,496 ms, and the 4 ms gap is invisible until a task with a tight deadline misses it. Multiply by 1,000, yes, but subtract the drift first.
What the Number Means for Anyone Building or Timing Things
The conversion's simplicity makes it easy to state as a rule and hard to apply when a tool reports something adjacent. When a device or API claims millisecond values, the question to ask is whether it's measuring milliseconds or just displaying them. A 1,000-to-1 ratio describes the units, not the accuracy.
Here's the decision path worth internalizing. If you're converting a duration, multiply seconds by 1,000 and you're done. If you're measuring one, use a monotonic clock, not the wall clock. If you're comparing across systems, check each system's effective tick before trusting the digits.
A worked example shows why this matters. Suppose you measure 12.5 seconds with a system clock that drifted backward by 4 ms during the run. The naive conversion reports 12,500 ms. The actual duration was 12,496 ms, and the 4 ms gap is invisible until a task with a tight deadline misses it. Multiply by 1,000, yes, but subtract the drift first.
FAQ
Is a millisecond exactly one thousandth of a second?
Yes, without adjustment. The SI brochure lists milli as 10⁻³, and that definition does not vary by country or year. The only time this becomes subtle is when a piece of software reports milliseconds while measuring something coarser or finer, in which case you're reading a rounded value, not a different conversion.
Why does my sleep(1) take about 15 milliseconds on Windows?
Because the default system timer resolution on Windows is roughly 15.6 milliseconds, and a requested sleep rounds up to the next tick. Programs that need finer granularity call the time-begin-period function to raise it. If you see a 15 ms floor, you're observing the timer, not evidence that a second contains fewer milliseconds.
Are milliseconds the same as .NET ticks?
No. A .NET tick is 100 nanoseconds, so one millisecond equals 10,000 .NET ticks. The naming overlap causes real bugs: code that treats ticks as milliseconds will be off by four orders of magnitude. Check the platform's own definition before converting a tick count to seconds.
How many milliseconds are in a minute, an hour, and a day?
A minute holds 60,000 ms, an hour holds 3,600,000 ms, and a day holds 86,400,000 ms, all derived by multiplying seconds by 1,000. These figures assume a 24-hour day and don't account for leap seconds, which add a single second rather than scaling the conversion. A leap second adds exactly 1,000 ms to that day's total.
Can a clock report negative elapsed milliseconds?
Yes, and it's a sign you're using a wall clock rather than a monotonic one. If NTP steps the system clock backward during a measurement, end-minus-start can go negative. The fix is to measure durations with a monotonic clock source, which never jumps backward by design.
How does the 100 ms sprinter rule relate to the conversion?
World Athletics flags a reaction faster than 100 milliseconds as a false start, treating it as anticipatory rather than reactive. That 100 ms equals a tenth of a second. The rule is a human performance threshold, so it doesn't change the conversion, but it's a useful sanity check on what a tenth of a second means physically.
Is the second defined by the Earth's rotation?
No. Since 1967 the SI second has been defined by 9,192,631,770 cycles of cesium-133 radiation, set by the International Bureau of Weights and Measures. Earth's rotation is no longer the reference, although leap seconds are occasionally inserted to keep atomic time aligned with it. The last one was added on December 31, 2016.
Does a 60 Hz display update once per millisecond?
No. A 60 Hz display refreshes about every 16.67 milliseconds, so it updates roughly 60 times a second rather than 1,000. That quantization matters for animation timing: motion finer than about 16 ms cannot be shown as a distinct frame, regardless of how many milliseconds the code computes.