A bit is the smallest possible unit of information, representing a single choice between two alternatives: yes or no, on or off, one or zero. The term, coined by mathematician Claude Shannon in 1948 as a contraction of “binary digit,” underpins virtually every system that stores, transmits, or processes information, from smartphones and fiber-optic cables to DNA and the human brain. But the concept runs deeper than computing, touching physics, biology, and even how we think.
What a Bit Actually Represents
A bit does not have to be electronic. It is any physical system that can exist in one of exactly two distinguishable states. Inside a computer, that might be a tiny capacitor that is either charged or discharged, or a magnetic region on a hard drive that points north or south. In fiber-optic communication, it could be a pulse of light that is either present or absent. In a coin flip, it is heads or tails. The medium does not matter; what matters is that exactly two outcomes are possible and you learn which one you got.
When you string bits together, they can represent much richer information. Two bits give you four possible combinations (00, 01, 10, 11). Three bits give you eight. Each bit you add doubles the number of possible values, which is why computing power scales so dramatically with seemingly small hardware improvements. A group of eight bits is called a byte, and a byte can represent 256 distinct values, enough to encode any character in the English alphabet (upper and lower case), all the digits, punctuation marks, and still have room left over.
Bits in Everyday Scale
Most of the file sizes and data plans you encounter daily are measured in bytes and their multiples, not raw bits. A kilobyte is roughly a thousand bytes (1,024, technically), a megabyte about a million, a gigabyte about a billion, and a terabyte about a trillion. A single high-resolution photograph on your phone might be five megabytes, which is 40 million bits. A two-hour movie streamed in high definition uses around four to six gigabytes, or somewhere north of 30 billion bits. Your brain does not need to track those numbers, but they give a sense of how quickly bits add up when you are encoding the richness of images, sound, and video.
Internet speeds, meanwhile, are typically quoted in bits per second rather than bytes per second. When your provider advertises “300 Mbps,” that is 300 megabits per second, which works out to roughly 37.5 megabytes per second. The distinction trips people up constantly: a file that is 100 megabytes will not download in one second on a 100 Mbps connection because megabits and megabytes are different by a factor of eight.
How Bits Travel Through Fiber-Optic Networks
Modern long-haul communication networks push bits through glass fibers as pulses of laser light, and engineers have gotten remarkably good at cramming more bits into each second of signal. One way they measure progress is spectral efficiency, which describes how many bits per second you can squeeze into each hertz of available bandwidth. Field demonstrations of high-capacity fiber systems have achieved spectral efficiencies around 4 bits per second per hertz using advanced modulation formats on tightly packed wavelength channels.1Optica Publishing Group. Field Transmission of 100 G and Beyond: Multiple Baud Rates and Mixed Line Rates Using Nyquist-WDM Technology That might not sound impressive as a raw number, but multiply it across the enormous bandwidths available in optical fiber and you get data rates measured in terabits per second on a single strand of glass.
The trick is that bits traveling through any real-world channel pick up noise, distortion, and timing errors along the way. At some point, packing more bits into the same bandwidth means each individual bit is harder to distinguish from its neighbors, and errors creep in. That tradeoff between speed and reliability is a fundamental constraint Shannon himself described, and it shapes the design of every communication system you use.
Catching Mistakes With Extra Bits
Because bits can flip accidentally, whether from electrical noise in a cable, cosmic rays hitting a satellite’s memory chip, or interference on a wireless link, engineers add extra bits whose sole purpose is to catch and fix those errors. The idea is straightforward: instead of sending just the data bits, you send a few additional bits calculated from the data in a specific pattern. The receiver runs the same calculation and checks whether the extra bits match. If they do not, the receiver knows something went wrong and can often figure out which bit flipped.
One of the classic approaches is the Hamming code, developed in the 1950s and still used in modern hardware. A version tested for use aboard small satellites in low Earth orbit, for example, can correct any single flipped bit and detect when two bits have flipped, all by adding just a handful of extra bits for every 11 bits of actual data.2Journal of Electrical and Computer Engineering. Error Detection and Correction On-Board Nanosatellites Using Hamming Codes That kind of overhead is small enough to be practical even for tiny spacecraft with tight power and space budgets, yet it makes the difference between reliable data and garbled transmissions in a harsh radiation environment.
More sophisticated error-correcting codes are used in everything from Wi-Fi and cell networks to QR codes and Blu-ray discs. The principle is always the same: spend a few extra bits to protect the ones that carry your actual information. It is one of the most quietly important ideas in all of information technology.
The Physical Cost of Erasing a Bit
A bit is an abstraction, but every real bit lives in some physical system, and physics imposes a minimum energy cost on manipulating it. In 1961, physicist Rolf Landauer argued that erasing a single bit of information, specifically the irreversible act of resetting a memory element to a known state, must release a tiny but nonzero amount of heat into the surroundings. That minimum, known as Landauer’s limit, is set by the temperature of the environment and a fundamental physical constant. At room temperature, it works out to an almost absurdly small amount of energy, roughly three-billionths of a trillionth of a joule per bit erased.3PubMed Central. Experimental test of Landauer’s principle in single-bit operations on nanomagnetic memory bits
Experiments using nanoscale magnetic memory elements have confirmed that this limit is real, not just theoretical. In practice, today’s computers dissipate vastly more energy per bit operation than Landauer’s minimum because of engineering inefficiencies, resistance in wires, leakage currents, and so on. But as transistors have shrunk and computing has grown more energy-hungry at global scale, the gap between practical energy costs and the theoretical floor has become more than an academic curiosity. It represents the ultimate boundary on how energy-efficient computation can ever become, and it connects the abstract world of bits directly to the thermodynamics of the physical universe.
How Many Bits English Text Contains
Shannon did not just define the bit; he also developed the mathematical framework for measuring how much information any source actually produces. One of his early applications was asking how many bits each character in written English really carries. Because English is heavily redundant (after reading “th,” you can almost certainly guess the next letter is “e” or some other vowel), each character conveys less than the theoretical maximum of about 4.7 bits you would need if all 26 letters and a space appeared with equal probability.
Researchers have applied Shannon’s methods to large samples of real English text, processing millions of characters to measure how much genuine surprise each new character delivers once you account for the patterns of the language.4arXiv. A New Look at the Classical Entropy of Written English The estimates typically fall in the range of about one to one and a half bits per character, meaning English is roughly 70 to 80 percent redundant. That redundancy is actually useful: it is the reason you can read a sentence with a few missing letters or tolerate a typo without losing the meaning. In information-theoretic terms, the “extra” bits in English text serve as a kind of natural error correction, similar in spirit to the engineered redundancy that protects data on a satellite.
Storing Bits in DNA
The four-letter alphabet of DNA (A, T, G, C) naturally encodes two bits per position, since four options gives you exactly two binary choices stacked together. But researchers working on synthetic DNA-based data storage have pushed beyond that baseline by exploiting additional chemical variations. By using degenerate bases, synthetic nucleotides that can represent more than one of the standard four letters, a team experimentally demonstrated an information capacity of about 3.4 bits per character in synthesized DNA.5Scientific Reports. High information capacity DNA-based data storage with augmented encoding characters using degenerate bases That is more than twice the roughly 1.6 bits per character achieved by previous DNA storage methods that stuck to the standard four-letter code.
DNA-based storage is appealing because DNA is extraordinarily dense and durable. A few grams of it could, in principle, hold all the data humanity has ever produced, and it remains readable after thousands of years under the right conditions. The field is still far from being commercially practical for everyday use, with synthesis and sequencing costs remaining high and read/write speeds orders of magnitude slower than electronic storage. But as a long-term archival medium, DNA highlights something interesting about the bit: information is substrate-agnostic. The same bit that lives as a voltage in a chip can live as a molecular bond in a strand of synthetic DNA.
How Fast Your Brain Processes Bits
Here is a genuinely surprising finding about bits and biology. Despite the enormous bandwidth of your sensory systems, which collectively gather data at roughly a billion bits per second from your eyes, ears, skin, and other inputs, your conscious information throughput runs at only about 10 bits per second.6PubMed Central. The unbearable slowness of being: Why do we live at 10 bits/s? That is not a rough estimate; it is a figure consistent across decades of measurements on tasks like reading, typing, speaking, and performing skilled motor actions.
To put that in perspective, 10 bits per second means your conscious experience processes information at roughly the rate of a 1950s teletype machine. Your retina alone sends about 10 million bits per second to your brain, but the vast majority of that data is processed unconsciously, filtered, compressed, and discarded before it ever reaches your awareness. The roughly hundred-million-fold gap between sensory input and conscious processing is one of the most striking numbers in all of neuroscience, and it raises deep questions about what consciousness actually does. If your brain’s unconscious systems handle the heavy data processing, your conscious mind seems to act more like a slow, careful executive, picking out a tiny sliver of what matters from a firehose of sensory data.
This also explains why you cannot meaningfully multitask at a conscious level. Ten bits per second is barely enough to follow one conversation or read one line of text. Trying to split that budget across two demanding tasks means each gets even less, which is why talking on the phone while driving is dangerous even without holding the phone.
Probabilistic Bits and the Frontier of Computing
Classical bits are deterministic: they are either 0 or 1 at any given moment. Quantum bits, or qubits, are the well-known alternative, exploiting quantum superposition to exist in both states simultaneously. But there is a less famous middle ground gaining traction in research labs: the probabilistic bit, or p-bit. A p-bit fluctuates randomly between 0 and 1, but its probability of being in one state versus the other can be precisely tuned.
Recent work has demonstrated p-bit devices that generate their randomness from physical processes. One approach uses light-induced charge switching in a semiconductor junction, where the randomness comes from photon-driven effects and the probability is controlled by applying a tiny voltage, less than a millivolt.7PubMed Central. Sub-mV tunable photonic p-bits for probabilistic computing The key insight is that the source of randomness and the control knob for probability are independent, which makes these devices stable and scalable. Another approach builds p-bits entirely from standard chip-fabrication technology, using a component called a bistable resistor paired with a comparator circuit to produce the same kind of tunable random output.8Advanced Functional Materials. Fully CMOS‐Based p‐Bits with a Bistable Resistor for Probabilistic Computing
Why bother? Certain classes of hard problems, like optimization, factoring large numbers, and solving satisfiability puzzles, are well suited to probabilistic search. Instead of grinding through every possibility or requiring the exotic hardware of a quantum computer, a network of p-bits can explore the solution space stochastically, often finding good answers with far less energy than brute-force approaches. Researchers have already demonstrated this by using p-bit networks to factor integers and solve combinatorial optimization tasks. Probabilistic computing will not replace classical or quantum computing, but it is carving out a niche for problems where controlled randomness is an asset rather than a nuisance.
When “Bit” Means Something Slightly Different
The word “bit” shows up in several contexts that look similar but carry subtle distinctions worth knowing. In information theory, a bit is a measure of surprise or uncertainty: receiving a bit of information means your uncertainty about something was cut in half. In computing, a bit is a concrete physical thing, a transistor state, a magnetic orientation, a voltage level. In communication, bits per second measures throughput. These three meanings overlap but are not identical. A file that is one megabyte in the computing sense does not necessarily contain one megabyte of information in the Shannon sense, because much of the file may be redundant or predictable. Compressed files are smaller precisely because compression strips out the redundancy, bringing the computing-bit count closer to the information-bit count.
You will also encounter bits in contexts that have nothing to do with computers. A musician might refer to a “bit” as a short comedic routine. A drill bit is a cutting tool. These are unrelated to the information-theoretic meaning, but the collision of terminology can cause momentary confusion when you are searching for information about, well, information. In technical contexts, if someone says “bit” without further qualification, they almost always mean the binary digit Shannon defined, the atomic unit that makes the digital world possible.