Most of a Bitcoin block is the list of transactions. The header is the short prefix that lets a node name the block, tie it to a parent, and check the proof of work without rereading every transaction first. The header is 80 bytes. Learning those bytes is the difference between “a block happened” and being able to say which claims the block is actually making.
The six fields
In order, the header is version, previous block hash, merkle root, time, nBits, and nonce. Version is 4 bytes. The previous block hash is 32 bytes. The merkle root is 32 bytes. Time, nBits, and nonce are 4 bytes each. Four plus 32 plus 32 plus 4 plus 4 plus 4 is 80.
- Version. Signals which block validation rules the miner claims to be using. Soft-fork coordination has also used bits inside this field. It is not a software marketing version.
- Previous block hash. The double SHA-256 hash of the parent header. This is the link in the chain. A header that names an unknown parent is not yet something a node can connect.
- Merkle root. A single hash that commits to the ordered list of transactions, including the coinbase. Change a transaction, or change its position, and this root changes. The white paper describes the tree. Light clients later use it to check that a transaction sits in a block without downloading the whole block.
- Time. A Unix timestamp chosen by the miner. Consensus limits how far it may sit from the median of earlier blocks and from the node’s own clock. It is evidence of ordering, not a certified clock.
- nBits. A compact encoding of the target. The header hash must be less than or equal to that target. This is the difficulty commitment for this block.
- Nonce. A 4-byte number the miner changes while searching for a header hash that meets the target. Extra search space also lives in the coinbase, because changing the coinbase changes the merkle root.
What the header hash is
The block’s identity is the double SHA-256 of those 80 bytes. Nodes do not ask a miner to sign the header with a company key. They check that the hash is numerically under the target encoded in nBits, that the parent exists, that the merkle root matches the transactions, and that the rest of the consensus rules hold for the block body.
Proof of work is therefore a statement about this header, not a statement about the miner’s legal identity. Anyone can assemble a header. The expensive part is finding bytes whose hash falls under the target. The cheap part, for everyone else, is hashing the header once and comparing.
What you still have to read in the body
The header does not list amounts. To know who was paid, you read the transactions, starting with the coinbase, which creates the subsidy and collects fees. The header’s merkle root tells you whether the list you were given is the list the miner committed to. It does not, by itself, tell you the list is valid. A full node checks the scripts, the amounts, and the consensus rules. A header-only view cannot.
Two practical mistakes follow from skipping that distinction. One is treating a header timestamp as the moment a payment became final. The other is treating a low header hash as proof that every transaction inside was honest. The hash proves work on that header. Validity of the payments is a separate check.
A short reading order
Start with the previous-block field and see which chain you are on. Check nBits against the difficulty you expect at that height. Hash the header and compare it with the target. Then, if the question is about a payment, demand the transaction and a merkle proof that connects it to the root, or the full block if you are checking the rules yourself. Eighty bytes are enough to name the block. They are not enough to finish the audit.
Display order is part of the reading
Block hashes are numbers. Explorers print them so the leading zeros, the visible sign of a low hash, sit on the left. The bytes inside the serialized header are not in that same left-to-right order. Bitcoin stores the previous-block hash and the merkle root in internal byte order, which is the reverse of the order most websites print. Copy a displayed hash straight into a hex editor and the header you build will not hash to the block you meant. The developer reference is the authority for the serialization. The habit is simply not to mix the two writings.
The same caution applies to nBits. It is a compact form of the target, not the difficulty number a chart shows. Difficulty is a rescaling of that target against the easiest target the protocol defined. A chart can say difficulty rose. The header says, in nBits, what target this block actually had to beat. If those two stories disagree, recompute from nBits before trusting the chart.
A valid header can still lose
A header that hashes under the target, names a real parent, and carries a merkle root is not yet “the” Bitcoin block at that height. Another header can do all of that and win, because nodes follow the valid chain with the most accumulated work. Two miners can find valid headers a few seconds apart. One of them becomes a stale block. Its transactions are not destroyed as ideas. They are simply not in the chain, unless a later block includes them. A payment that appeared in the losing header needs a later inclusion before a full node will treat it as confirmed.
That is also why a timestamp is a weak thing to hang a dispute on. The time field has consensus bounds, but it is chosen by the miner inside those bounds. It can be an hour off the clock on the wall and still be acceptable. Order comes from the parent pointer and from the work built on top, not from the header’s claim about what a kitchen clock said.
The genesis block is the one header every implementation already knows, which makes it a useful reference point rather than a mystery. Its previous-hash field is zero because it has no parent. Everything after it is a header that points at a parent and a transaction list. If you can name those six fields and say which question each one answers, you can read a block header. If the question is whether a payment was valid, you are no longer reading the header. You are reading the block.
One last distinction keeps the header from being asked to do the merkle tree’s job. The root is a commitment, not a list. A light client that wants a single payment still needs a path of sibling hashes from that payment up to the root. The header tells you the root the miner announced. The path tells you a transaction is under that root. The block body tells you the whole list. Reading them in that order is how an 80-byte object stays useful without being mistaken for the block it names.
Sources
- Bitcoin Developer Reference: block chain, including the serialized header.
- Bitcoin white paper, sections on proof of work, the Merkle tree, and simplified payment verification.
- NIST FIPS 180-4, the Secure Hash Standard that specifies SHA-256.




