Measurement in quantum computing extracts classical information from a quantum state, collapsing it probabilistically according to the Born rule: measuring a qubit in state alpha|0> + beta|1> yields outcome 0 with probability |alpha|^2 and outcome 1 with probability |beta|^2, with the post-measurement state being the corresponding basis state. Measurements can be performed in any orthonormal basis by applying a change-of-basis gate before measuring in the computational basis. Understanding measurement is essential because it is the only way to extract results from a quantum computation, and its probabilistic, destructive nature is the central constraint that quantum algorithm design must navigate.
From your study of the Born rule in quantum mechanics, you know that measurement outcomes are probabilistic and that measurement disturbs the system. In quantum computing, these facts become engineering constraints. A quantum computer performs a unitary computation on qubits, then measures some or all of them to extract a classical answer. The Born rule dictates the probability of each answer, and the post-measurement state is the projected (collapsed) state. The entire challenge of quantum algorithm design is arranging the unitary computation so that the desired answer has high measurement probability.
Projective measurement in the computational basis is the standard operation. For a single qubit in state alpha|0> + beta|1>, the measurement yields 0 with probability |alpha|^2 and 1 with probability |beta|^2. Afterward, the qubit is in the corresponding basis state — the superposition is gone. For multi-qubit systems, measuring one qubit collapses it and updates the remaining qubits' joint state accordingly. If two qubits are in the Bell state (|00> + |11>)/sqrt(2) and you measure the first qubit, getting 0 collapses the joint state to |00> and getting 1 collapses it to |11> — the second qubit's state is now determined.
You are not restricted to measuring in the computational basis. To measure in the X basis ({|+>, |->}), apply a Hadamard gate first, then measure in the computational basis. To measure in an arbitrary basis, apply the appropriate unitary rotation first. This is equivalent to measuring with projection operators onto the desired basis states. The choice of measurement basis is a powerful tool: the states |+> and |-> are indistinguishable in a Z-basis measurement (both give 50/50) but perfectly distinguishable in an X-basis measurement. Many quantum protocols, including BB84 key distribution, exploit exactly this basis-dependent distinguishability.
A fundamental limitation is that measurement provides at most one classical bit per qubit, and it is destructive. You cannot "peek" at a quantum state without disturbing it, and a single copy of an unknown state cannot be fully characterized. This connects to the no-cloning theorem: if you could copy an unknown quantum state, you could make many copies and measure each in a different basis to reconstruct the state, but cloning is forbidden. Quantum algorithms must therefore be designed to concentrate the answer into a high-probability measurement outcome, often by exploiting interference across many computational paths. The probabilistic nature of measurement also means many quantum algorithms are inherently probabilistic — they succeed with high probability but may need to be repeated a few times to boost confidence.