It estimates how well different parts of. Correct me, please, if i'm wrong: Essentially parallel is sufficient because mean differences (additive constants) do not matter for reliability (variance differences is what matters).
There are major difference between. By splitting a test into two halves and comparing the results, researchers can assess whether the test items are reliable in measuring a particular construct. It is also unsuitable for.
The methods in splithalf are built around split half reliability estimation. Split half reliability is very much similar to parallel form reliability in which items are divided into two sets and then administered to individuals.