Where the Rules Go Quiet
Contraction rules are built around common letter sequences in ordinary language. The word the earns a single cell; and, for, with follow suit. But the system's designers were always working with a statistical target: high-frequency patterns in continuous prose. Move away from that target and the rules either produce ambiguity or simply offer nothing.
Proper nouns expose this most plainly. A reader encountering the whole-word contraction for be inside the name Belize has a problem the code anticipates: contractions are applied to names much as to other words, but they are avoided wherever they would straddle a syllable boundary or obscure the pronunciation, and in those places the name is spelled out letter by letter. This is not an exception bolted on later — it is a deliberate design boundary. Names are low-frequency, non-predictable, and ambiguity there is costly in a way it rarely is in common vocabulary.
Technical and scientific vocabulary raises the same pressure from a different direction. A contraction that compresses con works tidily in connect or concert, but applied inside a chemical name or a trademarked product designation it can destroy the reader's ability to reconstruct the original string. Unified English Braille encodes explicit rules about where contraction is suspended — initialisms, abbreviations, and strings that mix letter cases signal a switch to a more literal rendering.
Mixed-script content — mathematical expressions embedded in prose, for instance — forces a more complete suspension. When a passage enters Nemeth Code or any specialised notation mode, literary contractions do not carry across the boundary. The mode switch resets the rules entirely.
What the code does in all these cases is the same thing: it defaults to transparency. Compression is abandoned when the cost of misreading outweighs the cost of bulk. Braille books grow larger; reading slows slightly. But the text remains unambiguous — which is, in the end, what a writing system is for.