Which shortforms can carry an ending?

A shortform is a whole word written with an abbreviated sequence of letters: about as ab, because as bec, o'clock as o'c. There are 73 of them, and the standard treatment adds that a shortform keeps its abbreviation when a common ending is bolted on — friends, goodness, letters.

It is a good rule and it is not quite true. Some shortforms take endings the list does not mention. Others take almost none. And the same three letters at the front of a word can keep the shortform in one word and lose it in the next, for reasons no ending list can express.

The method

The word list is a full dictionary, so it includes rare and technical words alongside common ones. That is deliberate: the question is what the rule does across English, not across the words someone thought to check.

The result

Three quarters keep it, which is why the rule is taught the way it is. The interesting quarter is the exception, and the exceptions are not spread evenly.

The largest group is the one no single list can handle. That is the finding: for 23 of the 51 shortforms, whether the abbreviation survives is a property of the whole word, not of the ending.

The shortforms the reference is inconsistent about

Lowest first — these are the ones where assuming the shortform carries over goes wrong most often.

Two pairs are worth sitting with.

must keeps its shortform in mustard and loses it in mustache. Neither word has anything to do with must; the reference treats one as a word that happens to start with those letters and the other as an exception. should keeps it in shouldest and loses it in shoulder — that one at least has an explanation a reader can feel, since shoulder would otherwise open with a sign meaning a different word.

And him keeps it exactly once out of twenty-two: in himself. Every other word in English beginning with those three letters is something like Himalaya.

What this site does with it

Rather than one list of endings for all 73 shortforms, this site's engine carries a measured list per shortform: 38 of them take at least one ending, and 35 take none at all. Across all of them the endings that ever appear number 14.

The direction of error decides how the lists are built. Writing a contraction the reference does not use is wrong braille and a reader is misled; spelling a word out where the reference contracts is longer and correct. So an ending goes on a shortform's list only where the reference actually uses it, and the test that reads this data checks one thing above all: that the false-positive count is zero.

If you are learning these

The practical version is shorter than the research. Use the shortform for the word itself, always. Use it with an ending only when you have seen that combination — and the page for each shortform now lists the endings measured safe for it, along with the English words that begin with the same letters and do not take it.

The contraction finder will show you what happens to any word you type, including which contractions were legal and passed over. All six write-ups are published with their method and their failures.