Floating gallows

[nablator]

nablator, do you make anything of the mildly astounding statistic that almost exactly half of all tokens contain a gallows, and half don't?

Yes, but my interpretation will be in another thread.

A teaser, to answer your question:
Since k/t gallows seem to work like some optional binary marks (I'm ignoring p/f for the moment because they are so unevenly distributed, unlike k/t) there is a possible good reason for marking about half of the words: it is to signal de-synchronization from a binary state. Imagine that each vord needs a binary state information to make sense (one ciphertext for two cleartexts). There is a 50% probability that the correct state can be deduced from the context with whatever rule is implemented, for example (not the rule that I will present in the other thread): keep the same state as the previous vord, or (better) alternate to make reduplication possible without reduplication in the cleartext.

One type of mark would be enough to re-synchronize the stream (this role was actually held by double space) but two are easier to manage, giving the correct state information, thus the evolution and insertion of k/t as disambiguation marks.

The system initially required double spaces, that were unpractical on line breaks, and could be lost or added to justify text on the left and right. They were mostly replaced and made obsolete by k/t gallows, that also made the text much more language-like. I hope to show evidence in the other thread that there are still remainders of this evolution in the artificial-looking uneven spacing.