A key to understand the VMS

[Torsten]

No, adjacency rules are not preserved from the seeding part they are a result of the autocopying process. If a word is copied and some glyphs are replaced with similar ones the order of the glyphs stays unchanged. 

Well... if your starting point is akin to that of a verbose cipher (e.g. where al ol ar or qo are all irreducible tokens, as per Curse 2006 etc), you're starting with (what to nearly everybody else's eyes looks like) a set of adjacency rules within those irreducible tokens.

And then you run into problems parsing a and o, because they appear both inside and outside your core set of irreducible tokens. What is someone to make of ok, ot, of, op?

I didn't say that they are irreducible tokens. I only say that 'al', 'ol', 'ar' and 'or' can replace each other. But this doesn't mean that they are irreducible. After the network graph of similar words for the VMS it is possible to write 'chal' without 'l'. But since most times 'a' is written as 'y' at the end of a word the result is most likely a 'chy':


cthaiin ( 13) ckhaiin ( 3) chaiin ( 45) chdaiin ( 16) chodaiin (44)
cthain  (  4) ckhain  ( 1) chain  ( 18) chdain  (  9) chodain  ( 9)
cthan   (  2) ckhan   (--) chan   ( 11) chdan   (---) chodan   (--)
cthair  (---) ckhair  (--) chair  (  1) chdair  (  2) chodair  ( 2)
cthar   ( 20) ckhar   ( 3) char   ( 72) chdar   ( 20) chodar   (14)
cthal   (  7) ckhal   ( 4) chal   ( 48) chdal   ( 19) chodal   ( 7)
ctham   (  1) ckham   ( 3) cham   ( 20) chdam   ( 10) chodam   ( 1)
cthos   (  1) ckhos   ( 3) chos   ( 38) chdos   (  1) chodos   (--)
cthor   ( 45) ckhor   ( 9) chor   (219) chdor   (  8) chodor   (--)
cthol   ( 60) ckhol   (22) chol   (396) chdol   (  2) chodol   ( 2)
cthy    (111) ckhy    (39) chy    (155) chdy    (150) chody    (94)
cthey   ( 50) ckhey   (32) chey   (311) chedy   (501) cheody   (89)
ctheey  ( 13) ckheey  (11) cheey  (174) cheedy  ( 59) cheeody  (12)

And - even more problematically - how about aiin and aiir family words? Shouldn't they all be irreducible tokens too?

No. The network of similar words for the VMS shows that 'aiin' and 'aiir' are just intermediate forms between 'daiin' and 'ar':
aiiin( 41) daiin  ( 17) odaiiin( 4) okaiiin(  4)
aiin (469) daiin  (863) odaiin (60) okaiin (212)
ain  ( 89) dain   (211) odain  (18) okain  (144)
an   (  7) dan    ( 20) odan   ( 2) okan   (  5)
aiir ( 23) daiir  ( 23) odaiir ( 2) okaiir (  6)
air  ( 74) dair   (106) odair  ( 5) okair  ( 22)
ar   (350) dar    (318) odar   (24) okar   (129)
ail  (  5) dail   (  2) odail  (--) okail  (  1)
al   (260) dal    (253) odal   (13) okal   (138)
am   ( 88) dam    ( 98) odam   ( 6) okam   ( 26)
os   ( 29) dos    (  1) odos   (--) okos   (  8)
or   (363) dor    ( 73) odor   ( 8) okor   ( 34)
ol   (537) dol    (117) odol   ( 2) okol   ( 82)
y    (151) dy     (270) ody    (46) oky    (102)

For me, the answer seems more likely to be yes than no: but having lots of carefully chosen irreducible token blocks formed of what look like sets of letters is something that is as close to verbose cipher as it is far away from everything else, linguistically.

Which would seem to lead you to the position that what you're describing is autocopying where the seeding text rules are indistinguishable from verbose cipher.

Which would rather reduce the need for autocopying at all, surely?

No, I didn't see any other way how it was possible to someone in the 15th century to write a text with a network of 6837 similar word types out 8026 word types [see You are not allowed to view links. Register or Login to view.].