Suppose we have a 3Cons (a Tri-Consonantal-Root language) that is descended from a 2Cons protolang (a Bi-Consonantal-Root proto-language).
Here are three things that might happen.
1) Maybe: For any six consonants C1, C2, C3, C4, C5, and C6;
If any three of C1-C2-C5, C1-C2-C6, C3-C4-C5, or C3-C4-C6, are attested roots in the 'lang;
then (a) "with (overwhelmingly?) greater-than-chance frequency", the fourth one is also a root in the 'lang,
AND: (1b) "with (overwhelmingly?) greater-than-chance frequency", the semantic relationship between C1-C2-C5 and C1-C2-C6, is the same (or almost the same?) as the semantic relationship between C3-C4-C5 and C3-C4-C6.
2) Maybe: For any six consonants C1, C2, C3, C4, C5, and C6;
If any three of C1-C3-C4, C1-C5-C6, C2-C3-C4, or C2-C5-C6, are attested roots in the 'lang;
then (a) "with (overwhelmingly?) greater-than-chance frequency", the fourth one is also a root in the 'lang,
AND: (2b) "with (overwhelmingly?) greater-than-chance frequency", the semantic relationship between C1-C3-C4 and C2-C3-C4, is the same (or almost the same?) as the semantic relationship between C1-C5-C6 and C2-C5-C6.
3) Maybe: For any six consonants C1, C2, C3, C4, C5, and C6;
If any three of C1-C2-C3, C1-C4-C5, C2-C3-C6, or C4-C5-C6, are attested roots in the 'lang;
then (a) "with (overwhelmingly?) greater-than-chance frequency", the fourth one is also a root in the 'lang,
AND: (3b) "with (overwhelmingly?) greater-than-chance frequency", the semantic relationship between C1-C2-C3 and C2-C3-C6, is the same (or almost the same?) as the semantic relationship between C1-C4-C5 and C4-C5-C6.
Questions:
I. For each of those three statements, is there a RL 3Cons Natlang for which that statement is true? (And if so, what is one such natlang?) (this is really three questions.)
II. For each pair of those three statements, is there a RL 3Cons Natlang for which both of those statements are true? (And if so, what is one such natlang?) (this also is really three questions.) (Any positive answer to II would imply two positive answers to I. Any two positive answers to II would imply three positive answers to I.)
III. Is there a RL 3Cons Natlang for which all three of those statements are true? (And if so, what is one such natlang?) (A positive answer to III would imply three positive answers to II.)
Questions I, II, and III above, are short-answer questions. They may require an awful lot of research, but no theorizing.
Questions IV and V below are more essay-questions.
IV. If there is one of those statements which is not true of any RL 3Cons natlang, (A) Is there a reason why not? AND (B) Is that reason such that no conlang satisfying that statement would be naturalistic nand/nor realistic?
V. If there are two of those statements which are never both simultaneoussly true of any RL 3Cons natlang, (A) Is there a reason why not? AND (B) Is that reason such that no conlang simultaneously satisfying both of those statements would be naturalistic nand/nor realistic?
My guesses, without any research, are that:
I. 1 probably happens in a natlang; and,
I. 2 somewhat-less-probably-but-still probably also happens in a natlang; and,
I. if one of them doesn't happen in any natlang it is probably 3; and,
II. that it's by no means improbable that 1 and 2 do not both happen together in any natlang.
III. If those guesses are true, then the answer to III is "no".
My guesses about IV and V are, that EITHER all 3Cons roots in the daughterlang that are derived from the protolang are derived from a root + a suffix, OR ELSE all 3Cons roots in the daughterlang that are derived from the protolang are derived from a prefix + a root.
If 1 and 2 can happen together in a natlang, then I guess 3 can also happen in (either the same or some different) natlang.
Conversely if 3 can happen in some natlang, then I guess the combinations 1+2, and 1+3, and 2+3, can also happen in some natlang.
Is anyone aware of any already-existing conlang that would be relevant for this question?
Some of us are working on (or have worked on) 3Cons conlangs; some of those are working on (or have worked on) deriving that conlang from a 2Cons con-proto-lang.
If you are one of those, which of statements 1 and/or 2 and/or 3 do you plan to make true of your conlang?
1) Maybe: For any six consonants C1, C2, C3, C4, C5, and C6;
If any three of C1-C3-C5, C1-C4-C5, C2-C3-C6, or C2-C4-C6, are attested roots in the 'lang;
then (a) "with (overwhelmingly?) greater-than-chance frequency", the fourth one is also a root in the 'lang,
AND: (1b) "with (overwhelmingly?) greater-than-chance frequency", the semantic relationship between C1-C3-C5 and C1-C4-C5, is the same (or almost the same?) as the semantic relationship between C2-C3-C6 and C2-C4-C6.
I didn't because I thought that had to be a minority phenomenon, if it occurred at all; I think it could happen only if the 2Cons protolang were heavy on infixes.
So it's not impossible but less likely than the other three IMO (or, rather, IMG).
*1 has to do with suffixes in the protolang, and 2 has to do with prefixes in the protolang, and 3 has to do with both together. Clearly I could have made a 5th and 6th statement about suffixes-with-infixes and respectively prefixes-with-infixes in the protolang. I didn't.