Editing Openai/69186d20-2d14-8013-b9c6-c09000f173f6 (section)

=== Assistant: 2511131754ET ===
251113_1754ET

short version: for a single consonant, Irish tries hard to avoid true “mixed nearest vowels”. Once you (a) group vowel digraphs and (b) respect syllable/morpheme boundaries, you almost always get a clear broad or slender answer. When you really do have broad on one side and slender on the other at the same distance, the winner is basically “whichever side the consonant belongs to” (syllable/morphology), i.e., you just have to know that word.

Let’s make this concrete.

==== For each consonant, do this: ====
# Group vowels into known digraphs: - ae, ao, eo, ia, io, oi, ui, ua, ea, ai (etc.) - These count as one vowel unit for “how many vowel sounds per syllable”. - But each component letter still has a class: - broad: a/o/u - slender: e/i
# For that consonant, look for the closest vowel letter (or digraph component) on each side within its syllable/morpheme. - Left side: walk left until you hit a vowel or a boundary. - Right side: walk right until you hit a vowel or a boundary.
# If nearest on one side is clearly closer, use that side to decide: - nearest left is one letter away (a), nearest right is three letters away (i) → broad.
# If both nearest vowels are equally close but same class (both broad or both slender) → easy. - consonant is that class.
# If both nearest vowels are equally close but different classes (true “mixed case”): - Ask: which side does the consonant really belong to syllable-wise / morpheme-wise? - That side wins. In practice this shows up in relatively few words and is often lexicalized: you just learn that consonant as broad or slender.

The key: once you respect digraph grouping and syllable/morpheme boundaries, almost everything beginners see stops being ambiguous.

==== ### ====

c a i f e → /ˈkafʲə/
* c: - nearest vowel on right: a (immediately next) → broad. - (i is further away and really grouped as [ai] with the vowel of the syllable.)
* f: - nearest left vowel: i (one step left) – slender - nearest right vowel: e (one step right) – slender → no conflict; f is slender.

So there’s no true mixed case for any single consonant here.

===== b a i n n e → /ˈbˠanʲə/ =====
* b: nearest vowel left/right is a → broad.
* n (first): nearest vowels are i left (slender), e right (slender) → slender.
* n (second): nearest left i or e (depending how you syllabify bain-ne) → still slender.

Again, no “one side broad, one side slender” for the same consonant.

===== m a i t h → /mˠah/ =====
* m: nearest vowel is a → broad.
* t (underlying, before lenition): nearest left vowel i and right none (th is one unit, no vowel after) → slender.

Still no real mixed.

===== t a e → /t̪ˠeː/ =====
* Digraph ae spells /eː/.
* For t, nearest vowel letter on the right is a (broad) – even though the sound is [eː].
* So orthographically t is broad, because its “class vowel” is a.

This is the kind of thing that looks like “mixed” at the level of phonetics, but at the level of spelling it’s coherent:
* the a in ae “belongs” to the left side and is broad;
* the e “belongs” to the right side and is slender;
* but for t we only care about the a side.

==== You only really get a conflict when: ====
* a consonant is exactly between broad on one side and slender on the other,
* both are equally close,
* and no digraph grouping or syllabification cleans it up.

In those cases, native speakers are basically relying on:
* underlying morphology (e.g. “this is the slender stem form”), or
* lexical memory (you just know “that n is slender”).

For learners, the practical rule is:

: 

==== When you look at a consonant and see “mixed” around it: ====
# First try to parse digraphs: is that ae/ao/ea/ia/io/oi/ui/ua/ai hiding there?
# Then syllabify (one vowel nucleus per syllable).
# Then assign each consonant to a syllable and ask: which side’s vowel is actually “its”?
# Only if all of that still leaves a tie do you treat it as a memorization / morphology case.

In practice, if you stick to that, most “what is this vowel doing?” puzzles in Irish resolve cleanly, and you only need to brute-memorize a modest residue of exceptions (especially loans and some irregular forms).