It helps a lot with the how I am comparing part, although if anyone guessed in advance that was how By vowels: sorted, I'd like to solicit their advice on the outcome of upcoming NFL games. It's still not exactly what I had in mind for bringing all identical terms together. The

$a->{RECNO} <=>  $b->{RECNO}
[download]

is unnecessary if the sort is stable, as sort() is, by default. Since different words can compare equal under the influence of

$a->{DICTFOLD} cmp  $b->{DICTFOLD}
[download]

(if I'm correctly guessing what DICTFOLD is), I still might see

word
Word
word
Word
[download]

in the sorted output, when I might have preferred

word
word
Word
Word
[download]

which makes it easier to determine if words are "unique" without repeating all the complicated logic. So I would prefer

$a->{ORIGINAL} cmp $b->{ORIGINAL}
[download]

as the final tie-breaker. That's neither better nor worse than your code, merely different.