On Fri, 1 Dec 2006 01:21:41 -0500, Steve Rice <[log in to unmask]> wrote:


>The design of Loglan in this regard was ingenious: it allowed a wordmaker
>to collapse several forms into one. Thus "hasfa" (house) combined elements
>of English, German, Spanish, and Chinese into one form. There were two
>major problems:

Ingenious, certainly, but by me the results, for your reasons below
and others, are disappointing.    For me, kangu will always mean
'kangaroo,' in analogy to your 'telfo'.

>
>1. Composite primitives were scored based on how much of a source word they
>contained. I think they should have been rated for how much of the C-Prim
>was accounted for by the source word. "Godzi" contains all of E "go"
>while "telfo" contains only most of "telephone"--but which is easier to
>learn? "Godzi" has the "-dzi" element for an anglophone to learn, but
>everything in "telfo" will be familiar.
>
>2. Phonemes followed Loglan rules, not source rules, so a source /j/ could
>show up in a C-Prim as [i]. "Cutri" got its "u" in part from "*w*ater," for
>example.
>
>But the major snag was the change in lexical philosophy. When Loglan came
>out, compounding was meant to be rare--the result of overwhelming Zipfean
>pressure to shorten a common phrase to a word. For predicates were supposed
>to be a string of modifiers before the main predicate, with all the
>syntactic relationships spelled out more clearly that is possible with a
>compound.
>
>But when compounds did arrive, their form and place-structure were
>unpredictable. This is why Loglan picked up djifoa--special abbreviated
>forms of predicates used only n compounding. It was a retrofit solution,
>and one of the things that I despised about Lojban was that it copied this
>feature instead of starting with a morphology designed for compounding.
>(Only guaspi had a proper design from this standpoint.)

Could you elaborate?  I assume the tone of the first element of a compound
does it, but I have a hell of a time understanding the material on guaspi.
I need an Intro to Guaspi by Steve Rice.

I hadn't actually realized that.  It certainly explains the attack of the
allomorphs, the Loglan thing that drives me craziest.

>
>My solution was to distinguish grammatical and lexical bases: the lexicals
>were compounded onto the grammar words, which formed the skeleton of the
>sentence. That way a sentence could be parsed easily by stripping out the
>lexicals (like Eo "o a as an on e"), but the lexicals gave the humanly
>necessary information. It would make it possible to predict both form and
>place-structure.
>
>I may take that up again someday.

I  hope you do.  How many grammaticals would be needed?  I had the notion
first of assigning the last vowel of C(CV/VC)CV's for a grammatical function,
but of course five wouldn't be enough.

I take it that your idea is that lexicals are automatically separated from
each other by grammaticals, and without a grammatical there, they
automatically form compounds, right?