Rules vs. Lists

[Rob Freeman]

Right Jouni. It is somewhat laborious to factor them out, but I think
we can find lots of rules once we start to look for them, even if we
restrict ourselves to one kind: morphological, phonological etc.

Whether a vowel-consonant inventory would count as this kind of system
is another question. Phonemes are not really examples. They are
classes. The right question would be to compare the number of phonemes
and phonotactic rules with the number of utterances. In this context I
refer you to suggestions such as Syd Lamb's that we need to relax the
"linearity requirement" for phonemes in combination. An issue which
goes right back to the core of the dispute between structural and
functional schools in linguistics.

As to why it should be so, why we should want to have more rules and
examples, it is perhaps not immediately obvious. But actually such a
system gives us lots of power. Power we have simply been throwing away
because we have assumed fewer rules than examples. If we can
constantly draw new rules out of the examples we can use all those
extra rules to parametrize our system. All we need to do is look for
them.

-Rob

On Sat, Jul 5, 2008 at 10:29 PM, Jouni Maho <vch468d at tninet.se> wrote:
> May a lurker butt in with a little thought experiment?
>
> Re Rob Freeman's:
>>
>> Can you show me a system where there are more rules implicit in the
>> examples than there are examples themselves, and explain why it must
>> be so?
>
> Assume the following (complete) lexicon of a hypothetical language:
>
> berama
> bilama
> butaba
> metama
> tilaba
>
> And the rules:
>
> C > b r m l t
> V > e a u i
> Root > CVCV
> V2 > a
> R+ba > agent
> R+ma > causative
>
> That is, 5 list items, 6 rules. This assumes, of course, that the types of
> rules can be of any "kind", i.e. morphological, phonological, etc. Or does
> the question suppose that there should be restrictions on type of rules
> (only morphological, only phonological, etc.)?
>
> I'm not sure how easy this would be if the lexicon's size was considerably
> larger, but at least it's possible to devise a less-items-more-rules system
> as a thought experiment. I have no idea why it should be so, but it's
> certainly pissoble.
>
> By the way, would a vowel-consonant inventory (list) with it's accompanying
> rules (phonotax, assimilation, etc.) count as a valid less-items-more-rules
> system?
>
> ---
> jouni maho