MORPHOLOGY DEBATE

The question: Are regular and irregular verb forms handled by one system 
or by two different systems? Whcih of the two views provides the better 
explanation of children's overgeneralizations?

1) dualist approach
- posites separate modules for regular and irregular forms
- default rule is needed, statistical learning cannot solve the problem

evidence:
- RT studies: regular verbs faster
- PET studies: different areas light up for regular vs irregular verbs

developmental model:

stage 1: associative memory
                |
                | 
              input
stage2:  associative memory     --> rules
                |                /
                |               / 
                |              /
stage3:       input           /
         associative memory    --> rules
                |                /
                |               /
                |              /
                      input 

2) single system approach

terminology:
- type frequency
- token frequency
- default: What one uses if all else fails
 
NOTE: The single system side's definition of default rule seems to be 
different from the dual system side's definition. For the dualists, the 
default rule seems to have greater significance and developmental 
priority. To be specific: The default is part of the innate equipment of 
each child, at least according to the dualist! It's not just something 
that you use if you run out of more interesting alternatives.

facts:

- irregular words are older
- only replaced by regular forms if not used frequently

- Overall, there is a small proportion of overgeneralization errors in 
  children's language (less than 5%)
- there is not only regularization, but also irregularization and blending

Furthermore, the single system side claims that:

- Irregulars are not at all irregular!
  Every part of language is highly structured, including the irregulars! 
  There is a lot of phonological and semantic regularity surrounding the 
  irregular forms.

[PICTURE: ATTRACTORS IN LEXICAL SPACE]

Question: 
How does overregularization develop?
How does the default-space in the attractor space develop?

Answer:
High type frequency of default forms is responsible for that.

Question:
Single system works fo production, but does it also work for comprehension?
How does the single system account for 
- U-shaped learning
- RT differences
- PET differences?

Answer: RT differences can be explained by attractor model.

The dual discussion group had serious doubts about this statement. It did 
not seem obvious why certain forms should be faster than other forms. 
Mike was particularly puzzled by the differences between plink and think, 
I think.

I guess we concluded that the empirical evidence about these RTs is still 
inconclusive.

Q: Has the dual side EVER tried to implement a model?
A: (obvious)

After this smart rhetorical move, the single system side made this 
puzzling statement: 

"Nature favors simplicity and economy."

On a personal note: No, I don't think that's simply a version of Ockham's
razor. Ockham's razor simply states that - if one has a choice - one
should always prefer the simpler theory (with equal explanatory power).
That's a rule for the selection of theories, and that's not equivalent to
a statement about THE NATURE of nature.

Next, the dual side gave a very surprising (namely, non-dualistic)
definition of rule.  They argued that rules can be realized in a system in
an explicit or an implicit way. Somehow, the dual discussion group did not
seem to feel very comfortable with a strictly dualistic view of things
(Who would?). But one should not forget - as Doug and Mike pointed out -
that Pinker & Co. are not joking when they talk about rules in the head.
They mean RULES, hard-wired and innate. 

At this point, the discussion shifted to children's metalinguistic
abilities. Does the fact that children are able to talk ABOUT the past
tense substantiate the claim that there is an explicit rule system in the
brain? Somehow, it seems to be easier to abstract explicit knowledge from
an explicit rule system. Eliana disagreed, I don't know what Linda thinks
(she brought it up), but I guess we cannot neglect this kind of evidence
entirely. 

To beat a dead horse: Pinker and Co. do not believe in development. Rules 
are there from the beginning, all that is learned are the exceptions.

So, is there an innate, eternal brain area for regular English past tense?

Doug reminded us that there are lots and lots of brain function (like
vision) which reliably appear, across people, in the same area. The place
where the vision system is does not vary greatly across people. Does that
mean there is an innate program for the architecture and location of the
vision system in the brain?

I leave it up to you guys to make your own inferences about PET studies of
regular and irregular verbs and what they mean.