Transcript Ellen`s slides on Goldberg, Chapter 4

Casenhiser and Goldberg (2005)

Ability to learn to pair novel constructional meaning with
novel form

Known nouns and nonsense verb arranged in nonEnglish word order

Presented present tense and then past tense



NP1, NP2, nonsense V
The spot the king moopos; The spot the king moopoed
Video showing a spot appearing on the king’s nose

51 kids, 5-7 years old

Training: presented 5 novel verbs and 16 examples (3
minutes)

BF: 5 verbs w/ low frequency (44422)

HF: 5 verbs w/ high frequency (82222)

Control: watched film with no sound

Forced choice comprehension


Two video clips shown simultaneously; subjects point to the
one described by the sentence
Six with novel pattern and verb, six with transitive pattern
and verb
C and G Results

Figure 4.1

Control: no better than chance

Balanced: significant improvement over control

Skewed (HF): significant improvement over balanced

Children can get novel abstract meaning from a novel
pattern with novel verbs; they can extend that to new
utterances with new novel verbs

Implicit learning

“…high token frequency of a single general exemplar
does indeed facilitate the acquisition of constructional
meaning” (p. 82)
Other Studies

Goldberg, Casenhiser, and Sethuraman (2004) had
found similar results with adults

Kidd, Lieven, and Tomasello (2005)

4-year olds with complement-taking verbs
 I say her give the present to her mom.
 Children changed the main verb 25% of the time
 70% of substitutions involved think
 Evidence that kids’ generalizations about construction
involve verbs that are frequent in those constructions
Fast Mapping and UG

Could this fast mapping be evidence that they’re innate?
Morphology and Word Order

Could subjects have been paying attention to the -o
suffix on the novel verbs?


Do children recognize the novel word order?


Use non -o verbs
Two scenes: appearance (SVO) and transitive (SOV)
Results


Children learned the novel construction without the
morphological cue (Fig. 4.3)
Children matched word orders with appropriate scenes
(Fig. 4.4)
Skewed Frequency in Non-Linguistic
Categorization

Strong correlation between frequency of token and the
likelihood it will be considered a prototype


Facilitates category learning
Less variability / distortion = faster category learning
Elio and Anderson (1984)

“Centered” condition – frequent, prototypical instances

“Representative” condition – fully representative samples

“The superiority of the centered condition over the
representative condition suggests that an initial, lowvariance sample of the most frequently occurring
members may allow the learner to get a ‘fix’ on what will
account for most of the category members.”
Gentner, Loewenstein, and Hung
(2002)

Martians and blicks
1

2
Those who get high similarity tasks first do better with
low similarity tasks later on
Goldberg and Casenhiser
(forthcoming)

High frequency and dot patterns

24 college undergraduates tested to see if they could
determine new variations in dot patterns over the
frequently occurring pattern

Skewed frequency group performed better than the
balanced frequency group

Figure 4.5
What’s the point?

Frequency and early use of one verb pattern should
facilitate the learning of the semantics of that pattern
1) She put a finger on that.
2) He done boots on. (28 months)
X causes Y to move Zloc is associated with
Subj V Obj Oblpath/loc

Other constructions center around nouns, adjectives,
complementizers, etc.

Double is construction with thing
Cognitive Anchoring

High-frequency type of example acts as an anchor (a
standard for comparison)

Number anchoring in cognitive psychology

Anchoring effects are stronger when the anchor is
perceived to be relevant to the task
High Frequency Tokens

Are they necessary? NO! Subjects in the balanced
condition performed better than those in the control
condition (also – natural language learning)

Do high frequency morphological tokens lead to
generalizations? Bybee (1995) says no – they become
routines that are not analyzed and can’t be extended
(went, am)

VP idioms (kick the bucket) are analyzed
Conservative Learning / Fast Mapping

Children stick with the forms they’ve heard with particular
verbs (Ch. 3)

Age-related?

Children vs. adults – experience with language

Learners may be simply making tentative generalizations
– after all, they’re just recognizing differences – there’s
no actual production involved in the experiments