Significant predictors of discrimination accuracy: Session number (p < .000),position in word (p = .001), identity of contrast (p < .000)

Significant post-hoc comparisons:• Velar-coronal contrast less accurate than fricative-glide or coronal-labial contrast.• Initial position more accurate than final position.

In terms of identity of phonemic contrast, B's perceptual bias parallels his errors inproduction.

• Consistent with hypothesis that markedness applies over both domains.

In terms of position, B's perceptual bias is the reverse of his bias in production.• Capturing this divergence in positional bias will require modifications to Pater’s

FAITH(LS) and FAITH(SL) model.

The perceptual advantage for initial over final contrasts reflects an intrinsic bias ofthe auditory system (Fujimura, Macchi, & Streeter, 1978).

But neutralization can be driven by articulatory as well as perceptual factors.Articulation has its own positional bias: Neutralization tends to apply preferentially ininitial position, where gestures are more forceful (Inkelas & Rose, 2008).

To encode the possibility that positional bias in perception may diverge frompositional bias in production, I propose a modified division of faithfulnessconstraints: PERCEPTUAL-FAITH versus GESTURAL-FAITH.

PERCEPTUAL-FAITH compares any perceived surface form vs the lexical representation.• Perceived form could be child’s own production or that of another speaker.• When P-FAITH has low weight, markedness can induce perceptual simplification.• P-FAITH has a positional component: Magnitude of violation is greater in contexts

where neutralization is more perceptually salient.

GESTURAL-FAITH penalizes any articulatory gesture not specified by the child’s LR.• G-FAITH constraints do not apply to perceptual mapping.• G-FAITH constraints are not inherently positional. Articulation-driven

asymmetries are encoded directly in the magnitude of the markedness violation.

The PERCEPTUAL-FAITH / GESTURAL-FAITH distinction will make it possible to model B’sdivergent patterns in perception and production.

A Harmonic Grammar framework will be used.

Production mapping

Fronting is driven by a high-weighted constraint, *K.• *K is a simplified cover for an articulatory-driven constraint detailed

elsewhere (McAllister, 2009).• Magnitude of violation of *K is greater in initial position.

Table 1: Violation of *K outweighs both G-FAITH and P-FAITH.Table 2: Smaller *K violation in final position is less than combined violation of P- and G-FAITH. Velar target emerges faithfully.

Table 1. An initial velar is fronted.

Table 2. A final velar is realized faithfully.

Perception mapping

Recall that G-FAITH does not apply. In Tables 3-4, *K violation trumpsfaithfulness in initial and final position, eliminating the asymmetry.

Table 3. An initial velar is perceived to be fronted.

Table 4. A final velar is perceived to be fronted.

The general perceptual advantage for initial over final contrasts willbecome apparent when contrasts not associated with a high-weightedmarkedness violation are also included.

Dividing faithfulness into perceptual and articulatory componentsmakes it possible to capture speech patterns in cases where theauditory and articulatory systems exert conflicting pressures.

In this system, how will the child learner arrive at a correct LR?• Stochastic variation in constraint weight will allow child to perceive

the correct target some percentage of the time. With sufficientexposure, the grammar will converge on an adultlike LR.

• Using nonwords presumably influenced extent of neutralization.

Convergence and Divergence of Perception and Production in Phonological AcquisitionTara McAllister

How does perception interact with production in phonological acquisition?

Case Study Subject

Analysis: Positional Velar Fronting

Allen, G. D., & Hawkins, S. (1978). The development of phonological rhythm. In A. Bell & J. B. Hooper (Eds.), Syllables and segments (pp. 173-185). Amsterdam: North Holland.Dinnsen, D. A., & Farris-Trimble, A. W. (2008). The prominence paradox. In D. A. Dinnsen & J. A. Gierut (Eds.), Optimality Theory, Phonological Acquisition and Disorders (pp. 277-308). London: Equinox Publishing Ltd.Fujimura, O., Macchi, M. J., and Streeter, L. (1978). Perception of stop consonants with conflicting transitional cues: A cross-linguistic study. Language and Speech, 21, 337-346.Inkelas, S., & Rose, Y. (2008). Positional Neutralization: A Case Study from Child Language. Language, 83, 707-736.Jusczyk, P., Houston, D., & Newsome, M. (1999). The Beginnings of Word Segmentation in English-Learning Infants. Cognitive Psychology, 39, 159-207.McAllister, T. (2009). The Articulatory Basis of Positional Asymmetries in Phonological Acquisition. Unpublished doctoral dissertation, Massachusetts Institute of Technology.Pater, J. (1999). The comprehension/production dilemma and the development of receptive competence. Ms., University of Alberta.Pater, J. (2004) Bridging the gap between receptive and productive development with minimally violable constraints. In R. Kager, J. Pater & W. Zonneveld (Eds.), Constraints in phonological acquisition (pp. 219-244). Cambridge: Cambridge University Press.

References

Pater (1999, 2004): Models of child phonology should accommodatethe finding of parallel restrictions across perception and production.

Example: Infants learn to detect words with trochaic foot structurebefore iambic words (Jusczyk, Houston, & Newsome, 1999). Thisparallels truncation in production, which affects the weak syllable inan iambic but not a trochaic foot (e.g. Allen & Hawkins, 1978).

Given these parallels, Pater proposes that both perception andproduction are governed by the same set of markedness constraints.

However, perception and production abilities unfold on distinct timeframes (8-9 months versus 18-24 months in the previous example).

To capture this fact, Pater posits two sets of faithfulness constraints:• FAITH(LS) constrains the mapping from lexical to surface forms.• FAITH(SL) constrains the mapping from perceived to lexical forms.

I present case study data documenting parallel constraints onperception and production in a child with phonological delay.New observation: Perception and production errors can divergein positional bias.

Subject B, a 4-year-old boy acquiring American English.Significant phonological delay/disorder.

Of note, B’s output featured positional processes that preferentiallyneutralize contrast in initial but not final contexts.

• Positional velar fronting:/dajæk/, “kayak” /teɪk/, “cake”

• Positional fricative gliding:/jak/, “sock” /bas/, “bus”

Question: Can the same positional bias be observed in perception?(Dinnsen & Farris-Trimble, 2008)

Nonword discrimination task: B responded "same" or "different" tophonetically controlled nonword pairs.

23 identical pairs:• “I can say woop. I can say woop.”24 pairs differing by a single phoneme in initial or final position:• “I can say tuv. I can say kuv.” / “I can say vud. I can say vug.”

Three contrasts tested:• Coronal versus velar. Actively neutralized in B's output at testing.• Fricative versus glide. Initial position only. Ceased to undergo

neutralization in B’s output roughly one month prior to testing.• Coronal versus labial. Never neutralized in B’s output.

Results were analyzed with a 5-factor logistic model.DV: Accuracy in detecting contrast when contrast was present.IV: Session number; Position in word; Identity of contrast; Voicingof contrast; Possible harmony context.

Results

Methods

Proposal: PERCEPTUAL-FAITH versus GESTURAL-FAITH

Discussion

THE 18TH MANCHESTER PHONOLOGY MEETING, MAY 2010

Bars show standard error.