Frequency effects in lenition and the challenge of lexicalized markedness

1

Frequency effects in lenitionand the challenge of

lexicalized markedness

James MyersNational Chung Cheng University

[email protected]

Workshop on Variation, Gradience and Frequency in PhonologyStanford University, July 2007

2

Acknowledgments

• Li Yingshing– For coauthoring the phonetic study

• National Science Council (Taiwan)– For paying the bills

• People like you

3

The argument

• Lenition is markedness reduction in the raw

• Yet lenition is lexicalized– Attempts to escape this conclusion don’t work

• Lexicalized markedness cannot be formalized in grammar insightfully*– It’s essentially a peripheral processing issue

• Therefore, markedness isn’t “inside” grammar– Hale & Reiss (2000), Boersma (2005)

*Rhetorical convenience doesn’t count as insight.

4

Frequency effects in lenition

• The more common a word or phrase, the more phonetically reduced it is in production

• Studied by (among many others):Aylett & Turk (2004), Berkenfield (2001), Bybee (2000ab, 2002), Cacoullos & Ferreira (2000), Cohn et al. (2005), Fidelholtz (1975), Hammond (1999, 2004), Hay (2000), Hooper (1976), Johnson (1983), Jurafsky et al. (2001, 2002), Kawamoto et al. (1999), Lavoie (2002), Li (2005), Munson & Solomon (2004), Myers & Guy (1997), Myers (1999), Myers & Li (2005), Patterson & Connie (2001), Phillips (1984, 1999), Pierrehumbert (2001), Pluymaekers et al. (2005), Tseng (1999), van Bergem (1995), van Son et al. (2004), Wright (1979)

5

Examples• English vowel reduction (Hooper 1976)

• Syllable contraction in Southern Min (Li 2005)

6

What causes this phenomenon?

• Speaker-oriented explanations– Articulatory targets become more automatized thro

ugh use (e.g. Bybee 2001, Pierrehumbert 2001, 2002)

• Listener-oriented explanations– Frequent words are more predictable, so speakers c

an afford to be less clear (e.g. Jurafsky et al. 2001)

• What will resolve this crucial debate?– Phonetic and psycholinguistic experimentation– Theoretical phonology can only play catch-up

7

Why phonologists worry anyway

• Phonetics is sensitive to lexical frequency– So phonetics isn’t really “post-lexical”?

• Gradient reduction is word-specific– So lexical representations aren’t categorical?

• Frequency-reduction correlation is universal– So lexical effects aren’t always idiosyncratic?

• Yet lenition begets “real” phonology– Deletion, stress shift, assimilation...

8

Escape hatch #1: It’s not lexical?

• Maybe it’s just an indirect effect– Frequent words are also more predictable in dis

course context (e.g. Jurafsky et al. 2001)

– Frequency eases lexical access, facilitating articulatory fluency (e.g. Pluymaekers et al. 2005)

• How to test this– Factor out contextual predictability, ease of acc

ess, speaking rate– Does frequency still affect lenition?

9

Escape hatch #2: It’s not gradient?

• Maybe it’s just stochastic “ordinary” phonology– Maybe frequency just increases the probability of ch

oosing lenited over full allomorphs, but both are categorical (cf. Pluymaekers et al. 2005)

– Variant: Probability of choosing prosodic frame

• How to test this– Use a continuous dependent measure (not allomorph

probability, as in many studies)– Control prosodic structure

10

Case study:Southern Min syllable contraction

• Phonologically regular– “Edge-in” preservation of segments– Tonal contours are merged– Output often respects sonority profile– Vowels of higher sonority are often favored

• Has been formalized with autosegmental notation and/or Optimality Theory– Chung (1996, 1997), Hsiao (1999, 2002), Hsu

(2003)

11

Measuring syllable contraction• Twenty native speakers of Southern Min• Shadowing task

– 120 items from spoken Southern Min corpus (Myers & Tsay 2003)

– Hear uncontracted forms, must repeat back naturally (not told explicitly to contract)

– Isolated items in random order (no contextual predictablity)

• Dependent measure of contraction– Trough depth

Myers & Li (submitted)

12Analysis using Praat (Boersma & Weenink 2007)

Trough depth

• Maximum depth of amplitude contour (syllable boundary detection algorithm of Mermelstein 1975)

13

Predicting syllable contraction

• Lexical frequency from corpus (log-normed)• Phonetic confounds

– Segment types– Duration– Maximum intensity

• Higher-level confounds– Reaction time (ease of lexical access)– Lexical category (whether or not word/phrase co

ntains a function morpheme)• Tests for the influence of prosodic structure

14

Sample stimuli

15

Regression analysis

factoring out phonetic influences

frequency

factoring out prosody

[linear mixed-effect modeling (e.g., Baayen forthcoming)]

factoring out lexical access

16

Frequency predicts lenition degree

0.0 0.5 1.0 1.5 2.0 2.5 3.0

010

20

30

40

Log Corpus Frequency

Tro

ugh d

epth

(dB

)

17

But is it gradient…?

0.0 0.5 1.0 1.5 2.0 2.5 3.0

010

20

30

40

Log Corpus Frequency

Tro

ugh d

epth

(dB

)

Linear fitLoess fit

frequency estimates less reliable down here…?

accidental run of obstruent onsets…

>

18

More evidence for gradience

• Categorical allomorphs predict bimodality– Only two targets: Shallow vs. deep troughs– So no frequency effect within trough categories

• Wrong: Frequency affects all trough depths– Increasing frequency always means shallower

troughs, even among already shallow troughs

19

Still more evidence...• ... for gradience, and for lexical status• Twenty new native speakers of S. Min• Familiarity judgment task

– Hear artificially contracted forms– Judge their familiarity (magnitude estimation)

• Do familiarity and trough depth correlate?– If so, acoustic detail is stored in perceptual lexicon

20

Artificially contracted stimuli

21

Production & perception correlate

0.5 1.0 1.5 2.0

010

20

30

40

Subjective Familiarity of Contracted Forms

Tro

ugh d

epth

(dB

)

Linear fitLoess fit

(correlation remains even if corpus frequency is taken into account) (and vice versa)

><

22

Maybe a listener-oriented effect?

• Twenty more native speakers of S. Min• Familiarity judgment task

– Hear uncontracted forms from first experiment– Judge their familiarity (magnitude estimation)

• Do familiarity and trough depth correlate?– If so, speakers are contracting to just the degree t

hat listeners can compensate for via their familiarity with the intended categorical targets

23

0.5 1.0 1.5 2.0

010

20

30

40

Subjective Familiarity of Uncontracted Forms

Tro

ugh d

epth

(dB

)

Linear fitLoess fit

Correlation is just so-so

(correlation disappears when corpus frequency is taken into account)

<

24

The story so far

• Lenition is lexical– Frequency affects it directly– Listeners store copies of lenited forms

• Lenition is phonetically gradient– Not merely selection of categorical allomorphs

• Yet frequency effects in lenition aren’t “deep”– Occur with shadowing (cf. Pluymaekers et al. 2005)

– Not reducible to lexical access effects– Don’t respect phonological units (syllables)– Speakers don’t care about listeners...?

25

Invasion of the phonologists

*StructureHighFreq >> Faith >> *StructureLowFreq

• Interface structure constraints (Hammond 2004)

HighFreq*Structure >> Faith

LowFreqFaith >> *Structure

• Cophonologies (Myers & Li 2005)

FaithLowFreq >> *Structure >> FaithHighFreq

• Interface faith constraints (Hammond 1999, Myers 1999, Coetzee 2007; cf. Boersma 1998, 2006)

26

Learning these rankings

• E.g. faith constraints for high-frequency items get violated more often, hence get demoted

• Boersma (2006) applies this to common vs. rarer gestures (e.g. Cor vs. Lab), but it also works for word frequency effects in gradient lenition

XLowFreq

X Y

X XHighFreq

X Y X Y

rarely violates Faith

often violates Faith

27

• In lenition, markedness is defined articulatorily– Why not FaithLowFreq >> *[+rd,-bk] >> FaithHighFreq?

• Lenition is phonetically gradient– Why can’t X & Y be categorical?

• One-mechanism-fits-all approach misses point– Lenition is an articulatory phenomenon, and so are i

ts frequency effects

Problem: Lenition isn’t Faith alone

XLowFreq

X Y

X XHighFreq

X Y X Y

28

Problem: Kids work backwards

• Higher-frequency words are pronounced more adult-like (Tyler & Edwards 1993; Gierut et al. 1999)

– Gierut et al. (1999) analyze this with interface faith constraints ranked the reverse of lenition

FaithHighFreq >> *Structure >> FaithLowFreq

• Boersma (p.c.) calls this learning “articulatory” – How does processing level affect the XY logic?– How can there be “extra-lexical” frequency effects?

29

Escape hatch # 3:Everything is lexical

• Memory resides in synapses, so everything the brain does is “memorized”– Frequency alone can’t diagnose processing stage

• E.g. whole-word frequency effects in the access of morphologically complex words– Whole-word storage in the mental lexicon?– ... or memory traces of the morpheme combination

process (Taft 2004, Myers et al. 2006)?

• Does grammar provide any insights here?

30

The argument (reprise)

• If all lenition is driven by articulation...

• ... and is expressed mentally by lexicalized peripheral processing, not grammar or the “linguist’s lexicon” ...

• ... and lenition is the source of some of the most interesting “real” phonology ...

• ... then what does a markedness-based grammar have left to do?

31

Frequency effects in lenitionand the challenge of

lexicalized markedness

James MyersNational Chung Cheng University

[email protected]

Workshop on Variation, Gradience and Frequency in PhonologyStanford University, July 2007

32

References (1/5)Aylett, M., & Turk, A. (2004).The smooth signal redundancy hypothesis: A functional explanatio

n for relationships between redundancy, prosodic prominence, and duration in spontaneous speech. Language & Speech, 47 (1), 31-56.

Baayen, R. H. (forthcoming). Analyzing linguistic data: A practical introduction to statistics. Cambridge University Press.

Berkenfield, C. (2001). The role of frequency in the realization of English that. In J. Bybee, & P. Hopper (Eds.), Frequency and the emergence of linguistic structure (pp. 137-157). John Benjamins.

Boersma, P. (1998). Functional phonology. The Hague: Holland Academic Graphics.Boersma, P. (2005). Phonology without markedness constraints. ICLaVE 3. http://www.fon.hum.

uva.nl/paul/presentations/ICLaVE3.pdf.Boersma, P. (2006). The acquisition and evolution of faithfulness rankings. 14th Manchester Pho

nology Meeting. http://www.fon.hum.uva.nl/paul/presentations/BoersmaMFM14.pdf.Boersma, P., & Weenick, D. (2007). Praat: Doing phonetics by computer. http://www.fon.hum.u

va.nl/praat.Bybee, J. L. (2000a). Lexicalization of sound change and alternating environments. In M. Broe,

& J. Pierrehumbert (Eds.), Papers in laboratory phonology V(pp. 250-269). CUP.Bybee, J. L. (2000b). The phonology of the lexicon: Evidence from lexical diffusion. In M. Barlo

w, & S. Kemmer (Eds.), Usage-based models of language (pp. 65-85). CSLI.Bybee, J. L. (2001). Phonology and language use. Cambridge: Cambridge University Press. Bybee, J. L. (2002). Word frequency and context of use in the lexical diffusion of phonetically co

nditioned sound change. Language Variation and Change, 14, 261-290.

33

References (2/5)Cacoullos, R. T., & Ferreira, F. (2000). Lexical frequency and voiced labiodental-bilabial variati

on in New Mexican Spanish. Southeast Journal of Linguistics, 19(2), 1-17.Chung, R.-F. (1996). The segmental phonology of Southern Min in Taiwan. Taipei: Crane Publis

hing.Chung, R.-F. (1997). Syllable contraction in Chinese. In F.-F. Tsao, & H. S. Wang (Eds), Chines

e languages and linguistics III: morphology and lexicon (pp. 199-235). Taipei: Academia Sinica.

Coetzee, A. (2007). A lexical theory of variation. Talk presented at the Workshop on Variation, Gradience and Frequency in Phonology. Stanford University.

Cohn, A., Brugman, Crawford, J. C., & Joseph, A. (2005). Phonetic duration of English homophones: An investigation of lexical frequency effects. LAS 79th meeting.

Fidelholtz, J. L. (1975). Word frequency and vowel reduction in English. CLS, 11, 200-213.Gierut, J. A., Morriette, M., & Champion, A. H. (1999). Lexical constraints in phonological acqu

isition. Journal of Child Language, 26, 261-294.Hale, M., & Reiss, C. (2000). Substance abuse and dysfunctionalism: current trends in phonology.

Linguistic Inquiry, 31 (1), 157-169.Hammond, M. (2004). Frequency, cyclicity, and optimality. Studies in Phonetics, Phonology, an

d Morphology, 10, 349-364.Hammond, M. (1999). Lexical frequency and rhythm. In M. Darnell, E. Moravcsik, F. J. Newme

yer, M. Noonan, & K. M. Wheatley (Eds.), Functionalism and formalism in linguistics I: General papers (pp. 329-358). John Benjamins.

Hay, J. B. (2000). Causes and consequences of word structure. PhD dissertation. Northwestern University.

34

References (3/5)Hooper, J. B. (1976). Word frequency in lexical diffusion and the source of morphophonological

change. In W. Christie (Ed.), Current progress in historical linguistics (pp. 96-105). Amsterdam: North Holland.

Hsiao, Y.-C. E. (1999). From Taiwanese syllable contraction to the relationship between phonology, morphology and syntax: A new direction for an old issue. In Y.-M. In, Y.-L. Yang, & H.-Z. Zhan (Eds.), Chinese languages and linguistics V: Interactions in language (pp. 251-288). Taipei: Academia Sinica.

Hsiao, Y.-C. E. (2002). Tone contraction. In Chinese languages and linguistics VIII (pp. 1-16). Taipei: Academia Sinica.

Hsu, H.-C. (2003). A sonority model of syllable contraction in Taiwanese Southern Min. Journal of East Asian Linguistics 12(4), 349-377.

Johnson, T. C. (1983). Phonological free variation, word frequency, and lexical diffusion. University of Washington PhD thesis.

Jurafsky, D., Bell, A., & Griand, C. (2002). The role of lemma in form variation. In C. Gussenhoven, & N. Warner (Eds.), Laboratory phonology 7 (pp. 3-34). Mouton de Gruyter.

Jurafsky, D., Bell, A., Gregory, M., & Raymond, W. D. (2001). Probabilistic relations between words: Evidence from reduction in lexical production. In J. Bybee, & P. Hopper (Eds.), Frequency and the emergence of linguistic structure (pp. 137-157). John Benjamins.

Kawamoto, A. H., Kello, C. T., Higareda, I., & Vu, J. V. Q. (1999). Parallel processing and initial phoneme criterion in naming words: Evidence from frequency effects on onset an rime duration. J. of Experimental Psychology: Learning, Memory, and Cognition, 25(2), 362-381.

Lavoie, L. (2002). Some influences on the realization of for and four in American English. Journal of the International Phonetic Association, 32(2), 175-202.

35

References (4/5)Li, Y. (2005). Lexical frequency effects in Taiwan Southern Min syllable contraction. National C

hung Cheng University MA thesis.Mermelstein, P. (1975). Automatic segmentation of speech into syllabic units. Journal of Acousti

c Society of America, 58(4), 880-883.Munson, B., & Solomon, N.P. (2004). The effect of phonological neighborhood density on vowel

articulation. Journal of Speech, Language, and Hearing Research, 47(5), 1048-1058.Myers, J. (1999). Lexical phonology and the lexicon. National Chung Cheng University ms. http:

//roa.rutgers.edu/view.php3?id=349.Myers, J. & Guy, G. R. (1997). Frequency effects in Variable Lexical Phonology. University of P

ennsylvania Working Papers in Linguistics, 4 (1), 215- 228.Myers, J., Huang, Y.-C., & Wang, W. (2006). Frequency effects in the processing of Chinese infl

ection. Journal of Memory and Language, 54, 300-323.Myers, J., & Li, Y. (2005). Frequency effects and Optimality Theory. Paper presented at the First

Theoretical Phonology Conference, Taipei, Taiwan.Myers, J. & Li, Y. (submitted). Frequency effects in Southern Min syllable contraction. National

Chung Cheng University ms.Myers, J., & Tsay, J. (2003). Phonological competence by analogy: computer modeling of experi

mentally elicited judgments of Chinese syllables (I). (Project report). National Cheng Chung University. Research Project funded by National Science Council, Taiwan. (NSC 91-2411-H-194-022).

Patterson, D., & Connine, C. M. (2001). Variant frequency in flap production: A corpus analysis of variant frequency in American English flap production. Phonetica, 58, 254-275.

Phillips, B. S. (1984). Word frequency and the actuation of sound change. Language, 45, 9-25.

36

References (5/5)Phillips, B. S. (1999). The mental lexicon: evidence from lexical diffusion. Brain and Language,

68, 104-109.Pierrehumbert, J. B. (2001). Exemplar dynamics: Word frequency, lenition and contrast. In J. By

bee, & P. Hopper (Eds.), Frequency and the emergence of linguistic structure (pp. 137-157). John Benjamins.

Pierrehumbert, J. B. (2002). Word-specific phonetics. In C. Gussenhoven, & N. Warner (Eds.), Laboratory phonology 7 (pp. 101-139). Mouton de Gruyter.

Pluymaekers, M., Ernestus, M., & Baayen, R. H. (2005). Lexical frequency and acoustic reduction in spoken Dutch. JASA, 118 (4), 2561-2569.

Taft, M. (2004). Morphological decomposition and the reverse base frequency effect. Quarterly Journal of Experimental Psychology, 57A (4), 745-765.

Tseng, C.-C. (1999). Contraction in Taiwanese: Synchronic analysis and its connection with diachronic change. In Y.-M. In, Y.-L. Yang, & H.-Z. Zhan (Eds.), Chinese languages and linguistics V: Interactions in language (pp. 205-232). Taipei: Academia Sinica.

Tyler, A. A., & Edwards, M. L. (1993). Lexical acquisition and acquisition of initial voiceless stops. Journal of Child Language, 20, 253-273.

van Bergem, D. R. (1995). Perceptual and acoustic aspects of lexical vowel reduction: A sound change in process. Speech Communication, 16, 329-358.

van Son, R. J. J. H., Bolotova, O., Lennes, M., & Pols, L. C. W. (2004). Frequency effects on vowel reduction in three typologically different languages (Dutch, Finnish, Russian). In Proceedings of INTERSPEECH 2004. (pp. 1277-1280).

Wright, C. E. (1979). Duration differences between rare and common words and their implications for the interpretation of word frequency effects. Memory & Cognition, 7(6), 411-419.