Li6 Phonology and MorphologyLi6 Phonology and Morphology

Features and natural classes

Roman Jakobson, 1896-1982

Today’s topicsToday’s topics

Natural classes features Evidence for features

Natural classesNatural classes

There is robust evidence from many domains that phonemes pattern in natural classes. Quick example: glide suppression in AmE

Tune, dew, news, lute, suit… Pure, bugle, cute, music…

Hypothesis: phonemes composed of features Jakobson, Trubetzkoy, and Karcevsky 1928 Drawing on decompositional theories of Plato, Democritus Analogous to elements being made of atoms, or more

accurately atoms being made of electrons, protons, etc. Basic idea: natural class behavior involved manipulation

of simple feature combos (typically 1 feature) Analysis of glide suppression: no glide / [coronal] _ /u:/

Evidence for featuresEvidence for features

Conventional evidence Synchronic alternations, Historical change

Psycholinguistic evidence Speech errors, priming studies…

Synchronic alternationsSynchronic alternations Spanish spirantization

noun definite glossbanca [baŋka] la banca [la βaŋka] bankdemora [demora] la demora [la emora] delaygana [gana] la gana [la γana] desire Why {b d g} ~ {β γ}, not e.g. *{r m e}? Spread of [+continuant]

English voicing assimilation e.g. plural /-z/ [s] / {ptkfθ(x)} _

Place assimilation i[n]ept i[n]decent i[m]plausible i[ŋ]capable

Historical changeHistorical change

Grimm’s Law

LatinLatin EnglishEnglish

pater father

ped- foot

trēs three

dent- tooth

canis hound

centum hundred

IE *p *t *k

Lat p t k

English f θ h

Innovation:Innovation: *[-voice] > [+continuant]

Which is shorthand for:“All voiceless phonemes become [+continuant]”

Potential problemPotential problem

One might argue that synchronic alternations are the residue of historical changes, which in turn are the result of physical constraints on perception and production, and therefore don’t need to be encoded in terms of features.

This, however, doesn’t account for the wide range of psycholinguistic evidence…

Psycholinguistic evidencePsycholinguistic evidence

Disorders Neurolinguistic studies Speech errors Child and adult acquisition facts Priming and masking studies

DisordersDisorders

Language disordersLanguage disorders

Stopping f v s z θ → p b t d t d [+continuant] → [-continuant]

Devoicing b d g → p t k [+voice] → [-voice]

[cons]: C-V dissociations[cons]: C-V dissociations

Dysgraphic patients show strong tendency for substitution and transposition errors to preserve the C/V status of letters E.g. 82% of AS’s substitutions and 62% of his transposition errors

preserved CV status

Cotelli et al. 2003 Patient recovered from fluent aphasia to selective dysgraphia The spelling disorder was selective for vowels

Ferreres et al. 2003 Patient with alexia secondary to cerebral lesion Errors in the reading of non-words affect vowels more than consonants.

(Cf. also Miceli et al. 2004, Buchwald and Rapp (forthcoming)) A case study…

Caramazza, Miceli, and Chialant 2000Caramazza, Miceli, and Chialant 2000 Alicia, 41, failed to correctly produce vowels three times

more often than consonants. Irena, 52, had the opposite problem; she failed three times

more often on consonants. For example, instead of saying "leggere" ("to read" in Italian),

Irena might say "seppere" and Alicia "loggiare." Both words are meaningless.

NeurolinguisticsNeurolinguistics

Mismatch fields and [voice]Mismatch fields and [voice]

pæ tæ tæ kæ dæ pæ kæ tæ pæ kæ bæ tæ ...

– – – – [+voi] – – – – – [+voi] – …

• Voiceless phonemes are in many-to-one ratio with [+voice] phonemes

• No other many-to-one ratio in this sequence

Phillips, Pellathy & Marantz 2000

Phillips, Pellathy & Marantz 2000

Mismatch fields and [voice]Mismatch fields and [voice]

Speech errorsSpeech errors

Speech errors (Fromkin 1971)Speech errors (Fromkin 1971)

voicing [voice] clear blue sky glear plue sky big and fat pig and vat is Pat a girl is bat a curl

place [coronal, dorsal, labial…] computation po[ŋ]kutation

nasality [nasal] Cedars of Lebanon Cedars of

Lemadon

Speech errors (Jaeger 1992)Speech errors (Jaeger 1992) Children as young as 1;7 make slips in which single

consonants or single vowels are substituted or exchanged, implying segmental organization in phonological representations.

Corpus of 366 consonant substitutions and reversals made by children aged 1;7-6;0 subjected to a multidimensional scaling analysis

Findings errors governed by patterns of phonetic similarity,

suggesting that these segments have phonetic (feature) structure.

Errors by feature (in decreasing order of frequency): place of articulation [voice] [nasal]

AcquisitionAcquisition

L1 Acquisition errorsL1 Acquisition errors

[continuant] p for f [maip] ‘knife’ t for s, θ [tuwt] ‘suit’; [ba:t] ‘bath’ d for z, [] ‘does’; [] ‘other’

place of articulation Dorsal harmony (English)

take → [gek] Labial harmony (Dutch, English)

snoep ‘candy’ → [fup], zeep ‘soap’ → [fep] Dorsal or coronal default

kiss → tiss (“positional velar fronting”) Japanese children often replace t with k

Acquisition of mannerAcquisition of manner

Jusczyk, Goodman, and Baumann 1999 Nine-month-old English-learners Headturn Preference Procedure Infants exposed to two types of lists of CVC syllables

Items in the experimental lists shared a particular phonetic property

Items in the control lists were unrelated Results

9-month-olds sensitive to shared features that occur at the beginnings, but not at the ends of syllables.

They had significant listening preferences for lists in which the items shared the same manner of articulation at syllable onsets.

AssociationAssociation

Žagar and Locke 1986 Method

Ten children, aged 4;6 to 5;5, trained to associate consonantal features of voicing, manner, and place of articulation with cups of a particular location and color.

Results Performance on untrained generalization trials

exceeded chance on voicing and manner, but not on place of articulation.

Conclusion Children store sounds with feature structure and can

target these features in generalization formation.

Learning nonce wordsLearning nonce words

Anisfeld, Barlow, and Frail 1968 Method

1st and 2nd graders (i.e. 6-8 yrs old) given CVC singular nonce words (e.g. NAR) orally

Asked to choose between two plurals (e.g. NARF vs NARK) Results

Kids preferred final sounds agreeing with /z/ for [continuant] or [strident]

Conclusion Kids’ plural rules are formulated in terms of features, not

segments When forced to pick a new plural, they pick the option that

shares the most features with their normal plural affix.

Adult acquisitionAdult acquisition

Pycha et al. 2003 Is a less formally complex grammar (with

complexity computed in terms of features) easier to learn? simple: Xα…X Xα…Xα

complex: Xα,β…X Xα,β…X

Adult acquisitionAdult acquisition

Design Subjects asked to learn patterns of non-local vowel interaction

by listening to nonce words. Subjects told they’re listening to singular-plural pairs in a new

language, and their task is to learn how to make plurals in this language.

Each word consists of a CVC stem followed by a VC suffix whose vowel alternates according to the vowel of the stem.

Subjects listen to pairs of stem and suffixed forms exhibiting a particular pattern, and are then asked to make judgments about novel pairs of stem and suffixed forms.

Correct judgments taken as an indication of learning.

Adult acquisitionAdult acquisition

Experimental condition [+simple]

Stem and suffix V agree in backness [-back] root: CiC-εk, CIC-εk, CæC-εk [+back] root: CuC-k, CUC-k, CaC-k

Control condition [-simple]

Stem and suffix V may agree or disagree in [back], depending upon what the stem V is CiC-εk, CUC-εk, CæC-εk CuC-k, CIC-k, CaC-k

Adult acquisitionAdult acquisition Results

Interpretation Confirmation of Simplicity Hypothesis:

naïve listeners learned phonological processes that exhibited single-feature predictability significantly better than processes that did not.

Priming and maskingPriming and masking

Priming and featuresPriming and features

Lukatela et al 2001 Question:

Does the priming of a word (e.g. sea, film, basic) by a rhyming non-word depend on the non-word's featural similarity to the word?

Method mask--prime--target--mask sequence with both brief (57 ms in

Experiments 1 and 2) and long (486 ms in Experiment 1) prime durations.

Results Non-word primes that differed from their targets by a single phonemic

feature (initial voicing as in ZEA, VILM, PASIC) led to faster target lexical decisions than non-word primes that differed by more than a single phonemic feature (e.g. VEA, JILM, SASIC).

Conclusion Visual word recognition seems to involve a feature-based level of

processing.

Backwards maskingBackwards masking

Bedoin and Chavand 2000 Question

Does phonetic similarity improve performance in backward-masking experiments?

Method Subjects had to recall a briefly presented target that had been immediately

replaced (masked) by another stimulus. This task is difficult because target processing is disrupted by the mask. This disruptive effect is reduced when the target and mask are

orthographically or phonemically similar (Perfetti & Bell 1991). Results

Featural similarity reduced deleterious masking effects e.g. DÉBUT /deby/ was better identified if masked by zévut (differs only in

[cont]) than by séfut (which differs in [cont] and [voice]). Conclusion

The features [voice] and [continuant] appear to be involved in processing.

Intraword priming/maskingIntraword priming/masking Krifi et al 2003 Method

Adult skilled readers and third graders with average reading level phonetic priming and masking effects were assessed within one

briefly presented CVCV printed nonce stimulus. The consonant target (C1 or C2) was either similar or different in voicing to the other consonant.

DUBA, DUPA, TOBI, TOPI C’s used: {dbgvzž}, {tpkfsš} (factors out gestalt acoustic similarity)

Results Voicing similarity impaired C2 consonant detection Voicing similarity improved C1 consonant detection

Conclusions Involvement of phoneme detectors organised by inhibitory

relations based on shared features.

SummarySummary We’ve seen evidence for these features:

[voice] vowels, glides, liquids, nasals, voiced stops, voiced fricatives

[continuant] vowels, liquids, glides, fricatives

[nasal] m, n, ŋ

[consonantal] all but glides, vowels, and laryngeals (h, ?)

As well as three places of articulation: coronal, dorsal, labial

Some other important features: [sonorant]

nasals, liquids, glides, vowels [high]

dorsal consonants, some vowels

ReferencesReferencesAnisfeld, M., J. Barlow, and C. Frail. 1968. Distinctive features in the pluralization rules of English speakers. Language and Speech 11:31-37.Bedoin, Nathalie. 1998. Phonetic features activation in visual word recognition: The case of voicing. Xth Congress of the European Society for Cognitive

Psychology (ESCOP X), Jerusalem.Bedoin Nathalie. 2003. Sensitivity to voicing similarity in printed stimuli: effect of a training programme in dyslexic children. Journal of Phonetics 31.3:541-546.Berent, Iris & Charles Perfetti. 1995. A rose is a REEZ: The two-cycles model of phonology assembly in reading English. Psychological Review 102:146-184.Buchwald, Adam and Brenda Rapp. [forthcoming]. Consonants and vowels in orthographic representations. Cognitive Neuropsychology 1–30.Caramazza, Alfonso, G. Miceli, & D. Chialant. 2000. Separable processing of consonants and vowels. Nature 403:428-30. Cotelli, Maria, Jubin Abutalebi, Marco Zorzi, and Stefano Cappa. 2003. Vowels in the buffer: a case study of acquired dysgraphia with selective vowel

substitutions. Cognitive Neuropsychology 20.2:99–114.Ferreres, A., C. Lopez, and N. China. 2003. Phonological alexia with vowel-consonant dissociation in non-word reading. Brain and Language 84.3:399-413.Fromkin, Victoria. 1971. The non-anomalous nature of anomalous utterances. Language 47.1:27-52.Hu, Chieh Fang and Hugh Catts. 1993. Phonological recoding as a universal process? Evidence from beginning readers of Chinese. Reading and Writing

5.3:325-337.Jaeger, Jeri. 1992. Phonetic features in young children’s slips of the tongue. Language and Speech 35.1-2:189-205.Jakobson, Roman, Serge Karcevsky and Nikolaj Trubetzkoy. 1928. Quelles sont les méthodes les mieux appropriés à un exposé complet et pratique d’une

langue quelconque? Reprinted in Jakobson 1971, pp. 3-6.Jusczyk, Peter, Mara Goodman, and Angela Baumann. 1999. Nine-month-olds’ attention to sound similarities in syllables. Journal of Memory & Language

40:62-82.Krifi, Sonia, Nathalie Bedoin, and A. Mérigot. 2003. Effects of Voicing Similarity Between Consonants in Printed Stimuli in Normal and Dyslexic Readers. Current

Psychology Letters 10.1.Lukatela, G., T. Eaton, C. Lee, and M. Turvey. 2001. Does visual word identification involve a sub-phonemic level? Cognition 78.3:B41-52.Obleser, Jonas, Aditi Lahiri, & Carsten Eulitz. 2004. Magnetic Brain Response Mirrors Extraction of Phonological Features from Spoken Vowels. Journal of

Cognitive Neuroscience 16:31-39. Phillips, Colin, Tom Pellathy, and Alec Marantz. 2000. Phonological Feature Representations in Auditory Cortex.Pycha, Anne, Pawel Novak, Ryan Shosted, and Eurie Shin. 2003. Phonological rule-learning and its implications for a theory of vowel harmony. In Proceedings

of WCCFL 22, Gina Garding and Mimu Tsujimura, eds., 423-435.Stevens, Kenneth. 2003. Acoustic and perceptual evidence for universal phonological features. Proceedings of the XVth International Congress of Phonetic

Sciences, Barcelona, 33-38. Vaux, Bert. 2006. The role of features in a symbolic theory of phonology. In Contemporary views on architecture and representations in phonological theory, Eric

Raimy and Charles Cairns, eds. Cambridge, MA: MIT Press.Žagar, Linda and John Locke. 1986. The Psychological Reality of Phonetic Features in Children. Language, Speech, and Hearing Services in Schools 17:56-62.

Conditioned Head TurnConditioned Head Turn

Infant ( 0;6) hears constant background stimulus, e.g. [ba], and target, e.g. [pa]

Conditioned to turn head only to target set, not background set (correct response reinforced by flashing lights, drumming bears)

drumming bear

GlossaryGlossary

Manner features [continuant], [nasal], [strident], [lateral], and

(according to some, not including me) [approximant]