290 likes | 470 Views
Debates about Language. Innateness Learnability Evolution. Components of Language. Phonetics Phonology Syntax Semantics Pragmatics. Components of Language Another View. Phonetics Minimal pairs Speech perception Syntax Phonology Morphology Case Agreement (etc…) Semantics
E N D
Debates about Language • Innateness • Learnability • Evolution
Components of Language • Phonetics • Phonology • Syntax • Semantics • Pragmatics
Components of LanguageAnother View • Phonetics • Minimal pairs • Speech perception • Syntax • Phonology • Morphology • Case • Agreement (etc…) • Semantics • Words and concepts • Pragmatics
Coarticulation • no 1-to-1 mapping from phonemes to muscle commands, articulator positions, or acoustics • co-articulation • influence of one speech segment on another • influence of a phonetic context upon a given segment • phonemic variation results from co-articulation
Motor Theory • Liberman et al. (1967) • Argument: Speech is special • Though phoneme realization is variable, there is a direct relationship between articulatory intent and phoneme production • What must be decoded are the articulatory positions and these will give the intended phonemes
Is speech special? • The most important single phenomenon in this respect is: Categorical perception (We perceive speech sounds as being in a limited number of categories)
Voice Onset Time (VOT) • The time delay between the beginning of a sound and the beginning of the vibration of the vocal cords that accompanies voicing
VOT in Speech • Consider the phonemes /d/ (voiced) and /t/ (unvoiced)—together with a following vowel phoneme, making the sounds /da/ and /ta/ • Timing is just subtly different for Voice Onset Time (VOT)—when the vocal cords start to vibrate—after lips open
VOT • The voiced /d/ causes /da/ to have a short VOT • The unvoiced /t/ causes /ta/ to have a long VOT
Categorical Perception: Categorization • First: synthesize sounds with VOTs increasing in equal steps from 0 to 80 ms (Eimas & Corbitt, 1973) • Second: Ask listeners to judge whether they hear a /da/ or /ta/ at each step • Listeners perceive two categories
Categorical Perception: Discrimination • Determine how well a listener can discriminate between 2 stimuli that differ in VOT • They can reliably discriminate between pairs of sounds only if they are from either side of the VOT boundary • No discrimination between stimuli with VOTS of 10 and 30 ms; reliable discrimination between stimuli with VOTs of 30 and 50 ms
Nonlinear Influence of VOT • Different versions of /t/ may vary as much as the 20 ms in VOT, equal to the difference between /t/ and /d/, but they are all perceived as /t/
Why are categories special? • Different languages have different categories • English has two on the voice onset dimension • Thai has three on that dimension • English has a boundary on the /l/-/r/ dimension • Japanese does not • So…in one language a difference in sound may make a difference between words; in another, it might not
Learned Categories? • Categorical perception turns out not to be learned by exposure to a language • Eimas et al. (1971) showed that 1 month old infants have it • Indeed, English exposed infants can distinguish between 3-way Thai sounds • extra-linguistic ability lost around 10 months • So…humans are born with ability to recognize potential phonemes of a human language • By analogy with work in vision, researchers concluded that we have special purpose phoneme detectors
Loss of Categorical Perception • Example: Hindi contrast
Perceptual Loss: Locus of Onset • When does perceptual loss take place? • Hindi contrasts tested on… • Hindi-speaking adults • Discrimination • English-speaking adults • No discrimination • English-exposed 7-month-olds(Werker, 1981) • Discrimination
Hindi (spoken in India): unvoiced unaspirated retroflex vs. dental stop Dental Stop Retroflex Stop Retroflex - tongue curled so tip is behind alveolar ridge Dental - tip of tongue on teeth (English /t/ is somewhere between the two)
Salish (Native North American—Canadian—language): glotalized voiceless stops Uvular Velar (note: they are actually ejectives - ejective is produced by obstructing the airflow by raising the back of the tongue against or behind the velum) Velar - tongue is raised against velum Uvular - tongue is raised behind the velum
Not-So-Special Perception • Three main issues undermine the view that categorical perception makes speech special: • Some speech sounds do not evoke categorical perception (e.g., vowels) • It has been shown that musical sounds (non-speech sounds) of various kinds (e.g., notes, bowing, plucking) appear to have categorical properties • Non-linguistic animals have been shown to perceive speech sounds categorically as well
The Importance of Chinchillas • Kuhl & Miller (1975) • Trained chinchillas to move to one side of their box when they heard a target sound and not to move when they did not hear it • If a sound provokes movement (however different in detail) it must sound the same to the chinchilla • K & M’s chinchillas demonstrated categorical perception for human speech sounds • Subsequently, birds and macaque monkeys have been shown to have it to
Possible Explanation • Probable that speech evolved to exploit only those sound differences that can reliably be perceived: • Chinchillas distinguish the sounds, not because they are phonemes but because they happen to sound different (given the auditory apparatus)