Phonology

Question 1

Psychological reality of the phoneme allophone distinction

Answer 1

Sapir 1933 found that speakers of Southern Paiute (which has a lenition rule after vowels) realised certain sounds differently if asked to articulate the first and second syllable of a word.

Cowan & Levitt studied a woman’s language game, and find that she is not reversing letters but rather phonemes - silent letters ignored, homographs pronounced differently, diphthongs and affricates preserved as units.

Loan Phonology - Korean word-final [t(no audible release] is an allophone of phoneme /s/. Loan words with word-final post-vocalic /t/ are invariably assigned the /s/ phoneme (Martin 1992), which becomes evident when morphological endings are added to the word and the t is realised as an s because it’s no longer word-final.

Question 2

Phoneme Theory

Answer 2

Phoneme theory posits that sounds for languages are stored as abstract units (phonemes) in the mind. They have underlying representations, but also ‘allophones’: phonetic variants of the underlying form which occur in specific contexts.

There are various views as to what the phoneme consists of (acoustic sound, articulatory gesture, feature matrix)

Question 3

Exemplar Theory

Answer 3

According to exemplar theory, mental storage of sounds is not abstract, but a collection of memories of phones (instances of a sound’s use), words, and multi-word utterances. These can be used to categorise new sounds encountered.

Question 4

DEFINITION OF: Phoneme

Answer 4

• form in which a sound is stored in the abstract mental grammar
• form underlying a set of surface allophones
• minimal unit of language that suffices to distinguish meaning

Question 5

DEFINITION OF: Allophone

Answer 5

• member of a set of speech sounds identified by a native speaker as the same sound
• a non-contrastive surface manifestation of a phoneme

Question 6

Complementary Distribution

Answer 6

When 2 sounds occur in non-overlapping phonetic environments
Phonetically similar phones in complementary distribution are usually allophones

Question 7

Free Variation

Answer 7

When 2 or more allophones freely occur in the same phonetic environment with no semantic effect

Question 8

Minimal Pair

Answer 8

2 words that differ only by a single sound in the same position and have different meanings

Question 9

HINTS FOR SOLVING PHONOLOGY PROBLEMS

Answer 9

1) Look for minimal pairs
2) List environments in which sounds occur (complementary distribution?)
3) are sounds in CD phonetically similar?
4) determine which form is the ‘elsewhere’ form
5) for the form conditioned by a rule, determine the rule: should describe a common process like assimilation in terms of natural classes of sounds interacting with neighbouring segments / syllable structure

Question 10

Features

Answer 10

Saussure thought humans divided the perceptual world into a binary opposites (+/-). This led to the proposal of ‘distinctive features’: phonetic features which distinguish phonemes. The idea is that a phoneme is composed of (strong view) / categorised by (weak view), a set of phonetic features (+/-).
As such, phonemes sharing features would form natural classes.

Question 11

Evidence that natural classes influence linguistic processes.

Answer 11

Coronal Consonants (made with front section of tongue). 
In AmE yods are dropped after coronal consonants (glide suppression) but not labials or velars.

Question 12

Evidence for features: Synchronic Alternations

Answer 12

the allophonic relation between voiced stops and fricatives (intervocalic) in spanish can be described as the spread of [+continuant] feature to [+voice, -sonorant] segments].

English place assimilation for can be described as the spread of the place of articulation (coronal, dorsal, labial) feature leftwards from the [-nasal] segment to the [+nasal] segment.

Question 13

Evidence for features: Historical Change

Answer 13

Grimm’s Law: Innovation where [-voice] segments become [+continuant]

ease of articulation?

u could say synchronic alternations are results of historical changes which happened due to ease of articulation requirements (e.g final devoicing from difficulty voicing without a following vowel), don’t need to be encoded in features - but for the weak view, this still works, and psycholinguistic evidence!

Question 14

Psycholinguistic evidence for features

Answer 14

Language disorders: stopping (+continuant to -continuant), devoicing (+voice>-voice)
could say it’s just muscle control failing
consonant

Consonant Vowel Dissociations
Dysgraphic patients show a strong tendency for substitution errors (82%) and transposition errors (62%) to preserve the C/V status of letters. (percentages for one patient)
Cotelli et al 2003 patient only struggled writing vowels (selective dysgraphia)
Caramazza, Miceli, Chialant 2000 - one patient made errors 3 times as often on vowels, other same pattern for consonants

Mismatch fields and Voice
pa ta ka da pa ka ta
(only the voiced one detected as different by Broca’s area - this is the only feature which appears in a many to one ratio)
(Phillips, Pellathy & Marantz 2000)

Speech Errors: Fromkin 1971
[voice] glear plue sky
[place] computation > pɒŋkjuːteɪʃn̩
[nasal] cedars of lebanon > cedars of lemadon

Question 15

Evidence for features from FLA

Answer 15

STOPPING: p for f [maip] ‘knife’
! t for s, θ [tuwt] ‘suit’; [ba:t] ‘bath’
! d for z, ð [dʌd] ‘does’; [ʌdʌ] ‘other’

Place harmony:
dorsal harmony - take → [gek]
labial harmony - zeep ‘soap’ → [fep] (dutch)
coronalisation/velar fronting - kiss > tiss

Question 16

Different positions on how sounds ‘phonemes’ are stored in the brain

Answer 16

• traditional PHONEME theory - an abstract underlying acoustic form
• Motor theory - a gestural score
• features (strong view) - feature matrices
  ^these ones (abstract theories) have top down and bottom up processing
• Exemplar theories - memories of words and multiword sequences are stored

Question 17

Evidence for exemplar based theories

Answer 17

(See Lent 1 essay for actual figures and reference)

French liaison - can’t be accounted for by simple before vowel rules bc in fact percentage liaison is correlated to frequency

thought to be an irregular variant so preserved in frequent elements

Question 18

What is the cohort model?

Answer 18

Marslen-Wilson’s cohort model is the idea that items in the lexicon are activated segment by segment during speech processing.

if a stimulus is presented up to or beyond the recognition point (the point where only one word contains that initial sequence of phonemes), only one word is primed, whereas if the recognition point of the stimulus is not reached, multiple words – the cohort – are primed. For example, in Pienie Zwitserlood’s study of Dutch ‘Kapit’ (the initial segments of both ‘Kapitein’ and ‘Kapitaal’), words relating to both words were primed, but using ‘Kapitein’, only words relating to ‘Kapitein’ were primed.

Question 19

Traxler notes that there is no 1-1 relationship between acoustic signal and phoneme. Is this a problem?

Answer 19

bottom up processing could involve a matching process (most similar phoneme), discarding unnecessary phonetic information.

Question 20

Evidence for the Motor Theory

Answer 20

This is evidenced by the McGurk Effect: if someone sees somebody’s lips saying ‘ga’, but the sound produced (in edited footage) is ‘ba’, the listener might hear ‘da’. This shows that when it is available, we use visual information related to production to perceive speech, implying that our understanding of speech sounds is related to their production rather than only simply the acoustic signal. In addition, nNeuroimaging data shows that there is a lot of activity in the motor cortex when listening to speech (Pulvermuller et al), and sounds produced in different places correspond to activity in different parts of the motor cortex.

However, there is also evidence to the contrary: Perception can occur by those who cannot produce speech. According to Eimas et al (1971) - Children are fully capable of perceiving the differences between many different speech sounds despite being incapable of producing them. In addition, Japanese Quail and Chinchillas are capable of perceiving speech sounds of phoneme size (learning to respond to some speech sounds and refrain from responding to others, and showing evidence of compensation for coarticulation) but they obviously do not have a human-like articulatory apparatus. So although motor cortex activity and the McGurk effect suggest a link between perception and production, the fundamental process of speech perception cannot rely on the ability to produce speech as shown by the cases of children and animals.

Question 21

Problem with proposing that phonemes ARE feature matrices.

Answer 21

doesn’t account for length

Question 22

The idea behind phonological rules

Answer 22

that we extract predictive generalisations rather than responding mechanically to statistical data.

This suggests you use rules (feature based) to map phonetic plan from underlying to surface form, rather than memorising all input data (you couldn’t deal with new loan words) or even using analogy (what if words ended with novel phonemes).

Take the pigeon study - they extracted the simplest rule possible, despite there being additional important data.

Question 23

What does + mean in rules

Answer 23

morpheme boundary