Introduction to the Study of Language
Functions of σ:
Regulate admissible combinations of consonants and vowels in a language
Provide the domain for many phonological rules (e.g. stress), and many rules are
sensitive to syllable structure (e.g. aspiration)
Cat /kæt/ 1 syllable
Lecture /lɛktʃər/ 2 syllables
Banana /bənænə/ 3 syllables
Hippopotamus /hɪpəpɑtəməs/ 5 syllables
The Maximal Onset Principle
Not: * ban.an.a
Onsets are favored cross-linguistically Maximal Onset Principle (universal): The consonants that fill the onset position of a
syllable are those that correspond to the maximal sequence found in word-initial
position in the particular language.
i. Transcribe the word phonemically.
ii. Nucleus/Rhyme-formation: assign Vs to nucleus; from each Nucleus, draw a
line to the rhyme node and to a ! node.
iii. Maximize Onset: put as many consonants as possible in the Onset of each
syllable, obeying language specific phonotactic constraints.
iv. Coda-formation: Assign remaining Cs to Coda; connect Coda to the Rhyme
and join Onset to !
v. Word-level construction: join syllables to Wd Stress: a prominence relation
Phonetic correlates of stress: pitch contour, length and/or loudness.
The language particular rules determining where word stress falls are usually
sensitive to syllabification.
Consequences of word stress in English: vowels are reduced to schwa, [ə], when
unstressed, but maintained when stressed.
Stress and Vowel Pronunciation in English
The big dots represent stressed syllables.
Vowels in unstressed
pronounced as [ə]
The English Vowel reduction
/i ɪ e ɛ æ ɑ u ʊ o/ are permitted only in syllables bearing stress in English.
English Vowel Reduction Rule: Vowel [ ə ] when occurring in unstressed syllable
Summary of syllabic phonology
Phonemes are grouped together into syllables.
The syllabification rules are universal.
Onsets are favored over codas.
What counts as a possible Onset depends on each language’s phonotactic
Select an inventory of phonemes, from amongst all possible human phones
Have phonological rules that derive various allophones from the underlying
Group segments into syllables according to universal syllabification rules.
Languages vary in ...
The specific inventory of their phonemes
The specific inventory of phonological rules The specific phonotactic constraints
The specific stress patterns (stress algorithms)
Computation In Phonology
Sequences of speech sounds are grouped into syllables.
Each vowel is the nucleus of the syllable.
Vowels are crucial in the construction of syllables in terms of their sounds, not
Different syllables have their own rhythm to it. There is a ‘stress’ in every one
other syllable in a word.
o For natives, stress in syllables come naturally
There’s a computational way or process to decide where or in which syllable the stress is
located (We first separate the syllables into 'foots’; how it works 9 input + process =
Count the vowels, and you know how many syllables there are:
In words that have more than one syllable, one syllable will be most prominent:
the stressed syllable.
Stress is realized on the vowels of a word, but it is traditional to attribute stress to
Assume that the stress system of I-languages is rule-based
Assume that the input to the stress system is a string of syllables with no stress
assigned, while the output is the same string with the correct stress pattern
Two ways of getting there?
Both algorithms correctly predict the observed stress patterns.
What is amiss? Having two such different algorithms misses a generalization -
namely that Maranungku is the mirror image of Weri!
Algorithm 2 for Maranungku stress:
You form pairs starting from left.
The left edge of each foot is stressed.
The leftmost stress gets to be primary stress. Extensional equivalence
Both algorithms give us the
Both predict the observed patterns correctly.
They are extensionally equivalent because they produce the same output
How to choose?
Elegance & simplicity?
Which one is more likely to reflect the mental process that actually occurs in the
mind of a Maranungku speaker?
More than one grammar can generate the output of certain language speakers, but the
goal of linguists is to model the system of knowledge that is actually in a person's mind
Evidence from one language’s stress system should bear on the best analysis of the
stress system in other languages
Languages may differ minimally in their stress systems but conform to the same
general pattern The empirical base of potentially relevant data (for generating an universal I-language) is
all linguistic output of all human languages
All language grammars are built from the same primitives part of a
computational system realized in a biological organism within the properties of
Data from other languages may bear on the issue:
o Pintupi supports the approach in terms of pairing syllables into feet!
o Aspects of cognitive plausibility that are independently established can
bear on the question (including learnability).
o Investigation points towards the fact that language is hierarchically
structured and that computation does not involve ”counting” (a highly
conscious process, contrary to subtilizing.)
Two algorithms that produce the same outputs are extensionally equivalent
If I-language are properties of individual minds, then there should be a correct
answer is the true grammar of a certain language, even if several grammars can
generate the same output
Humans are born with the capacity to learn any human (stress) system, but
experience, the data a child is exposed to, determines which system is encoded
in the learner's mind
Our goal should be to discover a single set of primitives of representation and
computation that can generate all the (stress) systems we observe Artificial intelligence: concerned with simulating human behavior
Cognitive intelligence: concerned with discovering what will generate human behavior
Subitizing: the act of abstracting knowing numbers (which is more, less, the immediate
perception of number) but not actually counting
Therefore, when figuring our stress systems, it is not useful to use the numbering
system vs. the grouping system because the numbers will go fairly high
Grouping process is part of an algorithm in syntax, morphological reduplication, and in
the phonology of stress in various languages
Targeted syllables may receive stress, or may be repeated
Some variability in grouping; starts in the beginning of the word/end of word etc.
Grouping equivalence classes together (eg. syllables) and use of computation
(grouping of two elements) similarities in a variety of languages
Internalistic approach assures that there is a correct answer since language
begins in the brain of an individual
Some people believe that there is a perfect language that exists out there somewhere
(people meaning linguists)
Other linguists also believe that millions of people consider themselves the
speaker of one language, and that is correct
External language: believes that a language is a collection of data, or a corpus (like a
collection of utterances in a specific language)
E—language: is an attempt to come up with a formal characterization of language
The corpus is often a collection of utterances produced by several speakers
Newspapers represent the output of many individuals, and so you can't
guarantee that a set of data can be modeled by a single grammar that
corresponds to a possible humanly attainable language
A single person can generate output that corresponds to multiple grammars
i.e. bilingual person
Difference between E-language and I-language is the fact that I-language is
rooted in biology
E—language only cares about the rules that can correctly generate the data
I-language teases apart the rules that can generate a corpus and the rules that
can underlie any corpus
E—language only looks at the external factors to determine what the rules are
to generate a corpus This concept sees language as independent of any psychological states of