Linguistics 200 Phonology

Syllabification rules & parameters

1. Syllable structure: A phonological phenomenon
It appears to be a universal characteristic of all languages to have syllables. (This has been
proposed even for signed languages, incidentally.) However, it is very difficult to give a
precise phonetic definition of a syllable, and it is not possible to look at a spectrogram of an
unknown language and “see” syllable boundaries in some physically measurable way.

This is because syllable structure is a phonological phenomenon — it is part of how the

mental grammar organizes speech sounds.

So, if syllable structure can’t be directly investigated through phonetics, how do we study it?
Native speaker judgements can help, especially since native speakers often have (relatively)
clear and stable intuitions about how many syllables are in a given word. However, native
speaker judgements about where syllables begin and end are not very reliable — partly because
pronouncing one syllable at a time puts them in a different phonological environment than
they would be in a word, so additional (word-level) factors interfere with the judgements.

Instead, what we need to do is use evidence about what sounds can occur where in a
language, or evidence about where certain rules apply in a language, to make a proposal for
how that language divides sounds into syllables. As always, this kind of proposal must be
backed up with argumentation based on the available evidence. (Ideally, the conclusions we
reach about syllable structure in one language based on multiple sources of evidence from
that language should be consistent.)

For example, a starting-point hypothesis would be that the left edge of a word is also the left
edge of a syllable, and the same for the right edge. Thus, patterns of consonants (or the lack
thereof) at the left or right edges of the word may give us some initial clues about possible left
or right edges of syllables. We have to be a little careful with this technique, however,
because sometimes there are special rules that apply at word edges and disrupt this word-
edge/ syllable-edge correlation. So this is a good way to get started, but be willing to revise
your initial word-edge-based hypotheses when necessary.

A final word of caution: It is important to remember that, while some aspects of syllable
structure seem to be universal, there are a number of aspects — especially related to assigning
consonants to syllables — that are language-specific. So, it is not safe to apply your own
intuitions about syllable structure in your native language to the question of syllable structure
in other languages. Every language must be examined based on the evidence it provides.

2. How the mental grammar assigns syllable structure: Overview

Once you have examined language data and formed a proposal about where the syllable
breaks are (i.e., which segments are syllabified together and which are split between

1
syllables), you can develop a phonological analysis that will assign syllable structure in such
a way as to put the syllable breaks in the right place.

As we saw above, syllable structure can vary from language to language. But within each
language, the syllable structure assigned to any given SR is predictable. This means that it
must be assigned by the phonological grammar.

As always in phonology, our goal is to propose a theory of syllabification and syllable

structure that is general enough to encompass the range of variation that we see between
languages, while being restrictive enough to account for the aspects of syllable structure that
do not vary from language to language.

However, while languages differ in the specific kinds of syllable structure that they allow, the
variation between languages actually seems to involve only a very limited number of possible
options. In other words, the cross-linguistic variation in possible syllable-building rules is a
lot more restricted than the cross-linguistic variation in possible segmental (feature-related)
rules. Our model needs to take this restrictiveness into account. One approach that has been
taken is a rule-and-parameter model of syllabification.

(1) Rule-and-parameter model of syllabification

(a) The syllable itself is modeled as a node in the phonological representation,

symbolized with the Greek letter σ (sigma). This node is the root node of a tree
structure; each tree shows which segments belong to a particular syllable.
(b) Segments are linked to a syllable node by association rules (see below).
(c) These association rules may be stated along with restrictions. For example: Does
a rule apply once, or does it apply iteratively? What natural classes of segments
are affected?
(d) The model takes into account the highly restricted possibilities for cross-
linguistic differences as follows:
• There are three syllabification rules: the Nucleus Rule, the Onset Rule, and
the Coda Rule. These rules always apply in this order.
• The Nucleus Rule and the Onset Rule are mandatory, so they apply in every
language.
• The Coda Rule is optional, so some languages do not use this rule at all.
• The restrictions on these rules are universally provided as a set of of
parameters, or “menu options,” which are part of the formalization of each
rule. Languages are free to select different settings for each parameter, but
that is the extent of their freedom.

3. Basic terminology for syllable structure

The following diagram shows some basic terminology for syllable structure. These terms are

2
further defined and discussed below as the relevant rules and parameters are introduced.

Syllable structure diagram

We can also define the following notational conventions for use in stating syllabification
rules and parameters:

Notational conventions
• X — any segment at all
• V — any [-cons] segment that is not a glide
• C — any [+cons] segment or any glide
(The handout “Representing glides and high vowels” will have more to say about C, V, and glides.)

4. Rules and parameters

A. Syllable nuclei
One universal aspect of syllable structure is that every syllable has a nucleus, sometimes also
called a peak. This is the core or main part of the syllable. We can incorporate the universal,
mandatory character of the syllable nucleus into our model by proposing that the first
syllabification rule in every language is the Nucleus Rule, which associates a syllable node
with each possible nucleus. The dashed line in the rule means “Form an association
between...”.

Nucleus Rule
(Associate a "possible nucleus" with a syllable node.)

The set of things that can be syllable nuclei differs somewhat from language to language,
however. All languages allow a single vowel to be a nucleus, but some languages are more
permissive than that. Specifically, some languages allow two vowel segments in the same
nucleus (=diphthongs). Other languages allow consonants, or particular types of consonants,
to be nuclei. And some languages, like English, allow both. So here are the first parameters
that we can include in our model.

3
 Parameters concerning nuclei:
• V nucleus is allowed (universal)
• Diphthongs (sequence of two V) allowed? (yes/no)
□ If yes, characterize the relevant natural class(es)
• Other natural classes allowed as possible nuclei? (yes/no)
□ If yes, characterize the relevant natural class(es)

In summary, the universal Nucleus Rule applies in each language to whatever elements are
designated as possible nuclei in that language, through its settings for the nucleus-related
parameters. This proposal captures the fact that all syllables have nuclei and that vowels can
always be nuclei. It also allows for the optional extension of possible nucleus status to
diphthongs and consonants, for languages that choose these parameter settings. Our system
(correctly) does not allow us to create a language where consonants are always nuclei and
vowels are not, because the parameters as stated do not include this option.

B. Syllable onsets
Another universal fact about syllable structure is that consonants are syllabified as onsets
whenever that is possible. In particular, a consonant between two vowels is universally
syllabified as an onset to the second syllable ([a.tu]), not a coda to the first syllable ([at.u]).
We can account for the universal preference for onsets over codas by proposing a universal
Onset Rule, which applies (in all languages) before any rule that creates codas. This way, all
consonants will have a chance to become an onset first, and will only end up as a coda as a
last resort.

Onset Rule

(For any C that precedes a syllabified segment, associate that C with the same syllable.)

Note that this rule does not specify whether the consonant is joining a syllable that currently
lacks an onset consonant, or one that already starts with a consonant (thereby forming an
onset cluster). This is a desired result, because it says, in essence, “Form the largest allowable
onset.” Once again, we can turn to language-specific parameters to specify what an allowable
onset will be in each language.

Here are the possible ways that languages can differ regarding onsets. Some languages allow
onsets to include more than one consonant, forming clusters, while other languages allow
only one onset consonant. Furthermore, languages that allow clusters often place restrictions
on what kinds of consonant combinations can be in a cluster. (If you need to state a cluster
restriction, remember to state it in terms of features and natural classes.)

4
 Onset Rule iterativity parameter:
• Onset Rule applies iteratively for each syllable? (yes/no)
□ If yes, any restrictions on possible onset clusters?

In summary, the universal status of the Onset Rule means that after nuclei are created, the
next step in every language is to create the largest possible onset for each syllable. If the
iterativity parameter is set at “no”, then the Onset Rule is permitted to apply at most once for
each syllable. But if the iterativity parameter is set at “yes”, then the Onset Rule will apply to
each new consonant it finds, as long as adding that consonant to the syllable will create an
allowable cluster. For example, the English word flee will undergo the Onset Rule twice
because English has the onset iterativity parameter set at “yes” and [fl] is an allowable onset
cluster. (We’ll see more about factors that determine allowable clusters in a future class.)

Derivation: Syllabification of English flee

C. Syllable codas
Many languages have one more syllabification rule that applies after the Onset Rule, namely,
the Coda Rule. This rule applies to syllabify remaining unsyllabified consonants as codas. It
is crucial to specify that the rule only applies to unsyllabified consonants (symbolized C'), so
that it will not affect any consonants that have already been syllabified as onsets.

5
 Coda Rule

(For any unsyllabified C that follows a syllabified segment, associate that C with the same syllable.)

Every language has syllables with nuclei and onsets, but not every language has syllables
with codas. In terms of our model, this means that every language has the Nucleus Rule and
the Onset Rule as part of its system of syllabification rules, but the Coda Rule is not a
universal part of syllabification. Therefore, there is a parameter that determines whether a
language will even apply the Coda Rule at all. There may also be restrictions on which
consonants are allowable codas in the language; it is often the case that not all of the
consonants found in a language can appear in coda position. Finally, as with onsets,
languages can vary as to whether they allow coda clusters, and if so, whether there are
restrictions on which consonants can combine to form a cluster. (Again, any coda or cluster
restrictions must be stated in terms of features and natural classes.)

Parameters concerning Coda Rule:

• Does the Coda Rule apply? (yes/no)
□ If yes, any restrictions on possible coda segments?
• Coda Rule applies iteratively for each syllable? (yes/no)
□ If yes, any restrictions on possible coda clusters?

5. Example: Setting parameters, applying syllabification rules

Like most phonological rules, syllabification rules apply to a phonological representation
whenever their environments are met. For example, a UR with seven possible nuclei will
probably end up with seven syllables in the SR (unless further rules apply to change things
around). In many languages, it doesn’t matter whether the syllabification rules apply to the
string of segments from left to right or from right to left, but sometimes the direction of
application of a particular rule does matter and needs to be specified.

There is one important way in which syllabification rules are different from most of the rules
that we have seen so far. They have a special status with respect to rule ordering in most
languages: they apply continuously, that is, they reapply every time another phonological
rule has applied (although this only has an observable effect if the other phonological rule
was one that inserted or deleted segments, or otherwise changed things that are relevant for
the application of the syllabification rules).

Here is an example, from the Cairene Arabic data set discussed in class.

In the course of the class discussion, we used various sources of evidence to discover what

6
syllables are like in Cairene Arabic. Then, we developed a formal account of syllabification in
this language, which consists of a list of syllable-structure parameter settings, as follows.
Remember that our analysis strategy is to propose the most restrictive parameter settings
that are compatible with the data set, thereby making our analysis of the syllabification
phenomena as simple as possible.

Parameter settings for Cairene Arabic

Diphthong allowed in nucleus? no
C nucleus allowed? no
Onset Rule applies iteratively? no
Coda Rule applies? yes
Restrictions on possible coda segments? no
Coda Rule applies iteratively? no

Now, here is a derivation that shows syllabification rules applying in Cairene, with these
parameter settings constraining the application of the rules (in particular, no clusters are
formed).

Derivation: Cairene Arabic

The two leftover consonants cannot be syllabified at this point: onset and coda clusters are not
allowed, so there is no place to put these consonants.

7
Different languages do different things to segments that cannot be syllabified. Some
languages apply a deletion rule. In Cairene, there is an epenthesis (insertion) rule, which
inserts a high front vowel after the first unsyllabified consonant (scanning the string of
segments from left to right).

Epenthesis rule applies after unsyllabified consonant

Interestingly, the epenthesis rule does not apply twice, even though there is a second
unsyllabified consonant. This shows us that as soon as epenthesis has applied once, the
syllabification rules must reapply. The new syllable [til] can now be formed; the [l] is no
longer unsyllabified, so the epenthesis rule does not apply a second time.

Syllabification rules reapply

We can also argue that the syllabification rules apply (for the first time) before the epenthesis
rule, because we need to know which consonants are left unsyllabified in order to apply the
epenthesis rule. This supports our claim that syllabification rules apply continuously.