Classical Phonemics

Within the tradition of structural linguistics, the study of sounds was designated by levels of analysis. The phonetic level represented the place where speech sounds were produced. The sounds described by this level were couched in terms of articulatory phonetics or phones and are represented by bracket, [phonetic sounds]. The next level of analysis represented the level of sound perception, i.e., the level of phonemics. Since sounds are perceived differently from one language to another, linguistics needed a level that represented the uniqueness of sound perception. The called this level phonemics and represented it by means of front slashes, /phonemics/. Finally, there was a level to represent words (morphology) and their patterns of affixation (prefixes, infixes, and suffices). The units of a word were called morphs and the study of word shapes was called morphology. This level was represented by means of braces, {morphology}.



The structure of word forms and patterns

Morphemes are composed of morphs

Morphs are composed of phonemes

 Phonemics  The perception of sounds within a language
Phonemes are composed of phones
 Phonetics  The pronunciation of speech sounds within a language

This system of phonemic analysis is now known as classical phonemics. It characterized a period from the turn of this century until the middle of the century. What is important about classical phonemics is the claim that there is a need to capture phonological regularities (phonemics) as well as morphological information (morphemics). Structuralists were preoccupied by methodological concerns and developed a means of designating phonemes on the basis of their phonological contrasts, minimal pair analysis.



[pit] "pit"contrasts with [bit] "bit." They constitute minimal pairs of sounds contrasts. Therefore /p/ and /b/ are separate phonemes. They are perceived as different sounds.


[phin] "pin" and [spin] "spin" both contain bilabial sounds. One is aspirated and the other is not. These words do not form "minimal pairs." They do not contrast. Therefore, they are not perceived as separate sounds. Instead, they are considered to be variants of the same sound. They are allophones of /p/. The allophones of the phoneme are in complementary distribution with each other. The aspirated bilabial appears in word initial position and the non-aspirated voiceless bilabial appears elsewhere. The distributional pattern of one complements the other.


The emergence of generative phonology began with the work of Chomsky and Halle (1968) in Sound Patterns of English. This approach to the study of the structure of sounds differed significantly from the model of classical phonemics. Phonology was no longer seen as an inductive system beginning with phonetics and culminating in phonemics, but it was now seen as a deductive system that began at the word level and ended in phonetic representations. A system of ordered phonological rules connected the morphological level (underlying representations0 with their phonetic forms (phonetic representations). This system of ordered rules was called its phonological history..

 Structure and Process Structural Level Symbology
 Initial String Deep Structure  / original string /
 Ordered Rules    
 Final String Surface Structure  [ modified string ]

What is important about this new approach to doing phonological analysis is that it no longer contained a separate level of phonemic representations. There are several reasons why this level was abandoned. Chomsky and Halle found numerous examples in which the criterion of minimal pairs led to errors in the prediction of how sounds are perceived in English. These examples are important and they merit recapitulations.


Andre MalJcot, a specialist in acoustic phonetics, disagreed with the way that American linguists, native speakers of English, transcribed the nasalized vowels in their language. What they transcribed is given below where the nasalized [F] is symbolized by the symbol [n and oral [F] is symbolized by the symbol [a]:

study of oral and nasal vowel contrasts


What is at issue here are the sounds of "can't" and "cat" MalJcot heard them as minimal pairs, a nasalized vowel and a pure non-nasal vowel. What he heard was "can't"[knt] and "cat" [kat]. Because he was a native speaker of French, MalJcot assumed that the problem was due to the perceptual influence of his native tongue. However, he was a trained phonetician and wanted to settle this issue once and for all. As a consequence, he subjected these forms to acoustic analysis. He was fortunate in having some of the best equipment at his campus at the University of California at Santa Barbara at that time. The results confirmed his suspicions. The sound machine produced a spectrogram that clearly showed "can't" and "cat" formed minimal pairs, they were exactly alike except for the nasal vowel in one differed from the oral vowel in the other. He wrote up his results and presented them to his colleagues as a phonemic problem that needed to be resolved.

Chomsky and Halle did not disagree with the analysis of the data. They argued that the problem could be found the assumption that minimal pairs at the phonetic level determine the existence of phonemic contrasts in a language. They reanalyzed the data within the contexts of generative phonology as follows:

cant vs can't and nsalization

Their resolution of the problem is interesting because it makes several claims about how their system of generative phonology differed from classical phonemics. First, the system is deductive rather than inductive. One begins with the underlying forms and through the use of ordered rules arrive at phonetic representations. Second, they note that /kant/ and /kat/ at the morphological level (the underlying forms) do not present a problem because it designates how native speakers of English perceive these words. They are not minimal pairs at this level. Nor are they nasalized vowels existing at this level. This means that native speakers of English are impervious to the perception of nasal versus oral vowels in their language. Hence, the underlying forms represent how native speakers of a language perceive the forms of that language. Third, there is no level of phonemic analysis as in new approach. They deemed such a level unnecessary and erroneous. It predicted phonemic contrasts that were contrary to what native speakers know about their language. Fourth, the rule of Nasal Loss is a common Germanic rule that continues to exist in Modern English. It is not an idiosyncratic rule. It reflects historical changes in the language. Finally, they began phonology at the morphological level. The difference between their system and classical phonemics is given below:


  Classical Phonemics

  Generative Phonology

 Morphology, the word level. It consists of morphemes. Phonemes are combined to constitute morps and morphs constitute morphemes. Free standing morphemes function as words in a language  The underlying forms of a language begin at the word level. They are underlying words that are made up of affixes and root forms", for example, /un+interest+ing/ uninteresting.
 The Phonemic Level, the level of minimal pair contrasts  No phonemic level is needed because minimal contrasts exist at the word level
 The phonetic level where phonemic analysis begins  The level of phonetic representation where phonological analysis ends.


Another example of how the methodology of classical phonemics failed can be found in the differences of vowel length before the flap "r" in English. Chomsky noted that certain dental forms are flapped intervocalically. He symbolized the flap "r" by means of the symbol [D].


Orthographic Form  Writer  rider  latter  ladder
 Symbolic Representation  [rayDr]  [ray:Dr]  [laDr]  [la:Dr]

What is interesting about these forms is that there is a contrast between long and short vowels before the flap r [D]. Given the assumption that minimal pairs designate phonemic contrasts in a language, one would be forced to conclude from this evidence that native speakers of English readily perceive differences between long and short vowels, but such is not the case. Under their system of reanalysis, Chomsky and Halle (1968) provided the following derivational history for these forms.

Underlying forms /.../ raytr  raydr  latr  ladr
Vowel Lengthening   -  ray:dr  la:dr   -
 Flap R  rayDr  ray:Dr  laDr  la:Dr
Phonetic Representation  rayDr  ray:Dr  laDr  la:Dr

Once again, the level of perception can be found at the morphological level of underling forms. The fact that there are minimal pair contrasts at the level of phonetic representation is irrelevant, as this is the level of sound production and not one of sound perception. Native speakers of English do not perceive differences between long and short vowels are phonemic. Furthermore, they perceive the underlying /t/ of "latter" and the underling /d/ of "ladder" even thought they both emerge in the form of a flap r [D] at the level of phonetic representation.



Linguists who attempted to understand phonological processes within the new system of generative grammar asked the question of what is the basic unit of sound. The problem occurred when linguists attempts to account for tone languages, pitch languages, intonational phrases in language, trisyllabic root languages, and stress patterns. At this time, they began to realize that generative phonology could be characterized by a preoccupation with the rule system and that other kinds of phonological information merited equal discussion. As they looked back at Classical Phonemics, they realized that the idea of having a level of phonology separate from one of morphology was important and that it should be reinstated. Instead of levels of analysis, they called these STRATA or DOMAINS.

Kahn (1976) broke with the tradition by arguing for the importance of syllable structure in phonology and how syllables were important for an adequate description of English phonology. He also addressed the issues of tone and stress and how they relate to syllable structure. Syllables are structured. Each syllable has a nucleus that is usually a vowel. The other elements in a syllable are non-syllabic. The internal structure of the syllable is provided below:


The nucleus, symbolized by (N), is the only obligatory member of the syllable. It forms the core of the syllable. The onset, symbolized by (O), is made up of elements that precede the nucleus, and the coda, symbolized by (C), is made up of elements that follow the nucleus. Finally, the Rhyme, symbolized by (R) is important for designating rhythmical patterns in language.

The old notion of FOOT, important to traditional phonemics, was re-instituted under the revised model of metrical phonology (Lieberman, 1975). The new system allowed linguists to explain how two words echo one another by means of rhyme, assonance, consonance, and alliteration.

Rhyme has to do with the relationship between words with different consonants immediately preceding the final accented vowels and identical sounds, for example "pillow/willow" and "go/know".

Assonance is the relationship between words with different consonants immediately preceding the and following the last vowel. The vowels have identical sounds. For example "hit/will" and ""disturb/bird.""

Consonance is the relationship between words whose final accented vowel sounds are different but with the same consonant frame. For example, "truck/trick" and "billion/bullion.

Alliteration is the relationship between words with identical consonants preceding the first accented vowel and possibly different sounds on that vowel in subsequent syllables. For example, "slip/slide" and "glowing/glare.

In syllabic terms, these can be summarized as follows:

 Name  Example  Onset  Nucleus  Coda
 Rhyme  Pillow, willow  Different  Same  Same
 Assonance  Hit, will  Different  Same  Different
 Consonance  Lip, lap  Same  Different  Same
 Alliteration  Slip, slow  Same  Different  Different

Since syllables had to be distinguished from phonology and morphology, Goldsmith (1976) argued that a new field of Autosegmental Phonology was needed. He referred to the old system characterized by the Sound Patterns of English (Chomsky and Halle, 1968) as Segmental Phonology. The new approach is non-linear. It involves moving between different strata within a grammatical system. Such a model is multi-tiered.


This multi-tiered approach to autosegmented phonology allows one to explain compensatory lengthening. When the voiceless velar fricative is lost, the syllabic structure along the skeletal tier is reassigned to the vowel, making it a long vowel. It also allows one to explain long monophthong vowels and how they differ from diphthongs.


Similarly, geminates can be distinguished from clusters of consonants.


Assimilation can be explained by delinking and spreading rules as when the cluster "kt" becomes assimilated to a geminate or long "tt."

The last change to be made in metrical phonology came from Siegel (1974) who wanted to specify the interaction between phonological rules and other modules in the grammar of a language. His research led him to recognize several strata beginning with the morphological module and its relationship to the phonological one.


Under this new framework, morphological representations are treated differently from phonemic ones. This is reminiscent of the levels of analysis within classical phonemics.


Morphophonemic representation  t  
 Phonemic representations  s  t
 Phonetic representation    th

Velar Softening and Spirantization are morphological processes. They must be dealt with within the morphological module of the language. It accounts for the following English alternations.


One of the first findings to emerge within this new framework was a resolution to the problems of affixation. Why does velar softening occur before some affixes and not others. For example, panic, panicking, flight, flighty, might, mightily, but not *panisy, *flighsy, nor *mighsily. The answer has to do with what Mohanan Class I and Class II affixes. He designates these in accordance to the modules in which they occur. Class I affixation operates in the Lexical Module; Class II affixes operate in the Postlexical Module. An example of this can be found in the rule of trisyllabic shortening that operates only in the Lexical Module and is consequently considered to be a morphological rule.


The rule of Trisyllabic Shortening accounts for the following words in English. An exception to this rule is "obscene / obscenity.

 vain vanity
serene serenity
divine divinity

"What is important about the morphological component or the lexical module is that it enables linguists to distinguish between word phonology and phrase phonology.

This similarity between phrase syntax and word syntax has led certain linguists to argue that both systems can be handled by the same principles. Notice that the shift in recent times is away from rules and how they are formed to principles and the parameters that surround rules. Hence, the concept of the head of a phrase in syntax has been applied to phonological structures by many within lexical morphology.


 N NP  The Noun is the functional head of a Noun Phrase
 V VP  The Verb is the functional head of a Verb Phrase
 A AP  The Adjective is the functional head of an Adjectival Phrase
 P PP  The Preposition is the functional head of a Prepositional Phrase
 ADV ADVP   The Adverb is the functional head of the Adverbial Phrase
 Q QP  The Quantifier is the functional head of a Quantifier Phrase

From this pattern of head constructions, linguists were able to generalize that X is the head of XP.


 X XP  X is the functional head of the X Phrase

This new way of categorizing phrase markers allows linguists to create three levels of categories: X, X-bar, and X-Double Bar. 

 X-Double Bar  X-Bar  X
 NP  N-bar  N
 VP  V-bar   V
 AP  A-Bar  A
 PP  P-Bar  P
 QP  Q-Bar  Q

Given this framework, linguists naturally asked is word phrases have head constructions. Selkirk (1980) has argued that the word structure follows the same system of principles and parameters as one finds in government binding theory. He employs X-bar theory in order to explain word syntax. Notice that the negative superscripts represent morphology and that the positive ones represent syntax.

Within this multi-tiered approach, lexical phonologists developed several domains or tiers within phonological structure.


 The Prosodic Hierarchy
   Phonological Phrase
 Phonological Word
 Syllabic Constituents
 Skeletal Positions


Mohanan (1986) argued that one does not need to invoke X-bar theory in order to account for morphological and phonological information. He devised a simpler system with just two tiers which he called Stratum I (Class I affixation) and Stratum II ( Class II affixation. Harris (1994) refers to these as Root Morphology and Word-Level Morphology respectively.

Siegel (1974) was the first to notice the behavior of Class I and Class II affixes. He noted that both classes are adjective and noun forming, but a class I affix cannot be attached to a stem containing a class II affix.


 Acceptable Affixed Form  Non-acceptable Affixed Form  Coimmentary
 Bountifulness   *boundtfullity  The suffix <ity> cannot be affixed the morph {bound}
 Guardedness   *guardedity  The suffix <ity> cannot be affixed to the morph {guarded}.

Hence, nouns can be formed with Class I or Class II affixes, but these affixes must be ordered in their units of operation. It is important to notice that in the older linear phonology, rules were defined by their operations. Morphological rules operated on morphological elements and phonological rules operated on phonetic elements. Under the new system, rules apply to strata. The same rule can apply to Stratum I or II. The language dictates how rules are assigned and to what domains they belong. Hence, in the lexicon, -ity is designated as a Stratum I ending and -ness is designated as a Stratum II ending.

The suffix -ity cannot be attached to "beautiful" because this word is formed at Stratum II. At this point, the -ity affixation is no longer available. In order to affix -ity to this word, one must go back to Stratum I and this is disallowed. Once a form leaves Stratum I and enters Stratum II, it cannot return to this domain. What is important about these levels is that they distinguish affixation from compounding, something that was not done in the earlier work of Chomsky and Halle (1968). Hence, "blackbird, redcap" and other endocentric compounds are found in Stratum I, whereas Exocentric compounds such as "maidenhood" are found in Stratum II.. It also designates that only in Stratum I can one find bound morphemes such as ""inert, intrepid, immaculate, insipid, and "cranberry."

There are several words in which the voiced velar stop [g] appears in some form root forms and are deleted in others.


 [n]  [n]  [gn]
 sign signing
 resign resigned resignation
malign maligning malignant
design designing designate
assign assignment assignation
paradigm   paradigmatic

The same kind of alternation can be found in alveolar nasals as evidenced by the following words.


 [m] [m] [mn]
 solemn   solemnity
damn damming damning damnation
hymn hymned
 column columns columnar
 dendemn condemning condemnation

Finally, there are words in which [mb] clusters are reduced to [m] after certain suffixes.


 [m] [m] [mn]
 bomb bomber
 crumb crummy crumble

Why do all of these forms undergo consonantal alternations? The answer can be found in the kind of suffixes that are attached to them. Mohanan (1986: 22) arrives at the following rule for g-deletion (sign, signature) and n-deletion (hymn, hymnal)


Mohanan (1986:26) argues that there is evidence for at least Four Strata in English. This is based on the fact that a third stratum is needed for compounds, and a fourth is where regular inflection occurs.

 Stratum I Class I derivation, irregular inflection
 Stratum II  Class II derivation
 Stratum III  Compounding
 Stratum IV  Regular inflection

Mohanan discovered a major problem in the Sound Patterns of English. He noted that Chomsky and Halle did not distinguish between vowel tensing and vowel lengthening. It was assumed that all tense vowels were long. Such is not the case. There are tense vowels that are long as in keep, reed, site"" and tense vowels that are short, as in the final vowel of "city." Hence, he found it necessary to discuss the rule of Stem Final tensing. This is the rule that stresses the final vowel of "city. This is a rule that tenses a non-low final vowel. Hence, it accounts for the final vowel of "city" [sIti] being tenser than the medial vowel.

What is happening in these forms is that the lax vowel [I] becomes tense [i] before Class II affixes and before another stem in a compound word. At first, Mohanan (1986:26) believed that the rule should be assigned to Stratum II. However, in investigating how that rule operates in compounds, he found it necessary to assign it to Stratum III, the level at which compounds are formed.

Stratum III is where compound nouns are form and Stratum IV is where inflectional endings are attached.


When words are put together, they may appear as compounds that are organized on Stratum III or they may be concatenated elsewhere such as in Stratum I. The following words provide evidence for levels of compounding in Stratum I.

 Vowel Shortening

 Receive   reception
 Perceive  perception
 Describe   description
 Reduce  reduction
 Scribe  scripture
 Retain  retentive
 Intervene  intervention
 Five  fifty
 Wise  wisdom
 Keep  kept
Deep   depth
 Wide  width
 Sleep  slept
 Deal  dealt
 Leave  left
 Thief  theft
 Heal  health
 Steal  stealth

Are these forms listed as complete units in the lexicon or are they derivable by means of a phonological or morphological rule? What do all of these forms have in common is that they undergo the process of vowel shortening. The next question is where do these rule applications occur. By looking at the affixes attached to these root forms, one can readily ascertain that they are Class I affixes. Hence, the rule of vowel shortening must take place in Stratum I.


Mohanan (1986: 47) uses the metaphor of the word-making factory to explain how strata are organized within a grammar. The factory has a storehouse in which all morphemes are listed. There is an entry gate from the storehouse to the factory work area. Thee is also an exit gate from which the finished products emerge. Within the facttory there are four rooms, these domains are called Strata I-IV. There is a conveyer belt that runs from the entry gate to the exit gate and it passes through each of the four rooms. The forms from room I automatically goes on to room II and so on. There are two types of workers in the factory. One group works on morphological rules and the other works on phonological rules (the Lexical Domain and the Post-Lexical Domain). Mohanan (1986: 48-49) found it necessary to include a loop from room III back to room II for certain lexical items. This loop is optional and accounts for variations with in words that undergo palatalization and yod deletion.

In order for these dialects to function, the lexical items must be recycled through the second stratum. The concept of the loop is still controversial in English metric phonology and not all languages participate in loop constructions.


One of the realizations of the new model of lexical phonology is that both morphological and phonological rules function as redundancy rules (Aronof, 1976). Their purpose is not to generate an output, but to maintain a distinction between potential words (lexical outputs) and actual words (word lists). The phonology of some words in the lexicon are completely transparent and others are rather opaque. For example, "went" is listed in the lexicon, but it is derived from the specification of [go+PAST]. Words such as "student union" and "government center" are exceptional in that the major stress is on the second compound. Hence, such forms are relegated to word lists in the lexicon. What is interesting about the new phonology is that no two speakers have the same lexicon. This is because generative grammar is a model of the knowledge of the language of a speaker and no two speakers have the same knowledge. A word may exist in the mind of the speaker as a listed item. Once it has been subjected to morphological analysis and rule behavior, it will be relegated to a lexical output. Consider, for example, the word "breakfast" which is semantically opaque for most speakers. Once an individual understands its etymology and sees it as coming from "to break the fast with food" then the meaning of the word is no long opaque.

The productivity of the strata in the lexicon is another aspect to consider when dealing with the mental representation of lexical entries. Mohanand (1986: 56) has derived a hierarchy of productivity base on the functions of production, recognition, and lexical storage.

A. Rules, which are used only in production and recognition, need not be stored and they constitute highly productive rules. Hence, aspiration, flapping, nasalization, and rhota deletion are productive rules. The rule of plural suffixation (-es) and past tense formation (-ed) belong to this category.

B. Rules which are used both in production, recognition, and storage are less productive. Vowel shortening, vowel lengthening, Vowel shift-diphthongization, g-deletion and n-deletion belongs to this category.

C. Rules that are used only in storage and recognition, but not in production are even less productive. The rules of ablaut (man/men, foot/feet) and the rule of velar softening belong to this category.

D. Rules which are used only in storage and not in production nor in recognition are the least productive. Many rules having to do with syllabic constraints (CCC onset) belong to this category. Similarly prefixes such as con- and per- are totally unproductive.

When one considers the four strata in English, it turns out that Stratum I is the least productive.

Inept, classic, grammarian, polity, illusionary, destruction, generate, dialectal, informant. warmth…

Stratum II is more productive.

Unkind, dislike, kingdom, kindness, statehood, friendship, beautiful, argument, orderly, takes, eaten, walked, and compounds.


There are many examples of morphological restructures undergoing re-syllabification at the syllable level of metrical phonology. An obvious example is provided below from Spanish.


 Morpheme Level:   l+o+s animale+s
 Syllable Level:
 Word Level  : los animales

Words may undergo syllabification when a high front vowel is transformed into a front glide (yod) before another vowel.


 Morpheme Level  : iat
 Syllable Level:   yat "yacht

"The more interesting cases of re-syllabification, however, can be found when nasal and liquid consonants alternate become syllabic under certain conditions within a syllable structure.


 Morpheme Level:   botm
 Syllable Level:  bot.m "bottom

"This property of syllabification, however, is not limited to liquids and nasals. They can also occur when a fricative functions as the peak of a syllable: /pst/
What appears to be at issue in these examples is the property of sonority. The nucleus of a syllable is the most sonorant sound in that syllable. Hence, the following sonority scale can be ascertained (Giegerich, 1992: 133):

What this chart signifies is that the most sonorous element in a segment functions as its syllabic peak or nucleus. The voiceless fricative /s/ functions as a syllabic peak in the vocative "pst." Hence, the sonority of a sound is based on its relative loudness in comparison to other sounds in its syllabic environment. These syllabic environment, it should be noted, are marked by morphological boundaries. If they were not, one would not be able to account for the differences in syllabic structure between the following lexical items:

 Hidden aims
 Hidden names

As noted earlier in this synopsis of metrical phonology, the syllable is structured and consists of an onset and a rhyme. The rhyme is further structured into a nucleus or peak and a coda. Each of these structures accounts for phonotactic asymmetries and each has its own phonotactic domain.

Many lexical items do not have onsets (eye, eat, ink). Others do contain an onset (pie, sleeve, swing, spring, spy, street). Phonologists have noted that there are many possible combinations of onsets in English, but they do not all appear in the language. The onset of /bn/, for example, is possible in English, but it does not appear. This leads them to hypothesize that the onsets in English are constrained so as to omit certain combinations of sounds. They do this by establishing a syllable template for English and in that template they build in filters that account for the onset patterns of English.

Each language will have its own syllable template to account for its onset phonotactics. An interesting situation occurs between onsets and the nucleus of the rhyme in words such as "view /vju/. The yod (front glide) may have originated in the rhyme of the syllable, but it is quickly associated with both the onset and the peak of the syllable.

The process of re-syllabification means that the front glide /j/ must belong to both the onset phonotactics and the rhyme phonotactics. Hence, the following filter is needed for English syllable structure.

It is now time to consider the Coda and how it functions alongside the nucleus of a syllable The ending of syllables in English can be /nt/ "lent"," /nd/ "land," /mt/ "dreamt,"/md/ "hummed," and /õd/ "hanged." Other combinations are not allowed. Hence, the template for the coda is provided below:


This syllabic structure accounts for "out" and "ant."

The final syllable template for English is provided below:


Not all words in English are monosyllabic. Hence, one needs to consider polysyllabic words and how they are re-syllabified.  at.las Ag.nes  hem.lock  de.cath.lon co.di.fy
 Al.ti.tude   a.pri.cot  ma.tron
 a.gen.da   stan.dard  nigh.tin.gale

Phonemic representation


Syllable Boundaries

  [at las]

 Syllabic Structure:  

The following words provide an interesting problem for the syllabic structure of the coda. They are ambisyllabic.


 Apple  pedestal  madonna  Petrol  camera
 Rubella  epic  labrador  confetti  Africa

How does one syllabify apple? Is it app.le or a.pple? The first syllable in "apple" should be [ap] because of the branching rule. The second syllable would contain the syllabic liquid [l]. Hence, the syllabification should be [ap][l]. The problem with this conclusion is that the syllabification of this word is ambiguous. The [p] in apple is ambisyllabic. A similar ambisyllabic lexical item can be found in petrol, i.e., pet.trol.


The cluster of consonants in the coda of a syllable presents a problem for metrical phonology. The kinds of clusters in a word final coda differs substantially from those that occur within a word. This is because morphological endings such as the past tense endings /t, d/ complicate the structure of final codas.


   VV  day, so, see, pie
   VC   pat, tip, den, gull
 Three-Position Final Coda  VVC  Late, light, town, feat
   VCC   fist, fact, desk, gulp
 Four-Position Final Coda  VVCC   paste, paint, rind
   VCCC   text

The problem of syllabification occurs with suffixation at the word level. The plural marker makes a VCCC [text] into a VCCC-C [text-s] structure. These final consonants resulting from suffixation fail to observe the expected final coda pattern. Hence, it has been proposed that these forms are extra-syllabic. What this means is that they remain unsyllabified until they are licensed at some other level within metrical phonology. Some have referred to this unassigned final coda as a degenerate syllable. It is degenerate because it lacks an audible nucleus. Hence, the final coda has its own phonotactic domain. It has its own phonotactic constraints. It differs from word internal coda in that it has to delay syllabic assignment until later in the phonology of the language where all morphological suffixes have been added and word final phonotactics are assigned.

The final /t/ is "mist" represents a stray onset because it is not followed by a nucleus. Some of the differences in syllabic assignment of codas can be seen in the following examples:

 Structure Final Medial
 VVnt   saint, mount, pint  fountain
 VVnd   rind, sound  flounder
 VVns  pounce, ounce council
 VVnd  ñ range, scrounge  angel
 VVlt   revolt, colt  poultry
 VVld   child, cold, field  shoulder
 VVst  paste, boast, boost  pastry
 (V)VsC   cast, clasp, cask  basket


The phonological template for a language such as Arabic would differ substantially from English because it contains tri-consonantal roots.

 TKR  (taraka, to leave
 FHM  fahima, to understand
 RKB  rakiba, to ride
 NSR  nasara , to help

These tri-consonantal roots are used as the underlying structures for other phonological processes such as derivation where RKB produces the infinitive rakiba (to ride) and the agentive form rakib (rider). Similarly, the root KTB is used to produce the nominal forms kitab (book) and katib (writer). Obviously, what is happening in this language is that the vowels of the language are intercalated among the tri-syllabic roots. This process differs substantially from the phonological pattern of English where a root form merely takes on a prefix or a suffix. In Arabic, the affixal morphology consists of intercalated vowels. Hence, the template for the Arabic root KTB consists of tri-syllabic consonants and intercalated vowels that are unspecified.

The V slots have been left vacant in the Arabic word template because they will be assigned values later. Such is the case in the following example where the root is KTB and the vowel melody is /a/.

The use of abstract underlying forms that are filled later in the phonological derivation is reminiscent of Trubetskoy's concept of the archiphoneme. His concept was used to explain vowel neutralization in word final position among German nouns.

 Tag [tak] day   Tage [tage] days
 Bund [bunt] alliance  Bunde [bunde] alliances
 bunt [bunt] colorful   bunte [bunte] colorful

In structural phonemics, phonemes were contrasted as opposites. Structuralists believed in the adage that ""once a phoneme, always a phoneme. What they meant by this is that phonemes were not allowed to overlap. The allophone [t] is assigned to the phoneme /t/ and the allophone [d] is assigned to the phoneme /d/ and they can never overlap. But what one finds in German is just such a case of phonemic overlapping.

To account for this problem of phonemic overlapping, Troubetskoy created the concept of the archiphoneme {D}. This element existed at the morphological level and had two variants at the phonemic level. Because this form existed at the level of morphology, it did not violate the restriction on phonemic overlapping. What was different about the archiphoneme and how did it avoid the problem of phonemic overlapping? It was different because it underwent a morphological rule of devoicing.


 Morphological Level  bunt  bunte  bunD  bunDe
 Word Final Devoicing  -  -  bunt  -
 Phonemic Level  bunt  bunte  bunt  bunde

One could now handle this same problem in metrical phonology by saying that the morphological archiphoneme {D} is unspecified for the distinctive feature of voicing. It is assigned the value of [-voiced] in word final position by the word final coda template.


The level above the syllable is the foot. This is an important timing unit in English. There are two kinds of syllables: strong (S) and weak (W). Stressed syllables must be heavy (S) while unstressed syllables may be light (W).

One could also represent this structure as follows:

There is a process known as foot assignment. (1) Assign a foot to the final syllable of a word if it contains a long vowel or if it is heavy. (2) Assign a bisyllabic foot to the penultimate syllable if it is heavy. (3) Assign a foot to the penultimate syllable if it is heavy and initial. (4). Assign a maximal bisyllabic or trisyllabic foot to any remaining string of syllables from right to left.


There is a whole tradition in poetry dedicated to the study of the rhythms of speech. Within this tradition, each line of poetry is measured and assigned a foot structure. Students of poetic literature even have their own prosodic symbols.

   An unaccented syllable, W syllable
 /  An accented syllable, S syllable
 X  A non-rhyming line

These poets looked for patterns of assonance, consonance, alliteration, and rhyme. For example, alliteration occurs when the onsets of a syllable are the same as in slip and slide.  

  Poetic Structures

 Name  Example  Onset Nucleus  Coda
 Rhyme  Pillow, willow  Different Same Same
 Assonance  Hit, will  Different Same Different
 Consonance  Lip, lap  Same Different Same
 Alliteration  Slip, slow  Same Different Different

Poetic structures have names. Many of these names were borrowed from Greek poetry and still reflect these original structures (Miller, 1986).


 Iamb   ( /  Source unknown
 Pyrrhus  ( ( .  From Ancient Greek war dance invented by Phyrrhus
 Spondee   / /  This is the Greek word for a solemn toast. Trochee / ( From the Greek for "to run trippingly"
 Änapest   ( ( /  From the Greek for "to hit back
 Dactyl  / ( (  From Latin for "finger," suggesting three joints
 Ionic Major   / / ( (   A spondee followed by a pyrrhus
 Iambic minor   ( ( / /   A pyrrhus followed by a spondee
 Bacchic foot   ( / /  After a dominant beat in the odes to Bacchus
 Antibachhic foot   / / (  The reverse of the Bacchic foot.

Literary scholars went on to define traditional stanza patterns. The short couplet, for example, consists of two rhymed lines in iambic or trochaic tetrameter (a four beat measure made up of trochee). This short couplet could also be made up of an iambic tetrameter ( ( / ).


( / ( / ( / ( / a
( / ( / ( / ( / a

Watching the birds, I think of Bach
Each of the distant wheeling flock

The split couplet is a version of the short couplet and consists of two rhymed lines one of them is made up of five stresses (pentameter) and the other consists only of two stresses.


( / ( / ( / ( / ( / a
( / ( / a

A book is coming out I wrote somehow
I could not now
(John Stone)

The heroic couplet consists of two rhymed lines of five stresses each (pentameter). It is made up traditionally of iambs and often with a full stop at the end of each stanza (Thompson, 1989).


( / ( / ( / ( / ( / a
( / ( / ( / ( / ( / a

Vice is a monster of so frightful mien
As, to be hated, needs but to be seen
(Alexander Pope)

There are many variations on English meter. One of them is the quatrain, the preferred form for hymns. It is usually iambic with alternating rhymes.


( / ( / ( / ( / a
( / ( / b
( / ( / ( / ( / a
( / ( / b

Amazing Grace, How Sweet the Sound
That Saved a wretch like me;
I once was lost, but now am found,
Was blind but now I see
(John Newton)


This concludes a brief introduction into lexical phonology. Many new texts are appearing in which this analysis is applied to other languages. This summation will assist you in understanding the basic concepts of lexical phonology.




 This word means "to make similar to". It comes from Latin ad (to) + simulare (be similar). Assimilation can take on many forms, but they are generally classified in terms of the direction of influence.

:Progressive assimilation, for example, happens when the process of assimilation is from left to right. The reverse is called regressive assimilation.
Progressive Place Assimilation
: Latin nokte notte (night)
Japanese yom+da yon+da
Regressive Voicing Assimilation: Japanese yom+ta yom+da

 Degemmination:  When a double consonant (a gemminate) is made into a single consonant, it undergoes degemmination. Portuguese general becomes geeral due to nasal loss. The long /e/ is rendered into a short vowel by the process of degemmination.
 Dissimulation:   This is when two similar sounds undergo change and become more dissimilar. German Tartoffeln (Potato) Kartoffel (k/t); Spanish colonel English Colonel (l/r)
 Coalescence or Merger:   This is when two sounds merge into one. Latin falta French faute (au merges into o), Latin casa+i Italian case (ai merges or coalesces into e)
 Phonemic Split:  This is when the allophones of a phoneme become separate phonemes. Old English had the phoneme /z/ had two allophones. The voiced interdental allphone [ x ] occurs intervocalically, and the voiceless interdental fricative [z] occurs elsewhere. In middle English these allophones contrasted intervocalically and emerged as separate phonemes in English.
 Compensatory Lengthening:  Long vowels are lengthened before a nasal consonant in many Indo-European languages. What is interesting about this process is that it is often accompanied by the loss of the nasal consonant. Many phonologists believe that the lengthening of the vowel is there to compensate for the loss of the nasal and they have named this process compensatory lengthening. Latin cas+a+ns cas+aa+s; Old English ham hoom "home".
 Haplology  This is the loss of a repeated sound. English engla (people) + lond (land) England; haplology occurs lexically in English because repeated root forms are avoided. Hence, one does not say "the builder built the building" but "the builder constructed the edifice.
 Ëpenthesis or Prothesis:   This is the introduction of a sound into a word: Latin scuola Spanish escuela. Epenthesis is usually used for the insertion of an extra vowel intermedially.
 Metathesis  : When two sounds exchange place, this is called metathesis. Old English dryd English third, cf. German dritte. English ask Dialectal English aks
 Apocope   This is when a final vowel is lost. Spanish has /flore/ which emerges as "flor"in the singular and "flores"in the plural. The loss of the final [e] in the singular form is called apocope.
 Syllabic Fortification:   Syllables are strengthened by the insertion of a consonant. Greek cassanra was borrowed into English as Cassandra. Latin homine became homre and was fortified as hombre in Spanish.
 Vowel Loss:  The thematic vowel in Spanish is lost before another vowel. Hence, Spanish am+a+o becomes amo.
 Vowel Insertion:  In Japanese a vowel is inserted between consonants because the rhythmical CV pattern must be maintained. Hence, Japanese yom+mas+ru yom+i+mas+ru.
 Consonant Loss  Sometimes a consonant must be deleted in order to maintain a syllabic pattern. In Japanese the 'r' consonant is lost in order to maintain a CV pattern. Hence, des+ru des+u. Another form of consonant loss occurs in French. Le petit garHon le peti garHon.
 Umlaut:   In the Germanic languages vowels underwent changes in vowel height or vowel fronting because of the influence of a high front vowel in the following syllable. German hand+e becomes hend+e "hands." Old English gans undergoes compensatory lengthening and appears as goos in Middle English. The plural of this form is goos+iz which quickly became goos+i. This high front vowel caused the long /o/ to umlaut into long /`/ which later was unrounded and emerged as long /e/. At this time the plural form goose became geese.
 Ablaut  : This is a grammatical term that signifies a change in tense by means of a vowel change. Hence, the present tense of English "drink" undergoes an ablaut in the past tense form "drank."
 Rhotacism  This is when /s/ changes into /r/. Rhota is the Greek word for the letter /r/. Latin tempus (time) becomes tempora in the plural.
 Metanalysis  This is when a word is derived from another word through error. Old English a napron became an apron. A naddre became an adder. Modern English Asia and Cuba become Asiar and Cubar (President Kennedy).


Aronoff, M. 1976 Word Formation in Generative Grammar. Cambridge, Mass.: MIT Press.

Bohas, Goerges. 1997 Matrices, Étymons, Racines: Ëlements d'une théorie lexicologique du vocabulaire arabe. Leuven et Paris: Peeters.

Carr, Philip. 1993 Phonology. NY: St. Martin's Press.

Chomsky, Noam. 1970 "Remarks on Nominalization," in Studies in Generative Grammar.Mouton: The Hague.

Chomsky, Noam and Morris Halle. 1968 Sound Patterns of English. NY: Harper and Row

Giegerich, Heinz. 1992 English Phonology: An Introduction. Cambridge: Cambridge University Press.

Gegerich, Heinz. 1995 Metrical Phonology and Phonological Structure: German and English. Cambridge, London, New York, New Rochelle, Melbourrne, and Sidney: Cambridge University Press.

Goldsmith, J. 1979 Autosegmental Phonology. New York: Garland Press.

Harris, John. 1994 English Sound Structure. London and New York: Basil Blackwells.

Kahn, D. 1976 Syllable Based Generalization in Enlgish Phonology. PH.D. dissertation, MIT., distributed by Indiana University Linguistics Club, Bloomington, Indiana.

Lieber, Rochelle. 1987 An Integrated Theory of Autosegmental Processes. Albany, NY: State University of New York Press.

Lieberman, M. 1975 he Intonational System of English. NY: Garland Press.

Mohanan, K. P. 1986 The Theory of Lexical Phonology. Dordrecht./Boston/Lancaster/Tokyo: D. Reidel Publishing Company.

Pesetsky, D. 1979 The Phonology and Phonetics of English Intonation. Cambridge, MA.: MIT Press.

Pulleybank, D.1983 Tone in Lexical Phonology. Dordrecht: Reidel.

Roca, Iggy. 1994 Generative Phonology. London and New York: Routledge

Selkirk, E.. 1980 The Phrase Phonology of English and French. NY: Garland Publishing Inc.

Selkir, E. 1983. The Syntax of Words. Cambridge, MA.: MIT Press.

Siegel, D. 1974 Topics in English Morphology. Unpublished Ph.D. dissertation, MIT.

Stampe, David. 1979 A Dissertation in Natural Phonology. NY: Garland Press.

Strauss, S. 1982 Lexical Phonology. Dordrecht: Foris

Thompson, John. 1989 The Founding of English Metre. NY: Columbia University Press.

Williams, Miller. 1986 Patterns of Poetry: An Encyclopedia of Forms. Baton Rouse: Louisiana State University.