The Lexical Phonology and Morphology Model

Incorporates strata from Strata theory

Levels in theory.

Lexical Strata

• neutral

  Have no phonological effect on the base to which they are adjoined.

  -ness: kindness, homliness, laziness, redness, hopelessness, eagerness, seriousness, productiveness.

  A --> N. productive with AS roots, unproductive with Latinate stems: *pregnantness

  -less: hopeless, carless, homeless, headless, treeless, idealess, legless, careless, beardless,

  N --> A, productive with AS roots, unproductive with Latinate stems. *developmentless.

• nonneutral

  Rule: The stress that is assigned to verbstems cannot go past the root.

  Have a phonological effect on the base to which they are adjoined.

  -ic: 'idyll, i'dyll+ic; 'phoneme, pho'nemic.

  the penultimate syllable of the derived word is stressed.

  -ee: refugee, employee, governee, abstentee, divorcee.

  '-ee' is inherently stressed. This means that stress is assigned to the morpheme, usurping the rule that otherwise would assign stress to the stem. For some speakers (Americans?) the stress doesn't always go on the -ee. It may go on the preceding syllable.

• -th: affects the vowel of the adjoinee (the stem)

• wide, width; broad, breadth; long, length; deep, depth; weal (obsolete), wealth, heal, health;
• Note: the suffix -th that is added to numerals, is productive and with one ora two exceptions does not affect the quality of the vowel.

• geminates: they never occur within a morpheme or across so-called primary (nonneuteral) affixes

• innate, immaterial, irregular, illegal, attest, commit are "primary."
• unnatural, unneighborly; subbase, subbituminous (referring to a lower rank coal); override, overripe, overrun; missend, misstate "secondary."
• The "secondary" are true morphemes, the "primary" affixes are pseudo-morphemes or base-extenders.

• gemination and lack of Latinate-assimilation occur in secondary affixes.
• Why should these, the true affixes, be considered secondary? They should be primary.
• the base-extenders occur first. They should; they are not true morphemes and it is questionable whether they are even base-extenders.
• We will continue to mark these with '+': the separator between true morphemes.
• un+nat-ur-al, un+neighbor+ly, sub+base, over+ride, over+run, mis+send, mis+state; and so forth.

 This does leave a problem the "-th."

It seems to be a suffix: width, length, breadth, wealth, health, depth, strength.

height, weight, heat: wright, blight, plight.

But it could be a base-extender,

This suffix is unproductive.

heigh-t, *tallth or*talth, *narrowth, *fatth, *shortth, *heavyth, and so forth.

There is a second suffix: it definitely is productive with ordinals: four+th, fif+th, six+th, zero+th, i+th.

Here '+th' is productive.

For some reason infinitive does not take +th: *infinitiveth.

It occurs in the context with certain vowel alternations mentioned above and in fif+th. This is irregular.

These alternations are Middle/Early modern English.

Morphophonemic alternations are valid with true morphemes.

The following morphophonemic sets are valid English ones; they occurred in the late Middle English or early Modern English period:

{aj, î}, {i, œ}. {e, æ}, {æw, &Mac215;}, {o, a}

bite, bit; flee, fled; sane, sanity; profound, profundity; morose, morosity;

These alternations are valid with true morphemes.

The alternations have little to nothing to do with primary and secondary affixes.

The base-extending suffix '-ity' has some rather interesting properties

Antepenultimate stress is assigned through it: seren'dip-ity, ca'lam-ity,

'proxim-al, pro'xim+ity; sagac-ious, sa'gac+ity. (suffix deletion).

in-equ-'al+ity, stu'p-id+ity; de'pendable, depend-i'bil+ity (no suffix deletion; here 'ity' looks more like a bona fide suffix).

Trisyllabic Laxing

A historical sound change that occurred in late Middle English or early Modern English, probably the former.

It leaves in its wake the morphophonemic sets mentioned above {aj, î}.

sane, sanity; profound, profundity;

/sa{e,æ}n/;

select /æ/ / ___ îti (and a few other contexts);

sæn+îti/.

The suffix '-ian'

The stress falls on the antepenultimate syllable

'custody, cu'stodian; 'comedy, co'median; fal'lopian; 'Croat, Cro'atian; 'Russia, 'Russian.

'ian' is a level 1 suffix since it affects stress.

Sometimes a true suffix, sometimes a stem extender.

It cannot be added to Anglo-Saxon roots

*bookian, *glassian, *womanian, *cupian, *fingerian, *headian.

The true suffix '+like' can be added to '-ian':

cust'odianlike, co'medianlike, Cro'atianlike, 'Russianlike.

Probably best to consider 'ian' a base extender since it has more properties of base-extenders with the occasional exception that it inherents meaning.

The suffix '-ive'

stress on the antepenultimate syllable if lax and possible, on the penultimate syllable if it is tense or the penultimate syllable is a root:

pro'gress, pro'gressive (root is 'gress') = pro-gress-ive;

causes the lax vowel variant of the {i, œ} class mentioned above (the so-called trisyllabic laxing rule).

r&Mac246;'pit, r&Mac246;'pœt&Mac246;tiv; com'pete, com'petitive {i, œ}

+y, (both as a noun and as an adjectrive)

+ful, (N -> A)

+wise, (N -> A)

Level 2 (stratum 2, neutral)

Anglo-Saxon suffix (level 2) '-ness' added to it with no effect on the phonological form of the stem:

pro'gressiveness, com'petitiveness, re'petitiveness.

competitivernessless.

Exceptions are typicall of level 1 or base-extenders.

type 2 or true suffixes

+ly, +ness, +less, +er, +like, +ing,

going canoeing, sailing, yatching, and so forth

(ceiling is one morpheme)

The suffix -ling is type 1, not type 2:

-ling (unproductive)

duck-ling, star-ling, gos-ling (goose), *hen-ling, *bird-ling, *crow-ling, *sparrow-ling.

Note: gos-ling. This illustrates a vowel alternation typical of type 1 affixes: goose.

It does not affect the stress of the stem but may cause a phonogical alternation: goose, gosling.

Sometimes the unproductive affixes do not inherit meaning: build+ing.

Perhaps 'building' should be considered a single morpheme, or perhaps the suffix can weaken to a base-extender. This seems better.

Verbal inflection

Irregular verbs could be seen as affected by level 1 suffixes

swim, swam, swum; there is no affix here.

buy, bought; sell, sold; there is an affix here. Is it level 1? We will return to this. It could be since it affects morphophonemic alternations in the base.

Nominal Inflection

Irregular nouns:

tooth, teeth; no affix.

child, child-r+en; brother, brethr+en.

-r- is definitely a stem extender in children.

but not in brother.

is '-en'? Hard to determine since it occurs only three times.

What triggers the alternation in 'brethren'? The suffix 'en'?

Try the grammatical feature [+Pl] for the context of these variant forms. It works.

child+[+Pl] -->

child-r+[+Pl] -->

child-r+en

(preferable)

Or

child+[+Pl] -->:

child+en -->

child-r+en

&Mac176;ajld ---> &Mac176;îld-r-&Mac246;n

{CH{î, aj} LD, Stem, 'rather young adults'}

The above is the three component analysis.

In this analysis the suffix '+en' would have the following entry in the grammaticon (the subcomponent of the lexicon where grammatical forms are stored):

[+Pl]	grammatical feature
+split +host N ____	form features
/n/; /z/ (default)	phonological form

2005 - STOP HERE.

Lexical Rules

Morphological rules are dependent on

the class of affected bases.

the affix that is attached.

linear order: prefix, root, suffix.

the class of the adjoinee.

the stratum to which the affix belongs. [debatable]

Morphological rules that affixes to bases:

@ stratum n

Insert A in environment [Y ____ Z]x

Prefer: ø -> A / Y _ Z.

Sample: either insert /´/ 'a' in environment:

[d{e, æ}t] + Noun Pl     (data)

or

[aks]+Npl. (oxen)

and so forth.

Bad rule writing

The lexical entry of each word contains all the irregular information about the word (stem).

The rules shouldn't contain the information of which items of an irregular verb are irregular, as Katamba does.

There is no insertion here.

The phonological form of the grammatical feature is spelled out:

Better rule writing

[+Pl.] --> /a/ / lexical item containing the feature denoting this spell out.

We could get fancier in rule writing but this will do.

Anything appropriate may be put in the context slot.

No insertion, but spell out; how do we spell out a morpheme?

And

the lexical entry

here, just the lexical feature

/a/ / [Pl]

n this means [+Pl] is spelled out as /a/ when true; the underscore means the context (can't adjust the position of the underscore).

so ox+[Pl] --> ox+en.

this basically the idea here on rule writing.

of course, it may not be a matter of rules, but of filters, or selection.

• (?2005: End of Course Requirements)
• Post Lexical Rules

• Rules that are not related to word-formation rules.

• Another name: phrasal phonology
• These rules deal with the production of speech rather than phonemic rules.
• Affects the allophonic output.
• post-cyclical (sic!)
• These forms are clitics.
• Some must be adjoined to a host:

• As speech production increases, several changes occur affect the output.
• I'll see you -> I'll seeya.
• Did you go -> Didya go? --> Didja go? --> ‡´'go? (‡ = 'dzh', voiced)
• Did he go -> didee go?
• Are you doing something -> arya doing something?
• What did you do -> What didya do -> whadja do?
• What are you up to? -> Chupta? (how many steps?)
• Cheatin' vs. t¢î'?i?´n?
• English /t/: -> glottal stop (British) or a flap, dialect dependent.

• Filters

  A filter marks a member of a set bad.

  A number of filters are required if the set contains more than two members.

  Recall a set is enclosed in {}s.

  Filters would be required in morphology if we go that rule.

  Filters account for regular rules as well as exceptions.

  An example of a voicing feature in phonology

  a voiced obstruent cannot follows voiceless obstruent at the end of a word

  The opposite is true, too.

  A simple filter would block one, but no the other.

  Voiced 'z' filter

  *z / [-Voice] ____ ##.

  We still need a filter for /s/ after a voiced consonant at the end of a word.

  Furthermore, we want the filter to be more general.

  Use the alpha convention

  *[aVoiced] / [-aVoice] ____ ##

  This is the 'alpha' convention.

  If the features of voice at the end of a word do not match, the example goes down in flames.

  */kætz/, */dags/.

  A morphological filter

  Given noun strems in English which take the stem extender -us or -um:

  * N-{um, us} / ____ [+Pl].

  This filter blocks the stem extenders -um and-us in the plural form.

  {radi, radi-us}, {dat, dat-um}

  *radiusi, *datuma.

  It doesn't block the lack of the base-extenders in the singular:

  *radi, *dat. What makes these ungrammatical?

  a universal convention

  in words, if a set contains two or more morphemes, each morpheme must be selected in at least one context.

  This block perfectly good morphemes from never appearing.

  In the above case, radi and dat are selected in the plural, then other two must be selected in the singular.

  This is the idea in principle.

  It works well in binary situations.

  Preliminary Version

  If a morpheme set contains two or more allomorphs, each morpheme must be assigned to an output form.

  This says that if we generate *radi, as well as radi+i, *radi will fail because there is a second allormph which must be assigned. In this case is must be assigned in the singular..

  This system works best in a dyadic situation.

  If we can make the distribution of allomorphs binary, then the principle should work rather well.

  Biggest problem here is there must be an ad hoc rule that checks to see if each allomorph has been assigned a form.

  Second version

  I think I found the answer.

  [-Pl] is assigned to all noun stems.

  In English singular nouns, the stem does not split into a lexical stem plus an inflectional suffix.

  However, there is a group or set of noun stems that split into a base plus a base-extender:

  {RADI, [-Pl]} --> {RADI-US}, {[-Pl]}.

  At this point, [-Pl] is a paradigmatic feature of the stem.

  Because the noun stem splits into two submorphs, [-Pl] is assigned to the base extender.

  {RADI-US}, {[-Pl]} --> {RADI-{US, [-Pl]} &Mac249;

  /rédi-´s/.

  In effect, the base-extender '-us' is deleted before another phonetic affix.

  Without going into deeper rules for now, let us assume the following partial lexical entry for radius:

  RADI - {US, [-Pl]}

  RADI - {US, [-Pl]}	allomorph
  RADI + {I, [+Pl]}	allomorph

  The lexical entry contains enough information to get the correct result.

  We do not need to apply a rule or a filter here.

  A common problem that linguists do is underspecify items.

  We should be able to predict now that {us, um} will be marked base-extenders that are deleted before a phontic suffix.

  We might be able to do the same thing for the plural.

  But for the time being, let us go with the table we have above.

  It will be the same for all members of these two classes.

Go to Keywords

Return to course outline.