From morphemes to Distributed Morphology
LING 449U, How words are built
2026-09-29
Last week: speakers can prepare morphological pieces, as well as sounds.
Today: what are those pieces, and how do they become a spoken word?
Compare smile, smiles, and smiled. What stays the same? What changes?
| Word | A first segmentation | Contributions |
|---|---|---|
| dogs | dog-s | dog + plural |
| unkind | un-kind | negation + kind |
| workers | work-er-s | work + person who does it + plural |
A morpheme, in a familiar introductory definition, is a minimal pairing of form with meaning or grammatical function.
work → worker → workers
| Term | What it identifies | Here |
|---|---|---|
| Root | The core left after removing affixes | work |
| Base | What an affix attaches to | work for -er; worker for -s |
| Affix | A bound piece attached to a base | -er, -s |
A base can already contain several morphemes.
unlockable can have two bracketings:
| Structure | Interpretation |
|---|---|
| [un-[lock-able]] | Not able to be locked |
| [[un-lock]-able] | Able to be unlocked |
Same linear pieces; different combinations and meanings.
These answer different questions.
| Lexical content | Grammatical function | |
|---|---|---|
| Free: can occur as a word | dog, kind | the, and |
| Bound: needs a host | cran- in cranberry | plural -s, past -ed |
Free/bound: how dependent is its form?
Lexical/functional: what kind of contribution does it make?
Why isn’t “root” another word for “free morpheme”?
| Pattern | Position | Example |
|---|---|---|
| Prefix | Before a base | un-kind |
| Suffix | After a base | kind-ness |
| Infix | Inside a base | Tagalog s-um-ulat, from sulat ‘write’ |
| Circumfix pattern | Around a base | German ge-spiel-t ‘played’, participle |
Position describes where the material appears.
| Typical job | Example | |
|---|---|---|
| Derivation | Make a new lexeme | kind → kindness; kind → unkind |
| Inflection | Make a grammatical form of a lexeme | dog → dogs; smile → smiled |
Derivation can change word class; unkind shows that it need not.
Inflection contributes meaning too: past and plural matter.
unhappiness · replayed · doghouses
For each word:
Which labels describe position, and which describe function?
| Plural form | Ending we hear |
|---|---|
| cats | [s] |
| dogs | [z] |
| buses | [ɪz] or [əz] |
These are allomorphs: different realizations of a morpheme.
Here, the surrounding sounds predict the variation.
| Contrast | What changes in the form? |
|---|---|
| cat → cats | Add an ending |
| sheep → sheep | No audible change |
| mouse → mice | Change the stem vowel |
| go → went | Use a radically different stem: suppletion |
We can identify a grammatical contrast without finding a separate sound segment for it.
She walk-s.
The ending occurs with a bundle of properties:
present tense + third person + singular
Compare I walk, they walk, she walked.
Where are the three separate sound pieces for those three properties?
DM separates these jobs and locates them in different parts of the grammar.
| A lexicalist architecture | Distributed Morphology |
|---|---|
| A word-building component supplies words to syntax. | Syntax builds structures within words as well as phrases. |
| Word formation and phrase formation have distinct machinery. | One syntactic system combines the abstract building blocks. |
DM: pronunciation is supplied after syntactic combination.
| First-pass definition | In the DM analysis |
|---|---|
| A minimal form–function pairing | A terminal node with grammatical content, or a root |
| Plural looks like a piece -s | Syntax contains Number[plural] |
| Its sound and function come together | A realization rule supplies its exponent |
Exponent = the form that realizes a grammatical unit.
| Store | Stored information | Used for |
|---|---|---|
| Syntactic inputs | Features and roots, e.g. [past], √SMILE | Building structure |
| Vocabulary | Rules linking roots/features to exponents | Pronunciation |
| Encyclopedia | Conventional meanings and world knowledge | Interpretation in context |
There is stored knowledge, spread across these lists.
The usual three-store presentation groups features and roots together.
Goryczka (2025) separates them into two input lists, giving four lists in total.
The central distinction remains:
syntactic inputs · pronunciations · conventional meanings
1. Select abstract ingredients
√SMILE + v + T[past]
2. Combine them hierarchically
[ ] ]
v makes a verbal structure; T contributes tense.
3. Vocabulary Insertion supplies exponents
| Abstract item | Exponent in this example |
|---|---|
| √SMILE | /smaɪl/ |
| v | ∅ |
| T[past] | /d/ |
4. Arrange the pieces and apply phonology: [smaɪld], smiled.
LF: combine the structural contributions, including past tense.
Conceptual interface: use conventional meaning and world knowledge.
For smiled: a smiling event located in the past.
The root label, its pronunciation, and its interpretation have different jobs in the analysis.
A simplified plural Vocabulary:
| Realization rule | Condition |
|---|---|
| Number[plural] ↔︎ ∅ | In the local context of √SHEEP |
| Number[plural] ↔︎ /z/ | Elsewhere |
two sheep: the specific rule applies.
two dogs: use the elsewhere rule.
I smiled. She smiled. They smiled.
The subjects differ in person and number; the past ending stays /d/.
A rule T[past] ↔︎ /d/ need not specify person or number.
The rule can be underspecified even when the structure is fully specified.
Syntax and pronunciation may package features differently.
These operations occur after syntax, toward PF.
Keep a record of the meeting.
Please record the meeting.
How would DM derive these two uses of record, using its different lists?
The files are separate.
Please separate the files.
How would DM derive these two uses of separate, using its different lists?
arrive → arrival
collect → collection
destroy → destruction
How would DM derive these nouns? What must be listed, and what must the morphology or phonology do?
| A possible error pattern | Candidate source of difficulty |
|---|---|
| Wrong root, intended tense retained | Root selection or realization |
| Intended root, wrong tense form | Feature selection or grammatical realization |
Root and inflection errors can help us investigate these distinctions.
What additional evidence would distinguish a feature error from an exponent-selection error?
Thank you to Faruk Akkuş and Pamela Goryczka for the materials that helped shape these slides.