Show class topics

Make sure you're familiar with the syllabus.

Questions for this quarter: What kinds of phenomena are part of morphology? How can different languages' morphological systems differ, and how are they similar?

Morphemes can take the form of prefixes, suffixes, infixes, circumfixes, changes to segmental or supersegmental features inside words, reduplication, and more.

Inflectional morphemes can express tense, case, noun class, verb agreement; derivational morphemes can change a word's part of speech and often add more meaning-rich content.

Show class topics

Undoable is ambiguous because morphemes are added to roots one at a time, here in two possible orders/structures; the meaning of undoable is therefore more complex than that of do or undo. Cranberry isn't assembled on the fly like this; the cran morpheme in e.g. cranapple is thus arguably more complex than the whole word cranberry.

Some words, like cat, have a simple 1:1 form-meaning relationship. In other words and morphemes, this 1:1 relationship is disrupted; cases include toward/towards, the plural/possessive/3sg -s suffixes, the null English present tense morpheme, roots like eat/ate, and more.

Make sure you understand how to format morphologically complex linguistic data (see 'Presenting linguistic data').

A resource for IPA enthusiasts: Hear the sounds on the IPA chart!

Show class topics

After going through general properties of morphemes last week, today we shift to looking at morphology in particular languages.

Tense, aspect, and mood: how are these categories different? How are they expressed in English (partially via morphology; also via syntax)?

ASL has a much richer system of productive aspect morphology; while we can express all the same meanings in English, we don't have nearly as many regular, productive, grammatical categories.

This description of a language's tense system raises some interesting grammatical points, but doesn't always seem to distinguish grammar from thought, and seems to think English is inherently more sensible; beware of these kinds of claims as you read about languages with wildly different structures and categories.

Show class topics

Make sure you understand why person and tense (but not number) are deictic features.

The English number system has singular and plural (expressed various ways) on count nouns; other nouns (e.g. rice are mass nouns.

English pronouns don't have internal morphology; pronouns in many Australian languages, however, are composed of productive (or semi-productive) person and number morphemes. Pronoun systems can be based on simple number (singular/dual/plural) or can be relational (basic form/one more than.../two or more more than...).

Show class topics

A common moral of working with descriptive grammars: Language data is extremely complicated!

Grammatical gender is a complex combination of semantically based (masculine and feminine can reflect biological gender), phonologically based (certain endings, for example, may correlate imperfectly with genders), and arbitrary (there's no synchronic explanation for why many words fall into particular gender category).

Jingulu gender is largely, though not entirely, shape-based (even though the 'gender' names are misleading). Patterns of possible agreement show the internal feature structure of these genders: vegetable is the most semantically complex, masculine the least, etc.

Show class topics

In noun-adjective agreement, features like gender are inherent on Ns and contextual on As; evidence for this comes from syntactic structure (N is the head of the NP), distribution (Ns occur without As and still have gender), and variability (Ns have only one gender each; As can be any gender). Similar arguments explain why e.g. number is inherent on Ns and contextual on Vs: Ns can have any gender wherever they occur; Vs reflect only the gender of e.g. the subject N.

Joseph Greenberg, former Stanford anthropologist and linguist, famously proposed 20 morphological language universals. Most of these restrict the universe of logically possible languages, and many are implicational (of the form "if...then..."). Many imply a markedness relation among elements (e.g. 36 suggests that number is more basic/less marked than gender), and many seem to share some sort of basic premises (e.g. 36 and 37).

Whether absolute or statistical, universals add richness to our investigation of morphology in showing how all of these individual systems fit into the greater generalizations. All are strong, falsifiable claims, and they invite questions about where these commonalities come from (e.g. cognitive or cultural limitations, Darwinian-style adaptivity, shared 'proto-Earth' origins...).

Show class topics

Nouns, verbs, adjectives, and adverbs are often called open (or lexical) categories; prepositions, determiners, and conjunctions are closed (or functional) categories. These categories aren't absolute: it's easy to create new nouns; adverbs are less flexible; sometimes there are new prepositions; determiners are pretty settled.

Inflectional morphemes are more like functional categories; derivational morphemes are more like lexical categories. New derivational morphemes include -gate, -palooza, -stock, etc.

As suggested here, inflection is generally obligatory, whereas derivation can instead be replaced by a word or phrase.

Show class topics

Please make sure to start data collection 3 early, as it may take some looking. Talking to each other could be very useful!

Booij defines inflection as "the morphological expression of morpho-syntactic categories on words" (p. 315) and derivation as all other morphology; this is why inflection is a small, tractable set of meaning types whereas derivation is best understood in contrast to inflection.

Some of these comparisons between derivation and inflection are relatively clear-cut, like expression at word periphery vs. near the base. Others are quite fuzzy; is plurality really more abstract than child-hood?.

Valence effects fall right at the intersection of derivation and inflection: they change meaning, but these changes are to relatively basic grammatical properties. They affect syntax, but don't respond to it (like case does). Some valence-changing morphemes are inflectional and others are derivational. To be continued next time.

Show class topics

Deface can be used to mean "vandalize by drawing faces on", illustrating how words formed by derivational morphology can have idisyncratic, non-compositional meanings.

Inflectional morphology is tightly connected to syntax. Passive and antipassive morphology, which change the mapping between semantic and syntactic structures, are typically inflectional; causative and anticausative, which change semantic structures (and syntax too, but only less directly), are typically derivational.

Morphology, as a field, is hard to define; there are many more self-declared morphosyntacticians and morphophonologists than strict morphologists. As a result, it's hard to identify phenomena which clearly are or aren't fundamentally morphological. Many morphological concepts (e.g. inherent vs. contextual) are similarly hard to define, and rely on other theory-internal assumptions. It's more important to understand that such a distinction might exist, and how it might be defined, than to know exactly which morphemes definitely belong to which categories.

Show class topics

Verb agreement, inverse marking, fixed word order, and case marking are all strategies for (among other things) distinguishing subjects and objects; each has cases where it's less helpful, or other downsides.

An intuition behind NOM-ACC case systems: all subjects are similar. Behind ERG-ABS: transitive agents are unique.

While there's a roughly universal animacy hierarchy, it's not as strict or lexical a system as e.g. gender. Languages intuitively think of subjects as more typically animate than objects; we see a small effect of this sort in English, in the increased acceptabilty of "I was hit by the truck" vs. "The truck was driven by me".

For next time: Actual animacy-based ergative split data; discussion of Comrie's "functional explanations", and of how a noun's case gets decided.

Show class topics

NOM-ACC and ERG-ABS are not the only logically possible S/A/P case systems, but they are by far the most common.

Comrie refers to a functional explanation for this: The function of case is to distinguish A and P; these two systems do it most efficiently. Functional explanations are common in some linguistic circles and loathed in others; they can be a useful way of finding unifying explanations, but can also be difficult to falsify and introduce bias. Here, while it's an intuitively appealing story, it's hard to know whether this is actually why these systems are most common.

In each of these Australian languages, objects show split case marking (make sure you understand the basis for each split; this version corrects two A-W typos from class). For next time: do the subjects seem to show a similar split?

Show class topics

I'm happy to talk with anyone about possible topics for the final paper; please feel free to make an appointment!

The data from last time doesn't have much data on subjects and agents, but there's just enough to identify Arabana-Wangkangurru as having fundamentally ERG-ABS subjects, and to suspect there's no animacy-based split in agent marking in Badjalang.

White House is a compound, whereas white house is a phrase. This is evident in their meaning and stress; if English had e.g. gender agreement, we might see a difference there, too.

Babysit seems to be a headless (exocentric) compound; it's neither a kind of baby nor a kind of sitting. Many exocentric compounds insist, surprisingly, on overly regular inflection, e.g. the Toronto Maple Leafs. The existence of irregular babysat is tricky; to be continued Wednesday.

Show class topics

Workman is an endocentric compound; walkman is exocentric. This explains why walkmans is possible, but not *workmans. The existence of possible alternative walkmen raises tricky questions about the ordering of morphosyntactic processes; this is the kind of thing morphologists and syntacticians spend lots of time discussing.

Mother in laws is most likely a compound; mothers in law is definitely a phrase. Similarly, White House travel office staff can be produced as either a compound or a phrase. This difference is fairly clear in listening to the two pronunciations, but it's hard to find other (e.g. semantic) properties that clearly correlate. It's easy to identify compound White House vs. phrase white house, but longer and less familiar examples are harder to distinguish, even for those producing them.

Phenomena at the morphology-syntax interface then include things like case, where morphology is determined by syntax as well as root features (e.g. animacy), and also compounds vs. phrases, which contrast in being morphological vs. syntactic structures, often with similar surface meanings. Clitics also fall at this interface; these morphemes are not quite full free words, but also not fully word-internal affixes.

Show class topics

A bit more on clitics: crucially, they're less bound to roots than affixes, but still not free words. Contractions (can't, could've, I'm) are the most intuitive English examples, since they look like parts of words. Other English function words also function as clitics in casual speech; reduced forms of go to town and vote for Betty each can have only two phonological words. Clitics are best diagnosed by identifying phonological properties of word edges.

Any morphophological phenomenon which is defined phonologically, or which triggers phonological processes, can be said to involve the morphophonology interface. This includes allomorphy (im-possible ~ in-distinguishable), affix-triggered root changes (politic ~ political), morphemes which are only root-internal changes (record(N) ~ record(V)), morphemes whose appearance is defined in phonological terms (Cali-fuckin-fornia), and those which only occur on roots of particular phonological shapes (?*Con-fuckin-necticut).

Show class topics

McCarthy and Prince 1986: Yes, this is a very, very difficult paper. We're focusing on working through basic generalizations about data in selected sections, to get a sense for some of the cases where morphology and phonology are most tightly intertwined. This is why I encourage you to skip most of the theory, analysis, weird terminology, and other painful paragraphs and just focus on understanding the examples. (See my version of today's homework.)

Moras are units of weight. In all languages, each vowel is counted as one mora. In some languages, each coda consonant counts as one mora (e.g. taa and tan are heavy; ta is light); in others, codas are weightless (e.g. taa is heavy; tan and ta are both light). Onset consonants are always weightless.

The morphology here is templatic because it's defined purely in terms of shapes: in Ilokano and Oykangand, reduplicants are single syllables; these don't necessarily match actual syllables in roots, and are preferably heavy. In Tagalog, reduplicants are single 'core' syllables: no more than one onset consonant; no coda ( (C)V ). This morphology is prosodic because the templates are defined in terms of prosodic units (syllables, moras, occasionally feet) rather than segments (most morphemes, e.g. normal affixes, are defined in terms of their strings of sounds). Previous work had generally discussed templates segmentally; McCarthy and Prince are trying to show that these descriptions can be made more general.

Next time we'll look at how Mokilese reduplication is different from those we discussed today, and at how truncation processes refer to the same kinds of template shapes as these reduplication processes.

Show class topics

Last time we saw that languages can demand that reduplicants be a light ('core') syllable, as in Tagalog, or a potentially larger (heavy if possible, given the shape of the base) syllable, as in Ilokano. While there are plenty of other non-monosyllabic forms of reduplication (reduplicants may be larger or smaller than a single syllable), there's only one other basic option for monosyllabic reduplication: reduplicants must be heavy in Mokilese; this may be satisfied by adding a mora to the vowel if necessary.

Truncation templates follow the same basic pattern: when truncating to a monosyllable, this may be a light 'core' syllable (Zuni), may be optionally heavy (Madurese), or must be bimoraic (some Japanese nicknames).

Other prosodically based phenomena are less rigid; for example, English nicknames often involve truncation, but don't hold to a fixed syllable/mora template (and can vary in whether they retain initial or stressed syllables). The Japanese secret language discussed in McCarthy & Prince has a rigid output form, but can get there via various processes.

Show class topics

While it's generally not possible to explain exactly why a particular morpheme has a particular surface form (e.g. why English uses -ed rather than -et or -en for past), there are some general frequency-based principles that can help identify singulars vs. plurals, etc. We looked through some of this data, taken mostly from this chapter, to explore these issues.

Some basic principles:
*Singular > plural > dual; 3rd person > 1st > 2nd; nominative > accusative > other cases; active > passive; etc. (">" means "is more frequent than")
* Frequent categories are often expressed by shorter (or null) forms.
* Frequent categories may use more frequent sounds (t s n > k p m, etc.).
* Frequent categories may have less syncretism.
* Frequent categories may, conversely, have more suppletive allomorphs.
* Frequent roots may have more irregular forms.

These aren't absolutes, as evidenced by e.g. local plural morphology reversals in Welsh. And like any functional explanation, it's tempting to try to unify these, but hard to know how far one can usefully go. Still, it's clear that phonological pressures exist in shaping the grammar at the morphological form level, in addition to being used in more active on-line processes.