In this section, we are going to read through an academic article together, Cliticization vs. inflection: English N’T by Arnold Zwicky and Geoffrey Pullum. You should be able to find the paper in your university library.

At the end of this page, there is a glossary of technical terms used in this paper that haven’t been covered yet in this textbook.

The first pass

As discussed in Section 6.9, the first thing you should do when reading an academic article is to read the abstract and keywords carefully. The abstract of this paper isn’t labeled, but is found in a smaller font right under the authors’ affiliations. This paper doesn’t have keywords, so we can skip that step.

Next, skim the introduction and conclusion. For this step in this paper, I would skim the unlabeled section at the beginning, Section 1: Background, and Section 5: Further Observations. When you’re scrolling or flipping through the pages between sections 1 and 5, pay attention to the titles of the headers and the kinds of data. Since this paper has uninformative section headers, you may wish to read the first sentence or two of each section as well as the header. When you get to Section 5, you will probably notice that it is not a conclusion, but rather in-depth analysis and contextualization. At this point, I would start skimming to find the conclusion—it turns out the true conclusion is just the final paragraph of the paper.

The second pass

Now go back to the beginning of the paper and read it through from beginning to end. In this guided reading section, we will do a deep dive into parts of Section 4, which is both the most important part of the paper and also where most of the more complex ideas are found. I will leave the rest for you to read on your own.

Beginning of section 4, page 506

At this point of the paper, the authors have presented 6 criteria for distinguishing inflectional affixes from clitics and have illustrated how those criteria work for some non-controversial examples of both inflection and clitics, to establish that these criteria are valid. Now, they are using these six criteria to test a more controversial morpheme: negative n’t, and will show that it behaves like an inflectional affix. Note that one of the authors of this paper, Zwicky, had previously adopted the opposite analysis—and has changed his mind when he considered the evidence more fully.

This section doesn’t go through the criteria in the same order as the rest of the paper. Instead, they start with Criterion E: “Syntactic rules can affect affixed words, but cannot affect clitic groups.” Back on page 504, the reason for this criterion are explained. They are working under the assumption that the model of grammar is structured in such a way that all syntactic operations occur before any phonological operations, and that cliticization also occurs after all syntactic operations. On the other hand, word formation processes, including inflection, happens before syntax.

To test this criterion, they look at Subject-auxiliary inversion (SAI), which is the process involved in yes-no question formation. SAI is unquestionably a syntactic operation, and so it should occur after inflection but before cliticization, as shown in (1). This predicts that inflectional affixes should move with the auxiliary in yes-no questions, as shown in (2). Here, different inflectional forms of DO can undergo SAI: do, does, and did.

It also predicts that clitics should not move with the auxiliary, which is also what happens, as shown in (3) (Zwicky and Pullums’s (8)).

What does n’t do? It can move, so it behaves like an affix according to criterion E. This is shown in (4) (Zwicky and Pullums’s (4)), where it undergoes SAI.

Table 1, page 508

Zwicky and Pullum’s Table 1 shows the complete list of auxiliaries that can host n’t in the authors’ grammars, providing evidence for criteria A, B, and C.. The table includes the bare form of the auxiliary, the pronunciation of the bare form, the form of the auxiliary with n’t, and the pronuniciation of the form of the auxiliary with n’t. Note that the transcriptions here aren’t in standard IPA, which is common in older papers like this when typing unusual characters was more difficult.

Criteria A states that “clitics can exhibit a low degree of selection with respect to their hosts, while affixes exhibit a high degree of selection with respect to their stems.” In other words, clitics can usually attach to anything, but affixes are more picky. The fact that this table is even possible provides evidence that n’t behaves like an affix—n’t can only attach to auxiliaries and the full list of possible stems is short enough that it can fit on one page!

Criteria B states that “arbitrary gaps in the set of combinations are more characteristic of affixed words than of clitic groups.” This table shows that n’t behaves like an affix because it cannot attach to auxiliaries like may and am. There is no phonological reason why may should not be able to host n’t—mayn’t is a possible word in English—making this an arbitrary gap. The auxiliary am is a bit more difficult to pronounce with n’t, but some dialects do allow it, suggesting that, again, this is an arbitrary gap.

Criteria C states that “morphophonological idiosyncrasies are more characteristic of affixed words than of clitic groups.” In other words, there is more lexically conditioned allomorphy for affixes than for clitics. This is shown throughout the table, in the many differences in pronunication between the bare form and the affixed form. For example, the vowel in do /du/ changes in don’t /dont/ and the /l/ in shall /ʃæl/ is dropped in shan’t /ʃænt/. There is also not a consistent pattern that can predict whether the negative form is one syllable or two—for can, both options are possible!

Criterion D, page 509

Criterion D states that “semantic idiosyncracies are more characteristic of affixed words than of clitic groups,” which means that words with affixes are more likely to have special non-predictable meanings than combinations of clitics and hosts. In this section, Zwicky and Pullum show that not and n’t don’t have the exact same meanings in all contexts, but the difference is subtle. This difference in meaning shows that n’t behaves more like an affix than a clitic. This section is one of the trickiest to understand, especially if you haven’t studied much semantics before.

The argument in this section depends on the notion of scope, which is a word used to describe which parts of a sentence are modified by a semantic operator. In this case, the semantic operator is negation. Compare sentences (4a) and (4b). In (4a), the word not only modifies rotten. That is, the scope of negation is only the word rotten. In (4a), Hentie did eat a fruit, but the fruit was fresh. In (4b), on the other hand, the word not modifies the entire sentence; in other words, not takes scope over the entire sentence. In (4b), we do not know whether Hentie ate a fresh fruit. All we know is that the action of eating a rotten fruit did not occur.

When there are two semantic operators in the same clause, they interact. Which order you compose them changes the meaning of the sentence. For example, in (5), we see the difference between must(not(P)) and not(must(P)). In semantics, P is used as a placeholder for a proposition, which is a semantics term meaning a statement that can be true or false.

The difference in scope between not and n’t is illustrated in (6) and (7) (from in-text examples from Zwicky and Pullum 1983: 509). The modal must not only has one possible scope meaning. The sentence you must not go home can only mean that you are not allowed to go to home. The sentence with n’t instead of not has the same meaning.

The pattern is different with can, though. In these sentences, an accent (é) is used to mark stress. With the full-word form in (7a), the sentence is ambiguous. It can mean that you have the option of not going home, but it can also mean that you are not allowed to go home. The sentences with n’t in (7b) and (7c), on the other hand, only have one possible meaning; they can only mean that you are not allowed to go home, which is the opposite scope order that was allowed with mustn’t.

The relationship between must not and mustn’t is predictable—they share the same pattern as each other. In contrast, the relationship between can not on the one hand and can’t and cannot on the other is not predictable. We then have a semantic idiosyncracy in the meaning of a word using n’t, indicating that n’t behaves more like an affix than a clitic.

Glossary of terms used in this paper

The following is a list of technical terms used in this paper that have not yet been introduced in this textbook, as well as some advanced words used in scientific discourse that aren’t quite technical terms.

Agreement: When the features on one word gets transferred to or shared with another word.

Auxiliary verbs: Verbs like have, be, and do that appear before a main verb.

Conjunctive: Having the meaning of ‘and’.

Constituent: A group of words behaving like a syntactic unit.

Copula: A verb used to introduce non-verbal predicates, such as am in I am a linguist.

Feed: When one rule or process must occur before another one, the first rule is said to feed the second. This is a term borrowed from phonology.

Finite verbs: Verbs which are inflected for tense (as opposed to infinitive verbs, which are not).

Formative: A word you can use in morphology to describe a linguistic element when you don’t want to be specific about whether it is a word, a morpheme, or something else.

Government: A particular structural relationship in syntax, largely no longer in use.

Emphatic stress: When you pronounce a word more forcefully for emphasis. When a word has a clitic variant, the clitic usually can’t be used if there is emphatic stress on it.

Idiosyncracy: A pattern or feature that is unique to an individual. In this paper, the individual is the morphological stem. Attachment of an affix to a stem is more likely to have exceptions or unusual patterns than attachment of a clitic to a host.

Inchoative/causative verb: An inchoative is a verb expressing that a certain state has begun. A causative is a verb indicating that the subject caused a resulting state. For example, verbs such as moisten and soften mean ’cause to become moist/soft.’

Lexical items: Items formed in the lexicon. That is, they are formed by morphological rules, not syntactic rules.

Lexicalization: The process by which words are added into the lexicon. The paper seems to be referring, here, to how auxiliary + n’t constructions were previously combined syntactically but were so frequent in combination that they got added to the lexicon. Once they had their own lexical entries, the combinations were able to pick up idiosyncratic meanings and forms.

Monosyllables: Words which contain only one syllable.

Neologism: Newly-formed word.

Node: The point at which several branches come together in a syntax tree.

Parochial: Limited in use and scope.

Readjustment rule: A rule that adjusts the output of syntactic rules to prepare it for the phonology.

Selection: When a word or morpheme is “picky” about what things can attach to it to form a new word or a phrase.

Sentential negation: When the entire sentence is negated, not just a word or a phrase inside the sentence.

Simple clitic: A clitic which has an equivalent full-word form that can appear in the same positions.

Special clitic: A clitic which either has no equivalent full-word form or it has an equivalent full-word form that appears in different positions.

Stipulate: To add an assumption to our model without any motivation other than that it makes the data work out.

Subject-auxiliary inversion (SAI): The word order change involved in the formation of yes-no questions in English, in which the subject and the auxiliary swap positions. For example, the statement “You are awake” becomes the yes-no question “Are you awake?” when the subject you swaps position with the auxiliary are.

Superlative: An inflectional affix on adjectives to mean ‘most’.

Syllabic: Behaving as the nucleus of a syllable (i.e., instead of a vowel).

Syncategoremic: Something that doesn’t quite have meaning on its own, but only has meaning when combined with something else.

Transformation: A rule of grammar that rearranges the order of words, used in some grammatical models of the 1960s-1980s. They have generally been replaced by movement rules in many modern models of grammar.

Check yourself!

References and further resources

Academic sources

Noyer, Rolf. n.d. Distributed Morphology: Frequently Asked Questions list. https://www.ling.upenn.edu/~rnoyer/dm/home.html

Zwicky, Arnold, and Geoffrey Pullum. 1983. Cliticization vs. inflection: English N’T. Language 59 (3): 502-513.