J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface IS MORPHOLOGY A SEPARATE COMPONENT? University of Newcastle, Structure of the Syntax-Lexicon Interface I. What is “Morphology”? Its empirical scope We should first decide what the term “Morphology” covers. For English, it seems to be the study of some but not all bound morphemes, such as those in (1): (1) -al, -(a)tion, -age, -(e)d, -en, -er, -(e)s, -ess, -est, -ic, -ify, -ing, -ism, -ity, -ize, -ly, -ment, -s, -th, -ton, -ward, -y, co-, de-, ex-, mis-, non-, re-, un-, etc. (2) national, admiration, linkage, linked, soften, softer, edges, priestess, shortest, sarcastic, solidify, linking, criticism, nudity, stabilize, safely, resentment, links, growth, Riverton, eastward, grassy, cooperate, de-stress, ex-friend, mis-speak, non-toxic, re-wrap, uneasy In practice, morphology has not included study of the following italicized morphemes, even though they are “bound.” They are rather “compounds,” which is confirmed by the fact that that these items exhibit the stress patterns of compounds. (Hyphens are morpheme boundaries.) (3) aero-space, astro-naut, catty corner, chock-full, cran-berry, e-market, epi-gram, fishmonger, helter-skelter, iso-morph, jay walk, luke-warm, multi-plex, neat-nik, pell-mell, shilly-shally, sleep-aholic, topsy-turvy, x-ray, geo-metry, gyno-phobe, micro-scopy, mono-maniac, necro-philia, phono-graph, sino-phile, taxo-nomic, tri-lingual, were-wolf As in (4), Morphology also typically treats some potentially free morphemes. So all we can say is: Morphology includes the study of “some” bound morphemes and “some” free morphemes. (4) -able (consumable), -ful(l) (harmful; spoonful), over- (over-exert), out- (out-swim) Nor does the division (1)-(3) correlate with productivity or semi-productivity. Many items in (1) are not productive. Some of compounding patterns found in (3) are fully productive (e.g. spy store, dessert free). Many of the others accept coinings and certain other extensions (5): (5) e-target, globocentric, micro-manage, neophilia, screen-oholic, pornophobe, pie-monger The essential difference between what tradition distinguishes as “morphology” (1) and “compounding” (3) lies elsewhere, in the types of features bound morphemes express, as defined in Chomsky (1965: Ch. II) as follows. His example of a feature F was ±ANIMATE. (6) a. Definition. Semantic features f (of high specificity and detail) that play a role in LF but no role in derivations can be called “purely semantic features.” b. Definition. Features F (of less specificity and detail) that play a role both in LF and in derivations can be called “(cognitive) syntactic features.” Throughout, I use lower case “f’ for purely semantic features, and upper case “F” for features of general meaning used in syntax (ANIM, DEF, LOC, NEG, PAST, PATH, PLUR, WH, etc.). Warning: at least 50% of attendees will DELUDE THEMSELVES into thinking that “syntactic features” means “not interpreted” in a vague sense, like “agreement features” or “declensions.” 1 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface An important distributional distinction: All grammatical categories have syntactic (F) features, but only the open classes N, V, A and P have purely semantic (f) features. Exercise: Think of English Modals. Each has its own semantics. So Emonds must not realize or think this. OR he thinks they are Vs (like I. Roberts). OR he is wrong that only open class items have semantic features. IF YOU REASON THUS, YOU ARE ALREADY MISTAKEN. Why? It is good that we are focusing on English, because this language provides a very informative diagnostic for which morphemes have purely semantic features: (7) English Morphology traditionally includes study of all and only bound morphemes that (i) lack purely semantic features f, and (ii) do not contribute their own stress to word stress, unlike members of compounds. (8) (English) Morphology: If a bound morpheme μ has no LF features except syntactic head features F, then μ entirely lacks stress prior to combining in larger PF domains. II. Separating this language-specific property of Morphology from its essence I use the term “affix” for English only for its morphemes μ that obey (8). (8) is more than just a definition because it claims to empirically correlate two logically independent properties (7i-ii). And English Morphology is indeed language-particular. French morphology, which English has borrowed so much of in the past, is not subject to (8). In (9), French compounds and phrases (col. 1) and suffixed forms (col. 2) have exactly the same intonations. (9) [N [N rapide ] [N été ] ] ‘summer express’ [N [α télé ] [N cité] ] ‘TV city’ [NP un [A beau ] [N thé] ] ‘a beautiful tea’ [NP [N cou ] [A rond ] ] ‘a round neck’ [NP son [N beau-frère ] [N Roger ] ] [N [A rapide ] [N ité ] ] ‘quickness’ [N [α téli ] [N cité ] ] ‘telicity’ une [N [A beau ] [N té ] ] ‘a beauty’ [I [V cour ] [I ons ] ] ‘let’s run’ [I sont ] [A [V interrog ] [A és ] ] The English combinations in (3) seem to be compounds because they all participate in compound stress; each member has at least secondary stress. French provides no such test because French vowels other than (unaccented) e don’t reduce, whatever the type of morpheme. And independently, all morphemes in (3) also seem to have features f of semantic detail that don’t participate in syntactic derivations; that is, none of them are “affixes” that lack f. Status as an “affix” is thus entirely predictable from the lexical feature make up of a morpheme; “affix” has no special formal status as a category (Lieber, 1992). So we can say that in differently stressed pairs as in (10), internal category structures are identical. (10) high class, highness: stress-free, stressful: deep fry, deepen: afterbirth, rebirth: [N [A high ] [N class/ -ness ] ] [A [N stress ] [A free/ -ful ] ] [V [A deep ] [V fry/ -en ] ] [N [P after/ re- ] [N birth ] ] 2 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface When each member of a branching X0 category contributes an independent stress to the contour for X0, as in the first of each pair, the resulting English structure is called a “compound.” So, what determines English stress in branching X0 is whether a right hand member has a purely semantic f. In (10), -ness, -ful and –en have no such feature. A few items in English that lead to compound stress are often wrongly called suffixes. Consider –esque and –ee. The adjectives in (11a) are stressed like other compound adjectives: picturefrée, noise-frée, visitor-frée. Those in (11b) are like some irregular heads of N-N compounds with stress: apple pie, meat pie. In thedr examples, hyphens show morpheme boundaries. (11) a. picture-ésque, statu-ésque, carnival-ésque, Roman-ésque b. stand-ée, recommend-ée, divorc-ée, grant-ée The combinations in (12) are stressed like compound nouns: stéamship, báttleship, férry boat (12) friend-ship, assistant-ship, receiver-ship, survivor-ship, neighbor-hood, knight-hood, bachelor-hood, maiden- hood Thus in the perspective here, -esque, -ee, -ship and –hood are not part of English morphology; they are not “affixes.” Rather, they have features of semantic detail and are instances of bound morphemes in compounds. As (11a) and (12) show, their stresses are usually well-behaved. The calculation of word stress in compounds. I assume that the account of English compounds in Chomsky and Halle (1968) is correct in its essentials. We return to this in Topic 2. III. The scope of some other putative “principles of Morphology” With Principle (8), we have delimited—and made a strong claim about—English morphology. (8) (English) Morphology: If a bound morpheme μ has no LF features except syntactic head features F, then μ entirely lacks stress prior to combining in larger PF domains. Now there is a major question: Should (8) be considered a principle specific to morphology, or perhaps some kind of “defining characteristic” of morphology? The next lecture argues for a “no” answer, maintaining that (8) is an aspect of the English lexicon-syntax interface.. That is, (8) constitutes an interface principle that says how to construct a syntactic object based on lexical information about stored morphemes. Notice for example how it crucially depends on the notion “syntactic feature” defined in (6b). It belongs to no special morphological component. It is widely thought that other distinct principles apply to “morphology,” the study of a class of bound items defined more or less as in (7ii). However, it goes unnoticed that such principles apply equally well to compounds, i.e. they cover both types of X0 constructs in e.g. (10): (13) Stricter acceptability judgments in X0. Speakers perceive violations of word/ morpheme order in X0 domains more strongly than in Xj domains (j>0). (14) This car is drivable long distances This car can be driven long distances. 3 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface The wet darkness moved eastward. The shadows from trees moved eastward. Those that argue for an “Autonomous Morphology” point out that judgments of unacceptability in morphological domains X0 (15) are unmistakably stronger than in syntactic domains XP (16): (15) Stronger ordering violations inside X0: **This car is [X able-drive ] long distances. **The wet [X ness-dark ] moved eastward. (16) Weaker ordering violations inside XP: *This car can [XP driven be long distances ]. *[XP The from trees shadows ] moved eastward. But if we idealize away from the irrelevant factor of potential (un)pronounceability, the impression of unacceptability in compounds with order reversed as in (18) is more akin to (15): (17) This car is now { road-worthy/ waterproof }. The park was full of { street lights/ night fog }. (18) **This car is now { worthy-road / proofwater }. **The park was full of { lights street / fog night }. So it can be concluded: Acceptability judgments distinguish X0 from XP domains, but acceptability judgments do not distinguish syntax from morphology. Here is another putative principle of “Autonomous Morphology”: The following general constraint on the form of X0 domains appears in the literature under various guises. (19) “No Phrase Constraint” or “Nonmaximality” or “Domain Size Restriction.” A productive X0 domain cannot contain categories characteristic of maximal projections, such as D, I, Degree Words, Intensifiers, Focus Particles, etc. Hoeksema (1988) is a useful discussion of this restriction. By virtue of it, elements characteristic of maximal phrases, bold in (20), are excluded inside X0 domains with bound morpheme heads. (20) She can’t tolerate a [N (*very) [A high]-t] of more than five stories. Try not to behave in such a [A (*your/ *any) [N baby]-ish] way. Lecturers needing tenure end up [N (*must) [V think]-ers] in spite of themselves. However, they are equally well prevented from appearing inside X0 compounds (21). (21) My son attended a [N city [N (*very) high school]]. She has rarely [V (*your/ *any) [N baby] sat] in the last year. That new [N (*must) [V think] tank] outside town is currently hiring professors. Constraint (19) is presumably part of the definition of “possible projections of X0 and XP.” Bound morphology indeed conforms to both (13) and (19). However, mrophology has no special relation to them, since these principles also cover compounds. That is, (13) and (19) are not candidates for some specialized Morphological Component or “Morphological Structure.” 4 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface IV. Digression: predicting the stress split among Germanic (including English) prefixes Items traditionally termed “prefixes” in English contribute to word stress in two different ways. A prefix with an interpreted syntactic feature F has at least secondary stress: (22) 2 – 1: de-plane, ex-wife, fore-tell, mis-speak, re-think, trans-figure, un-happy, under-play In this 2 – 1 stress pattern, the prefixes are stressed like the first elements in certain compound patterns such as spring fresh, ice cold, apple pie, Lincoln Avenue, iron mask, etc. However, prefixes with no inherent LF content (with no interpreted features, neither F nor f) are totally unstressed. This contrast is systematically remarked in grammars of Old English for ancestors of prefixes as in (23a). Modern English extends the list of such prefixes as in (23b): (23) 0 – 1 stress pattern; the prefixes are unstressed: a. be-take, enact, for-get, , under-stand, with-hold b. de-tain, con-fuse ex-hale, re-view, sub-mit The same contrast in German distinguishes stressed, “separable” prefixes with compositional meanings (24a) from unstressed “inseparable” prefixes (24b), which lack any inherent features. (24) a. a. auf-stehen ‘get up,’ ab-steigen ‘dismount,’ teil-nehmen ‘take part’ b. verstehen ‘understand,’ besteigen ‘climb,’ entnehmen ‘take’ The “separable prefixes” in (24a) are actually incorporated prepositions (Maylor, 2002: Ch. 1). In both English and German compounds and “stem + affix” combinations, the right hand member is generally the head (Lieber, 1980). Therefore the interpretable features of the bound prefixes μ in English (22) and German (24a) are not “syntactic head features.” By the formulation (8), these meaningful but purely grammatical prefixes with interpreted F then contribute to compound stress exactly as if they were open class items with interpreted f. In contrast, the “meaningless” English prefixes (23) and the inseparable German prefixes (24b) have no intrinsic content (=syntactic head features), and so all are unstressed. Principle (8) thus correctly predicts that (only) prefixes with fixed content as in (22) and (24a) have stress. Lieber (1980) claims that a few English prefixes e.g. be- and en- are exceptionally heads. This gets repeated endlessly, though both Walinska de Hackbeil (1985) and Maylor (2002: section 5.4) argue convincingly against their special status and conclude that no prefixes are heads. Recapitulating, Germanic prefixes exhibit the following split: If a prefix carries any interpretable feature F, it will consequently be stressed like in a compound. If a prefix has no intrinsic LF content it is unstressed. Principle (8) as stated correctly captures this dichotomy. Exercise: Consider English verbs prefixed with –en. How might one argue against –en as a derivational, category-changing head? What might be a head of e.g. enchain, if –en is not? 5 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface V. No “types of categories” or “types of irregularity” are specific to Morphology There are two types of proposals for morphology-specific inventories of categories, (i) one for categories with content and (ii) another for “purely formal” categories like declension classes. Re (i): Is there some special set of meanings expressed only by bound morphemes? Proposals of this type generally attempt more or less exhaustive inventories of morphological categories. Advocates of Autonomous Morphology may thus sometimes claim that certain syntactic features F, or perhaps semantic features f, can be expressed only by bound morphemes. Some cross-linguistic candidates exemplified in English might be PAST tense, PLURAL, an ordinal suffix –th, and “conversion” suffixes like –ness (AN), -ing (VN), -ic (NA), etc. Yet the quantifier some expresses little else than plural. Even if –s and some differ by a feature, in an example like (25a) the free morpheme some essentially expresses only “PLUR.”. Similarly, the morphemes number and –th in (25b) seem to mean the same thing. (25) a. Are there chairs on the patio? Well, some. b. Will any more cyclists be passing here after these five? The sixth will arrive soon. Number six will arrive soon. Overall, there is no clear evidence that certain meanings are expressed only by free morphemes. Re (ii): Are there some special categories of morphology (other than of phonology) that never occur in syntax? It is difficult to convince morphologists enamored of tradition to dispense with cherished declension classes and conjugations of languages like Latin, Bantu, etc. But these classes can be better defined using only categories independently justified in syntax and phonology (Emonds and Spaelti 2005: 12). Latin seems to be a widely known case in point. (26) Declension Classes. The Latin noun declensions simply reflect the last phoneme of the noun’s lexical entry. This yields the five vowel declensions and a consonant declension. In my view, I-E declensions and conjugations are always best viewed as just opaque names for stem-final segments, which can in turn can phonologically condition allomorph choices among suffixes. Czech noun classes, in contrast to Latin, are often explicitly based on final segments. (27) Underlying forms of nominative, genitive and oblique singulars and accusative plurals: Nom. Sing. terra + Ø re + s servo + s servus classi + s manu + s duc + s (dux) Gen. Sing. terra + i re + i servo + i servi classi + s manu + s duc + is Obl. Sing. terrá ré servó classí manú duc + e Acc. Plur. terrá + s ré + s servó + s classí + s manú + s duc + és Even though the conditioning factors are not simple, arbitrary morphological classes divorced from phonology, rather than reducing complexity, invariably obscure the actual generalizations. (iii) Is there some type of irregularity or idiosyncrasy specific to (derivational) morphology? 6 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface It is frequently remarked that large numbers of forms morphologically derived from open class stems, such as result nominals and causative verbs, can have strongly idiosyncratic meanings. (28) a. Building does not just mean something that is built: Johnny constructed a fantastic cari from his Lego set. *What a complex buildingi it is! b. Containment does not mean just the act or event of containing: *They arranged for the new house’s {containing/ *containment of} more space for kids. c. Nationalize and internationalize have very different senses: Starbuck’s management made the chain nationwide and even internationalized it. (True) In its early years, Starbuck’s management nationalized the chain. (False) From the sheer mass of idiosyncrasies, analysts often feel there is some systematic link between elements of bound (“derivational”) morphology and the function of expressing semantic idioms. However, there are equally numerous quite semantically specific syntactic pairings of the type {open class item(s) + free morpheme}: That is, idiosyncratic meanings of particular derived forms as in (28) don’t differ in kind from thousands of idiomatic free form combinations: (29) bring up ‘íntroduce’ or ‘raise’, hold up ‘delay’ or ‘rob’, face up, feel up, fire up, flame up, free up, freshen up, take up, talk up, think up, tie up, tool up, track up, trip up, turn up Exercise for native speakers: How many idioms with distinct meanings can you find for the verb hit? Like hit home, hit the bottle, hit the books, hit the hay. To take another case, the derivational suffix –ism at first seems specially used for ideologies. (30) atheism, Buddhism, communism, Darwinism, determininism, feminism, Fourierism, hedonism, Judaism, Maoism, Mormonism, nationalism, Naziism, nudism, PanAfricanism, Protestantism, rationalism, Satanism, Shintoism, Taylorism, Trotskyism Such “semi-productive” links, based on multiple pairs {open class item(s) + derivational morpheme}, might then differentiate morphological pairings from those of productive syntax. I nonetheless maintain, for example, that “X–ism” means little more than “a mental creation related to X.” This broad characterization is probably expresses all that is purely linguistic for this suffix in (30). It suffices for the examples in (31) as well, which are by no stretch ideologies. (31) antagonism, authoritarianism, atomism, autism, baptism, cannibalism, criticism Exercise: Can you find 10-20 more such words that are not ideologies? (Some on next handout.) Semi-productive links equally involve free morphemes, such as “take an X,” where X suggests some form of “non-work.” The large set (32) is about as systematic to the set in (30). (32) take a { bath/ break/ breather/ holiday/ nap/ rest / snooze/ stretch/ vacation / week off } Thus, semi-productive combinations using derivational morphology, as in (30), are not different in kind from similar combinations with free forms, as in (32). Thus, no special property of “idiosyncrasy” or “irregularity” can be associated with morphology more than with syntax. 7 J. Emonds, May 2009, Topic 1 Structure of the Syntax-Lexicon Interface VI. Distributed Morphology and “Morphological Structure” (“MS”) This handout has examined initially plausible principles or properties for differentiating syntax and morphology: English Morphology (8), Acceptability Judgments (13), the No Phrase Constraint (19), differences in category inventories, and degree of idiosyncrasy. Only (8) remains a candidate for Autonomous Morphology or a “Morphological Structure” component. These results support a different approach to bound morphemes, called Distributed Morphology. (33) Distributed Morphology (DM). “...the machinery of…morphology is not concentrated in a single component of the grammar, but rather is distributed among several different components.” (Halle and Marantz, 1993: 111-112) DM therefore expects results whereby principles like (13) and (19) are not “concentrated” in Morphology. So I emphatically concur: DM is on the right track. But the original source of DM is not the 1993 article. The first representatives of this approach to this approach were the revolutionary affix movement rule and passive transformation of Syntactic Structures. More generally and later, Baker’s classic (1985) Mirror Principle article and his massive study Incorporation (1988) demonstrate that morphology and syntax must be somehow integrated and constructed in tandem. As words are constructed “from the inside out,” so also are phrases. Thus, due to Baker’s work, generative syntax has had to face this question: Since much of morphology is nothing but the word-internal face of syntax, what if anything remains separate? (34) How much should we distribute morphology? Despite (33), Halle and Marantz (115) maintain a Morphological Component. “DM recognizes a level of grammatical representation with its own principles and properties,” namely “MS.” To answer (34) properly, the first topics of this course focus on prototypical behaviors of bound morphemes, including the statement of English Morphology (8). Should principles for these behaviors belong to some Morphology Proper (MS) or be distributed to other components? (i) Word order. Head placement principles that determine word order are more general than the division between syntactic and morphological domains. (My conference talk) (ii) Topic 2. English Morphology (8) is a purely syntactic interface principle determining how the outputs of syntactic structure are processed in English PF. (iii) Topic 3. Proper formal specifications of lexical items can account for several behaviors that some DM advocates have supposedly limited to and hence identified with bound morphemes, such as local dislocation, impoverishment and fission. No MS is needed. (iv) Topic 4. A fundamental characteristic of bound morphology, what H and M label “Merger,” is a special case of a more general syntactic principle of “Alternative Realization.” No formal aspect of “Merger” is actually limited to bound morphology. This course’s answer to question (34) is thus: Completely. There is no MS component or level. 8