Antisymmetry And The Lefthand In Morphology

Abstract As Kayne (1994) has shown, the theory of antisymmetry of syntax also provides an explanation of a structural property of morphological complexes, the Righthand Head Rule. In this paper we show that an antisymmetry approach to the Righthand Head Rule eventually is to be preferred on empirical grounds, because it describes and explains the properties of a set of hitherto puzzling morphological processes —known as discontinuous affixation, circumfixation or parasynthesis. In considering these and a number of more standard morphological structures, we argue that one difference bearing on the proper balance between morphology and syntax should be re-installed (rewith respect to Kayne), a difference between the antisymmetry of the syntax of morphology and the antisymmetry of the syntax of syntax proper. Key words: antisymmetry, Righthand Head Rule, circumfixation, parasynthesis, prefixation, category-changing prefixation, discontinuities in morphology. Resum. L’antisimetria i el costat esquerre en morfologia Com Kayne (1994) mostra, la teoria de l’antisimetria en la sintaxi també ens dóna una explicació d’una propietat estructural de complexos morfològics, la Regla del Nucli a la Dreta. En aquest article mostrem que un tractament antisimètric de la Regla del Nucli a la Dreta es prefereix eventualment en dominis empírics, perquè descriu i explica les propietats d’una sèrie de processos fins ara morfològics —coneguts com afixació discontínua, circumfixació o parasíntesi. Considerant aquestes i altres estructures morfològiques més estàndards, proposem que una diferència que té a veure amb l’equilibri propi entre morfologia i sintaxi s’hauria de reprendre (rerespecte a Kayne), una diferència entre la antisimetria de la sintaxi de la morfologia i la antisimetria de la sintaxi de la pròpia sintaxi. Paraules clau: antisimetria, Regla de Nucli a la Dreta, circumfixació, parasíntesi, prefixació, prefixació amb canvi de categoria, discontinuïtats en morfologia. *The research reported here is part of the research program defined and financed by the Utrecht institute of Linguistics-OTS. I wish to single out Peter Ackema, Hans den Besten, Reineke Bok, Hans Broekhuis and Jan Don for inspiring discussion: without them, section 4 would not have existed. CatWPL 7, 1999 71-87 Antisymmetry And The Lefthand In Morphology* Frank Drijkoningen Utrecht Institute of Linguistics-OTS. Department of Foreign Languages Kromme Nieuwegracht 29. 3512 HD Utrecht. The Netherlands [email protected] Received: December 13th 1998 Accepted: March 17th 1999 CatWPL 7 071-087 13/6/00 12:26 Página 71 1. Introduction This paper is inspired by Kayne’s work on the antisymmetry of syntax (Kayne 1994), and more in particular the section in which he claims that his syntactic theory derives the ordering facts that are ordinarily accounted for by the morphological Righthand Head Rule. In this paper we look into a number of standard morphological structures which are generally discussed by morphologists working with the Righthand Head Rule. The basic question is whether there are empirical facts that might favour or disfavour the antisymmetry approach. We have put theemphasis on the empirical side because —after all— Kayne shows that the Righthand Head Rule (henceforth RHR) can be subsumed under the antisymmetric principles, but he does not fully elaborate the morphological issues concerning righthand / lefthand members of morphological structures. In the first section we will discuss two classical formulations of the RHR, and we show that antisymmetry cannot deal directly with most of the traditional lefthand members of a morphologically complex word. The conclusion of this section is that the syntactic antisymmetry approach is —untill further notice— empirically inferior to the morphological approach. In the second section we will discuss one particular group of morphological affixes, discontinuous affixes. They cannot be explained by the RHR, while we show that the discontinuity can be explained by the antisymmetry approach. The conclusion of this section is that the syntactic antisymmetry approach is empirically superior to the morphological RHR approach. In elaborating the theory for these particular discontinuous affixes, we encountered a number of problems, both of a conceptual and a technical nature. Section 4 discusses these in a brief way; eventually the solutions proposed in this section bear on all properties of morphology. In section 5, finally, we show that the proposals made for independent reasons in section 4 also allow us to explain the problematic cases of the first section in a relatively straightforward way. The grand total of all this will be that the antisymmetry approach is empirically superior to the morphological RHR, but on the other hand also that the morphological RHR is strongly confirmed as a property relevant at a surface-level —as a property of the string passed to the phonological component. 72 CatWPL 7, 1999 Frank Drijkoningen Table of Contents CatWPL 7 071-087 13/6/00 12:26 Página 72 1. Introduction 2. Two classical formulations of the Righthand Head Rule 3. Circumfixation 4. Problems and solutions 5. Prefixation 6. Conclusion References 2. Two classical formulations of the Righthand Head Rule 2.1. Williams’ RHR The first formalization of the (ancient) idea that the suffixal part of morphologically complex words is the element that determines the category of the word can be found in Williams (1981), cf. (1). Two examples are given in (2). (1) In morphology, we define the head of a morphologically complex word to be the righthand member of that word. The rule in (1) is part of the formalism, in the sense that it determines the label of the top node, in a «bottom-up» manner. In more recent terminology we would say that, during the merger of two morphological constituents, the righthand constituent is the constituent that is determining the label of the constituent created by merge. On the empirical level, Williams (1981) acknowledges the existence of «systematic exceptions», like ennoble and enlarge —the more traditional class of «category-changing prefixation». It is clear that these form a rather fundamental problem for the RHR: they also exist in the Romance languages, and in Germanic languages they are abundant. On the empirical level, it is also well-known that English and Germanic compounding is directly accounted for. Compounding in Romance, however, presents another class of rather fundamental problems. Examples are timbre-poste (litt. Stamp-post, post-stamp) or essuie-glace (litt. Whipe-windshield, windshield-whiper). Finally, the rule is limited to prefixation and suffixation, but does not deal with the traditional cases of «infixation» or «circumfixation». We will return to circumfixation below —it is also known as «parasynthesis». 2.2. Selkirk’s RHR The criticisms directed against Williams (1981) led Selkirk (1982) to the formulation of the Righhand Head Rule given in (3). Antisymmetry And The Lefthand In Morphology CatWPL 7, 1999 73 (2) a. A b. A V Asuf --pref A read able im proper (3) In a word-internal configuration: Xn P XmQ where X stands for a syntactic feature complex and where Q contains no category with the feature X, Xm is the head of Xn. CatWPL 7 071-087 13/6/00 12:26 Página 73 The examples given above in (2) are treated in the following way. If the topcategory is A in (2), then the head of this constituent is the rightmost element with the category A, hence the suffix (in 2a), or the word (in 2b). This rule is formally different from the one proposed by Williams. It works in tandem with an X-bar rule system, which works in a «top-down» manner. Eventually, it is this top-down property that enables Selkirk to account for category-changing prefixation and a number of the Romance compounds, as illustrated in (4). With respect to (4a), if the top node is known (V), then it is necessary for X-bar theory that there be a verbal head, and it is true that this verbal head (the prefix) is the rightmost element that has a category matching the category of the top node. In essence, category-changing prefixation is solved, but the formalism looks very rich (it lets intervene both X-bar theory and RHR). With respect to (4b), this example is solved if one takes only into consideration the category. However, it does not explain the more interpretative properties of the compound (the windshield remains the object at an interpretational level). In addition, Selkirk’s approach does not solve all Romance compounds: if the categories of the two members are identical, as in the case of timbre-poste, the prediction is wrong (poste would be the head according to (3)). Hence, Romance compounding is not really solved. 2.1.3. Di Sciullo & Williams’ Relativized RHR After these two classical formulations of the RHR, Di Sciullo & Williams (1987) defined another one, a Relativized RHR. In their analysis, the morphological order in (4b) reflects the syntactic word order; but they do not address examples of the type in (4a). We will not further discuss this rule here. The reason for the absence of discussion here is that the notion of «relative» head in morphology is incompatible with syntactic structures in the first place (syntax in the GB sense, the situation could be different in a HPSG framework). In other words, «relativeness of heads» presents an issue that goes beyond properties of linear order. 2.4. Kayne’s approach to the RHR Kayne (1994) claims that the antisymmetry approach that he developed for fully independent syntactic structures subsumes the RHR. As an illustration he gives the structure in (5). 74 CatWPL 7, 1999 Frank Drijkoningen (4) a. V b. N Vpref A V N en noble essuie glace whipe(r) windshield CatWPL 7 071-087 13/6/00 12:26 Página 74 The string can only be linearized properly (in the formal interpretation) if the result of moving J to K results in the label K for the morphological complex. It is his Linear Correspondence Axiom ultimately that assures (5). On a less abstract level, the LCA assures (6a). In practice, words like the one in (6b) derive from a deeper source with two independent heads. (6) a. [H1P H1 [H2P H2]] => [H1P H2i+H1 [H2P ti]] b. modern+ize <= [ ize [ modern ]] The claim is a very interesting one, in the sense that the quite general observation that all syntactic movement is upward-leftward, is directly relevant for the righthand headedness of structures making use of Head-movement. It cannot be accidental that leftward-upward movement of Heads generates orders that morphologists tend to explain by the RHR. Let us first reconsider the cases that Williams (1981) claimed to be «systematic exceptions» —category-changing prefixation. Basically, there are two possibilities, each illustrated in (7). (7a) gives the correct category for the prefixed verb: it is a verb derived from an adjective. However, Kayne’s formalisations work in a way in which only leftadjunction is allowed. Right-adjunction is barred, and therefore the correct order en+noble cannot be derived. Conversely, if one resorts to the permitted left-adjunction, (7a) surfaces as the ungrammatical noble+en. In other words, this class of verbs cannot be derived by using (7a). (7b) on the other hand can give the right order of elements through left-adjunction (en+noble); but (7b) is not the correct representation because the category of the whole is not correct. All in all, category-changing prefixation cannot be dealt with in this fashion. In addition, questions may be raised with respect to all prefixes. One of the attractive properties of Williams’ rule was that it explained why it is impossible to determine the category of a prefix —the prefix is always to the left of the head. Consider then (8). Antisymmetry And The Lefthand In Morphology CatWPL 7, 1999 75 (5) Q K/*R S J K T over turn (7) a. VP b. AP V AP A VP en noble A V noble en CatWPL 7 071-087 13/6/00 12:26 Página 75 (8a) illustrates the same problem as (7a). Moving proper to the prefix in the allowed way —left-adjunction— gives the wrong linear order: * proper+im. Rightadjunction is not allowed, so the ordering problem remains. In addition, a decision has to be taken as to the category of the prefix; perhaps Neg for this case (but this is quite troublesome: improper does not seem to function in syntax as a NegP; it simply seems to function as an adjective). In fact, (8b) fares better: left-adjunction derives the correct order, and the category of the complex word is correct. Next to a decision to take for the category of the prefix, one would be forced to conclude that prefixes are in fact the original head of the complement —that a NegP is neither an adjunct, nor a specifier, but a complement. This has a counter-intuitive flavour we will not elaborate here (e.g. the scope of negation). And the counter-intuitive flavour specifically holds for this class of prefixes, not for all cases (it can be a viable solution for a number of cases (see last section of this paper). As a conclusion to this section, Kayne’s antisymmetry explanation of the RHR is at the very best empirically equivalent to Williams’ formulation of this rule. It is subject to the same type of counter-examples, and looks less valid if the ordinary prefixes are not treated more extensively. In section 5, we will return to the prefixes. First, in section 3, we will show that the antisymmetry approach is capable of treating a class of examples which has been beyond the scope of the RHR to begin with. 3. Circumfixation In a phrase structure rule system, circumfixation (also called «parasynthesis») has been described as a morphological instantiation of the general schema of (9). (9) A -> B A C One of the clearest examples is Past Participle Formation in Dutch and German. Initially, this morphological rule simply looks like (10). One takes a verb, adds a prefix and simultaneously also a suffix. (10) Past Participle Formation: V -> ge + V + t ex. pak -> ge + pak + t It is obligatory to add both affixes, as illustrated in (11). This requirement can be formulated in a generalized way, as in (12). (11) ge+huil+d *ge+huil *huil+d (‘cried’) ge+werk+t *ge+werk *werk+t (‘worked’) (12) B A C * B A *A C 76 CatWPL 7, 1999 Frank Drijkoningen (8) a. ?P b. AP ? AP A ?P im A proper ? proper im CatWPL 7 071-087 13/6/00 12:26 Página 76 The subject has been commented as follows, by Spencer (1991). (13) Some morphologists regard such duets as a special kind of discontinous affix, a circumfix. Many linguists would argue that all cases of alleged circumfixation can be reduced to suffixation and concomitant prefixation. With respect to these comments, which tend to deny the relevance and the mere existence of circumfixation, it is relevant to note two things. Although the informal description «concomitant» is close to the intuition of the native speakers, it is not formalized. That is, although one might agree on the fact that suffixation and prefixation are concomitant in circumfixation, morphological theory never has given a specific formalisation of the relation between the two affixes that apparently exists. In that sense, «many linguists» should try to formalize the very notion of «concomitant». We will argue in what follows that «concomitant» can be interpreted as a Spec-head relation. In these comments it is also rather clear that the problem posed by circumfixation stems largely from the discontinuity of the process. It is known that morphological structures cannot deal with discontinuities. But —crucially— syntactic theory can deal with discontinuities; in fact, one might say that large parts of movement theory are motivated by (linearly) discontinuous relations in structures. At this point, one of the crucial properties of Kayne’s antisymmetry approach is important: it is based on movement. Hence, deriving the RHR from movement theory opens the road for the explanation of morphological discontinuities. This is in fact what we argue for here. Past participle formation in Dutch and German has been the topic of discussion in syntax, in phonology and in morphology. Although not all problems have been solved, the syntactic properties of past participles are clear in isolation; the same holds for the phonological properties of past participles. It is the morphology (and the position of the morphological component) that is subject to discussion. In syntax, past participles are considered to consist of one form, but with several interpretations. In this respect, consider (14). (14) a. Hij heeft het boek gelezen (perfective participle) He has read the book b. Het boek wordt gelezen (passive participle) The book is being read c. Het boek is aangekomen (perfective participle ergative) The book has arrived d. Het door iedereen geciteerde werk (perfective participle passive/adjectival) The [by everyone cited] work Current syntactic theory postulates a functional head PART (or P.M.) —independently of issues of checking vs. building. In most analyses, an additional AGR relationship is postulated —independently of the issue of whether this AGR is an Antisymmetry And The Lefthand In Morphology CatWPL 7, 1999 77 CatWPL 7 071-087 13/6/00 12:26 Página 77 independent head or not. In general, the interpretative differences are formulated in terms of absorption of case / theta features —no absorption in perfectives, the Burzio-generalisation for the others. There is also discussion about the categorial status, V or A, which we do not engage upon here (for sake of concreteness, assume A for PART, licensing an Event-variable in the sense of Grimshaw (1990) for Event-nominalizations, see also Drijkoningen (1997)). In phonology, on the other hand, PART is neither one head nor one element. There are simply two affixes. As far as (morpho-)phonology proper is involved, there is no evidence whatsoever for the link between the affixes. All rules function properly by taking the affixes in isolation. In this respect, consider (15). (15) a. De gewerkte (*gewerkde) dagen. (‘the worked days’) Het vertelde (*vertelte) verhaal. (‘the told story’) b. Hij heeft het hout bewerkt (*ge+be+werk+t) He has the wood be-work-ed (GE+BE+work+D) = He has worked on the wood For (15a) one needs a phonological rule that deals with the choice between /d/ and /t/. This rule is dependent on the voice of the final consonant of the verbal root, but ignores the prefix. In fact, the rule generalizes with the voice-alternation found with the suffixes for past tense (without prefix in the first place). In (15b) there is a phonological rule that may delete the prefix part of the past participle; this rule ignores the suffix, and only takes into consideration the other prefix in the immediate environment. To sum up, syntax functions properly while having one functional head (although there are two affixes), while phonology also functions properly while having two independent affixes (although they relate to one head). In other words, past participle formation is a typical morphological problem —or an interface problem. Shortly above we noted two important problematic aspects of circumfixation —the «concomitant» nature of the affixation process and the «discontinuity» of the affixes. Another problematic aspect can be added to these. In general, it has been impossible to draw a structure of the standard type. That is, one might propose [[ge+V]+t], but this one cannot be correct because [ge+V] is not an independent constituent. Alternatively one might propose [ge+[V+t]], but this one also cannot be correct for the same reason, [V+t] is not an independent constituent either. Should one deny the relevance of the criterion —the constituenthood of the part— then the problem comes back in taking a decision as to which one of the two should be correct. The cause is simple: Both subparts in isolation are ungrammatical —as shown in (11). As a consequence of this, one might propose ternary branching. But ternary branching is no longer standard, i.e. theoretically disallowed. Hence, there is no tenable structure available for past participles. Now, if there is no structure available, past participles might illustrate a typical interface situation. This track has been defended by Don (1993). He gives the following picture. 78 CatWPL 7, 1999 Frank Drijkoningen CatWPL 7 071-087 13/6/00 12:26 Página 78 As far as syntax goes, there is no problem, there is one functional head. As far as phonology goes, there is no problem either, there are two independent affixes. The problem is the linking between the two components. Don’s proposal illustrates crossing association lines. For a theory of this type, it is worthwhile to note that crossing association lines are not attested elsewhere and probably only occur with circumfixation. Again, the solution does not look principled. We conclude by saying that although phonology and syntax function properly in their own independent way, there is no principled truely morphological analysis of circumfixation. As we have indicated above, deriving and explaining morphological ordering by linearization axioms includes the possibility of movement —hence, that movement might explain discontinuities in morphology. As we have indicated also above, using syntactic relationships like Spec-Head Agreement in morphology might explain the «concomitant» nature of the circumfixation process. This gives us the following proposal —which is subject to additional comments in section 4. As discontinuity is explained by movement, there are necessarily two representations. Movement falls within the boundaries set by the theory (Head-Movement Constraint, and for antisymmetry: left-adjunction). For syntax, the one functional projection that is needed is present; it has the complement it normally takes in syntax. For phonology, there are two affixal pieces, as desired. The relationship between the two pieces is encoded by Spec-Head Agreement, by using the Specifier position of PART. This means that we formalize the intuition «concomitant» as a syntactic relation between two morphological pieces, not a phonological one. As a result, the two most important problems dissolve. In addition, the additional morphological problem also dissolves. The suggestion of ternary branching need no longer be entertained; the structure is binary. What led to the suggestion of ternary branching (the problem of the constituenthood of the parts) also has a clear answer: at surface-level the structure is [ge+[V+t]]; [[ge+V]+t] is excluded. Antisymmetry And The Lefthand In Morphology CatWPL 7, 1999 79 (16) Morpho-syntax V-STEM PART Morpho-phonology Pref V-STEM Suf (17) Proposal (first version) PART” →PART” SpecjPART’ SpecjPART’ prefix prefix PARTjVP PARTjVP suffix V V PART V huil huilisuffix ti CatWPL 7 071-087 13/6/00 12:26 Página 79 complement. More in particular, an update of the analysis by Hoekstra et al. (1987) gives the possibility in (27a), for the example in (27b). (27) a. [ Xpref + Y [ tx]] b. dat hij [AGROP de wegi[bejwandelt] [SC titj]] that he the roads BE +walks ‘that he walks the road’ Second, the difference in compounding between Romance and English / Germanic that remained problematic also in previous accounts, can be linked up to a difference between overt and covert movement, under both Kayne’s and my approach. Under our interpretation of morphological structures (as discussed in section 3), the order difference in (28a) can be analyzed by (28b) and (28c). (28) a. screwdriver - tourne-vis b. [N0max screwj driveier [V0max ti[N0max tj]] c. [N0max ——— tournie [V0max ti[N0max vis]] That is, compounding of this type illustrates N0min movement to the Specifier of a higher N0max, overt in French, covert in English. The basic difference between Kayne’s analysis of these and ours is that we do agree with e.g. Di Sciullo & Williams (1987) in considering morphologically complex words different from objects constructed by syntactic rules at some level. With respect to (28), data of the type in (29) provide the distinction between Kayne’s interpretation of the LCA with respect to the RHR and ours, given in section 3. (29) a. I bought a * [lot of screws - driver] b. J’ai acheté un * [tourne-beaucoup de vis] c. I need it. (it = screwdriver) (*it = screw) d. J’en ai besoin. (en = tourne-vis) (*en = vis) In sum, suggestions as to the existence of «maximal» heads vs. «minimal» heads (e.g. Chomsky 1995) prove useful for morphology —and allow us to make use of the LCA inside morphological structures while maintaining a theory in which morphology and syntax remain distinct modules. References Ackema, P. (1995). Syntax below zero. Utrecht: Ph.D.Diss. Bok-Bennema, R. (1996). «Dutch Verbalization and the Theory of Morphology». Yearbook CLCG 1995, Groningen. Chomsky, N. (1995). The minimalist program. Cambridge Mass: MIT Press. 86 CatWPL 7, 1999 Frank Drijkoningen CatWPL 7 071-087 13/6/00 12:26 Página 86 Di Sciullo & Williams (1987). On the defintion of Word. Cambridge Mass: MIT Press. Don, J. (1993). Conversion. Utrecht: Ph.D. Diss. Drijkoningen, F. (1997). «Morphological Strength: NP positions in French». In Beermann, D.; Leblanc, D.; van Riemsdijk, H. (eds.). Rightward Movement. Amsterdam: Benjamins, p. 81-114. Grimshaw, J. (1990). Argument Structure. Cambridge Mass: MIT Press. Hoekstra, T., et al. (1987). «Complexe verba». GLOT 10. Kayne, R. (1994). The antisymmetry of Syntax. Cambridge Mass: MIT Press. Neeleman, A.; Schipper, J. (1992). «Verbal prefixation in Dutch: Thematic evidence for conversion». Yearbook of Morphology 1991. Dordrecht: Foris. Selkirk, E. (1982). The syntax of Words. Cambridge Mass: MIT Press. Spencer, A. (1991). Morphological Theory. Oxford: Blackwell. Williams, E. (1981). «On the notions “lexically related” and “head of a word”». Linguistic Inquiry. Antisymmetry And The Lefthand In Morphology CatWPL 7, 1999 87 CatWPL 7 071-087 13/6/00 12:26 Página 87