Parameters of Inflectional Heads

Wim de Geest (K.U.Brussel)
Eric Hoekstra (P.J. Meertensinstituut)
Guido Vanden Wyngaerd (K.U.Brussel/FWO)

Linguistics 35, 995-1001.

The title of this special issue, ‘Parameters of Inflectional Heads’, reflects a number of changes in attitude toward linguistic description. On a global level, linguists are no longer in a position where they can afford to describe facts from one language without paying attention to insights gained from the analysis of other languages. In comparing different languages, we resort to our semantic intuitions. Thus if two sentences mean the same thing, we in principle assume that they must have certain elements in common, such as elements of the syntactic structures underlying them, as well as elements of their semantic representation. We may see a parallel here with certain branches of traditional grammar, which held that there is a uniform semantic system underlying all natural languages (see e.g. Chomsky 1995). The question which then naturally arises is: how are the differences between parallel structures in different languages accommodated? This question received an answer with the introduction of the concept of parametrization. This brings us to a second change, which has taken place at a more detailed level.
Let us consider this second change by looking at the model outlined in the work that introduced the concept of parametrization, Lectures on Government and Binding (LGB) (Chomsky 1981). The idea was that the child came equipped with a Universal Grammar (UG), which had a number of switches that needed setting, the parameters. For example, word order differences between languages were accounted for by assuming a directionality parameter for government: in VO languages the verb governed to the right, in OV languages it was assumed to govern to the left. Furthermore, the theory of parametrization in LGB had a highly deductive character. The idea was that everything was connected to everything else, so that, say, a different value for the pro-drop parameter not only caused different behavior with respect to overt or covert pronominal subjects, but also affected WH-movement, postverbal subject placement, expletives, and perhaps other properties as well. Empirically, a parameter would make itself visible as a clustering of seemingly unrelated facts. Linguists thus expected to find conditional universals of the type sketched above. However, no convincing clustering of facts survived the test of time. Pro-drop itself turned out not to be a binary property. There are all sorts of in-between cases, e.g. some languages only drop weather subjects and expletive subjects but not referential subjects, other languages only drop certain types of expletive subjects, etc. Theoretically, although the LGB theory of parameters had rich deductive structure, it was also unconstrained in one important respect: the framework offered no format for parametrization, in so far as anything could be a parameter.
Minimalism (Chomsky 1995, Kayne 1994) embodies a radical departure from this theory of parametrization. On the positive side, all parameters have a uniform format, but this is matched on the negative side by the near-total absence of deductive structure. Let us discuss these two issues in turn, starting with the former, the uniform format of parameters. Parameters in minimalist theory specify whether or not movement to a functional head or specifier position takes place in PF (overt movement) or in LF (covert movement). The values for the parameters are associated with these functional heads or their specifiers. For example, word order differences between languages no longer follow from the parameter setting the directionality of government, but rather reduce to some fairly epiphenomenal property of movement rules, viz. their locus of application (pre- or post-Spell-Out). Apart from this, there is very little variation between languages. At the underlying level, all languages are assumed to have Spec-head-complement order (Kayne 1994), and the difference between VO and OV order is assumed to be a result of covert or overt movement of the object to a specifier position to the left of V, required to check off Case features against a functional head. At the final level on the semantic branch of the grammar, all languages are alike as well, movement rules applying uniformly across languages; thus in a language with VO order, the object will likewise move to the left in order to check off its Case features against a functional head, but since this movement takes place covertly, its effects are not visible on the PF side. Whether movement has to occur overtly or covertly is determined by the strength of the features involved: strong features trigger overt movement, weak features require covert movement. Hence all parametric differences between languages are ultimately expressible in terms of the properties of features of heads. The further question that then arises is how strength and weakness of features is determined. This property of features does not seem to be subject to independent confirmation, other than the fact of whether the feature triggers overt movement or not: hence a feature is strong because it triggers overt movement, and it triggers overt movement because it is strong. If indeed features are the ultimate primitives of syntactic change and differentiation, this is not necessarily undesirable: one in fact expects certain differences between languages to be arbitrary, and therefore unreducible.
This discussion has revealed a further property of parameters. The principle of Full Interpretation entails that all moved elements will have to participate in feature-elimination, which can only take place if the moved element lands sufficiently close to the inflectional head against which it is checked. The idea is that features are illegitimate objects at the interface levels, and must consequently delete; this they can do by finding an identical feature elsewhere in the tree. We may refer to feature-elimination as Head-agreement, since both moved maximal projections and moved heads will agree with the head of the projection to which they adjoin. Head agreement may thus be viewed as a uniform property in the theory of parametrization, since parametrization always involves properties of Head agreement. Essential, then, to the minimalist program, is the investigation of the properties of inflectional heads, both which respect to the features they host and with respect to the question of what the syntactic effects are of these features. Inflectional heads are not infrequently realized as affixes on stems or lexical heads, and it should therefore come as no surprise that there has been a renewed interest in the detailed properties of morphological paradigms.
Let us then turn to the weakness in the minimalist theory of parameters, its low performance on the issue of deductive structure. Very few of the parameters proposed in minimalist writings manage to link various empirical phenomena in the way the pro-drop parameter did, for example. This is in part a consequence of the approach that takes properties of heads to be the primitives of syntactic change, in that it has almost become the default case to write extremely construction-specific parameters. That is not necessarily bad: construction-specific parameters are the first step towards discovering construction-independent ones, once we see that a common pattern shows up in the description of several construction-specific parameters. However, this process will take some time: in the meanwhile there is no theoretical view on this matter which can be evaluated. On the negative side, then, it seems that minimalism does not incorporate any hypothesis about the ways in which parameters are linked, i.e. there is as yet little deductive structure to the parameters that sail under the minimalist flag. However, this is not a necessary property of minimalist parameters. In fact, some of the more interesting proposals, of which this issue contains a few specimen, attempt to relate differences between languages, as well as language-internal differences for that matter, to different types of derivations that may or may not be available. In this way, we are again beginning to see the outlines again of a theory of parametrization with rich deductive structure.
Such an evolution toward theoretically more interesting analyses is not unprecedented. As is frequently the case when a new framework for analysis is put in place, there is an initial bias for data-oriented work, as one tries to fit both old and new data into the new theoretical mould. In the field of syntactic categories we have over recent years witnessed a huge explosion of primitive elements and features. Thus in the wake of Pollock’s seminal proposal to assign tense and agreement features to separate syntactic head positions (those heading TP and AgrP, respectively), the theory has seen an explosion of such functional heads: apart from the familiar AgrP (later split into AgrSP and AgrOP), NegP and TP, more recent additions to this inventory include Asp(ect)P, Num(ber)P, Deg(ree)P, Foc(us)P, M(ood)&M(odality)P, etc. In the same vein, the minimalist program has lead to a seemingly unchecked increase in the number of new features proposed to account for the various movement rules. This has worried some linguists. ‘Do we need so many functional categories/features?’ they ask. The table of chemists, which after all deals with dull matter and not with human thought, counts over a hundred elements. It is only to be expected that the stuff human minds are made of will incorporate many more than a hundred elements. Most functional structures do not have more than ten elements anyway. Furthermore, the number of primitives can be reduced, as in the chemists’ table, by attributing to each element an internal structure, which represents a particular combination of an extremely restricted set of primitives. As far as linguistics is concerned, we have as yet little idea what these primitives would look like, but the possibility cannot be dismissed offhand. A more adequate objection to the explosion of functional projections and features is not that there are too many, but that they do not give us more insight into various complex empirical phenomena, but merely describe the structure of a simple sentence. As times goes by, it is to be expected that valuable insights stated in the new terms will make their appearance.
The articles in this issue are all concerned with shedding light on the parametric properties of inflectional heads. They have been written from a generative point of view, making use of the formal apparatus supplied by generative theory. The patterns which are traced in the facts, however, are accessible also for syntacticians with a different background. A mere survey of the example sentences will reveal a treasure of facts and generalizations, which puts to shame any claim, sometimes heard over a glass of beer, that generative syntacticians are not interested in facts. The problem rather is that we have too many facts crying to be explained. Theoretical apparatus is intended to aid us in the process of gaining some insight in the patterns which are present in the facts and which we are otherwise unable to see. If too much emphasis is sometimes put on the theoretical apparatus, then this is logical reaction to the fact that theoretical linguistics is suffering from the sheer amount of facts (a point also made by Muysken 1984:293), more than for example particle physics or sociolinguistics, in which it takes four years and a lot of money to conduct the experiments providing a few new facts which will confirm or falsify a relatively small and simple set of hypotheses. As Muysken notes, investigation of hypotheses of comparable complexity within syntax requires a concentrated effort taking up less than a full working day.
In the area of the properties of functional heads, then, Hans Bennis, Frits Beukema and Marcel den Dikken, in their paper Getting Verb Movement, compare the diachrony of English with subsequent developmental stages in the acquisition of French, noting that the acquisitional order of the placement of negative elements in French is opposite to the order of the placement of such elements in the historical development English. To account for the data, they propose two parameters. On the one hand, they make use of a weak or strong specification for the feature [+finite]. This feature is in Tense, and correlates with the presence or absence of overt V-movement to Infl. In addition, the relation between inflectional functional heads and V is parametrized: the licensing head for V is T in the adult languages, but Mod(ality) or Asp(ect) in early child language.
Jonathan Bobaljik’s paper If the Head Fits contains an analytical investigation of subject agreement paradigms in Icelandic and English, with some extensions to other Germanic languages (German, Dutch, Yiddish, Afrikaans and Faroese). It is claimed that the type of agreement (analytic or synthetic) correlates with the type of syntactic derivation which a sentence allows, such that the presence of fused or synthetic morphology in a language correlates with the presence of object shift and the transitive expletive construction. The relevant parameter thus involves the type of agreement which the verb exhibits, which correlates with the different types of derivation. The virtue of Bobaljik’s paper is that it proposes a parameter linking various empirical phenomena in a way reminiscent of the early LGB days.
Marcel den Dikken and Eric Hoekstra present a detailed study of the phenomenon of Parasitic Participles in Frisian, arguing that there is a parameter differentiating languages with recursive feature checking for a given feature from languages without feature checking. Recursive feature checking is morphologically directly observable since it leads to parasitic morphology: in Frisian we find in irrealis constructions two (or more) past participles licensed by one occurrence of HAVE. The Frisian facts provide a counterexample to a rigid conception of the Head Movement Constraint (HMC). This does not lead the authors to abandon the HMC completely: they propose a relativized HMC which covers both the classical cases which the original HMC was designed to capture as well as the counterexamples to it (including the ones from Serbo-Croatian).
Joost Zwarts investigates the behavior of prepositions in Dutch in his Complex Prepositions and P-stranding in Dutch. Two classes of prepositions are distinguishable on empirical grounds. This leads to a theory in which prepositions can either be simple or complex; in the latter case, the preposition incorporates the lexical head of its complement. Complex prepositions can be nominal or verbal, depending on the category of the lexical item which was incorporated. Simple prepositions in Dutch are parametrically specified for a strong R-feature, which accounts for the occurrence of the so-called R-pronouns, which are morphologically distinct from ordinary pronouns (e.g. *op het ‘on it’ is realized as er op ‘there on’).
Overlooking the series of four articles brought together in this issue, one cannot fail to notice that each of it is concerned with proposing a specification of a parameter for some specific inflectional functional head. Thus the concept of grammar now dominant is one in which every relevant feature is given its own place in syntactic structure, usually, but not necessarily, as a functional head. This is different from the early eighties, in which a modular view of grammar reigned, which was not accompanied by any theory of parametrization. Conceivably, too much emphasis is perhaps put nowadays on parametrization. However, we believe that this is a necessary stage in the development of syntactic research, as we are only just beginning to come to grips with the dimensions along which languages and dialects differ from each other.
The focus on heads in syntax began emerging at the end of the eighties, with the advent of minimalism, as is clear from the introduction to the collection of papers brought together in Den Dikken & Beukema (1991). However, in that collection the focus is not on the parametric properties of heads. Instead, the presence or absence of specific inflectional heads itself is hotly debated, especially in relation to the question of whether words are formed in syntax or not. In the present collection, the focus is rather on the parametric properties of inflectional heads, and the results of intra- or interlanguage typological comparison which they are designed to explain. This difference in focus is presumably at least partly due to the influence of the minimalist program, of which the effects are gradually affecting syntactic research in the nineties.
The papers contained in this issue are a selection of the papers presented at the tenth Comparative Germanic Syntax Workshop organized jointly by the P.J. Meertensinstituut of Amsterdam and the Katholieke Universiteit Brussel at the latter institution in 1995. For a variety of reasons, the papers of the following speakers at the workshop have not been included in this issue: Sjef Barbiers, Daniel Büring and Katharina Hartmann, Martin Everaert, Erich Groat, Eric Haeberli, Alison Henry, Teun Hoekstra and Johan Rooryck, Ellen-Petra Kester, Sergio Menuzzi, Christer Platzack, Tarald Taraldsen, Fred Weerman, and Jan-Wouter Zwart.


Chomsky, N. (1991) Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, N. (1995) The Minimalist Program. MIT Press, Cambridge, Massachusetts.
Dikken, M. den & F. Beukema (1991) “Heads—An Introduction”. The Linguistic Review 8, 107-117.
Kayne, R. (1994) The Antisymmetry of Syntax. MIT Press, Cambridge.
Muysken, P. (1984) “Repliek” [Reply]. Interdisciplinair Tijdschrift voor Taal- & Tekstwetenschap 4, 291-293.