06nm,   PDF
Mark Newson and Vili Maunula

0

Word Order in Finnish: whose side is the focus on?*

Introduction

In this paper we discuss certain facts concerning the organisation of basic elements of Finnish sentences. Like Hungarian, Finnish demonstrates relatively free word order of the major grammatical functions (S, V and O), though, also like Hungarian, most of these alternate patterns come with semantic consequences. At that point the similarity between Hungarian and Finnish stops. We will show that the semantic effects achieved by the various word orders in Finnish are based on rather different notions to Hungarian and also that the principles that determine the ordering of elements tend to be of a different type. This is of interest not only because of the comparison between the two genetically related languages, but also because what it has to tell us about possible organising principles within linguistic systems and how these may vary from one language to another. The paper is organised into three parts. In the first we discuss some general aspects of grammatical systems and propose a restrictive typology of constraints, detailing their properties and the kind of linguistic phenomena they are responsible for. In the second part we present the facts about Finnish word order, and in the third we propose an analysis of the data using constraints of the proposed kinds.

1

Constraint types

If we take a linguistic system to be a ranking of the kind of alignment constraints argued for in Newson (2000, 2002, 2003, 2004), Gáspár (2005) and Newson and Gáspár (2001), we expect word order to result from the competition of elements to achieve their optimal position with respect to other elements. Essentially there are two kinds of alignment: elemental constraints, in which an element, the target, is aligned with respect to another, the host, either to its left or its right (1a); or domain based constraints in which the target

We would like to express our thanks to the audience of the Budapest Phonology Circle and Linguistics Discussion Group, where we first presented a version of this paper. In the present version, we have tried to respond to comments made there and hope that in doing so we have clarified a number of issues which we previously either did not express well or, more likely, had not thought through enough. Any improvements in this paper therefore owes much to the searching questions of that audience. Any remaining inadequacies are entirely our own fault. The Even Yearbook 7 (2006), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 1218–8808, http://seas3.elte.hu/delg/publications/even, © 2006, Mark Newson and Vili Maunula

Word order in Finnish 22

is aligned with respect to a set of elements, the domain, either in first or last position (1b): (1) a target host or host target b target x1 … xn or x1 … xn target (where target, x1 – xn ∈ D) Let us adopt the following conventions for naming these constraints. In constraint names, lower case letters refer to elements; targets or hosts, and upper case letters identify domains. The symbols P, F, 1 and Ω are reserved for indicating the relationships ‘precedes’, ‘follows’, ‘is first in’ and ‘is last in’ respectively. Thus we have the following types of constraints: (2) tPh tFh t1D tΩD

target precedes host target follows host target is first in domain target is last in domain

In the competition, some elements will get to win and they will stand in the position determined for them by some highly ranked constraint. Inevitably there will be losers too. It is interesting to ask what happens to the losers: if an element cannot occupy its preferred position, what position does it occupy? There are two possibilities. In one, the element will accept the second (or lower) best position, but one that is still determined by the constraint that it fails to satisfy completely. For example, we know that objects in English like to be immediately behind the verb and will not accept second position to an adjunct: (3) a I spent the money yesterday b *I spent yesterday the money Thus, in a competition for the post verbal position, the object wins. However, when there is an indirect object, things do not go the direct object’s way and it has to give up its preferred post verbal position to the indirect object: (4) a I gave Smith the money yesterday b *I gave the money Smith yesterday Interestingly enough, however, the direct object still wants to be as near to the right side of the verb as possible, and although it has to give up first place and accept second, it will not give up entirely and accept third place to the adjunct: The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 33

(5) * I gave Smith yesterday the money This is expected if the constraint determining the object’s position is violable by degree. Thus, although being next to the verb is optimal, being two steps away is worse than being one step away: obFv1

(6) V Obj x y z V x Obj y z V x y Obj z V x y z Obj

The other possible strategy for a loser to adopt is to completely give up the competition for the position in question and to go and compete for another position determined by a lower ranking constraint. An example of this strategy might be the behaviour of the second wh-element in an English multiple whquestion. Wh-elements, as is well known, want to be first in the clause. Thus a wh-object will appear at the front of an expression despite the fact that as an object it wants to be after the verb. This shows that the wh-requirement is stronger than the object requirement: wh1S2 obFv you saw who

• !

who (did) you see

(7)

However, when there are more than one wh-element, they cannot all be first and thus, whatever the order, the wh-constraint will be violated. In this case it seems that the object wh-element gives up and behaves like an object instead:

1

This is meant as a demonstration only and in reality there probably is no such constraint which targets objects as objects are simply arguments which are not subjects. As such they conform to general conditions to do with the placement of arguments, but no condition to do specifically with objects. The ordering of objects amongst themselves is probably based on a thematic hierarchy. 2 The domain of the wh-element is the set of elements which it has scope over. This can be defined as the head of the interrogation, a predicate, and all its dependants. Thus, in who likes whistling, the domain is {likes, who, whistling} as likes is the head of the question and the other elements are dependent on this. In I asked who likes whistling the domain is still {likes, who, whistling} as the interrogation relates to the subordinate predicate and its dependents. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 44

(8) a b c d e

• what who did say
• who what did say
• what did who say
• who did what say

who said what

This indicates that the wh-constraint which positions wh-elements in English is not a gradient constraint but one which is either satisfied or violated: (9)

wh1S obFv who what saw

• !

what who saw

• !

who saw what

If the wh-constraint were a gradient constraint, then it would be better to have all wh-elements near the front of the clause and the object constraint would have no role in determining the position of the object. But what makes one constraint a gradient constraint and another not? We want to argue that this distinction is not random, but rather one which falls out from the nature of the constraints themselves. This is the point we will argue in the following section. 1.1 Precedence and adjacency Let us consider the case of direct object placement in English again, but this time including the case where the object appears on the other side of the verb: (10) obFv
give Smith the money yesterday

give Smith yesterday the money !
the money give Smith yesterday

By switching sides the object manages to get closer to its preferred position, to the immediate right of the verb, than it can by staying on the right but following the indirect object. Thus if all else were equal (which it of course is not) we might expect that the direct object would on occasions ‘switch sides’ in order to satisfy the relevant constraint. A vague feeling that this simply is not how language works and the observation that, in these cases, side switching just never happens lead Newson (2000) to suggest that side switching violates a constraint to a greater degree than if the element is further removed from its preferred position but still on The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 55

the appropriate side. It was proposed that being further from the preferred position, but on the appropriate side, earned the standard violation marks – one violation for each element that appeared between the aligned element and its preferred position. However, being on the wrong side earned another kind of violation mark, the ‘big star’ , and this was evaluated as being worse than any number of ‘little star’ violations: (11) obFv
give Smith the money yesterday

give Smith yesterday the money ! the money give Smith yesterday ! However, there is a general tendency in OT to dislike the idea of differential violation of constraints by different elements. That is, although, in a set of candidate expressions, any given element might violate certain constraints to a greater or lesser degree, or a constraint may be violated by more or fewer elements, the situation in which element x violates a constraint more than does element y under the same conditions is generally frowned upon. It is argued that this situation is better handled by having two different constraints, one relevant for element x and the other for element y, with the latter being ranked higher than the former. In this way, constraint violation calculations are made in a uniform way throughout the system, and the treatment of x and y is determined by constraint ranking. The big and little star system is somewhat like differentially violated constraints, and is therefore subject to similar criticisms. The way around this problem is to rethink alignment constraints in a way that, instead of having one constraint violated differently under different conditions, we propose two different constraints that are specifically violated by the two different conditions. Thus, instead of having constraints violated with big stars when the element is on the wrong side, and by little stars when it is not, we now propose two different types of constraint: one which is violated by the element being on the wrong side, the precedence constraints, and one which is violated by the element being further away from its host, an adjacency constraint. For the purpose of naming adjacency constraints, we reserve the symbol A to indicate the relationship ‘is adjacent to’. Here is an example of how the data presented in (11) could be handled with precedence and adjacency constraints:

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 66

(12) obFv obAv
give Smith the money yesterday

give Smith yesterday the money ! the money give Smith yesterday

• !

The higher-ranking precedence constraint ensures that the object remains following the verb and the adjacency constraint forces it not to stray too far from the verb. Note that the adjacency constraint is a gradient constraint in that the further removed the two elements are, the worse the violation incurred is. In principle, however, an adjacency requirement could also be non-gradient. Yet, this would lead to a more complex situation, requiring an extra stipulation that the relationship be not only adjacency, but immediate adjacency. We will shortly see that there is a reason why adjacency constraints are gradient, and hence there is no need to introduce more complex immediate adjacency constraints to the system. Precedence constraints, on the other hand, are absolute rather than gradient: you either precede or you do not. If we forced precedence constraints to be gradient, this would not only be more complex, introducing a notion of immediate precedence, but it would also be tantamount to building into them an adjacency requirement. As the adjacency requirement is already present in the form of the adjacency constraint, this would not only be pointless, but also detrimental to the system, as it would rob the adjacency constraints of their effects under certain rankings. Obviously, the above system replicates the effects of the big/little star system. However, without further stipulation the precedence/adjacency system predicts phenomena which were assumed not to be possible and indeed were part of the motivation for the big and little star system in the first place: side switching. This is due to the fact that with two constraints ranking becomes an issue. If we rank adjacency above precedence, it becomes better for the target to switch sides in order to remain adjacent to the host, rather than to remain on the dominant side but be further away: obAv obFv give Smith the money yesterday

• !

give Smith yesterday the money !
the money give Smith yesterday

(13)

This is not the way English works. But it is an empirical issue whether any language works in this way. If there is evidence that side switching does take place, then this would support the precedence/adjacency approach over the The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 77

big/little star system. We will see that the behaviour of the Finnish focus provides the crucial data. 1.2 Domains and adjacency The system we have so far built has taken the elemental alignment constraints, involving the positioning of a target with respect to a host, and identified three basic subtypes: xPy, xFy and xAy. This raises a question concerning the domain-based constraints, namely whether anything like an adjacency constraint could exist with respect to domains? Empirical evidence in favour of such constraints will be discussed later in this paper, but for now let us consider the technical aspects of such a possibility. Let us first broach the issue of what it would actually mean to be adjacent to a domain. The intuitive notion that we want to build on views domain adjacency as similar to element adjacency, but with the domain as a whole, rather than its individual members, functioning as the host. Consequently, being adjacent to a domain involves being adjacent to its edges. This would appear to somewhat resemble the first and last constraints, but with two important differences. First, domain adjacency, like elemental adjacency, is nondirectional. With elemental alignment, the direction of the alignment, precedence or succession, is a matter for the precedence constraints to determine. Similarly, with domain adjacency the ordering of the target with respect to the domain is a matter for the first and last constraints to decide, and adjacency merely places the target at either edge of the domain. Secondly, first and last constraints are naturally gradient in as much as for the first constraint the most optimal position is the domain-initial position, while the second best position is the position behind it, and so on. Domain adjacency, however, is not naturally gradual: a target is either adjacent to a domain (i.e. at its edges) or it is not. One might feel that this is not necessarily the case as a target can be placed nearer to or further away from a domain edge. Yet, this overlooks the fact that domain adjacency is not directional, and therefore not oriented towards one edge or another. If we view domain adjacency as gradual, we end up with a rather useless constraint that is violated to the same degree no matter where in the domain a target is positioned. Consider the situation in which there are n elements in a domain, including the target. Suppose that the target is placed at the leading edge of the domain, in front of all other elements of the domain. In this case, although it is adjacent to the leading edge, it is n-1 elements away from the following edge, incurring n-1 violations of the adjacency constraint. Suppose we then move the target to the following edge. Now, the target is n-1 elements away from the leading edge, again incurring the same number of violations. Finally, suppose that we move the target m (m<n) number of places back from the leading edge. It now has m The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 88

number of violations of the adjacency constraint with respect to the leading edge, but n-1-m violations with respect to the following edge. The total number of violations is therefore n-1-m+m, i.e. n-1, the exact same number of violations as in the previous cases. Clearly, such a constraint would never play a role in determining the optimality of any candidate as it treats all candidates equivalently, and therefore we must reject the view of domain adjacency in which the relevant constraints follow a gradient violation pattern. The conclusion that we therefore reach is that as a consequence of being non-oriented with respect to particular edges, domain adjacency constraints must be viewed as non-gradient and, as first and last constraints are so oriented, they may be gradient. Obviously, it would still be possible to stipulate that some first and last constraints are non-gradient, violated if the target is not first (or last). This situation, however, would be comparable to making precedence constraints non-gradient by building into them some notion of adjacency, complicating the system pointlessly. If first and last constraints are non-gradient, we then automatically build into them some notion of domain adjacency, which is already present in terms of the specific domain adjacency constraints. It follows that while domain adjacency constraints are nongradient, the first and last constraints are necessarily gradient. A more general conclusion we can reach is that any alignment constraint that is evaluated with respect to a single point, either a specific host or a particular edge of a domain, will be gradient, whereas those which are evaluate with respect to stretches of an expression, either a domain or the stretch of the expression preceding or following a host, will not. The following table summarises the discussion of this section showing the six constraint types along with their conditions of violation: (14) Gradient constraints xAh x1D xΩD Non-gradient constraints xPh xFh xAD

conditions of violation violated by every z between x and h violated by every z ∈ D such that z precedes x violated by every z ∈ D such that z follows x violated if x follows h violated if x precedes h violated if x is preceded by a and followed by b, a and b∈D

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 99

2

Finnish data

Finnish, as an agglutinative language, like Hungarian, is comparatively rich in its case and agreement morphology, allowing it to employ what could and has been termed a relatively free word order. Indeed, all permutations of the major elements of the sentence produce grammatical expressions. However, as will be demonstrated in this section, the rules governing the ordering of words in the language in reality play a major and systematic role in identifying both the discourse prominence and the functions of various elements, as well as (to a lesser degree) identifying their grammatical relations. 2.1 Topic, Focus and Contrast Before continuing to investigate the details of Finnish word order, it is important that we give some thought to concepts that play a major role in determining this: topic, focus and contrast. It is especially crucial that we establish ways of manipulating these concepts so that a clear interpretation of the data will be possible. The notions of topic and focus are often expressed as opposites. They are ‘given’ and ‘new’ elements respectively. The topic in an obvious way has to be present in the current discourse, either by already having been mentioned or by having been ‘conjured up’ by elements of the discourse. The focus, on the other hand, is often at the heart of information being transmitted from the speaker to the hearer, and contains information that is assumed not to already be encompassed by the hearer. These two discourse functions therefore do not combine without contradiction: one cannot speak of a topic focus, for example. The notion of contrast, however, readily combines with both topics and foci. Contrast always involves a set of elements that are contrasted. So, for example, a situation in which there is a contrastive topic will involve a set of topics already established for example in a preceding utterance, and which are then rotated as the topics of a sequence of statements in which something different is predicated of each one. The following short discourse demonstrates typical contrastive topics: (15) John introduced Mary to his three friends. Bill, Mary immediately liked. Fred made little impression, but Steve, she sensed was openly hostile The first sentence sets the scene and introduces a number of elements that can act as the topics for subsequent sentences: John, Mary and John’s three friends. In this case, it is the three friends which are the actual topics of the The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 10 10

following three clauses, in which they are contrasted with respect to some predication, i.e. Mary’s evaluation of them. Contrastive focus also involves the formation of a set within which the contrast is carried out. In this case, however, a newly introduced element is contrasted with something else, which may already be part of the discourse. A typical case of contrastive focus is in corrective situations. Suppose speaker A believes P to be predicated of x, but speaker B knows that it is y that P is predicated of, not x. If B corrects A’s belief by asserting P(y), B constructs a set consisting of x and y and contrasts these with respect to P. To A, y is a new element introduced by B and so has the status of a focus. The following short dialogue exemplifies this kind of contrastive focus: (16) A Bill left B No, Steve left Of course, it is possible to have topic and focus interpretations without any contrastive set being involved. In these cases, we get non-contrastive topic and focus readings. A non-contrastive topic is simply a discourse old element that is predicated of in such a way that no comparison is made between it and anything else. So if we introduce an element into a conversation and then continue to predicate of this element it is a topic and there will be no contrastive interpretation associated with it. In the following, the pronoun subject of the second sentence is a non-contrastive topic: (17) Steve left the party. He didn’t like the hostess. Finally, a non-contrastive focus is simply a new item entered into a discourse without including it in a set. For example, a simple answer to a question might qualify as a non-contrastive focus under normal discourse conditions of questioning and answering. Presumably, the questioner lacks a certain piece of information and an answer to the question will provide this information. Hence the answer is new information, and so focus, although it is not contrasted with any other element: (18) A who left? B Steve did While nothing that we have said in this section is original, the importance of the above discussion has been to identify a number of discourse conditions under which the relevant concepts can be elicited. We will be using these contexts to investigate the effects of the relevant discourse functions on The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 11 11

Finnish word order. The following table sets out the contexts we will be using in a convenient way: (19) discourse function non-contrastive topic non-contrastive focus contrastive topic contrastive focus

eliciting context sentence which introduces a single individual question concerning a single individual sentence which introduces a set of individuals sentence containing information to be corrected

2.2 Basic word order in Finnish The basic word order in Finnish is generally understood to be SVO. For example, the following sentence might be claimed to express a neutral meaning: (20) Mikko juo teetä Mikko drinks tea ‘Mike drinks tea’ However, the fact that a similar sentence in Hungarian with the same ‘neutral’ interpretation would also be expressed with the same word order has not unequivocally led to the conclusion that Hungarian is also basically SVO: (21) Misi iszik teát Misi drinks tea É. Kiss (1994), for instance, has claimed that Hungarian is a topic oriented language in which grammatical function has very little role in determining word order. Thus in a sentence such as (21), the subject should be seen as topicalised3. If the same is true of the Finnish sentence in (20), then we might also be led to the conclusion that, contrary to popular belief, grammatical function is not a major contributor in determining basic word order. That Finnish is not like Hungarian in this sense can be demonstrated by manipulating the discourse conditions of the sentence. We may, for example, topicalise the object, which in Hungarian would have the effect of bringing the 3

Gáspár (2005) has claimed that topics are universally present and hence there is no such thing as a ‘neutral’ interpretation, by which what is meant is an interpretation in which discourse functions have no role. The apparent neutral interpretation of the Hungarian SVO word order might therefore be seen as a reflection of the fact that there is a tendency for subjects and topics to coincide. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 12 12

object to the front of the clause. But in Finnish a simple topicalised object remains in its position behind the verb: (22)

Lisa walks past the teahouse a Mikko näkee Liisan Mikko sees Liisa ‘Mike sees Lisa’ b * other word orders

Although the other word orders indicated in (22b) may not be ungrammatical in other contexts, they do not express the intended meaning, i.e. a topicalised object and a standard subject. The grammaticality of (22a) and the inappropriateness of (22b), then, appears to indicate that Finnish does indeed rely on notions of subjecthood and objecthood, as something other than the notion of topic must be responsible for the ordering of the elements.4 The case is similar, although slightly more complicated, when it comes to (non-contrastive) focus interpretations. As illustrated by (23) and (24), focusing on an element does not in itself result in word orders that deviate from the standard SVO pattern. The focused element is marked here with a stress mark indicating its slightly stressed pronunciation. (23) Q A1 A2 (24) Q A1 A2

Kuka näki Liisan? ‘Who saw Lisa?’ ‘Mikko näki Liisan ‘Mike saw Lisa’

• other possible orders

Mitä Mikko näki? what Mikko saw ‘What did Mike see?’ Mikko näki ‘Liisan ‘Mike saw Lisa’

• other possible orders

4

Although agreeing with claims made in Vallduví & Vilkuna (1998) according to whom all OSV word orders in Finnish have contrastive objects, this is at odds with Kaiser’s (2000) findings, as in her data about 35% of OSV structures were seen to express a non-contrastive topic object. As this does not accord with our intuition, we can only assume that Kaiser’s result is either an effect of dialectal differences (although this seems unlikely), or the type of data used by her, namely a corpus of written literary sources. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 13 13

(25) Q A1 A2

Kuka näki mitä? ‘Who saw what?’ ‘Mikko näki ‘Liisan ‘Mike saw Lisa’

• other possible orders

Once again, although other possible word orders might be grammatical given other contexts, they are ungrammatical here as they do not express the intended meanings. Matters appear to become somewhat more complicated when adverbials are introduced to the sentence. With no focus laid on any of the elements, SVAO can be argued to be the least marked word order: (26) Liisa tervehtii iloisesti Mikkoa Liisa greets happily Mikko ‘Lisa greets Mike happily’ With one crucial exception, it is also possible to focus on each of the elements individually without altering the word order. This is achieved by slightly stressing the word in question, again suggesting that, just like topic, focus in itself does not alter the underlying word order. The relevant discourse contexts for sentences in (27) could for example be questions asking to identify the person Lisa greeted (27a), the way the greeting was done (27b), and the person greeting Mike (27c): (27) a b b’ c

Liisa tervehtii iloisesti ‘Mikkoa

• Liisa tervehtii ‘iloisesti Mikkoa

Liisa tervehtii Mikkoa ‘iloisesti ‘Liisa tervehtii iloisesti Mikkoa

The reason behind (27b)’s ungrammaticality will be discussed in a moment. Before that, however, we must also note that SVOA is a possible word order in Finnish, with focus possible on any of the elements, or no element at all. Again the focus is indicated by the stress mark: (28) a b c d

Liisa tervehtii ‘Mikkoa iloisesti Liisa tervehtii Mikkoa ‘iloisesti ‘Liisa tervehtii Mikkoa iloisesti Liisa tervehtii Mikkoa iloisesti

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 14 14

By thus inverting the order of the object and the adverbial, we consequently also slightly alter the interpretation of the sentence. The difference between (27a) and (28a) is in the intended discourse prominence of the adverbial, with the element in (27a) being clearly interpreted as providing important additional information, while in (28a) its contribution to the information provided by the sentence is much smaller, indeed almost insignificant. In order to focus on the adverbial, it cannot precede the object as shown by the ungrammatical (27b), but has to follow it as in (28b). In sentences (28c-d) both the object and the adverbial are interpreted as being prominent, with more prominence given to the adverbial than the object. Finally, the two word orders with no prosodic indication of focus, (26) and (28d), differ in terms of what is regarded as the most prominent non-subject element: in (26), which we have mentioned to be the least marked of the two word orders, the object is given more prominence than the adverbial, while in (28d) it is the adverbial that is interpreted as being the most important piece of new information. These observations, although they may initially seem puzzling, are in fact not that difficult to account for. It is a fairly common view of the Finnish word order that the most prominent post-verbal element is placed at the end of the utterance (see for example Vainio et al. 2003), with the interpreted prominence rising towards it. Consequently, in a linear string of post-verbal elements such as is presented in (29), the prominence hierarchy can be shown to correspond to the one given in (30). (29) X1 X2 … Xn-1 Xn (30) Xn > Xn-1 > … > X2 > X1 This, however, appears to hold true only of instances where no special focus is laid on any of the post-verbal elements (our examples (26), (27c), and (28cd)). In case one of those elements is given a focus interpretation, as is true of our examples (27a-b) and (28a-b), whatever follows the focused element appears to be interpreted as non-prominent information. This is illustrated again in sentence (31), with the prominence hierarchy of the sentence laid out in (32): (31) Mikko tilaa nopeasti teetä ‘Liisalle tarjoilijalta Mikko orders quickly tea for-Liisa from-waiter ‘Mike quickly orders tea for Lisa from the waiter’ (32) Liisalle > teetä > nopeasti Non-prominent (or very low prominence): tarjoilijalta The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 15 15

Consequently, instead of prominence peaking at the end of the utterance, the data in fact suggests that the peak is situated either where a post-verbal focus element lies or, in case no such element is present, at the end of the utterance. This is schematically illustrated by the diagram in (33). (33)

Prominence

focus / end of utterance

VERB

Linear word order In order to distinguish between the domain that starts with the verb and ends with the peak of prominence on the one hand, and the domain that follows the prominence peak on the other, we will from here on label them the “centre domain” and the “margin”, respectively. Note that the ungrammaticality of sentence (27b) suggests that arguments cannot be marginalized. The distinction we have introduced here is not unheard of in the literature. Haegeman (2002, 2003), for one, distinguishes between what she calls central adverbial clauses and peripheral adverbial clauses, showing that these are distinct in English in as much as the former do not allow topicalisation, whereas the latter do: (34) a * if these exams you don’t pass, you can’t get a degree b if these problems we cannot solve, there are many others we can tackle Haegeman differentiates between the two types of modifying clause in that the central ones modify the event or state expressed by the main clause, whereas peripheral ones “serve to provide a discourse frame against which the proposition of the host clause is evaluated” (Haegeman 2003 p.5). In many ways this is comparable to the restrictive/non-restrictive distinction. Huddleston (1984) for example, points out that this distinction has to do with the salience of the information provided by the modifier: The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 16 16

(35) a we were surprised to see she was wearing shoes which were red b we were surprised to see she was wearing shoes, which were red With the restrictive relative in (35a), the surprising thing is the colour of the shoes and hence the relative clause introduces semantically salient information. The non-restrictive relative in (35b) however does not work in this way. What is surprising there is that shoes were being worn and this has nothing to do with their colour. The relative clause here simply adds extra non-salient information This is also essentially the distinction we make between central and marginal elements. For example, consider adverbial placement discussed above: (36) a Liisa tervehtii iloisesti Mikkoa Liisa greets happily Mikko ‘Lisa greets Mike happily’ b Liisa tervehtii Mikkoa iloisesti The difference between these sentences concerns the role played by the adverb. In the first case its contribution is one that modifies the event in a way that makes it an event involving ‘happy greeting’, and the fact that the greeting was done ‘happily’ is thus a salient part of the proposition. (36b), on the other hand, is the description of a ‘greeting’ event. That this greeting was done happily is proffered as extra information about it, but this is not a salient part of the proposition. This distinction seems to be restricted to modifiers, and there appears to be nothing like a marginal argument: (37) a we were surprised to see that he was eating an apple b we were surprised to see that he was eating Here, what is communicated as being surprising is that ‘he is eating an apple’ not that ‘he is eating’. The latter interpretation is impossible while the object is part of the clause, and can hence only be achieved if the object is unmentioned, as in (37b). 2.3 Contrast We have so far looked at neutral Finnish sentences, as well as those involving topics and foci, and concluded that neither topicalisation nor focalisation is capable of altering the basic underlying SVO word order found in the lanThe Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 17 17

guage. However, when the notion of contrast is added to the mix, the word order becomes more dynamic. In Finnish, a contrastive element, topic or focus, tends to come at the beginning of the clause. To illustrate this (38) is a sentence involving a subject, and (39) an object, with a contrastive topic reading. Bear in mind that once again, although all permutations are possible, the ones marked as ungrammatical here do not fit the contexts provided. (38)

We are talking about a group of people and what they study. a … ja Risto opiskelee astrofysiikkaa and Risto studies astrophysics ‘and as for Christopher, he studies astrophysics’ b * other orders

(39)

We are talking about the various physics departments and what kind of people study in them. a … ja astrofysiikkaa nörtit opiskelee and astrophysics nerds study ‘and as for astrophysics, nerds study it’ b * other orders

Meanwhile, the responses in (40) and (41) illustrate contrastive foci: (40) A Matti opiskelee astrofysiikkaa B1 Ei, ‘Risto opiskelee astrofysiikkaa! no Risto studies astrophysics ‘No, it is Christopher who studies astrophysics (not Matthew)’ B2 * other orders5

5

The sentence Ei, ‘Risto astrofysiikkaa opiskelee would, in fact, be both a grammatical and a suitable response to the original query. However, there is a difference in meaning between it and B1 in that in this sentence the object is given more prominence, and hence is a more marked word order than the one presented in B1. This is possibly the result of an ordering of the verb, rather than the object, a topic we will not cover in this paper. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 18 18

(41) A Risto opiskelee kvanttimekaniikkaa B1 Ei, ‘astrofysiikkaa Risto opiskelee! no astrophysics Risto studies ‘No, it is astrophysics that Christopher studies (not quantum mechanics)’ B2 * other orders6 While it is no surprise to find the subject in a preverbal position in these cases, as that is the canonical position of the subject, the fact that the object is pulled to the front of the clause by a contrastive reading indicates that this is the contrastive position in Finnish7. The contrastive position seems to be defined with respect to the beginning of the clause rather than as a particular pre-verbal position, unlike the case with the subject. We can see this from the fact that a contrastive object precedes the subject and could not follow it: (42) * Ei, Risto ‘astrofysiikkaa opiskelee! Of course, a contrastive subject can at the same time satisfy its subject and contrastive requirements in one pre-verbal position, providing there are no other preverbal elements. In the case of two contrastive elements, one focalised and another topicalised, it is the topicalised element that has to come first, immediately followed by the focalised one:

6

Ei, Risto opiskelee ‘astrofysiikkaa would again be a possible response to the sentence, involving what appears to be an in-situ contrastive focus. This is possibly the result of the repetition of the word order employed in the original utterance (A). We acknowledge the need for a further study of in-situ contrasts, which in this paper have been largely ignored. 7 Kaiser (2000) comes to the same conclusion, providing an X-bar analysis in which the Finnish clause has a ‘Contrast Phrase’ as its top functional projection; the contrastive element being moved to the specifier of which in much the same way as wh-elements move to spec CP and topics and foci move to spec TopP and FocP respectively under similar assumptions. However, apart from the fact that contrasted elements sit at the front of the clause, i.e. the very fact that needs accounting for, there is no other evidence to favour the structural architecture that Kaiser proposes. The head position of the Contrast Phrase, it seems, is never filled either by an overt contrast morpheme or by head movement. The proposal then only serves to exasperate an already over burdened mechanism without much in the way of positive gain. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 19 19

(43)

We are talking about a group of people and what they did on Wednesday evening. Anna usually meets Mary, but not this time around. a … Anna ‘Petrin tapasi keskiviikkona Anna Petri-acc. met Wednesday-on ‘As for Anna, it was a Peter she met on Wednesday (and not Mary)’ b * … ‘Petrin Anna tapasi keskiviikkona.

(44)

We are talking about a group of people and what they did on Wednesday evening. Mary usually comes to meet Anna, but not this time around. a … Annan. ‘Petri tapasi keskiviikkona Anna-acc. Petri met Wednesday-on ‘As for Anna, it was Peter who met her on Wednesday (and not Mary)’ b * ‘Petri Annan tapasi keskiviikkona.

In these examples, as Anna is introduced as one of the set of people being discussed, this element has the status of a contrastive topic. Peter, on the other hand, is introduced as a surprise element, contrasting with the expected individual (Mary) and hence has a contrastive focus reading. Finally, with two contrastive topics, the ordering between the two topics does not seem to be restricted in any way: (45)

We are talking about a group of people and a group of fruits that are on the table. a … Jessika omenan haluaa Jessica the apple wants ‘As for Jessica, she wants the apple b … omenan Jessika haluaa the apple Jessica wants ‘As for the apple, Jessica wants it’

More interesting, as well as more theoretically challenging, is the behaviour of focalised subjects when interacting with contrastive elements in the same clause. We have so far illustrated that focus does not in itself alter the canonical SVO word order, and that contrastive elements are rather straightforwardly pulled to the front of the clause. However, with both a contrastive object element and a focused subject in the same clause, as in examples (46) and (47), the subject is suddenly sent to the end of the centre domain. (46) inThe Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 20 20

volves a contrastive topic object, while (47) illustrates the situation with a contrastive focus object: (46)

We are at a party. Someone asks about who bought the drinks. The responder knows that various people bought different drinks, and that John bought all the beer. Q Kuka osti juomat? who bought drinks-acc. ‘Who bought the drinks?’ A Oluet osti ‘Jukka beers-acc. bought Jukka ‘As for the beer, John bought it’

(47)

There is a paper bag in the corner with bottles in it. The questioner incorrectly assumes that the bottles contain wine, and wants to ask who bought them. Q Kuka osti viiniä? ‘Who bought wine?’ A ‘Olutta osti ‘Jukka ‘The beer was bought by John’ (and it is not wine) A’ * ‘Olutta ‘Jukka osti

That the focused subject indeed moves to the end of the centre domain, and not for example to an immediately post-verbal position or to a clause-final one is suggested by (48), which is a modified version of the situation presented in (47): (48)

There is a paper bag in the corner with bottles in it. The questioner correctly assumes that they were bought at the supermarket and that they were bought today, but incorrectly supposes that they contain wine. Supermarket is deemed unimportant information both in the reply as well as the question because it is the place where the shopping is always done. Q Kuka osti tänään ‘viiniä supermarketista? who bought today wine supermarket-from ‘Who bought wine today from the supermarket?’ A ‘Olutta osti tänään ‘Jukka supermarketista beer bought today Jukka supermarket-from ‘The beer was bought today by John (at the supermarket)’ (and it is not wine) The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 21 21

Note also that it is only a focused subject that is moved to the end of the centre domain when contrastive elements are present. Example (39), reprinted partly here as (49), has already shown that plain subjects are not moved from their canonical pre-verbal subject position in the presence of a fronted contrastive element: (49)

We are talking about the various physics departments and what kind of people study in them. a … ja astrofysiikkaa nörtit opiskelee ‘and as for astrophysics, nerds study it’ b * … ja astrofysiikkaa opiskelee nörtit

Topicalised subjects are not moved, either: (50)

Liisa kävelee teetalon ohi. Liisa tervehtii Petriä. ‘Lisa walks past a teahouse. Lisa greets Peter.’ a Ei, ‘Mikkoa Liisa tervehtii no Mikko-part. Liisa greets ‘No, as for Lisa, it is Mike she greets’ b * Ei, ‘Mikkoa tervehtii Liisa

It appears, then, that while topicalisation and focalisation in themselves are not individually able to change the canonical SVO word order, and the contrastive elements are always fronted, the co-appearance of contrast and a focused subject makes the latter switch sides from a pre-verbal position to the end of the centre domain. The following table summarises the word order observations we have made in this section: (51) element subject object topic subject topic object focus subject focus object all contrasts

position pre-verbal post-verbal pre-verbal post-verbal pre-verbal if no contrastive element present post-verbal if contrastive element present post-verbal pre-verbal

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 22 22

3

Analysis

To analyse the data presented in the previous section, we will develop a set of ranked alignment constraints of the types discussed in the first section. These will align elements bearing certain features to given hosts or domains. As both the features and the domains are given in the inputs, some preliminary discussion of inputs will be necessary. 3.1 Input We will be making fairly standard assumptions about inputs, i.e. that they contain all the information relevant for the semantic interpretation of the expression. Also, following standard Alignment Syntax practise we will assume that the input is directly interpreted by the semantic component of the language faculty. 3.1.1

Features

Besides lexical elements themselves and the semantic material that they introduce, we assume that other meaningful aspects of the input are included as features assigned input elements. Feature assignment is free, modulo a small number of redundancy rules, such as discussed below, though it of course has semantic consequences which may render an input uninterpretable. For the present study, the important features are: (52) Subject (S) Contrast (K) Prominence (P) Aboutness (A) The subject feature distinguishes its assignee from all other arguments. It is the only grammatical function feature we assume. Other arguments can be distinguished in terms of the θ-roles they bear. While it is unclear what semantic content it carries, it appears to be an essential part of linguistic systems. The contrast feature confers on its assignee a contrastive interpretation, which as discussed above involves the construction of a set from which to interpret the contrast. Prominence is a feature associated with focus interpretation, but not equivalent to it. Thus, while foci are prominent, not all prominent elements are foci. Presumably a true focus interpretation also includes some notion of newness, though whether this is included as a feature is not a question we investigate in the present paper. Aboutness is the feature of the topic and the two notions can be equated. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 23 23

We assume a redundancy rule that ensures that all contrastive elements are prominent: (53) K  P This will be important for accounting for the competition between contrastive elements and foci for the leading edge of the relevant domain. It is this assumption that forces us to separate the notion of prominence from focality as a contrastive topic will also be prominent, by (53), though we would not want to say that such elements are also foci. The combination of the three discourse oriented features produce the following input elements indicated with their intended interpretations: (54)

features ∅ P A KP KPA

-

interpretation8 neutral focus topic contrastive focus contrastive topic

In addition, the subject feature may be added to any of these elements. 3.1.2

Centre and Margins

Domains are defined as sets of input elements which are related in a particular way. For example, a straightforward domain consists of a predicate and all its dependents. We call this the predicate domain. A dependent of a predicate is any input element that bears a feature which is interpreted with respect to that predicate. Thus, arguments and modifiers are all dependents of a predicate, and therefore part of the predicate domain. In the case of subordination, the dependents of a dependent are also included in a domain. In Finnish, as was illustrated in the diagram in (28), there appears to be a sub-domain of a predicate domain which includes all those dependents of the predicate which carry ‘important’ information. We call this domain the centre domain. Elements in the predicate domain which are not in the centre are said to be in the margin. As it affects the interpretation, the status of a modifier as central or marginal must be marked in the input. The central domain can therefore be 8

The system allows the combination of features PA, a ‘prominent topic’. Whether or not there exists such an element is unclear. However, if there is such an element, our prediction is that is would behave like a focus. The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 24 24

defined as the set of elements dependent on a predicate not marked as marginal. We assume, therefore, a feature which marks marginal elements: (55)

Marginal (M)

3.2 Grammatical Functions As argued above, Finnish has a basic SVO word order. From this we can deduce that the language is predicate initial, with arguments tending to follow the predicate, but with the subject in a special position before the verb. Thus we have the following constraints: (56) aFp > aPp (arguments Follow their predicate) sPp > sFp (subject Precedes its predicate) sPp > aFp The first ranking ensure that in general arguments follow their predicates while the second and third ensure that subject behave differently. First, that subject precede their predicate is captured by ranking sPp above sFp and second that the higher ranking subject constraint can overcome the dominant argument alignment is achieved by ranking the former above the latter. The following table shows how this works to select the SVO word order: (57) sPp aFp aPp sFp
XS V X

XS X V !

V XS X *! V X XS *! X V XS *!

X XS V !

In this table, and those that follow, X represents an argument and the subscripts represent the features it carries. Thus XS is an argument baring the subject feature. An argument without any features is a simple object. (58) a Faith(S) > sPp

sFp (Finnish ranking) b sPp, sFp > Faith(S) (Hungarian ranking)

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 25 25

(59) Faith(S) sPp sFp
XS V

V XS

• !

XV

• !

VX

• !

(60)

sPp sFp Faith(S) XS V

• !

V XS *!
XV

VX

The tables show the difference between the Finnish and Hungarian rankings. In Finnish the subject feature remains as a consequence of the high ranking of the faithfulness constraint. Therefore the subject alignments are effective and the subject occupies a preverbal position, unless some higher ranked constraint determines otherwise. In Hungarian, however, the subject feature is underparsed and so plays no role in determining the distribution of this argument, which is therefore entirely at the mercy of other constraints. The Finnish object, however, prefers a position towards the back of the predicate domain and is only followed by a focus or by marginal material (see examples (28)-(33)). While we will return to the distribution of the focus later, let us first consider the order of the object with respect to central and marginal adverbs. It is clear that both objects and adverbs want to follow the verb, but their order with respect to each other is dependant on the status of the adverb. If objects not only have to follow the verb, but they should also be at the edge of some domain, both conditions will place the object as near to the final edge of this domain as possible. As marginal adverbs can follow the object (see for example (28a) and the related discussion), we conclude that it is the centre domain that the object is aligned to. Thus we have the following: (61) aAC (an argument is adjacent to the edge of the central domain) Marginal elements, however, always follow central elements. This might be the result of a fairly low ranked requirement that central elements should be adjacent to the verb:

The Even Yearbook 7 (2006), © Mark Newson and Vili Maunula

Word order in Finnish 26 26

(62) cAp (a non-marginal element is adjacent to the predicate)9 Consider, then, what happens when there is an object (X), a central adverb (M

modifier) and a marginal adverb (MM):