Vendler (1974, p. 96) intended to show that a proper linguistic analysis of quantificational expressions could help to capture the “logically relevant features involved in the vernacular use of the particles of quantification”. He tried to show that each, every, all and any did not have the same universal meaning. Even though his analysis was properly linguistic, he wanted to provide an empirical basis for some of the traditional problems of analytic philosophy (such as the problem of reference, existential import, and lawlike generalizations). In this short essay, we will show that the application of the alternative semantics framework provided by Menéndez-Benito (2005, 2010) and Aloni (2007, 2019) –which are inspired in Kratzer and Shimyoma (2002)– can account for Vendler’s insights on Free Choice Items (FCI). We will also defend that if we take into account a speech act theory, as in Searle (1985), we may better understand how FCI work.

First, we will present some of Vendler’s main findings with regards to FCI any. Second, we will show how we can apply a contemporary alternative semantics framework to Vendler’s proposal. Third, we will see how a speech act theory can help us to better understand free choice effects. Fourth, we will present some final remarks.

In his very thorough analysis of universal expressions in natural languages, Vendler (1962 and 1974) described the behavior of any. In English, any may function as a negative polarity item (1), and as a free choice item (2). Vendler described both behaviors, but since our research is focused on the notion of freedom of choice, we will only take into account what he said with regards to FCI any.1

I don't have any potatoes.

FCIs have been on the front of the linguistic discussion ever since Vendler (1962). The discussions have been focused on whether they are universal, existential, both, or indefinites (see KADMON, 1993, DAYAL, 1998, HORN, 2000, MENÉNDEZ-BENITO, 2010, among others). Vendler's extensive analysis of any has been responsible for some of those discussions.

In this section, we will focus on some of the modal contexts in which any can sometimes appear that were noted by Vendler (1962): subtrigged in declarative statements, unsubtrigged in permissions and generic contexts.2 The first context that we will consider is related to subtrigging, first analyzed in Legrand (1975), which allows a modified any NP to appear in episodic statements (DAYAL, 1998). Even though this was not central for Vendler (1962), he did note that subtrigged any could appear in declarative contexts with the addition of a modal clause, as in (5):

3. *Any doctor told me…

4. *I asked any doctor…

5. Any raven you may select will be black.

In Vendler (1974, p.92), he argued that one of the most striking characteristics of any is that it cannot appear in “simple declarative sentences", though it can be corrected by introducing a modal clause, as in (5). Basically, he noted that when any was subtrigged it could be allowed in contexts where it would not normally be allowed.3 How subtrigging allows FCI in episodic sentences is something that is still discussed (see DAYAL, 1998, CHIERCHIA, 2013, among others).4 Even if Vendler was not properly interested in subtrigging, he did note that it was important to understand the behavior of FCI any, and the example selected seems to take into account a modal environment.

The second context we will analyze is the one that involves a permission in the sense that it expands the set of permissible worlds to “include at least one in which the content of the permission holds”, but not a command which would “involve the elimination of all those worlds in which the content of the command does not hold” (DAYAL, 1998, p. 455). Some of the examples Vendler analyzed involved the presence of an indefinite, but count nonetheless as permissions:

6. Take any one of them [offering apples from an apples basket]

7. Take any two (three, etc.) of them

These examples are the ones that lead Vendler to propose that any involves what he called freedom of choice which: “...succeeds in blending indetermination with generality"(VENDLER, 1962, p. 151). Basically, what he is saying may be modernized by defending that the FCI any is a universal indeterminate pronoun (MENÉNDEZ-BENITO, 2005). Thus, we do not get a definite reading, and we can make a choice from all the available possibilities.5

The last context we will consider has to do with the generic contexts in which the FCI any may appear. Vendler (1962) was interested in characterizing what he called lawlike propositions, as (8). They are analyzed as statements which lack “existential import”. In those type of propositions, according to him, we are not committed to the concrete existence of the entity that is being selected by any. That is, they do not involve anything concrete. He is talking about generic statements.6 Vendler (1962) argues that these propositions, including examples (9) and (10), may be accepted even though there is nothing or no one who has done the things that they are predicated of doing, and they do not involve a commitment nor a belief that it will come into being, as in (11):

8. Anybody trespassing on the premises will be prosecuted.

9. Any nation that conquers the moon will rule the earth.

10. Any perpetual-motion engine would violate the laws of thermodinamics, which is impossible.

11. Anybody who would do that would perform a miracle.

He argues that what any does in those cases is introduce a lawlike assertion.7 According to him, all general statements involving any lack not only existential import, but also definite reference. In those cases, it may be analyzed as a synonym of all, which basically restates its universal behavior. He did not explore the issue further, but the apparition of any in generic statements has been discussed by many (see KADMON, 1993, CARLSON et al., 1995, HORN, 2000, DAYAL, 1998, MENENDEZ-BENITO, 2010, among others).

In the next section, we will provide an approach to FCI from an alternative semantics framework which also takes into account the behavior of the Spanish FCI cualquier, we will argue that the application of the framework can account for the characteristics Vendler noted, related to subtrigging, permissions, and lawlike propositions.8

We will apply an alternative semantics framework inspired in the work of Menéndez-Benito (2005; 2010), and Aloni (2007; 2019). They assume that FCI cualquier is a universal indeterminate pronoun which involves two covert operators ∀ (in MENÉNDEZ-BENITO, 2010), and exh (ALONI, 2019) in an alternative semantics framework (KRATZER & SHIMOYOMA, 2002).9 The indeterminate pronoun cualquier introduces alternatives and is associated at Logical Form to a universal operator ∀ that is taken to operate at the propositional level, which is what we see in (12.1), in which A is the set of alternative propositions. Exhaustification is an operation that takes an expression α of type e, which provides the domain A, a predicate which provides the property P, and returns an expression exh[α , P], of type <e, <s, t>>, as in (12.2) where α _w,g = A, and [[P]]_w,g = {P} (ALONI, 2007):10

12. Covert operators:

1. [∀](A)={the proposition that is true in all worlds in which every proposition in A is true}

2. [exh[α,P]]]_w,g = {λxλv.x ∈ A&P (x)(v)&∀y ∈ A : P (y)(v) ⇒ P (x) ⊆ P (y)}11

Since Exh is of type <e, <s, t>>, we need two type shifts for it to be applied at the IP level, and at the DP level, to explain the behavior of FCI: one which will be propositional, and another which will be entity denoting.12 When the FCI is unsubtrigged, exh must apply at the IP and SHIFT_<s,t> to generate a set of mutually exclusive propositions. When the FCI is subtrigged, it will apply to the DP level, it will SHIFT_e and ↓ will be applied to generate a set of individuals. This analysis predicts the ungrammaticality of FCI cualquier and any in episodic sentences, as (13), because exhaustification applies at the IP level, where SHIFT_<s,t> yields the partitions represented in (13.2). Since each alternative in the partition is stated to be true, it results in a contradiction. It also predicts their grammaticality in modal sentences, as (14), and subtrigged cases, with a SHIFT_e, as (15) (examples taken from ALONI, 2019, pp. 5-7).

13. # Anyone walked/ Cualquiera caminó.

1. [∀] ((SHIFT_<s,t> (exh[anybody, walked]))

2. | nobody walked | only d1 walked | only d2 walked | … |

14. Anyone may walk/ Cualquiera puede caminar.

1. [∀] (◊ (SHIFT_<s,t> (exh[anybody, walk]))

2. |◊ nobody walked | ◊ only d1 walked | ◊ only d2 walked | …|

15. Anyone who tried to walk walked / Cualquiera que intentara caminar caminó.

1. [∀] (↓ (SHIFT_e (exh[anyone, who tried to jump]) fell)

2. | d1 fell | d2 fell | …|

In (13) and (14), exhaustification undergoes SHIFT_<s,t> which explains the partitions we get in both cases. In (13), the partition results in a contradiction, because there is an operation akin to only which yields a contradiction if it interacts with ∀ (MENÉNDEZ-BENITO, 2010). In (14), the derivation is good, because the modal operator expands the possibilities. In (15), exhaustification occurs within the DP, which is why it undergoes SHIFT_e. It yields as an output the sum of all the people that tried to jump in w. ↓ is applied to avoid trivial quantification, and produces a set of singular individuals. This final example accounts for the correction Vendler notes is possible when a modal clause is introduced, in cases as (5), repeated here as (16):13

16. Any raven you may select will be black.

Aloni (2007; 2019), and Menéndez-Benito (2010) provide an approach to FCI cualquier that may deal with modal statements, and can account partially for the data Vendler described. Menéndez-Benito (2010) provides an explanation of the behaviour of cualquiera in permissions, such as (17), with the use of the exclusiveness operator. We will follow Aloni (2007) and use exh, in (18):

17. Juan puede coger cualquiera de las cartas del mazo.

‘Juan can take any of the cards in the discard pile.’

18. [∀] (◊ (SHIFT_<s,t> (exh[Juan coge, cualquiera de las cartas del mazo]))

19. |◊ Juan took no card | ◊ Juan took only card1 | ◊ Juan took only card2 | ... |

In this case, exhaustification undergoes SHIFT_<s,t> which explains the partitions we get in (19) in which the alternatives include only one card. Again, we do not get a contradiction because the modal operator expands the possibilities, and Juan can choose any cards in the deck. This analysis allows us to formalize Vendler's idea with regard to permissions, and it also allows us to take into account his freedom of choice. This is may happen thanks to the presence of ◊ (SHIFT_<s,t> (exh[Juan coge, cualquiera de las cartas del mazo]) in which the exclusive propositional alternatives generated by the application of SHIFT_<s,t> to exh are expanded by the possibility modal. Thus, we can apply the universal propositional quantifier that provides us with a given amount of possibilities to choose from.

This proposal does not seem to allow us to analyze the cases which involve an indefinite quantifier such as two. However, by modifying the generated alternatives it could be accommodated by this proposal, as in:

20. Take any two of them [offering apples]

21. [∀] (◊ (SHIFT_<s,t> (exh[Take, any two of them]))

22. |◊ He took no apples |◊ Juan took only apple1 & apple2 | ◊ Juan took only a2 & a3| ...|

Even though Aloni (2007) did not take into account this possibility, the alternative semantics framework can accommodate Vendler's analysis with such small modifications.

Finally, we should explain how an alternative semantics approach can account for the generic contexts Vendler (1962) encountered, such as (23). We will propose that, once we have the partition of the logical space provided by the application of SHIFT_<s,t> to Exh, the generic operator applies to the set of propositional alternatives generated by Exh and provides us with a generic interpretation similar to the one found with indefinite generics. GEN is a modal, and does not commit ourselves with the actual existence of a singular proposition (see CHIERCHIA, 1998: 381). The idea is that, in a sentence like (23), there are two sources of genericity, one that will be indicated by the verbal aspect, which is the one that triggers GEN, and another with the FCI cualquier, which is the one that triggers ∀, as in Dayal (1998, p. 447). Thus, the universal quantifier is quantifying over possible propositions, which are generated by the application of GEN to the alternatives in (24) and no contradiction arises:

23. Any violation will be prosecuted.

24. [∀] GEN ((SHIFT_<s,t> (exh[any violation,prosecuted]))

25. |no violation prosecuted | only violation1 prosecuted | only violation2 prosecuted |...|

This representation allows us to explain the availability of the universal sentence above which has a sort of generic flavor, but involves nonetheless universal quantification. Thus, it is not just the application of GEN, which is modal in nature and a part of the verbal aspect (KRIFKA et al.,1995), CHIERCHIA, 1998, among others). It is the addition of GEN in these contexts, while also having ∀, what allows us to explain the behavior of any NP in these generic contexts. This combination can also explain the behavior of negative generic sentences (POLAKOF, 2021, p. 11), such as No cualquier jugador puede jugar en Boca /‘Not any player can play in Boca.’, as well as the necessity generic sentences analyzed in Menendez-Benito (2010, p. 54), such as Tienes que contestar cualquier pregunta./‘You must answer any question.’, among others.

In addition to this, the possibility of the derivation being made on the basis of the propositional alternatives generated, which includes no violation is prosecuted seems to reflect the fact noted by Vendler (1962, p.156) that lawlike generalizations are not rendered false even if there is no one or nothing that satisfies the predicate. Thus, his “lack of existential import” may be accounted for in the alternative semantics framework which we have applied in this discussion.14

Vendler showed that a proper description of how natural language quantifiers such as any, all, each and every had to be taken into account to give a correct characterization of the logical features which are relevant in natural language. His descriptions were taken into account by some logicians, such as Hintikka (1980), who proposed that any and all differed regarding their scope properties. We have shown that nowadays we can account for the behavior of FCI with an alternative semantics framework, but we have said nothing with regard to reference. The account here presented assumes a semantic approach to FCI, which introduces universal propositional quantification, and exhaustification which may be used to account for the relevant logical features of natural language quantifiers. This can explain the semantic behavior of FCI, such as any and cualquier. However, it does not explain the role the FCI plays in the understanding of natural language nor does it explain anything about the relationship between the use of language and the world.

If we assume that FCI involve universal propositional quantification, it cannot be seen as a variable, as indefinites in Heim (1982). We can ask ourselves, then, whether it plays any role at all with regard to reference (understood here as a relationship between the use of an expression and a concrete particular in the actual world) or not. And the answer must be that, since it involves universal quantification, it does not refer to a concrete particular in the actual world. It is not involved in definite referential phrases when it is subtrigged, it does not refer in permissions or in generic contexts (see RECANATI, 2004). Thus, if we are confronted with the philosophical question of reference and referential expressions, we may defend that FCI do not form referential expressions.

Even though FCI do not involve referential expressions (in the sense that they do not relate to a concrete particular in the actual world), there is something to be said with regard to how the use of a statement involving a FCI interacts with the actual world. That is, there is something to be said with regard to the speech acts that we make with FCI. Those speech acts can be related to agency and free choice. The first examples that we will consider are taken from Rioplatense Spanish (in POLAKOF, 2021), they appear in episodic statements, and involve the interaction of negation and cualquier:15

26. No agarré cualquier historia y la produje aunque no tuviera nada que ver con lo que creo.

‘I did not grab any story and produced it even though it had nothing to do with what I believe’

27. No enviaron a cualquier periodista a cubrir el viaje de el candidato, sino a sus informativistas principales.

‘They did not send any journalist to cover the trip of the candidate, but their main newsreporters.’

We proposed (Polakof, 2021, inspired in DAYAL, 1998) that these examples can be reinterpreted as involving a covert modifier which presupposes a selection, such as that I could choose. In the alternative semantics framework we are using, it would amount for the following:

28. No agarré cualquier historia (que pudiera elegir)…

1. [NEG][∀] (Yo agarré ↓ (SHIFT e (exh[cualquier historia, que pudiera elegir]))

29. No enviaron a cualquier periodista (que pudieran elegir).

1. [NEG][∀] (Ellos enviaron ↓ (SHIFT e (exh[cualquier periodista, que pudieran elegir]))

SHIFT_e (exh[cualquier historia, que pudiera elegir]) yields the maximal collection of histories I could choose in w₀, and SHIFT_e (exh[cualquier periodista, que pudieran elegir]) the maximal collection of journalists that they could choose in w₀.16 To avoid vacuous quantification, ↓ is applied and produces a set of individuals (following ALONI, 2007). Then, it combines with the denotation of the verb to produce the set of alternatives presented before. Since the set occurs in the scope of a negated universal operator, we get the desired interpretation: not every story was chosen by me or not every every journalist was sent by the bosses.

The previous representation correctly predicts that the agent did not select all of the possible alternatives. She selected the stories that she wanted to select, and they sent the journalists that they wanted to send. Thus, the free choiceness effect is maintained within the scope of negation. Even when all of the possibilities are available, we may choose between them, and this is reflected by the application of exh. However, this strictly semantic analysis, does not shed light about the actual use the speaker makes of these statements.

These statements may be classified as assertives (SEARLE, 1985). They commit the speaker to the truth of the asserted proposition (SEARLE, 1985). The asserted proposition is connected to a concrete entity which is freely chosen by the agents of the statement, and that fact differentiates FCI assertions from non-FCI assertions in which the freedom of choice is not relevant.17 In addition to this, the assertion involving a FCI also reflects the fact that there was a concrete entity that was chosen: there was a particular story, and there were some journalists. Thus, even though, there is no definite reference, we may find a relation to the actual world.

The second set of examples that we will deal with involve permissions, as in (in POLAKOF, 2021):

30. Podés sacar cualquier auto.

`You can take-out any car.'

Permissions involve a different speech act, which could be classified in Searle's speech act theory as a directive. In a directive, the speaker wants the hearer to do something (SEARLE, 1985, p. 13). The speaker says something to the hearer and expects the hearer to act in accordance to what she said. In addition to it being a directive/permission, it seems reasonable to assume that, with the use of the FCI, the speaker may be signaling that any conceivable possibility is an option, and instructs the hearer to consider a wide domain of quantification (PENKA, 2016, p. 316).

Thus, when I give someone the permission to take any car they want, I'm expecting a perlocutionary effect which involves the selection of any car in the actual world (or in that particular situation). Thus, as a speaker, I influence the hearer to do something, by saying something (DAVIS, 1980, p. 43). I intend the hearer to choose the particular car that she wants. And, in languages such as English, it may be even more than one, as in the examples provided by Vendler (1974):

31. Take any one of them [offering apples from an apples basket]

32. Take any two (three, etc.) of them

Thus, even if FCI does not involve any definite reference at all, they do involve a link to the world which should be differentiated from others. In the case of assertives (made possible by subtrigging), the use of a FCI commits the speaker to the truth of a proposition which involves her as a free choosing agent of an action in which she chooses freely a particular entity she wants to choose. In the case of directives, there may be a particular perlocutionary effect, and the speaker has the intention that the hearer knows that any conceivable possibility is an option.

In the case of generic contexts, or lawlike propositions, there is no link to the actual world, but a relation to possible worlds which could explain why Vendler noted that lawlike propositions, as example (8) Anybody trespassing on the premises will be prosecuted, are not rendered false if no one enters the premises. As we noted earlier, since GEN is a modal, it allows us to speak about possible propositions, and it does not commit ourselves with their existence in the actual world nor does it commit ourselves with the realization of any of the arguments of the proposition.

If we leave the semantic domain, and move into the speech act theory, we can try to classify examples as (8) in Searle’s speech act theory. They cannot be assertives because they do not involve a description of an actual state of affairs.18 They cannot be directives because they do not intend to have a perlocutionary effect. We do try, when we use it, to match the language to the world. Thus, they can be classified as declarations (SEARLE, 1985, p. 16). We can assume that, when someone uses a sentence as (8) they declare that if someone trespasses the premises, she will be prosecuted. Nonetheless, the declaration in itself does not involve a commitment with the existence of someone who trespasses. Thus, they do not involve reference to the actual world. If they are classified as declarations, we can account for the pragmatic effects Vendler noted of statements such as (8) which he calls a “lack of existential import”. This can be translated to a more contemporary terminology, and it can imply that when we use a generic statement we do not need to commit ourselves with the existence of any of the entities involved in the proposition. Since existence is not important in declarations, it seems reasonable to classify Vendler’s lawlike propositions as declarations.

Vendler (1962; 1974) described many of the characteristics that FC any has. He did not have the proper resources to provide a formal approach to his insights. We have shown that an alternative semantics framework, as the one proposed in Menendez-Benito (2010) and Aloni (2007), can account for such properties. It can explain why it can sometimes appear in episodic contexts, in modal contexts and in generic contexts. It can explain the freedom of choice by introducing alternatives. Overall, we have shown that from a linguistic perspective the developments made have been such that Vendler's insight were taken into account, which is something that cannot be denied.

We have shown that an alternative semantics framework can explain the lack of definiteness that Vendler saw in FC any. Free Choice Items are indeterminate pronouns which may be linked to a universal propositional operator, and to exh. Their combination with these covert operators allows us to explain most of Vendler's observations, and in addition we may defend as he did that the linguistic behavior of natural language quantifiers should be taken into account by philosophers who research language related philosophical problems.

We have defended that definite reference is not at stake when FCI are taken into account, because they involve indeterminate universal quantification. We have shown that if we take into account a speech act theory, we can explain how FCI can be related to the world. In the case of assertives, not only are speakers committed to the truth of the asserted propositions, but to their freedom of choice. When we use a subtrigged any NP, we are specifying to the hearer that we had the option to choose freely. In the case of directives, not only is there a perlocutionary effect wanted, but also the speaker signals the hearer that she may consider any conceivable option. Thus, the hearer, as an agent, can choose whatever she wants (of the conceivable options). In the case of declarations, which involve generic contexts, the speaker tries to match the speech act to the world. Thus, the use of a generic statement does not involve a commitment with the existence of concrete entities.

Overall, we have tried to show that we have contemporary tools to take into account Vendler’s insights with regard to freedom of choice. We have also tried to show that semantics does not exhaust the effects of Vendler’s freedom of choice. Pragmatics needs to be considered, if we want to fully understand that said phenomenon. This short essay also tries to provide a pragmatic approach to freedom of choice, by analyzing different speech acts in which FCI appear (in Spanish and English). Nonetheless, more research is needed to explain all the phenomena that FCI involve.. Finally, we would like to defend, as Vendler did, that Philosophy should consider the behavior of FCI to try to solve problems which are related not only to reference, but to free will, to intention, and to agency.

This research was possible thanks to the grant FCE_3_2018_1_148810 by ANII (Uruguay). I would also like to thank the people at Chá das 5, organized by Marco Ruffino at UniCamp.

ALONI, Maria. Free choice and exhaustification: An account of subtrigging effects. In Proceedings of Sinn Und Bedeutung, volume 11, pages 16–30, 2007

ALONI, Maria. Indefinites as fossils: The case of wh-based free choice. 2019, ms.

CARLSON, Greg. Reference to kinds. New York: Garland. Revised version of author’s thesis, University of Massachusetts, Amherst. 1977

CARLSON, Greg; PELLETIER, Francis, editors. The Generic Book. University of Chicago Press. 1995.

CHIERCHIA, G. Reference to kinds across language. Natural language semantics, 6(4):339–405, 1998.

CHIERCHIA, Gennaro; OLIVEIRA, Roberta Pires de. Contemporary Issues in Natural Language Semantics: an interview with Gennaro Chierchia. DELTA: Documentação de Estudos em Lingüística Teórica e Aplicada, 2020, vol. 36.

DAYAL, Veneeta. Any as Inherently Modal. Linguistics and Philosophy, 21(5):433–476, 1998.

DAYAL, Veneeta. The universal force of free choice <any>. Linguistic Variation Yearbook, 4:5–40, 2004.

HASPELMATH, Martin. Indefinite pronouns. Oxford: Oxford University Press, 1997.

HORN, Laurence. Pick a theory, not just any theory. In Horn, L. and Kato, Y., editors, Negation and Polarity. Syntactic and Semantic Perspectives, pages 147–192. Oxford University Press, Oxford, 2000

KADMON, Nirit; LANDMAN, Fred. Any. Linguistics and Philosophy, 16(4):353–422, 1993.

KRATZER, Angelika; SHIMOYAMA, Junko. Indefinite pronouns: The view from Japanese. Paper Delivered at the Third Tokyo Conference on Psycholinguistics, 2002

KRIFKA, Manfred; PELLETIER, Francis; CARLSON, Greg; TER MEULEN, Alice; LINK, Godehard; CHIERCHIA, Gennaro. Genericity: An introduction. In CARLSON, Greg; PELLETIER, Francis, editors, The Generic Book, pages 1–124. University of Chicago Press, 1995

LEGRAND, Jean. Or and Any: The Semantics and Syntax of two Logical Operators (Ph. D. dissertation). University of Chicago, 1975

MENÉNDEZ-BENITO, Paula. The Grammar of Choice. University of Massachusetts Amherst Amherst, MA, 2005

MENÉNDEZ-BENITO, Paula. On universal free choice items. Natural Language Semantics, 18(1):33–64, 2010

POLAKOF, Ana. The negation of cualquier NP. Isogloss. Open Journal of Romance Linguistics, 2021, 7/8: 1-24

RECANATI, Francois. Descriptions and Situations. In : A. Bezuidenhout and M. Reimer ( eds . ) , Descriptions and Beyond. Oxford : Oxford University, pp. 28-53, 2004.

RIVERO, Maria. Un Desconocido Cualquiera. Cuadernos de la Asociación de Lingüıstica y Filologıa de la América Latina 3, pp. 60–80, 2011

RYLE, Gilbert. The concept of mind. Hutchinson. London, UK. 1949

SEARLE, John. Expression and meaning: Studies in the theory of speech acts. Cambridge University Press, 1985.

VENDLER, Zeno. Each and Every, Any and All. Mind, 71(282):145–160, 1962

VENDLER, Zeno. Each and every, any and all. In Linguistics in Philosophy. Cornell University Press, 1974.

ZEEVAT, HENK. Applying an Exhaustivity Operator in Update Semantic. In H. Kamp, (editor) Ellipsis, Tense and Questions. ILLC, Amsterdam. DYANA2 deliverable R 2.2.b. 1994.