We claim that eigentlich is an example of a particle that has no semantic (truthconditional) meaning, but signals which role a speech act plays within the given discourse by blocking contextually salient conclusions that otherwise might be drawn by the recipient.

The German particle eigentlich is syntactically used as a sentential modifier, which is usually translated as actually. (We consider the translation as a rough approximation only. We do not wish to make any claims on the use of actually.)

Peter: Kommst Du heute abend zur Party? (Will you come to the party tonight?)

Paul: (a) Ich möchte kommen. (I'd like to come.)
(b) Ich habe wenig Zeit. (I don't have much time.)
(c) Ich möchte kommen, aber ich habe wenig Zeit.
(I'd like to come, but I don't have much time.)

(d) Ich möchte eigentlich kommen, aber ich habe wenig Zeit.
(Eigentlich, I'd like to come, but I don't have much time.)
(e) Ich möchte kommen, aber ich habe eigentlich wenig Zeit.
(I'd like to come, but, eigentlich, I don't have much time.)

By uttering (a) or (b) Paul gives a hint whether he will come (in case of (a)) or not (in case of (b)). By uttering (c) Paul leaves this widely undecided. (d) gives rise to the expectation that Peter will not come, while (e) gives rise to the expectation that he will come. In each case Peter expects a relevant answer to his question (cf. Grice's maxim of relation) and therefore is inclined to interpret Paul's answer accordingly. From (a) Peter concludes that Paul will most probably come, because Paul gives only a fact supporting this conclusion. (b) works in the opposite way. From the clauses of (c) Peter can draw contradictory conclusions. We claim that (d) has the same truthconditonal meaning as (c), but eigentlich blocks the conclusion from the first clause, that Paul will most probably come. (e) works accordingly by blocking the contrary conclusion from the second clause.

This and similar examples lead to the hypothesis that eigentlich(p), where p is a proposition,

• has no other semantic (truthconditional) meaning than p itself,
• does not presuppose anything else than p itself,
• but blocks contextually salient conclusions that may be drawn from p.

The hypothesis was successfully tested against the 170 occurrences of eigentlich in the Limas corpus, and it can be shown that it can be applied even to those cases where eigentlich seems to mean a negation (Eigentlich, I finished my paper, but I have still to write the introduction).

We will propose a framework where agents are characterized by their intentional and belief states and where the role of eigentlich can be formally explicated. Meanings of speech acts are regarded as functions changing the intentional and belief states of agents. A speaker S asserting p is understood if the hearer H knows that the speaker want him to believe that p (ISBHp is added to the hearer's knowledge). Grice's maxim of relation allows the hearer to draw all contextually salient conclusions r from p, i.e. all conclusions r from p for which it is common ground that the hearer wants to know whether they are the case or not (CG IH (BHr Ú BHØr)). A speaker asserting eigentlich(p) is also understood if the hearer knows that the speaker want him to believe that p (ISBHp is added to the hearer's knowledge). But this time the speaker does not want the hearer to draw all salient conclusions (Ø"r[CG IH (BHr Ú BHØr) Ù CG(p ~> r) Ù ISBHp ® ISBHr]).