@b00n Het’s answer is short and sweet and correct; but I’ll expand on it a bit, to back it up and consider the complications raised in @ryang’s answer.
The key issue is that only in natural English is a bit subtle in several ways, and sometimes ambiguous. The main ambiguity is that only doesn’t have to directly precede the specific item it’s governing — it usually precedes the main verb. So for instance William only mows the lawn on Mondays can potentially be read as:
- [only governs Mondays] It’s only on Mondays that William mows the lawn (not on other days).
- [only governs the lawn] It’s only the lawn that William mows on Mondays (not the field beyond).
- [only governs mows] The only thing William does to the lawn on Mondays is mowing it (not raking it).
- [only governs mows the lawn] The only thing William does on Mondays is mows the lawn (then he goes straight back to bed).
In practice, usually it’s clear from context which reading is the intended one — and when it wouldn’t be clear, then speakers do place only directly with the phrase it governs, and say for instance William mows the lawn only on Mondays. However, as long the intended reading is clear from context, only prefers to precede the main verb — and so when only is in this position, it should be read as governing whichever subsequent part of the phrase is semantically most natural. So in William only eats ice cream on Mondays, other readings are in principle possible, but it’s semantically clear that the intended reading is William eats ice cream only on Mondays.
The other is to what extent only entails that the thing in question happens at all. If I say “I’ve only been to Brazil twice”, you will rightly conclude that I have been to Brazil twice. If I say “I only visit democratic countries”, you will conclude that I have probably visited some democratic countries — but not that I’ve visited all of them. So this positive entailment is present but subtle; as such, it’s hard to judge whether it’s part of the logical content of the only sentence, or just implicature. In formal usage (mathematical and some legal), it’s well-established that the positive entailment is not taken as part of the logical content; in other contexts, the positive entailment is usually analysed as implicature, but is at least arguably part of the logical content. But in any case, the positive claim entailed is only existential, not universal.
So William only eats ice cream on Mondays has a couple of reasonable readings; writing $P(t)$ for “William is eating ice cream at time $t$ and $Q(t)$ for “$t$ is a Monday” as you suggest, these are:
$\forall t\ P(t) \Rightarrow Q(t)$, “the only time William may eat ice cream is on Mondays”
$(\forall t\ P(t) \Rightarrow Q(t)) \wedge (\exists t\ Q(t) \wedge P(t))$ “the only time William may eat ice cream is on Mondays; and he sometimes does eat it on Mondays”
The first of these is the more standard reading, especially in formal or semi-formal contexts. The second reading is also defensible in natural contexts. But neither of these is a bi-implication; the potential bi-implicative reading $(\forall t\ P(t) \Leftrightarrow Q(t))$, “at all times, William is eating ice cream precisely if it’s Monday”, is not justified either by natural usage or formal conventions.