Kai,

There's indeed no difference between putting pin0 to ground by means of the switch or directly via a wire. But that's not the real problem...
The problem is that the level doesn't go high anymore, after I opened the switch or (in the second case without the switch) removed the direct wire from ground.

The only way to bring the level high again, is connecting pin0 directly to +5Vdc (which isn't the correct way, of course... that I know).

So, the problem is not 'going low', but 'going high' again.
I just gave the extra test with connecting pin0 directly to +5Vdc/ground as an extra input, in case it might help someone to understand what's going wrong.

The only difference, though, is that in one case the +5Vdc is 'hard' connected to the pin, in the other case the current through pin0 is limited by the resistor (which isn't too high to me).

My switch isn't damaged: if I remove the wire going to pin0 of the IO expander and I'm measuring between the resistor and the switch, then the level goes to 0V when pressed and comes back to +5Vdc when released. So, nothing's wrong with the switch.

I will do the check by only reading the PCF8574, that's a good idea (although I prevent the input pin from being written with a '0', see the answer I gave to Eric's remark above).

Indeed, never connect any IO directly to +5Vdc. That's obvious, but as already said, I only did it to see if there's a different behaviour. And there is, but I can't explain for the moment.

As you can see on the circuit, I'm definitely using a resistor between the +5Vdc and pin0... ;-)

Best rgds,

--Geert