When you go out to do a recording session of something you control like dropping rocks, do you press record say take one "dropping rocks", then drop the rocks, press stop. Or do you keep recording and cut up the takes later when editing? I tend to just keep rolling and say take one, take two.... until that session is done. Your preference?
If I have to reset between takes, then I will usually stop the machine and start a new take. When I recorded the sounds for my rocks library, this is the way I did it. We had to set up the rocks for each take, so I stopped the machine and reslated each one. When I recorded my dog library, I ran a take for each dog and only stopped if I need to wait for a plane or something. These takes would be 5 plus minutes long.
On somethings, like guns, I like to listen back to the recordings often, so I'll do multiple takes. For example, a take of single shots, a take of short bursts, a take of long bursts for each weapon. So I guess, I do it either way depending on what/how I'm recording.
I think the "slate-perform action-stop take" approach was more of a requirement when the physical media (1/4", DAT) was limited. Now with 8GB and upwards so readily available and affordable, I tend to record very long takes (esp. ambience, nature, traffic, etc) and use voice slates to identify what I'm doing, what take it is, what's different this time, etc.
The byproduct of this approach, however, is a longer mastering session!
Sounds like a trend brewing with these answers! :-)
If it's easy to repeat the sound performance and the takes are short, I too will just roll every take into one big audio file and verbally slate what I think is a good vs. bad take. My verbal slates for bad takes are usually four letters in length. :-)
Jay's also right that you then need to "spot" your own SFX takes to find the good/bad takes...if you can, written notes help a TON.
However, for some takes where the actual take is long or takes a while to turn around for another take (vehicle 'bys, longer actions), I'll record single takes per file. This lets me use the insanely useful circle-take and false-take markers on my Sound Devices recorder to help only transfer the good takes when I enter the editing mode.
I'll do long takes but if I change the action of creating the sound i.e. from dropping the rock to rolling the rock or changing the items i.e small glass bottle smash to big glass bottle smash, I will announce it. I still have to listen through the takes to find the one I want.
This of course all depends on the level of detail you want to make note of when you record the sound.
Always leave the recorder running, otherwise you will miss something. Make lots of notes on paper with the timings. It is incredibly useful to verbally slate but not always appropriate, especially if you have limited time. Also avoid wiping your memory cards until you absolutely have to. What you thought was a bad take is often a good take for a completely unrelated project.
I always do long takes and almost never do any voice announcing. For one, when you go to edit the file you are increasing your workload because you have to edit out the voice in addition to editing the sfx. Also, if you think about it ... when you drop a rock 5 times, you will see this in the .wav files when you go to edit, so you can name them as you edit them
A long file is better because you only need to import one file into your DAW, instead of 20 or whatever.
The only time I've announced is when the sound is very specific, like an exact type of antique car engine, or a specific type of gun for example
Otherwise, you'll most likely recognize and remember what is was you recorded so don't bother wasting time announcing and stopping and restarting the record button. You can do all the naming and organizing later in the process.