Todd -
If your Temp tablespace is 3.5 Gb then your queries have needed to do some pretty big disk sorts.
There could be a variety of reasons why that's the case
- is your system big and 3.5Gb is a reasonable TEMP ts size?
- are there queries with poor joins in them ?
- are you letting users loose with reporting tools such as BO, Crystal or Actuate ?
(I don't actually want answers to those questions, I'm just suggesting that they're questions you should be asking yourself !!

)
Anyway, if you reckon a 3.5Gb Temp tablespace is outrageous for your system, then Santa's suggestion will limit the size and you'll soon find out who's running the dodgy queries when they phone up with 'failed to extend' errors.
Personally, I prefer Automatic UNDO since
- it is the Oracle default from 9 onwards,
- it's less hassle
(we don't always have other DBAs to hand to extend it manually)
- occasionally users do have to run big one-off queries
- if it gets really big, it's a 1-minute job to shrink it
1) log on to OEM
2) create a new tablespace TEMP2, as type 'temporary', and checking on 'set as default temporary tablespace', say about 200Mb
3) make sure the datafile is set to autoextend
4) as soon as queries using the original TEMP have finished, drop TEMP and delete the OS file(s) for it.
The best approach will depend on your own circumstances.
Best of Luck
Steve