Since 2008 U.S. banks have been able to earn interest on their reserves: This pins down a minimum interest rate that banks can earn: They never lend for less than that amount. For decades longer, U.S. banks have been able to borrow funds from the Federal Reserve's discount window: This pins down a maximum interest rate at which banks will borrow (plus or minus some details).
This system, where no bank pays more than the discount rate to borrow funds and earns no less than the interest rate paid on excess reserves to lend funds is known as a "channel system" or "corridor system" of monetary policy. The Fed's lending and borrowing rates pin down a corridor, and short-term rates typically stay within that corridor. Why pay more than the Fed's rates to borrow? Why earn less then the Fed's rates when lending? The corridor sets rates in the safe-return market.
The rise of the channel/corridor system is one reason I suspect that it's reasonably fair to think that under our current policy regime the short term interest rate is an instrument. As David correctly notes, the Federal Reserve does buy and sell government bonds in order to influence short term interest rates. It's possible that this channel is still crucial, I don't claim to know whether it still is. But I do know that our Fed uses the rate on excess reserves as one instrument of monetary policy and uses the slightly higher discount rate as another instrument of monetary policy; together they do a lot to pin down our new corridor system of short term interest rates.
Fun Fact: There's evidence that even before the rise of the corridor approach, central banks could sometimes move short-term interest rates without buying or selling government bonds. Central banks could move interest rates just by announcing their new target.
This approach is known as "open mouth operations"; Guthrie and Wright had a nice paper in the Journal of Monetary Economics showing how open mouth operations worked in New Zealand: Rates routinely moved to where the head of the central bank told them to move without any change in bank reserves. The St. Louis Fed has a one-pager on a Swiss example as well. So sometimes you don't need to buy bonds to cut interest rates, sometimes you don't need to sell bonds to raise them.
In my dissertation I attempted to find evidence of open mouth operations in the U.S. Yes, I did find some evidence, but just as in the Swiss case, it was a "sometimes" situation: Sometimes the Fed bought a lot of bonds when it wanted to cut rates, and sometimes it didn't. A noisy relationship--the kind that central bankers can discuss amongst themselves but which leads to an empirical muddle. Some interesting ideas are difficult to test, and I'd like to think that the theory of open mouth operations falls into that category....