There have been many improvements in motor construction over the last hundred years. I'd guess you're probably using a two-pole motor, since many project books describe such motors; a three-pole motor will likely work much better. In a two-pole motor, you must make sure there's a big enough gap between poles to ensure that the commutator never poses a short across the supply. Assuming that's the case, you may be able to reduce arcing by installing a pair of back-to-back Zener diodes across the coil (on the rotor). Something like 12-15 volt diodes should probably be pretty helpful. Make sure to keep an eye on their temperature, though; depending upon various factors, they may end up absorbing a fair amount of energy. Zener diodes can tolerate getting pretty hot, but beyond a certain point they may fail shorted, and the current from a lantern battery may be sufficient to make them fail dramatically. Still, I would expect that 12-15 volt diodes may help with your arcing problem.
Another thing I would suggest is that you avoid metal-on-metal contact; carbon brushes are apt to be far less problematical. Pure copper tends to oxidize quickly, and copper oxide is an insulator. If you can put a reasonably smooth layer of solder on your commutator that may improve things considerably.
An alternative approach (since this is an electronics forum after all) might be to add to the motor a position sensor, and then use electronics to switch the current to it. Two approaches may be useful here: (1) keep the mechanical commutator to reverse polarity, but use electronics to ramp down the current flowing to the motor before the commutator switches, so that the commutator never has to switch under load; (2) replace the commutator with side-by-side solid rings, and then use a circuit called an "H-bridge" to control the polarity. Both approaches would enable some explorations which would not be possible with a purely-mechanically-commutated motor.