Carl,
Within a region, you may have multiple MSTP instances (up to 16), each having its own number. Additionally, the MST Instance 0 is called Internal Spanning Tree, IST. All these instances run in the whole region.
Things start to differ when we look at the behavior of these instances at the region boundary. The IST is the only instance that is allowed to span beyond the region boundary. All other instances remain confined within the region and never "leak" beyond its boundary but the IST talks to the outside world and decides the state and role of ports on the region boundary. To the outside world, the boundary port speaks with spanning tree values derived from the IST, thus it appears as a port in a single spanning tree. Also, the state of the boundary port derived by the IST interaction with the outside world is replicated onto all internal MSTP instances, i.e. if a boundary port is root forwarding because of IST negotiation, it is going to be root forwarding for all internal instances, and analogously for all other possible combinations of roles/states.
If you plug an MSTP switch to an RSTP switch, the MSTP switch will talk to the RSTP switch as if there was only a single RSTP instance running inside the region - all BPDUs it will send will be initialized with values taken from the IST. The external RSTP switch will think it speaks with another RSTP switch with just one spanning tree. If the MSTP switch decides to block the boundary port, it will be blocked for all VLANs, and if it decides to make some port a forwarding port, it will again be forwarding for all VLANs (because the state of the IST is replicated onto all MSTP instances)
I am not sure if I managed to put this simply... you are more than welcome to ask further.
The actual behavior is more complicated because of Cisco's PVST/RPVST but let's for now remain only with pristine 802.1Q/802.1D without proprietary additions.
Best regards,
Peter
