Have a look here for an explanation for different OS's:
PPP capture setup
I assume you are using Linux, here's the relevant portion:
On Linux, you won't be able to capture PPP control protocol traffic in the usual manner (via libpcap) as that traffic is not supplied to the networking stack. You will be able to capture IP traffic, for example, but you won't be able to see the PPP headers, as the PPP code doesn't supply them to the networking stack.
The PPP control protocol traffic can be captured by configuring the ppp daemon to 'record' to a capture file all the data the daemon sends and receives. Wireshark can then be used to display the created capture file.
On Fedora Core 6 the pppd capture file is created if 'record filename-of-your-choice' is added as a line in /etc/ppp/options (YMMV: see 'man pppd').
Note that all traffic on the PPP port is captured to the file so this option, if left on, may cause a large capture file to be generated.