I am working on a raw socket ICMP message receiver and I am having a bit of trouble with knowing when to byte swap. I already know that any field longer than a byte in the IP header have to be swapped from network byte order to hosts byte order. The thing that confuses me is that the ICMP header appears to be byte swapped by word. Does that mean all the data in the IP payload is byte swapped by word? Basically the way I found out that the ICMP header was getting swapped was that the Code and Type are coming up in the wrong place in the destination unknown messages. Any help you guys could throw my way would be ... helpful.

Just look at the RFC, and use htons/ntohs (16-bit) and htonl/ntohl (32-bit) as needed.

For example, here is RFC792:
http://www.ietf.org/rfc/rfc792.txt

I have already looked into the RFC for both IP and ICMP. Neither mention why the header for ICMP is being byte swapped by word. Neighter RFC mentiones byte swapping in the case I mentioned above. IP doesn't mention byte swapping payload data, only that in multi-byte numbers have the most signifigant bytes first. And of ICMP doesn't mention byte order at all.
http://www.faqs.org/rfcs/rfc791.html
http://www.faqs.org/rfcs/rfc792.html

Fixed. I'm sorry, my ignorance of sockets caused me to place the blame on unknown bytes being swapped when in fact my IP header struct had 1 short as a long which resulted in a shift in the entire structure. As a result, when I byte swapped the IP addresses, I was really byte swapping the first 4 bytes of my header.