The outputs of the '138 are low-active; an LCD E input (high for >700nS) latches stuff on its *falling* edge - this is probably the error. Following this, are you sure your display initialises correctly?

I am currently working on a C routine to drive an LCD attached to an Atmel evaluation board (chopped about to work with 8752s). The LCD is memory-mapped. If you give me a fax number, I can send you the sketch of the circuit.

BTW my LCD "bible" is Densitron application note AP-LC-0, dated August '83 - still valid!

Regards,
Richard