Effect of large text field data on queries

[Glen Parker <glenebob(at)nwlink(dot)com>]

I am having some query problems on a table with large text fields.

The table contains 6.7M rows.  It is vacuumed every night, and since the 
last vacuum, rows have been inserted only, never updated or deleted.

There are many large text field values in one text field, some in excess 
of 6MB.

I have another table with 13.8M rows, several times as many columns, but 
no large text values.

I am testing select count() queries on both tables, counting on an 
integer field only.  The first table takes almost 300 seconds to 
complete, while the second table takes about 90 seconds.

Why would very large text values effect the speed of a seq scan that 
does not actually evaluate those values?

Oh, PG 8.1.3 on FC 5, 64-bit.

TIA
-Glen Parker