A more robust approach to testing the findings at http://rickystewart.wordpress.com/2013/09/03/why-sorting-an-array-makes-a-python-loop-faster/ that factors out all the "buts" that immediately came to mind.

compare_sorted_unsorted_iteration.py

import timeit

sorted_setup = """
a = range(1000000)
"""

unsorted_setup = """
from random import shuffle
{}
shuffle(a)
""".format(sorted_setup)

forloop_stmt = """
for i in a:
    pass
"""

comprehension_stmt = """
[i for i in a]
"""

print "Checking timings."

# How many times to run the statement.
n = 1000

# For loop
t_sorted_forloop = timeit.timeit(forloop_stmt, sorted_setup, number=n)
t_unsorted_forloop = timeit.timeit(forloop_stmt, unsorted_setup, number=n)

# List comprehension
t_sorted_comprehension = timeit.timeit(comprehension_stmt, sorted_setup, number=n)
t_unsorted_comprehension = timeit.timeit(comprehension_stmt, unsorted_setup, number=n)

print "Sorted for loop:"
print t_sorted_forloop
print
print "Unsorted for loop:"
print t_unsorted_forloop
print
print "Sorted comprehension:"
print t_sorted_comprehension
print
print "Unsorted comprehension:"
print t_unsorted_comprehension
print

results.txt

Results from running this on my Mac, OS X 10.6.8, 2.7GHz Intel Core i7 yadda yadda:

Sorted for loop:
18.2923810482

Unsorted for loop:
56.756278038

Sorted comprehension:
124.576835871

Unsorted comprehension:
126.68524313