Each sublist forms a row and each column has matching index number elements from each sublist. The index is zero based, so column one has index 0. The value in row three and column one is matrix34 --> 3.5
#-*- coding:utf-8 -*-
matirx = 
for i in xrange(3):
print "Enter a 3-by 4 matrix row for row %d" % i
row = map(float, raw_input().split(' '))
print "The matrix is"
for row in matirx:
revmatirx = map(list, zip(*matirx))
for i in xrange(4):
print "Sum of elements for column %d is" % i
Also, if this is just an exercise, then doing it with native python datastructures (like lists) is a nice way to go. In reality, you should use numpy to make a matrix. It's a library for just this, and will be about 1000 times faster than a matrix made out of python datastructures, due to its c-level optimizations.