There is 2 behavior of numpy.where(condition, [x ,y]). Theano always support you provide 3 parameter to where(). As said in NumPy doc[1], numpy.where(cond) is equivalent to nonzero().
You can do it like this in Theano:
import theano
import numpy as np
v = np.arange(10)
var = theano.tensor.vector()
out = theano.tensor.eq(var, 2).nonzero()[0]
print out.eval({var: v})
Check line 5. NumPy nonzero() return a tuple. Theano do the same. There is one vector in that tuple per dimensions in the input of nonzero().
[1] http://docs.scipy.org/doc/numpy/reference/generated/numpy.where.html