on unbelievable ell student subgroup is the most honest able among other subgroup of students as soon as they become proficient they are out of ell categories so ell category by nature stays low performing always because again as the the the moment they become proficient they are out of these categories and then the new arrival are added so i don't know how el students can perform at higher level of proficiency if that's the case if more professions are out of these categories and less profession can come in fact number six these students their teachers and their parents are often blamed for the lack of progress in the schools why because we don't provide them enough instruction appropriate accommodation why should they become blamed for their lack of proficiency fact number seven assessments that are constructed and field tested for native english speakers may not be relevant for ell students why is that i'm going to discuss and give you a couple of evidence some data here these are reliability coefficients for english language learners and and regular students so i wanted to you to pay attention to this and compare these are the reliabilities here as you know reliabilities range between zero to one if it has has a reliability of zero meaning that the test is not reliable at all hundred percent measurement error it has two reliability of one means that there is no measurement error is a perfect test we don't have any tests at either of those ends it's something degree between zero to one so a good test usually should have a reliability of 0.8 or above as you can see for english language learners for whom the tests or tests were constructed all of them has above 0.8 there is even not one instance that you don't have anything above 0.8 which is good and that serves the purpose because these tests are constructed for for these students but take a look at these tests for english language learners and take a look at some of the reliability reliability of 0.5 means 50 percent of the score is measurement error of the results of outcome imagine that these tests these are a status these tests are used for accountability purposes for no child left behind purposes once and higher sakes decision making are based on these assessment with almost like a flipping a coin 50 that's exactly what it is so these assessments are used for classification for ayp reporting for uh placement for graduation for all of these this and this is the situation again these data have been cross-validated for many different states okay uh we did a study this is a more psychometric issues we did a study and we tried to see test items how they load we use factor analysis and try to see test items how they load on first factor the assumption is that if the test items measure the same construct then all of them should load on the first factor and this is the case for non-ele students as you can see for instance we have a test for this case we have a test of 50 items in math and what it says all individual test items are 50 items measuring the math and all of them saying the same thing a common underlying construct but when we go to ells students test items measure only 50 percent of the variance so that means there are some other factors involved what are those factors those factors are language complexity and cultural issues so when you give assessment to a group of students ell students and when they have low performance we don't know whether that low performance is due to lack of understanding of the question or whether it is lack of content knowledge that's the issue it's interesting i gave a i wrote a paper submitted for publication the title of the paper was impact of language on assessment the editor actually changed the title and when i saw my paper the title was completely changed but i liked it you know what she said what the editor said would you explain the question to me she put that as the title of the test so ella that's it their question would you explain what do you want me to do because i don't understand the question i have the content knowledge but i don't understand the question because of complex linguistic structure of the question so that's what what it says only 50 of the construct is measured by this test uh but the rest of them is language which is construct irrelevant the concept relevant is the math in this test and language is construct irrelevant you