American Educator has a 19-page article titled Teaching English Language Learners: What the Research Does and Does Not Say (pdf) by Claude Goldenberg, Professor of Education at Stanford University. The article looks at the findings of two reviews of the research, one by the National Literacy Panel and the other by the Center for Education, Diversity, and Excellence. Here's an excerpt from the conclusion:

Although there are numerous areas in which there is insufficient research to guide policy and practice, we can lay claim to some things that matter for the education of ELLs. Chief among these is that 1) teaching children to read in their primary language promotes reading achievement in English; 2) in many important respects, what works for learners in general also works for ELLs; and 3) teachers must make instructional modifications when ELLs are taught in English, primarily because of the students’ language limitations.

Practically, what do these findings and conclusions mean? In spite of the many gaps in what we know, the following is the sort of instructional framework to which our current state of knowledge points:

• If feasible, children should be taught reading in their primary language. Primary language reading instruction a) develops first language skills, b) promotes reading in English, and c) can be carried out as children are also learning to read, and learning other academic content, in English.
• As needed, students should be helped to transfer what they know in their first language to learning tasks presented in English; teachers should not assume that transfer is automatic.
• Teaching in the first and second languages can be approached similarly. However, adjustments or modifications will be necessary, probably for several years and at least for some students, until they reach sufficient familiarity with academic English to permit them to be successful in mainstream instruction; more complex learning might require more instructional adjustments.
• ELLs need intensive oral English language development (ELD), especially vocabulary and academic English instruction. However, as the sidebar on critical unanswered questions explains (see p. 12), we have much to learn about what type of ELD instruction is most beneficial. Effective ELD provides both explicit teaching of features of English (such as syntax, grammar, vocabulary, pronunciation, and norms of social usage) and ample, meaningful opportunities to use English—but we do not know whether there is an optimal balance between the two (much less what it might be).
• ELLs also need academic content instruction, just as all students do; although ELD is crucial, it must be in addition to—not instead of—instruction designed to promote content knowledge.