--- a
+++ b/docs/explanation/backend-tables.md
@@ -0,0 +1,100 @@
+OpenSAFELY provides secure access to real data
+via different OpenSAFELY backends.
+
+For example, there is a TPP backend,
+for querying patient data held by TPP.
+
+The tables that are available for use in ehrQL
+depend on which OpenSAFELY backend that you access.
+
+We can consider two kinds of table:
+
+* core tables
+* backend-specific tables
+
+## Core tables
+
+If you have read these documentation pages in the suggested order,
+then all the examples you will have seen access the _core_ tables.
+
+The core tables are intended to be compatible with any OpenSAFELY backend providing primary care data.
+If you *only* use the core tables,
+then your ehrQL dataset definition will be compatible with different OpenSAFELY backends,
+without requiring any changes.
+
+### Importing core tables
+
+Make the core tables available for use in a dataset definition
+with import statements like:
+
+```python
+from ehrql.tables.core import medications, patients
+```
+
+where the `ehrql.tables.core` specifies that we are using the core tables.
+
+## Backend-specific tables
+
+Different OpenSAFELY backends may opt to provide additional data tables other than the core tables.
+You may wish to access these tables if the core tables are not sufficient
+to answer the research question of interest.
+
+:warning: When using backend-specific tables in a dataset definition,
+your dataset definition will be only compatible with that backend.
+
+:notepad_spiral: Implementation of other backends is still in development.
+But all data in the EMIS OpenSAFELY backend is available via cohort-extractor.
+
+### Importing backend-specific tables
+
+Instead of referring to `core` in the import statement,
+we use the name of the backend.
+
+For example, for TPP-specific tables,
+we use `tpp` in the import statement:
+
+```python
+from ehrql.tables.tpp import addresses, patients
+```
+
+:notepad_spiral: In this example,
+the `addresses` table is specific to the TPP backend.
+The `patients` table is a core ehrQL table.
+We import both core and backend-specific tables within a single import statement.
+
+## Using the table schema reference
+
+In the examples given so far,
+the names of the table schemas, tables and columns
+have been provided for you.
+
+For example,
+in the initial [dataset definition](../tutorial/working-with-data-with-ehrql/index.md) of the tutorial,
+we had the following line:
+
+```python
+from ehrql.tables.core import patients, practice_registrations, clinical_events, medications
+```
+
+* `core` is the *table schema*
+* `patients`, `practice_registrations`, `clinical_events` and `medications` are the *table names*
+
+We also used the `show()` command to display the `patients` table and saw that the *table columns*
+available were `date_of_birth`, `sex` and `date_of_death`.
+
+Use the [table schema reference](../reference/schemas.md)
+to look up which schemas and columns are available.
+
+The table schema reference explains:
+
+* which backends support the schema
+* the table column names
+* the table column data types
+* any additional constraints on table column values
+* additional contextual information about table columns
+* whether table columns contain at most one row per patient,
+  or may contain multiple rows per patient
+
+:grey_question: Consult the [`core`](../reference/schemas/core.md) schema.
+Choose any of the tables there
+and understand its structure from the schema.