diff --git a/pkg-r/NEWS.md b/pkg-r/NEWS.md
index e5f5213f..02d1e891 100644
--- a/pkg-r/NEWS.md
+++ b/pkg-r/NEWS.md
@@ -1,5 +1,7 @@
 # commons (development version)
 
+* A Snowflake semantic view whose base tables are named with quoted identifiers, such as `"orders"`, now scopes to those tables. The quotes were kept as part of each name, so `data_source()` could not find them (#446).
+
 * A definition that adds `NULL` to a date, such as `created + NULL`, now compiles to SQL that Snowflake accepts. It compiles to a typed `NULL` timestamp (#441).
 
 * A definition that divides by a compound expression, such as `amount / (ratio * 2)`, now compiles to the correct Snowflake and Databricks SQL. The parentheses around the divisor were dropped, so the query divided by `ratio` alone (#439).
diff --git a/pkg-r/R/catalog-snowflake.R b/pkg-r/R/catalog-snowflake.R
index a52b89f0..3f2a44bf 100644
--- a/pkg-r/R/catalog-snowflake.R
+++ b/pkg-r/R/catalog-snowflake.R
@@ -680,9 +680,12 @@ snowflake_semantic_dependencies <- function(id, specification) {
   complete <- TRUE
   for (table in semantic_spec_entries(specification$tables)) {
     base <- table$base_table %||% list()
-    table_name <- base$table
-    catalog <- base$database %||% base$catalog %||% model[["catalog"]]
-    schema <- base$schema %||% model[["schema"]]
+    # A name the YAML gives follows SQL identifier rules; one it leaves out
+    # comes from the view's own id, which is already exact.
+    table_name <- snowflake_yaml_identifier(base$table)
+    catalog <- snowflake_yaml_identifier(base$database %||% base$catalog) %||%
+      model[["catalog"]]
+    schema <- snowflake_yaml_identifier(base$schema) %||% model[["schema"]]
     if (
       !rlang::is_string(catalog) ||
         !rlang::is_string(schema) ||
@@ -706,6 +709,19 @@ snowflake_semantic_dependencies <- function(id, specification) {
   list(ids = ids[!duplicated(keys)], complete = complete && length(ids) > 0L)
 }
 
+# A name in a semantic view's YAML follows SQL identifier rules: a quoted one
+# is exact, with a doubled quote standing for one, and an unquoted one folds to
+# upper case.
+snowflake_yaml_identifier <- function(name) {
+  if (!rlang::is_string(name)) {
+    return(NULL)
+  }
+  if (grepl('^".*"$', name)) {
+    return(gsub('""', '"', substr(name, 2L, nchar(name) - 1L), fixed = TRUE))
+  }
+  toupper(name)
+}
+
 snowflake_semantic_context <- function(
   specification,
   dimensions,
diff --git a/pkg-r/tests/testthat/test-catalog-snowflake.R b/pkg-r/tests/testthat/test-catalog-snowflake.R
index cdb8456f..fb2184be 100644
--- a/pkg-r/tests/testthat/test-catalog-snowflake.R
+++ b/pkg-r/tests/testthat/test-catalog-snowflake.R
@@ -93,6 +93,36 @@ test_that("Snowflake semantic YAML imports only public dimensions and metrics",
   expect_equal(model$metrics[[1]]$parent, "orders")
 })
 
+test_that("Snowflake semantic dependencies read base tables as SQL identifiers", {
+  # The YAML keeps a quoted name's quotes, and folds an unquoted one to upper
+  # case the way the SQL that created it would.
+  specification <- list(tables = list(
+    list(base_table = list(
+      database = '"Commons"',
+      schema = "live",
+      table = '"order ""lines"""'
+    ))
+  ))
+  id <- DBI::Id(catalog = "COMMONS", schema = "LIVE", table = "MODEL")
+
+  model <- snowflake_semantic_model_from_spec(id, specification)
+
+  expect_identical(
+    model$dependencies[[1]],
+    DBI::Id(catalog = "Commons", schema = "LIVE", table = 'order "lines"')
+  )
+
+  # A namespace the YAML leaves out is the view's own, which is exact.
+  inherited <- snowflake_semantic_model_from_spec(
+    DBI::Id(catalog = "MyDb", schema = "MySchema", table = "MODEL"),
+    list(tables = list(list(base_table = list(table = "orders"))))
+  )
+  expect_identical(
+    inherited$dependencies[[1]],
+    DBI::Id(catalog = "MyDb", schema = "MySchema", table = "ORDERS")
+  )
+})
+
 test_that("Snowflake semantic scope retains dependencies and public context", {
   specification <- list(
     custom_instructions = "Use booked revenue for finance questions.",