diff --git a/conservancy_beancount/reports/ledger.py b/conservancy_beancount/reports/ledger.py
index 3dfe589..10409f1 100644
--- a/conservancy_beancount/reports/ledger.py
+++ b/conservancy_beancount/reports/ledger.py
@@ -60,6 +60,7 @@ from typing import (
     Mapping,
     Optional,
     Sequence,
+    Set,
     TextIO,
     Tuple,
     Union,
@@ -401,12 +402,13 @@ date was also not specified.
 """)
     parser.add_argument(
         '--account', '-a',
-        dest='sheet_names',
+        dest='accounts',
         metavar='ACCOUNT',
         action='append',
         help="""Show this account in the report. You can specify this option
-multiple times. If not specified, the default set adapts to your search
-criteria.
+multiple times. You can specify a part of the account hierarchy, or an account
+classification from metadata. If not specified, the default set adapts to your
+search criteria.
 """)
     parser.add_argument(
         '--sheet-size', '--size',
@@ -436,18 +438,18 @@ metadata to match. A single ticket number is a shortcut for
 `rt-id=rt:NUMBER`. Any other word is a shortcut for `project=TERM`.
 """)
     args = parser.parse_args(arglist)
-    if args.sheet_names is None:
+    if args.accounts is None:
         if any(term.meta_key == 'project' for term in args.search_terms):
-            args.sheet_names = [
+            args.accounts = [
                 'Income',
                 'Expenses',
                 'Assets:Receivable',
+                'Liabilities:Payable',
                 'Assets:Prepaid',
                 'Liabilities:UnearnedIncome',
-                'Liabilities:Payable',
             ]
         else:
-            args.sheet_names = list(LedgerODS.ACCOUNT_COLUMNS)
+            args.accounts = list(LedgerODS.ACCOUNT_COLUMNS)
     return args
 
 def diff_year(date: datetime.date, diff: int) -> datetime.date:
@@ -483,14 +485,26 @@ def main(arglist: Optional[Sequence[str]]=None,
     returncode = 0
     books_loader = config.books_loader()
     if books_loader is None:
-        entries, load_errors, _ = books.Loader.load_none(config.config_file_path())
+        entries, load_errors, options = books.Loader.load_none(config.config_file_path())
     else:
-        entries, load_errors, _ = books_loader.load_fy_range(args.start_date, args.stop_date)
+        entries, load_errors, options = books_loader.load_fy_range(args.start_date, args.stop_date)
     for error in load_errors:
         bc_printer.print_error(error, file=stderr)
         returncode |= ReturnFlag.LOAD_ERRORS
 
-    postings = data.Posting.from_entries(entries)
+    data.Account.load_from_books(entries, options)
+    accounts: Set[data.Account] = set()
+    sheet_names: Dict[str, None] = collections.OrderedDict()
+    for acct_arg in args.accounts:
+        for account in data.Account.iter_accounts(acct_arg):
+            accounts.add(account)
+            if not account.is_under(*sheet_names):
+                new_sheet = account.is_under(*LedgerODS.ACCOUNT_COLUMNS)
+                assert new_sheet is not None
+                sheet_names[new_sheet] = None
+
+    postings = (post for post in data.Posting.from_entries(entries)
+                if post.account in accounts)
     for search_term in args.search_terms:
         postings = search_term.filter_postings(postings)
 
@@ -500,7 +514,7 @@ def main(arglist: Optional[Sequence[str]]=None,
     report = LedgerODS(
         args.start_date,
         args.stop_date,
-        args.sheet_names,
+        list(sheet_names),
         rt_wrapper,
         args.sheet_size,
     )
diff --git a/tests/books/ledger.beancount b/tests/books/ledger.beancount
index a941379..eba4c54 100644
--- a/tests/books/ledger.beancount
+++ b/tests/books/ledger.beancount
@@ -1,10 +1,16 @@
 2018-01-01 open Equity:OpeningBalance
 2018-01-01 open Assets:Checking
+  classification: "Cash"
 2018-01-01 open Assets:Receivable:Accounts
+  classification: "Accounts receivable"
 2018-01-01 open Expenses:Other
+  classification: "Other expenses"
 2018-01-01 open Income:Other
+  classification: "Other income"
 2018-01-01 open Liabilities:CreditCard
+  classification: "Accounts payable"
 2018-01-01 open Liabilities:Payable:Accounts
+  classification: "Accounts payable"
 
 2018-02-28 * "Opening balance"
   Equity:OpeningBalance  -10,000 USD
diff --git a/tests/test_reports_ledger.py b/tests/test_reports_ledger.py
index 1613424..34559da 100644
--- a/tests/test_reports_ledger.py
+++ b/tests/test_reports_ledger.py
@@ -15,6 +15,7 @@
 # along with this program.  If not, see <https://www.gnu.org/licenses/>.
 
 import collections
+import contextlib
 import copy
 import datetime
 import io
@@ -33,6 +34,8 @@ from conservancy_beancount import data
 from conservancy_beancount.reports import core
 from conservancy_beancount.reports import ledger
 
+clean_account_meta = contextlib.contextmanager(testutil.clean_account_meta)
+
 Acct = data.Account
 
 _ledger_load = bc_loader.load_file(testutil.test_path('books/ledger.beancount'))
@@ -43,19 +46,11 @@ DEFAULT_REPORT_SHEETS = [
     'Equity',
     'Assets:Receivable',
     'Liabilities:Payable',
-    'Assets:PayPal',
     'Assets',
     'Liabilities',
 ]
-PROJECT_REPORT_SHEETS = [
-    'Balance',
-    'Income',
-    'Expenses',
-    'Assets:Receivable',
-    'Assets:Prepaid',
-    'Liabilities:UnearnedIncome',
-    'Liabilities:Payable',
-]
+PROJECT_REPORT_SHEETS = DEFAULT_REPORT_SHEETS[:6]
+del PROJECT_REPORT_SHEETS[3]
 OVERSIZE_RE = re.compile(
     r'^([A-Za-z0-9:]+) has ([0-9,]+) rows, over size ([0-9,]+)$'
 )
@@ -275,7 +270,8 @@ def run_main(arglist, config=None):
     arglist.insert(0, '--output-file=-')
     output = io.BytesIO()
     errors = io.StringIO()
-    retcode = ledger.main(arglist, output, errors, config)
+    with clean_account_meta():
+        retcode = ledger.main(arglist, output, errors, config)
     output.seek(0)
     return retcode, output, errors
 
@@ -292,6 +288,30 @@ def test_main(ledger_entries):
     for _, expected in ExpectedPostings.group_by_account(postings):
         expected.check_report(ods, START_DATE, STOP_DATE)
 
+@pytest.mark.parametrize('acct_arg', [
+    'Liabilities',
+    'Accounts payable',
+])
+def test_main_account_limit(ledger_entries, acct_arg):
+    retcode, output, errors = run_main([
+        '-a', acct_arg,
+        '-b', START_DATE.isoformat(),
+        '-e', STOP_DATE.isoformat(),
+    ])
+    assert not errors.getvalue()
+    assert retcode == 0
+    ods = odf.opendocument.load(output)
+    assert get_sheet_names(ods) == ['Balance', 'Liabilities']
+    postings = data.Posting.from_entries(ledger_entries)
+    for account, expected in ExpectedPostings.group_by_account(postings):
+        should_find = account.startswith('Liabilities')
+        try:
+            expected.check_report(ods, START_DATE, STOP_DATE)
+        except NotFound:
+            assert not should_find
+        else:
+            assert should_find
+
 @pytest.mark.parametrize('project,start_date,stop_date', [
     ('eighteen', START_DATE, MID_DATE.replace(day=30)),
     ('nineteen', MID_DATE, STOP_DATE),
diff --git a/tests/testutil.py b/tests/testutil.py
index f46b1fc..9534bff 100644
--- a/tests/testutil.py
+++ b/tests/testutil.py
@@ -44,9 +44,11 @@ TESTS_DIR = Path(__file__).parent
 # it with different scopes. Typical usage looks like:
 #   clean_account_meta = pytest.fixture([options])(testutil.clean_account_meta)
 def clean_account_meta():
-    yield
-    data.Account.load_options_map(bc_options.OPTIONS_DEFAULTS)
-    data.Account._meta_map.clear()
+    try:
+        yield
+    finally:
+        data.Account.load_options_map(bc_options.OPTIONS_DEFAULTS)
+        data.Account._meta_map.clear()
 
 def _ods_cell_value_type(cell):
     assert cell.tagName == 'table:table-cell'