diff --git a/parquet/file/record_reader.go b/parquet/file/record_reader.go
index 87ae20f3..62090443 100644
--- a/parquet/file/record_reader.go
+++ b/parquet/file/record_reader.go
@@ -522,7 +522,7 @@ func (rr *recordReader) Reset() {
 // process no more levels than necessary to delimit the indicated
 // number of logical records. updates internal state of recordreader
 // returns number of records delimited
-func (rr *recordReader) delimitRecords(numRecords int64) (recordsRead, valsToRead int64) {
+func (rr *recordReader) delimitRecords(numRecords int64) (recordsRead, valsToRead int64, err error) {
 	var (
 		curRep int16
 		curDef int16
@@ -533,6 +533,9 @@ func (rr *recordReader) delimitRecords(numRecords int64) (recordsRead, valsToRea
 
 	for rr.levelsPos < rr.levelsWritten {
 		curRep, repLevels = repLevels[0], repLevels[1:]
+		if rr.atRecStart && curRep != 0 {
+			return 0, 0, errors.New("parquet: record starts with a nonzero repetition level")
+		}
 		if curRep == 0 {
 			// if at record start, we are seeing the start of a record
 			// for the second time, such as after repeated calls to delimitrecords.
@@ -576,7 +579,10 @@ func (rr *recordReader) ReadRecordData(numRecords int64) (int64, error) {
 	)
 
 	if rr.Descriptor().MaxRepetitionLevel() > 0 {
-		recordsRead, valuesToRead = rr.delimitRecords(numRecords)
+		recordsRead, valuesToRead, err = rr.delimitRecords(numRecords)
+		if err != nil {
+			return 0, err
+		}
 	} else if rr.Descriptor().MaxDefinitionLevel() > 0 {
 		// no repetition levels, skip delimiting logic. each level
 		// represents null or not null entry
diff --git a/parquet/file/record_reader_internal_test.go b/parquet/file/record_reader_internal_test.go
new file mode 100644
index 00000000..e9285edd
--- /dev/null
+++ b/parquet/file/record_reader_internal_test.go
@@ -0,0 +1,46 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package file
+
+import (
+	"testing"
+
+	"github.com/apache/arrow-go/v18/arrow"
+	"github.com/apache/arrow-go/v18/arrow/memory"
+	"github.com/apache/arrow-go/v18/parquet"
+	"github.com/apache/arrow-go/v18/parquet/schema"
+	"github.com/stretchr/testify/require"
+)
+
+func TestDelimitRecordsRejectsNonzeroInitialRepetitionLevel(t *testing.T) {
+	descr := schema.NewColumn(schema.NewInt32Node("values", parquet.Repetitions.Repeated, -1), 1, 1)
+	rr := newRecordReader(descr, LevelInfo{DefLevel: 1}, memory.DefaultAllocator, nil).(*recordReader)
+	defer rr.Release()
+
+	rr.repLevels.ResizeNoShrink(arrow.Int16SizeBytes)
+	rr.defLevels.ResizeNoShrink(arrow.Int16SizeBytes)
+	rr.RepLevels()[0] = 1
+	rr.DefLevels()[0] = 1
+	rr.levelsWritten = 1
+	rr.atRecStart = true
+
+	records, values, err := rr.delimitRecords(1)
+	require.Error(t, err)
+	require.Zero(t, records)
+	require.Zero(t, values)
+	require.Zero(t, rr.levelsPos)
+}