Cognition　of　soundscape　information　affects　an　individual＇s　preference　and　understanding　soundscapes　in　urban　parks.　This　study　explored　the　cognitive　persistence　of　soundscape　and　the　factors　that　influence　cognitive　soundscape　in　urban　parks.　Landscape　information　was　collected　at　36　observation　sites　in　three　parks.　The　duration　of　soundscape　samples　at　each　site　heard　by　the　observers　were　divided　into　10　s,　20　s,　and　30　s.　The　observers　in　different　groups　were　required　to　start　filling　out　the　questionnaire　following　the　different　time.　Principle　Component　Analysis　(PCA)　produced　three　principal　components　of　cognitive　soundscape,　loudness　cognition,　harmonious　cognition,　and　directional　cognition.　Three　hierarchical　Cluster　Analyses　contributed　to　finding　the　variation　characteristics　of　the　factors　from　PCA　in　different　intervals.　A　model　using　Artificial　Neural　Network　(ANN)　provided　a　prediction　of　cognitive　soundscape　attenuation.　Results　indicate　that　the　optimal　cognitive　persistence　of　soundscape　is　20　s　to　maintain　the　cognitive　soundscape,　and　the　influencing　factors　include　soundscape　and　lightscape.　The　classified　indicators　from　PCA　were　more　effective　in　forecasting　than　those　that　were　unclassified　after　ANN　training.　Overall,　the　soundscape,　lightscape,　and　space　together　play　a　key　role　in　the　cognitive　persistence　of　soundscape.　©　2019　Elsevier　Ltd