Window　opening　plays　a　decisive　role　in　indoor　environments　in　residential　buildings.　To　date　great　efforts　have　been　conducted　to　explore　the　driving　characteristics　of　window　opening　behavior　based　on　a　group　of　samples.　In　this　context,　at　the　level　of　one　individual　residential　unit,　more　understanding　of　occupant　interactions　with　windows　is　still　required.　Based　on　oneyear　field　measurements　of　indoor　and　outdoor　environmental　parameters　and　window　status　of　three　apartments　in　Beijing　and　semi-structured　interviews　with　occupants,　this　study　provides　an　engineering　and　sociological　analysis　of　the　realistic　characteristics　of　window　opening　behavior　during　the　heating,　transition,　and　cooling　seasons　for　each　apartment.　These　single　apartmentbased　findings　show　that　window　opening　is　a　complex　occupant　behavior　under　the　combined　influence　of　numerous　environmental　and　non-environmental　factors,　which　cannot　be　exhaustively　listed.　Moreover,　the　environmental　and　non-environmental　factors　interact　with　each　other　on　their　impacts　on　window　opening　behavior,　and　the　interaction　between　the　two　kinds　of　factors　is　difficult　to　quantify.　It　is　thus　difficult　to　generalize　the　driving　characteristics　of　window　opening　even　based　on　window　state　transitions.　Also　due　to　this,　window　opening　exhibited　significant　individual　differences　among　these　apartments.　Nevertheless,　it　is　more　important　to　note　that　the　indoor　environment　of　these　apartments,　regulated　only　by　opening　the　window,　is　statistically　the　same　during　the　transition　season.　Based　on　thermal　adaptive　approach,　these　environmental　outcomes　illustrate　that　the　indoor　environment　regulated　by　window　airing　alone　can　represent　the　environments　acceptable　to　occupants　under　natural　ventilation.　Accordingly,　comprehensively　accurate　investigation　of　the　driving　characteristics　of　window　opening　behavior　would　not　be　practical.　In　this　context,　this　study　suggests　that　a　probabilistic　model　of　window　states　is　necessary　to　describe　window　opening　behavior　when　focusing　on　window　states　and　their　effects　on　indoor　environment.　As　long　as　the　simulated　indoor　environment　is　guaranteed　to　be　acceptable　to　occupants,　the　inferred　model　based　on　window　state　can　be　considered　to　be　appropriate.　More　importantly,　based　on　these　new　insights,　this　study　presents　a　new　research　perspective　compared　to　existing　studies　on　occupants＇　interaction　with　windows.　It　will　contribute　to　more　comprehensive　understanding　about　the　environments　acceptable　to　occupants　under　natural　ventilation,　and　more　accurate　predictions　of　indoor　environments　and　energy　consumptions　in　the　real　buildings,　and　thereby　enhancing　the　application　values　of　the　study　about　occupants＇　interaction　with　windows.