Shear　modulus　is　an　essential　parameter　for　deformation　prediction　and　site　response　analysis　of　geological　problems.　This　paper　considers　the　dynamic　shear　modulus　degradation　of　saturated　marine　coral　sand　containing　fine　particles　from　a　reef　in　the　Nansha　Islands.　Resonant-column　tests　were　carried　out　on　specimens　of　the　saturated　coral　sand　with　varying　non-plastic　fines　content　(FC)　and　relative　density　(Dr)　subjected　to　initial　effective　confining　pressure　(sigma　＇　m).　The　test　results　show　that　the　stress　exponent　n-which　reflects　the　rate　of　increase　of　the　maximum　dynamic　shear　modulus　(Gmax)　with　increasing　sigma　＇　m-is　a　soil-specific　constant.　A　significant　finding　is　that　the　equivalent　skeleton　void　ratio　e*sk　has　a　unique　power　relationship　with　the　normalised　maximum　dynamic　shear　modulus　[Gmax/(sigma　＇　m/Pa)n].　An　improved　Hardin　model　based　on　binary　packing　theory　is　proposed　to　evaluate　Gmax　of　the　coral　sand,　and　its　applicability　is　verified　using　experimental　data　for　three　different　sands　from　the　literature.　For　a　given　shear　strain,　dynamic　shear　modulus　G　decreases　with　increasing　FC　and　increases　with　increasing　Dr　and　sigma　＇　m.　The　curves　of　the　maximum　dynamic　shear　modulus　rate　(G/Gmax)　tend　to　fall　with　increasing　FC,　while　Dr　and　sigma　＇　m　have　little　influence　on　G/Gmax.　The　fitting　parameters　A　and　B　of　the　Davidenkov　model　are　insensitive　to　sigma　＇　m,　Dr,　and　FC　and　are　fixed　at　1.08　and　0.42,　respectively,　for　the　saturated　coral　sand.　Furthermore,　the　reference　shear　strain　gamma　0　in　the　Davidenkov　model-corresponding　to　the　shear　strain　at　G/Gmax　=　0.5-can　be　evaluated　uniformly　by　e*sk　regardless　of　sigma　＇　m,　Dr,　and　FC.　A　unified　form　of　shear-strain-dependent　G　prediction　method　is　established　and　provides　a　significant　advantage　in　evaluating　G　for　marine　coral　sand　in　practice.