Traffic　speed　variance　is　defined　as　a　measure　of　the　dispersion　of　space　mean　speeds　among　drivers.　Empirical　speed-density　observations　exhibit　a　structured　traffic　speed　variance,　which　has　been　found　to　be　associated　to　the　roadway　crash　rate,　the　fatality　rate,　and　travel　time　variability.　The　objective　of　this　paper　is　to　propose　a　generalized　traffic　speed　variance　function　to　describe　this　structured　variance.　The　proposed　speed　variance　function　is　a　response　of　the　speed-density　curve　with　two　additional　parameters.　The　estimation　of　the　model　parameters　in　the　proposed　traffic　speed　variance　function　can　be　carried　out　through　an　iterative　nonlinear　least-square　algorithm　(i.e.,　LevenbergMarquardt).　A　series　of　logistic　speed-density　curve　with　varying　parameters　are　used　in　the　proposed　traffic　speed　variance　function　with　different　levels　of　performance.　The　proposed　traffic　speed　variance　model　can　potentially　help　to　unveil　the　underlying　mechanism　of　empirical　traffic　phenomenon　such　as　spontaneous　congestion　or　capacity　reduction.　Copyright　(c)　2013　John　Wiley　&　Sons,　Ltd.