The present study was conducted during three consecutive winter seasons (2015 to 2017). Assessment of the variability and genetic associations within a population of the local cauliflower cv. Soultani along with the response to phenotypic positive mass-selection for days to visible curd (DVC) and curd diameter (CD) were studied. Further, field performance of concurrently evaluated base (C0) and both first (C1) and second (C2) cycle populations was conducted. Data analysis showed a wide range of variation in the primary (DVC and CD) and secondary (plant length PL, number of leaves NL, curd weight CW and total plant weight TPW). All traits had low negative correlation coefficients (rg) with DVC while they showed high positive values with CD. The response (R) of DVC to selection, as percent of mean, was low due to the existence of a predominant non-additive gene action. However, CD exhibited a moderate R to selection in C1 that declined in C2. The highest correlated response (CR) was for the CW in C1 but no appreciable R occurred in C2. Estimated indirect selection efficiency for all secondary traits was less than one. Concurrent field evaluation showed significant differences among the three populations (C0, C1 and C2). Relative to C0, the C2 population had 57% higher total yield and 114% early yield. The DVC showed a reduction of 11.7% and CD elevated 47.6%. We concluded that directive positive mass-selection can practically be efficient in improving our cauliflower landraces.