Findings from TIMSS-2003 in Israel reveal a significant increase in the national average score in mathematics and science compared to that in the previous TIMSS study in 1999. This increase occurred in all three subpopulations of the sample: Hebrew-speaking students in secular state schools (47%) and in religious state schools (25%) and Arabic-speaking students in state schools (27%). However, the gains were much more profound in the Arabic-speaking sector. These uneven gains resulted in a significant decrease in the long-standing huge achievement gaps between the two ethnic sectors (Bashi, Chan, Davise, 1981; Ariav, Kafir, Ben-Simon, 1996; Kafir, Ariav, & Ben-Simon, 1999) and causing a decline from about a full standard deviation in both school subjects in 1999 to about half a standard deviation in mathematics and 0.4 in science in 2003. Closing achievement gaps between subpopulations in Israel, and amongst them between students in the Hebrew- speaking schools and Arabic-speaking schools, continues to be one of the most prioritized goals of the Israeli educational system. TIMSS-2003 findings provide first evidence that efforts made during the 1990s to close these gaps were in the right direction and, although inequality in input still remains, gaps in learning outcomes have narrowed. This paper intends to highlight factors that still perpetuate the achievement gap on the one hand and on the other, to detect factors that explain why these gaps have narrowed. School inputs such as favorable students' body composition, school resources and learning climate positively associated with achievement in the two sectors, but less frequent in the Arabic-speaking sector, explain the perpetuation of the achievement gap. However, on the other side, two mechanisms can explain the narrowing of the achievement gaps between the two sectors. The first is the interaction effect found between most variables describing instruction and students' ethnic affiliation. These variables seem to be more effective in the Arabic-speaking sector resulting in relatively higher achievement gains in this sector. The second mechanism, known as the Heynemann & Loxley effect, is based on the relative higher effects of school and class variables versus home and family variables that were found in less economically developed countries (Fuller, 1987; Heynemann & Loxley, 1983; Meyer & Baker, 1996; Kastilis & Robinson, 1990). Regarding the Arabic- speaking sector as less economically developed and conducting separate regression analyses in Israel's two ethnic sectors indeed reveals higher school and schooling effect in the Arabic-speaking sector, which explains the narrowing achievement gap.