Reading is crucial for functioning in our modern societies, and reading failure is associated with reduced happiness and success. Indeed, learning-to-read an alphabetic script requires the explicit teaching of the correspondences between letters and sounds (phonics). Memorizing thousands of visual words by heart is no alternative as it would be like memorizing a telephone directory. Although most teachers indicate that they systematically teach letter-sound correspondences in grade 1, there is evidence that many of them complement explicit letter-sound teaching (strict phonics) with various whole word recognition strategies (mixed phonics). The goal of the present study was to nd out whether strict phonics methods produce better results than less strict or mixed phonics methods. To do so, the responses of 9,340 grade 1 teachers concerning their teaching practice and the textbook they used were matched to the performance of their 139,288 students assessed in mid-grade 1 and early-grade 2 through the national evaluations in France. The results of hierarchical mixed effects modelling showed a clear advantage of strict phonics textbooks and strict phonics teaching methods over mixed, strongly mixed, or very strongly mixed (non-systematic) phonics methods while controlling for pre-reading skills, class reading level, socio-economic status, and teacher experience. The advantage of strict phonics methods over all other methods was even stronger for pupils with initially weak pre-reading skills and students from socio-economically disadvantaged neighborhoods. These results have far-reaching implications for educational policy, teacher training, prevention of reading di culties, and social equality.