Purpose To investigate whether the presence of large fragment (LF) and abnormal cell divisions (ACDs) has influenced the correlation between live birth rate and number of blastomeres detected on day 2 by conventional scoring. Methods This study included 578 embryos cultured in time lapse and selected for transfer by conventional scoring on day 2. By time-lapse recordings, embryos were reassessed to identify ACDs and/or LFs mistaken as blastomeres. The latter identifications were used to recalculate fragmentation rate and the number of blastomeres. Life birth rate according to number of blastomeres was compared in (a) embryos selected by conventional scoring and (b) embryos reassessed by time lapse. Results After conventional scoring, embryos with four cells had a significantly higher pregnancy rate than embryos with less than four cells and embryos with more than four cells. By time-lapse assessment, ACDs and/or recalculated fragmentation >25 % was recognized in 106/578 (18.3 %) of transferred embryos. None of them resulted in a live birth. After exclusion of these embryos, the number of blastomeres on the day of transfer did not have any impact on life birth rate. Conclusion Conventional scoring on day 2 did not detect ACDs and LFs mistaken as blastomeres. LFs can lead to a recalculated fragmentation rate to >25 %. No significant correlation between live birth rate and number of blastomeres in day 2 embryos was observed when embryos with ACDs and fragmentation >25 % were excluded. Recognition of ACDs and fragmentation >25 % is more predictive of live birth than number of blastomeres.