Objectives: To evaluate the clinical severity and proportionality of small for gestational age, very low birth weight neonates (< 1,500 g) and to estimate the neonatal mortality risk associated with the condition of being small for gestational age according to the degree of severity and proportionality. Methods: Observational design. Ail of the NEOCOSUR Collaborative Group's very low birth weight infants (25-36 weeks' gestation) were included (n = 1,518). Anthropometric indices: birth weight < 3rd and 10th percentile. Severity (fetal growth ratio = observed weight/mean birth weight for gestational age); no growth restriction: fetal growth ratio 0.90-1.10, mild: fetal growth ratio 0.80-0.89, moderate: fetal growth ratio 0.75-0.79 and severe: fetal growth ratio < 0.75. Proportionality: coefficient of bimodality and z score for ponderal index (PI = g/cm3 × 100). Neonatal mortality until discharge. Results: < 3rd percentile: 13.5% (p < 0.001); < 10th percentile: 31% (p < 0.001); fetal growth ratio: 0.90±0.21 (p < 0.001), mild restriction: 20.8%, moderate restriction: 8.7% and severe restriction: 32.6%. Coefficient of bimodality: 0.53; PI z score < -1: 8%. Maternal hypertensive disease was systematically associated with being small for gestational age (aOR 1.20, 95% CI 0.86-1.67), fetal growth ratio ≤. 0.89 (aOR 1.71, 1.24-2.36) and PI z score < -1 (aOR 1.60, 1.03-2.41). Adjusted odds ratios for neonatal mortality were: 2.64 (95% CI 1.71-3.92) for being small for gestational age, 2.76 (95% CI 1.85-4.10) for fetal growth ratio ≤ 0.89, and 1.37 (95% CI 0.80-2.32) for z score PI < -1. Conclusions: Small for gestationai age, mostly symmetric and severe restriction is a frequent condition in < 1,500 g neonates and is associated with higher mortality rates. Copyright © 2005 by Sociedade Brasileira de Pediatria.