The tumor stage is the most powerful prognostic tool for predicting the survival rates of lung carcinoma patients. However, prognosis of individual patients is difficult in part because of the marked clinical heterogeneity among such patients. Galectins are involved in cell growth, apoptosis and cell migration features, and their diagnostic and prognostic values have already been demonstrated in various types of cancers. In the present paper we analyze the potential prognostic value of immunohistochemical galectin-3 expression in lung adenocarcinomas and squamous cell carcinomas. In all, 165 squamous cell carcinomas and 121 adenocarcinomas were immunostained for galectin-3. In each case the immunohistochemical analyses consisted of an evaluation of the percentage of tumor cells stained and the intensity of staining. An IP score (ie Intensity x Percentage) was thus determined for each lung carcinoma. A large majority of cases displayed galectin-3 expression. While the cytoplasmic staining in the squamous cell carcinomas was focal and moderately intense, the staining in the adenocarcinomas was diffuse and intense. The IP scores were significantly (P=0.0001) higher in the adenocarcinomas than in the squamous cell carcinomas. The difference in nuclear expression profiles between the two cancer types was statistically significant (P=0.0005). Cox multivariate analysis carried out on the patients' genders, the TNM classification and the galectin-3-related variables showed that of the galectin-3-related variables, only the nuclear location of galectin-3 was identified as a prognostic indicator of recurrence independent of the clinicopathological features characterizing the patients (P=0.02). The prognostic contribution of this latter variable was enhanced when the patients with relapse-free follow-ups longer than 8 months were considered (P=0.005). Galectin-3 immunohistochemical expression differs between squamous cell carcinomas and adenocarcinomas, but the nuclear expression of galectin-3 behaves as a significant prognostic predictor for all the cases as a group.