Reading is a complex activity that requires activation of many cognitive-linguistic processes, processes that are interrelated and among whom, we highlight as the subject of study, recognition of words. The variety of definitions contributes to access to a wide range of concepts about what is word recognition. Being the most general the conversion of the graphemes in phonological patterns, identifying through a process of decoding the written word (Morais, 1997). The recognition of words as part of the reading process gains importance in currents researches. Nonetheless, it's a fact that there aren't, nationally, valid and standardized assessment instruments. In this way, is part of this dissertation aims to describe the construction and validation of the Word Recognition Inventory - WRI. The test consists of reading aloud words that vary in the regularity, frequency and extent. It is a computerized test using the program Superlab 4.0 which presents the 40 items on the screen. During application the answer is recorded as right or wrong. The test was applied individually to 146 children in second, third and fourth year of the first cycle, studying in a group of schools in the north of the country. Were carried out, statistical analysis revealed a level of difficulty of the items relatively high, giving evidence of the test easiness. The problem items were excluded, in the following analysis were revealed differences statistically significant between the years of schooling. We also found moderate correlations with the reading assessment conducted by teachers and the results of the PRP. Considering the importance of the WRI presented in the analysis of the psychometric characteristics of spoken words, it is suggested to continue the research on overcoming the limitations alluded to.