The aim of the current study was to explore the integration processes of gender and number morphological features, since it has been proposed that grammatical gender and number features might be associated with different strength with the word stem in lexical representation. Event related potentials (ERPs) were recorded using a 128-channel sensor net while twenty-four volunteers read Spanish word pairs and performed a syntactic judgment task. The word pairs which could agree or disagree in gender or number or in gender and number at the same time, were formed by a noun and an adjective (e.g. faro-alto [lighthouse-high]). A negativity around 400 msec with posterior distribution, which has been related to lexical integration processes, was found in response to both gender and number violations. No differences were found between gender disagreement, number disagreement and the double disagreement. Therefore, ERPs suggest that integration of gender and number features may not be different, and that the detection of disagreement may work under a binary state, since the double disagreement condition did not differ from the others. In addition, a subsequent component (identified as P3) showed delayed latencies in the gender disagreement condition as compared to the number disagreement condition, while the double disagreement conditions showed a shorter peak latency than the other two disagreement conditions and similar to the agreement condition. The variations in the latency of the P3 component, which has been related to categorization processes, suggest that these processes are quickly triggered by the accumulation of two incongruent as compared to one disagreement features, and that reanalysis is costlier in the case of gender disagreement as compared to the number disagreement.