In the process of reading compound words, those with high-frequency second constituents are recognized faster than the ones with low-frequency second constituents. However, the role of the first constituent still remains unclear. In the present study, the time course of the frequency effects for both constituents was assessed using Basque compound words embedded in sentences while electrophysiological measures (ERPs) were recorded (Basque is a language with a high frequency of compound words, both right and left-headed). Subjects responded to comprehension questions that were not focused on the compound words. The results revealed that high-frequency first constituents elicited larger negativities starting very early (100–300 ms time window), while low-frequency second constituents elicited larger N400 amplitudes than high-frequency second constituents. Following an activation–verification framework, we argue that the early negativity difference reflects candidate triggering, whereas the N400 difference for the second constituent reflects the cost of its selection and integration for the whole-word meaning to be accessed.