This paper focuses on a well-known pattern of systematic syncretism in Spanish 'se' constructions. Detailed syntactic and semantic analyses are provided with the aim of sustaining two main theses. First, I conceive of 'se' as a probe for A-movement. This probe is merged with Voice in order to satisfy a subcategorization restriction. Yet, being defective, it cannot receive a 'θ'-role from Voice. As a probe it looks for a goal in its complement domain. If there is such a goal, then it A-moves to Spec,VoiceP, position in which it agrees with 'se' and receives an additional agent 'θ'-role from Voice, if there is one. This results in most, if not all, instances of the so-called “paradigmatic” 'se' ('se' reflexives, inherent 'se', benefactive 'se' and so on). There are cases in which there is no such a goal for 'se'. In those scenarios, 'Agree' fails and the clitic receives third person singular by default. This results in the so-called “non-paradigmatic” 'se' (essentially, passive/impersonal 'se'). Second, at LF, these two syntactic scenarios feed two different LF realizations. Whenever 'se' has a goal with which it agrees, 'se' itself is realized as a 'λ'-abstractor, but as an indefinite variable whenever 'Agree' fails, as in the case of passive/impersonal 'se'. This theory dispenses, then, with particular Voice features (e.g., Active vs. Non-active) and with different types of 'se' (paradigmatic vs. non-paradigmatic) but, more importantly, it does so by appealing to well-motivated restrictions on A-dependencies, namely, Activity and Miminality.