Our goal consists in establishing the analytical nature of French temporal sequences sharing the morphosyntactic pattern [Preposition – Definite article – Noun - Preposition]. These sequences, frequently classified as “prepositional locutions” in the French linguistic literature, appear to be regular prepositional phrases comprising a predicative nominal nucleus which represents an interval of a temporal event. Thus, the absolute use of the PP, due to the omission of the complement representing the temporal whole, can no longer be unconditionally attributed to ellipsis, but should rather be ascribed to associative anaphora. Are only stage level predicates (SLPs) acceptable as secondary predicates (SPs), and are all individual-level predicates ruled out in this position ? The paper shows that the first generalisation is wrong and argues in favour of a distinction between two types of SLPs. The first kind (e.g.,“malade” ’sick’, acceptable as SPs) denotes a “pure state”, while the second one (e.g., “bavard” ’talkative’, unacceptable as SPs) denotes an “action-dependent state”. These two kinds of SLPs are defined through properties independent of the SP construction, and arguments are offered against the assimilation of “bavard”-SLPs to agentive predicates.