I too believe the "correct" answer would be classification.
I also agree with @David Ruess that the question is less then ideally worded. For starters, traffic "flows" are usually not identified/classified, usually packets (if L3) are. This because within a single traffic flow, a flow's packets might be classified differently. (Usually for drop treatment, not prioritization.)
Second, DSCP/IPPrec/MPLS-EXP markings might not be used for priority at all, or priority is but one attribute of such classified traffic treatment.
Marking, or tagging, as David describes is setting/changing a specific field set aside for optimal identification/classification purpose. Basically, a tag or marking allows for a very efficient identification (short cut) of what treatment "class" a packet (or frame, etc.) belongs too.
For example, we what to identify (classify) VoIP bearer packets so they can be provided minimal latency, minimal drops and minimal jitter at each transit egress interface. So, what must you do to identify such traffic? It can get rather involved, and (relatively speaking) time consuming to analyze each packet is a VoIP packet, or not, especially if further restricted to source and destination hosts/networks too. Each egress interfaces needs correct classification to provide necessary egress treatment.
Suppose, though, a trusted host transmits its packets, for VoIP bearer packets, with the DSCP marking EF. Each hop still needs to classify this transit traffic, but now it only needs to check the L3 ToS field for a DSCP value of EF.
Classification might not use frame markings/tags at all, use them exclusively, or in conjunction with other packet attributes.
A transit hop can, optionally, based on classification/identification set/reset a marking/tag, but that's falls under treatment.
