We often remember information without its source (e.g. word or picture format). This phenomenon was termed source amnesia, which has been studied extensively in long-term memory, but rarely in the context of short-term/working memory. It is an open question as to whether source amnesia is always the result of forgetting over a prolonged period of time. This study attempted to address this issue by directly testing the possibility whether source amnesia could happen in the context of working memory by using surprise tests developed in the fields of inattentional blindness, change blindness, and attribute amnesia. Specifically, participants repeatedly judged the congruency between two color representations from one single object (i.e., ink color and identity of a color word) or two distinct objects (i.e., color of a square and identity of a color word), and then were unexpectedly asked to report the source of one color representation. The results consistently showed participants’ inability to report the source of a color representation, even though the color had just been attended and/or encoded into working memory. To the best of our knowledge, this is the first study showing source amnesia in a context of working memory with normal young participants, despite it is typically assumed only existing in long-term memory. These counterintuitive findings suggest a cognitive-economy view of memory function, in which source information is only stored when it is expected to be useful.