The idea that some recently encountered items reside in a special state where they do not have to be retrieved has come to be a critical component of short-term memory theories. In the current work, the existence of such a special state was tested using the probe-recognition paradigm followed by a delayed recognition test. Across two experiments participants received a series of probe recognition trials where list lengths of 1-, 4- and 8-items were intermixed. Delayed recognition performance for non-target probes was poorer for the only item in 1-item lists than for the last item in multi-item lists. At the same time, the delayed recognition of studied-but-not probed items was better for the 1-item list, compared to the last item in a multi-item list, indicating that some form of a retrieval effect was involved and not lower levels of attention/initial learning. An examination of the size of the testing effect as it varied across list lengths and experiments also indicated that residence in a special state was not playing an important role. Overall, the data are not in support of the assumption that items at the focus of attention are in a special state that do not require retrieval. Our conclusions are that special states cannot be used to define STM memory and that the probe recognition paradigm may be useful in determining how testing affects memory.