The recall of 90 magazine advertisements, consisting of a Brand, a Picture and a Slogan, was tested by presenting one of these components as cue and asking for the other two. The advertisements were tested in groups of ten, each group requiring a sequence of 30 test trials (10 advertisements × 3 cues). Backwards analysis of the sequences of responses-that is, starting with the last trial and comparing it successively with the responses on earlier trials-identified many responses as repetitions of previous errors, replicating many of the findings in Laming, D. On the recall of errors in recall. JSMC Brain Science, 2019, 3: 21, https://www.researchgate.net/publication/333448328_On_The_Recall_Of_Errors_In_Recall . Nine different groups of advertisements were tested after various lapses of time up to 4 months. A comparison is made between the (short term) probability of repeating a previous error as a function of lag within a single test sequence and the (longer term) probability of retrieving an advertisement as a function of elapsed time. Both these empirical relationships can be characterised by reciprocal functions, but they are otherwise quite different. Extrapolation of long-term accessibility to short intervals suggests recall in the short term could be much better than it is; extrapolation of short-term accessibility to longer intervals poses the contrary problem. Even though the only relevant variable in this comparison appears to be lapse of time, there is a question whether this comparison between short and long term is truly like with like.