The standard strategy for analysis by tandem mass spectrometry of protein phosphorylation at serine or threonine utilizes the neutral loss of H3PO4 (= 97.977/z) from proteolytic peptide molecular ions as marker fragmentation. Manual control of automatically performed neutral loss-based phosphopeptide identifications is strongly recommended, since these data may contain false-positive results. These are connected to the experimental neutral loss m/z error, to competing peptide fragmentation pathways, to limitations in data interpretation software, and to the general growth of protein sequence databases. The fragmentation-related limitations of the neutral loss approach cover (i) the occurrence of abundant 'close-to-98/z' neutral loss fragmentations, (ii) the erroneous assignment of a neutral loss other than loss of H3PO4 due to charge state mix-up, and (iii) the accidental occurrence of any fragment ion in the m/z windows of interest in combination with a charge-state mix-up. The 'close-to-98/z' losses comprise loss of proline (97.053/z), valine (99.068/z), threonine (101.048/z), or cysteine (103.009/z) preferably from peptides with N-terminal sequences PP, VP, TP, or CP, and loss of 105.025/z from alkylated methionine. Confusion with other neutral losses may occur, when their m/z window coincides with a 98/z window as result of a charge state mix-up. Neutral loss of sulfenic acid from oxidized methionine originating from a doubly charged precursor (63.998/2 = 31.999) may thus mimic the loss of phosphoric acid from a triply charged phosphopeptide (97.977/3 = 32.659). As a consequence of the large complexity of proteomes, peptide sequence ions may occur in one of the mass windows of H3PO4 loss around 97.977/z. Practical examples for false-positive annotations of phosphopeptides are given for the first two groups of error. The majority of these can be readily recognized using the guidelines presented in this study.