Conventional 2-D tactical displays depict military objects (platforms) using non-realistic symbols. A new generation of 3-D tactical displays is being developed that depict platforms using miniature realistic icons of ships and planes. These icons are very compelling, but do they translate into performance benefits for users? In four experiments we measured identification and classification performance for a set of common military platforms displayed as either 3-D icons, 2-D icons, or conventional 2-D symbols (from Mil. Std 2525B). 3-D icons were identified consistently slower and less accurately than 2-D symbols, and no better than computationally simpler 2-D icons. Icons are poor because they put the work of discriminating between subtly different real-world platforms onto the user. Symbols, on the other hand, can be engineered to have arbitrarily high discriminability for similar-looking real-world platforms. Users may appreciate, and clamor for, a familiar faithfully realistic representation of the world on their display, but they may be much better off operating on a pre-processed discriminable, if unrealistic, version of it.