Pack and Herman (1995) revealed that in ``cross-modal'' matching tests, a dolphin could spontaneously recognize the shape of an object through one sense (either echolocation or vision) that it had earlier interrogated through the alternate sense [A. A. Pack and L. M. Herman, J. Acoust. Soc. Am. 98, 722--733 (1995)]. Objects exposed to the echolocation sense were presented underwater in an anechoic box fronted by opaque Plexiglas. Objects exposed to the visual sense were presented in air ``live'' or on a television monitor. Eight unique pairs of 16 shapes were tested. Recognition was immediate across these pairs both going from echolocation to vision (E--V matching) and from vision to echolocation (V--E matching)---no learning of shape was required. The present study re-paired subsets of these shapes, thereby contrasting many more features. Twenty-five completely new pairings were tested. Performance on the first two unique trials with each pair provided the measure of immediacy of recognition. The dolphin chose correctly on 49 of 50 of these trials under E--V matching and on 44 of 50 during V--E matching. Together, the results of both studies strongly support the hypothesis that the dolphin's echoic perceptual system can construct rich spatial representations of objects.