Gold evaporated on air-cleaved (00.1) graphite was observed by scanning tunneling microscopy. Most individual gold islands were not stable during scanning; some islands even appeared to translate. These observations suggest that the interaction between the gold islands and the scanning tunneling microscope tip is stronger than that between the gold and the air-cleaved graphite substrate. Some gold islands were observed to be stable. It was deduced that stable islands at steps were located on the upper layer rather than on the lower layer. It is also deduced that the alignment of the elongated gold grains in the continuous film may be related to cleavage steps.