The strengths of thermocompression bonds made between gold plated copper lead frames and gold metallized thin film circuits decrease in time when aged at 200-300°C in air or vacuum unless there is a diffusion barrier (such as nickel) between the copper and gold. After pulling bonds to destruction, failure modes and failed surfaces were characterized by scanning electron microscopy, Auger spectroscopy, electron microprobe, stylus probe, and X-ray diffraction. The failed surface on the substrate side of the bond was very rough, appearing to consist of "mountains" of almost pure Cu resting on a relatively smooth layer of grains, identified as Cu 3 Au, where the Cu 3 Au regions were Visible in the valleys between the copper mountains. The mating failed surface on the lead side of the bond was similarly rough, but with only a few scattered grains of Cu 3 Au evidently pulled out from the opposite surface. Oxidation, pore formation (Kirkendall effect), and ordered phase formation were considered as possible mechanisms for the degradation. Oxidation was ruled out by the observation that bonds degraded as fast in vacuum as in air. Ordered phase formation was judged more likely than pore formation as the primary mechanism. A parameter, \tau , to describe the time required for measurable degradation was defined as the time for the average 90° peel strength of a 0.25 X 0.75 mm lead to decrease from six or seven pounds to four pounds. This parameter was found to be temperature activated with an activation energy of 0.8 eV. Since ordering and pore formation are both dominated by interdiffusion, \tau was extrapolated to lower temperatures using the diffusion coefficient measurements of Pinnel and Bennett, resulting in a value of ten years at 50°C. However, with a 5000A ± 2500A nickel diffusion barrier between the copper lead and the gold plate, the value of \tau extrapolated to 50°C was considerably greater than 40 years, as desired in Bell System equipment.