The conversion of silent synapses into active sites is hypothesized to be a primary mechanism underlying learning and memory processes. Here, we used an in vitro model of classical conditioning from turtles that demonstrates a neural correlate of eyeblink conditioning to examine whether the conversion of silent synapses has a role in this form of associative learning. This was accomplished by direct visualization of AMPAR and NMDAR subunits colocalized with synaptophysin (Syn) using immunofluorescence and confocal microscopy. In naive preparations, there was a relatively high level of synapses immunopositive for NR1-Syn alone interpretated to be silent synapses. After early stages of conditioning during acquisition of conditioned responses (CRs), there was a significant increase in the colocalization of GluR1-3 AMPAR subunits at NR1-immunopositive synaptic sites. Later in conditioning, levels of GluR1-3 declined and enhanced colocalization of GluR4-containing AMPAR subunits at synapses was observed. The trafficking of these subunits during conditioning was NMDAR-mediated and was accompanied by protein synthesis of GluR4 subunits. Examination of the PSD fraction confirmed the early and late synaptic insertion of GluR1-3 and GluR4, respectively, during conditioning. These findings suggest that there is differential trafficking of synaptic AMPARs during classical conditioning. Existing GluR1-3 AMPAR subunits are initially delivered to silent synapses early in conditioning to unsilence them followed by synthesis and insertion of GluR4 AMPAR subunits that are required for acquisition and expression of CRs.