Objective: In the vertebrate olfactory bulb, axonless granule cells (GC) mediate self- and lateral inhibitory interactions between mitral/tufted cells via reciprocal dendrodendritic synapses. Synaptic GC output occurs on both fast and slow time scales, allowing for multiple GC functions during olfactory processing. Here we describe a possible mechanism for delayed GC output. Methods: We use current-clamp recordings and two-photon imaging in acute brain slices from adult mice (PND 36-45; dye: 100 mM OGB-1; room temperature) and sensory-like stimulation, i.e. extracellular activation of a glomerulus. We also studied responses in GCs where either GluRN1 or GluRA2 were deleted via viral transfection. Results: We observed 15 evoked responses in 12 spines with an average amplitude of 53 ± 43 %DF/F. The average release probability was estimated as p [asymp]0.11 ± 0.06. While the onset of the evoked synaptic responses was tightly correlated to the stimulus, in most cases the peak amplitude occurred several 100 ms later (average time to peak ttp = 450 ± 500 ms, n = 11; 2 events with peak past the duration of the scan, i.e. 1000 ms). This slow rise was independent of Ca²⁺ entry via NMDARs, since similar signals occurred in GluN1 knockout GCs with reduced Ca²⁺ entry (34 ± 10%, n = 12 spines; average DF/F ttp = 550 ± 300 ms, n = 13 events; 4 more with peak 1000 ms). Additional Ca²⁺ entry due to deletion of GluA2, however, resulted in larger DF/Fs that rose even more slowly (73 ± 17%, n = 8 spines; DF/F ttp = 700 ± 250 ms, n = 7 events; 13 more with peak 1000 ms; P = 0.03 vs WT for all responses, Mann-Whitney test). Conclusions: The slowly evolving events constituted the majority of synaptic responses and may allow for to GC outputs with long latency. Since GluA2^DGCL animals performed significantly faster in a difficult odor discrimination task than WT, the observed increase in rise times does not interfere with fast odor discrimination.