Neurons in cortical sensory regions receive modality-specific information through synapses that are located on their dendrites. The use of two-photon microscopy combined with whole-cell recordings has helped identify visually-evoked dendritic calcium signals in mouse visual cortical neurons in vivo. The calcium signals were restricted to small dendritic domains (‘hotspots’) and represented visual synaptic inputs that were highly-tuned for orientation and direction (Jia et al., Nature, 2010). A new variant of two-photon imaging with an improved sensitivity for fluorescence signaling and a reduced risk of phototoxicity, termed LOTOS (= low power temporal oversampling), allowed the detection of sensory-evoked calcium transients in dendritic spines of mouse neurons in the auditory (Chen et al., Nature, 2011). Individual spines on the same dendrite were functionally heterogeneous. Even neighboring spines were often tuned to different sound frequencies. In line with these observations, we found in the vibrissal cortex that stimulation of different whiskers activated distinct spines on the same dendrite (Varga et al., PNAS, 2012). Some spines were activated uniquely by single whiskers, but many spines were activated by multiple whiskers. These shared spines indicate the existence of presynaptic ‘feeder’ neurons that integrate and transmit activity arising from multiple whiskers. Together, our results suggest that in different cortices (visual, auditory, vibrissal), afferent sensory inputs to layer 2/3 neurons are widely distributed throughout the entire dendritic tree in a ‘salt-and-pepper’-like manner. In contrast to previous suggestions stressing the role converging inputs generating dendritic spikes, we find that the ‘computational units’ in layer 2/3 neurons in vivo are highly tuned individual spines.