The spatiotemporal organization of neuronal firing is crucial for information processing, but how thousands of synaptic inputs to the dendritic spines drive the firing remains a central question in neuroscience. Although a change in synaptic density and strength have been a responsible factor for the pathophysiology of various psychiatric disorders, it is entirely unclear what changes in neuronal computation and subsequent dynamism of neuronal circuits would be evoked when such change occurs. To address this question, we performed multi-scale synapse analyses, in which dendrite and somatic events are simultaneously assessed during precise stimulation of identified spines by two-photon glutamate uncaging. Interestingly, mice with knockdown of SETD1A and DISC1, both well-established animal models for schizophrenia, exhibited a significant number of extra-large (XL) spines (corresponding to more than three standard deviations from the average of normal mice). We found XL spines evoke a marked supralinear synaptic amplification, and clustered inputs to a few XL spines were sufficient to drive neuronal firing. We experimentally and theoretically observed that the cluster density of XL spines negatively correlated with working memory, which can contribute to psychiatric pathophysiology. Furthermore, the cluster density of XL spines was significantly overrepresented in the postmortem brain of schizophrenia than age- and gender-matched control. The currently dominant hypothesis of schizophrenia pathophysiology is a reduction in spine density (the over-pruning hypothesis). However, our results presented here suggest that the hypothesis may need to be revised and that more in-depth investigations into the interactions between intra-spine and dendritic computations and their effect on brain (dys)function are needed.