Conclusion

Reading the eleven body sections together, the review establishes a position on VIP-expressing cortical interneurons that is more restrictive than the textbook reading and more structured than the catalogue of exceptions. At the subclass level, VIP-INs are a CGE-derived, 5-HT3AR+/Adarb2+ GABAergic population whose identity is reproduced across rodent, primate, and human single-cell atlases, recovered as one of four conserved cortical GABAergic subclasses Rudy et al., 2010Tasic et al., 2018Hodge et al., 2019Bakken et al., 2021Lee et al., 2023Yao et al., 2023. Within the subclass, the lineage decomposes into multiple t-types whose discreteness depends on dataset, taxonomic resolution, and integration method, and whose intrinsic, morphological, and connectivity profiles are graded along the bipolar/bitufted/multipolar and irregular-spiking/continuous-adapting axes that Morphological Diversity and Intrinsic Electrophysiology chart Tasic et al., 2016Tasic et al., 2018Gouwens et al., 2020Prönneke et al., 2015Cauli et al., 1997Cauli et al., 2014Apicella & Marchionni, 2022Schneider-Mizell et al., 2025. The disinhibitory reading of VIP function — VIP cells, recruited by behavioural state, suppress SST cells and so release pyramidal dendrites from inhibition — is supported as a population-average operating mode in primary visual, auditory, somatosensory, prefrontal, and cingulate cortex Pi et al., 2013Lee et al., 2013Pfeffer et al., 2013Fu et al., 2014Cichon et al., 2017Williams & Holtmaat, 2019Veit et al., 2023, but it is one operating mode among several, and the literature provides equally well-documented preparations in which VIP recruitment activates SST cells, inverts pyramidal modulation, or fails to modulate pyramidal output at all Pakan et al., 2016Dipoppa et al., 2018Garrett et al., 2020Yavorska & Wehr, 2021Bigelow et al., 2019Anastasiades et al., 2021Bastos et al., 2023.

The five cross-cutting tensions previewed in Introduction admit, on this reading, partial resolutions rather than dissolutions. The discrete-versus-continuous question for VIP t-types is best read as resolution-dependent: supervised clustering on transcriptomic data recovers nominally discrete groups, but cross-modal Patch-seq integration, morphological metrics, and intrinsic-property continua reveal graded boundaries between many of those groups, with continuity predominant within the VIP Lamp5, VIP Sncg, and irregular-spiking branches Tasic et al., 2018Gouwens et al., 2020Prönneke et al., 2015Apicella & Marchionni, 2022Emmenegger et al., 2018Schneider-Mizell et al., 2025. The subtractive-versus-divisive gain ambiguity reduces to operating regime: VIP perturbation is divisive when SST recruitment is concurrent and dendritic inhibition is dominant, subtractive when the disinhibitory release falls on perisomatic compartments, and approximately mixed when both are engaged, with the rate-model literature reproducing each of these outcomes from the same wiring under different parameter draws Kuchibhotla et al., 2016Hertäg & Sprekeler, 2019Litwin-Kumar et al., 2016Molino et al., 2017Veit et al., 2023Tahvili et al., 2025Tahvili et al., 2025. The net-disinhibition versus paradoxical-effect conflict reduces to whether the local circuit is in the inhibition-stabilised regime: outside it, VIP activation produces the textbook disinhibitory gain; inside it, the same wiring produces paradoxical pyramidal suppression on VIP activation, and the same connectomic prior is consistent with both behaviours Molino et al., 2017Sanzeni et al., 2020Reimann et al., 2026Hertäg & Sprekeler, 2019Tahvili et al., 2025Tahvili et al., 2025.

The two remaining tensions are less tractable from the present evidence. The identifiability of the VIP→SST weight in four-population rate equations is, on the present data, an unresolved bottleneck: VIP→SST and VIP→PV cross-couplings trade off against SST→Pyr and PV→Pyr terms in a way that connectomic priors do not constrain, and several published models with incompatible VIP→SST weights fit the same in-vivo locomotion, attention, and reward data Hertäg & Sprekeler, 2019Litwin-Kumar et al., 2016Molino et al., 2017Veit et al., 2023Reimann et al., 2026Sabri & Batista-Brito, 2024. Computational Models of VIP Circuit Function argues that this is the field’s principal next-decade problem and that simultaneous SST/PV/VIP/Pyr recordings under structured perturbations — paired with electron-microscopy connectomes that report not only contacts but also synapse weights — are the minimum required to break the degeneracy Reimann et al., 2026Schneider-Mizell et al., 2025Hertäg & Sprekeler, 2019Bos et al., 2025. Cross-species translation runs along an analogous axis: VIP-IN identity is conserved at the subclass level between rodent and primate cortex, but the human VIP subclass expands to twenty-one t-types in middle temporal gyrus, includes upper-layer enrichment and human-specific VIP PCDH20 and MC4R morphologies absent in mouse, and over-represents in disease-associated programmes that the rodent atlases do not capture Hodge et al., 2019Boldog et al., 2018Chartrand et al., 2023Bakken et al., 2021Bakken et al., 2021Krienen et al., 2020Lee et al., 2023.

These conflicts converge on a single integrated reading. VIP function is best treated not as a fixed circuit operation but as the output of an area×state×modulator triple that In Vivo Function During Behavior and VIP Interneurons Across Brain Regions describe in cortical terms and that Oscillatory Dynamics and Temporal Coordination extends into the temporal domain via gamma gain modulation, theta-locked IS-3 firing, Up-state biasing, and predictive-coding/attention frames Veit et al., 2023Cichon et al., 2017Hertäg & Sprekeler, 2020Hertäg & Clopath, 2021Shipp, 2016Francavilla et al., 2018Tyan et al., 2014Tricoire et al., 2011Gulyás et al., 1996. Within that triple, primary visual cortex VIP cells implement contrast-dependent, locomotion-gated gain enhancement Fu et al., 2014Dipoppa et al., 2018Pakan et al., 2016Millman et al., 2020; auditory cortex deploys them for reinforcement-linked disinhibition and critical-period plasticity, with a polarity puzzle around movement-driven SST activation Pi et al., 2013Yavorska & Wehr, 2021Bigelow et al., 2019; somatosensory cortex recruits them via motor copy during active sensing Lee et al., 2013Sermet et al., 2019; medial prefrontal and anterior cingulate cortex implement top-down control of attention, working memory, and pain Bastos et al., 2023Anastasiades et al., 2021Kamigaki & Dan, 2017Li et al., 2022; and hippocampal IS-3 cells assemble a circuit motif that targets other interneurons rather than principal cells, breaking the cortical disinhibitory assumption at the wiring level Gulyás et al., 1996Tyan et al., 2014Tricoire et al., 2011Kawaguchi & Kubota, 1996Francavilla et al., 2018. The same triple maps onto the disease literature reviewed in Species Differences, Human Relevance, and Disease: VIP-IN dysfunction has been documented in autism, Rett, Dravet, schizophrenia, and Alzheimer-related models, with the sign of the effect depending on perturbation polarity, brain area, and disease model rather than on a uniform vulnerability of the subclass Mossner et al., 2020Goff et al., 2023McFarlan et al., 2024Bhandari et al., 2024Hall et al., 2015Goel et al., 2025Kranz et al., 2025Goff & Goldberg, 2019.

What the corpus does not yet support is also identifiable. It does not support a single rate or spiking model that reproduces all in-vivo VIP recordings with one set of cross-coupling weights Hertäg & Sprekeler, 2019Molino et al., 2017Veit et al., 2023Bos et al., 2025Reimann et al., 2026. It does not support an unambiguous mapping from t-type to behavioural function: most in-vivo recordings target VIP cells via Cre-driver labelling that pools multiple t-types, so the t-type composition of the imaged or photo-stimulated population is rarely specified in the studies that motivate the disinhibitory reading Pi et al., 2013Fu et al., 2014Pakan et al., 2016Dipoppa et al., 2018Prönneke et al., 2015Schneider-Mizell et al., 2025. It does not yet support direct transfer of mouse circuit conclusions to human cortex, given the primate expansion of CGE-derived diversity and the human-specific morphologies in upper layers Hodge et al., 2019Boldog et al., 2018Chartrand et al., 2023Bakken et al., 2021Krienen et al., 2020. And it does not yet support a settled account of how VIP-mediated disinhibition implements credit assignment or predictive coding: the disinhibitory motif has been recruited as the substrate for both, but the empirical tests that would discriminate the two readings have not been performed Hertäg & Sprekeler, 2019Hertäg & Sprekeler, 2020Hertäg & Clopath, 2021Wagatsuma et al., 2022Lee et al., 2025Rossbroich & Zenke, 2025Nemati et al., 2025Nemati et al., 2025Sabri & Batista-Brito, 2024Shipp, 2016.

The forward agenda the review supports follows from these limits rather than from the strengths. Three lines of work would discriminate the unresolved cases. First, simultaneous multi-class recordings (VIP, SST, PV, Pyr) under structured perturbations, paired with synapse-weight-resolved connectomes, would constrain the VIP→SST weight and break the rate-equation degeneracy that Computational Models of VIP Circuit Function exposes Schneider-Mizell et al., 2025Reimann et al., 2026Hertäg & Sprekeler, 2019Bos et al., 2025Sabri & Batista-Brito, 2024. Second, t-type-resolved in-vivo recordings — Patch-seq-tagged or driver-line-restricted to single VIP t-types rather than to the pooled subclass — would allow the area×state×modulator triple to be decomposed by t-type and would test whether the gain regime, polarity, and target compartment of VIP modulation are t-type properties or only area properties Gouwens et al., 2020Prönneke et al., 2015Apicella & Marchionni, 2022Schneider-Mizell et al., 2025Tasic et al., 2018. Third, comparative recordings in primate and human ex-vivo and in-vivo preparations — leveraging the cross-species t-type alignments already in hand — would test whether the rodent disinhibitory reading transfers and how the human-specific VIP types behave functionally Hodge et al., 2019Boldog et al., 2018Chartrand et al., 2023Bakken et al., 2021Krienen et al., 2020Lee et al., 2023. Until those discriminating experiments are performed, the calibrated map this review provides is the safest intermediate position: the disinhibitory motif holds as a population-average operating mode under specifiable area, state, and modulator conditions; it fails or runs in reverse under specifiable other conditions; the wiring is reproducible at the subclass average but ambiguous at the parameter level; and the cross-species transfer is partial, with disease-relevant divergences clustered in upper-layer human-specific VIP types Pi et al., 2013Pakan et al., 2016Dipoppa et al., 2018Molino et al., 2017Reimann et al., 2026Hodge et al., 2019Chartrand et al., 2023McFarlan et al., 2024.