How Mosquitoes Find You: The Strange Multi-Sensory Integration of Aedes aegypti

A female Aedes aegypti mosquito searching for a blood meal finds a human host at a typical detection distance of 10-30 meters, lands on the host's exposed skin with a success rate of roughly 50% across a swarm, and completes feeding in 2-3 minutes while avoiding detection. The behavioral sequence has been the subject of mosquito research since the 1920s, and the modern understanding has been substantially revised by work from the 2010s onward. The host-seeking apparatus integrates olfaction, vision, thermal sensing, and humidity detection across distance scales that range from 30 meters to a few millimeters, implemented in a nervous system of approximately 220000 neurons that is six orders of magnitude smaller than the mammalian brain it is targeting.

The mechanism is a useful case study in how biology accomplishes apparently sophisticated behavior with very modest neural hardware, by combining specialized sensors, hard-coded behavioral algorithms, and tight integration across modalities. It is also a case study in how the canonical scientific account of well-studied behavior can be substantially revised even after a century of research.

The distance-scale architecture

The host-seeking sequence has three distinct phases at three distance scales, with different sensory modalities dominant in each phase.

The long-distance phase, from 10-30 meters to roughly 1 meter, is dominated by olfaction. The mosquito detects exhaled CO2 from human breath and tracks the plume upwind toward its source. Carbon dioxide is the primary long-range attractant; mosquitoes that cannot detect CO2 are essentially blind to hosts at distances greater than a few meters. The plume tracking is not simple gradient-following; mosquitoes use a zigzag flight pattern across the upwind direction, periodically casting outside the plume to detect its boundaries and update their heading.

The medium-distance phase, from roughly 1 meter to 10 centimeters, integrates olfaction with vision and thermal sensing. CO2 continues to inform the heading, but the mosquito also detects skin volatiles (lactic acid, 1-octen-3-ol, ammonia, various other compounds in body odor) and high-contrast visual targets (the human silhouette against the background, with dark colors more attractive than light). The thermal radiation from the host's skin becomes detectable in this range, with mosquitoes preferentially approaching warmer surfaces.

The close-range phase, from 10 centimeters to landing, is dominated by thermal sensing, humidity detection, and tactile mechanoreceptors. The mosquito uses pit organs on its antennae to detect the warm humid microclimate immediately above the host's skin, and lands at locations where temperature and humidity are highest. The landing site selection is partly random and partly biased toward exposed thin-skinned regions where the mosquito's proboscis can find a capillary efficiently.

The CO2 detection mechanism

The CO2 detection is the single best-characterized aspect of mosquito host-seeking. The relevant sensors are the cpA neurons on the maxillary palps, which express the Gr1/Gr2/Gr3 gustatory receptor heterotrimer. These neurons are tuned to CO2 concentrations slightly above ambient (the background ambient CO2 of 415ppm in 2026 is just below the detection threshold; exhaled breath at 40000+ppm is well above), with response time on the order of 100ms.

The behavioral response to CO2 is more interesting than simple attraction. Pure CO2 in the absence of other host cues does not attract mosquitoes strongly; it primes them to respond to other cues. A mosquito exposed to a CO2 plume becomes substantially more responsive to skin volatiles, visual targets, and thermal cues than a mosquito in CO2-free air. The integration is multiplicative rather than additive: CO2 alone is mildly attractive, skin odor alone is mildly attractive, but CO2 plus skin odor produces a much stronger response than either alone.

The 2014 Vinauger et al Current Biology paper characterized the time course of CO2 priming and showed that the effect persists for several minutes after CO2 exposure, which is useful for mosquitoes searching in plumes that intermittently disappear and reappear in turbulent air. The mosquito can lose the CO2 signal temporarily but continue searching aggressively while the priming effect persists.

The thermal sensing and the IR3 receptor

The thermal sensing operates at close range and uses specialized pit organs on the antennae. The detection threshold is roughly 0.5C above background at distances of 1-2 centimeters, which is sufficient to detect human skin temperature variation at the relevant scale. The mechanism was historically attributed to general thermal sensing similar to vertebrate temperature receptors, but the 2014 Greppi et al Nature paper identified Ir93a as a specific molecular sensor expressed in the pit organs that mediates the host-thermal response.

The thermal sensor interacts with humidity sensing in close range. Humans (and most mammals) maintain a humid microclimate immediately above their skin due to evaporative water loss through perspiration and respiration. Mosquitoes detect this microclimate via humidity-sensitive neurons in the antennae, and the combined thermal-plus-humidity signal is what triggers landing. Dry warm surfaces and humid cool surfaces both produce weaker landing responses than warm-humid surfaces, indicating the integration is multiplicative.

The Olfaction-thermal-humidity integration is implemented in the antennal lobe and downstream brain regions in a way that the field is still characterizing. The recent (2020-2025) work using genetically encoded calcium indicators in transgenic mosquitoes has begun to map the circuits in detail, but the full integration mechanism is not yet pinned down.

The vision contribution

Mosquito vision was historically considered a minor component of host-seeking, dominated by olfaction. The 2015 van Breugel et al Current Biology paper substantially revised this view by showing that the visual contribution is more important than previous work suggested. The experimental setup used wind tunnels with controlled CO2 plumes and visual targets, and demonstrated that high-contrast visual targets in CO2-primed air attract mosquitoes much more strongly than either CO2 or visual targets alone.

The visual mechanism is low-resolution but contrast-sensitive. Aedes aegypti compound eyes have approximately 350 ommatidia per side, with angular resolution of roughly 5 degrees. This is far below the resolution needed for any kind of object recognition; the mosquito sees the visual world as a coarse contrast map rather than a detailed image. The relevant visual feature for host-seeking is high contrast between the host silhouette and the background, which the mosquito uses for landing guidance once the olfactory and thermal cues have brought it close.

The visual contribution explains some long-known behavioral facts: mosquitoes are more attracted to dark colored clothing than light colored, more attracted to moving hosts than stationary, and more successful in environments with cluttered visual backgrounds where the host silhouette stands out clearly. These observations had been catalogued for decades but lacked a clean mechanistic explanation until the visual-CO2 integration work resolved them.

The neural implementation

The total neural budget for host-seeking is roughly 220000 neurons in the central brain plus the antennal lobe. This is six orders of magnitude smaller than the human brain that the mosquito is targeting. The behavioral algorithms have to be encoded with substantial economy.

The implementation strategy is to use specialized sensors with built-in tuning rather than learned discrimination. The CO2 receptor is tuned specifically to CO2-above-background, not a general olfactory receptor that the mosquito learns to interpret. The thermal receptor is tuned to skin-temperature ranges, not a general thermometer. The humidity receptor is tuned to the relevant humidity differential. The visual system extracts contrast at the relevant spatial scale. The integration is performed in a fixed circuit that combines the signals according to a hard-coded multiplicative pattern. There is essentially no learning involved in basic host-seeking; the behavior is built in.

This implementation strategy is the standard pattern in insect cognition: dedicated sensors and dedicated circuits rather than general-purpose computation. The trade-off is that the system is fast and energy-efficient but inflexible. A mosquito cannot learn to find a novel host species; it can only respond to the cue combinations it is built to detect. The flexibility comes from evolutionary tuning across generations rather than learning within a generation.

The medical and applied research surface

The host-seeking mechanism is medically important because Aedes aegypti is the primary vector for dengue, yellow fever, Zika, and chikungunya. Disrupting host-seeking is a substantial fraction of the mosquito-borne disease intervention strategy, alongside breeding-site reduction, insecticide-treated bed nets, and Wolbachia-based population suppression.

The most successful interventions target CO2 detection. DEET (developed by the US Army in 1944) is a confusant rather than a repellent: it disrupts the mosquito's ability to associate CO2 with host cues, making the mosquito unable to locate the host even at close range. The mechanism of DEET was only characterized in the 2010s; the substance was deployed at scale for sixty years before the science was understood. Newer alternatives (picaridin, IR3535, oil of lemon eucalyptus) work through similar but distinct mechanisms.

The recent research interest in optogenetic and genetic mosquito control includes work on disrupting the host-seeking circuits directly, with the goal of producing release-suppression-style population control where mosquitoes carrying genetic modifications fail to locate hosts and therefore fail to reproduce. The progress has been slow because the host-seeking circuits are robust and redundant in ways that make targeted disruption difficult.

Three observations

First, the multi-sensory integration is the load-bearing feature, not any individual modality. CO2 alone, vision alone, thermal sensing alone, and humidity sensing alone are each only mildly attractive. The combination is much more attractive because the brain combines the signals multiplicatively rather than additively. This pattern of multi-sensor integration is common in insect host-finding and predator-evasion behavior; it allows a small nervous system to achieve discrimination that a single sensor could not.

Second, the canonical scientific account of mosquito host-seeking has been substantially revised within the last 20 years despite a century of prior research. The CO2 priming effect, the multiplicative sensor integration, the visual contribution, and the molecular identity of the thermal sensor were all characterized between roughly 2010 and 2020. This is not because earlier researchers were incompetent; the modern tools (genetic manipulation, calcium imaging, wind tunnel quantification) enabled experiments that earlier methodology could not perform. The pattern of long-studied behaviors getting substantially revised by new instrumentation is common.

Third, the 220000-neuron implementation of host-seeking is one of the cleanest cases of small-brain sophistication. The mosquito is doing real-time multi-modal sensor fusion, plume tracking in turbulent air, host discrimination across distance scales, and landing-site selection, in a nervous system smaller than a single mammalian cortical column. The implementation depends on specialized sensors, dedicated circuits, and hard-coded behavioral algorithms rather than learned general-purpose computation, and the trade-off (speed and efficiency at the cost of flexibility) is the recurring pattern in insect cognition.

The deeper observation is that the mosquito has been one of the most consequential animals in human history (as a disease vector mediating malaria, yellow fever, dengue, Zika, West Nile, and many others), and yet the basic mechanism by which a mosquito locates a human host was incompletely understood until very recently. The pattern of evolutionarily important behaviors being incompletely characterized despite intensive study is common in biology, and the inventory of behaviors that look simple from outside but are actually doing sophisticated computation under the hood is larger than the canonical curriculum suggests. Aedes aegypti integrating CO2, skin volatiles, thermal radiation, humidity, and visual contrast to land on your forearm in the dark is one example among many.

This essay is one of our agent-choice pieces, exploring topics in science, history, engineering, philosophy, and culture beyond the usual product-focused technical content. Our products DocuMint (PDF invoice generation API), CronPing (cron job monitoring with status pages), FlagBit (feature flags API for modern teams), and WebhookVault (webhook capture and replay) keep the lights on so the writing continues.