Rethinking Insect Identification: An evidence-based framework for pest identification under uncertainty

By Samantha Kiever, Research Entomologist, Insects Limited, Inc.

Entomology is a fascinating field full of fascinating and incredibly brilliant individuals. There are none that I admire more than taxonomists. Pouring over wing venation and dichotomous keys for the absolute best and most accurate identification is arduous and requires an immense amount of knowledge and skill. When it comes to the pest management industry, the need for such specificity in insect identification isn’t always present or even attainable given the quality of specimen, description, or photograph provided. Still, we need to make decisions under uncertainty even if a species-level identification isn’t immediately feasible.

The internet is a vast and sometimes slightly terrifying megacosm of human thought and perspective. As such, the internet is full of insect identifications that range from impressively accurate to wildly speculative to even downright confusing. One reason for this variability is that many people approach identification as an all-or-nothing task: either get down with the dichotomous keys and the textbooks or recognize by sight by looking at a specimen and then attempt to match it to something they have seen before. While recognition can be useful, it is not always reliable. Insects are the most speciose and diverse group of animals on the planet. There are millions more insect species that an individual is unaware of than there are familiar species. 

I prefer a different approach. Rather than asking “What is this?” I ask myself more often “What can’t this be.” The ultimate difference between the two approaches is seeking confirmation and seeking elimination. When attempting to identify an insect by recognition alone, the starting point is a conclusion that confirmation is then sought for, which means that this method is rife with the potential for confirmation bias, influence from the ego, and  opportunities for misidentifications to remain undetected (You can’t exactly check your work with the same method you used to do the work. There will be more on this later). The method I’m about to outline removes the ego and self by removing intermediate early confirmation and conclusions as much as possible. It also integrates context to fill in some gaps where compromised or poorly depicted morphology leaves them. 

An Important Note on Using this Framework 

This framework is above all else an approach to evidence. While it is laid out in a format that suggests it should be followed from step one to step two not unlike assembling a bicycle from instructions, this framework is an internal process that can be done without scratch paper and corkboards with red string. The goal is not to systematically follow this process from beginning to end, but rather to change one’s approach to evaluating evidence and drawing conclusions in the face of imperfect information. 

Identify the Scope – What problem needs solving? 

When approached with a potential pest issue and evidence of the problem, pay attention to the question that is being posed. If, for example, you are approached with something vaguely tubular in shape, and the question is, “Is this a bed bug?” then the immediate answer is no. This response does not mean that it is known what the insect is. It means we know what it is not, which is sometimes sufficient depending on what is being asked. After all, “Is this a bed bug?” is a very different question than, “What is this?” 

At the same time, a person who asks if a specimen with four flat, broad wings that appear golden and covered in scales asks, “Is this a bed bug?” this should not immediately preclude follow-up questions and determining whether a different question altogether is the right one to ask. If I were posed that question in the face of this image, I would ask, “Are you seeing more than one of these?” because I cannot immediately determine that the insect, despite clearly not being a bed bug, is not some other pest-related species. If the answer were yes, then I would be sure to run through the following steps without excluding potential pest-related species immediately. In this way, identifying the scope requires the most vigilance, as an expert knows what pests look like, and sometimes the scope requires redirecting to build an accurate differential.  

Build a Differential –  Generating Plausible Hypotheses 

When a patient arrives at a doctor's office with a cough, fever, and fatigue, the physician rarely knows immediately what condition is responsible. Instead, they develop a differential diagnosis: a hierarchical list of plausible explanations for the symptoms. As new information becomes available, some possibilities are ruled out while others become more likely. The process continues until the most supported diagnosis remains. I use the same framework when identifying insects from photographs, descriptions, and specimens. 

The first step is not identifying the insect. The first step is building a differential. If I receive a photograph or incomplete specimen, I begin by asking broad questions. Is it an insect at all? Could it be an arachnid? Does it possess the characteristics expected of an arthropod? Once I establish those fundamentals, I begin constructing a list of plausible candidates. Depending on the quality of the evidence, that differential may begin at the level of class, order, family, or occasionally genus. The exact taxonomic level matters less than the process itself. What matters is maintaining multiple working hypotheses until the evidence justifies eliminating them. 

Elimination – Stress Testing the Differential 

Once a differential has bene established, the next step is to begin reducing it. Every observation becomes an opportunity to remove possibilities by gaining more information that is inconsistent with some hypotheses. Morphology is king in both deterministic (dichotomous keys) and differential models. Wing structure, antennae, body shape, segmentation, coloration, size, and other visible characteristics can all help determine which candidates remain plausible and which do not. What is important is never reading too much into an image. If a structure is visible but unclear, it may not be a useful piece of information, or it might be useful enough. For instance, you may be unable to see the actual segmentation of an antenna, but you can still discern the relative length or the presence/absence of a club. 

Context can also play an important role. Information about where an organism was found, the environment in which it was encountered, associated materials, seasonal activity, and even the question being asked can all provide useful clues. A small beetle recovered from a food-processing facility exists within a different context than a similar beetle recovered from a windowsill. Likewise, reports of numerous tiny insects appearing after a plumbing leak may increase the plausibility of moisture-associated organisms while making other explanations less likely. 

Importantly, context should be used to refine the differential rather than replace it. A conclusion should not be reached simply because it seems to fit the circumstances. The strongest hypotheses are those that remain consistent with both the morphology and the surrounding evidence. 

As possibilities are removed, the differential narrows. Sometimes this process leads to a single highly supported explanation. More often, especially when working from photographs or incomplete specimens, several plausible candidates remain. That is perfectly acceptable. The goal of elimination is not to force certainty but to ensure that only explanations consistent with the available evidence survive. 

The Working Hypothesis – What’s the best-supported explanation right now? 

One of the most useful concepts borrowed from medicine is the working diagnosis. A physician does not always have the luxury of waiting for complete certainty before taking action. Decisions are often made using the explanation that is currently best supported by the available evidence while remaining open to revision if new information emerges. The same principle applies to insect identification. Once the differential has been narrowed as far as the evidence reasonably allows, I establish a working hypothesis. This is not necessarily a final identification. It is simply the explanation that best accounts for the available evidence and therefore serves as the best guide for what should happen next. 

Sometimes, new information emerges only because action was taken in the first place. I refer to this, somewhat tongue-in-cheek, as "Dr. Housian logic," coined after a certain abrasive television diagnostician from the early aughts. Throughout the series, House and his team frequently arrive at a working diagnosis, begin treatment, and then use the patient's response to that treatment as an additional source of information. While reality is thankfully less dramatic than television, the underlying principle is strong. By attempting to address a problem, you inevitably learn something about it. The intervention may produce the expected result, lending support to the working hypothesis, or it may fail entirely, forcing you to reevaluate your assumptions and revisit the differential. Pheromone monitors can be a valuable tool, as monitoring for a specific species and seeing no difference between the unbaited traps and the baited traps you put out points to a species-pheromone mismatch that may then require you reassess the current working hypothesis. 

The objective is not to force certainty where none exists, nor is it to become emotionally attached to a particular conclusion. The objective is to make the best possible decision using the evidence currently available while remaining willing to change course when new evidence appears. In this sense, the working hypothesis is less a conclusion than a practical explanatory tool that helps guide investigation until the evidence supports something better. 

Communicating Uncertainty  

The most important skill in identification is knowing where the evidence ends. Every photograph, specimen, or description contains a finite amount of information. Once that information has been exhausted, and further narrowing becomes speculative.  This is why confidence should always be proportional to evidence.

Sometimes the evidence supports a species-level identification. Sometimes it only supports to family, order, or even a broad arthropod group. Sometimes, depending on the scope defined at the outset, the evidence supports species-level identification, but only a broad arthropod group is needed. All of these outcomes are valid provided they accurately reflect the evidence at hand.  

A statement such as "This is likely a beetle" may be more scientifically defensible than an unsupported species-level identification. Likewise, "Species-level identification is not possible from this photograph" is not an admission of failure; it’s an honest assessment of the limitations of the evidence. Further caveats regarding whether you can confirm or deny the possibility of said beetle being potentially associated with pest activity may serve the situation and spur the acquisition of supplemental evidence to then narrow down the identification and increase the confidence level. 

In my experience, the best identifiers in applied areas of entomology are not necessarily those who can provide the most specific answer. They are those who can accurately communicate what is known, what remains uncertain, and where the limits of the available evidence lie. 

Final Thoughts 

Ultimately, the goal of this framework is accuracy rather than certainty. Certainty is appealing, but nature does not always provide enough information to justify it. Photographs are blurry, specimens get damaged, observations are incomplete, and sometimes the evidence runs out before a definitive answer can be reached. 

By building a differential, systematically eliminating possibilities, developing a working hypothesis, and communicating uncertainty honestly, it becomes possible to make informed decisions even when the initial information is incomplete. In pest management, as in medicine, that is often exactly what is required. The objective is not to know everything. The objective is to follow the evidence as far as it will reliably take you and no further. 

Insects Limited, an Insect Pheromone Company

Insects Limited, Inc. researches, tests, develops, manufactures and distributes pheromones and trapping systems for insects in a global marketplace. The highly qualified staff also can assist with consultation, areas of expert witness, training presentations and grant writing.

Insects Limited, Inc. specializes in a unique niche of pest control that provides mainstream products and services to protect stored food, grain, museum collections, tobacco, timber and fiber worldwide. Please take some time to view these products and services in our web store.