Insect Behaviour and the Design of Laboratory Bioassays

Studying insect behaviour

Design of laboratory bioassays

Dr Joe RobertsReader in Integrated Pest Management

MIBTP Masterclass July 2026

A bioassay is a chain of four decisions

A bioassay converts a biological question into a measurable response under defined conditions.

01

Define

State the contrast and one scoreable behaviour.

02

Match

Choose an arena that permits the required behaviour.

03

Protect

Identify the unit and control other variation.

04

Claim

Plan the analysis and limit the conclusion.

The question determines the arena. The arena determines the response. The response determines the claim.

Roberts et al. (2023) Entomologia Experimentalis et Applicata · Percie du Sert et al. (2020) PLOS Biology

Foundations

PART ONE

Behaviour, evidence and the insect being tested

Behaviour is an observable response

External stimulus

Odour · contact · light · humidity · host texture

Internal state

Age · hunger · experience · reproductive state · time

OBSERVABLE
RESPONSE

defined · detectable · repeatable

Operational definition

Set explicit start, stop and scoring rules.

Choice · latency · path · transition · duration · frequency

Variation in the stimulus or the insect's state can create bias or hide a real response.

The evidence ladder limits the claim

01Cue at the insect

Was the odour, light or contact present?

→

02Sensory detection

Did a receptor system respond?

→

03Behaviour

What does the insect do?

→

04Biological consequence

Did survival, reproduction or damage change?

ONE RESULT, ONE RUNG

A result supports the rung that was measured. Later rungs require separate evidence.

Roberts et al. (2023) Entomologia Experimentalis et Applicata

Validate the stimulus reaching the insect

Treatment labelApplied dose

≠

At the insectStimulus received

Sensory channelPreserve or validateResponse enabled

Airborne odourRelease rate, airflow, surface loss and source ageOrientation

Surface contactDose, texture, residues and access to plant tissueAcceptance

VisionSpectrum, intensity, contrast and geometryDirected movement

Touch and vibrationBench vibration and unintended air movementStartle or movement

Temperature and humidityThe gradient across the arena, not the room settingActivity

An uncontrolled sensory gradient becomes an alternative explanation.

Roberts et al. (2023) Entomologia Experimentalis et Applicata · Döring (2014) Annals of Applied Biology · Powell et al. (2006) Annual Review of Entomology

Different paths can produce the same count

ORTHOKINESIS Speed changes

Speed changes with stimulus intensity.

KLINOKINESIS Turning changes

Turning rate changes without directed movement.

TAXIS Direction changes

Movement follows a directional cue or gradient.

ARRESTMENT Progress slows

Speed falls and/or turning rises after encounter.

A terminal occupancy count cannot distinguish these processes. Record paths, speed, turning, transitions or latency.

Roberts et al. (2023) Entomologia Experimentalis et Applicata

State and history define the insect you test

CURRENT STATE

Development

Narrow age or moult-stage window

Form and reproduction

Sex, wing form and mating state

Feeding and activity

Deprivation, recovery and time of day

POPULATION + EXPERIENCE

Origin

Clone, population and collection history

Culture

Generation, density and rearing environment

Prior exposure

Host, cue and learning history

Keep these factors consistent or balance them among treatments. The tested population sets the boundary of the claim.

Ramírez & Niemeyer (2000) Journal of Insect Behavior · McLean et al. (2009) Ecological Entomology

Handling can become the treatment

01 · TRANSFER

Use one defined method

Brush contact, aspiration and dislodgement can trigger immobility or escape walking.

02 · RECOVERY

Do not trust appearance alone

CO₂ or chilling can alter behaviour after visible movement returns.

03 · ACCLIMATION

Pilot a fixed interval

Use blank trials to find when baseline behaviour stabilises.

04 · RECORD

Make handling reproducible

Record the tool, duration, time off host, recovery, conditions and operator.

There is no universal acclimation time. Pre-specify it and validate it for the assay.

Johnson (1958) Animal Behaviour · Phelan et al. (1976) Annals of the Entomological Society of America · Powell & Bale (2006) Physiological Entomology · Bartholomew et al. (2015) Scientific Reports

From question to assay

PART TWO

Endpoints, observation and apparatus

Define one primary endpoint

INFORMAL OBSERVATION

“The aphid seemed to prefer the untreated leaf.”

↓

OPERATIONAL DEFINITION

Specify start and stop rules, observable criteria and treatment of ambiguous cases.

Observable: trained observers apply the same rule
Bounded: the window and zones are explicit
Aligned: the measure represents the tested process

Candidate endpointData structure

First option enteredcategory

Latency to contacttime until response

Residence timecontinuous

Response by the time limityes / no

Probe countcount

Turning ratecontinuous

Nominate the endpoint that answers the question. Treat the remainder as secondary or exploratory.

Percie du Sert et al. (2020) PLOS Biology

Choice and no-choice tests differ

Choice test

Relative selection

Which option is selected when alternatives are simultaneously available?

Tests discrimination among offered options
Responses within an arena are dependent
A dominant option can mask acceptance

No-choice test

Acceptance under constraint

Is an option accepted when no alternative is available?

Tests acceptance under confinement
Can reveal delayed acceptance
Confinement can exaggerate field use

Preference, acceptance and suitability are different claims.

Murray et al. (2010) Biological Control

Observation is part of the instrument

01

Record

Video or timestamped event log

02

Score

Explicit start, stop, zones and ambiguity rules

03

Validate

Blind rescoring and tracking checks

EVENT LOGPredefine what the observer can enter

00:00Releasedtrial starts

00:18First contactzone boundary crossed

01:42Settledcriterion held for 30 s

Reliability checkRescore 20% blind · Cohen's κ (categories) · intraclass correlation (continuous scores)

Automation increases throughput only when tracking and event classification are validated.

Noldus et al. (2001) Behavior Research Methods, Instruments, & Computers · Cohen (1960) Educational and Psychological Measurement · Koo & Li (2016) Journal of Chiropractic Medicine

See what each assay system can measure

Scientific illustration of a moving-air Y-tube olfactometer with treatment and control source chambers, balanced airflow and an aphid release chamber — RELATIVE ORIENTATIONY-tube olfactometer
First arm entered · equal flow · side balance

Top-down scientific illustration of a four-arm olfactometer with four discrete curved odour fields and central extraction — MULTIPLE ODOUR FIELDSFour-arm olfactometer
Zone use over time · check airflow and field separation

Scientific illustration of an electrical penetration graph system with a wired aphid, plant electrode, amplifier and waveform display — FEEDING PROCESSElectrical penetration graph
Probing waveforms · tethering may alter feeding

Scientific illustration of a wind tunnel with an odour source, visualised plume and insect release platform — DIRECTIONAL MOVEMENTWind-tunnel plume
Upwind progress · measure the plume at the insect

ACCEPTANCE ARENASLeaf disc or Petri dish

Settling, departure or feeding initiation

CONSEQUENCE ARENASCage or semi-field assay

Survival, reproduction, damage or transmission

Match the method to the claim

INTENDED CLAIM SUITABLE STARTING METHOD WHAT THE RESULT CAN SUPPORT

Sensory detection

Electroantennography (EAG)
GC–EAD · single-sensillum recording

Sensory activity, not preference. A cue may be detected without an EAG response.

Relative orientation

Y-tube · four-arm olfactometer

Choice relative to the alternatives supplied in a validated odour field

Directed movement

Wind tunnel · tracked gradient

Movement through a measured plume or gradient

Acceptance or feeding

Petri dish · leaf disc · EPG (feeding)

Behaviour in that arena; confinement or tethering can alter it

Biological consequence

Cage · whole plant · semi-field

Survival, reproduction or damage under the tested conditions

Define the claim first. Then choose the arena, endpoint and analysis that can support it.

Roberts et al. (2023) Entomologia Experimentalis et Applicata · Jacob (2018) Frontiers in Cellular Neuroscience · Walker et al. (2024) Entomologia Experimentalis et Applicata

Start with the biological claim rather than a favourite apparatus. Electroantennography (EAG) records a summed whole-antenna response. Gas chromatography–electroantennographic detection (GC–EAD) links separated compounds to antennal responses, while single-sensillum recording samples individual sensory units. None establishes preference, and no EAG response does not prove that the insect cannot detect the compound. Choice arenas estimate relative orientation under the alternatives and stimulus field supplied. Wind tunnels and gradient arenas test directed movement only when that plume or gradient has been measured. Petri dishes and leaf discs can test acceptance or departure; electrical penetration graph (EPG) recording resolves probing and feeding processes, but handling, wiring and tethering can alter behaviour. Survival, reproduction and crop effects require a later-rung cage, whole-plant or semi-field assay.

The endpoint and analysis must make the same match. Examples include a binomial or mixed logistic model for a binary choice, survival analysis for latency, and a count model for offspring or feeding events. These are starting families, not automatic choices: the experimental unit and design still govern the analysis. (Roberts et al., 2023; Jacob, 2018; Walker et al., 2024)

Protect the comparison

PART THREE

Find the flaw, identify the unit and diagnose failure

Diagnose the flawed experiment

2 MINUTES IN PAIRS

Does a botanical extract repel green peach aphid (Myzus persicae)?

Insects · mixed-age culture; brushed directly into dishes; no recovery or acclimation

Contrast · extract in acetone always on the left; right side untreated; no evaporation interval

Replication · five dishes; twenty aphids released together in each dish

Endpoint · count each side at 1 hour; discard aphids in the centre

Scope · one day, one cohort and one extract batch; no validated known-response control

Analysis · treat 100 aphids as independent and claim repellency

Find at least six faults. For each, name the conclusion it threatens and the minimum repair.

One treated dish is one experimental unit

ILLUSTRATIVE SIMULATION · NO TREATMENT EFFECT WAS BUILT IN

Treatment was assigned once per dish. The 20 aphids within a dish share the same treatment and conditions.

10 INDEPENDENT ASSIGNMENTS

Five treated dishes + five control dishes

TTTTT CCCCC

Experimental unit: one dish
Subsamples: 20 aphids per dish

INCORRECT · POOL THE APHIDS81% versus 59%Two-sample z-test of proportionsn = 200 · p < 0.001

Assumes aphids sharing a dish are independent.

CORRECT · COMPARE DISHESFive dish proportions per groupTwo-sided exact permutation testn = 10 · p = 0.26

Matches the level at which treatment was assigned.

Count independently assigned units—not every observation collected within them.

Percie du Sert et al. (2020) PLOS Biology · Lakens (2022) Collabra: Psychology

Treatment was assigned once per dish, so the independent units are the ten dishes. The 20 aphids within a dish are subsamples because they share treatment and conditions. No treatment effect was built into this simulation; the aggregate difference arose from dish-to-dish variation.

The incorrect analysis pools 81 of 100 versus 59 of 100 responding aphids and uses a two-sided pooled two-sample z-test of proportions. That test assumes 200 independent yes/no observations. Here that assumption is false, giving p = 0.000687, shown as p < 0.001.

The dish-level analysis first calculates one response proportion for each dish. The illustrative counts were 19, 20, 14, 20 and 8 responses out of 20 in the treated dishes, and 17, 13, 8, 3 and 18 out of 20 in the control dishes. Their mean response proportions differ by 0.22. A two-sided exact randomisation, or permutation, test considers all 252 ways to allocate these ten dish values five versus five. Sixty-six allocations produce an absolute mean difference of at least 0.22, so p = 66/252 = 0.262, shown as p = 0.26.

This example demonstrates that the analysis must preserve the level of treatment assignment. It is not a general instruction to use a permutation test: blocked or repeated bioassays may require a binomial GLMM, beta-binomial model or another design-appropriate cluster-aware analysis.

Identifying the experimental unit tells us what to count, not how many units are enough. Choose the smallest effect that would change the biological conclusion, estimate variation among experimental units from a pilot, and calculate or simulate power for the model that will actually be fitted. The support guide continues this example. (Percie du Sert et al., 2020; Lakens, 2022)

Randomise, block and blind

Design actionWhat it protects againstBioassay example

RandomisePredictable treatment assignmentTreatment side and run order

BlockKnown temporal or apparatus variationDay, cohort, apparatus or source batch

BalanceChance imbalance in small experimentsEqual treatments within each block

BlindObserver and analysis expectationsCoded videos and treatment labels

Record the allocation and actual run order, including departures from the plan.

Percie du Sert et al. (2020) PLOS Biology

Non-response is part of the result

DEFINE BEFORE DATA COLLECTION

What counts as a response?
How will an insect that reaches the time limit be recorded?
Which equipment or handling failures justify exclusion?

REPORT IT, ALWAYS

Numbers released, responding and excluded
Reasons and rates for each treatment
Whether reasonable alternative rules change the result

Immobility, delayed response or dropping may be biology, not equipment failure.

Percie du Sert et al. (2020) PLOS Biology

Diagnose non-response

Do not assume the treatment has no effect—or simply add more insects.

Use the checks below to separate a genuine biological non-response from a bioassay that is not functioning as intended.

CheckEvidence to examineWhat the finding means

01 · CUE DELIVERY

Measure odour, light, temperature, airflow or surface dose inside the arena.

If the intended cue was absent or unstable, repair delivery before repeating the test.

02 · INSECT CONDITION

In untreated trials, check activity after handling, recovery and acclimation.

Low baseline activity points to insect state or handling rather than the treatment.

03 · ARENA DESIGN

Review recordings for wall effects, confinement and unintended gradients.

If the arena suppresses the expected movement, change the arena or endpoint.

04 · RESPONSE RULE

Look for behaviour that occurred but was not counted by the endpoint.

Revise the rule only before a new confirmatory run; retain the original result.

05 · KNOWN-RESPONSE CONTROL

Where one exists, test a cue already known to elicit the measured response.

If that control also fails, the assay is not yet functioning reliably.

Diagnose first. Keep valid non-responses in the dataset, and increase replication only after the assay passes these checks.

Roberts et al. (2023) Entomologia Experimentalis et Applicata

Preregister the plan and check permissions

TIME-STAMP THE DECISIONS

Preregister before collecting data

State the endpoint, experimental unit, smallest worthwhile effect, sample-size and stopping rule, exclusions, allocation and analysis.

OSF Registrations · ARRIVE Study Plan

RESPONSIBLE PRACTICE

Legal scope is not the standard

ASPA covers vertebrates and cephalopods, not insects. The Animal Welfare (Sentience) Act 2022 currently defines animals as vertebrates. Neither boundary removes ethical duties: minimise numbers and avoidable harm; define handling, endpoints and disposal; check local expectations.

PLANT HEALTH

Start authorisation early

In England and Wales, ask APHA CIT whether scientific authorisation is required. Apply through eDomero well in advance: technical assessment and a site visit may be needed. Involve your biological safety committee.

Defra scientific authorisations

Record deviations from the registered plan. Define containment, decontamination, disposal and incident procedures.

Home Office (2024) GOV.UK · UK Parliament (2022) Legislation.gov.uk · Defra (2024) UK Plant Health Information Portal

Preregistration means time-stamping the decisions that could otherwise drift after seeing the data. OSF can embargo a registration. ARRIVE covers living invertebrate models and is useful as a planning checklist even where ASPA does not apply.

ASPA currently protects living vertebrates and cephalopods. The Animal Welfare (Sentience) Act 2022 concerns recognition of animal sentience in government policy and currently defines animals as vertebrates; it is not a laboratory licensing statute. These legal boundaries are not an ethical standard. Students should minimise numbers while preserving valid inference, avoid unnecessary harm, justify handling and confinement, predefine welfare endpoints, and plan containment and disposal. They must also check the current requirements of their institution, funder and target journal.

In England and Wales, ask APHA’s Centre for International Trade (CIT) whether scientific authorisation is required before obtaining the organism. Applications use Defra’s eDomero route. Current guidance says to apply in plenty of time because technical assessment and a site visit may be needed. Scotland and Northern Ireland use different competent authorities. (Nosek et al., 2018; Percie du Sert et al., 2020; Home Office, 2024; UK Parliament, 2022; Defra, 2024)

Four questions before the first trial

1

What exactly is the claim?

DEFINE · Name the response, comparison, conditions and target population.

2

Can this arena test it?

MATCH · Confirm that the cue reaches the insect and the predicted behaviour is possible.

3

What else could explain it?

PROTECT · Identify the experimental unit and control plausible alternatives.

4

What result would support it?

CLAIM · Predefine the endpoint, analysis, exclusions and claim boundary.

The question determines the arena. The arena determines the response. The response determines the claim.

Your challenge: design the bioassay

PRIMARY HYPOTHESIS

Across the two supplied clones, a standardised brush touch increases the probability that a settled aphid leaves a field-bean leaf disc within 60 seconds, compared with an approach-only control.

Side-view schematic of a 90-mm Petri dish with a field-bean leaf disc lying flat on damp filter paper and one aphid on the disc

90-mm dishLeaf disc flat on baseDamp filter paper

LEAVES crosses onto the paperREMAINS stays on the disc

AVAILABLE TO EVERY GROUP

Basic equipment · two hours · one supplied pea aphid clone · one supplied black bean aphid clone · one aphid, leaf disc and dish per trial.

YOUR GROUP MUST SPECIFY

the brush-touch and approach-only procedures;
settling, acclimation and technical-failure rules;
the experimental unit, blocks and randomisation;
the primary response, data table and analysis;
the conclusion this assay could—and could not—support.

COMMON STRUCTUREOne complete block = four dishes: touch and control for each clone. Total dishes = 4 × blocks; n per cell = blocks.Primary contrast: touch versus control across both clones.

Deliverable: a protocol another group could run without asking what you meant.

Braendle & Weisser (2001) Journal of Insect Behavior · Gish (2021) Journal of Pest Science · Rasekh et al. (2010) Journal of Insect Science

Place the leaf disc flat on damp filter paper in the Petri-dish base. This arena measures departure from the disc; it cannot reproduce or identify a natural drop from an intact plant. Keep the claim at that level. Use one named clone of each species. A clone is a genetically matched lineage; with one clone per species, the result cannot establish a general species difference.

Groups should decide the details rather than copy a completed design from the slide. Require them to define a reproducible touch and approach-only control, settling and technical-failure rules, the experimental unit, randomisation, one primary response, the planned analysis and the boundary of the claim. Before data collection, agree one common protocol so that class data can be pooled. The primary contrast is touch versus approach-only control, averaged equally across the two supplied clones. Treat the treatment-by-clone interaction as exploratory unless the number of complete blocks supports it. For the pooled class exercise, start with a binomial GLM containing treatment, clone and complete block as fixed effects. Report model-predicted departure probabilities and their treatment contrast with a 95% confidence interval. The worked protocol and alternatives are in the companion support guide. (Braendle and Weisser, 2001; Gish, 2021; Rasekh et al., 2010)

Questions?

Open the student support guide ↗

Dr Joe Roberts · jroberts@harper-adams.ac.uk