Aggression Bouts
Number of aggression bouts in the session
ConductVision · Behavioral Analysis
Aggression & Fighting
Time aggression bouts between two animals and see which one attacked.
RodentHamsterSocialAuto Export
ConductVision / Aggression & Fighting
Recording / Trial 3subject tracked
Aggression Bouts14
Attack Latency7.2s
Bout Duration27%
Key Parameters
Metrics automatically extracted by ConductVision.
Attack Latency
Time from the start of the session to the first bout
Bout Duration
Length of each bout, in seconds
Total Aggression Time
Sum of all bout durations in the session
Attacker per Bout
Which of the two animals attacked in each bout, with the model confidence
Bout Timeline
Start and end of every bout, in seconds and in video frames
+ 3 more parameters trackedShow allHide
Frame-by-Frame Labels
Behavior label and model probability for every video frame
Two-Animal Pose Tracking
9 body points on each animal: head, shoulder, two spine points, four paws, and tail base
Pose Export
Body-point coordinates for both animals in DeepLabCut-format CSV
Example output
From frame labels to a bout table
A five-minute resident-intruder session with two animals. The model labels every frame, the clean-up rules turn those labels into bouts, and each bout gets an attacker.
How the clean-up works, 86 to 99 s
The same 13 seconds at each step.
Model outputFrame by frame
Short gaps joined10 frames or less
Bouts kept30 frames or more
- Scored as aggression
- Gap joined
- Animal out of view
- Dropped as too short
- 86.6 s: a 15-frame detection is dropped, because bouts need at least 30 frames.
- 91.2 s: a 6-frame gap is joined, so the fight stays one bout.
- 93.9 s: one animal is out of view for 8 frames. Those frames are not scored, then joined like any short gap.
- 97.4 s: a 15-frame gap is too long to join, so the short detection after it is dropped.
The whole session
Each bout sits in the row of the animal that attacked. Select a bout for its details.
Animal 1 attacking
Animal 2 attacking
- Aggression bouts
- 7
- Attack latency
- 42.0 s
- To the first bout
- Total aggression time
- 30.7 s
- Mean bout length
- 4.4 s
- Bouts started by Animal 1
- 6 of 7
| Bout | Start | End | Duration | Attacker | Confidence |
|---|---|---|---|---|---|
| 1 | 42.0 s | 45.8 s | 3.8 s | Animal 1 | 92% |
| 2 | 61.3 s | 63.0 s | 1.7 s | Animal 1 | 88% |
| 3 | 88.0 s | 97.4 s | 9.4 s | Animal 1 | 90% |
| 4 | 131.5 s | 133.1 s | 1.6 s | Animal 1 | 79% |
| 5 | 170.2 s | 176.9 s | 6.7 s | Animal 2 | 86% |
| 6 | 214.0 s | 219.6 s | 5.6 s | Animal 1 | 86% |
| 7 | 262.4 s | 264.3 s | 1.9 s | Animal 1 | 82% |
The download uses the column names of the ConductVision bout table. The animal scored as attacking is in the intruder column and the other animal is in resident.
What is the Aggression Test?
Aggression in rodents is usually studied with the resident-intruder test. An unfamiliar animal is placed in a resident's home cage, and attacks are scored over a short session. Attack latency, the number of attacks, and time spent fighting measure territorial aggression. Repeated defeat in the same setup is also used as a model of social stress.
ConductVision finds both animals in each video frame and tracks 9 body points on each. After the session, a trained neural network (an LSTM) reads how both animals move over time and marks the frames that belong to an aggression bout. Frames where either animal is out of view are not scored. Short gaps inside a bout, including brief dropouts, are joined so one fight is not split into several, and detections shorter than 30 frames (1 second at 30 fps) are dropped. For each bout you get its start, end, and duration, and which animal attacked.
Two model sets are included: one for lab rodents and one for hamsters. Results export as CSV, with one row per bout and a frame-by-frame file that holds the behavior label and model probability for each frame.
Analysis Parameters
| Parameter | Description | Default |
|---|---|---|
| Animals per Recording | Animals found and tracked in each video | 2 |
| Body Points | Points tracked on each animal | 9 |
| Species Models | Choose the model that matches your animals | Lab rodent or hamster |
| Minimum Bout Length | Shorter detections are dropped | 30 frames (1 s at 30 fps) |
| Gap Bridging | Gaps this short inside a bout are joined | 10 frames (0.33 s at 30 fps) |
| Frame Rate | Read from the camera or video file; the bout rules count frames | 30 fps if not reported |
| Session Length | Set by your protocol | 10 min is common for resident-intruder |
Before you record
Keep both animals in view
Frames where either animal is out of view are not scored. Only dropouts of 10 frames or less inside a bout are joined back in.
Pick the species model
Choose the lab rodent or hamster model to match the animals you record.
Know your frame rate
The bout rules count frames: at least 30 frames per bout, and gaps of up to 10 frames joined. That is 1 s and 0.33 s at 30 fps, or 0.5 s and 0.17 s at 60 fps.
Two animals per recording
The analysis finds and tracks two animals in each video.
Even lighting
Light without glare or flicker helps the software find both animals in every frame.
Methods text
A starting paragraph for your methods section. Fill in the bracketed values from your study.
Aggressive behavior was scored automatically with ConductVision (ConductScience, version [version]) using the [lab rodent or hamster] model. Both animals were detected in every video frame and tracked with nine body points each (head, shoulder, two spine points, four paws, and tail base). A trained LSTM classifier labeled the behavior in each frame. Frames in which either animal was not detected were not scored, gaps of 10 frames or fewer within a bout were joined, and runs shorter than 30 frames were discarded ([frame rate] fps video). For each aggression bout, the start time, end time, duration, and attacking animal were recorded.
Interpreting Results
↑
More Aggression Bouts
Higher territorial aggression. Longer isolation housing and some genetic lines raise attack frequency.
↓
Shorter Attack Latency
The resident attacks sooner, a common sign of heightened aggression.
↑
Longer Attack Latency
Reduced aggression, or reduced activity. Sedation also delays attacks, so check locomotion before calling a drug anti-aggressive.
↑
Longer Bouts
More sustained fighting within each encounter.
→
Attacks Split Between Both Animals
In a standard resident-intruder test one animal does most of the attacking. Attacks from both animals suggest the roles were not set up, for example an intruder larger than the resident.
Research Applications
Resident-Intruder Test
- Territorial aggression: attack latency and number of bouts
- Isolation-induced aggression: effects of single housing
- Strain and genotype comparisons of aggressive behavior
- Hamster aggression, including Syrian hamster resident-intruder tests
Social Defeat Stress
- Chronic social defeat: scoring attacks in each defeat session
- Susceptible versus resilient animals after repeated defeat
Pharmacology
- Anti-aggressive (serenic) compound screening, with activity controls
- SSRI effects on impulsive aggression
- Oxytocin and vasopressin effects on social and aggressive behavior
Related paradigms
Continue with rodents analysis
RodentSocial Behavior
Social Interaction Test
Quantify social approach, avoidance, and interaction behaviors.
RodentSocial Behavior
Sociability Chamber Maze
Three-chamber paradigm for social motivation and social novelty preference.
RodentSocial Learning
Empathy Assay
Observational fear learning and emotional contagion between a demonstrator and an observer.
RodentSpatial Memory
Barnes Maze
Track escape latency, search strategy, and head direction with automated analysis.
RodentAnxiety
Elevated Plus Maze
Measure anxiety-like behavior through open vs. closed arm exploration.
RodentSpatial Learning
Morris Water Maze
Gold-standard spatial learning and memory test with automated swim tracking.
Ready to automate your behavioral analysis?
Request a demo or contact our team to discuss how ConductVision can accelerate your research.
