Interspecific variation in cleaning behaviour and cheating among coral reef cleaner fishes

This study reveals that dedicated and non-dedicated coral reef cleaner fish exhibit distinct interspecific variations in cheating behavior and responses to social context, with specific species like *Labroides bicolor* adjusting their exploitation strategies based on client type and the presence of bystanders, thereby highlighting species-specific strategies that influence mutualism stability.

The Gist

Imagine the coral reef as a bustling, underwater city. In this city, there are specialized "barbers" (cleaner fish) whose job is to pick parasites and dead skin off their neighbors (client fish). In exchange for a free haircut and a clean body, the clients give the barbers a meal. It's a classic deal: You scratch my back, I scratch yours.

However, just like in any business, there's a temptation to cheat. Instead of picking off the annoying bugs (the parasites), a barber might decide to take a bite of the client's healthy skin or mucus. It's a tasty, high-calorie snack, but it hurts the client and damages their trust. This is called cheating.

This paper is like a detective story where scientists put seven different species of these "barbers" in a controlled aquarium lab to see: Who is the honest barber, who is the cheat, and does it matter who is watching?

Here is the breakdown of their findings, using some everyday analogies:

1. The "Full-Time" Barbers vs. The "Side-Hustlers"

The researchers split the fish into two groups:

• The Dedicated Cleaners (The Full-Time Barbers): These fish (like the famous Labroides species) rely only on cleaning for food. It's their 9-to-5 job.
• The Non-Dedicated Cleaners (The Side-Hustlers): These fish clean occasionally but mostly eat other things. Cleaning is just a quick snack on the side.

The Finding: The Full-Time Barbers were all over the map. Some were incredibly honest, while others were notorious cheaters. They had very different "personalities." The Side-Hustlers, however, were all pretty much the same: they didn't clean for very long, and they almost never cheated. They were like part-time workers who just want to clock in, do the bare minimum, and clock out.

2. The "Client" Test: Who are you cleaning?

The scientists tested the barbers with three types of "customers":

• The Predator: A big, scary fish that could eat the barber if it got annoyed. (Think of a customer holding a knife).
• The Visitor: A fish that just passes by.
• The Resident: A small, local fish that hangs around the barber's station.

The Finding: The Full-Time Barbers were smart strategists.

• When cleaning the Predator, they were very careful. They knew if they bit the wrong spot, they'd get eaten. So, they acted like perfect, polite barbers.
• When cleaning the Resident (the small guy), they were a bit more relaxed.
• But the star of the show was Labroides bicolor. This fish was the "rude barber." It cheated the most, biting the clients' skin regardless of who they were. It was like a barber who knows the customer can't afford to leave, so they take advantage of them.

3. The "Bystander" Test: Is someone watching?

In the wild, if a barber cheats, other fish might see it and decide never to come back. This is called "reputation management." The scientists tested this by having a "bystander" fish watch the cleaning session.

The Finding: Surprisingly, most barbers didn't care if someone was watching. They kept doing what they were doing.

• The Exception: Labroides bicolor (the rude barber mentioned earlier) actually behaved better when a bystander was watching! It stopped cheating. This suggests that even the worst cheaters know how to put on a good show if an audience is present. It's like a shoplifter who suddenly becomes very polite when the manager walks by.

4. The "Captive" Twist

The paper notes something interesting: The fish in the lab behaved differently than they do in the wild. In the wild, cheating is often much more frequent. In the lab, the fish were less stressed and had plenty of food, so they didn't need to cheat as much. It's like how a person might be more honest when they aren't starving.

The Big Picture Takeaway

This study teaches us that not all "good guys" are the same.

• Specialists (Full-time cleaners) have evolved complex strategies. They know exactly how to balance being nice and being greedy, depending on who they are dealing with and who is watching. They are the "master negotiators" of the reef.
• Generalists (Side-hustlers) don't bother with complex strategies. They just do a quick job and move on.

The coral reef isn't a place where everyone follows the same rules. It's a complex society where different species have evolved different "personalities" to survive, cheat, and cooperate in their own unique ways. The next time you see a cleaner fish, remember: you're looking at a master strategist who knows exactly how to play the game!

Technical Summary

1. Problem Statement

Mutualistic relationships, such as coral reef cleaning interactions, are inherently vulnerable to conflict when one partner (the cleaner) cheats by consuming the client's mucus or scales instead of ectoparasites. While the behavior of the well-studied Labroides dimidiatus is known to be highly sophisticated (e.g., reputation management, tactile stimulation), it remains unclear whether:

• Interspecific variation exists among other dedicated cleaners (other Labroides species) and non-dedicated (opportunistic) cleaners.
• Context-dependent strategies (adjusting cheating based on client identity or the presence of a bystander) are widespread or species-specific.
• The ecological reliance on cleaning (dedicated vs. non-dedicated) drives the evolution of different behavioral strategies regarding service investment and exploitation.

2. Methodology

The study employed a standardized experimental framework to compare seven cleaner fish species:

• Subjects:
  • Dedicated Cleaners (4 species): Labroides dimidiatus, L. bicolor, L. pectoralis, L. rubrolabiatus.
  • Non-Dedicated Cleaners (3 species): Larabicus quadrilineatus, Halichoeres poeyi, Thalassoma lunare.
• Clients: Three distinct client types were used:
  • Predator: Paracirrhites forsteri.
  • Non-resident: Zebrasoma scopas.
  • Resident: Dascyllus trimaculatus.
• Experimental Design:
  • Interaction Test: Each cleaner interacted sequentially with the three client types in a controlled tank. Interactions were recorded for 45 minutes. The order was fixed (Predator $\to$ Non-resident $\to$ Resident) to manage stress levels.
  • Bystander Test: Cleaners interacted with a client (Z. scopas) in the presence and absence of a bystander fish (another Z. scopas) in an adjacent tank.
• Data Collection & Analysis:
  • Video Analysis: Videos were processed using DeepLabCut (pose estimation) and SimBA (supervised machine learning) to automatically detect and quantify two key metrics:
    1. Interaction Time: Duration of active cleaning.
    2. Client Jolts: Rapid movements by the client away from the cleaner, used as a behavioral proxy for cheating (mucus feeding).
  • Statistical Modeling:
    • Interaction Time: Analyzed using Gamma generalized linear models (GLM).
    • Jolt Occurrence: Modeled using bias-reduced binomial logistic regression (hurdle approach) to handle zero-inflated data.
    • Jolt Rates: Analyzed using negative binomial GLMs for videos where jolts occurred, with interaction time as an offset.
    • Bystander Effects: Assessed using Wilcoxon signed-rank tests.

3. Key Contributions

• Comparative Framework: This is one of the first controlled studies to simultaneously compare multiple dedicated and non-dedicated cleaner species under identical social stimuli, moving beyond single-species case studies.
• Machine Learning Integration: The study utilized advanced pose-estimation and supervised machine learning (SimBA/DeepLabCut) to objectively quantify complex social behaviors (jolts and interaction time) across multiple species, reducing observer bias.
• Decoupling Investment and Quality: The research explicitly separates "service investment" (time spent) from "service quality" (frequency of cheating), revealing that these traits do not always correlate linearly across species.

4. Key Results

• Interspecific Variation:
  • Dedicated Cleaners: Showed high behavioral variability. L. bicolor exhibited the highest cheating rates (up to 8.6 jolts/100s) and longest interaction times. L. rubrolabiatus was the most cooperative (lowest jolts).
  • Non-Dedicated Cleaners: Showed uniform, low-investment profiles. They had significantly shorter interaction times and extremely rare jolts (near-zero cheating), suggesting opportunistic rather than specialized strategies.
• Client-Dependent Modulation:
  • Cheating and interaction times varied significantly based on client identity.
  • Predator Clients: Generally elicited higher jolt rates in L. bicolor (contrary to wild expectations where predators usually suppress cheating), suggesting captive conditions may alter risk assessment.
  • Resident Clients: Elicited the lowest investment and cheating, likely due to their small size and lower value as a food source.
• Bystander Effects:
  • Overall, bystander effects were weak across most species.
  • Species-Specific Adjustment: Only L. bicolor significantly reduced client jolts when a bystander was present, indicating a capacity for reputation management. L. dimidiatus showed a non-significant trend in the same direction. Non-dedicated cleaners showed no bystander sensitivity.
• Captive vs. Wild Discrepancies:
  • Jolt rates in captivity were generally lower than wild reports, and interaction times were longer. This suggests that the absence of natural ecological constraints (e.g., predation risk, competition) in the lab allows for different behavioral expressions.

5. Significance

• Evolution of Cooperation: The findings suggest that strong ecological reliance on cleaning (dedicated species) drives the evolution of diverse, species-specific strategies to balance exploitation and cooperation. In contrast, non-dedicated cleaners face weaker selection pressures, resulting in uniform, low-risk strategies.
• Stability of Mutualism: The study highlights that mutualisms are not monolithic; the stability of the interaction depends on the specific behavioral phenotype of the cleaner species involved. High variance in dedicated cleaners may facilitate coexistence by reducing direct competition for specific client types.
• Cognitive Complexity: The species-specific response to bystanders (notably in L. bicolor) challenges the assumption that only L. dimidiatus possesses advanced social cognition, suggesting that reputation management may be more widespread but context-dependent within the genus.
• Methodological Advance: The successful application of machine learning to quantify "cheating" in multi-species interactions provides a robust template for future behavioral ecology studies.

In conclusion, the paper demonstrates that cleaning mutualisms are shaped by a complex interplay of species-specific traits, ecological dependence, and social context, with dedicated cleaners exhibiting a broader spectrum of strategic behaviors than their opportunistic counterparts.