Experimental setting affects the performance of guppies in a numerical discrimination task

Area of Science:

• Comparative cognition
• Animal behavior
• Neuroethology

Background:

• Guppies (Poecilia reticulata) previously demonstrated numerical discrimination (4 vs. 5 objects), a skill rare outside mammals and birds.
• This prior study involved foraging for rewards in patches of objects, raising questions about the task's ecological relevance.
• It remained unclear if guppies possess exceptional numerical accuracy or a specific foraging-related bias.

Purpose of the Study:

• To investigate guppy numerical discrimination abilities in a standard cognitive task.
• To determine if performance differences were due to the task setting or inherent numerical capacity.
• To identify factors influencing numerical discrimination in guppies.

Main Methods:

• Guppies were trained on numerical discriminations (e.g., 2 vs. 3 objects) using a two-choice task with dot stimuli on a screen.
• Stimuli were presented on a computer screen, with rewards for approaching the larger numerosity.
• Variations included decision time, stimulus type (2D vs. 3D), training length, and error correction.

Main Results:

• Guppy performance drastically decreased in the computer-based task, with most failing even the simplest discrimination (2 vs. 3).
• Modifying task parameters (time, stimuli, training, correction) did not significantly improve performance to previous levels.
• A group trained with 3D stimuli showed improved performance on the easiest numerical task.

Conclusions:

• Guppy numerical discrimination performance is highly dependent on the experimental setting.
• The results suggest potential learning constraints rather than inherent limitations in numerical acuity.
• Guppies may be predisposed to estimate patch quality in foraging contexts, not abstract numerical associations.