PALMS: Pavlovian Associative Learning Models Simulator

TL;DR

PALMS addresses the need for a unified, open-source platform to compare multiple Pavlovian learning theories under a single experimental design. It implements five attentional-learning models, including a unified MLAB extension, and adds configural cues computation to assess complex cue interactions. The suite enables high-throughput, phase-based simulations with per-stimulus parameters, randomisation, and comprehensive plotting/export options, demonstrating capabilities through canonical and novel designs. This tool advances theoretical testing, reproducibility, and rapid hypothesis generation in associative learning research. By facilitating direct model comparisons and large-scale stimulus simulations, PALMS has practical impact for computational neuroscience, psychology, and related fields seeking precise, shareable, and extensible simulation infrastructure.

Abstract

Simulations are an indispensable step in the cycle of theory development and refinement, helping researchers formulate precise definitions, generate models, and make accurate predictions. This paper introduces the Pavlovian Associative Learning Models Simulator (PALMS), a Python environment to simulate Pavlovian conditioning experiments. In addition to the canonical Rescorla-Wagner model, PALMS incorporates several attentional learning approaches, including Pearce-Kaye-Hall, Mackintosh Extended, Le Pelley's Hybrid, and a novel extension of the Rescorla-Wagner model with a unified variable learning rate that integrates Mackintosh's and Pearce and Hall's opposing conceptualisations. The simulator's graphical interface allows for the input of entire experimental designs in an alphanumeric format, akin to that used by experimental neuroscientists. Moreover, it uniquely enables the simulation of experiments involving hundreds of stimuli, as well as the computation of configural cues and configural-cue compounds across all models, thereby considerably expanding their predictive capabilities. PALMS operates efficiently, providing instant visualisation of results, supporting rapid, precise comparisons of various models' predictions within a single architecture and environment. Furthermore, graphic displays can be easily saved, and simulated data can be exported to spreadsheets. To illustrate the simulator's capabilities and functionalities, we provide a detailed description of the software and examples of use, reproducing published experiments in the associative learning literature. PALMS is licensed under the open-source GNU Lesser General Public License 3.0. The simulator source code and the latest multiplatform release build are accessible as a GitHub repository at https://github.com/cal-r/PALMS-Simulator

Figures (6)

• Figure 1: PALMS layout displaying a MLAB model simulation of Haselgrove et al., Experiment 3 two_kinds_of_attention. The design includes three Phases and one Group. The different sections of the layout are identified and framed by a black rounded rectangle, namely, the Design input (top of the GUI), and from left to right, the Model selection buttons, model Parameters area, Per CS Parameters area, the Plot area, the simulation Functional options section, and at the bottom, the Phase selection arrows.
• Figure 2: Simulation results for a within-subjects blocking design for Groups Blk Exp1 McN and Group Blk HS Target across phases. From left to right, Phase 1, Phase 2, and Phase 3. Individual stimuli are shown with different coloured lines and marker shapes. The top panel display RW simulations, whereas the bottom panel presents Mackintosh Extended simulations of the same hypothetical experiment. Random trials included 500 sequences.
• Figure 3: Simulations of a simplified Learned Irrelevance lepelley2004. The plots on the left show associative strength across 8 training trials during Phase 2 for the Learned Irrelevance, CS-Prexposure, and Novel groups. On the left side, the corresponding $\alpha$ value. From top to bottom, simulations of the Pearce, Kaye and Hall model, the Mackintosh Extended model, Le Pelleys' Hybrid model and the MLAB model. Random trials included 500 sequences.
• Figure 4: A simulation of Haselgrove et al. 2025 Experiment 2 novelty_mismatch_haselgrove. The top panel shows the results predicted by the Pearce-Kaye-Hall model during the test in Phase 2, for Group D-novel (left panel) and Group D-repeated (right panel). The middle panel displays corresponding results as simulated by Le Pelley’s Hybrid model, using an out-of-range $\alpha^H$ value (0.05). On the bottom panel, Le Pelley’s Hybrid model predictions are displayed with parameters within the model's range ($\alpha^H=0.9$). Random trials included 200 sequences.
• Figure 5: Simulation of Byrom and Murphy 2019 Experiment 1 byrom2019cue run with 1500 random sequences. The left panel shows the results predicted by the RW model with configural cues for Group Bicond (middle lines) and Group Unicond (top and bottom lines). The right panel displays Mackintosh Extended simulated associative strength for the same groups and conditions.