Parameters of prediction: Multidimensional characterization of top-down influence in visual perception

Abstract

Despite the recent popularity of predictive processing models of brain function, the term prediction is often instantiated very differently across studies. These differences in definition can substantially change the type of cognitive or neural operation hypothesized and thus have critical implications for the corresponding behavioral and neural correlates during visual perception. Here, we propose a five-dimensional scheme to characterize different parameters of prediction: flow of information, mnemonic origin, specificity, complexity, and temporal precision. We describe these dimensions and provide examples of their application to previous work. Such a characterization not only facilitates the integration of findings across studies but also helps stimulate new research questions.

Main Findings

• Flow of Information: Predictions can be recursive, involving constant interaction between predictions and inputs, or sequential, triggered by specific cues with assumed time delays.
• Mnemonic Origins: Predictions stem from episodic or semantic memory, impacting the nature of information conveyed and neural substrates involved.
• Specificity: Predictions range from specific expected inputs to a broad range of potential inputs, influencing the precision of prior expectations.
• Complexity: Predictive processing involves varying levels of information, from low-level sensory details to high-level object information, represented across different cortical sites.
• Temporal Precision: Predictions can provide precise temporal information essential for actions or imprecise timing about future events, supported by different neural mechanisms.

Conclusion

The broad use of prediction across studies poses challenges for aggregating findings. The proposed five-dimensional scheme offers a framework to highlight similarities and differences across studies, helping to avoid unwarranted generalizations and guiding new research questions. The multidimensional characterization presented can shape future cognition and behavior research by targeting unexplored parameter combinations.

Recommended citation: Ortiz-Tudela, J., Nicholls, V. I., & Clarke, A. (2023). "Parameters of prediction: Multidimensional characterization of top-down influence in visual perception." Neuroscience and Biobehavioral Reviews. 153. 105369.
Download Paper | Download Slides