Skip to main content

Searching high and low: domestic dogs’ understanding of solidity


Physical reasoning appears central to understanding how the world works, suggesting adaptive function across the animal kingdom. However, conclusive evidence for inferential reasoning about physical objects is limited to primates. We systematically tested a central feature—understanding of solidity—in domestic dogs, by adapting a validated procedure (the shelf task) previously used to test children and non-human primates. Dogs watched a treat dropped into an apparatus with a shelf either present (treat landing on the shelf) or absent (treat landing on the bottom surface) and chose where to search for it (above or below the shelf). Across four studies (n = 64), we manipulated visual access to the treat trajectory and apparatus interior. Dogs correctly inferred the location of treats using physical cues when the shelf was present (Study 1), and learned rapidly when visual cues of continuity were limited (Study 2), and when the shelf was absent (Study 3). Dogs were at chance when the apparatus was fully occluded, and the presence and absence of the shelf varied across trials within subjects, and showed no evidence of learning (Study 4). The findings of these four studies suggest that dogs may be able to make some inferences using solidity and continuity and do not exhibit proximity or gravity biases. However, dogs did not always search correctly from Trial 1, and failed to search correctly when the rewarded location varied within-subjects, suggesting a role for learning, and possible limits to their ability to make inferences about physical objects.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3


  1. We recognize that testing performance on Trial 1 is often considered the most stringent assessment of inferential ability (though see e.g., Povinelli and Henley (2020) for a recent argument against privileging Trial 1 data), resource limitations did not permit collecting a sample large enough to power first trial analyses and thus we elected to use repeated trials in order to increase power with a conventional sample size. For instance, to detect at least 70% correct performance at 80% power with α = 0.05 on Trial 1 alone, a minimum of 47 dogs per study would be required, instead of the current 16 dogs per study.


Download references


Natural Sciences and Engineering Research Council of Canada, Grand number (2016-05552).

Author information

Authors and Affiliations


Corresponding author

Correspondence to Julia Espinosa.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 192 KB)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Espinosa, J., Tecwyn, E.C. & Buchsbaum, D. Searching high and low: domestic dogs’ understanding of solidity. Anim Cogn 25, 555–570 (2022).

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Domestic dog
  • Canine science
  • Solidity
  • Physical inference
  • Naive physics
  • Comparative cognition