Studies in spatial learning. I. Orientation and the short-cut.

E. C. Tolman, B. F. Ritchie, D. KalishJournal of Experimental Psychology 1946 · DOI

TL;DR

Tolman and his colleagues challenge the prevailing stimulus-response (S-R) dogmas of the 1940s by demonstrating that rats learn more than just a sequence of turns; they learn the spatial location of a goal. By showing that rats can navigate a novel "short-cut" to a food source when their habitual path is blocked, the authors provide foundational evidence for what would later be known as "cognitive maps."

What Problem Does This Paper Address?

In the mid-20th century, behaviorism was dominated by Stimulus-Response (S-R) theories, which posited that learning consists of strengthening connections between specific environmental stimuli and motor responses (e.g., "when at the junction, turn left"). Tolman argued for a more sophisticated "Expectancy Theory."

The challenge was that earlier versions of Expectancy Theory were so vaguely defined that they were indistinguishable from S-R theories. If "expectation" is just a "tendency to respond in a certain way," then it’s just S-R theory with different labels. Tolman, Ritchie, and Kalish sought to redefine "expectation" as a disposition to orient toward a physical location in space, regardless of the specific motor path taken. They designed an experiment to prove that rats possess this abstract spatial knowledge.

Methods

The researchers used a custom-built elevated maze to test 56 female rats (a mix of Tryon "bright" and "dull" strains).

  1. Preliminary Training: Rats were trained to run a specific, indirect route to a food box. As seen in Figure 1, the path involved a circular table, a short alley (CD), and a series of turns (DE, EF, FG) leading to the goal (G). A dim light (H) was placed behind the food box to serve as a landmark.

    Figure 1. Apparatus used in preliminary training.
  2. The Test Trial: On the fifth day, the original path was blocked at the end of the alley (CD). The researchers then added 18 new paths radiating from the circular table like spokes on a wheel (see Figure 2).

    Figure 2. Apparatus used in the test trial.
  3. The Goal: Path No. 6 pointed directly toward the location where the food box had been during training. If rats only learned a sequence of turns, they should have been lost. If they learned the location of the food, they should choose Path 6.

Datasets

The "dataset" consisted of the behavioral choices of 56 female rats. 26 were from the Tryon "bright" stock and 30 were from the Tryon "dull" stock (strains bred for high and low maze-learning performance, respectively).

Key Insights and Ideas

  • Redefining Expectation: The authors reject the idea that expectation is a simple response tendency. Instead, they define it as a "conditioned definition": if a rat is motivated and its path is blocked, it will choose a novel path that points toward the goal.
  • Spatial vs. Sequence Learning: The experiment distinguishes between "response learning" (memorizing a chain of muscle movements) and "place learning" (understanding the Euclidean coordinates of a goal).
  • Landmarks as Relational Cues: The authors argue that the light at H wasn't a simple "conditioned stimulus" that pulled the rat toward it. Instead, it served as a landmark that allowed the rat to judge its own location relative to the goal within the total environment.
  • Human-Animal Continuity: The authors explicitly compare the rat's behavior to human symbolic behavior (e.g., understanding the sign "There is bread in the kitchen"), suggesting that both involve abstracting a goal's location from specific paths.

Key Results

The results strongly supported the Expectancy Theory:

  • The Short-Cut Preference: 36% of the rats (19 out of 53) chose Path No. 6, the direct route to the goal. This was the most frequent choice by a wide margin (see Figure 3).
  • Failure of S-R Generalization: Only 9% of rats chose paths 9 and 10, which were spatially closest to the original path. This suggests the rats weren't just trying to get back on their old route; they were aiming for the goal itself.
  • Cognitive Parity: Interestingly, there was no significant difference between the "bright" and "dull" rats; 10 brights and 9 dulls chose the short-cut.
Figure 3. Numbers of rats which chose each of the paths.

Limitations and Future Work

  • Incomplete Success: While Path 6 was the clear winner, the majority of rats (64%) chose other paths. The authors hypothesize that these rats may have been under-trained and had not yet fully learned the location of the food.
  • The "Light" Criticism: Critics might argue the rats were simply running toward the light. The authors counter this by noting that the angle of the light on Path 6 (50 degrees) was different from the training path (head-on), and that rats didn't show similar preferences for adjacent paths (5 and 7) where the light angle was only slightly different.
  • Future Directions: This paper was the first in a series intended to explore the nuances of spatial learning, setting the stage for decades of research into the hippocampus and the neural basis of cognitive maps.

Summary updated

BibTeX
@article{tolman1946studies,
  author = {E. C. Tolman and B. F. Ritchie and D. Kalish},
  title = {{Studies in spatial learning. I. Orientation and the short-cut.}},
  journal = {Journal of Experimental Psychology},
  year = {1946},
  doi = {10.1037/h0053944},
}
Figure