Spatial prediction of dynamic interactions in rats

Abstract

Animals and humans receive the most critical information from parts of the environment that are immediately inaccessible and highly dynamic. The brain must effectively process potential interactions between elements in such an environment to make appropriate decisions in critical situations. We trained male Long-Evans rats to discriminate static and dynamic spatial stimuli and to generalize novel dynamic spatial stimuli displayed on an inaccessible computer screen. We provide behavioral evidence indicating that rats encode dynamic visuospatial situations by constructing internal static representations that capture meaningful future interactions between objects. These observations support previous findings in humans that such internal static representations can encapsulate relevant spatiotemporal information of dynamic environments. This mechanism would allow animals and humans to process complex time-changing situations neatly.

Fig 1. Prediction of a dynamic interaction.

(A) Time compaction theory proposes that a dynamic scenario is internally represented by statically mapping the predicted future interactions. This compact internal representation (CIR) enables efficient learning and memorization of dynamic situations [19], allowing real-time decision-making, a primary requirement for survival. (B) Visuospatial prediction of future interaction of dynamic stimuli presented on the distant computer screen during the visuospatial discrimination (VSD) task. (C) Scheme of the behavioral apparatus used for the VSD task with an example of dynamic stimulus. The time grayscale indicates the direction of movement of the circles (from bottom to top and from right to left in the figure), which disappear before colliding. The red circle denotes the position of the future collision between the circles, which was not displayed to the rat. (D) Experimental hypothesis. After training, the rat associates the particular position of a static circle with a reward obtained after pressing the lever (left). Then, when a novel dynamic stimulus is presented with two moving circles that would collide at the position of the static circle in the familiar rewarded stimulus, the rat should tend to press the lever since the CIR of the novel stimulus is similar to the familiar rewarded stimulus (right; circle grayscale and stimulus red circle as in panel C).

Fig 2. Compact internal representation of future interactions in rats.

(A) VSD task - static version: Generalization test. We used two familiar rewarded static stimuli (Static REW1 and Static REW2, in red rectangle) and two familiar non-rewarded static stimuli (Static non-REW1 and Static non-REW2) in this generalization test. In addition, we presented the rats with four novel stimuli pseudorandomly displayed between the presentations of the four familiar training stimuli. The novel dynamic stimuli had the location of the predicted collision identical to the static positions of familiar training stimuli (red circle, not displayed to the rats): Compl. to REW1 to Static REW1, Compl. to REW2 to Static REW2, Compl. to non-REW1 to Static non-REW1, Compl. to non-REW2 to Static non-REW2. No novel stimulus was rewarded. The time grayscale denotes the direction of movement of the circles in the novel stimuli, so they move from gray to white positions. The rats discriminated between rewarded and non-rewarded familiar static object positions. (B) Mean rate of pressing the lever during the presentations of the two types of familiar stimuli in the generalization test, with the solid lines depicting the regression curves that model the temporal structure of the lever-pressing rate (odds ratio (OR) = 6.53). (C) Lever-pressing rate distribution throughout the two types of familiar stimuli duration (histograms binned at 200 ms). (D) Probability of the first lever press throughout the duration of the two types of familiar stimuli after the stimulus onset. The orange asterisk marks significant differences (p < 0.0001). The rats pressed the lever more often during the novel dynamic stimuli in which two circles approached the familiar static rewarded positions. (E) Main rate of pressing the lever duringthe presentations of the two types of novel stimuli in the generalization test (odds ratio (OR) = 1.24). (F) Lever-pressing rate distribution throughout the two types of novel stimuli duration (histograms binned at 200 ms), with the regression curve modeling the temporal structure given by the solid line. (G) Probability of the first lever press throughout the duration of the two types of novel stimuli after the stimulus onset. The purple asterisk marks significant differences (p < 0.0001). Familiar stimuli lasted 15 sec. Novel dynamic stimuli lasted 10 sec to prevent learning during the session that these stimuli were not rewarded. Data in panels B, D, E, and G are shown as means ± SEM. The rat and session were introduced as random effect (grouping) factors and time (trial) and stimulus type (rewarded/non-rewarded) as fixed effect factors (the same for subsequent GLMM analyses; see Materials and methods).

Fig 3. Supporting evidence of the compact internal representation of future interactions in rats.

(A) VSD task - static version: Generalization test. During this generalization test, we presented the rat with four novel dynamic stimuli pseudorandomly displayed between the presentations of the four familiar training stimuli (shown in panel A, upper). One novel stimulus shared the location of the predicted collision with the Dynamic REW stimulus (Compl. to REW), another with the Dynamic non-REW stimulus (Compl. to non-REW), and two more stimuli had no predicted collision point (Control 1 and Control 2). The rats pressed the lever more often during the rewarded dynamic stimulus than the static one when novel non-rewarded dynamic stimuli were introduced. (B) Mean rate of pressing the lever during the presentations of individual familiar stimuli in the generalization test (odds ratio (OR) = 1.37). (C) Lever-pressing rate distribution throughout individual familiar stimuli duration (histograms binned at 200 ms and solid lines as in Fig 2). (D) Probability of the first lever press throughout individual familiar stimuli duration after the stimulus onset. The red asterisk marks significant differences (p < 0.0001) between the Dynamic REW stimulus and all other familiar stimuli; further statistical differences are described in the text. The rats pressed the lever with equal probability during the novel dynamic stimulus that shares the spatial information with rewarded familiar stimuli and during the rewarded familiar dynamic stimulus, and more often than during other novel stimuli. (E) Mean rate of pressing the lever during the presentations of individual novel stimuli in the generalization test (odds ratios: Compl. to REW vs. Compl. to non-REW OR = 2.6, Compl. to REW vs. Control 1 OR = 1.95, Compl. to REW vs. Control 2 OR = 2.2). (F) Lever-pressing rate distribution throughout individual novel stimuli duration (histograms binned at 200 ms and solid lines as in Fig 2). (G) Probability of the first lever press throughout individual novel stimuli duration after the stimulus onset. The purple asterisk marks significant differences (p < 0.0001) between the Compl. to REW stimulus and all other novel stimuli; further statistical differences are described in the text. Familiar stimuli lasted 15 sec. Novel dynamic stimuli lasted 10 sec to prevent learning during the session that these stimuli were not rewarded. Data in panels B, D, E, and G are shown as means ± SEM.

Fig 4. Faster learning and shorter reaction time of dynamic versus static stimulus discrimination.

(A) VSD task - dynamic version. We presented the rats with two complementary rewarded stimuli – one static and one dynamic (Static REW and Dynamic REW, in red rectangle) and two complementary non-rewarded stimuli (Static non-REW and Dynamic non-REW). The static stimuli were stable white circles, and the complementary dynamic stimuli consisted of two circles moving toward the static circle location (red circle – predicted collision point; not displayed) but disappearing half of the way (gray-to-white color of the circles denote their movement direction, given by the time grayscale). (B) Lever-pressing rate for individual stimuli during the first training configuration (odds ratio (OR) = 1.18). (C) Hazard ratios of the first lever press after the stimulus onset for individual stimuli during the first training configuration. The red asterisk marks significant differences (p < 0.01 for panel B, p < 0.05 for panel C) between the Dynamic REW stimulus and all other stimuli. Further statistical differences are described in the text. Data in panels B and C are shown as means ± SEM (Dynamic-REW vs. Static-REW hazard ratio (HR) = 1.19).