Framing spatial cognition: neural representations of proximal and distal frames of reference and their roles in navigation

Abstract

The most common behavioral test of hippocampus-dependent, spatial learning and memory is the Morris water task, and the most commonly studied behavioral correlate of hippocampal neurons is the spatial specificity of place cells. Despite decades of intensive research, it is not completely understood how animals solve the water task and how place cells generate their spatially specific firing fields. Based on early work, it has become the accepted wisdom in the general neuroscience community that distal spatial cues are the primary sources of information used by animals to solve the water task (and similar spatial tasks) and by place cells to generate their spatial specificity. More recent research, along with earlier studies that were overshadowed by the emphasis on distal cues, put this common view into question by demonstrating primary influences of local cues and local boundaries on spatial behavior and place-cell firing. This paper first reviews the historical underpinnings of the "standard" view from a behavioral perspective, and then reviews newer results demonstrating that an animal's behavior in such spatial tasks is more strongly controlled by a local-apparatus frame of reference than by distal landmarks. The paper then reviews similar findings from the literature on the neurophysiological correlates of place cells and other spatially correlated cells from related brain areas. A model is proposed by which distal cues primarily set the orientation of the animal's internal spatial coordinate system, via the head direction cell system, whereas local cues and apparatus boundaries primarily set the translation and scale of that coordinate system.

Figure 1

Diagram of the apparatus used by Tolman, Ritchie, and Kalish (238) to contrast response and place learning. Two start locations (S₁ and S₂) and two goal locations (food boxes, F₁ and F₂) were used throughout training. Start locations and/or food locations were changed from trial to trial so that performance of particular responses (left or right) or navigation to a particular place was reinforced. For example, a rat trained to perform a right turn response (Response Group) would be trained to navigate to F₁ when released from S₁, and to navigate to F₂ from S₂, as indicated by dotted lines in the left panel. This strategy required the animals to navigate to different places on each trial. A rat trained to navigate to location F₁ (Place Group) would be trained to navigate to F₁ when released from either S₁ or S₂, as indicated by the dotted lines in the right panel. This strategy required animals to perform different motor responses from trial to trial.

Figure 2

Diagram of the T-maze locations and orientations used by Blodgett, McCutchan, and Mathews (12) to contrast place, response, and direction learning. The T-maze was translated and/or rotated from trial to trial and could occupy four locations in the room. The top left panel illustrates the T-maze in one location (solid lines) and the other three possible positions (dashed lines). There were four possible start locations (S₁, S₂, S₃, and S₄) and three possible reinforcement locations (F₁, F₂, and F₃). The maze position and orientation and the start and/or food locations were changed from trial to trial so that particular responses, navigation to a particular place, or navigation in a particular direction were reinforced. Examples of specific conditions among three groups are illustrated. One condition from the place group (top right) involved translating the maze between two positions from trial to trial. Rats were released from S₁ or S₃, and reinforced for navigating to F₂, requiring different responses and navigation in opposite directions. One condition from the direction group (bottom left) involved translating and rotating the maze from trial to trial. When released from S₁, navigation to F₁ was reinforced. When released at S₄, navigation to F₂ was reinforced. This required the animal to navigate in the same direction in the room and apparatus, but to navigate to different locations and perform different responses from trial to trial. One condition from the response group (bottom right) involved translating and rotating the maze from trial to trial. If released at S₁, navigation to F₁ was reinforced and when released from S₂ navigation to F₂ was reinforced. This required the animal to perform the same response (left turn) while navigating to different places and in different directions from trial to trial. Rats learned the direction and response tasks more readily than the place task.

Figure 3

Diagrams of an eight-arm radial maze illustrating the types of distal cue manipulations used by Suzuki, Augineros, and Black (228). Reinforced (baited) arms are indicated by +. Four distinct distal cues are represented by circles. After training (top left), the critical manipulations involved rotating the distal cues (top right), transposing the distal cues (bottom left), and systematically removing (deleting) distal cues (bottom right).

Figure 4

Layout of the room used in the experiments of Hamilton and colleagues showing the two locations where the pool could be positioned in the room. Pool positions 1 (black) and 2 (gray) represent the same pool positions used by Hamilton et al. (72-74) and Akers et al. (1; 2). Each pool position was separated by 75 cm (the radius of the pool). During training, the platform was typically located at the black rectangle labeled B; the gray rectangles (locations A and C) mark comparison locations (relative/opposite) used for probe trial analyses. The small circles (SW, SE, NW, and NE) represent release points used during training trials and the rectangles (north-most and south-most points) represent release points used during no-platform probe trials. Black indicates release points for pool position 1 and gray indicates release points used for pool position. (Reproduced from Akers KG, Candelaria FT, and Hamilton DA. Preweanling rats solve the Morris water task via directional navigation. Behavioral Neuroscience 121: 1426-1430, 2007.)

Figure 5

Diagram of the Direction, Place 1, and Place 2 conditions utilized by Skinner et al. (211). The squares represent the boundaries of a single square open field that was positioned in one of two locations in the room. The grey circles indicate potential locations of reinforcement. In each condition, the apparatus was moved (translated) between the two possible positions for each trial. The dotted lines indicate the paths from the release point to the reinforced location for each apparatus position. In the Direction condition, navigation in the direction of reinforcement within the apparatus and room was reinforced. In the Place 1 and Place 2 conditions navigation to a particular location in the room regardless of the apparatus position was reinforced. Rats learned the Direction and Place 2 tasks much faster than the Place 1 task.

Figure 6

Place cells maintain spatial selectivity in the absence of controlling distal cues. Three cells are shown during a spatial learning task on a plus maze (170). The top row shows the firing locations of the cells as a contour map when controlled distal cues were present on the curtains surrounding the maze (Perceptual trials). The cues were rotated between trials, and the rats were required to visit a goal arm that rotated along with the cues. The contour maps were aligned such that the variable goal arm was placed at the top of the figure, and the place fields are seen to be sharply tuned relative to the goal arm and the distal cues. The bottom row shows the firing locations when the distal cues were removed after the rat was placed on the maze (Memory trials). The sharply defined contour maps show that the cells maintained place fields under these conditions, and that the locations of the place fields were consistent with the locations when the distal cues were present. (Reproduced with permission from O'Keefe J and Speakman A. Single unit activity in the rat hippocampus during a spatial memory task. Exp Brain Res 68: 1-27, 1987.)

Figure 7

Place cells are more strongly controlled by local, platform cues than by distal cues in a spatial task. (Top) Protocol. Rats were trained for 8-12 sessions to go an unmarked location (orange square) on a square platform when a tone sounded, in order to release food from an overhead dispenser. On the first test day, rats performed the task with the platform in the standard location in the room, and then the platform was shifted in the room. Analogous to Hamilton et al. (74), rats had to choose to go to the goal location either in room-based or platform-based reference frames. The platform shift manipulation was repeated after 3 more days of training in the standard condition. (b) Place field results. After the first Shift manipulation (left pie chart), the majority of place fields remapped between the standard and shift conditions (gray). Of the cells that did not remap, fields were almost 6x more likely to shift with the platform (red) than to remain in the room reference frame (blue). Similar results were obtained in the second shift manipulation (right pie chart). Modified from Siegel et al. (208).

Figure 8

(A) Affine transformations of two coordinate systems. (B) The grid cells (red) can be considered an internal coordinate system that must be aligned with the coordinate system of the external world (green, representing a circular platform in a square room with distal cues on the walls) via translation, rotation, dilation, or a combination.

Figure 9

Flow chart showing hypothesized relationships between local cues, distal cues, and physiological cell types in the control of neural spatial representations. Head direction cells and grid cells are presumed to be updated primarily by idiothetically driven, path integration mechanisms (not reviewed here). Distal cues are primarily used to orient the head direction cell system, which is an internally coherent representation that integrates angular velocity signals to continuously update a directional signal. The head direction cells set the orientation of the grid cells of the MEC and subicular regions and also disambiguate apparatus and environmental boundaries (i.e., disambiguate the north wall from the east wall) so that boundary cells can fire selectively at a single boundary. The boundary cells are used to phase-align and scale the grids to the environment (they may also directly influence place cells). Place cells derive their spatial firing from grid cells, and thus they are controlled by the distal cues and apparatus boundaries via these input cells. The major functions of place cells are to create context-specific, spatial representations (3; 160; 214); to perform the mnemonic operations of pattern completion and pattern separation (138; 146; 171; 191); and to store representations of event sequences (45; 59; 104; 124; 175; 210). Information about specific local and distal cues can be incorporated into the place-cell representation via the lateral entorhinal cortex, and then fed back to the grid cells via backprojections from the hippocampus to aid in map alignment between the areas and to send conjunctive item + context representations back to the neocortex for potential use as an organizing framework for episodic memory (115; 121; 134; 169).