Ensemble codes involving hippocampal neurons are at risk during delayed performance tests
-
Figure 1Performance in the DNMS task. (A) Delay-dependent DNMS performance is indicated as mean (± SEM) percent of correct trials per session (100–200 trials), averaged across 7 different animals. Trials were grouped and plotted by 5-sec intervals across the 1- to 30-sec delay intervals. Horizontal dashed lines indicate a 6% delay-dependent decrement in DNMS performance contributed by trials with delays of <15 sec (left of vertical dashed line), and an additional 14% delay-dependent decrement contributed by trials with delays ≥15 sec. (B) Effect of proactive interference on DNMS performance. The data shown in A were sorted, on the basis of preceding trial type (which was determined with respect to lever press position in the sample phase of the trial), into the Same category or the Different category. Mean (± SEM) percent of correct DNMS performance by delay is plotted separately for each of the sorted trial types. The dashed line is the overall mean DNMS performance curve from A. Proactive interference is demonstrated by the consistent 6–8% decrease in DNMS performance at each delay (relative to Same trials) for Different trials.
-
Figure 2Influence of preceding trial delay and proactive interference. DNMS performance represented in Fig. 1 was analyzed according to the preceding trial type (Same or Different) and the length of the delay interval in the preceding trial. (A) Same plot that in Fig. 1 B except only DNMS trials that were preceded by trials with short delays (<15 sec) were included. Each point represents mean (± SEM) DNMS performance from the same seven animals represented in Fig. 1. (B) Same graph as A except for preceding trials with long delays (≥15 sec). Note increase in proactive interference effect Same versus Different. Dashed line represents the mean of both Same and Different curves irrespective of preceding trial type. (C) Same data as in A and B plotted to show separation between Same versus Different pairs of trials as a function of the duration of delay on the previous trial. Each point depicts mean (± SEM) DNMS performance across trial blocks with 1- to 30-sec delays in the current trial (n = 7 animals).
-
Figure 3Illustration of hippocampal ensemble activity in correct and error trials in the DNMS task. (A) Mean ensemble firing pattern for the right nonmatch correct trial type is displayed as a three-dimensional firing surface (larger surface, Upper left) constructed from multiple perievent histograms (100 trials each) for 10 different hippocampal neurons recorded ±1.5 sec before and after a right correct lever press (R and thick white line). All neurons were recorded simultaneously from the electrode array shown at the top of the figure. Display of recording sites from CA1 and CA3 alternate along the electrode location axis (Elect. Locn.) of the three-dimensional surface plot to preserve anatomic relationships. S, Septal location; T, temporal location. Color contour plot beneath the firing surface is a two-dimensional projection of the surface activity. Color bar indicates firing rate (1–8 Hz). The three smaller surface patterns (Lower left and Upper and Lower right) depict the same ensemble firing rates “adjusted” by linear discriminant functions (derived by canonical discriminant analysis) corresponding to task phase (root 1), error trials (root 2) and lever position (root 3), as described (21). Note that activity associated with roots 1 and 3 occupy different regions of the surface, and that root 2 shows very little activity. (B) Mean firing pattern for a right “match” error trial for the same ensemble shown in A. Extracted root firing patterns for the same roots are displayed in the Lower left and Upper and Lower right as in A. Adjusted firing surfaces for task phase (nonmatch) and lever position (right) are similar to correct right lever (roots 1 and 3). Note the increase in activity associated with root 2 which is the source of the difference in overall mean ensemble firing (above) between correct and error trials.
-
Figure 4Comparison of within-trial and between-trial interactions of roots 2 and 3 on correct versus error trials. Scatter diagrams show root 1 versus root 3 and root 1 versus root 2 ensemble discriminant scores as two-dimensional plots. Each point reflects the intersection of two ensemble discriminant scores, one for each root for a single behavioral event (i.e., upper left cluster of red points on the correct trial plot reflects single trial scores during left sample lever presses). Colors indicate events related to the same trial types (blue, right sample trials; red, left sample trials). In correct trials, the same color clusters are diagonally opposite, reflecting opposite lever positions in the two different phases of the task as dictated by the nonmatch contingency. (Top Left) Correct trial: Normal distribution of root 3 discriminant scores in both types of correct DNMS trials. (Middle Left) Root 2 scores (plotted against root 1) on the same trials. There was no separation in scores with respect to lever position (horizontal axis) in either phase of the task on correct trials. (Bottom Left) Same as the top plot for root 3 except that discriminant scores are plotted for the following (Middle) trial as a function of the previous trial type (Top). Mixture of colors indicates the lack of relationship between encoding on the previous trial (Top) and encoding on the next trial (Bottom). (Top Right) Error (LDE) Trial: Illustration of (sequential) error trial “cascade.” (Top Right) Same as Top left plot except that ensemble scores are from LDE trials. Note weak encoding in root 3 during the sample phase illustrated by the mixture of colors and the lack of separation into distinct color-coded clusters. Nonmatch phase scores reflect the “match” error response. Root 2 scores on the same LDE trials show distinctly separated clusters, but only in the nonmatch phase, which correspond to the same lever position as root 3 scores in the nonmatch phase of the same trial (orange arrow). (Bottom Right) Miscode trial: Distinctly separated color-coded clusters of root 3 ensemble scores consistently reflect that same root 2 information (Middle Right) in the sample phase on the next (Mis) trial following LDE trials. The same information is encoded in the sample phase on the next (Mis) trial as in the preceding nonmatch phase regardless of the current trial type. This illustrates “carryover” of information (green arrow) between trials as the basis for proactive interference. The reversal of color clusters along the diagonal indicates that the nonmatch decision rule was applied to the miscoded sample information that caused a match error with respect to the nonmatch contingency.
-
Figure 5Schematic illustration of dynamic properties of hippocampal memory. Opposing influences on DNMS performance, strength of hippocampal ensemble encoding (solid lines) and proactive interference (dotted line). The horizontal axis represents the relationship between strength of encoding and DNMS delay intervals. At short delays, encoding can be relatively weak and resistant to interference, but only for a certain delay duration. At long delays, only strong encoding prevents interference across the delay and results in correct performance. The intersection between two curves defines the critical point at which strength of encoding (vertical axis) is able to counter proactive influences (dotted line). Note that the time constant of decay also varies with strength of encoding. The proactive interference curve was purposely plotted with the same relative decay as the strong encoding curve, and represents encoding from the previous trial. The dynamic range (strong to weak) of hippocampal encoding is illustrated by the three solid curves and the oblique arrow. Encoding in the absence of a hippocampus is indicated by the dashed line.
Footnotes
- Copyright © 1996, The National Academy of Sciences of the USA
