Our research focuses on studying the psychological and neural processes involved in rat cognition with a special interest in how rats organize complex behavior through time in what is known as serial pattern learning. Our method for studying serial pattern learning in rats is a functional analogue of human pattern learning tasks that require subjects to learn to choose items from an array in the proper sequential order .  Rats learn to press levers in an array in the proper sequential order.  Our current procedure involves training rats in an octagonal Plexiglas box (see Figure 1) equipped with a retractable lever mounted on each wall.  [The levers are designated Levers 1 through 8 in clockwise order.  It should be noted that Lever 8 is adjacent to Lever 1.] All levers are presented at the beginning of each trial and the rat may press any of the 8 levers.  If the correct lever is pressed, the rats receive reward.  If an incorrect lever is pressed, all levers except the correct lever are removed from the box and the animal is not reinforced until the correct response is emitted.  This method is easily learned by the rat without pretraining procedures other than leverpress shaping, provides the rat with a continuous (circular) stimulus array, and allows us to record response latency (in addition to accuracy measures) on a trial-by-trial basis while the rat performs the task at its own pace.  Rats have been trained with up to fifty 24-element patterns per daily session (e.g., Fountain et al., 2000). Our method is an improvement over earlier methods used with rats because it allows us to study how rats learn long, elaborate serial patterns and because it provides measures of correct-response rates, error rates, and "intrusion" rates (i.e., the number of specific kinds of errors produced at particular locations in the pattern) on a trial-by-trial basis throughout the serial pattern.

THE SYMBOLIC ORGANIZATION OF SEQUENTIAL BEHAVIOR

            Hierarchically organized serial patterns.  One prediction from the rule-learning view is that a highly organized, hierarchically structured sequence should be easier to learn than a sequence having little or no higher-order structure.  We designed several studies to explore whether pattern structure would determine the ease or difficulty of learning long and elaborate patterns.  

            In one experiment (Fountain et al., 1995b), we tested whether pattern structure described as “runs” (e.g., 1-2-3-4-5-…) or “trills” (e.g., 1-2-1-2-1-…) would determine the ease or difficulty of anticipating a final sequence item that either conformed to the implied structure of the sequence or violated pattern structure.  Rats received patterns having either perfect structure or one sequence element (the last in the series) that violated an otherwise perfect structure:

"Perfect Runs"            123 234 345 456 567 678 781 812 ...

"Violation Runs"          123 234 345 456 567 678 781 818 ... (violation element indicated)

"Perfect Trills"             121 232 343 454 565 676 787 818 ...

"Violation Trills"           121 232 343 454 565 676 787 812 ... (violation element indicated)

A 1-s ITI was used except where spaces indicate 3-s phrasing cues.  [Note once again that Lever 1 is immediately to the right of Lever 8, so that, for example, a 6-7-8-1-2 sequence would be a quite natural "run" series.]  As shown in Figure 2, high error rates were observed in acquisition on the violation trial (the last trial of the pattern) for both Violation Runs and Violation Trills patterns, despite the fact that one view might predict generalization of associations from other parts of the pattern should have predisposed animals to learn the violation patterns easily.  For example, in the Violation Trills pattern a correct response on lever 1 should always predict that the next response should be to lever 2, yet rats had great difficulty learning to respond on Lever 2 on the last trial of the pattern but not the second trial of the pattern.  No comparable errors were observed for the Perfect Runs or Perfect Trills patterns. An alternative view is that rats learned about the highly repetitive structure of the sequence that resulted in highly repetitive patterns of response to learn the sequence even when doing so produced errors at the violation element, even though these errors might have been avoided by adopting another strategy.  CF1 mice show the same pattern of results as rats when learning the perfect and violation runs patterns described here(Fountain et al., 1999).           

            In another set of studies, we tested whether pattern structure would determine the ease or difficulty of pattern learning by developing patterns with hierarchical structure, then reordering chunks of the pattern to produce “linear” structure, that is, a sequence of unrelated chunks.  The Hierarchical (H) and Linear (L) patterns were:

H Pattern: 123 234 345 456 567 876 765 654 543 432...

L Pattern: 123 234 543 456 567 876 765 654 345 432...

For both groups, the digits indicate the clockwise position of the correct response on successive trials and spaces indicate brief pauses. 

            The completely nested H pattern is described by a simple hierarchical rule structure: elements within 3-element chunks are related by first-order rules, chunks within the first and second halves of the pattern, respectively, are related to each other by second-order rules, and the first half of the pattern is related to the second half of the pattern by a third-order rule.  A formal description of this pattern is (M(T₊₁⁴(T₊₁²(1)))),

where “1” refers to the starting lever, T_+n represents a “transpose” rule (i.e., to move n units in the indicated direction, where + indicates clockwise), M represents a “mirror image” rule, and superscripts reflect the number of repeated applications of the rule that are required.  Because of the nested structural organization, the second-order T₊₁ rule applies a “+1” rule to each item in the first chunk to generate the second chunk, and so on.  The third-order M rule produces a “mirror image” (a more complex form of a “reverse” rule) of the first half of the pattern to generate the entire second half of the pattern. 

            The incompletely nested L pattern was generated by exchanging the two underlined 3-element chunks in the H pattern.  In so doing, however, it should be noted that all pairwise associations were maintained; rats were always required to press a lever immediately to the left or right of the last correct response in both patterns, and the number of transitions from a given lever to any other was conserved across patterns.  In this structure, elements within any chunk are related by a rule, but chunks are not related to each other in any systematic way.  A formal description of this pattern is (T₊₁(T₊₁²(1))) - (T_-1²(5)) - (T₊₁(T₊₁²(4))) - (T_-1²(T_-1²(8))) - (T₊₁²(3)) - (T_-1²(4)), where T_+n and T_-n represent rules indicating to “transpose” clockwise and counter-clockwise, respectively.  Dashed lines indicate connections that must be learned by non-hierarchical rules or non-rule-based associations between rule-governed chunks or chunk subsets.  Note that this complicated structure resulted from changing the serial positions of only 4 of 30 pattern elements compared to the H pattern. 

            The results showed that, for rats, pattern complexity predicted pattern learning difficulty (Fountain & Rowan, 1995a).  The formally simpler H pattern was easier to learn than the formally more complex L pattern.  In addition, rats in H were sensitive to the hierarchical structure of their pattern.  For rats, as in humans, in the H pattern groups, the difficulty of learning to respond appropriately on any trial was a function of the hierarchical level of the rule required to predict the item.  Figure 3 shows rats' group mean element-by-element error rates for Week 1 of the experiment for the H group (top panel) and the L group (bottom panel).  As shown in Figure 3, during Week 1, rats produced significantly more errors on the first trial of Chunks 1 and 6 than on all other trials.  These trials corresponded to the highest-order rule transitions in the pattern structure (i.e., third-order rule transitions).  Fewer errors were observed on the first trial of other chunks; these trials corresponded to second-order rule transitions.  The fewest errors occurred within chunks, where trials corresponded to first-order rule transitions.  Thus, in the completely hierarchical pattern, the difficulty of learning to respond appropriately on any trial was a function of the hierarchical level of the rule required to predict the item.

            Rats in L did not show the 3-level hierarchical pattern of errors observed for H rats.  L rats found trials within chunks easier than the first trial of each chunk, but their response to the first trial of chunks was disorganized.  That is, L rats, unlike H rats, showed no differential response for the first trial of Chunks 1 and 6 (corresponding to third-order rule transitions in the completely nested pattern) versus the first trials of other chunks  (corresponding to second-order  rule  transitions  in  the  completely nested pattern).   However, error rates for the second and third elements of each chunk (with the exception of the second element of Chunks 3 and 10) were significantly lower than for the first element of each chunk.  These results support the view that L rats were sensitive to the actual pattern structure; they recognized that elements within a given chunk were related by a single rule, but that chunks were somewhat haphazardly arranged. 

            In the hierarchical versus linear structure experiment just described, rats demonstrated sensitivity to multi-level hierarchically-organized pattern structure.  Rats found learning completely nested hierarchical patterns easier than learning less organized patterns even when pairwise associations and pattern length were conserved across patterns.  In another study from the same series (Fountain et al., 1995a), a 3-level hierarchy was easier to learn than a 4-level hierarchy when pattern length was conserved across patterns.  As a rule, then, pattern complexity was a better predictor of acquisition difficulty in these studies than was pattern length.  These acquisition results alone are strong evidence that pattern organization, that is, pattern complexity, was the primary determinant of pattern difficulty, as argued by a rule-learning view of sequential learning. 

Interleaved serial patterns.  One question of significance for animal sequential learning research is whether animals are constrained to learn sequences on the basis of pair-wise associations between successive elements, for example, as in chaining (Skinner, 1934).  A significant body of evidence suggests that animals are able to be more flexible in representing sequential events, conceivably by coding hierarchical representations characterized by relations for nonadjacent events.  The mechanisms of learning involving nonadjacent events are not well-understood.  Terrace(1987), for example, indicated that little evidence existed that animals are able to spontaneously reorganize sequentially-presented items into chunks not presented by the experimenter.  As Terrace noted, such processes are readily observed in human free-recall (Tulving, 1962).  Additionally, it should be noted that chunking of nonadjacent items in human serial-pattern learning has been studied extensively using patterns of letters and digits (e.g., Hersh, 1974).  We have previously shown that rats, when presented a sequence of reward quantities, can spontaneously sort quantities from nonadjacent serial positions into chunks to facilitate learning (Fountain & Annau, 1984). A comparable strategy in humans would be to learn the pattern 2555455565558 by sorting pattern elements into 555 chunks and a 2468 chunk.  Other work also supports the view that rats have this capacity.  For example, Capaldi and Miller (1988) have shown that rats can keep count of different kinds of rewards by chunking nonadjacent items in series into different food categories. In two recent studies in our laboratory (Fountain, Rowan, & Benson, Jr., 1999), rats learned either a structured (ST) or unstructured (UNST) sequence interleaved with elements of a repeating (R) sequence in one experiment or an alternation (A) sequence in another experiment. The question was whether rats would learn the interleaved subpatterns at different rates as a function of subpattern complexity. 

The first experiment sought to determine whether rats would show signs of being sensitive to the organization of nonadjacent items from interleaved subpatterns when one subpattern was a composed of simple, repeating element and the second subpattern was either highly structured or not.  For rats in the Structured (ST) subpattern condition, a 123 234 345 456 567 subpattern was interleaved with a Repeating (R) subpattern, 888 888 888 888 888, resulting in the ST-R pattern that rats were required to learn: 

182838 283848 384858 485868 586878.

For rats in the Unstructured (UNST) subpattern condition, a 153 236 345 426 547 subpattern was interleaved with the same R subpattern to create the UNST-R pattern in the same manner.  For both patterns, integers represent the clockwise position of correct levers in the octagonal chamber on successive trials and spaces represent pauses that served as phrasing cues. 

Acquisition of the interleaved structured pattern (i.e., ST-R) was significantly faster than for the interleaved unstructured pattern (i.e., UNST-R).  The unstructured pattern was generated by exchanging only two pairs of elements in the structured pattern, as described above.  In so doing, however, all pair‑wise associations in the interleaved patterns were maintained because all of the relocated items were preceded by “8” trials.  Nevertheless, Figure 4 shows that the effects of disrupting pattern structure were apparent throughout the pattern.  This was so even in the third (middle) chunk that was not altered in producing the unstructured pattern; rats found this chunk, 384858, harder to learn in the context of the UNST-R pattern than in the ST-R pattern. 

In the second experiment, rats learned two interleaved sequences where both were created from sets composed of more than one element.  As before, longer patterns were composed of two interleaved subpatterns; either a structured or unstructured subpattern was interleaved with a subpattern of two alternating elements.  For one group of rats, the structured (ST) subpattern, 1 2 3 4 5 6, was interleaved with the alternating (A) subpattern, 7 8 7 8 7 8 to create the ST-A pattern.  For another group of rats, the unstructured (UNST) subpattern, 1 5 3 4 2 6, was likewise interleaved with the same alternating subpattern to produce the UNST-A pattern.  Note that the unstructured subpattern was generated by exchanging two items of the structured subpattern.  Rats learned the subpatterns of their interleaved patterns at different rates both within and between pattern groups.  As predicted based on subpattern structure, in the case of the UNST-A pattern, the A subpattern was acquired faster than the UNST subpattern.  The A subpattern would be expected to be acquired faster because it is formally simple whereas the UNST subpattern has little structure.  Based on similar reasoning, it was expected that the ST subpattern should be easier to learn than the UNST subpattern, and this result was obtained.  In the case of the ST-A pattern, since both subpatterns were structured, it might be difficult to predict in advance based on subpattern structure alone whether rats should find either the ST or A subpattern easier to learn than the other.  However, if structural complexity is equated (i.e., if the same number of rules are needed to describe subpattern structure), rats might show the same predisposition that humans do (Kotovsky & Simon, 1973) to detect repeating items before other structural features of patterns.  In fact, evidence for the latter assertion was obtained in this experiment.  Rats in the ST-A pattern group showed better acquisition for A with its repeating “7” and “8” elements than ST subpatterns of their interleaved pattern despite the fact that both ST and A subpatterns have simple structure that can be described by a single rule (viz., a “+1” rule for the 123456 ST subpattern versus an “alternate” rule for the 787878 A subpattern).  The results of differential acquisition of ST and UNST subpatterns support the notion that accurate performance on these interleaved subpatterns was dependent on a mnemonic representation characterized by relations for nonadjacent events.  The results indicate that rats are sensitive to the organization of nonadjacent elements in serial patterns and that they can detect and sort structural relationships in interleaved patterns.  Pattern and subpattern structure appear to drive how animals sort, chunk together, and represent nonadjacent pattern elements that are related by common rules or features.

THE NEURAL ORGANIZATION OF SEQUENTIAL BEHAVIOR

Several recent studies have begun to explore the neurobiological basis of serial learning and memory in rats.  They suggest how serial learning processes might be integrated into more general neurobiologically based models of learning and memory.  In one study, Olton, Shapiro, and Hulse (1984) tested rats' sequential memory.  Four quantities of food 14, 7, 1, and 0 pellets of food were placed in the goal boxes at the ends of the four arms of a plus maze.  Rats were allowed to choose freely among the arms and over a period of days learned to choose the large quantities of food first and the smallest quantities of food last.  Thus rats had encoded a stimulus alphabet of four elements and had also learned an orderly response to the four elements as they were distributed in four spatial locations represented by the four goal boxes of the plus maze.  Once rats had learned this task, they were given lesions of the fimbria‑fornix (FFx), the major extrinsic pathway of the hippocampus.  Subsequent testing showed that the rats could remember to search out the quantities in the order in which they had previously learned them.  In other words, rats would go first to 14, then to 7, then to 1, then to the arm containing 0 pellets of food.  However, if before given a free choice in the maze the rats were required to sample one or more of the quantities out of order in a forced choice procedure, they subsequently failed to remember having sampled the quantities when they were tested in the free choice test.  For example, if a rat were allowed to retrieve the 1‑pellet quantity before being given the free choice test, when the rat was allowed to make a free choice among the arms, the rat went first to 14, then to 7, then to 1 just as if it had never sampled the 1‑pellet quantity in the preexposure.  These kinds of mistakes indicated that rats had no memory for previously sampling food quantities from the maze before the free choice.  However, subsequent tests showed that rats could remember elements sampled in preexposure as long as the quantities were received in the order in which they had originally learned them.  For example, if a rat was first preexposed to the arms containing 14 and 7 pellets of food, when given a free choice, the rat would not run down the arms previously containing 14 and 7, but would go immediately to 1 and then 0.  Thus FFx‑lesioned rats could remember elements presented in order, but could not remember elements presented out of the order originally learned.  These results are consistent with Olton et al.’s (1984) interpretation  that the impairment produced by FFx lesions was an impairment of working memory, but not reference memory.  They are also consistent with Eichenbaum et al.’s (1992) idea that hippocampus mediates representational flexibility, the ability to use declarative memories flexibly in new configurations, situations, or tasks.  This idea predicts that FFx-lesioned rats should be inflexible in their use of sequential information learned before surgery, and therefore they should be impaired in their ability to respond to probe situations where patterns differed from the training pattern.  Yet another interpretation of Olton et al.’s (1984) results is consistent with the view that hippocampus mediates item associative processes in SPL, but not rule-induction and memory for pattern structure (Fountain, Schenk, & Annau, 1985).  According to this latter RL view, FFx lesions spare information about pattern structure that mediates responding according to the rule learned in training prior to surgery.

A second experiment shows that hippocampal lesions produce results predicted by the RL view of serial-pattern learning if one views rule-induction as a process potentially dissociable from item association formation.  Fountain, Schenk, and Annau (1985) trained rats with long monotonic and nonmonotonic patterns created from quantities of brain‑stimulation reward.  The monotonic pattern was 18‑10‑6‑3‑1‑0 and the nonmonotonic pattern was 18‑1‑3‑6‑10‑0.  Prior to training, one group of rats was exposed to trimethyltin (TMT), a neurotoxic organometal which produces damage in the limbic system, primarily in the hippocampus.  TMT‑exposed rats learned the formally simple monotonic 18‑10‑6‑3‑1‑0 pattern as fast as control rats, but learned the nonmonotonic 18‑1‑3‑6‑10‑0 pattern slower than controls. 

According to Olton et al.(1984), hippocampal damage should have impaired working memory, but for both groups of rats reference memory should have been intact.  However, the results indicate differential impairment for two different kinds of patterns despite the fact that reference memory should have been intact and available for learning both kinds of patterns.  Similarly, according to Eichenbaum et al. (1992), flexible declarative memory should have been impaired, but inflexible nondeclarative memory should have been spared.  This view suggests that since learning for both groups involved learning to respond to a consistent repeating pattern, learning for both should have been spared following TMT damage to the hippocampus.  The results fit best with the notion that rule-induction processes were spared following TMT damage, whereas item associative processes were impaired by TMT damage.  These, along with other data, suggested that item association formation is a hippocampal-dependent process, whereas rule induction is not.  A dissociation of this sort may be difficult to model with SPAM, though Metcalfe (1993)  has succeeded in simulating characteristics of Korsakoff's amnesia using a closely related model, CHARM.

In a third series of experiments (Fountain & Rowan, 2000), we sought additional evidence for this distinction between item associative and rule induction processes.  In the first study of the series, rats were trained on two patterns, one which was structurally “perfect” and a second virtually identical to the first, but containing a single element that violated the otherwise simple structure.  The Perfect (P) and Violation (V) patterns were:

P Pattern:  123 234 345 456 567 678 781 812

V Pattern:  123 234 345 456 567 678 781 818

As before, the digits indicate the reinforced lever for successive trials.  The last “8” item of the V pattern (underlined) was the violation element.  Rats from one group for each pattern condition were injected with MK-801 daily before training.  MK-801 is a systemically administered NMDA receptor antagonist that blocks neuronal plasticity, known as  long-term potentiation,

in the hippocampus.  It is thought that MK-801 should impair any hippocampal-dependent learning.  As shown in Figure 5, MK-801 had little effect on learning to respond to rule-based items within chunks.  However, it did impair responding at points where rules were violated, namely, on the first trial of each new chunk and, most dramatically, for the violation element.  Although rats showed no signs of learning to respond to the violation element, throughout the experiment they produced rule-based errors on the violation trial by responding “2” instead of “8” at the end of the sequence (Fountain & Rowan, 2000).  The results are strong evidence that hippocampal damage impaired learning the item associations necessary to track violations of pattern structure while sparing the rule induction processes necessary to induce pattern structure and extrapolate the sequence on the violation trial.

In a later study in the same series (Fountain & Rowan, 2000), we examined the role of hippocampus when new serial pattern information is added to old.  Rats were first trained to a high criterion on a pattern consisting of the first 7 chunks of the P pattern above: 123 234 345 456 567 678 781.  After rats learned the pattern, they were transferred to one of two new patterns that contained all elements of the first pattern and an additional chunk of three additional elements.  The three added elements were either structurally consistent with the first pattern (viz., 812), making it structurally “perfect” (P), or they contained a violation (V) of the pattern structure learned in training (viz., 818).  On the day of transfer, half the rats were injected with MK-801 to determine the effects of hippocampal dysfunction on the rats’ ability to integrate structurally consistent or inconsistent new information with an already learned pattern.

As shown in Figure 6, when a structurally consistent chunk was added in the P transfer, the effects of MK-801 were very similar to the effects of the drug on acquisition (Fountain & Rowan, 2000).  That is, the drug produced a selective decrease in the animals' accuracy on the first elements of each chunk of the original pattern, but produced virtually no change in accuracy on the remaining two elements of the 3-element chunks.  The most interesting result occurred when a structurally inconsistent chunk was added in the V transfer.  As shown in Figure 6, although saline controls showed difficulty in learning the new chunk, there was little effect on the rest of the pattern.    However, MK-801 dramatically disrupted performance for elements both in the new chunk to be learned and throughout the rest of the pattern (Fountain & Rowan, 2000).  When this effect is compared to the effects of MK-801 in the P transfer, the effect can only be accounted for by the addition of the terminal violation element.  One interpretation of these results is that adding new information to a pattern representation is possible under MK-801, but only if the information is consistent with pattern structure that has already been encoded.  In fact, this initial evidence indicates that for rats with hippocampal impairment, new information that is structurally inconsistent can disrupt previously well-learned response patterns.  This suggests that, in intact animals, nonhippocampal systems mediate rule induction whereas hippocampus may play a role in the successful integration of new rule-inconsistent SPL information with already   encoded   information  about  pattern  structure.     These  ideas  are reminiscent of the distinction between flexible and inflexible memory processes proposed by Eichenbaum et al.(1992), but our MK-801 results suggest that what constitutes “representational flexibility” is far from resolved.  Under MK-801, rats were able to add a rule-consistent chunk to their already learned pattern with relatively little difficulty, but not a rule-inconsistent chunk.