Louis D. Matzel

The Comparator Hypothesis: A Response Rule for The Expression of Associations

Publisher Summary This chapter describes the potential explanatory power of a specific response rule and its implications for models of acquisition. This response rule is called the “comparator hypothesis.” It was originally inspired by Rescorlas contingency theory. Rescorla noted that if the number and frequency of conditioned stimulus–unconditioned stimulus (CS–US) pairings are held constant, unsignaled presentations of the US during training attenuate conditioned responding. This observation complemented the long recognized fact that the delivery of nonreinforced presentations of the CS during training also attenuates conditioned responding. The symmetry of the two findings prompted Rescorla to propose that during training, subjects inferred both the probability of the US in the presence of the CS and the probability of the US in the absence of the CS and they then established a CS–US association based upon a comparison of these quantities. The comparator hypothesis is a qualitative response rule, which, in principle, can complement any model of acquisition.

Acquisition of conditioned associations in Hermissenda : additive effects of contiguity and the forward interstimulus interval

Conditioned suppression of photokinesis by the marine mollusc Hermissenda was examined in 3 experiments. In each experiment, groups of animals received light (the conditioned stimulus, CS) that was paired with high-speed orbital rotation (the unconditioned stimulus, UCS), light and rotation explicitly unpaired, or no exposure to these stimuli. Twenty-four hours after training, all animals were tested for suppression of photokinesis in the presence of the light. To establish the effectiveness of our conditioning procedure, in Experiment 1 individual groups of animals received either 50, 100, or 150 CS-UCS pairings. Fifty pairings resulted in a marginal suppression of photokinesis, whereas 100 and 150 pairings produced strong suppression. In Experiment 2, the delay between CS onset and UCS onset was varied between 1 and 10 s. The 10-s interstimulus interval (ISI) did not support conditioning, whereas 1-s and 2-s ISIs were effective. As predicted by the current understanding of Hermissendas neural network, in Experiment 3 it was found that CS-UCS pairings in which the CS preceded the onset of the UCS and terminated with the offset of the UCS evoked stronger conditioned suppression than either a CS that preceded the UCS and terminated with its onset or a CS that was paired in simultaneous compound with the UCS. This result indicates that CS-UCS contiguity as well as the forward ISI act additively to establish the CS-UCS association. In none of the 3 experiments were any differences observed between groups that were untreated and that received the CS and UCS unpaired. In total, these experiments suggest strong similarities in the temporal characteristics of associative learning in Hermissenda and vertebrate species.

Commentary |[mdash]| Reconsolidation: Memory involves farmore than 'consolidation'

The observation that retrieval returns a stable memory into a labile state cannot be readily explained by any simple version of consolidation theory. This finding has been interpreted as evidence for the need to reconsolidate a memory after reactivating it. However, as we discuss in this commentary, other behavioural observations indicate that even this modification to consolidation theory may be insufficient to describe the dynamic properties of memory.

Pavlovian conditioning of distinct components of Hermissenda's responses to rotation

In two experiments with the nudibranch mollusk Hermissenda, distinct characteristics of conditioned and unconditioned responses to high-speed orbital rotation were examined. In Experiment 1, two principle unconditioned responses to rotation were identified, namely, reduced rate of locomotion and contraction of the foot. The magnitude of the foot contraction increased throughout a 20-s period of rotation, whereas locomotion was reduced immediately after the onset of the rotation and was maintained at this constant low rate throughout the stimulus presentation. These divergent response patterns suggest that the two responses emerge independently. In Experiment 2, a classical conditioning procedure was employed in which a light (CS) was paired with the rotation (US) employed in Experiment 1. In a subsequent test, it was found that the light had acquired the capability to evoke both foot contraction and decreased locomotion. Although the magnitude of these conditioned responses was reduced relative to the corresponding unconditioned response, the patterns of responding were virtually identical; that is, foot contraction developed gradually whereas locomotion decreased immediately. In contrast, animals that received unpaired presentations of the light and rotation, light alone, or no prior exposure to those stimuli exhibited foot extension in response to the light. These results illustrate a transfer of some of the response-evoking properties of the US to the CS as a result of conditioning, as well as the emergence of two independent conditioned responses. Moreover, these results suggest modulation of at least two distinct motor pathways as a function of learning.

Impaired Working Memory Duration But Normal Learning Abilities Found in Mice That Are Conditionally Deficient in the Close Homolog of L1

In addition to its role in axon growth and neuronal migration, the close homolog of L1 (CHL1), a member of the L1 family of cell adhesion molecules, is involved in synaptic plasticity. To date, little has been done to disassociate the role of CHL1 during adulthood from its role during development. To address this issue, mice conditionally deficient in CHL1 (lacking CHL1 only after the third postnatal week) were tested relative to littermate controls as adults in five learning tasks and several tests of working memory (including duration and selective attention). CHL1-deficient mice showed no impairments in the learning tasks compared with wild-type controls. CHL1 deletion had no effect on selective attention despite its widespread impairment of working memory duration. These results suggest a role for CHL1 in the adult-brain in the short-term maintenance of information.

Outgrowths fromHermissenda photoreceptor somata are associated with activation of protein kinase C

We have found changes in the morphology of photoreceptor somata from the molluscHermissenda that are produced by application of 12,13-phorbol dibutyrate (PDBU), an activator of PKC, in combination with elevated intracellular Ca2+ levels. The changes in morphology were expressed as rapid and repetitive outgrowths and additionally as more general changes in shape of the soma. Application of4a-PMA< a phorbol ester which does not activate PKC, did not produce these changes. The functional integrity of the photoreceptors in these dissociated eye prepaations was maintained throughout the period of incubation with PDBU according to standard electrophysiological criteria. It has previously been shown that classical conditioning produced a reduction of dendritic volume in the type B photoreceptor ofHermissenda, a specific locus for associative memory storage. These changes in dendritic morphology were correlated with increased resistance across the cell membrane caused by learning-induced reductions of outward somatic K+ currents. Such conditioning-specific reductions of somatic K+ currents appear to depend on the phosphorylation of a 20-kDa G-protein (CP20) mediated by the Ca2+ and phospholipid-dependent kinase, protein kinase C (PKC). Thus PKC activity may be important in structural changes of the synaptic region of specific neurons involved in associative memory. The results of the present study suggest that the effects of PKC activation may also include structural changes in the soma of these same neurons.

Testing response generation rules.

Robbins (1988) reported data that he viewed as inconsistent with Miller and Schachtmans (1985a) comparator hypothesis of conditioned response generation. Here we explain why we do not find his experiments a compelling test of the comparator hypothesis. We also briefly review other studies that tested the same predictions of the comparator hypothesis that Robbins examined. We conclude that there is considerable evidence that following excitatory or inhibitory conditioning with a target conditioned stimulus (CS) and unconditioned stimulus (US), extinction of other cues that were present during CS training ordinarily increases excitatory responding and decreases inhibitory responding to the CS. However, consistent with Robbinss conclusion, there is scant evidence that after CS-US training, enhancing the associative value of other cues that were present during CS training influences excitatory or inhibitory responding to the CS. The implications of these conclusions for the comparator hypothesis as an explanation of differences in acquired behavior and as a heuristic tool are considered.

Parsing storage from retrieval in experimentally induced amnesia

By 1970, a new framework had begun to emerge with which to understand both memory storage and memory loss. During this period, a number of independent laboratories had demonstrated that both experimentally induced and naturally occurring instances of ‘‘memory failure’’ could be reversed by exposing amnestic subjects to isolated aspects of the training conditions upon which the forgotten memory had been based (Campbell and Jaynes 1966; Lewis et al. 1968; Miller and Springer 1972). While these data suggested a nominally novel interpretation of memory ‘‘loss’’ (better described as a ‘‘lapse’’ in these cases), they were, in some respect, a reassertion of what had long been known: Seemingly forgotten memories are often recovered. Colloquially, what is forgotten is not necessarily gone. To experience this effect outside of the laboratory, one need only look up a long ‘‘forgotten’’ phone number. The phone number will be instantly recognizable and easily relearned, indicative of the intact storage of a memory that seems, superficially, to have been ‘‘lost.’’ Empirical observations from the 1960s led many to conclude that experimentally induced amnesia, as well as naturally occurring memory loss, might reflect instances of retrieval rather than storage failure (Lewis 1969; Spear 1978). Despite the compelling nature of the experimental results and the fundamental significance of the theory that they encouraged, available data have left unresolved this long-standing controversy, and this conundrum may have limited the acceptance of ‘‘retrieval theory’’ by the larger field of memory researchers. In their recent paper, Hardt et al. (2009) correctly identify this enduring stalemate, arguing that prevailing methodologies have left unanswered the most basic of questions, i.e., does experimental amnesia reflect a memory loss or a retrieval failure? As Hardt et al. note, if an instance of memory failure seems immune to recovery, it cannot be said with certainty whether the memory is actually absent or simply that the retrieval blockade is too strong to be overcome with the applied technique (e.g., a specific ‘‘retrieval cue’’). On the other hand, the recovery of a seemingly lost memory could simply reflect an intact, but subthreshold (or degraded), memory being nudged across some previously unreached threshold for activation. Hardt et al. argue that a true test of the opposing views would require a test in which retrieval and storage interpretations of memory failure make opposite predictions regarding a tangible result (i.e., a result that is not dependent on the observation of a null effect). It is that approach that Hardt et al. (2009) take in their recent paper in Learning & Memory. Hart et al.’s approach to this problem is a welcome point of departure from prior efforts, and brings forth a strategy and techniques that were not available when this vein of research was originally struck. It has been previously observed that contextual fear conditioning (in which a context, or place, comes to evoke fear owing to its association with foot shock) is dependent on plasticity within the dorsal hippocampus, and that this plasticity requires the participation of N-methyl-D-aspartate (NMDA) receptors. However, it has also been established that this requirement for NMDA receptors subsides after the first training trial, such that NMDA receptor antagonists (such as 2-amino-5phosphonopentanoic acid; AP5) block learning on the first conditioning event, but not on subsequent exposures to the context paired with shock (Sanders and Fanselow 2003). Hardt et al. exploit this observation for the present purpose. They argue that if inducing experimental amnesia (through the post-training administration of the protein synthesis inhibitor anisomycin; ANI) impairs memory storage after the first training event, then no residual memory will persist into the second training event, and consequently an NMDA receptor antagonist (e.g., AP5) should effectively block learning on the second event, as though prior learning had never occurred. In contrast, if ANI disrupts retrieval of the memory effectively stored after the first training event, then the memory of that event will remain intact into the second training event, and therefore AP5 should be ineffective in blocking learning during the second event; that is, learning in the presence of AP5 during the second training would suggest that a memory of the first training event had survived the amnesic effects of ANI. When these predictions were tested, Hardt et al. find that AP5 impairs learning of the second training event despite ANI having been infused into the dorsal hippocampus after the first training event. This led them to conclude that ANI-induced experimental amnesia impairs memory via its impact on the storage, not retrieval, of the previously coded memory. Despite the virtues of their approach, the results of Hardt et al. are not immune from alternative interpretations, and in fact, introduce a new set of complications akin to those that have plagued prior efforts to parse storage failures from retrieval failures in experimental amnesia. First is the issue raised by the assumption that contextual fear memories reside exclusively in the dorsal hippocampus (dHPC). While it is clear that the dHPC plays a critical role in the mediation of contextually mediated learning, available data suggest that the HPC is not the sole storage site of nominally context-based memories (Good and Honey 1991; Good et al. 2007). Of course it is also possible that learned fear responses may depend intimately on the integrity of the HPC, while actual storage (or some component of the stored memory) may reside in other brain structures (Berger et al. 1976, 1980; Talk et al. 2002). It is indisputable that exactly these types of considerations have complicated any definitive statement regarding the function of the HPC in memory storage. If one accepts even the remote possibility that contextual fear memories reside (even to a functionally trivial degree) in some brain areas in addition to the dHPC, then the dHPC-limited ANI injections administered by Hardt et al. will have left those residual memory traces intact. Were such the case, then those residual memories would have been available to the animal on Hardt et al.’s second training event, and Corresponding author. E-mail matzel@rci.rutgers.edu; fax (732) 445-2263. Article is online at http://www.learnmem.org/cgi/doi/10.1101/lm.1478709.

Memory involves far more than 'consolidation': Commentary — reconsolidation

The observation that retrieval returns a stable memory into a labile state cannot be readily explained by any simple version of consolidation theory. This finding has been interpreted as evidence for the need to reconsolidate a memory after reactivating it. However, as we discuss in this commentary, other behavioural observations indicate that even this modification to consolidation theory may be insufficient to describe the dynamic properties of memory.