Access the full text.
Sign up today, get DeepDyve free for 14 days.
The transitive inference (TI) task assesses the ability to generalize learned knowledge to new contexts, and is thought to depend on the hippocampus (Dusek & Eichenbaum, 1997). Animals or humans learn in separate trials to choose stimulus A over B, B over C, C over D and D over E, via reinforcement feedback. Transitive responding based on the hierarchical structure A > B > C > D > E is then tested with the novel BD pair. We and others have argued that successful BD performance by animals – and even humans in some implicit studies – can be explained by simple reinforcement learning processes which do not depend critically on the hippocampus, but rather on the striatal dopamine system. We recently showed that the benzodiazepene midazolam, which is thought to disrupt hippocampal function, profoundly impaired human memory recall performance but actually enhanced implicit TI performance (Frank, O'Reilly & Curran, 2006). We posited that midazolam biased participants to recruit striatum during learning due to dysfunctional hippocampal processing, and that this change actually supported generalization of reinforcement values. Greene (2007) questions the validity of our pharmacological assumptions and argues that our conclusions are unfounded. Here we stand by our original hypothesis, which remains the most parsimonious account of the data, and is grounded by multiple lines of evidence. for genuine, explicit logical inference), neural models sug- Background Our interpretation of our findings was based on the fol- gest that the elemental associative learning process itself lowing premises. When humans are prevented from does not depend on the hippocampus, but rather on the becoming explicitly aware of the hierarchy they can still striatal dopamine system [6,7]. Moreover, the striatal and perform better than chance at the novel "inference" test hippocampal systems often compete, such that disruption using an implicit reinforcement learning system . Sev- of one system can lead to enhanced performance of tasks eral mathematical and neural models show that when that depend on the other (for review: [8,9]). trained with the TI task procedure, stimulus B develops a higher associative strength then stimulus D, and that tran- We therefore posited that midazolam would disrupt the sitive responding can be achieved simply by comparing hippocampus [10-18], thereby removing its normal these implicit values [1-5]. Thus although under some cir- inhibitory interaction with the striatum, and allowing that cumstances the hippocampus can play a subtle modula- system to dominate. The resulting impaired hippocampal tory role in setting up these values  (and is likely critical learning but enhanced implicit reinforcement encoding Page 1 of 5 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:5 http://www.behavioralandbrainfunctions.com/content/4/1/5 led to (i) impaired explicit memory recall; (ii) a pattern of Discussion learning in the initial training phase of the TI task that is a Below we elaborate these and other issues in response to characteristic signature of associative models: greater per- Greene's three criticisms, and identify some new avenues formance on the end anchor pairs AB and DE, which can for research to more directly address the question. be solved purely on the basis of reinforcement values, and worse performance on the conjunctive inner pairs BC and Assumption 1: other areas critical to TI account for the effect CD; (iii) substantially enhanced generalization of rein- forcement values in the transitive test . These results Greene takes issue with our suggestion that midazolam support predictions from more abstract mathematical improved TI performance by deactivating the hippocam- models in which conjunctive learning is detrimental to pus , stating that "it is entirely possible that the effects subsequent transitive responding . Finally, in another of midazolam on the TI task are attributable to deactiva- probabilistic learning task associated with the striatal tions of areas other than the hippocampus". He cites PET dopamine system , midazolam led to spared perform- evidence that midazolam inactivates some of the very ance and only caused deficits in the very first few trials of same frontal and parietal regions that have been shown to the task , when the hippocampus is usually most be activated during explicit TI performance in other stud- active in such tasks , consistent with other accounts ies [23,24]. We are puzzled by this point, as it is difficult on the role of the hippocampus in the early acquisition of to imagine how deactivation of a region that is normally probabilistic learning . activated during TI would lead to enhanced performance. Admitting possible additional frontal cortical effects of In his critique, Greene argued that our theory entails three midazolam, our key point was that thus far no studies critical assumptions that are "required but not met"  implicate striatal deactivation following midazolam. (pp. 1–2): Other evidence from Parkinson's patients and dopamine medication manipulations support predictions from 1. that other areas critical to TI are not affected by mida- computational models of the basal ganglia, which suggest zolam administration, so only hippocampal deactivation that this system is involved in learning reinforcement can explain the effect; associations in the TI task . These data provide converg- ing evidence for a role of striatal DA system in learning 2. that midazolam deactivates the hippocampus so that it implicit associative values in the TI task. no longer functions in a mnemonic capacity; and We note here that Greene is undeniably correct that by 3. that midazolam influences explicit but not implicit administering a systemic drug one cannot know for cer- memory. tain that either the observed behavioral decrements in recall (amnesia) and the associated improvement in TI, It is reasonable to question the assumptions of our logic. are related specifically to hippocampal deactivation. Nevertheless, on the whole we believe this criticism is mis- However, our central hypothesis is that the critical brain guided. First, there is ample evidence to suggest that hip- system supporting reinforcement-based learning in the TI pocampal function is dramatically impaired by and other feedback learning tasks is the striatal dopamine midazolam administration, and critically, no indication system, and to the best of our knowledge there is no evi- that the striatal reinforcement learning system – strongly dence to suggest this system is inhibited by midazolam, predicted to support reinforcement-based generalization either at the neural level or in learning tasks that depend in this task – is negatively affected. Second, while it is cer- on this system. Moreover, there is some evidence in rats tainly possible that areas other than the hippocampus are that the two main peptide markers of activity in the direct affected by the systemic drug, it is entirely unclear why and indirect pathways of the basal ganglia (dynorphin this would result in an enhancement in TI performance as and enkephalin; ) are actually substantially enhanced was observed . In contrast, our theoretical framework under systemic midazolam administration [26,27], as are clearly accounts for the effect, and is supported by other striatal dopamine levels . These effects could conceiv- diagnostic converging data in our study. Third, the notion ably result from the removal of the normal inhibitory that midazolam impairs explicit memory substantially interaction with hippocampus, which have been observed more than implicit memory, while well supported by psy- under a variety of conditions in both animals and humans chological data, is actually not a critical assumption of our [8,9]. As one recent (non-pharmacological) example in logic, which depends more specifically on striatal- humans, when dual task conditions were introduced to dependent implicit associative learning rather than on interfere with explicit memory, procedural learning nor- implicit learning in general. mally correlated with hippocampal activation and associ- ated declarative knowledge proceeded instead in a Page 2 of 5 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:5 http://www.behavioralandbrainfunctions.com/content/4/1/5 habitual fashion that correlated with striatal activation support generalization of learned behavior to novel situa- . tions [33,34]. Assumption 2: midazolam obliterates hippocampal Assumption 3: midazolam impairs explicit but not implicit function memory Greene also questions whether midazolam disrupts hip- Finally, Greene claims that our argument requires mida- pocampal function at all. That is, despite the dense expres- zolam to affect explicit but not implicit memory, but that sion of benzodiazepene receptors in the hippocampus there are a few studies showing some impairments in , some human neuroimaging studies fail to detect implicit memory. We note that first, there is substantial hippocampal deactivation under midazolam. While one evidence to suggest that midazolam has a far greater pro- study did show a dose-dependent hippocampal effect portional impairing effect on explicit than implicit mem- , others using low doses did not. However, the effects ory [35-38]. Other recent studies showed that the same of drugs on BOLD activity are not straightforward. First, experimental procedures (including midazolam dose) BOLD is a very indirect and imprecise measure of neural used in our study selectively abolished event-related brain activity, whereas direct administration of midazolam in potentials associated with recall, which had been previ- the hippocampus show robust disruptions of both neural ously linked to hippocampal function, without affecting activity and plasticity [10-13]. These findings are not other memory components . However, the need for restricted to local injections into the hippocampus. For midazolam to spare implicit memories of all types is actu- example, systemic midazolam administration has been ally not a requirement of our theory at all. Indeed, we shown to decrease hippocampal cholinergic activity . posit that midazolam impairs the conjunctive encoding of While this action should clearly interfere with hippocam- stimulus features in the hippocampus, and while this pal function and memory encoding, it is much less clear function is critical for several aspects of explicit memory, what effect, if any, this would have on BOLD activity – it it can also support some aspects of implicit memory . could simply add noise to the hippocampal system. Thus again our argument relies on the notion that mida- Indeed, this hippocampal noise hypothesis has been zolam has not been shown to impair striatal-dependent modeled to account for various detrimental memory memories, i.e. in so-called procedural learning tasks. encoding effects of midazolam ). It is similarly unclear how midazolam's inhibitory effects on hippoc- Future studies ampal long term potentiation  – the central mecha- To help further resolve this controversy, midazolam nism thought to give rise to hippocampal memories – effects on hippocampal activity ought to be examined in would translate into BOLD. situations for which the hippocampus is thought to be critical according to the same computational principles Furthermore, Greene also asserts that our theory requires used to motivate our original hypothesis. It would be par- midazolam to "deactivate the hippocampus to the point ticularly informative to explore the effects of midazolam that it no longer functions in a mnemonic capacity". This on a contrast between conjunctive and elemental memory is clearly not the case – while the low dose used in our encoding in the TI or other task showing hippocampal study robustly impaired memory encoding/recall, this sensitivity to conjunctions . Indeed, Greene's own memory was far from obliterated altogether. These find- study  showed hippocampal activation for the con- ings are similar to other studies purporting that mida- junctive "inner pairs" in the TI task, and midazolam zolam impairs hippocampal function and substantially impaired acquisition of these same pairs in our study , degrades, but does not eliminate, episodic memory and impaired conjunctive encoding in other tasks . [16,17,31]. In contrast, hippocampal amnesics were sub- stantially impaired at learning the training pairs, particu- Conclusion larly the inner pairs that require conjunctive encoding, Our original study was motivated by an explicit theoreti- never reaching training criteria . Thus although these cal and neurocomputational framework for why this pat- amnesics were impaired at the TI probe test, these data are tern should be observed, and is consistent with other not meaningful given that they never acquired the rele- psychological and mathematical analysis. In contrast, vant premises in the first place. In contrast, participants on while legitimately questioning some of our assumptions, midazolam were impaired initially at the conjunctive Greene provides no alternative explanation for the finding inner pairs relative to control subjects, but this deficit was that midazolam substantially improved TI performance in subtle enough that they were able to eventually overcome our study. Until such an interpretation arises, with sup- it. Other studies show that both humans and animals porting data, the most parsimonious explanation of our with hippocampal damage are completely unimpaired at data is that motivated by a priori theoretical predictions. re-combining learned elemental reinforcement values to Greene's commentary does reaffirm that alternative opin- Page 3 of 5 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:5 http://www.behavioralandbrainfunctions.com/content/4/1/5 of episodic recognition memory by midazolam. J Exp Psychol ions are valid and that as is always the case, further Learn Mem Cogn 2004, 30(2):540-549. research is necessary. 18. Evans MS, Viola-McCabe KE: Midazolam inhibits long-term potentiation through modulation of GABA-A receptors. Neuropharmacology 1996, 35:347-57. Abbreviations 19. Frank MJ, OReilly RC, Curran T: When memory fails, intuition TI: transitive inference reigns: Midazolam enhances implicit inference in humans. Psychol Sci 2006, 17:700-707. 20. Poldrack RA, Clark J, PareBlagoev EJ, Shohamy D, Moyano JC, Myers Competing interests C, Gluck MA: Interactive memory systems in the human The author(s) declare that they have no competing inter- brain. Nature 2001, 414:546-549. 21. Shohamy D, Myers CE, Kalanithi J, Gluck MA: Basal ganglia and ests. dopamine contributions to probabilistic category learning. Neurosci Biobehav Rev 2007. Authors' contributions 22. Greene AJ: Implicit transitive inference and the human hip- pocampus: does intravenous midazolam function as a M.J.F., R.C.O. and T.C wrote the paper. reversible hippocampal lesion? Behav Brain Funct 2007, 3:51. 23. Nagode JC, Pardo JV: Human hippocampal activation during transitive inference. Neuroreport 2002, 13:939-944. References 24. Acuna BD, Eliassen JC, Donoghue JP, Sanes JN: Frontal and Pari- 1. Frank MJ, Rudy JW, Levy WB, O'Reilly RC: When Logic Fails: etal Lobe Activation during Transitive Inference in Humans. Implicit Transitive Inference in Humans. Mem Cognit 2005, Cereb Cortex 2002, 12:1312-1321. 33:742-50. 25. Gerfen CR, Wilson C: The basal ganglia. In Handbook of chemical 2. von Fersen L, Wynne CDL, Delius JD, Staddon JER: Transitive neuroanatomy. Integrated systems of the CNS Volume 12. Edited by: inference in pigeons. J Exp Psychol Anim Behav Proc 1991, Swanson L, Bjorkland A, Hokfelt T. Amsterdam: Elsevier; 17:334-341. 1996:371-468. 3. Wynne CD: A minimal model of transitive inference. In Models 26. Rattan AK, Tejwani GA: Effect of chronic treatment with mor- of Action Edited by: Wynne CD, Staddon JE. New Jersey: Lawrence phine, midazolam and both together on dynorphin(1–13) Erlbaum Associates; 1998:269-307. levels in the rat. Brain Res 1997, 754(1–2):239-244. 4. Frank MJ, Rudy JW, O'Reilly RC: Transitivity, flexibility, conjunc- 27. Tejwani GA, Rattan AK: Met-enkephalin alteration in the rat tive representations and the hippocampus: II. A Computa- during chronic injection of morphine and/or midazolam. tional analysis. Hippocampus 2003, 13:341-54. Brain Res 1997, 775(1–2):119-126. 5. Siemann M, Delius JD: Algebraic learning and neural network 28. Gil E, Colado I, Lopez F, Fernandez-Briera A, Fernandez-Lopez A, models for transitive and non-transitive responding. Eur J Calvo P: Effects of chronic treatment with ethanol and with- Cogn Psychol 1998, 10:307-334. drawal of ethanol on levels of dopamine, 3,4-dihydroxyphe- 6. Frank MJ: Dynamic dopamine modulation in the basal ganglia: nylacetic acid and homovanillic acid in the striatum of the A neurocomputational account of cognitive deficits in medi- rat. Influence of benzodiazepines, barbiturate and somato- cated and non-medicated Parkinsonism. J Cogn Neurosci 2005, statin. Neuropharmacology 1992, 31(11):1151-1156. 17:51-72. 29. Foerde K, Knowlton BJ, Poldrack RA: Modulation of competing 7. Frank MJ, Seeberger LC, O'Reilly RC: By carrot or by stick: Cog- memory systems by distraction. Proc Natl Acad Sci USA 2006, nitive reinforcement learning in Parkinsonism. Science 2004, 103(31):11778-11783. 306:1940-3. 30. Montpied P, Martin BM, Cottingham SL, Stubblefield BK, Ginns EI, 8. Poldrack RA, Packard MG: Competition among multiple mem- Paul SM: Regional distribution of the GABAA/benzodi- ory systems: converging evidence from animal and human azepene receptor (alpha subunit) mRNA in rat brain. J Neu- brain studies. Neuropsychologia 2003, 41:245-251. rochem 1988, 51:1651-4. 9. Atallah HE, Frank MJ, O'Reilly RC: Hippocampus cortex and 31. Curran T, DeBuse C, Woroch B, Hirshman E: Combined pharma- basal ganglia: insights from computational models of com- cological and electrophysiological dissociation of familiarity plementary learning systems. Neurobiol Learn Mem 2004, and recollection. J Neurosci 2006, 26(7):1979-1985. 82(3):253-267. 32. Smith C, Squire LR: Declarative memory, awareness, and tran- 10. Rovira C, Ben-Ari Y: Developmental study of benzodiazepene sitive inference. J Neurosci 2005, 25(44):10138-10146. effects on monosynaptic GABAA-mediated IPSPs of rat hip- 33. Driscoll I, Sutherland RJ, Prusky GT, Rudy JW: Damage to the Hip- pocampal neurons. J Neurophysiol 1993, 70(3):1076-1085. pocampal Formation Does Not Disrupt Representational 11. Poncer JC, Durr R, Gahwiler BH, Thompson SM: Modulation of Flexibility as Measured by a Novelty Transfer Test. Behav synaptic GABAA receptor function by benzodiazepenes in Neurosci 2004, 118(6):1427-1432. area CA3 of rat hippocampal slice cultures. Neuropharmacology 34. Bayley PJ, Frascino JC, Squire LR: Robust habit learning in the 1996, 35:1169-79. absence of awareness and independent of the medial tempo- 12. Kristiansen U, Lambert JD: Benzodiazepene and barbituate lig- ral lobe. Nature 2005, 436:550-3. ands modulate responses of cultured hippocampal neurones 35. Hirshman E, Passannante A, Arndt J: Midazolam Amnesia and to the GABAA receptor partial agonist, 4-PIOL. Neurophar- Conceptual Processing in Implicit Memory. J Exp Psychol Gen macology 1996, 35:1181-91. 2001, 130:453-465. 13. Kobayashi S, Fujito Y, Matsuyama K, Aoki M: Differential effects of 36. Arndt J, Passannante A, Hirshman E: The effect of midazolam on midazolam on inhibitory postsynaptic currents in CA1 implicit and explicit memory in category exemplar produc- pyramidal cells and dentate gyrus granule cells of rat hippoc- tion and category cued recall. Memory 2004, 12:158-73. ampal slices. Brain Res 2004, 1003:176-82. 37. Thomas-Anterion C, Koenig O, Navez M, Laurent B: Midazolam 14. Imperato A, Dazzi L, Obinu MC, Gessa GL, Biggio G: Inhibition of effects on implicit and explicit memory processes in healthy hippocampal acetylcholine release by benzodiazepines: subjects. Psychopharmacology 1999, 145:139-43. antagonism by flumazenil. Eur J Pharmacol 1993, 238:135-137. 38. Hirshman E, Fisher J, Henthorn T, Arndt J, Passannante A: Mida- 15. Reinsel RA, Veselis RA, Dnistrian AM, Feshchenko VA, Beattle BJ, zolam amnesia and dual-process models of the word-fre- Duff MR: Midazolam decreases cerebral blood flow in the left quency mirror effect. J Mem Lang 2002, 47:499-516. prefrontal cortex in a dose-dependent fashion. Int J Neuropsy- 39. Park H, Quinlan J, Thornton E, Reder LM: The effect of mida- chopharmacol 2000, 3:117-27. zolam on visual search: Implications for understanding 16. Merritt P, Hirshman E, Zamani S, Hsu J, Berrigan M: Episodic repre- amnesia. Proc Natl Acad Sci USA 2004, 101:17879-83. sentations support early semantic learning: evidence from 40. Giovanello KS, Schnyer DM, Verfaeilie M: A critical role for the midazolam induced amnesia. Brain Cogn 2006, 61(2):219-223. anterior hippocampus in relational memory: evidence from 17. Malmberg KJ, Zeelenberg R, Shiffrin RM: Turning up the noise or an fMRI study comparing associative and item recognition. turning down the volume? On the nature of the impairment Hippocampus 2004, 14:5-8. Page 4 of 5 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:5 http://www.behavioralandbrainfunctions.com/content/4/1/5 41. Greene AJ, Gross WL, Elsinger CL, Rao SM: An FMRI analysis of the human hippocampus: inference, context, and task awareness. J Cogn Neurosci 2006, 18(7):1156-1173. Publish with Bio Med Central and every scientist can read your work free of charge "BioMed Central will be the most significant development for disseminating the results of biomedical researc h in our lifetime." Sir Paul Nurse, Cancer Research UK Your research papers will be: available free of charge to the entire biomedical community peer reviewed and published immediately upon acceptance cited in PubMed and archived on PubMed Central yours — you keep the copyright BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 5 of 5 (page number not for citation purposes)
Behavioral and Brain Functions – Springer Journals
Published: Jan 30, 2008
Access the full text.
Sign up today, get DeepDyve free for 14 days.