Lynch et al. Cognition enhancement in normal subjects in neuronal excitability cannot be excluded in these in vivo cursor,a previously seen real world cue from a group of similar studies. objects.Performance on the task was increased dramatically with ampakine pretreatment.Brain imaging studies then uncovered a FUNCTIONAL EFFECTS OF ENHANCED NETWORKS remarkable result:the ampakine intensified activity in frontal and There are relatively straightforward results showing the effects of temporal cortices but also led to the engagement of a superior increasing throughput in cortical networks in simple experimen- parietal region,the precuneus,which was inactive during vehi- tal paradigms.For example,unilateral lesions of the nigro-striatal cle trials.The precuneus is thought to be critical for envisioning projections result in a circling response to dopamine agonists; future actions by humans.In any case,these results in a primate ampakines significantly expand activation of the motor cortex provide an example in which expansion of cortical networks is on the side of the lesion and this is associated with a suppres- associated with a lifting of limits on performance in a cognitively sion of rotations (Hess et al.,2003).Note that in this case circuits demanding problem. are selectively brought into play that are directly germane to the These few studies using overtrained animals,exciting though problem faced by the animal.The expectation of a more subtle the findings may be,are hardly sufficient to establish the general version of this effect under conditions in which cortex is perform- point that increases in network throughput result in beyond nor- ing complex calculations constitutes one basis for hypothesizing mal performance on challenging problems.Experiments of this that improved network throughput will result in acute enhance- type are not common because they involve major investments ment of cognition.It should be noted here that the development in time and technology.And it will be noted that they focus on of very fast algorithms for extracting core spatio-temporal activ- problems that are sharply defined with regard to cues and appro- ity patterns from multi-electrode recordings has made it possible priate responses.One can fairly ask if these conditions capture to insert,via stimulation at many sites,information rich patterns the essence of cognition as a free flowing processing of the enor- to networks in behaving animals.These advances have opened mous complexities generated by the exterior and interior worlds the way to experimental testing of fundamental,long-standing of humans.This point is picked up in the following section. assumptions about how cortical circuits process complex signals from the environment.One recent study of this kind that is par- FUTURE NEUROBIOLOGICAL STUDIES ON COGNITIVE ENHANCEMENT ticularly germane to the present discussion showed that delivery A question running through this review,and alluded to imme- to CAl of a"correct"pattern of activation predicted from CA3 diately above,concerns the extent to which we can consider recordings during the sample phase of a match to sample problem problem solving by animals as a fair descriptor of cognition.One markedly enhanced performance of monkeys on the subsequent can hardly question the proposition that the analyses of different decision phase in difficult versions of the task (Hampson et al., computations performed by distinct frontal subfields in rats(e.g., 2013).These dramatic findings encourage the idea that facilitat- credit assignment to particular serial actions,set shifting,focusing ing partial or "weak"network patterns can lead to pronounced of attention;Turner et al.,2004;Sugrue et al.,2005;Demeter and improvements in the ability of animals,including primates,to Sarter,2013)will provide deep insights into how humans resolve deal with complexity. real world issues.But here we encounter the problem of how to As mentioned earlier,enhancement could take the form of define cognition and whether or not it can be understood in sim- acceleration of cognitive activities,and thus allowing for more ple computational terms.To be specific,what might be needed are computations in the same time frame,or expansion of networks studies testing whether putative enhancers improve the perfor- and potentially new types of operations.One route for testing mance of sophisticated (highly experienced)subjects dealing with the latter possibility would be to overtrain animals to the point novel circumstances of great complexity and without the benefit at which optimal performance is fully established and then to of external supervision.If nothing else,this would bring exper- determine if network facilitation through enhanced transmission iments closer to the human condition and thereby help explain allows the subject to go beyond normal limits.There is very little why animal studies on cognition and memory have such a poor work of this kind for ampakines or any other putative enhancer record in predicting human outcomes. but suggestive results have been described.The study noted above Much of the present discussion centered on the proposition in which ampakines expanded the hippocampal response during that enhancing network throughput will have positive effects learning(Figure 6)also found that overtrained rats significantly on cognition (Figures 7A,B).It will be recognized that most improved their learning scores under the influence of the drug.of the material presented in support of this idea dealt with Remarkably,the animals then continued to perform at supra- specific neuronal circuits or opportunistic discoveries of net- normal levels in the absence of the ampakine.Detailed analyses work expansion.A more systematic,agnostic description of how showed that the animals shifted response strategies in a man- experimental compounds affect the vast number of forebrain ner that reduced proactive interference between trials(Hampson circuits is badly needed.This could be obtained using activity- et al.,1998a).In essence,the drugs opened the way to expanded regulated immediate early gene expression to provide an index networks and the development of higher order rules that can-of the recent history of neuronal firing.Such analyses would not otherwise be acquired even with weeks of training.Another provide a picture of the networks assembled to deal with com- example of going well beyond normal limits has been described plex circumstances,surely an initial step toward a mechanism for monkeys performing a challenging delayed match-to-sample based theory of cognitive operations,and add an informa- problem (Porrino et al,2005).The animals were trained to tion rich step for the screening of experimental compounds. asymptote to identify,as indicated by movement of a computer There is also the possibility that network maps of drug effects Frontiers in Systems Neuroscience www.frontiersin.org May 2014 Volume 8 Article 90 12Lynch et al. Cognition enhancement in normal subjects in neuronal excitability cannot be excluded in these in vivo studies. FUNCTIONAL EFFECTS OF ENHANCED NETWORKS There are relatively straightforward results showing the effects of increasing throughput in cortical networks in simple experimen￾tal paradigms. For example, unilateral lesions of the nigro-striatal projections result in a circling response to dopamine agonists; ampakines significantly expand activation of the motor cortex on the side of the lesion and this is associated with a suppres￾sion of rotations (Hess et al., 2003). Note that in this case circuits are selectively brought into play that are directly germane to the problem faced by the animal. The expectation of a more subtle version of this effect under conditions in which cortex is perform￾ing complex calculations constitutes one basis for hypothesizing that improved network throughput will result in acute enhance￾ment of cognition. It should be noted here that the development of very fast algorithms for extracting core spatio-temporal activ￾ity patterns from multi-electrode recordings has made it possible to insert, via stimulation at many sites, information rich patterns to networks in behaving animals. These advances have opened the way to experimental testing of fundamental, long-standing assumptions about how cortical circuits process complex signals from the environment. One recent study of this kind that is par￾ticularly germane to the present discussion showed that delivery to CA1 of a “correct” pattern of activation predicted from CA3 recordings during the sample phase of a match to sample problem markedly enhanced performance of monkeys on the subsequent decision phase in difficult versions of the task (Hampson et al., 2013). These dramatic findings encourage the idea that facilitat￾ing partial or “weak” network patterns can lead to pronounced improvements in the ability of animals, including primates, to deal with complexity. As mentioned earlier, enhancement could take the form of acceleration of cognitive activities, and thus allowing for more computations in the same time frame, or expansion of networks and potentially new types of operations. One route for testing the latter possibility would be to overtrain animals to the point at which optimal performance is fully established and then to determine if network facilitation through enhanced transmission allows the subject to go beyond normal limits. There is very little work of this kind for ampakines or any other putative enhancer but suggestive results have been described. The study noted above in which ampakines expanded the hippocampal response during learning (Figure 6) also found that overtrained rats significantly improved their learning scores under the influence of the drug. Remarkably, the animals then continued to perform at supra￾normal levels in the absence of the ampakine. Detailed analyses showed that the animals shifted response strategies in a man￾ner that reduced proactive interference between trials (Hampson et al., 1998a). In essence, the drugs opened the way to expanded networks and the development of higher order rules that can￾not otherwise be acquired even with weeks of training. Another example of going well beyond normal limits has been described for monkeys performing a challenging delayed match-to-sample problem (Porrino et al., 2005). The animals were trained to asymptote to identify, as indicated by movement of a computer cursor, a previously seen real world cue from a group of similar objects. Performance on the task was increased dramatically with ampakine pretreatment. Brain imaging studies then uncovered a remarkable result: the ampakine intensified activity in frontal and temporal cortices but also led to the engagement of a superior parietal region, the precuneus, which was inactive during vehi￾cle trials. The precuneus is thought to be critical for envisioning future actions by humans. In any case, these results in a primate provide an example in which expansion of cortical networks is associated with a lifting of limits on performance in a cognitively demanding problem. These few studies using overtrained animals, exciting though the findings may be, are hardly sufficient to establish the general point that increases in network throughput result in beyond nor￾mal performance on challenging problems. Experiments of this type are not common because they involve major investments in time and technology. And it will be noted that they focus on problems that are sharply defined with regard to cues and appro￾priate responses. One can fairly ask if these conditions capture the essence of cognition as a free flowing processing of the enor￾mous complexities generated by the exterior and interior worlds of humans. This point is picked up in the following section. FUTURE NEUROBIOLOGICAL STUDIES ON COGNITIVE ENHANCEMENT A question running through this review, and alluded to imme￾diately above, concerns the extent to which we can consider problem solving by animals as a fair descriptor of cognition. One can hardly question the proposition that the analyses of different computations performed by distinct frontal subfields in rats (e.g., credit assignment to particular serial actions, set shifting, focusing of attention; Turner et al., 2004; Sugrue et al., 2005; Demeter and Sarter, 2013) will provide deep insights into how humans resolve real world issues. But here we encounter the problem of how to define cognition and whether or not it can be understood in sim￾ple computational terms. To be specific, what might be needed are studies testing whether putative enhancers improve the perfor￾mance of sophisticated (highly experienced) subjects dealing with novel circumstances of great complexity and without the benefit of external supervision. If nothing else, this would bring exper￾iments closer to the human condition and thereby help explain why animal studies on cognition and memory have such a poor record in predicting human outcomes. Much of the present discussion centered on the proposition that enhancing network throughput will have positive effects on cognition (Figures 7A,B). It will be recognized that most of the material presented in support of this idea dealt with specific neuronal circuits or opportunistic discoveries of net￾work expansion. A more systematic, agnostic description of how experimental compounds affect the vast number of forebrain circuits is badly needed. This could be obtained using activity￾regulated immediate early gene expression to provide an index of the recent history of neuronal firing. Such analyses would provide a picture of the networks assembled to deal with com￾plex circumstances, surely an initial step toward a mechanism based theory of cognitive operations, and add an informa￾tion rich step for the screening of experimental compounds. There is also the possibility that network maps of drug effects Frontiers in Systems Neuroscience www.frontiersin.org May 2014 | Volume 8 | Article 90 | 12