Intertrial interval as a contextual stimulus

Four experiments with rats investigated whether the time between appetitive conditioning trials can serve as a discriminative cue for responding during the next conditional stimulus (CS). In Experiment 1, rats that received extinction trials with a 4-min intertrial interval (ITI) showed spontaneous recovery after a retention interval of 16 min, whereas rats that received extinction with a 16-min ITI did not. Experiments 2 and 3 investigated more explicit discriminations between the 4- and 16-min ITIs. When a 16-min ITI signaled that the CS would be reinforced and a 4-min ITI signaled that it would not, the ITIs modulated responding to the CS. But when the 4-min ITI signaled reinforcement and the 16-min ITI did not, there was less evidence of modulation by the ITIs. This asymmetry was due at least partly to a difficulty in performance rather than learning. Experiment 4 investigated similar discriminations with 1- and 4-min ITIs. Here the results took a different form: time in the reinforced ITI elicited responding directly, but did not modulate responding to the CS. ITI can function as a contextual cue, and the results suggest new similarities between the processes behind interval timing and associative learning.     

Memory of a choice direction in a T maze as measured by spontaneous alternation in mice: Effects of intertrial interval and reward

Spontaneous alternation in a T maze was studied as a one trial learning paradigm in mice of the BALB/c strain. In the first experiment the combined effects of time interval between the first and second trial (intertrial interval: ITI), food deprivation and feeding given during the first trial, were shown to affect performance. Thus, on the one hand, the percentage of spontaneous alternation decreased as ITI increased; on the other hand, food reward dramatically improved spontaneous alternation for the 24-h ITI, but had no significant effect for 30-sec and 1-h ITI. Since the effect of feeding might be due either to an increase of arousal, thus favoring input of informations associated with the first choice, or to an improvement in memory consolidation, a second experiment was aimed at testing the effect of food given after the first trial. It was shown that, as in the first experiment, post-trial feeding improved spontaneous alternation on the second trial given 24 hours later with a temporal gradient of effect less than 30 min. These results clearly showed that the reinforcement of run to one side (first trial) increased the tendency to go to the other side 24 hours later. It is concluded that reinforcement might have two distinct effects: (i) according to SR theory, reinforcement increases conditioned responses and (ii), as shown here, acts on memory processes by preventing memory traces from fading. The fact that this last effect was only observed for long ITI suggests that short-term or transient memory and long-term memory are two relatively independent processes.     

A parametric study of the stimulation variables affecting the magnitude of the olfactory nerve response

The magnitude of olfactory responses can be related to three primary variables [number of odorant molecules (N), sniff volume (V), and sniff duration (T)] and three derived variables [concentration (C = N/V), flow rate (F = V/T), and delivery rate (D = N/T)]. To evaluate the effects of these interdependent variables upon the olfactory response, the summated multiunit discharges were recorded from the olfactory nerves of nine frogs in response to octane presented at two levels (in 2:1 ratio) of each primary variable. This presentation defined eight "sniff" combinations representing three levels of each derived variable. In an ANOVA of the logs of the responses, the effect of each primary variable was highly significant, with no significant interactions. A multiplicative regression model incorporating the effects of the three primary variables represented responses exceedingly well, with positive effects of N and T and a negative effect of V. When, with this model, the effect of each of the derived variables was isolated from the effects of all other variables, the analysis showed a positive effect for C, a near-zero positive effect for D, and a negative effect for F. Placing certain constraints upon the model parameters generates 13 distinct one- and two-variable models (e.g., the [C, T] model requires N and V to have equal but opposite effects). In ranking these reduced models in terms of their ability to predict the neural response, the predictive ability of [F, N] and [C, T] was at least as good as that of the three-variable model.     

The effect of intertrial interval food presentions on pigeons' delayed matching to sample accuracy

Parallel effects of temporal variables on autoshaping and on delayed matching to sample performance suggest that delayed matching, like autoshaping, might depend upon the within-trial expectancy of reinforcement relative to the overall expectancy of reinforcement in the session. This possibility was assessed by presenting free food at different times during a 30-sec intertrial interval (ITI) in a delayed matching to sample procedure with pigeons. In three conditions a single free food presentation occurred, either early, mid-way, or late in each ITI; in another condition, three food presentations occurred during each ITI, one at each time location. Relative to a baseline condition, in which free food never occurred during the ITI, only food presentations late in the ITI produced a significant disruption in accuracy, and this effect occurred only at the longest of three delays tested. Three free food presentations in each ITI disrupted accuracy only to the same degree as a single, late, ITI food presentation. Thus, accuracy was affected by the temporal location rather than the frequency of ITI food presentations. These effects appear to differ from those of ITI food presentations on autoshaping and do not seem to be understandable in terms of changes in the background expectancy of reinforcement. It was suggested instead that food presented late in the ITI might disrupt subsequent memory processes.     

Elucidation of the order of oxidations and identification of an intermediate in the multistep clavaminate synthase reaction

The enzyme clavaminate synthase (CS) catalyzes the formation of the first bicyclic intermediate in the biosynthetic pathway to the potent beta-lactamase inhibitor clavulanic acid. Our previous work has led to the proposal that the cyclization/desaturation of the substrate proclavaminate proceeds in two oxidative steps, each coupled to a decarboxylation of alpha-ketoglutarate and a reduction of dioxygen to water [Salowe, S. P., Marsh, E. N., & Townsend, C. A. (1990) Biochemistry 29, 6499-6508]. We have now employed kinetic isotope effect studies to determine the order of oxidations for CS purified from Streptomyces clavuligerus. By using (4'RS)-[4'-3H,1-14C]-rac-proclavaminate, a primary T(V/K) = 8.3 +/- 0.2 was measured from [3H]water release data, while an alpha-secondary T(V/K) = 1.06 +/- 0.01 was determined from the changing 3H/14C ratio of the product clavaminate. Values for the primary and alpha-secondary effects of 11.9 +/- 1.7 and 1.12 +/- 0.07, respectively, were obtained from the changing 3H/14C ratio of the residual proclavaminate by using new equations derived for a racemic substrate bearing isotopic label at both primary and alpha-secondary positions. Since only the first step of consecutive irreversible reactions will exhibit a V/K isotope effect, we conclude that C-4' is the initial site of oxidation in proclavaminate. As expected, no significant changes in the 3H/14C ratio of residual substrate were observed with [3-3H,1-14C]-rac-proclavaminate. However, two new tritiated compounds were produced in this incubation, apparently the result of isotope-induced branching brought about by the presence of tritium at the site of the second oxidation. One of these compounds was identified by comparison to authentic material as dihydroclavaminate, a stable intermediate that normally remains enzyme-bound. On the basis of the body of information available and the similarities to alpha-ketoglutarate-dependent dioxygenases, a comprehensive mechanistic scheme for CS is proposed to account for this unusual enzymatic transformation.     

Does the terminal-link effect depend on duration or reinforcement rate?

Preference in concurrent chains for the richer terminal-link schedule becomes more extreme as the schedule values increase with their ratio held constant, a result known as the terminal-link effect. We report two experiments that attempt to determine whether this effect is related to terminal-link duration or the overall rate of reinforcement. These variables have been confounded in prior studies, but can be separated by comparing variable-duration schedules that end after a single reinforcer has been earned, with constant-duration schedules during which a variable number of reinforcers may be earned. In Experiment 1, the terminal-link effect was obtained with variable-duration schedules when duration and overall reinforcement rate were manipulated, but not with constant-duration schedules when overall reinforcement rate was changed with duration held constant. In Experiment 2, the terminal-link effect was obtained with constant-duration schedules when duration was manipulated with overall reinforcement rate held constant. Taken together, these results show that the terminal-link effect depends on changes in terminal-link duration, not overall reinforcement rate (or equivalently, average time to reinforcement). This accords with the account of the terminal-link effect provided by the contextual choice model [J. Exp. Anal. Behav. 61 (1994) 113] but not delay-reduction theory [J. Exp. Anal. Behav. 12 (1969) 723].     

Relative temporal representations in Pavlovian conditioning

The transfer of relative temporal representations was assessed in a series of three experiments. In each experiment, rats (Rattus norvegicus) received one set of conditioned stimulus (CS) and intertrial interval (ITI) durations in Phase 1 and another set in Phase 2. The ratio between the CS and ITI intervals was either changed or maintained across phases. On the hypothesis that relative temporal representations are learned, groups receiving maintained temporal ratios across phases were expected to display greater change in responding upon encountering the new intervals. When the CS duration decreased across phases, maintaining the temporal ratio did lead to greater change in Day 1 of Phase 2 towards the final pattern of responding. However, when the CS increased across phases, maintaining the temporal ratio across phases did not facilitate adjustment to the new intervals, suggesting that extinction of previously reinforced times induced new learning. These results provide evidence that under some conditions, relative relationships in temporal maps may survive transformation-of-scale, like relative relationships in spatial maps.     

Assessment of methods for improving tracer estimation of non-steady-state rate of appearance.

The most common approach for estimating substrate rate of appearance (R(a)) is use of the single-pool model first proposed by R. W. Steele, J. S. Wall, R. C. DeBodo, and N. Altszuler. (Am. J. Physiol. 187: 15-24, 1956). To overcome the model error during highly non-steady-state conditions due to the assumption of a constant volume of distribution (V), two strategies have been proposed: 1) use of a variable tracer infusion rate to minimize tracer-to-tracee ratio (TTR) variations (fixed-volume approach) or 2) use of two tracers of the same substrate with one infused at a constant rate and the other at a variable rate (variable-volume approach or approach of T. Issekutz, R. Issekutz, and D. Elahi. Can. J. Physiol. Pharmacol. 52: 215-224, 1974). The goal of this study was to compare the results of these two strategies for the analysis of the kinetics of glycerol and glucose under the non-steady-state condition created by a constant infusion of epinephrine (50 ng. kg(-1). min(-1)) with the traditional approach of Steele et al., which uses a constant infusion and fixed volume. The results showed that for glucose and glycerol the estimates of R(a) obtained with the constant and the variable tracer infusion rate and the equation of Steele et al. were comparable. The variable tracer infusion approach was less sensitive to the choice of V in estimating R(a) for glycerol and glucose, although the advantage of changing the tracer infusion rate was greater for glucose than for glycerol. The model of Issekutz et al. showed instability when the ratio TTR(1)/TTR(2) approaches a constant value, and the model is more sensitive to measurement error than the constant-volume model for glucose and glycerol. We conclude that the one-tracer constant-infusion technique is sufficient in most cases for glycerol, whereas the one-tracer variable-infusion technique is preferable for glucose. Reasonable values for glucose R(a) can be obtained with the constant-infusion technique if V = 145 ml/kg.     

Effects of inter-trial interval length on food-hoarding partial reinforcement of running behaviour in the golden hamster

Food-hoarding provides an adequate motivation in sated hamsters for the acquisition of a two-way running response. This learning was studied using a discrete-trial procedure, in continuous (CR) and partial reinforcement (PR) conditions, with two different inter-trial interval (ITI) lengths. The dependent variables were: the time spent by Ss in the goal section of the runway, and the number of their runs during extinction.The PR training had two effects on goal times: a slow decrease in acquisition on non-reinforced trials, and then a stabilization at this level during the extinction phase, as contrasted with the sudden increase found in CR-trained Ss when they were switched from acquisition to extinction conditions. However, the PR effects on number of runs depended upon ITI length: resistance to extinction of PR-trained Ss was superior to that of CR-trained Ss with spaced, but not with massed, trials. In the latter case, CR-trained Ss persisted as much as did PR-trained Ss. An hypothesis is offered, along the lines of the Frustration theory.     

Thermostabilizing mutations preferentially occur at structural weak spots with a high mutation ratio

We apply Constraint Network Analysis (CNA) to investigate the relationship between structural rigidity and thermostability of five citrate synthase (CS) structures over a temperature range from 37 °C to 100 °C. For the first time, we introduce an ensemble-based variant of CNA and model the temperature-dependence of hydrophobic interactions in the constraint network. A very good correlation between the predicted thermostabilities of CS and optimal growth temperatures of their source organisms (R2=0.88, p=0.017) is obtained, which validates that CNA is able to quantitatively discriminate between less and more thermostable proteins even within a series of orthologs. Structural weak spots on a less thermostable CS, predicted by CNA to be in the top 5% with respect to the frequency of occurrence over an ensemble, have a higher mutation ratio in a more thermostable CS than other sequence positions. Furthermore, highly ranked weak spots that are also highly conserved with respect to the amino acid type found at that sequence position are nevertheless found to be mutated in the more stable CS. As for mechanisms at an atomic level that lead to a reinforcement of weak spots in more stable CS, we observe that the thermophilic CS achieve a higher thermostability by better hydrogen bonding networks whereas hyperthermophilic CS incorporate more hydrophobic contacts to reach the same goal. Overall, these findings suggest that CNA can be applied as a pre-filter in data-driven protein engineering to focus on residues that are highly likely to improve thermostability upon mutation.     

A Spiking Neural Network Model of Model-Free Reinforcement Learning with High-Dimensional Sensory Input and Perceptual Ambiguity

A theoretical framework of reinforcement learning plays an important role in understanding action selection in animals. Spiking neural networks provide a theoretically grounded means to test computational hypotheses on neurally plausible algorithms of reinforcement learning through numerical simulation. However, most of these models cannot handle observations which are noisy, or occurred in the past, even though these are inevitable and constraining features of learning in real environments. This class of problem is formally known as partially observable reinforcement learning (PORL) problems. It provides a generalization of reinforcement learning to partially observable domains. In addition, observations in the real world tend to be rich and high-dimensional. In this work, we use a spiking neural network model to approximate the free energy of a restricted Boltzmann machine and apply it to the solution of PORL problems with high-dimensional observations. Our spiking network model solves maze tasks with perceptually ambiguous high-dimensional observations without knowledge of the true environment. An extended model with working memory also solves history-dependent tasks. The way spiking neural networks handle PORL problems may provide a glimpse into the underlying laws of neural information processing which can only be discovered through such a top-down approach.     

Application of neural networks for controlling and predicting quality parameters in beer fermentation

The biochemical pathways involved in the production of ethyl caproate, a secondary product of the beer fermentation process, are not well established. Hence, there are no phenomenological models available to control and predict the production of this particular compound as with other related products. In this work, neural networks have been used to fit experimental results with constant and variable pH, giving a good fit of laboratory and industrial scale data. The results at constant pH were also used to predict results at variable pH. Finally, the application of neural networks obtained from laboratory experiments gave excellent predictions of results in industrial breweries and so could be used in the control of industrial operations. The input pattern to the neural network included the accumulated fermentation time, cell dry weight, consumption of sugars and aminoacids and, in some cases, the pH. The output from the neural network was an estimation of quantity of the ethyl caproate ester.     

Relevance of various formulations of phytoplankton chlorophyll a:carbon ratio in a 3D marine ecosystem model

Numerous experimental studies showed that the phytoplankton Chla-to-Carbon ratio (Chla:C) is highly variable, whereas most of the marine ecosystem models use a constant ratio. In this work, we tested three different formulations for computing the modelled Chla in a 3D coupled hydrodynamical-biogeochemical model of the Southwest lagoon of New Caledonia. The first formulation considers a constant Chla:C ratio. In the second one, Chla is a diagnostic variable related to the variable phytoplankton nitrogen-to-carbon ratio. In the last formulation, Chla is a state variable of the model, which is dynamically simulated. Results showed important differences between the formulations, the first leading to overestimate the Chla concentration in low nutrients conditions. Thus, this study strengthens the importance of the Chla modelling in a coupled model in order to better estimate a crucial variable for validation of ecosystem models.     

Effects of intertrial reinforcers on rats' timing behavior

Two experiments were conducted to assess the effects of non-contingent intertrial interval (ITI) reinforcers on rats' discrimination of duration. In the first experiment, rats' discrimination of a 2 vs. 8 s of light was significantly disrupted when reinforcers were presented in the ITI. Disruption was not different on short (2 s) and long (8 s) trials. The second experiment showed that this disruptive effect was not specific to trials preceded by ITI reinforcers; responding on empty ITI trials run in the same session as ITI-reinforcer trials was also disrupted. This disruption however was not as great as on the ITI-reinforcer trials. The results of these experiments show that ITI reinforcers affect timing discriminations in much the same way they affect classical conditioning and delayed matching to sample. However, detailed examination of the results suggests that the deleterious effects of ITI reinforcers in these different paradigms might be produced by different rather than the same mechanism. The results also support the conclusion that pre-trial reinforcement “priming” produces disruption rather than facilitation in complex tasks.     

A comparison of statistical approaches for modelling fish species distributions

SUMMARY 1. The prediction of species distributions is of primary importance in ecology and conservation biology. Statistical models play an important role in this regard; however, researchers have little guidance when choosing between competing methodologies because few comparative studies have been conducted. 2. We provide a comprehensive comparison of traditional and alternative techniques for predicting species distributions using logistic regression analysis, linear discriminant analysis, classification trees and artificial neural networks to model: (1) the presence/absence of 27 fish species as a function of habitat conditions in 286 temperate lakes located in south‐central Ontario, Canada and (2) simulated data sets exhibiting deterministic, linear and non‐linear species response curves. 3. Detailed evaluation of model predictive power showed that approaches produced species models that differed in overall correct classification, specificity (i.e. ability to correctly predict species absence) and sensitivity (i.e. ability to correctly predict speciespresence) and in terms of which of the study lakes they correctly classified. Onaverage, neural networks outperformed the other modelling approaches, although all approaches predicted species presence/absence with moderate to excellent success. 4. Based on simulated non‐linear data, classification trees and neural networks greatly outperformed traditional approaches, whereas all approaches exhibited similar correct classification rates when modelling simulated linear data. 5. Detailed evaluation of model explanatory insight showed that the relative importance of the habitat variables in the species models varied among the approaches, where habitat variable importance was similar among approaches for some species and very different for others. 6. In general, differences in predictive power (both correct classification rate and identity of the lakes correctly classified) among the approaches corresponded with differences in habitat variable importance, suggesting that non‐linear modelling approaches (i.e. classification trees and neural networks) are better able to capture and model complex, non‐linear patterns found in ecological data. The results from the comparisons using simulated data further support this notion. 7. By employing parallel modelling approaches with the same set of data and focusing on comparing multiple metrics of predictive performance, researchers can begin to choose predictive models that not only provide the greatest predictive power, but also best fit the proposed application.     

A genetic algorithm for the optimization of the thermoinduction protocol for high-level production of recombinant human-like collagen from Escherichia coli

Production of recombinant human-like collagen (RHLC) by thermoinduction of recombinant Escherichia coli BL 21 during high cell density cultivation was investigated in a 30 L bioreactor. The effects of induction temperature (T), pH, and carbon-to-nitrogen molar ratio of the nutrient medium (C/N) were examined. The optimal thermoinduction protocol for RHLC production was determined by using a model coupling genetic algorithm and artificial neural networks. The optimal operating conditions were as follows: maintenance of induction temperature at 42°C for 3 H and then at 39.4°C until the end, induction pH at 7.03, and C/N at 4.8 (mol/mol). The theoretical maximum concentration of RHLC was 12.5 g/L, whereas the experimental value was 12.1 g/L under the optimal induction conditions.     

Gut T1R3 sweet taste receptors do not mediate sucrose-conditioned flavor preferences in mice

Most mammals prefer the sweet taste of sugars, which is mediated by the heterodimeric T1R2+T1R3 taste receptor. Sugar appetite is also enhanced by the post-oral reinforcing actions of the nutrient in the gut. Here, we examined the contribution of gut T1R3 (either alone or as part of the T1R3+T1R3 receptor) to post-oral sugar reinforcement using a flavor-conditioning paradigm. We trained mice to associate consumption of a flavored solution (CS+) with intragastric (IG) infusions of a sweetener, and a different flavored solution (CS-) with IG infusions of water (23 h/day); then, we measured preference in a CS+ vs. CS- choice test. In experiment 1, we predicted that if activation of gut T1R3 mediates sugar reinforcement, then IG infusions of a nutritive (sucrose) or nonnutritive (sucralose) ligand for this receptor should condition a preference for the CS+ in B6 wild-type (WT) mice. While the mice that received IG sucrose infusions developed a strong preference for the CS+, those that received IG sucralose infusions developed a weak avoidance of the CS+. In experiment 2, we used T1R3 knockout (KO) mice to examine the necessity of gut T1R2+T1R3 receptors for conditioned flavor preferences. If intact gut T1R3 (or T1R2+T1R3) receptors are necessary for flavor-sugar conditioning, then T1R3 KO mice should not develop a sugar-conditioned flavor preference. We found that T1R3 KO mice, like WT mice, acquired a strong preference for the CS+ paired with IG sucrose infusions. The KO mice were also like WT mice in avoiding a CS+ flavor paired with IG sucralose infusions These findings provide clear evidence that gut T1R3 receptors are not necessary for sugar-conditioned flavor preferences or sucralose-induced flavor avoidance in mice.     

一类中立型Hopfield神经网络的全局吸引集

讨论了中立型Hopfield神经网络模型,利用矩阵谱的性质和微分不等式分析等技巧,给出了其不变集和全局吸引集的判别准则．特别地,当系统有平衡点时,我们也得到了平衡点全局稳定的判别条件．     

Transmission delays and frequency detuning can regulate information flow between brain regions

Brain networks exhibit very variable and dynamical functional connectivity and flexible configurations of information exchange despite their overall fixed structure. Brain oscillations are hypothesized to underlie time-dependent functional connectivity by periodically changing the excitability of neural populations. In this paper, we investigate the role of the connection delay and the detuning between the natural frequencies of neural populations in the transmission of signals. Based on numerical simulations and analytical arguments, we show that the amount of information transfer between two oscillating neural populations could be determined by their connection delay and the mismatch in their oscillation frequencies. Our results highlight the role of the collective phase response curve of the oscillating neural populations for the efficacy of signal transmission and the quality of the information transfer in brain networks.     

Stepping outside the box in considering the C/T ratio

Rate expectancy theory (RET) predicts that in Pavlovian procedures conditioned responding will be directly related to the ratio of time spent in the experimental context (C) relative to the trial time (T) or duration of the conditioned stimulus (CS). This prediction was discussed in the context of three experiments. The first and third experiments involved sexual conditioning in quail [Learn Motiv 31 (2000) 211; J Exp Psychol Anim Behav Processes 27 (2001) 269]; the second experiment involved conditioning rats with food as the unconditioned stimulus (US) [Learn Motiv 28(1997) 465]. In each experiment, one type of conditioned response was directly related to the C/T ratio (as predicted by RET) but another conditioned response was inversely related. In addition, the conditioned behavior that occurred with low C/T ratios was controlled by contextual cues rather than the putative CS. The implications of these findings for possible boundary conditions of RET were discussed. The discussion revealed serious shortcomings in the characterization of context conditioning by RET.