Robustly estimating the marginal likelihood for cognitive models via importance sampling.

Recent advances in Markov chain Monte Carlo (MCMC) extend the scope of Bayesian inference to models for which the likelihood function is intractable. Although these developments allow us to estimate model parameters, other basic problems such as estimating the marginal likelihood, a fundamental tool in Bayesian model selection, remain challenging. This is an important scientific limitation because testing psychological hypotheses with hierarchical models has proven difficult with current model selection methods. We propose an efficient method for estimating the marginal likelihood for models where the likelihood is intractable, but can be estimated unbiasedly. It is based on first running a sampling method such as MCMC to obtain samples for the model parameters, and then using these samples to construct the proposal density in an importance sampling (IS) framework with an unbiased estimate of the likelihood. Our method has several attractive properties: it generates an unbiased estimate of the marginal likelihood, it is robust to the quality and target of the sampling method used to form the IS proposals, and it is computationally cheap to estimate the variance of the marginal likelihood estimator. We also obtain the convergence properties of the method and provide guidelines on maximizing computational efficiency. The method is illustrated in two challenging cases involving hierarchical models: identifying the form of individual differences in an applied choice scenario, and evaluating the best parameterization of a cognitive model in a speeded decision making context. Freely available code to implement the methods is provided. Extensions to posterior moment estimation and parallelization are also discussed.

Wound-healing ability is conserved during periods of chronic stress and costly life history events in a wild-caught bird.

Chronic stress, potentially through the actions of corticosterone, is thought to directly impair the function of immune cells. However, chronic stress may also have an indirect effect by influencing allocation of energy, ultimately shifting resources away from the immune system. If so, the effects of chronic stress on immune responses may be greater during energetically-costly life history events. To test whether the effects of chronic stress on immune responses differ during expensive life history events we measured wound healing rate in molting and non-molting European starlings (Sturnus vulgaris) exposed to control or chronic stress conditions. To determine whether corticosterone correlated with wound healing rates before starting chronic stress, we measured baseline and stress-induced corticosterone and two estimates of corticosterone release and regulation, negative feedback (using dexamethasone injection), and maximal capacity of the adrenals to secrete corticosterone (using adrenocorticotropin hormone [ACTH] injection). After 8days of exposure to chronic stress, we wounded both control and chronically stressed birds and monitored healing daily. We monitored nighttime heart rate, which strongly correlates with energy expenditure, and body mass throughout the study. Measures of corticosterone did not differ with molt status. Contrary to work on lizards and small mammals, all birds, regardless of stress or molt status, fully-healed wounds at similar rates. Although chronic stress did not influence healing rates, individuals with low baseline corticosterone or strong negative feedback had faster healing rates than individuals with high baseline corticosterone or weak negative feedback. In addition, wound healing does appear to be linked to energy expenditure and body mass. Non-molting, chronically stressed birds decreased nighttime heart rate during healing, but this pattern did not exist in molting birds. Additionally, birds of heavier body mass at the start of the experiment healed wounds more rapidly than lighter birds. Finally, chronically stressed birds lost body mass at the start of chronic stress, but after wounding all birds regardless of stress or molt status started gaining weight, which continued for the remainder of the study. Increased body mass could suggest compensatory feeding to offset energetic or resource demands (e.g., proteins) of wound healing. Although chronic stress did not inhibit healing, our data suggest that corticosterone may play an important role in mediating healing processes and that molt could influence energy saving tactics during periods of chronic stress. Although the experiment was designed to test allostasis, interpretation of data through reactive scope appears to be a better fit.

Ground-state fluctuations in finite Fermi systems.

We consider a small and fixed number of fermions in a trap. The ground state of the system is defined at T=0. For a given excitation energy, there are several ways of exciting the particles from this ground state. We formulate a method for calculating the number fluctuation in the ground state using microcanonical counting, and implement it for noninteracting fermions in harmonic confinement. This exact calculation for fluctuation, when compared with canonical or grand canonical ensemble averaging, gives considerably different results. This difference is expected to persist at low excitation even when the fermion number in the trap is large. For comparison, the well-known bosonic results are also given.

Distinct ionotropic GABA receptors mediate presynaptic and postsynaptic inhibition in retinal bipolar cells.

Ionotropic GABA receptors can mediate presynaptic and postsynaptic inhibition. We assessed the contributions of GABA(A) and GABA(C) receptors to inhibition at the dendrites and axon terminals of ferret retinal bipolar cells by recording currents evoked by focal application of GABA in the retinal slice. Currents elicited at the dendrites were mediated predominantly by GABA(A) receptors, whereas responses evoked at the terminals had GABA(A) and GABA(C) components. The ratio of GABA(C) to GABA(A) (GABA(C):GABA(A)) was highest in rod bipolar cell terminals and variable among cone bipolars, but generally was lower in OFF than in ON classes. Our results also suggest that the GABA(C):GABA(A) could influence the time course of responses. Currents evoked at the terminals decayed slowly in cell types for which the GABA(C):GABA(A) was high, but decayed relatively rapidly in cells for which this ratio was low. Immunohistochemical studies corroborated our physiological results. GABA(A) beta2/3 subunit immunoreactivity was intense in the outer and inner plexiform layers (OPL and IPL, respectively). GABA(C) rho subunit labeling was weak in the OPL but strong in the IPL in which puncta colocalized with terminals of rod bipolars immunoreactive for protein kinase C and of cone bipolars immunoreactive for calbindin or recoverin. These data demonstrate that GABA(A) receptors mediate GABAergic inhibition on bipolar cell dendrites in the OPL, that GABA(A) and GABA(C) receptors mediate inhibition on axon terminals in the IPL, and that the GABA(C):GABA(A) on the terminals may tune the response characteristics of the bipolar cell.

Morphological differentiation of bipolar cells in the ferret retina.

Bipolar cells are not only important for visual processing but input from these cells may underlie the reorganization of ganglion cell dendrites in the inner plexiform layer (IPL) during development. Because little is known about the development of bipolar cells, here we have used immunocytochemical markers and dye labeling to identify and follow their differentiation in the neonatal ferret retina. Putative cone bipolar cells were immunoreacted for calbindin and recoverin, and rod bipolar cells were immunostained for protein kinase C (PKC). Our results show that calbindin-immunoreactive cone bipolar cells appear at postnatal day 15 (P15), at which time their axonal terminals are already localized to the inner half of the IPL. By contrast, recoverin-immunoreactive cells with terminals in the IPL are present at birth, but many of these cells may be immature photoreceptors. By the second postnatal week, recoverin-positive cells resembling cone bipolar cells were clearly present, and with increasing age, two distinct strata of immunolabeled processes occupied the IPL. PKC-containing rod bipolar cells emerged by the fourth postnatal week and at this age have stratified arbors in the inner IPL. The early bias of bipolar axonal arbors in terminating in the inner or outer half of the IPL is confirmed by dye labeling of cells with somata in the inner nuclear layer. At P10, several days before ribbon synapses have been previously observed in the ferret IPL, the axon terminals of all dye-labeled bipolar cells were clearly stratified. The results suggest that bipolar cells could provide spatially localized interactions that are suitable for guiding dendritic lamination in the inner retina.

AMPA receptor kinetics limit retinal amacrine cell excitatory synaptic responses.

Amacrine cells that respond transiently to maintained illumination are thought to mediate transient inhibitory input to ganglion cells. The excitation of these transient amacrine cells is thought to be limited by inhibitory feedback to bipolar cells. We investigated the possibility that desensitizing AMPA and/or kainate (KA) receptors on amacrine cells might also limit the duration of amacrine cell excitation. To determine how these receptors might affect amacrine cell input and output, we made whole-cell recordings from amacrine and ganglion cells in the salamander retinal slice. The specific AMPA receptor antagonist GYKI-53655 blocked non-NMDA receptor-mediated amacrine cell excitatory postsynaptic currents (EPSCs) and kainate puff-elicited currents, indicating that AMPA, and not KA, receptors mediated the responses. Cyclothiazide, an agent that reduces AMPA receptor desensitization, increased the amplitude and duration of amacrine cell EPSCs. To measure the output of transient amacrine cells, we recorded glycinergic inhibitory postsynaptic currents (IPSCs) from ganglion cells, and found that these were also enhanced by cyclothiazide. Thus, prolongation of amacrine cell AMPA receptor activation enhanced amacrine cell output. Current responses elicited by puffing glycine onto ganglion cell dendrites were not affected by cyclothiazide, indicating that the enhancement of glycinergic IPSCs was not due to a direct effect on glycine receptors. These data suggest that rapid AMPA receptor desensitization and/or deactivation limits glycinergic amacrine cell excitation and the resulting inhibitory synaptic output.

Lidocaine-loaded biodegradable nanospheres. I. Optimization Of the drug incorporation into the polymer matrix.

Spherical nanoparticulate drug carriers made of poly(d,l-lactic acid) with controlled size were designed. A local anesthetic, lidocaine, a small hydrophobic molecule, was incorporated in the core with loadings varying from about 7 to 32% (w/w) and increasing with the particle size. Particles with sizes from about 250 to 820 nm and low polydispersity were prepared with good reproducibility; the polymer concentration (at constant surfactant concentration) governed the particle size. The large particles with a high loading ( approximately 30%) showed under in vitro conditions a slow release over 24-30 h, the medium sized carriers (loading of approximately 13%) released the drug over about 15 h, whereas the small particles with small loading ( approximately 7%) exhibited a rapid release over a couple of hours. It seems that the drug release rate is related to the state (crystallized or dispersed) of the drug incorporated in the polymer matrix.

Persistence of deafferentation-induced presynaptic dendrites in the cerebellar cortex of adult rats.

The paper provides morphological evidence for a long-lasting morphogenetic potential of the excitatory granule cells and the inhibitory small Golgi neurons to manufacture and to maintain presynaptic sites on their otherwise exclusively postsynaptic dendritic processes and somata. The development of new dendrodendritic and somato-dendritic synapses follows and is the direct consequence of the mossy fiber deafferentation of the cerebellar cortex. By GABA-immunocytochemistry it was demonstrated that not only Golgi axons but also the Golgi somata and presynaptic dendrites contain GABA, suggesting that this inhibitory transmitter is an operating factor in many of the newly formed synapses.