Meclizine enhancement of sensorimotor gating in healthy male subjects with high startle responses and low prepulse inhibition.

Histamine H1 receptor systems have been shown in animal studies to have important roles in the reversal of sensorimotor gating deficits, as measured by prepulse inhibition (PPI). H1-antagonist treatment attenuates the PPI impairments caused by either blockade of NMDA glutamate receptors or facilitation of dopamine transmission. The current experiment brought the investigation of H1 effects on sensorimotor gating to human studies. The effects of the histamine H1 antagonist meclizine on the startle response and PPI were investigated in healthy male subjects with high baseline startle responses and low PPI levels. Meclizine was administered to participants (n=24) using a within-subjects design with each participant receiving 0, 12.5, and 25 mg of meclizine in a counterbalanced order. Startle response, PPI, heart rate response, galvanic skin response, and changes in self-report ratings of alertness levels and affective states (arousal and valence) were assessed. When compared with the control (placebo) condition, the two doses of meclizine analyzed (12.5 and 25 mg) produced significant increases in PPI without affecting the magnitude of the startle response or other physiological variables. Meclizine also caused a significant increase in overall self-reported arousal levels, which was not correlated with the observed increase in PPI. These results are in agreement with previous reports in the animal literature and suggest that H1 antagonists may have beneficial effects in the treatment of subjects with compromised sensorimotor gating and enhanced motor responses to sensory stimuli.

Explaining neural signals in human visual cortex with an associative learning model.

"Predictive coding" models posit a key role for associative learning in visual cognition, viewing perceptual inference as a process of matching (learned) top-down predictions (or expectations) against bottom-up sensory evidence. At the neural level, these models propose that each region along the visual processing hierarchy entails one set of processing units encoding predictions of bottom-up input, and another set computing mismatches (prediction error or surprise) between predictions and evidence. This contrasts with traditional views of visual neurons operating purely as bottom-up feature detectors. In support of the predictive coding hypothesis, a recent human neuroimaging study (Egner, Monti, & Summerfield, 2010) showed that neural population responses to expected and unexpected face and house stimuli in the "fusiform face area" (FFA) could be well-described as a summation of hypothetical face-expectation and -surprise signals, but not by feature detector responses. Here, we used computer simulations to test whether these imaging data could be formally explained within the broader framework of a mathematical neural network model of associative learning (Schmajuk, Gray, & Lam, 1996). Results show that FFA responses could be fit very closely by model variables coding for conditional predictions (and their violations) of stimuli that unconditionally activate the FFA. These data document that neural population signals in the ventral visual stream that deviate from classic feature detection responses can formally be explained by associative prediction and surprise signals.

Special issue on computational models of classical conditioning guest editors' introduction.

In the present special issue, the performance of current computational models of classical conditioning was evaluated under three requirements: (1) Models were to be tested against a list of previously agreed-upon phenomena; (2) the parameters were fixed across simulations; and (3) the simulations used to test the models had to be made available. These requirements resulted in three major products: (a) a list of fundamental classical-conditioning results for which there is a consensus about their reliability; (b) the necessary information to evaluate each of the models on the basis of its ordinal successes in accounting for the experimental data; and (c) a repository of computational models ready to generate simulations. We believe that the contents of this issue represent the 2012 state of the art in computational modeling of classical conditioning and provide a way to find promising avenues for future model development.

Solving Pavlov's puzzle: attentional, associative, and flexible configural mechanisms in classical conditioning.

This article introduces a new "real-time" model of classical conditioning that combines attentional, associative, and "flexible" configural mechanisms. In the model, attention to both conditioned (CS) and configural (CN) stimuli are modulated by the novelty detected in the environment. Novelty increases with the unpredicted presence or absence of any CS, unconditioned stimulus (US), or context. Attention regulates the magnitude of the associations CSs and CNs form with other CSs and the US. We incorporate a flexible configural mechanism in which attention to the CN stimuli increases only after the model has unsuccessfully attempted learn input-output combinations with CS-US associations. That is, CSs become associated with the US and other CSs on fewer trials than they do CNs. Because the CSs activate the CNs through unmodifiable connections, a CS can become directly and indirectly (through the CN) associated with the US or other CSs. In order to simulate timing processes, we simply assume that a CS is formed by a temporal spectrum of short-duration CSs that are activated by the nominal CS trace. The model accurately describes 94 % of the basic properties of classical conditioning, using fixed model parameters and simulation values in all simulations.

Classical conditioning mechanisms can differentiate between seeing and doing in rats.

We show that the attentional-associative SLG model of classical conditioning, based on the 1996 research of Schmajuk, Lam, and Gray, correctly describes experimental results regarded as evidence of causal learning in rats: (a) interventions attenuate responding following common-cause training but do not interfere on subsequent responding during observation, and (b) interventions do not affect responding after direct-cause training or (c) causal-chain training. According to the model, responding to the weakly attended test stimulus is strongly inhibited by the intervention in the common-cause case. Instead, in the direct-cause and causal-chain cases, the strongly attended test stimulus becomes inhibitory, thereby overshadowing the inhibitory effect of interventions. Most importantly, the model predicted that with relatively few test trials (a) the 2008 results of Experiment 3 by Leising, Wong, Waldmann, and Blaisdell should be similar to those of Dwyer, Starns, and Honey's 2009 Experiment 1, showing that interventions equally affect responding after common-cause and direct-cause training; and (b) the 2006 results of Experiment 2a by Blaisdell, Sawa, Leising, and Waldmann should be similar to those of Dwyer, Starns, and Honey's 2009 Experiment 2, showing that interventions equally affect responding after common-cause and causal-chain training. When those data were made available to us, we confirmed those predictions. In agreement with the SLG associative model, but not with causal model theory, this evidence supports the notion that the attenuation of responding by interventions only following common-cause training is the consequence of well-known learning processes-latent inhibition, sensory preconditioning, conditioned inhibition, protection from extinction, and overshadowing.

Deactivation and reactivation of the inhibitory power of a conditioned inhibitor: testing the predictions of an attentional-associative model.

An attentional-associative model (Schmajuk, Lam, & Gray Journal of Experimental Psychology: Animal Behavior Processes, 22, 321-349, 1996) assumes that nonreinforced presentations of an inhibitory conditioned stimulus (CS) do not decrease its inhibitory associations. However, the model predicts that extended presentations will decrease attention to the inhibitor, thereby decreasing both (1) the expression of its inhibitory power in a summation test and (2) the rate of acquisition in a retardation test. The model also predicts that subsequent presentations of the inhibitory CS with a novel CS will increase both (1) and (2). Using a predictive learning design in humans, Experiment 1 examined the predictions involving the summation tests, whereas Experiments 2 and 3 examined the predictions involving the retardation tests. Experimental results were in agreement with the predictions of the model.

Latent inhibition and compound conditioning: a reply to Holmes and Harris (2009).

Schmajuk, Lam, and Gray (SLG, 1996) presented a neural network model of classical conditioning that addresses the multiple properties of latent inhibition (LI). According to the model, LI is the result of the decreased attention to the target stimulus during preexposure and testing. Recently, Holmes and Harris (2009) suggested that, although the model was able to describe their experimental results showing that LI to a preexposed stimulus disappears with extended compound conditioning, it could not describe the fact that LI is not affected by a delay following compound conditioning. However, computer simulations demonstrate that the SLG model describes and explains both results. Because the model also explains both the deleterious and the facilitating effects on LI of a delay following simple conditioning, the SLG model seems unique in explaining the complete range of reported effects of temporal delays on LI as well as most of the properties of LI.

A computational model reveals classical conditioning mechanisms underlying visual signal detection in rats.

We applied a neural network model of classical conditioning proposed by Schmajuk et al. (1996) to visual signal detection and discrimination tasks designed to assess sustained attention in rats (Bushnell, 1999). We used a sign-tracking description of signal detection behavior by assuming that rats approach the location of the lever that they expect will be rewarded. We also assumed that although the visual signals contribute to set the occasion for the approach response to be generated, they do not become strongly associated with reward. The model accurately described Bushnell's (1999) results showing an increased proportion of hits with increasing signal intensity, decreased proportion of hits with increasing trial rate, and lower accuracy in a discrimination task compared to a detection task. In addition, observation of the behavior of rats performing the task confirmed assumptions and predictions of the model: (a) rats learn to approach the location of the lever they expect to be rewarded; (b) during the pre-signal interval, rats approach the location of the blank lever because it matches the intensity of the light they experience during that interval; and (c) the rats' behavior is directed to the location of the levers and not towards the light, which acts only as an "occasion setter" for the lever to be selected and pressed.

Interpreting patterns of brain activation in human fear conditioning with an attentional-associative learning model.

J. E. Dunsmoor, P. A. Bandettini, and D. C. Knight conducted a neuroimaging study of human fear conditioning and analyzed brain activity under various pairing rates between a conditioned and an unconditioned stimulus. Computer simulations with an attentional-associative model introduced by N. A. Schmajuk, Y. W. Lam, and J. A. Gray (1996) show that activity in the amygdala and anterior cingulate cortex is well described by a variable representing the prediction of the unconditioned stimulus, whereas activity in the dorsolateral prefrontal cortex and insula is well captured by a variable coding the attentional-modulated representation of conditioned stimuli. In addition, the model explains how those variables control behavior and provides a clear framework in which those variables play important roles in the description of numerous classical conditioning paradigms. Also, the model offers a number of predictions related to stimulus novelty for future neuroimaging studies of associative learning.

Attentional and error-correcting associative mechanisms in classical conditioning.

R. A. Rescorla (2000, Rescorla, 2001, Rescorla, 2002) reported that the associative changes undergone by 2 conditioned stimuli that are reinforced or not reinforced in compound depend on their initial associations. The results contradict the predictions of simple error-correction models but can be explained by models that incorporate a "constrained" error-correction rule. A model of classical conditioning presented by N. A. Schmajuk, Y. Lam, and J. A. Gray (1996) suggests that attentional mechanisms, acting during both compound training and testing, have an important role in producing those results. Moreover, the model suggests that those attentional mechanisms might obscure the evaluation of the associative changes undergone by the conditioned stimuli during compound training. Two experiments that differentiate our model from competing theories are proposed.

Attentional, associative, and configural mechanisms in extinction.

The participation of attentional and associative mechanisms in extinction, spontaneous recovery, external disinhibition, renewal, reinstatement, and reacquisition was evaluated through computer simulations with an extant computational model of classical conditioning (N. A. Schmajuk, Y. Lam, & J. A. Gray, 1996; N. A. Schmajuk & J. A. Larrauri, 2006). The model assumes that attention to stimuli (controlled by environmental novelty) and associations between stimuli interact during memory storage (learning) and retrieval (performance). Computer simulations indicated that a combination of attentional and associative mechanisms might be sufficient to describe most of the properties of extinction. However, configural mechanisms seem necessary to describe the properties of cues that precede the target stimulus during extinction (extinction cues) and might improve the description of some experimental results regarding the associative properties of the extinction context. These configural mechanisms can be easily integrated into the present version of the model.

Reinstatement of conditioned fear and the hippocampus: an attentional-associative model.

An existing attentional-associative model of classical conditioning [Schmajuk N, Lam Y, Gray JA. Latent inhibition: a neural network approach. J Exp Psychol: Anim Behav Process 1996;22:321-49] is applied to the description of reinstatement in animals and humans. According to the model, inhibitory associations between the context (CX) and unconditioned stimulus (US) are formed during extinction, which help preserve the association between the conditioned stimulus (CS) and the US. However, summation and retardation tests fail to reveal these associations because (a) the CX is not attended or (b) a CX-CS configural stimulus formed during extinction is both poorly attended and weakly active during testing. When US presentations and testing occur in the same context, reinstatement is the consequence of a decreased CX inhibition and the increased attention to the CS, which activates the remaining CS-US association. When US presentations occur in the context of extinction but the CS is tested in a different context, reinstatement results from an increased attention to the CS and the combination of CS-CX and CX-US excitatory associations. The assumption that associations between CSs are impaired following neurotoxic hippocampal lesions or in amnesia, is sufficient to describe absence of reinstatement in those cases. However, additional assumptions might be needed to describe the effect of hippocampal lesions on other postextinction manipulations.

Experimental challenges to theories of classical conditioning: application of an attentional model of storage and retrieval.

Several studies have recently challenged the accuracy of traditional models of classical conditioning that account for some experimental data in terms of a storage deficit. Among other results, it has been reported that extinction of the blocking or overshadowing stimulus results in the recovery of the response to the blocked or overshadowed stimulus, backward blocking shows spontaneous recovery, extinction of the training context results in the recovery from latent inhibition, interposing a delay between conditioning and testing in latent inhibition increases latent inhibition, and latent inhibition antagonizes overshadowing. An existing neural network model of classical conditioning (N. A. Schmajuk, Y. Lam, & J. A. Gray, 1996), which includes an attentional mechanism controlling both storage and retrieval of associations, is able to quantitatively describe these results.

Brain-behaviour relationships in latent inhibition: a computational model.

In a series of studies, we applied a neural network to study the neural bases of latent inhibition. We first designed a model able to handle behavioral data and then we investigated whether structures and neural elements in the brain were able to carry out the operations described by network. We demonstrated that the network was able to describe many of the behavioral properties of LI, and elucidate the effects of several manipulations of the dopaminergic system, the hippocampus, and the nucleus accumbens on LI, as well as some of the positive symptoms of schizophrenia. The results support the idea that a 'conceptual nervous system' can be successfully used to relate brain and behavior.

Neural network model of prepulse inhibition.

The authors introduce a real-time model of acoustic prepulse inhibition (PPI) and facilitation (PPF) in animals and humans. The model incorporates excitatory and facilitatory pathways activated by the positive value of changes in noise level in the environment and an inhibitory pathway activated by the absolute value of changes in noise level. Whereas excitation and facilitation are exponential functions, inhibition is a linear function of the input noise expressed in decibels. The model describes many properties of PPI and PPF that include, among others, their dependency on prepulse intensity and duration, duration of the lead interval, and changes in background noise. The model also describes how specific brain lesions enhance the strength of the startle response and impair PPI. Finally, the model correctly predicts how PPI depends on pulse intensity.

Latent learning, shortcuts and detours: a computational model.

Voicu and Schmajuk (Rob. Auto. Syst. 35 (2001a) 23) described a model of spatial navigation and exploration that includes an action system capable of guiding, with the help of a cognitive system, the search for specific goals as determined by a motivation system. Whereas in the original model the cognitive map stores information about the connectivity between places in the environment, in the present version the cognitive map also stores information about the paths traversed by the agent. Computer simulations show that the network correctly describes experimental results including latent learning in a maze, detours in a maze, and shortcuts in an open field. In addition, the model generates novel predictions about detours and shortcuts in an open field.

The Transition from Automatic to Controlled Processing.

The transition from automatic (unconscious) to controlled (conscious) processing is described in terms of a neural network model of classical conditioning ([Schmajuk et al., 1996]). In the framework of the network, an environmental stimulus is processed in controlled or conscious mode when Novelty and attention to the stimulus are large, and in automatic or unconscious mode otherwise. In the model, indirect dopamine (DA) agonists, such as amphetamine or nicotine, enhance the DA representation of Novelty, thereby increasing attention and engaging conscious processing of environmental stimuli. By contrast, DA receptor antagonists, such as haloperidol, reduce the DA representation of Novelty, thereby decreasing attention, and engaging unconscious processing of the stimuli.