Dissecting the Roles of Supervised and Unsupervised Learning in Perceptual Discrimination Judgments

Our ability to compare sensory stimuli is a fundamental cognitive function, which is known to be affected by two biases: choice bias, which reflects a preference for a given response, and contraction bias, which reflects a tendency to perceive stimuli as similar to previous ones. To test whether both reflect supervised processes, we designed feedback protocols aimed to modify them and tested them in human participants. Choice bias was readily modifiable. However, contraction bias was not. To compare these results to those predicted from an optimal supervised process, we studied a noise-matched optimal linear discriminator (Perceptron). In this model, both biases were substantially modified, indicating that the “resilience” of contraction bias to feedback does not maximize performance. These results suggest that perceptual discrimination is a hierarchical, two-stage process. In the first, stimulus statistics are learned and integrated with representations in an unsupervised process that is impenetrable to external feedback. In the second, a binary judgment, learned in a supervised way, is applied to the combined percept.SIGNIFICANCE STATEMENT The seemingly effortless process of inferring physical reality from the sensory input is highly influenced by previous knowledge, leading to perceptual biases. Two common ones are contraction bias (the tendency to perceive stimuli as similar to previous ones) and choice bias (the tendency to prefer a specific response). Combining human psychophysical experiments with computational modeling we show that they reflect two different learning processes. Contraction bias reflects unsupervised learning of stimuli statistics, whereas choice bias results from supervised or reinforcement learning. This dissociation reveals a hierarchical, two-stage process. The first, where stimuli statistics are learned and integrated with representations, is unsupervised. The second, where a binary judgment is applied to the combined percept, is learned in a supervised way.     

Isolated foveal hypoplasia with secondary nystagmus and low vision is associated with a homozygous SLC38A8 mutation

Foveal hypoplasia, always accompanied by nystagmus, is found as part of the clinical spectrum of various eye disorders such as aniridia, albinism and achromatopsia. However, the molecular basis of isolated autosomal recessive foveal hypoplasia is yet unknown. Individuals of apparently unrelated non consanguineous Israeli families of Jewish Indian (Mumbai) ancestry presented with isolated foveal hypoplasia associated with congenital nystagmus and reduced visual acuity. Genome-wide homozygosity mapping followed by fine mapping defined a 830 Kb disease-associated locus (LOD score 3.5). Whole-exome sequencing identified a single missense mutation in the homozygosity region: c.95T>G, p.(Ile32Ser), in a conserved amino acid within the first predicted transmembrane domain of SLC38A8. The mutation fully segregated with the disease-associated phenotype, demonstrating an ∼10% carrier rate in Mumbai Jews. SLC38A8 encodes a putative sodium-dependent amino-acid/proton antiporter, which we showed to be expressed solely in the eye. Thus, a homozygous SLC38A8 mutation likely underlies isolated foveal hypoplasia.     

Interaction of Different Antidepressants with Acute and Chronic Methadone in Mice,and Possible Clinical Implications

We studied the interaction of a single dose of different antidepressant medications with a single (acute) dose or implanted mini-pump (chronic) methadone administration in mice, using the hotplate assay. For the acute experiment, subthreshold doses of six antidepressant drugs were administered separately with a single dose of methadone. The addition of a subthreshold dose of desipramine or clomipramine to methadone produced significant augmentation of the methadone effect with each drug (p?<?0.05). Fluvoxamine given at a fixed subthreshold dose induced a synergistic effect only with a low methadone dose. Escitalopram, reboxetine and venlafaxine given separately, each at a fixed subthreshold dose, induced no interaction. Possible clinical implications of these findings are that while escitalopram, reboxetine and venlafaxine do not affect methadone’s antinociception in mice and are safe to be given together with methadone when indicated, fluvoxamine, clomipramine and desipramine considerably augment methadone-induced effects and should be avoided in this population due to the risk of inducing opiate overdose. For the chromic experiment, when a subthreshold dose of either escitalopram, desipramine or clomipramine was injected to mice following 2 weeks of methadone administration with the mini-pump, none of the antidepressant drugs strengthened methadone’s analgesic effect. Further studies are needed before possible clinical implications can be drawn.     

In vivo atrial excitability and anti-arrhythmic drugs

The threshold of excitability of the atrial muscle was studied in the in vivo beating canine heart. Unipolar cathodal and anodal strength-interval curves were constructed and found to be dissimilar in shape. It was found that at any interval within the relative refractory period of the atrium, as in the ventricle, there is a wide range of current levels delineated by an upper (Tu) and lower (Tl) limit of threshold which can stimulate the atrial myocardium. Within these limits the threshold varies spontaneously and can be reduced to Tl level by a run of extrasystoles.Such Tu and Tl curves were repeatedly determined following administration of therapeutic doses of quinidine, procaine amide or lidocaine. It was observed that all three drugs prolonged the refractory period. The Tu values increased following each of the drugs, and mostly after quinidine, while the Tl curve was less affected by quinidine. It is suggested that the exit block thus produced is the principal mechanism whereby quinidine depresses atrial disrhythmias.     

Immunization with the glutamate receptor-derived peptide GluR3B induces neuronal death and reactive gliosis, but confers partial protection from pentylenetetrazole-induced seizures

Do autoantibodies (Ab's) against glutamate/AMPA receptor subtype 3 affect the severity of seizures? Rats immunized with the GluR3B-peptide (amino acids (aa) 372-395) or with the control GluR3A-peptide (aa 245-274) produced the respective anti-GluR3B and anti-GluR3A Ab's (both types of Ab's found in some epilepsy patients). The GluR3B-immunized rats exhibited neuronal death and reactive gliosis in the brain, but not overt spontaneous seizures. Surprisingly, in response to the chemoconvulsant pentylenetetrazole, the GluR3B-immunized rats displayed fewer jerks, a lower percentage of generalized seizures, and a lower overall seizure-severity score than GluR3A-immunized, scrambled GluR3B-immunized or non-immunized control rats. These findings, combined with the previously demonstrated ability of anti-GluR3B Ab's to bind, activate, and kill neurons and glia, suggest that if these Ab's are present in the brain they may cause neuronal death, which by itself may be pro-epileptic, but they may also decrease the excitability of seizure-related neural circuits, thereby conferring partial protection from seizures induced by other exogenously applied epileptogenic stimuli. The present results could have clinical implications for epilepsy.     

Pitch vs. spectral encoding of harmonic complex tones in primary auditory cortex of the awake monkey

Neuromagnetic studies in humans and single-unit studies in monkeys have provided conflicting views regarding the role of primary auditory cortex (A1) in pitch encoding. While the former support a topographic organization based on the pitch of complex tones, single-unit studies support the classical tonotopic organization of A1 defined by the spectral composition of the stimulus. It is unclear whether the incongruity of these findings is due to limitations of noninvasive recordings or whether the discrepancy genuinely reflects pitch representation based on population encoding. To bridge these experimental approaches, we examined neuronal ensemble responses in A1 of the awake monkey using auditory evoked potential (AEP), multiple-unit activity (MUA) and current source density (CSD) techniques. Macaque monkeys can perceive the missing fundamental of harmonic complex tones and therefore serve as suitable animal models for studying neural encoding of pitch. Pure tones and harmonic complex tones missing the fundamental frequency (f₀) were presented at 60 dB SPL to the ear contralateral to the hemisphere from which recordings were obtained. Laminar response profiles in A1 reflected the spectral content rather than the pitch (missing f₀) of the compound stimuli. These findings are consistent with single-unit data and indicate that the cochleotopic organization is preserved at the level of A1. Thus, it appears that pitch encoding of multi-component sounds is more complex than suggested by noninvasive studies, which are based on the assumption of a single dipole generator within the superior temporal gyrus. These results support a pattern recognition mechanism of pitch encoding based on a topographic representation of stimulus spectral composition at the level of A1.     

A novel homozygous SLC25A1 mutation with impaired mitochondrial complex V: Possible phenotypic expansion

SLC25A1 mutations are associated with combined D,L‐2‐hydroxyglutaric aciduria (DL‐ 2HGA; OMIM #615182), characterized by muscular hypotonia, severe neurodevelopmental dysfunction and intractable seizures. SLC25A1 encodes the mitochondrial citrate carrier (CIC), which mediates efflux of the mitochondrial tricarboxylic acid (TCA) cycle intermediates citrate and isocitrate in exchange for cytosolic malate. Only a single family with an SLC25A1 mutation has been described in which mitochondrial respiratory chain dysfunction was documented, specifically in complex IV. Five infants of two consanguineous Bedouin families of the same tribe presented with small head circumference and neonatal‐onset encephalopathy with severe muscular weakness, intractable seizures, respiratory distress, and lack of psychomotor development culminating in early death. Ventricular septal defects (VSD) were demonstrated in three patients. Blood and CSF lactate were elevated with normal levels of plasma amino acids and free carnitine and increased 2‐OH‐glutaric acid urinary exertion. EEG was compatible with white matter disorder. Brain MRI revealed ventriculomegaly, thin corpus callosum with increased lactate peak on spectroscopy. Mitochondrial complex V deficiency was demonstrated in skeletal muscle biopsy of one infant. Homozygosity mapping and sequencing ruled out homozygosity of affected individuals in all known complex V‐associated genes. Whole exome sequencing identified a novel homozygous SLC25A1 c.713A>G (p.Asn238Ser) mutation, segregating as expected in the affected kindred and not found in 220 control alleles. Thus, SLC25A1 mutations might be associated with mitochondrial complex V deficiency and should be considered in the differential diagnosis of mitochondrial respiratory chain defects.

     

Overlapping parietal activity in memory and perception: evidence for the attention to memory model

The specific role of different parietal regions to episodic retrieval is a topic of intense debate. According to the Attention to Memory (AtoM) model, dorsal parietal cortex (DPC) mediates top-down attention processes guided by retrieval goals, whereas ventral parietal cortex (VPC) mediates bottom-up attention processes captured by the retrieval output or the retrieval cue. This model also hypothesizes that the attentional functions of DPC and VPC are similar for memory and perception. To investigate this last hypothesis, we scanned participants with event-related fMRI whereas they performed memory and perception tasks, each comprising an orienting phase (top-down attention) and a detection phase (bottom-up attention). The study yielded two main findings. First, consistent with the AtoM model, orienting-related activity for memory and perception overlapped in DPC, whereas detection-related activity for memory and perception overlapped in VPC. The DPC overlap was greater in the left intraparietal sulcus, and the VPC overlap in the left TPJ. Around overlapping areas, there were differences in the spatial distribution of memory and perception activations, which were consistent with trends reported in the literature. Second, both DPC and VPC showed stronger connectivity with medial-temporal lobe during the memory task and with visual cortex during the perception task. These findings suggest that, during memory tasks, some parietal regions mediate similar attentional control processes to those involved in perception tasks (orienting in DPC vs. detection in VPC), although on different types of information (mnemonic vs. sensory).