neuro and furture exam 4 lecture notes

numerosity

the perception of the number of items in a set

millers “Law” (1956)

people can only remember about 5-9 pieces of information at a time

a continuous code

if small and large numbers are represented by the same neuronal mechanism.

two different codes

If small and large numbers engage in distinct mechanisms, then discontinuity reflecting the change from one mechanism to the other is expected.

what is the general approach (probe)

It records single-neuron activity in the medial temporal lobe (MTL)

Support vector machine (SVM) classification analysis

finds the best possible curve separating the open circles from the filled circles, called the decision boundary, and the one that gives the biggest space between the two groups is called the margin.

State space

Imagine each neuron’s activity as a number. if you have many neurons, you can think of their combined activity as a point in a multi-dimensional space. as time passes, the point moves tracing a trajectory through space.

each dimension in a state-space

one neurons activity

each point in a state-space

the state of the whole population at a moment in time

Each colored line in a state space

It represents how a group of neurons changes its activity over time during different tasks or conditions

dimensionality reduction

a 3D data set being reduced to a 2d scatterplot.

What are two points about the subitizing system?

tuning selectively and surround suppression

What was the main point of the study?

when they asked participants to judge the number of items in a set, they should a behavioral dichotomy where they responded fast and accurately for small numbers up to 4.

When participants answered to items less than four it was _____, but if it was more than 4 it was______.

to numbers less than 4 it was subitzing as it was fast and accurate and if it was more than 4 it was an estimation as it was slower and imprecise.

what type of participants did they use for the study?

they recorded single-neuron activity in the medial temporal lobe (MTL) of neurosurgical patients who judged numerical quantity.

What was the parity task used for?

to test a broad range of explicit number representations devoid of other cognitive factors such as working memory in a short time.

what was the response of the neurons to certain numbers

they responded strongest to their respective numerosity and decreased their activity progressively with increasing numerical distance

what did they use the Mann-Whitney U test for?

to exclude neurons responsive to parity judgments

What did they find in proportions of neurons with a main effect for a number across different areas in the medial temporal lobe?

There is a significant main effect exclusively for numbers (15%)

What did they find in the proportion of neurons tuned to different preferred numbers and the firing rates?

the individual neurons fire most strongly to a specific prefered number and decrease their activity as the number and decrease their activity as the numbers move further away from that preferred value.

What did they find in surround suppression in neurons tuned to small numbers(0-3)?

They exhibit significant surround suppression, and the firing rates drop below baseline activity for non-preferred numbers.

What did they find in surround suppression in neurons tuned to large numbers (4-9)?

They would return to baseline

when analyzing the tuning curves, which aspects did they include?

The amplitude (response strength), variance (response precision), and firing rate.

What did the z-scores show the researchers?

The firing rates between all pairs of numbers for number-selective neurons

What did they hypothesize in terms of the firing rates and correlation coefficients from the z-scores for small numbers?

They would show more similar firing rates to other small numbers and thus higher correlation coefficients.

What did they hypothesize in terms of the firing rates and correlation coefficients from the z-scores for large numbers?

they would show higher correlation coefficient values suggesting radically different coding for numerosity, but also categorical differences between small and large countable numbers

What is the purpose of the confusion matrix?

it tries to predict which number was shown based on the firing patterns. The grid shows that the model frequently confused numbers within small and large numbers but did not mix them across categories.

What was the purpose of the boundary test?

it found that the sharpest, most significant split in the neural data happens between 4 and 5.

what did the state-space trajectories show?

The numbers 0-4 stayed grouped together and numbers 5-9 formed another group together.

What did the k-means clustering show?

without telling the computer which numbers were small or large (unsupervised), they naturally sorted all the data and sorted them in exactly 2 clusters with one holding 0-4 and the other holding 5-9/