People can report summary statistics for various features about a group of objects. One theory is that different abilities support ensemble judgments about low-level features like color vs. high-level features like identity. Existing research mostly evaluates such claims based on evidence of correlations within and between feature domains. However, correlations between two identical tasks that only differ in the type of feature for ensemble judgments can be inflated by method variance. Another concern is that conclusions about high-level features are mostly based on faces. We used latent variable methods on data from 237 participants to investigate the abilities supporting low-level and high-level feature ensemble judgments. Ensemble judgment was measured with six distinct tests, each requiring judgments for a distinct low-level (orientation, lightness, aspect ratio) or high-level feature (bird species, Ziggerin identity, Transformer identity), using different task requirements in each task (mean estimation, mean matching, diversity comparison). We also controlled for other general visual abilities when examining how low-level and high-level ensemble abilities relate to each other. Confirmatory factor analyses showed a perfect correlation between the two factors, suggesting a single ability. There was a strong unique relationship (r = .9) between these two factors, beyond the influence of object recognition and perceptual speed. Similar results were obtained controlling for working memory. Our results demonstrate that the ability common to a variety of ensemble judgments with low-level features is the same as that common to a variety of ensemble judgments with high-level features.
9:30 - 10:00 AM
- Gary Cottrell "Euclidean coordinates are the wrong prior for models of primate vision"
Convolutional Neural Networks (CNNs) are currently the best models we have of the ventral temporal lobe – the part of cortex engaged in recognizing objects. They have been effective at predicting the firing rates of neurons in monkey cortex, as well as fMRI and MEG responses in human subjects. They are based on several observations concerning the visual world: 1) pixels are most correlated with nearby pixels, leading to local receptive fields; 2) stationary statistics – the statistics of image pixels are relatively invariant across the visual field, leading to replicated features 3) objects do not change identity depending on their location in the image, leading to pooling of responses, making CNNs relatively translation invariant; and 4) objects are made of parts, leading to increasing receptive field sizes in deeper layers, so smaller parts are recognized in shallower layers, and larger composites in later layers. However, compared to the primate visual system, there are a couple of striking differences. CNNs have high resolution everywhere, whereas primates have a foveated retina, with high resolution for humans only about the size of your thumbnail at arm’s length, and steep dropoff in resolution towards the periphery. The mapping from the visual field to V1 is a log-polar transform. This has two main advantages: scale is just a left-right translation, and rotation in the image plane is a vertical translation. When these are given as input to a standard CNN, scale and rotation invariance is obtained. However, translation invariance is lost, which we make up for by moving our eyes about 3 times a second. We present results from a model with these constraints, and show that, despite rotation invariance, the model is able to capture the inverted face effect, while standard CNNs do not.
10:00 - 10:30 AM
- Break at Massachusetts Hall
10:30 - 11:00 AM
- Tom Palmeri "Using cognitive models and CNNs to understand individual differences in visual cognition"
11:00 - 11:30 AM
- Jim Tanaka "Shadows on the cave wall: Representational space, category learning and performance"
In his famous allegory, Plato describes a scene in which prisoners are chained together in a line and positioned in such a way they can only see the shadows of objects cast by a burning fire. From these shadow forms, the prisoners make inferences about the nature of reality. Similarly, for cognitive scientists, we do not have direct access to the workings of the human mind but make inferences about its contents and operations based on its outward manifestations, such as brain activity, response time and detection thresholds. In the Different Minds Lab, we have employed the PsiZ platform developed by Roads & Mozer (2019) to bridge the gap between mind and behaviour. PsiZ is a set of machine learning tools that transforms an observer’s similarity judgments into an internal representation (i.e., psychological embedding) of their category space. In my talk, I will describe three PsiZ projects where we investigate the psychological embeddings of participants who are trained in recognizing species of Warbler birds, students who are enrolled in an introductory geology course and people who are experts and novices in the domain of NBA basketball. The overarching goal of our work is to draw a tighter connection between the categorizer’s internal representation and their external category behaviors.
- Lunch at Thorne Dining Hall
1:00 - 2:00 PM
- Maria Kharitonova and Matt Mollison "Careers in data science"
A discussion of types of jobs, training, and career trajectories in data science. Open to STEM students on campus.
2:00 - 2:30 PM
- Heida Sigurdardottir "Why is a raven like a writing desk? The role of semantics, high-level visual information, and low-level visual information in object discrimination"
Why is a raven like a writing desk? The answer to this riddle, famously posed by the Mad Hatter in Alice’s Adventures in Wonderland, may simply be that it is not. A raven is a living, moving, behaving, breathing being. A writing desk is none of those things. However, objects differ in several ways, and these are often hard to tease apart. Not only are ravens and writing desks conceptually different, but they also look nothing alike. And if ravens looked like something else, then they would hardly be ravens anymore. In this talk, I will tell you about some of our work on how we are trying to disentangle semantics, high-level visual information, and low-level visual information, to estimate their role in visual object perception.
2:30 - 3:00 PM
- Lisa Scott "Out of the brains of babes"
Infants are immersed in a complex world and use domain- and task-relevant associative learning mechanisms to form appropriate and useful representations of their world. To learn more about this process, we examined a cross-sectional sample of parent-infant dyads at 6, 9, and 12 months of age both during and after they read a short book together. The book included novel objects labeled with individual names, category labels, or no labels. During infant-parent shared book reading joint attention and infant attention were measured using dual head-mounted eye-tracking. Infants and parents then returned the next day and visuocortical responses were measured during an EEG frequency tagging task in which dyads viewed the trained objects across label conditions (i.e., objects from the book) concurrently with spatially overlapping novel objects that varied in similarity (high, medium, low) from the trained objects. Dual eye-tracking results show increased infant attention and joint attention for 6- compared to 9- and 12-month-old infants and an overall longer duration of joint attention for labeled compared to unlabeled objects. For the frequency tagging EEG task, infant signal-to-noise ratios (SNR) were examined using several models. Infant SNRs were greater for trained relative to untrained stimuli at 9 and 12 months of age and the difference between trained and untrained stimuli increased from 6 to 9 to 12 months of age. This increase is best explained by a model of neural competition (relative to a model of specific amplification). An analysis of the infant SNR values across similarity revealed a pattern of neural sharpening for all conditions and additional evidence of neural generalization for objects labeled at the category-level. Finally, an analysis of the simultaneous parent-infant dual EEG data demonstrated increased synchrony for untrained stimuli relative to trained stimuli, with parent occipitotemporal regions showing high synchrony with the whole scalp of infants. These findings highlight the domain- and task-relevant nature of early learning such that learning from shared book reading results in (1) competitive visuocortical responses to concurrently presented trained and untrained objects and (2) differential patterns of neural sharpening and generalization depending on label condition.
3:00 - 3:30 PM
- Conor Smithson "Individual differences in object recognition ability: Measurement and the nomological network"
Individuals differ considerably in their ability to process and recognize objects. While much research has manipulated or exploited the effects of varying levels and kinds of experience with a particular object category, differences in recognition accuracy exist even in the absence of differing experience levels. This tendency for some to recognize objects more aptly than others remains consistent across diverse object categories. Understanding such individual differences may be key to understanding the development of expertise. To begin, I discuss the development of tools to measure object recognition ability. Then, I describe how these measures have made it possible to draw upon powerful methods from psychology’s correlational tradition to further explore the generality of this ability across modalities, its stability over time, and its relationships with other cognitive and perceptual abilities.
3:30 - 4:00 PM
- Rosie Cowell "The effects of normal aging on visual recall of objects and scenes
Amnesic patients suffer impaired explicit recall but preserved priming (e.g., word-stem completion). This is interpreted as impaired effortful retrieval, with preserved implicit retrieval. But this interpretation is incompatible with representational accounts of memory, which suggest that circumscribed cognitive processes (e.g., recollection) provide inadequate labels – the wrong “explanatory currency” – for the deficits caused by brain damage. Instead, representational accounts claim that brain damage impairs not “recall ability” per se, but the ability to solve any cognitive task that relies on representations residing in the damaged region, e.g., hippocampus. Thus, in amnesics, recall deficits should be reduced if the to-be-remembered content does not rely on hippocampal representations. We tested this in older adults, who often suffer incipient loss of hippocampal function. We used a non-associative recall task in which images (of objects and scenes) were studied in isolation, and circular image patches were used as retrieval cues. In a previous fMRI study using this task, patch-cued recall of scenes induced hippocampal activation, but patch-cued recall of objects did not, presumably because scenes but not objects rely on hippocampal representations. We therefore predicted that older adults would be more impaired at patch-cued recall of scenes than of objects. This is the pattern we found. We conclude that age-related impairments in recall depend not on whether retrieval is explicit or implicit, but on the content of the to-be-retrieved memory.
4:00 - 4:30 PM
A theory and neurocomputational model are presented that explain grid cell responses as the byproduct of equally dissimilar hippocampal memories. On this account, place and grid cells are not best understood as providing a navigational system. Instead, place cells represent memories that are conjunctions of both spatial and non-spatial attributes, and grid cells primarily represent the non-spatial attributes (e.g., odors, surface texture, etc.) found throughout the two-dimensional recording enclosure. Place cells support memories of the locations where non-spatial attributes can be found (e.g., positions with a particular odor), which are arranged in a hexagonal lattice owing to memory encoding and consolidation processes (pattern separation) as applied to situations in which the non-spatial attributes are found at all locations of a two-dimensional surface. Grid cells exhibit their spatial firing pattern owing to feedback from hippocampal place cells (i.e., a hexagonal pattern of remembered locations for the non-spatial attribute represented by a grid cell). The model explains: 1) grid fields that appear to be centered outside the box; 2) the toroidal nature of grid field representations; 3) grid field alignment with the enclosure borders; 4) modules in which grid cells have the same orientation and spacing but different phases; 5) head direction conjunctive grid cells that become simple head direction cells in the absence of hippocampal feedback; 6) the instant existence of grid fields in a novel environment; 7) the slower learning of place cells; 8) the manner in which head direction sensitivity of place cells changes near borders and in narrow passages; 9) the kinds of changes that underlie remapping of place cells; and 10) grid-like responses for two-dimensional coordinate systems other than navigation.
4:30 - 5:00 PM
- Tim Curran "ERP studies of acute influences of THC and CBD on recognition memory"
Previous research has documented acute harmful effects of cannabis use on verbal episodic memory, but prior work has not sufficiently considered that the memory effects of cannabis are the compound action of different cannabinoids acting on different memory processes. Specifically, cannabidiol (CBD), a non-psychotomimetic component of cannabis (doesn’t produce a “high”), is thought to have cognitively protective properties and may mitigate some harmful effects of ∆9-tetrahydrocannabinol (THC). Preliminary data, including our own, suggest that THC and CBD render differential effects on memory. Further, few prior studies have tested high potency strains that are commonly available. Our global hypothesis is that the effects of cannabis on memory vary as a function of the ratio of CBD to THC, with THC having adverse effects that may be counteracted by CBD. We will present preliminary results from a new study to test the effects of three real-world commercially available cannabis strains that differ markedly in their ratio of CBD to THC. To that end, we will test the effects of -THC/+CBD (0% THC/16% CBD), +THC/-CBD (16% THC/0% CBD), and +THC/+CBD (16% THC/16% CBD) strains on recognition memory as well as event-related brain potentials (ERPs) that have previously been found to be related to different underlying memory processes. We use a naturalistic observational design in which each participant will complete the same memory task while intoxicated one day and not intoxicated another day (order counterbalanced). We will assess recognition memory performance and memory-related ERP components in cannabis users after self-administration of one of three randomly assigned cannabis strains (+THC/-CBD vs. -THC/+CBD vs. +THC/+CBD) during both memory encoding (learning) and memory retrieval. We hypothesized a step wise effect of strain such that the +THC/-CBD group will demonstrate the largest decrement in memory accuracy, as compared to the +THC/+CBD group, which will show a larger memory decrement than the -THC/+CBD group. In addition to strain assignment, CBD and THC blood levels will also be tested in relation to memory accuracy, with greater CBD/THC levels associated with higher/lower memory accuracy.
5:00 -7:00PM (EST)
- Dinner at Cram Alumni House