ozimac

[ozimac]

Problem Solving, Tool Use, & Mental Imagery

Mental imagery in animals: Learning, memory, and decision making in the face of missing information, Blaisdell, A. P. (2019)

Intelligences and brains: An evolutionary bird’s-eye view, Delius & Delius (2012)

Tool-related cognition in New Caledonian crows, Bluff et al. (2007)

The review article by Blaisdell (2019) discusses mental imagery in animals, particularly focusing on learning, memory, and decision-making processes when faced with missing information. The introduction highlights that mental imagery is not just used by children during play (which is also important!) but is also a crucial element in the development of cognition, problem-solving, and empathy in both children and adults. It poses the central question of whether animals possess minds that function similarly to humans and discusses the importance of investigating cognitive processes shared between humans and nonhuman animals. Blaisdell defines mental imagery as the ability to maintain sensory/perceptual representations of events or objects in the absence of actual sensory input. This involves neural processes similar to those used in sensation and perception. The article reviews literature on mental imagery in animals, primarily focusing on visual imagery. It discusses various experiments exploring mental rotation tasks in animals, such as pigeons (e.g., Hollard & Delius, 1982), which yielded intriguing results suggesting differences in cognitive processing compared to humans.

Furthermore, the article explores associative processes involved in mental imagery and decision-making. Animals, including humans, learn about their environment through associative and causal learning. Blaisdell discusses studies demonstrating how animals infer hidden causes and make decisions based on associative processes. For instance, rats show recognition of ambiguous cues and can distinguish between explicit and ambiguous absence of events, indicating their ability to reason about missing information (e.g., Blaisdell et al., 2009; Fast and Blaisdell, 2011; Blaisdell and Waldmann, 2012). The review also delves into the neural basis of reasoning about missing information, particularly focusing on the role of the hippocampus. Studies suggest that the hippocampus is critical for processing ambiguous stimulus events and is recruited during mental imagery tasks in both humans and animals. This suggests a common neural substrate underlying memory, imagination, and the simulation of future events. 

It is proposed that associative retrieval of event representations can influence learning and decision-making, similar to how humans form mental images that drive behavior. However, alternative nonrepresentational accounts have also been suggested, challenging the idea of mental imagery as the driving force behind decision-making (however, there is more evidence for the representational account!). Specifically, experiments, such as those by Fast, Biedermann, and Blaisdell (2016), have provided evidence supporting the representational account. This raises questions about the psychological basis of these representations, whether they are mere expectations, memories, or possibly even mental images. The article discusses the implications of these findings for understanding the evolution of causal reasoning and the unique elements of human thought processes compared to other species. Additionally, it explores how the ability to differentiate between real events and mentally retrieved representations contributes to learning and inference processes, offering a clue into the fundamental nature of reasoning.

-

The chapter "Intelligences and brains: an evolutionary bird’s-eye view" by Delius & Delius (2012)* provides an interdisciplinary overview of intelligence across various animal species, including humans. It begins by discussing the varying definitions of intelligence, highlighting its association with the ability to adapt behavior to novel situations. While human intelligence has been extensively studied, animal intelligence, although often overlooked, has been studied in various species. Researchers have found evidence of individual differences in cognitive abilities among animals, with some individuals performing better than others on certain tasks. For example, tamarin monkeys and songbirds have shown variations in cognitive performance (Banerjee et al., 2009), suggesting the presence of a general intelligence factor. Learning sets and concept formation tasks have been used to assess intelligence in animals which reveal differences in learning abilities across species. 

Transitive responding, rotation invariance, and social intelligence are also discussed as components of animal intelligence. These abilities are believed to have evolved over millions of years, driven by selective pressures and environmental factors. Birds, in particular, exhibit complex neural structures that facilitate sophisticated information processing operations, despite lacking the layered organization of the mammalian neocortex. Finally, the chapter also explores the evolution of human intelligence, attributing it partly to social factors such as the demands of cooperative hunting and social interaction. It discusses practical intelligence and general intelligence as key components of human cognitive abilities, emphasizing the role of working memory in problem-solving tasks. Furthermore, it suggests that the development of language competence in humans may have contributed to the evolution of mental representation and conscious awareness of thought processes.

-

The article by Bluff et al. (2007) provides a comprehensive overview of tool-related cognition in New Caledonian crows. It spans six years of research, primarily focusing on experimental studies with captive crows in Oxford. The authors address the challenges that come with uncovering aspects of physical understanding and reasoning in non-human species. While tool-related behavior (TOB) is not necessarily indicative of unusually sophisticated cognition, it gives information into the cognitive processes involved in animals' manipulation of physical objects.

New Caledonian crows are unique among non-mammals in their ability to make and use various tools, both in the wild and in captivity. The authors explore hypotheses regarding the development of TOB, considering factors such as genetic predispositions, advanced reasoning abilities, and social learning. Through experiments with captive crows, they investigate the development of tool use and manufacture. The article also delves into the cognitive mechanisms behind the use of TOB in adult crows, exploring their understanding of physical causality and their ability to select appropriate tools for novel tasks. While some evidence suggests a degree of causal understanding in these birds, particularly demonstrated through tasks like the trap-tube test and hook making experiments, the authors are cautious to extend these findings to the entire Corvidae family. They emphasize the need for further research to understand the evolutionary origins and selective pressures shaping tool use behavior in New Caledonian crows.

My thoughts

To begin, the Blaisdell (2019) review article is a great resource for previous literature involving mental imagery in nonhuman animals. Additionally, in the article, it mentions the role of mental imagery in causal relations and hidden causes. A lot of the time, not all relevant information is perceptually available to us, thus we must make inferences about hidden causes. That being said, given my lab’s previous work on both causality and mental imagery, I wonder how the rats would respond to an ambiguous cue that starts off a causal chain. Akin to Blaisdell, Sawa, Leising, & Waldmann (2006)’s causal chain and Blaisdell et al. (2009)’s sensory preconditioning procedure, instead of order tone -> light -> food, we could use light -> tone -> food. Then, at test, along with a control, uncovered condition, the rats receive a session where the light is occluded via a metal block. Will the rats still predict food after the tone (as seen by nose poke behavior), or will they not correlate the tone with food because it is not predicted by the light? Given the results of all of the Blaisdell lab’s mental imagery studies series, along with previous pilot work I’ve done in the lab, I would suspect that they would believe the light was on underneath the block and that it caused the tone which then causes food to be dispensed. Or would they only form a direct association between the tone and the food and not worry whether the light was present at all (because food reward is salient enough)?

One of the main takeaways I got from the Delius & Delius (2012) chapter is the idea of control by systematic variation in comparative studies. That is, using functional relationships between variables across the different species that you are studying in order to maximize attention and motivation to the task at hand. As stated throughout a number of studies mentioned in the chapter (e.g., Warren (1973)’s evidence of little learning-to-learn ability in rats using visual cues, whereas Slotnik & Katz (1974)’s evidence of considerable learning-to-learn ability in rats using odor stimuli), it is important to assess the particular species that you are studying by the most appropriate modality (i.e., vision, odor, auditory). It is unfair to test an animal and deem it incapable of something because you are assessing them in a way that is not catered to their specific sensory capabilities. Another thought I had while reading this article is in reference to the individual abilities across all species. We definitely see this in our pigeons via individual differences in our experimental tasks. Some need to be dropped from certain experiments, and ones that do well in one type of experiment may not do as well in another. I think a good mindset shift that I had recently that had come as a result of my lab’s conversations, is understanding that our experiments are not necessarily always conducted under the premise of do pigeons do x, but rather can pigeons do x. Along with that, what strategies potentially are birds using to achieve success in doing x, and do those strategies differ depending on bird?

In the Bluff et al. (2007) article, it discusses the only New Caledonian crow (Betty) to be tested using the trap-tube test, which is designed to test for causal understanding in the physical domain. It says she didn't instantly solve the task, but improved gradually, however it also states: “After around 60 trials, she spontaneously developed a two-stage technique. This involved first inserting the stick in the tube from the ‘safe’ side (leaving the trap between herself and the food) until its distal end protruded from the opposite side, then walking to the opposite side and pulling the stick so as to extract the food in a controlled manner.” Thus, would this “spontaneous” strategy development be considered indicative of insightful behavior?

*Chapter from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press.

Blaisdell, A. P., Sawa, K., Leising, K. J., & Waldmann, M. R. (2006). Causal reasoning in rats. Science (New York, N.Y.), 311(5763), 1020–1022. https://doi.org/10.1126/science.1121872

[ozimac]

Variability, Creativity, and Insight

Do animals have insight, and what is insight anyway?, Shettleworth, S. A. (2012)

The Law of Expect or a Modified Law of Effect? Outcome expectation and variation in learned behavior, Blaisdell, A. P., Stolyarova, A., & Stahlman, W. D. (2016)

There is room for conditioning in the creative process: Associative learning and the control of behavioral variability, Stahlman, W. D., Leising, K. J., Garlick, D., & Blaisdell, A. P. (2013)

The chapter "Do animals have insight, and what is insight anyway?" by Shettleworth (2012) discusses the concept of insight both in human and animal cognition. It begins by discussing historical studies, notably Kohler's experiments with chimpanzees, which introduced the idea of insight in comparative psychology. Insightful problem-solving is defined by sudden solutions to problems after an impasse, accompanied by a subjective feeling of surprise and restructuring of the problem. This contrasts with gradual, analytical problem-solving approaches.The discussion then transitions to contemporary research on human insight, emphasizing the importance of restructuring problems and the role of past experiences in problem-solving. Brain imaging studies are also mentioned as a tool to understand the neural correlates of insight. However, there is no consensus on the best approach to study insight in humans.

In the section on animal insight, historical experiments with primates like chimps and birds are examined, particularly in tasks like string pulling and tool-making. While some behaviors initially seemed to indicate insight, further research suggested that these behaviors could often be explained by associative learning or simple trial and error. Contemporary studies with birds and rodents also question whether sudden changes in behavior truly indicate insight or simply reflect different learning mechanisms.

The chapter concludes by highlighting the need for clearer definitions and methodologies in studying animal insight. Recent research with rodents suggests that insight may not be exclusive to certain species or complex tasks, opening avenues for comparative studies that link behavioral observations with neural mechanisms. This integration could provide further insights into the nature of insight across species and on the evolution of cognitive processes.

-

The article "The law of expect or a modified law of effect? Outcome expectation and variation in learned behavior" by Blaisdell et al. (2016) explores the role of outcome expectation in modifying behavioral variation, proposing a new theoretical framework called the Law of Expect. Traditionally, learning theory has focused on selection processes rather than behavioral variation mechanisms. However, the authors argue that outcome expectation plays a crucial role in shaping learned behavior. They present empirical evidence supporting the idea that the expectation of an outcome modulates response variation, with higher expectations leading to less variability and lower expectations resulting in more variability.

The authors conducted several empirical studies across different species, including rats and pigeons, to investigate the relationship between outcome expectation and behavioral variation. They found that manipulations affecting reward expectation, such as reward magnitude and delay to reward, systematically influenced response variability. Additionally, they also suggested the Modified Law of Effect, which incorporates mutual inhibition between potential responses and replaces the weakening of S-R connections from nonreward with increasing inhibitory S-R connections. This modified framework successfully accounted for many empirical phenomena involving behavioral variation. However, the Modified Law of Effect was unable to explain anticipatory contrast effects observed in mammals, suggesting that outcome expectation might still have more explanatory power in certain contexts. Despite its limitations, the Law of Expect provides insights into how individuals balance exploration and exploitation in learning situations. The authors conclude by highlighting the importance of further research into the psychological and neural mechanisms underlying results indicated by the models of both Laws of Effect and the Law of Expect to gain a deeper understanding of learning processes.

-

The chapter “There is room for conditioning in the creative process: associative learning and the control of behavioral variability” by Stahlman et al. (2013) explores the mechanisms driving creativity, particularly focusing on how associative conditioning shapes novel and variable behavior in animals. They argue that creativity is not exclusive to humans but is evident in everyday actions of animals facing survival challenges. The authors propose three key principles, emphasizing the role of associative learning mechanisms in generating novelty and variability in behavior. They challenge the traditional view that behavior is solely determined by genetics and environmental history, highlighting the often-overlooked variability inherent in spontaneous behavior.

The chapter discusses studies demonstrating how animals can be trained to produce novel actions through explicit reinforcement of variability. It also examines how expectation influences behavioral variability, suggesting that animals exhibit greater variability when reinforcement is unlikely. Additionally, the neurobiological basis of variability and stereotypy is explored, emphasizing the involvement of basal ganglia and reward circuitry. Understanding the mechanisms underlying novel and variable action in both human and nonhuman animals is crucial for a comprehensive understanding of creativity and its adaptive significance.

My thoughts

As mentioned in all three of these readings, there is undeniable evidence that suggests that behavioral variation increases markedly during extinction. This makes me wonder if this behavioral variation coincides with this “activity burst” as seen in the pigeon art project*? For potential readers who may not know, our lab is currently studying pigeon art creation; where a “canvas” screen is given to our experimental pigeons after their daily experimental session where pecks to the screen create different lines and shapes in different colors. They are not reinforced for these pecks and seem to be intrinsically motivated. Perhaps, the birds are motivated to respond to the screen (e.g., with much spatial variability) after losing the food reward that they just had when make correct choice pecks to the screen once they’ve completed their experimental session for the day. This burst in behavior (i.e., pecking on the screen) is caused by some sort of extinction (i.e., loss of food reward for instrumentally responding to the screen via pecking) process, but then this behavioral variability is what facilitates this new, creative process (i.e., creation of “art”) that is, then, intrinsically satisfying to them so they continue to do it, despite the lack of reinforcement. Thus, initially, they began interacting with the “canvas” because of this reward extinction, but then they “enjoy” (for lack of a better term) the creative process of the creation of art and continue to do it, as seen oftentimes by the long amount of time spent pecking at the canvas post-experiment. That being said, we should also observe and make note of their other behaviors outside of screen interaction when the canvas is up (e.g., walking around the operant box, preening, looking around, turning around, etc.) to see if there is any consistent behaviors across sessions or even stereotyped behaviors. 

The Stahlman et al (2013) chapter also states that “[t]he observation that vari­ability tends to decrease with approach to reinforcers certainly suggests that reinforcement interferes with production of novel behavior”. This leads me to think that if we did reinforce their art production, we wouldn’t have the beautiful arrays of art that we see as created by the spatially variable (i.e., widely distributed) pecks across the entire screen. That is, if we did reinforce their behavior, my guess (as seen in the results of the instrumental condition in Stahlman, Young, & Blaisdell; 2010) is that their pecks would be concentrated in only one portion of the screen. 

We also saw this variable behavior during extinction in a rat pilot study we conducted this past year. In this study, we used a 2-lever choice task to study a previous finding from our lab of a conjunction fallacy committed by rats. In this task (Gonzalez et al., 2023), we concurrently trained rats on a feature positive instrumental discrimination involving sound A and light X (A-/AX+) and on a feature negative instrumental discrimination involving sound B and light Y (B+/BY-) using just one lever. During a nonreinforced (extinction) test session where just sounds A and B were presented, the light was occluded via a metal block that made it ambiguous as to whether the light underneath the block was on or off. Results showed that rats overwhelmingly assumed the light was on (as seen by lever pressing on A trials and withholding of LP during B trials), therefore committing a conjunction fallacy (assuming the trials were "compound" [AX+, BY-] trials). Last year, we attempted to replicate this result using two levers (A -> Right lever, AX -> Left, B -> Left, and BY -> Right), although initially, our extinction test results wherein the light was covered showed about the same amount of lever pressing for both cues. Thus, a possible explanation for these results that our lab discussed was this increased behavioral variability (resurgence) that has been consistently seen during extinction. Perhaps, the rats had initially committed the fallacy (i.e., pressing of the left lever during the A cue and right lever during the B cue) or not (vice versa) in the beginning of the first extinction trials, but then immediately began pressing the other lever once they had realized that they were no longer being reinforced – directly evidencing behavioral variation during extinction. A caveat to this study, which I will make sure to remedy in the future, is that we were limited to data of just the total LP per each lever per trial and did not program into Med-PC (unfortunately, if you know, you know) time stamps as to when each lever was pressed, which would show us their initial lever choice before the extinction behavior burst. However, we did see the fallacy once we only presented one of the levers at test (i.e., if left lever out, only LP for A trials, and if right lever out, only LP for B trials).

Another thought I had while reading the Blaisdell et al. (2016) article involves Gharib et al. (2004)’s description of the adaptive functionality of the relationship between variability and reward expectation: “If an animal’s actions vary too little, it will not find better ways of doing things; if they vary too much, rewarded actions will not be repeated. So at any time there is an optimal amount of variation, which changes as the costs and benefits of variation change. Animals that learn instrumentally would profit from a mechanism that regulates variation so that the actual amount is close to the optimal amount. (p. 271).” This reminds me of last week’s readings and the logical errors discussed by Terrace (2010) that I mentioned in my last post. Evolutionarily, it makes sense (adaptive functionality) of these logical errors made through their variation in behavior in order to succeed in solving a problem. Again, stated in the article, “[l]ikewise, if one initially successful method of extracting embedded food (mollusks buried under sand, grubs hiding in the bark of a tree, etc.) stops being successful, perhaps changing the form of the extraction response (e.g., digging more forcefully, twisting the beak into a crack in the bark a different way) may increase the probability of success.” I remember using a similar guess and check kind of strategy in computer games like Poptropica as a kid, where I would just click on anything and everything (objects, people in the game) to see if I could gain some sort of information or hint as to how I could solve whatever problem or task I was currently facing. Whereas, when I knew exactly what I needed to do to solve a problem, my “mouse clicking” behavior was straightforward and succinct (e.g., one mouse click per object that was necessary in solving the task/problem).

González, V., Sadeghi, S., Tran, L., & Blaisdell, A. (2023). The conjunction fallacy in rats. Psychonomic bulletin & review, 30. https://doi.org/10.3758/s13423-023-02251-z

*Check out our pigeons' artwork here!

[ozimac]

Pattern Learning

The organization of sequential behavior: conditioning, memory, and abstraction, Fountain et al. (2012)

The comparative psychology of ordinal knowledge, Terrace (2010)

The chapter "The organization of sequential behavior: conditioning, memory, and abstraction" by Fountain et al. (2012)* discusses the processes  that underly sequential behavior in rats. It begins by presenting evidence that rats not only encode sequential items but also abstract rules for sequences, like hierarchical patterns using chunking processes and phrasing cues. The chapter delves into various models of sequential behavior, highlighting the debate surrounding rule-learning vs. associative mechanisms. It suggests that sequential learning likely involves multiple concurrent processes rather than a single overarching theory.

A significant portion of the chapter explores experiments conducted with rats using a serial multi-choice task. This task involves rats learning to choose items in a proper sequential order in order to get reinforcement, demonstrating their ability to navigate complex serial patterns utilizing multiple cues and behavioral processes simultaneously. Additionally, the chapter examines how rats abstract rules from hierarchical and interleaved patterns, wherein they are often successful. There is also a role that serial position in item memory are also discussed, which tells how rats learn and remember these sequential patterns. The chapter further explores neurobehavioral studies, particularly the effects of drugs on sequential learning, revealing insights into the underlying brain mechanisms involved. Finally, it concludes by emphasizing rats' ability as multimodally sequential processors and discusses the broader implications of this research for understanding sequential behavior, cognitive processing, and neural circuits.

-

The review chapter "The comparative psychology of serially organized behavior" by Terrace (2010) explores the phenomenon of serial learning, which is fundamental to various aspects of behavior, from basic skills like navigation to complex abilities like language comprehension. It critiques traditional chaining theory, highlighting its inability to account for the relationships between non-adjacent items in serial behavior, as observed in language and other cognitive tasks. Recent research demonstrates that non-human primates possess a capacity for acquiring ordinal knowledge, challenging previous assumptions about the limits of non-verbal animals' cognitive abilities.

Terrace discusses experiments employing simultaneous chaining paradigms in monkeys, which shows their ability to learn arbitrary sequences without verbal cues. These studies suggest that monkeys form cognitive maps akin to spatial representations to navigate through sequences, a capability observed predominantly in non-human primates. Furthermore, investigations into serial expertise indicate that monkeys can generalize strategies across different lists, which is indicative of the ability to form general learning strategies rather than memorization of specific sequences. The chapter also explores the distinction between ordinal and cardinal knowledge, with evidence supporting non-human primates' proficiency in learning the ordinal properties of numbers. Studies on symbolic distance and magnitude effects further show monkeys' ability to judge relative magnitudes and ordinal positions, suggesting complex cognitive mechanisms for serial processing in these animals.

Finally, neurological investigations indicate the involvement of brain regions such as the hippocampus and parietal lobe in ordinal knowledge processing, suggesting commonalities between human and non-human primate cognition. These findings highlight the evolutionary significance of non-verbal mechanisms for learning arbitrary sequences and numerical rules, which allow us to consider the origins of human cognitive abilities such as counting and language comprehension. Overall, the chapter provides compelling evidence for the cognitive complexity of non-human primates in serially organized behavior.

My thoughts

What these two articles highlight the most for me is the role of the brain as a predictive organ. What makes sense to me is that, of course, we have evolved to recognize patterns and, like seen with many of the rat studies in the Fountain et al. (2012) article (e.g., Fountain et al., 2006), it would be the most efficient for organisms to follow a pattern until it is violated and then go in and update our existing pattern or idea of our model in our heads (e.g., Bayesian nets; MePMoS). The brain is constantly integrating new information with prior knowledge and experiences to refine its future predictions about the world, and I think this is especially seen in pattern learning and recognition. 

That being said, as mentioned by the “logical errors” by Terrace (2010), in terms of sequential learning, it seems as if more prediction errors initially would actually benefit an organism. This would allow it to learn the pattern quicker. These errors act as feedback signals, indicating deviations from expected outcomes and prompting the brain to adjust its predictive models accordingly. Essentially, each error serves as a valuable piece of information that guides the refinement of the internal representation of the pattern or sequence being learned. I wonder, then, ecologically if different species have evolved to gauge when it is most appropriate (least risky) to allow themselves to make errors to learn a pattern quicker and when it is least appropriate (most risky) to do so and therefore, make more intentional and meaningful actions (e.g., proximity to a predator)? This could be seen as some sort of trial and error process vs. intentionally focusing on that prior knowledge to make the best (most effective) decision.

In relatively stable and predictable environments where the cost of errors is low, like when foraging for familiar food sources or navigating familiar territories around safe conspecifics, an organism might adopt a more exploratory approach that can tolerate a higher rate of logical errors to accelerate pattern learning. In these contexts, the animal may know that the benefits of quickly acquiring new information may outweigh the risks associated with occasional mistakes. Conversely, in environments with high levels of unpredictability or where errors could have severe consequences (e.g., the presence of predators or during crucial social interactions), organisms may prioritize cautious and deliberate actions over rapid learning through trial and error. In this situation, the cost of errors might outweigh the benefits of accelerated learning. Therefore, these organisms may exhibit more conservative behaviors, relying on existing knowledge and minimizing the likelihood of making errors that could jeopardize their survival. While I know much research on this sort of behavior/risk weighing exists, I wonder about the direct cross-species differences that occur as a function of defensive behaviors and pattern learning. I wonder how the organisms utilized in these papers (non-human primates, monkeys, humans, rats, etc.) would perform in similar tasks under more risk? 

* Chapter from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press. 

[ozimac]

Categorization & Concept Learning

Relational discrimination learning in pigeons, Cook & Wasserman (2012)

Concept learning in animals, Zentall et al. (2008)

The chapter "Relational discrimination learning in pigeons" by Cook and Wasserman (2012)* investigates of concept learning in nonhuman animals, particularly focusing on whether pigeons can learn and transfer same/different discriminations to new stimuli. Both sameness and differentness are crucial concepts for survival and reproduction, lying at the core of various adaptive behaviors. Pigeons have been shown to successfully discriminate between same and different stimuli across various displays (e.g., Wasserman, Young, & Dalrymple (1997); Cook et al. (2003)), suggesting the potential formation of a relational same/different concept. However, there are variations in how pigeons process these stimuli, with some responding categorically while others are more sensitive to factors like entropy (measure of variability in a display) or oddity. There also seemed to be an emphasis in large training set size (e.g., Katz & Wright (2006)), although Blaisdell & Cook (2005) found evidence in pigeons that S/D relations could be judged when only two stimuli were used (see also Katz, Wright, and Bodily (2007)). 

Various experimental procedures, including simultaneous and successive same/different discriminations, have demonstrated pigeons' ability to generalize this behavior to novel stimuli. Differences in training methodologies and stimulus presentations have led to distinct patterns of responding. Additionally, human studies have mirrored some aspects of pigeons' responses (e.g., Young and Wasserman (2001a)), suggesting that humans, like pigeons, can exhibit both categorical and continuous processing of same/different relations. Categorical and continuous dimensions in stimulus control suggests that pigeons and humans may flexibly employ different strategies depending on contextual factors. The authors conclude with many questions that still persist involving competition for attention among these different dimensions, the longevity of prior experience effects, and the hierarchical nature of these conceptual relations. Overall, Cook and Wasserman, through their many paralleled experiments, show the complex cognitive processes involved in relational discrimination learning in both pigeons and humans, while specifically highlighting how animals navigate the multitude of stimulus relations present in their environments.

-

The review article “Concept learning in animals” by Zentall et al. (2008) challenges the idea that conceptual abilities are unique to humans, suggesting that many of the major varieties of conceptual classes claimed to be uniquely human are also exhibited by nonhuman animals. The authors present evidence for the formation of several types of conceptual stimulus classes in animals, including perceptual, associative, relational, and analogical classes. They argue that the ability to sort objects, events, and relations into classes allows for efficient learning and transfer of knowledge to new stimuli or contexts. The authors discuss perceptual concept learning, where animals sort stimuli into classes based on shared physical characteristics, such as shape or color. They provide examples of studies demonstrating animals' ability to categorize complex stimuli, showing that even pigeons can learn to discriminate between categories with varying degrees of success depending on training set size and stimulus repetition (e.g., Bhatt et al., 1988; Wasserman et al. (1988). There is extensive evidence that nonhuman animals are able to master perceptual or basic level concepts.

Associative concept learning involves forming categories based on arbitrary associations, such as between objects and words. The authors describe experiments demonstrating animals' ability to establish associative classes through matching-to-sample, many-to-one tasks and symmetry training (Zentall, Clement, & Weaver (2003)). Notably, Zentall et al. (2003) found that such training allows for the transfer of learning to a new association, where one of the stimuli previously involved in symmetry training is paired with a novel stimulus, while the other stimulus from the original training session replaces the first. The authors also discuss functional classes, where the common association is an acquired function shared by all members of the class (e.g., Vaughan (1988)). Finally, relational concept learning focuses on relationships among stimuli, such as larger than or brighter than. The authors explore studies on transitive inference across species (mentioning rats, pigeons, crows, monkeys, and chimpanzees) and same/different learning, showing that animals can make inferences based on relational properties. However, animals' performance in tasks requiring higher-order relational concepts like analogies varies, with some species demonstrating proficiency (e.g., Sarah the chimpanzee) while others struggle (e.g., pigeons), suggesting limitations in symbolic understanding depending on species.

Overall, the article highlights that animals possess various conceptual abilities comparable to those of humans, challenging the anthropocentric view of many of the cognitive sciences. By studying animals' cognitive processes under appropriate conditions, researchers are able uncover evidence of their conceptual abilities, demonstrating that these capacities are not unique to humans but lie at the heart of the phylogenetic history across all species. 

My thoughts

To begin, I really enjoyed both of these readings and highly recommend them. 

The first point I’d like to make is a thought I had while reading Castro, Kennedy, and Wasserman (2010)’s experiment findings in the Cook and Wasserman (2012) study, in which they used the Blaisdell & Cook (2005)’s two-item simultaneous S/D discrimination task but added a conditional component. In this task, pigeons were trained to respond conditionally based on different background colors in a display. They found that pigeons demonstrated the ability to learn and transfer this conditional discrimination task to novel displays effectively, but notably, they learned to peck at different arrays quicker than learning to peck at same arrays. The authors conclude that it is unclear why but could be because the birds prefer to look for difference or variability, which made me think of the Blough (2012)’s “Reaction-time Explorations of Visual Perception, Attention, and Decision in Pigeons” chapter that confidently stated, “pigeons like to search.” This is evidenced through a number of studies (e.g., “pop out” task in Allan & Blough, 1989; pigeons successfuly trained to find and peck a punched red diamond among complete diamonds in Fujita & Ushitani, 2005), thus this must have some ecological validity. In the field, pigeons must constantly scan their surroundings for food, potential threats, and other relevant stimuli. Thus, they must be vigilant and attentive to changes in their environment, which may be demonstrated by their notice and increased response to differences. This behavior aligns with their ecological role as adaptable and opportunistic foragers, where the ability to detect subtle changes in their environment is advantageous for survival and navigation.

In the Zentall et al. (2008) article, when reading on perceptual classes, the authors discuss how  “[i]t can be difficult to specify the particular common elements among the concept members that might be used to classify them”, although they are still easily classified to their respective classes. This made me think of the idea of essence and a study I read a little while back. Keil (1986) showed children photos of a raccoon whose appearance was altered to look like that of a skunk. The raccoon turned skunk had all the prototypical features of a skunk so, if determined by a category prototype, the children should have considered it a skunk. However, they did not, and rather categorized it still as a raccoon, due its dissimilarity to past skunk examples, similarity to past racoon examples, and what the authors concluded was “essence”. I wonder if there have been any studies studying essence in nonhuman animals?

I also found a lot of related concepts in the Zentall et al (2008) that coincide with a current study in my lab that investigates fast mapping pigeons, or rather fast object association mapping (FOAM), which uses a choice/learning by exclusion process as mentioned in the article (e.g., Clement and Zentall (2003); Kaminski, Call, & Fisher (2004). That being said, I wonder, with FOAM, are we showing learning by exclusion instead of choice by exclusion? Our birds are learning the novel stimuli pairs by exclusion in some capacity better than if they were not using a choice by exclusion process, as shown by the increase in accuracy on familiar foil trials compared to novel foil trials. On the same note, later in the article when discussing the Zentall & Hogan (1976) study looking at same/different discrimination learning, the authors mentioned “[u]nfortunately, pigeons tend to be neophobic and the presentation of novel stimuli appears to interfere with immediately transfer of matching”. This is interesting because I’ve also read that pigeons are neophilic, or rather choose by novelty (Aust et al., 2008). This seems to be contradictory. I would say, however, that the neophobic argument could be evidenced in the FOAM study currently, where after completing tests and moving onto a next phase, our birds will initially show a preference for familiar (incorrect foil) comparisons when presented with a novel sample and (correct) novel comparison. Is this, too, interfering with immediate transfer of choice by exclusion when beginning in new phases? Surprisingly, however, the pigeons switch this strategy shortly after and choose by exclusion (higher accuracy on familiar foil trials). I’ve been likening this to a preference for familiarity specifically because of its prior history of reinforcement but I wonder if this could more-so be due to this neophobia (i.e., avoidance of novelty), as evidenced by the birds switching once the novel comparison isn’t so novel anymore (e.g., after x amount of presentations). (To that same idea, I’m also now realizing that it could not be a preference for familiarity due to its history of reinforcement because it has only ever been reinforced in the context of the sample that it is paired with and has never been reinforced in the presence of any other sample. Thus, it could not be due to its previous history of reinforcement because a "familiar" comparison has never and will never be reinforced in the context of a “novel” sample.)

To finish, I absolutely loved one of the concluding paragraphs in Zentall et al. (2008)’s article and I will leave it here in case anyone outside of my lab ever reads this:

“Animals have evolved various behavioral capacities that, when studied under the appropriate conditions, rival or even exceed those of humans. No one would question the sense of smell of the bloodhound or the strength of an elephant. And, no one would argue that humans are inherently better at navigation than pigeons or migrating birds. Often, however, we assume that our conceptual ability is not only better than that of other animals, but that it is unique to our own species. Yet, other species have undoubtedly had to overcome environmental problems that required the deployment of an assortment of cognitive processes; if asked in the appropriate way, then these animals too can provide convincing evidence of their conceptual abilities.” 

* Chapter from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press. 

[ozimac]

Spatial Cognition

Arthropod navigation: Ants, bees, crabs, spiders, finding their way, Cheng (2012)

Corvid caching. The role of cognition, De Kort, Emery, & Clayton (2012)

Spatial inference without a cognitive map: the role of higher-order path integration, Bouchekioua, Y., Blaisdell, A. P., Kosaki, Y., Tsutsui-Kimura, I., Craddock, P., Mimura, M., & Shigeru, W. (2021) 

In the chapter “Arthropod navigation: ants, bees, crabs, spiders finding their way” by Cheng (2012)*, the diverse navigation mechanisms employed by species such as ants, bees, crabs, and spiders in their environments are explored. The chapter covers four main topics: integration, route behavior, landmark usage, and map-like navigational behavior. It emphasizes the significance of place-finding servomechanisms in navigation, of where a system aims towards a specified target place. Path integration is also discussed, as well as the limitations arising from error accumulation during their journeys. The concept of global vectors and search patterns, particularly in ants, is explained, illustrating how an ant's initial straight run home, based on the global vector, is complemented by a search system to improve accuracy. The role of sensory mechanisms like the sky-compass and odometry in determining direction and distance is explored, along with the integration of landmarks and beacons into navigation strategies. The passage also delves into the learning aspect of path integration, emphasizing the need for metric information from outbound journeys and the potential role of working memory in this one-trial learning process.

Furthermore, the discussion covers route following, where the inaccuracies of the global vector are addressed by breaking the journey into shorter segments guided by local vectors and sensorimotor vectors. The use of panoramic landmarks and landmark-based image matching is explored, showcasing how ants and bees reduce discrepancies between visual images to navigate effectively. The chapter also touches on the debate between map-based and route-based navigation, suggesting that arthropods may employ a combination of both mechanisms. The overarching conclusion emphasizes the widespread use of various navigation strategies in arthropods and outlines experiments mimicking their natural habitats for a realistic understanding of their navigational abilities. Finally, the chapter suggests that path integration requires metric encoding and emphasizes the role of multiple memories, retrieved based on contextual cues, in guiding arthropods through their journeys.

-

The chapter "Corvid Caching: The Role of Cognition" by De Kort, Emery, & Clayton (2012)* explores the cognitive aspects of food caching behavior in birds, focusing primarily on the corvidae (ravens, crows, magpies) and paridae (tits, chickadees) families. Food caching, a naturally occurring behavior, is considered in the context of cognitive abilities such as episodic-like memory, future planning, and social cognition. The study emphasizes the influence of environmental conditions on caching strategies, exemplified by the increased efficiency of Alaskan black-capped chickadees in harsh winters. The chapter delves into the decision-making processes involved in caching, covering factors such as what, when, and where to cache. Corvids exhibit varying caching propensities and recovery behaviors, with species differences linked to fluctuations in food availability depending on the season and ecological conditions. 

Social aspects of caching are also explored, highlighting the strategies used by birds to prevent cache pilfering. The potential for pilfering and the dynamics of living in social groups introduce additional cognitive dimensions to caching behavior. The chapter addresses the question of whether caching reflects prospective cognition, delving into experiments with scrub jays that suggest a dual-process model involving both a compulsion to cache and flexible behavior influenced by past caching experiences. The discussion extends to the cognitive mechanisms involved in cache recovery, emphasizing the role of spatial memory, episodic-like memory, and social strategies. The authors argue for an evolutionary perspective in understanding cognitive abilities, considering the phylogenetic relationships among species, and challenging hypotheses about adaptive specialization in brain size related to ecological demands. Overall, the chapter stresses the importance of an evolutionary approach in studying comparative cognition, especially in the context of complex behaviors like food caching in corvids.

-

The article "Spatial inference without a cognitive map: the role of higher-order path integration" by Bouchekioua et al. (2021) challenges the traditional cognitive map model for route planning. It introduces the concept of Higher-Order Path Integration (HOPI) as an alternative mechanism for spatial inference. The cognitive map model suggests that animals create mental representations of their environment for flexible navigation. However, the article argues that first-order path integration, which relies on direct experience and sensory stimuli, is insufficient to explain novel route taking observed in experiments where path integration is neutralized.

The authors propose HOPI as a new model that combines higher-order conditioning and path integration. HOPI involves vector arithmetic, where vectors representing spatial relationships are stored in memory and combined through common elements. This process allows animals to mentally connect places that have never been directly linked during navigation. The model suggests that animals anchor vectors to external cues and use higher-order conditioning to integrate them, resulting in the formation of higher-order vectors connecting familiar places. In suggesting HOPI, the article highlights three experiments supporting spatial inference without direct path integration. In pigeons and rats, the animals demonstrated the ability to choose novel routes even when path integration was prevented. The proposed HOPI model provides a theoretical framework to explain these observations, emphasizing the role of stored vectors and their integration through conditioning and path integration, challenging some traditional cognitive mapping theories. 

My thoughts

While reading about navigation behavior in arthropods, mammals, and birds, I can’t help but consider these mechanisms in reptiles as well. Specifically, I wonder how reptiles’ ectothermy would affect servomechanistic processes, as endotherms can afford to expense more movement whereas ectotherms must be more particular with their movements? If endotherms can, in terms of energy expenditure, afford to make more potential mistakes, would reptiles, perhaps, have evolved more efficient servomechanistic processes than endotherms? That being said, however, I’m sure escape and foraging behaviors have also led to other adapt navigation processes to develop in endotherms as well. I personally have worked in studying navigation behavior in leopard geckos, and observation for even just one minute would show the drastic difference in the amount of movement in geckos as opposed to the pigeons (birds) and rats (mammals) that I have also worked with closely. I wonder how these thermoregulatory mechanisms have impacted adaptations of servomechanistic processes involved in reptilian movements and, consequently, their navigational strategies? Would they be exclusive to homing species (e.g., sea turtles)? 

Moreover, would HOPI also be applicable to reptile navigation? Consider this scenario:

In the morning, a lizard starts at its burrow where it is quite safe from predators, represented by place A. Throughout the day, it moves in a particular direction and distance in hunting insects for food, creating a vector A->B. Later in the day, the lizard must find a hiding place from a predator, creating another vector B->C. To return to its safe burrow A, the lizard adds the vectors A->B and B->C, resulting in A->C, and symmetrically, C->A?

Although their learning capabilities are not thought to be as nearly cognitively developed as mammals and birds, I wonder if reptiles could show evidence of HOPI in a similarly developed task as suggested by Bouchekioua et al. (2021). That is, would HOPI just be applicable to mammals and birds, or would the phylogeny between birds and reptiles also suggest that reptiles could show evidence of HOPI if tested in a systematically varied way?

Turtles have been found to be able to navigate to a goal using visual cues in a place task using an arm maze (see Lopez et al. (2000)), so they must use visual landmarks to navigate in some capacity (at least when it is all that is available to them). That being said, as optic flow is a crucial factor in visually-dominant avian species, do reptiles also use optic flow as a prominent sensory input for spatial orientation and movement control? Would the reliance on visual cues and sensory inputs for navigation vary significantly between ectothermic reptiles and endothermic birds? I’m assuming, of course, this would also vary depending on reptile species as well. 

López, J. C., Rodríguez, F., Gómez, Y., Vargas, J. P., Broglio, C., & Salas, C. (2000). Place and cue learning in turtles. Animal Learning & Behavior, 28(4), 360–372. https://doi.org/10.3758/BF03200270

* Chapters from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press. 

[ozimac]

Memory Processes

Methodological Issues in Comparative Memory Research, Zentall (2012)

Memory Processing, Wright (2012)

Animal Models of Episodic Memory, Crystal (2018)

The chapter "Methodological Issues in Comparative Memory Research" by Zentall (2012)* focuses on exploring methodological challenges in studying working memory in animals. It emphasizes the need to address confounding factors and remove these methodological issues to accurately assess animal working memory and make sure the questions that are being asked are being answered in the correct way. Various memory phenomena, such as directed forgetting, rehearsal, base-rate neglect, episodic memory, meta-memory, prospective coding, and memory for time, are discussed in the context of animal studies, mostly in pigeons. The chapter highlights the historical development of research methods, including the Delayed Matching-to-Sample (DMTS) procedure, and its modifications in studying memory in animals, ranging from pigeons to monkeys.

Issues related to interference, both proactive and retroactive, are examined, shedding light on how events between the offset of the sample stimulus and onset of the comparison stimulus impact memory retention. Theories of working memory, including trace strength and temporal discrimination hypothesis, are discussed along with the role of delay of reinforcement in affecting memory. The chapter also delves into active memory processes such as directed forgetting, reallocation of memory, and rehearsal. It explores how memory processes in animals can be affected by different cues and instructions, with a particular emphasis on the significance of clear task instructions for obtaining accurate measures of cognitive abilities in animals, especially when transferring human procedures to animals. Generally, throughout the chapter, Zentall stresses the importance of refining experimental designs to enhance the validity of memory assessments in comparative studies and offers insights into potential pitfalls and considerations for future research.

-

In the chapter "Memory Processing" by Wright (2012)*, the author compares memory processing in nonhuman animals, specifically monkeys and pigeons, with that in humans. The focus is on multiple-item memory tests, utilizing either sequential lists or simultaneous spatial arrays of items. The chapter emphasizes the importance of studying list memory over single-item memory, allowing researchers to investigate changes in memory for different list positions as retention intervals increase. The concept of serial position functions, which include primacy and recency effects, is explored across various species, demonstrating that animals exhibit similar dynamic changes in these effects as retention delays are increased.

The author discusses visual list memory, drawing parallels to Ebbinghaus' work in humans and introducing the modal model as an early attempt to understand animal list memory. The findings indicate that monkeys and pigeons show comparable primacy and recency effects, challenging the view that these effects are unique to humans. The study extends to auditory memory in rhesus monkeys, revealing intriguing differences in serial position functions compared to visual memory. The involvement of inhibition in auditory list memory is highlighted, with proactive and retroactive inhibition playing distinct roles at short and long retention delays. It also discusses interference effects from previous memory items and the impact of repeated-item interference on memory performance. The study suggests that large training sets can minimize item repetition and improve memory performance. Familiarity, identity, and episodic memory are also explored, emphasizing the challenges in distinguishing between these memory types in animal studies. The chapter concludes by underscoring the value of animal models in memory research, providing insights into nonverbal memory processing and enabling direct species comparisons to enhance our understanding of how memory works.

-

The article "Animal Models of Episodic Memory" by Crystal (2018) explores the application of animal models to understand episodic memory, a form of memory that involves the ability to remember specific events in one's own past. The central hypothesis posits that at the moment of a memory assessment, animals can recall past events and retrieve episodic memories. The article discusses several case studies involving different memory paradigms in animals (rats), such as what-where-when memory, source memory, item-in-context memory, and incidental encoding. The research compares the animals' performance against a familiarity hypothesis, which suggests that memory traces passively decay over time. The article emphasizes the importance of focusing on the content of episodic memories rather than attempting to assess subjective experiences in animals. The findings suggest that the rats exhibit behaviors consistent with the retrieval of episodic memories, again, challenging the notion that episodic memory is unique to humans. The article concludes by highlighting the value of animal models in studying episodic memory, contributing to evolutionary research, and providing insights into cognitive elements relevant to understanding human diseases, such as Alzheimer's. The integration of advanced assessments of episodic memory in animals is also mentioned to be considered crucial for translational research in the context of neurological disorders.

My thoughts

When reading about episodic memory vs. familiarity, I was surprised that none of the articles, which all touched on these concepts, did not mention or consider the idea of memory reconsolidation. Episodic memory, as self-referential in nature (Crystal, 2018), is by definition, subject to the phenomenon of memory reconsolidation. As these memories are retrieved repeatedly, they can undergo modifications and integrations with new information. I wonder, would a strategy heavily reliant on episodic memory might, over time and repeated recalls, shift towards more-so a reliance of familiarity (less cognitive load?). That is, could the continual reconstruction of a memory eventually lead to a more generalized, familiarity-based representation? Or, would each repeated memory of that source memory instead strengthen the memory itself? Can the two exist on a spectrum, rather than being completely different frameworks? Could that explain the gradual improvement of performance when monkeys had over 30,000 (D’Amato, 1973) and pigeons had over 15,000 (Grant, 1975, 1976, Roberts, 1998) DMTS training trials with a stimulus set? Did that allow them to perhaps, switch strategies? 

The accuracy of episodic recall may diminish with each retrieval, potentially giving rise to a more gist-like or familiarity-based representation. Could this be a cause of the lower accuracy for the short-item lists that had repeated exposure prior to test compared to long item lists that were seen once before test, as mentioned in Wright (2012). That being said, I also wonder how if longer retention intervals may provide more opportunities for memory reconsolidation, or even rumination, would that process lead to better (or worse) accuracy as a function of increasing retention interval? Alternatively, shorter intervals might afford less time for reconsolidation, allowing the original episodic details to persist. This would of course be inconsistent with the literature (e.g., the longer the interval between the offset of the sample and the onset of the comparison, the poorer matching has been found), and hinges on if the animals are thinking retrospectively, at all. 

Similarly, on the topic of rumination, these papers also led me to consider a topic that has been floating around my lab recently – this idea of an incubation period. In the problem-solving literature, an incubation period is a time of diversion away from a problem or task at hand, thus allowing the subconscious to reorganize the problem and lead to a (n insightful) solution. Again, although inconsistent with the retention interval literature, could extended periods of potential rumination be as fruitful to memory processes as an incubation period is to problem solving, at least if facilitated in the right way? That is, could animals, given extended time away from a learning task (with perhaps, something similar to do or involving similar stimuli in the meantime), exhibit changes in memory recall accuracy, and could an incubation-like period contribute to that enhanced memory performance? 

* Chapters from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press. 

[ozimac]

Causality

Associative Accounts of Causality Judgments, Escobar & Miller (2012)

Rational Rats: Causal Inference and Representation, Blaisdell & Waldmann (2012)

The chapter "Associative Accounts of Causality Judgments" by Escobar & Miller (2012)* critically examines the efficacy of associative models in understanding causality. It begins by highlighting the mental processes involved in comprehending causal connections, emphasizing Thorndike's law of effect and the reinforcement of actions with positive outcomes. The text delves into key concepts such as contiguity, contingency, and association, drawing from Hume's perspective on inferring events and addressing the challenges posed by temporal contiguity. The Rescorla-Wagner model is introduced as an efficient tool for explaining conditioning phenomena. The narrative explores associative phenomena in causal learning, tackling acquisition, extinction, and contingent models. Stimulus interaction effects like conditioned inhibition, stimulus competition, and retroactive interference are scrutinized. Retroactive revaluation is an associative learning phenomenon where the perceived strength or significance of a cue, not present during training, is retrospectively altered based on subsequent learning experiences with other cues or events. The article also discusses retrospective revaluation as a phenomenon challenging traditional models, prompting modifications to better account for future evidence. It raises philosophical questions about why some relationships are interpreted as causal and introduces the concept of causal power, suggesting that knowledge about causal relationships differ from covariation.

Overall, the chapter discusses considerations in previous methodology, potential biases in assessing causality, and challenges in generalizing past findings to real-world scenarios. It questions the applicability of the associative framework, the conceptualization of causality as a special instance of associative learning, and the implications of retrospective revaluation effects. Finally, it invites exploration of biological relevance and belongingness concepts when studying causation, and specifically considers instances where stimuli of high biological relevance may deviate from typical associative principles in causal judgments.

-

The chapter "Rational Rats: Causal Inference and Representation" by Blaisdell & Waldmann (2012)* explores causal model theory using the study of rat behavior and its parallel to human studies. It begins by highlighting the importance of causal reasoning in comprehending the world, drawing from Humean philosophy and modern learning theories. It also discusses the advantages of causal models over associative knowledge, emphasizing their ability to represent causal directionality, power, differentiate between causal and noncausal covariations, infer hidden causes, exhibit parsimony, and make predictions.

The focus then shifts to investigating whether rats engage in rational causal reasoning processes or rely solely on covariations. Experimental evidence (Blaisdell et al., 2006) suggests that rats demonstrate basic forms of causal reasoning beyond associative accounts, aligning with causal model theory. The chapter explores whether rats form causal models, understand actions as causal interventions, and differentiate between observed and intervened events in their causal inferences. The authors stress the concept of self-generated actions having a special status in causal reasoning, supported by experiments demonstrating rats' recognition of the deterministic and independent nature of their actions (see Leising et al., 2008). Transfer effects and hidden event cognition are also discussed, revealing rats' ability to deduce hidden causes in various scenarios. The chapter emphasizes that rats' causal reasoning surpasses simple associative theories, advocating for an interpretation of causal model theory in nonhuman animals’ causal reasoning over associative interpretations (use of covariations).

My thoughts

Now, the Escobar & Miller (2012) chapter was a bit dense for me, as I have a very general understanding of causality. However, both articles touch on this idea of observation vs. intervention in causal learning which I find quite interesting. An observation of event A consistently occurring before events B and D might be perceived as the common cause of both effects (B and D). However, the authors point out that such observations alone may not enable the subject to establish whether A directly causes B. To explore the causal structure more conclusively, one potential approach is to intervene on one of the potential causal links while keeping all other factors constant. “Actions can also be viewed as a special type of event; the consequences of actions may be learned faster than if the situation were merely observed” (e.g., Leising, Wong, Waldmann, & Blaisdell, 2008).This makes sense to me, as I feel like you’d be more likely to remember, or learn, an answer in a class if you raised your hand and got it wrong (or right) than if you merely observed that happen to another student. I wonder, though, if observation of that circumstance (i.e., another student raising their hand and getting an answer wrong) would help more than if a professor just said the answer themselves. That being said, I wonder if another rat observing the intervening rat in the Leising et al. study would also produce the same response as that intervening rat themselves. That is, would the observing rat recognize the causal relationship between the lever -> tone without intervening themselves, or is this understanding unique to personal actions and their consequences? 

If rats are capable of recognizing causal relationships by observing interventions, it prompts the question of transferability of knowledge across individuals within the same species and even across species. How do animals perceive and process causal information that revolve around a different species – such as, a pet cat watching its owner pour his or her food from a plastic bag into their empty food dish. Will the cat understand this causal relationship (i.e., owner -> food bag -> pour into food dish) even when they do not see it happening (e.g., they come across their previously empty food dish, now full)? Would that be an example of hidden event cognition? Of course, this would be without considering any contextual signs (e.g., hearing the food rustling in the bag and being poured into the dish). I wonder too, if a cat observed this causal chain once and only once, and then just continued to find its food bowl full after last seeing it empty, would they ever forget what caused that to happen (transfer effects)? Would this count as transfer effects because the cat itself did not intervene, but rather, observe? Or, would their knowledge of that causal chain be strengthened with each time that they found the full food bowl? That is, will they be reminded of (and further, strengthen) that causal model with each presentation of just the outcome? 

On a similar note, to me, when speaking of causality, I can’t help but think of Bayesian inference. Does causal discounting occur when you’re weighing out your priors and judging them by their probabilities? To discuss this further, let’s refer to the example about watering your front yard used in the Blaisdell & Waldmann (2012) paper when discussing transfer effects:

“[i]f I water my front yard and then notice that the sidewalk is wet, I infer that it was I and not rain that caused the wet sidewalk. If on the very next day, however, I notice that the sidewalk is wet and I know that I haven’t watered my lawn that day, then I infer that it must have recently rained. Discounting of rain occurs in the instance in which I intervened, but that inference does not carry over to the next day on which I did not intervene; thus, no discounting is expected”

The initial intervention of watering the front yard establishes a causal belief, acting as a prior. On the next day, the absence of personal intervention shifts the inference, reflecting an update in the Bayesian model based on new evidence (wet sidewalk) and prior knowledge (previous day's intervention). Bayesian reasoning allows for this dynamic adjustment of causal beliefs in response to changing circumstances (contexts) and we must call upon these different adjusted beliefs when making causal inferences. Of course, our prior knowledge influences our causal inferences, but how exactly is that knowledge weighed in our knowledge bank? Are causal relationships that have been observed (or intervened) more often hold more weight, or does context triumph all? Do causal models encompass all the potential causes and their effects, in some sort of flow chart-like representation? Or is it more-so like a ranking (ordered from most to least likely)?

* Chapters from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press. 

[ozimac]

Perception & Attention

Amodal Completion & Illusory Perception in Birds and Primates, Fujita et al. (2012)

Reaction-time Explorations of Visual Perception, Attention, and Decision in Pigeons, Blough (2012)

Experimental Divergences in the Visual Cognition of Birds and Mammals, Qadri & Cook (2015)

The article "Amodal completion & illusory perception in birds and primate" by Fujita et al. (2012) explores the perceptual processes of nonhuman animals, emphasizing the importance of understanding their cognitive abilities. It discusses the role of perception in providing information for higher cognitive processes and highlights potential evolutionary constraints that shape species-specific perceptual characteristics. 

The study involves experiments with pigeons and primates, investigating their responses to completion tasks and visual illusions. Pigeons have even been observed to have difficulty completing ecologically relevant stimuli, which proves the relevance of using ecologically meaningful stimuli, such as seeds and artifacts, in testing pigeons' completion abilities not necessary (e.g., Ushitani & Fujita, 2005). Another question that arises is if pigeons lack a visual perceptual system to complete an image at all, or do they just decline the completed image. Fujita & Ushitani (2005) developed a task to test pure perception, not just completion, where they trained pigeons to find and peck a punched red diamond among complete diamonds. Then put white square next to diamond with a small gap. Finally, tested pigeons with punched diamonds that had a white square filling the punched part using reaction time as a measure of perceived confidence in their choice. Their results found that this was easy for pigeons (with shorter RTs!), but difficult for humans (longer RTs)! Which, to me, suggests employment of an exclusionary process, and with the authors concluding that it seems as if completion seems to be the last resort of pigeons, whereas it is the first choice for other species tested. Further exploration questions whether pigeons lack an early perceptual system for completion or if they choose not to complete images. Other experiments, including investigation of the Ponzo illusion and Muller-Lyer illusion, reveal that pigeons exhibit biases towards expected illusions, though individual variations exist. The article concludes by emphasizing the concept of "umwelt," highlighting differences in sensory processes among species and suggesting that a species' visual perceptual system is tuned to its own ecological niche in the wild.

-

The chapter "Reaction-time Explorations of Visual Perception, Attention, and Decision in Pigeons" by Blough (2012)* aims to highlight the utility of reaction times (RTs) in unraveling psychological processes in pigeons during their interactions with the environment. The experiments primarily involve discrimination tasks with computer-generated stimuli, and the emphasis is on using RTs as a key behavioral measure. The introduction underscores the unique properties of RTs, serving as a valuable measure to understand the duration of processes underlying discriminative responses.

Throughout the chapter, the author explores manipulations of sensory, perception, attention, and incentive variables. The chapter highlights the connection between research conducted on pigeons and studies involving humans, enhancing our comprehensive understanding of these subjects across both species. Specifically, citing that pigeons “like to search”, which like the article above, tells me that there has to be some sort of exclusionary process (e.g., choice by exclusion) that is integral to pigeons’ survival in their ecological niche. Various other topics, such as inhibitory interactions between rods and cones, search processes, display size, and similarity, are discussed. In terms of attention, Blough (1991) found pigeons’ search RTs were shorter after a run of trials with the same target letter than during equivalent sequences in which alterative targets appeared in random order. Perhaps, expectancy makes a perceptual object stand out in the same way that a distinctive feature does. Expectation and priming could be a remedy. The chapter explores the distinction between serial and parallel processing in pigeons, noting similarities and differences with humans.

It also discusses search asymmetry, examining the role of distinctive features and potential influences of experience on search behaviors. Attention, expectation, and search RTs are investigated, with evidence supporting attentional influences on pigeon search RTs. Attention becomes a focus, with evidence supporting two attentional influences on pigeon search RTs, used to compare the pigeon's search-image with a recognition-controlling representation. The chapter concludes by developing complex models of discriminative processes using RT distributions. Models include a mixture of response types, a two-component model for visual-search RTs, and a random-walk model based on reward and stimulus similarity variations. The findings contribute to understanding pigeon discriminative processes, with implications for related processes in other species, including humans. Some results surpass comparable human work (e.g., assymetry in Blough & Blough, 1997), some align, and others suggest intriguing differences in visual information processing between humans and pigeons. Overall, the chapter provides a comprehensive exploration of visual perception, attention, and decision processes in pigeons through a measure of reaction time.

-

The article "Experimental Divergences in the Visual Cognition of Birds and Mammals" by Qadri & Cook (2015) again, explores visual cognition in birds, particularly pigeons, and compares it to human visual cognition. The introduction highlights the evolutionary divergence in diurnal and nocturnal pathways between birds and mammals, leading to distinct nervous system structures for visually guided behavior. Despite the smaller avian brain, pigeons serve as a model for understanding visual cognition, presenting both similarities that rival and differences with human visual behavior.

One significant difference highlighted is pigeons' propensity to attend to smaller local features rather than grasping the larger global form, possibly linked to their ecological needs for searching in dense environments. The article discusses various experimental findings, such as search asymmetry, vertices and edges processing, glass patterns, and perceptual completion, revealing discrepancies and divergences in perceptual behaviors across species. Pigeons exhibit complexities in processing lines, edges, and object completion, often showing resistance to perceptual completion tasks.

It also delves into potential explanations for these differences, including attentional mechanisms, stimulus size, feature weighting, and the ecological relevance of completing separated objects (e.g., Watanabe & Furuya, 1997; Aust & Huber, 2006; Ushitani & Fujita, 2005). The analysis suggests that pigeons may process spatially extended and disconnected information differently from humans, potentially due to limitations in connective or grouping processes. The ecological context of pigeons is considered, questioning whether completing separated objects is crucial in their natural environment at all. 

My thoughts

All three articles touch on the same things, with considerable overlap. All three articles revolve around studies or experiments related to the visual perception and cognition of animals. They involve investigations into how different species, such as birds (pigeons specifically) and primates process visual stimuli, make perceptual decisions, and exhibit behaviors related to visual cognition. The studies explore aspects like amodal completion, illusions, reaction times, attention, and decision-making processes. While each work has its specific focus, they all explore visual cognition in animals from various perspectives and using different experimental approaches.

Throughout the three articles, my current particular interest lies within visual perceptual completion and occlusion. Naturally, it is difficult to replicate a 3D image, or concept (i.e., food), on a 2D space, and furthermore, reflect some sort of depth perception on it. That is, unless subjects are given targeted depth perception training. To just display images on a screen without this training could lead to a number of interpretations: e.g., a “partially occluded” image of a star by a rectangle the same color as the background for example could merely be representative of a new shape, some sort of asymmetrical object with multiple vertices. Even if the occluder was a different color than the background, it could still be seen as an entirely new object, just with this new shape as a feature of it. This could be achieved through a number of ways. 

Engaging in mental imagery through occlusion not only relies on working memory and attention, possibly influenced by long-term memory of the stimulus, but also involves a sense of object permanence. The concept of object permanence becomes complex in a 2D space, where partially occluded images might be perceived as entirely new to birds rather than as the same object covered (object unity). To address this, priming birds with a potentially occluded stimulus during training could be explored as a means to prompt them to mentally imagine it during testing.

In comparing humans' capacity for mental imagery with pigeons, humans have received extensive training through various life experiences which leads them to have a more developed depth perception. This is evident in our ability to distinguish between landscape, midground, and forefront elements in any form of art medium, aided by experiences like viewing paintings and developing depth perception through real-life exposures. In contrast, pigeons are primarily exposed to 2-dimensional shapes, posing depth perception as a challenge. Perhaps, pigeons may benefit from prior depth perception training to show evidence of understanding occlusion on the 2D space. This training could involve introducing moving occluders over static objects to simulate dimensions in the 2D space - an approach I plan on exploring.

Paralleling occlusion literature with rats (e.g., Fast & Blaisdell, 2011; Gonzalez et al., 2022), which demonstrate mental imagery in a 3D space where they can touch and interact with objects, occlusion experiments with birds are presented stimuli solely in a 2D space. Thus, there is a need for exploring potential methods to enhance pigeons' depth perception training, suggesting the possibility of a moving occluder to introduce dimensions in the 2D space that could perhaps facilitate mental imagery.

On another note, I do think there are issues with the MTS (matching-to-sample) procedure, particularly when studying occlusion or the aforementioned illusions, as suggested in the Fujita et al. (2012) paper. Similar to challenges observed in symmetry studies, there's ambiguity in interpreting results because it remains uncertain whether birds are genuinely matching to the sample (e.g., choosing a “long line” comparison when presented with a “long line” sample) at test and were relying on habitual responses during training, (e.g., when A, pick B). An alternative approach could incorporate use of a reward contingency, akin to the procedure employed in our lab's rat mental imagery experiments. This reward contingency, previously utilized in Fujita (2001) for investigating line length estimation, moves away from interpreting MTS information and instead involves inferences from a distinct type of outcome, which could lead to cleaner data and provide a more insightful understanding of how birds categorize and perceive stimuli.

* Chapter from Zentall, T. R., & Wasserman, E. A. (2012). The Oxford handbook of Comparative Cognition. Oxford U. Press.