Cognitive mechanisms for transitive inference performance in rhesus monkeys: measuring the influence of associative strength and inferred order

Hippocampal Pallium Lesion Impairs Transitive Inference in Goldfish

Transitive inference, a process that involves drawing logical conclusions based on preliminary information, is considered a cornerstone of human deductive reasoning. Furthermore, transitive inference is a clear instance of representational flexibility as it implies the novel expression of learned information. In mammals and birds, both episodic memory and transitive inference critically depend on the integrity of the hippocampus. Comparative neurobiological evidence indicates that a hippocampus homologue can also be found in the telencephalic pallium of teleost fish. Here, we investigated whether goldfish demonstrate inferential behavior in a standard transitive inference task, and whether the hippocampal pallium of goldfish, akin to the hippocampus in mammals and birds, plays a role in transitive responding. We trained goldfish with hippocampal pallium lesions and sham-operated controls on a series of overlapping two-item visual premise pairs: A+B-, B+C-, C+D-, D+E-. The sham-operated animals readily learned the premise pair discriminations and responded transitively during the crucial test involving a novel pair of nonadjacent elements (B vs. D). However, hippocampal pallium-lesioned goldfish were impaired in the critical transitive inference test, although they successfully learned to discriminate the premise pairs. These findings suggest that a relational memory function, which supports the novel expression of learned information, could be a primitive feature of the vertebrate hippocampus. Such outcome contributes significantly to the ongoing debate regarding the evolutionary origins of episodic memory in vertebrates.

Flexible Learning and Re-ordering of Context-dependent Object Sequences in Nonhuman Primates

Intelligent behavior involves mentally arranging learned information in novel ways and is particularly well developed in humans. While nonhuman primates (NHP) will learn to arrange new items in complex serial order and re-arrange neighboring items within that order, it has remained contentious whether they are capable to re-assign items more flexibly to non-adjacent positions. Such mental re-indexing is facilitated by inferring the latent temporal structure of experiences as opposed to learning serial chains of item-item associations. Here, we tested the ability for flexible mental re-indexing in rhesus macaques. Subjects learned to serially order five objects. A change of the background context indicated when the object order changed, probing the subjects to mentally re-arrange objects to non-adjacent positions of the learned serial structure. Subjects successfully used the context cue to pro-actively re-index items to new, non-adjacent positions. Mental re-indexing was more likely when the initial order had been learned at a higher level, improved with more experience of the re-indexing rule and correlated with working memory performance in a delayed match-to-sample task. These findings suggest that NHPs inferred the latent serial structure of experiences beyond a chaining of item-item associations and mentally rearrange items within that structure. The pattern of results indicates that NHPs form non-spatial cognitive maps of their experiences, which is a hallmark for flexible mental operations in many serially ordered behaviors including communication, counting or foraging.

A mathematical theory of relational generalization in transitive inference

Humans and animals routinely infer relations between different items or events and generalize these relations to novel combinations of items. This allows them to respond appropriately to radically novel circumstances and is fundamental to advanced cognition. However, how learning systems (including the brain) can implement the necessary inductive biases has been unclear. Here we investigated transitive inference (TI), a classic relational task paradigm in which subjects must learn a relation (A > B and B > C) and generalize it to new combinations of items (A > C). Through mathematical analysis, we found that a broad range of biologically relevant learning models (e.g. gradient flow or ridge regression) perform TI successfully and recapitulate signature behavioral patterns long observed in living subjects. First, we found that models with item-wise additive representations automatically encode transitive relations. Second, for more general representations, a single scalar "conjunctivity factor" determines model behavior on TI and, further, the principle of norm minimization (a standard statistical inductive bias) enables models with fixed, partly conjunctive representations to generalize transitively. Finally, neural networks in the "rich regime," which enables representation learning and has been found to improve generalization, unexpectedly show poor generalization and anomalous behavior. We find that such networks implement a form of norm minimization (over hidden weights) that yields a local encoding mechanism lacking transitivity. Our findings show how minimal statistical learning principles give rise to a classical relational inductive bias (transitivity), explain empirically observed behaviors, and establish a formal approach to understanding the neural basis of relational abstraction.

Breast Cancer and Bone Mineral Density in a U.S. Cohort of Middle-Aged Women: Associations with Phosphate Toxicity

Breast cancer is associated with phosphate toxicity, the toxic effect from dysregulated phosphate metabolism that can stimulate tumorigenesis. Phosphate toxicity and dysregulated phosphate metabolism are also associated with bone mineral abnormalities, including excessive bone mineral loss and deposition. Based on shared associations with dysregulated phosphate metabolism and phosphate toxicity, a hypothesis proposed in the present mixed methods-grounded theory study posits that middle-aged women with incidence of breast cancer had a greater magnitude of changes in bone mineral density over time compared with women who remained cancer-free. To test this hypothesis, a mixed-effects model was used to analyze the associations of breast cancer incidence with spinal bone mineral density changes in the U.S. Study of Women's Health Across the Nation. Compared with women in the cohort who remained cancer-free, women who self-reported breast cancer had higher bone mineral density at baseline, but had more rapid losses in bone mineral density during follow-up visits. These findings agree with the hypothesis that a greater magnitude of changes in bone mineral density over time is associated with breast cancer in a cohort of middle-aged women. The findings also have implications for studies investigating dysregulated phosphate metabolism and phosphate toxicity as causative factors of bone metastasis in metastatic breast cancer. Additionally, the authors previously found increased breast cancer risk associated with high dietary phosphate intake in the same cohort of middle-aged women, and more studies should investigate a low-phosphorus diet to reduce bone mineral abnormalities and tumorigenesis in breast cancer patients.

The transitive inference task to study the neuronal correlates of memory-driven decision making: A monkey neurophysiology perspective

A vast amount of literature agrees that rank-ordered information as A>B>C>D>E>F is mentally represented in spatially organized schemas after learning. This organization significantly influences the process of decision-making, using the acquired premises, i.e. deciding if B is higher than D is equivalent to comparing their position in this space. The implementation of non-verbal versions of the transitive inference task has provided the basis for ascertaining that different animal species explore a mental space when deciding among hierarchically organized memories. In the present work, we reviewed several studies of transitive inference that highlighted this ability in animals and, consequently, the animal models developed to study the underlying cognitive processes and the main neural structures supporting this ability. Further, we present the literature investigating which are the underlying neuronal mechanisms. Then we discuss how non-human primates represent an excellent model for future studies, providing ideal resources for better understanding the neuronal correlates of decision-making through transitive inference tasks.

Thinking about order: a review of common processing of magnitude and learned orders in animals

Rich behavioral and neurobiological evidence suggests cognitive and neural overlap in how quantitatively comparable dimensions such as quantity, time, and space are processed in humans and animals. While magnitude domains such as physical magnitude, time, and space represent information that can be quantitatively compared (4 "is half of" 8), they also represent information that can be organized ordinally (1→2→3→4). Recent evidence suggests that the common representations seen across physical magnitude, time, and space domains in humans may be due to their common ordinal features rather than their common quantitative features, as these common representations appear to extend beyond magnitude domains to include learned orders. In this review, we bring together separate lines of research on multiple ordinal domains including magnitude-based and learned orders in animals to explore the extent to which there is support for a common cognitive process underlying ordinal processing. Animals show similarities in performance patterns across natural quantitatively comparable ordered domains (physical magnitude, time, space, dominance) and learned orders (acquired through transitive inference or simultaneous chaining). Additionally, they show transfer and interference across tasks within and between ordinal domains that support the theory of a common ordinal representation across domains. This review provides some support for the development of a unified theory of ordinality and suggests areas for future research to better characterize the extent to which there are commonalities in cognitive processing of ordinal information generally.

Rhesus monkeys manipulate mental images

Humans form mental images and manipulate them in ways that mirror physical transformations of objects. Studies of nonhuman animals will inform our understanding of the evolution and distribution among species of mental imagery. Across three experiments, we found mostly converging evidence that rhesus monkeys formed and rotated mental images. In Experiment 1, monkeys discriminated rotations of sample images from mirror images, and showed longer response latencies with greater rotation as is characteristic of human mental rotation. In Experiment 2 monkeys used a rotation cue that indicated how far to mentally rotate sample images before tests, indicating a precision of better than 30° in discriminating rotations. Experiment 3 yielded mixed evidence on whether the rotation cue shortened decision times as has been found in humans. These results show that rhesus monkeys manipulate mental images.

Hippocampal and medial prefrontal cortices encode structural task representations following progressive and interleaved training schedules

Memory generalisations may be underpinned by either encoding- or retrieval-based generalisation mechanisms and different training schedules may bias some learners to favour one of these mechanisms over the other. We used a transitive inference task to investigate whether generalisation is influenced by progressive vs randomly interleaved training, and overnight consolidation. On consecutive days, participants learnt pairwise discriminations from two transitive hierarchies before being tested during fMRI. Inference performance was consistently better following progressive training, and for pairs further apart in the transitive hierarchy. BOLD pattern similarity correlated with hierarchical distances in the left hippocampus (HIP) and medial prefrontal cortex (MPFC) following both training schedules. These results are consistent with the use of structural representations that directly encode hierarchical relationships between task features. However, such effects were only observed in the MPFC for recently learnt relationships. Furthermore, the MPFC appeared to maintain structural representations in participants who performed at chance on the inference task. We conclude that humans preferentially employ encoding-based mechanisms to store map-like relational codes that can be used for memory generalisation. These codes are expressed in the HIP and MPFC following both progressive and interleaved training but are not sufficient for accurate inference.

Superstitious learning of abstract order from random reinforcement

Humans and other animals often infer spurious associations among unrelated events. However, such superstitious learning is usually accounted for by conditioned associations, raising the question of whether an animal could develop more complex cognitive structures independent of reinforcement. Here, we tasked monkeys with discovering the serial order of two pictorial sets: a "learnable" set in which the stimuli were implicitly ordered and monkeys were rewarded for choosing the higher-rank stimulus and an "unlearnable" set in which stimuli were unordered and feedback was random regardless of the choice. We replicated prior results that monkeys reliably learned the implicit order of the learnable set. Surprisingly, the monkeys behaved as though some ordering also existed in the unlearnable set, showing consistent choice preference that transferred to novel untrained pairs in this set, even under a preference-discouraging reward schedule that gave rewards more frequently to the stimulus that was selected less often. In simulations, a model-free reinforcement learning algorithm (Q-learning) displayed a degree of consistent ordering among the unlearnable set but, unlike the monkeys, failed to do so under the preference-discouraging reward schedule. Our results suggest that monkeys infer abstract structures from objectively random events using heuristics that extend beyond stimulus-outcome conditional learning to more cognitive model-based learning mechanisms.

Influence of Rule- and Reward-based Strategies on Inferences of Serial Order by Monkeys

Knowledge of transitive relationships between items can contribute to learning the order of a set of stimuli from pairwise comparisons. However, cognitive mechanisms of transitive inferences based on rank order remain unclear, as are relative contributions of reward associations and rule-based inference. To explore these issues, we created a conflict between rule- and reward-based learning during a serial ordering task. Rhesus macaques learned two lists, each containing five stimuli that were trained exclusively with adjacent pairs. Selection of the higher-ranked item resulted in rewards. "Small reward" lists yielded two drops of fluid reward, whereas "large reward" lists yielded five drops. Following training of adjacent pairs, monkeys were tested on novels pairs. One item was selected from each list, such that a ranking rule could conflict with preferences for large rewards. Differences between the corresponding reward magnitudes had a strong influence on accuracy, but we also observed a symbolic distance effect. That provided evidence of a rule-based influence on decisions. RT comparisons suggested a conflict between rule- and reward-based processes. We conclude that performance reflects the contributions of two strategies and that a model-based strategy is employed in the face of a strong countervailing reward incentive.

This should help with that: A behavioral investigation into self-derivation of knowledge about prescription medications

Self-derivation of new factual knowledge is crucial for building a knowledge base. In three experiments, we investigated self-derivation about prescription medications. In Experiment 1, adults self-derived new knowledge across textual materials on 40% of trials. Participants in Experiment 2 performed similarly (42%), even when half the information was presented in videos. It was crucial that participants received both learning episodes to successfully self-derive: control condition participants received half the necessary information and performed significantly lower. When a delay was imposed between related facts in Experiment 3, participants self-derived on only 33% of trials and performance did not differ from the control condition. The present research expanded our understanding of adults' learning and self-derivation across media about medications. It revealed room for improvement in adults' learning and self-derivation about health information. This work suggests the need to identify factors that alter performance, including better understanding of the properties of information sources.

Transitive inference after minimal training in rhesus macaques (Macaca mulatta)

Rhesus macaques, when trained for several hundred trials on adjacent items in an ordered list (e.g., A > B, B > C, C > D), are able to make accurate transitive inferences (TI) about previously untrained pairs (e.g., A > C, B > D). How that learning unfolds during training, however, is not well understood. We sought to measure the relationship between the amount of TI training and the resulting response accuracy in 4 rhesus macaques using seven-item lists. The training conditions included the absolute minimal case of presenting each of the six adjacent pairs only once prior to testing. We also tested transfer to nonadjacent pairs with 24 and 114 training trials. Because performance during and after small amounts of training is expected to be near chance levels, we developed a descriptive statistical model to estimate potentially subtle learning effects in the presence of much larger random response variability and systematic bias. These results suggest that subjects learned serial order in an incremental fashion. Thus, rather than performing transitive inference by a logical process, serial learning in rhesus macaques proceeds in a manner more akin to a statistical inference, with an initial uncertainty about list position that gradually becomes more accurate as evidence accumulates. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Positional inference in rhesus macaques

Understanding how organisms make transitive inferences is critical to understanding their general ability to learn serial relationships. In this context, transitive inference (TI) can be understood as a specific heuristic that applies broadly to many different serial learning tasks, which have been the focus of hundreds of studies involving dozens of species. In the present study, monkeys learned the order of 7-item lists of photographic stimuli by trial and error, and were then tested on "derived" lists. These derived test lists combined stimuli from multiple training lists in ambiguous ways, sometimes changing their order relative to training. We found that subjects displayed strong preferences when presented with novel test pairs, even when those pairs were drawn from different training lists. These preferences were helpful when test pairs had an ordering congruent with their ranks during training, but yielded consistently below-chance performance when pairs had an incongruent order relative to training. This behavior can be explained by the joint contributions of transitive inference and another heuristic that we refer to as "positional inference." Positional inferences play a complementary role to transitive inferences in facilitating choices between novel pairs of stimuli. The theoretical framework that best explains both transitive and positional inferences is a spatial model that represents both the position of each stimulus and its uncertainty. A computational implementation of this framework yields accurate predictions about both correct responses and errors on derived lists.

Category learning in a transitive inference paradigm

The implied order of a ranked set of visual images can be learned without reliance on information that explicitly signals their order. Such learning is difficult to explain by associative mechanisms, but can be accounted for by cognitive representations and processes such as transitive inference. Our study sought to determine if those processes also apply to learning categories of images. We asked whether participants can (a) infer that stimulus images belonged to familiar categories, even when the images for each trial were unique, and (b) sort those categories into an ordering that obeys transitivity. Participants received minimal verbal instruction and a single session of training. Despite this, they learned the implied order of lists of fixed stimuli and lists of ordered categories, using trial-unique exemplars. We trained two groups, one for which stimuli were constant throughout training and testing (n = 60), and one for which exemplars of each category were trial-unique (n = 50). Our findings suggest that differing cognitive processes may underpin serial learning when learning about specific stimuli as opposed to stimulus categories.

Absolute and relative knowledge of ordinal position on implied lists

Does serial learning result in specific associations between pairs of items, or does it result in a cognitive map based on relations of all items? In 2 experiments, we trained human participants to learn various lists of photographic images. We then tested the participants on new lists of photographic images. These new lists were constructed by selecting only 1 image from each list learned during training. In Experiment 1, participants were trained to choose the earlier (experimenter defined) item when presented with adjacent pairs of items on each of 5 different 5-item lists. Participants were then tested on derived lists, in which each item retained its original ordinal position, even though each of the presented pairs was novel. Participants performed above chance on all of the derived lists. In Experiment 2, a different group of participants received the same training as those of Experiment 1, but the ordinal positions of items were systematically changed on each derived list. The response accuracy for Experiment 2 varied inversely with the degree to which an item's original ordinal position was changed. These results can be explained by a model in which participants learned to make both positional inferences about the absolute rank of each stimulus, and transitive inferences about the relative ranks of pairs of stimuli. These inferences enhanced response accuracy when ordinal position was maintained, but not when it was changed. Our results demonstrate quantitatively that, in addition to item-item associations that participants acquire while learning a list of arbitrary items, they form a cognitive map that represents both experienced and inferred relationships. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Associative models fail to characterize transitive inference performance in rhesus monkeys (Macaca mulatta)

It has been suggested that non-verbal transitive inference (if A > B and B > C, then A > C) can be accounted for by associative models. However, little is known about the applicability of such models to primate data. In Experiment 1, we tested the fit of two associative models to primate data from both sequential training, in which the training pairs were presented in a backward order, and simultaneous training, in which all training pairs are presented intermixed from the beginning. We found that the models provided an equally poor fit for both sequential and simultaneous training presentations, contrary to the case with data from pigeons. The models were also unable to predict the robust symbolic distance effects characteristic of primate transitive choices. In Experiment 2, we used the models to fit a list-linking design in which two seven-item transitive lists were first trained independently (A > B…. > F > G and H > I …. > M > N) then combined via a linking pair (G+ H-) into a single, 14-item list. The model produced accurate predictions for between-list pairs, but did not predict transitive responses for within-list pairs from list 2. Overall, our results support research indicating that associative strength does not adequately account for the behavior of primates in transitive inference tasks. The results also suggest that transitive choices may result from different processes, or different weighting of multiple processes, across species.

Preserved visual memory and relational cognition performance in monkeys with selective hippocampal lesions

The theory that the hippocampus is critical for visual memory and relational cognition has been challenged by discovery of more spared hippocampal tissue than previously reported in H.M., previously unreported extra-hippocampal damage in developmental amnesiacs, and findings that the hippocampus is unnecessary for object-in-context memory in monkeys. These challenges highlight the need for causal tests of hippocampal function in nonhuman primate models. Here, we tested rhesus monkeys on a battery of cognitive tasks including transitive inference, temporal order memory, shape recall, source memory, and image recognition. Contrary to predictions, we observed no robust impairments in memory or relational cognition either within- or between-groups following hippocampal damage. These results caution against over-generalizing from human correlational studies or rodent experimental studies, compel a new generation of nonhuman primate studies, and indicate that we should reassess the relative contributions of the hippocampus proper compared to other regions in visual memory and relational cognition.

Learned Representation of Implied Serial Order in Posterior Parietal Cortex

Monkeys can learn the implied ranking of pairs of images drawn from an ordered set, despite never seeing all of the images simultaneously and without explicit spatial or temporal cues. We recorded the activity of posterior parietal cortex (including lateral intraparietal area LIP) neurons while monkeys learned 7-item transitive inference (TI) lists with 2 items presented on each trial. Behavior and neuronal activity were significantly influenced by the ordinal relationship of the stimulus pairs, specifically symbolic distance (the difference in rank) and joint rank (the sum of the ranks). Symbolic distance strongly predicted decision accuracy and learning rate. An effect of joint rank on performance was found nested within the symbolic distance effect. Across the population of neurons, there was significant modulation of firing correlated with the relative ranks of the two stimuli presented on each trial. Neurons exhibited selectivity for stimulus rank during learning, but not before or after. The observed behavior is poorly explained by associative or reward mechanisms, and appears more consistent with a mental workspace model in which implied serial order is mapped within a spatial framework. The neural data suggest that posterior parietal cortex supports serial learning by representing information about the ordinal relationship of the stimuli presented during a given trial.

Monkey Metacognition Could Generate More Insight

Monkeys demonstrate metacognition by avoiding memory tests when they forget, seeking information when ignorant, and gambling sensibly after making judgments. Some of this metacognition appears to be based on introspection of private mental states. It is likely that nonhuman cognitive systems, like human systems, differ in accessibility to such introspective metacognition, and the extent to which differences in access map to explicit and implicit cognition will be an important topic for future work. It will be exciting to learn more about the distribution of metacognition among species, and the conditions under which metacognition evolves.

Discovering Implied Serial Order Through Model-Free and Model-Based Learning

Humans and animals can learn to order a list of items without relying on explicit spatial or temporal cues. To do so, they appear to make use of transitivity, a property of all ordered sets. Here, we summarize relevant research on the transitive inference (TI) paradigm and its relationship to learning the underlying order of an arbitrary set of items. We compare six computational models of TI performance, three of which are model-free (Q-learning, Value Transfer, and REMERGE) and three of which are model-based (RL-Elo, Sequential Monte Carlo, and Betasort). Our goal is to assess the ability of these models to produce empirically observed features of TI behavior. Model-based approaches perform better under a wider range of scenarios, but no single model explains the full scope of behaviors reported in the TI literature.

Co-operation of long-term and working memory representations in simultaneous chaining by rhesus monkeys (Macaca mulatta)

We studied the memory representations that control execution of action sequences by training rhesus monkeys (Macaca mulatta) to touch sets of five images in a predetermined arbitrary order (simultaneous chaining). In Experiment 1, we found that this training resulted in mental representations of ordinal position rather than learning associative chains, replicating the work of others. We conducted novel analyses of performance on probe tests consisting of two images "derived" from the full five-image lists (i.e., test B, D from list A→B→C→D→E). We found a "first item effect" such that monkeys responded most quickly to images that occurred early in the list in which they had been learned, indicating that monkeys covertly execute known lists mentally until an image on the screen matches the one stored in memory. Monkeys also made an ordinal comparison of the two images presented at test based on long-term memory of positional information, resulting in a "symbolic distance effect." Experiment 2 indicated that ordinal representations were based on absolute, rather than on relative, positional information because subjects did not link two lists into one large list after linking training, unlike what occurs in transitive inference. We further examined the contents of working memory during list execution in Experiments 3 and 4 and found evidence for a prospective, rather than a retrospective, coding of position in the lists. These results indicate that serial expertise in simultaneous chaining results in robust absolute ordinal coding in long-term memory, with rapidly updating prospective coding of position in working memory during list execution.

Nonnavigational spatial memory performance is unaffected by hippocampal damage in monkeys

Evidence that the hippocampus is critical for spatial memory in nonnavigational tests is mixed. A recent study reported that temporary hippocampal inactivation impaired spatial memory in the nonnavigational Hamilton Search Task in monkeys. However, several studies have documented no impairment on other nonnavigational spatial memory tests following permanent hippocampal lesions. It was hypothesized that transient, but not permanent, hippocampal disruption produces deficits because monkeys undergoing transient inactivation continue to try to use a hippocampal-dependent strategy, whereas monkeys with permanent lesions use a nonhippocampal-dependent strategy. We evaluated this hypothesis by testing five rhesus monkeys with hippocampal lesions and five controls on a computerized analogue of the Hamilton Search Task. On each trial, monkeys saw an array of squares on a touchscreen, each of which "hid" one reward. Retrieving a reward depleted that location and monkeys continued selecting squares until they found all rewards. The optimal strategy is to remember chosen locations and choose each square once. Unlike the inactivation study, monkeys with hippocampal damage were as accurate as controls regardless of retention interval. Critically, we found no evidence that the groups used different strategies, as measured by learning rates, spatial search biases, perseverative win-stay errors, or inter-choice distance. This discrepancy between the effect of inactivations and lesions may result from off-target effects of inactivations or as-yet-unidentified differences between the physical and computerized tasks. Combined with previous evidence that hippocampal damage impairs navigational memory in monkeys, this evidence constrains the role of the hippocampus in spatial memory as being critical for navigational tests that likely involve allocentric spatial memory but not nonnavigational tests that likely involve egocentric spatial memory.

Transitive inference in humans (Homo sapiens) and rhesus macaques (Macaca mulatta) after massed training of the last two list items

Transitive inference (TI) is a classic learning paradigm for which the relative contributions of experienced rewards and representation-driven inference have been vigorously debated, particularly with regard to the notion that animals are capable of logic and reasoning. Rhesus macaque subjects and human participants performed a TI task in which, prior to learning a seven-item list ABCDEFG, a block of trials presented exclusively the pair FG. Contrary to the expectation of associative models, the high prior rate of reward for F did not disrupt learning of the entire list. Monkeys (who each completed many sessions) learned to anticipate that novel stimuli should be preferred over F. We interpret this as evidence of a task representation of TI that generalizes beyond learning about specific stimuli. Humans (who were task-naïve) showed a transitory bias to F when it was paired with novel stimuli, but very rapidly unlearned that bias. Performance with respect to the remaining stimuli was consistent with past reports of TI in both species. These results are difficult to reconcile with any account that seeks to assign the strength of association between individual stimuli and rewards. Instead, they support both sophisticated cognitive processes in both species, albeit with some species differences.

Neuronal Modulation in the Prefrontal Cortex in a Transitive Inference Task: Evidence of Neuronal Correlates of Mental Schema Management

When informed that A > B and B > C, humans and other animals can easily conclude that A > C. This remarkable trait of advanced animals, which allows them to manipulate knowledge flexibly to infer logical relations, has only recently garnered interest in mainstream neuroscience. How the brain controls these logical processes remains an unanswered question that has been merely superficially addressed in neuroimaging and lesion studies, which are unable to identify the underlying neuronal computations. We observed that the activation pattern of neurons in the prefrontal cortex (PFC) during pair comparisons in a highly demanding transitive inference task fully supports the behavioral performance of the two monkeys that we tested. Our results indicate that the PFC contributes to the construction and use of a mental schema to represent premises. This evidence provides a novel framework for understanding the function of various areas of brain in logic processes and impairments to them in degenerative, traumatic, and psychiatric pathologies.

SIGNIFICANCE STATEMENT

In cognitive neuroscience, it is unknown how information that leads to inferential deductions are encoded and manipulated at the neuronal level. We addressed this question by recording single-unit activity from the dorsolateral prefrontal cortex of monkeys that were performing a transitive inference (TI) task. The TI required one to choose the higher ranked of two items, based on previous, indirect experience. Our results demonstrated that single-neuron activity supports the construction of an abstract, mental schema of ordered items in solving the task and that this representation is independent of the reward value that is experienced for the single items. These findings identify the neural substrates of abstract mental representations that support inferential thinking.

Transitive inference of social dominance by human infants

It is surprising that there are inconsistent findings of transitive inference (TI) in young infants given that non-linguistic species succeed on TI tests. To conclusively test for TI in infants, we developed a task within the social domain, with which infants are known to show sophistication. We familiarized 10- to 13-month-olds (M = 11.53 months) to a video of two dominance interactions between three puppets (bear > elephant; hippo > bear) consistent with a dominance hierarchy (hippo > bear > elephant; where '>' denotes greater dominance). Infants then viewed interactions between the two puppets that had not interacted during familiarization. These interactions were either congruent (hippo > elephant) or incongruent (elephant > hippo) with the inferred hierarchy. Consistent with TI, infants looked longer to incongruent than congruent displays. Control conditions ruled out the possibility that infants' expectations were based on stable behaviors specific to individual puppets rather than their inferred transitive dominance relations. We suggest that TI may be supported by phylogenetically ancient mechanisms of ordinal representation and visuospatial processing that come online early in human development.

Implicit Value Updating Explains Transitive Inference Performance: The Betasort Model

Transitive inference (the ability to infer that B > D given that B > C and C > D) is a widespread characteristic of serial learning, observed in dozens of species. Despite these robust behavioral effects, reinforcement learning models reliant on reward prediction error or associative strength routinely fail to perform these inferences. We propose an algorithm called betasort, inspired by cognitive processes, which performs transitive inference at low computational cost. This is accomplished by (1) representing stimulus positions along a unit span using beta distributions, (2) treating positive and negative feedback asymmetrically, and (3) updating the position of every stimulus during every trial, whether that stimulus was visible or not. Performance was compared for rhesus macaques, humans, and the betasort algorithm, as well as Q-learning, an established reward-prediction error (RPE) model. Of these, only Q-learning failed to respond above chance during critical test trials. Betasort’s success (when compared to RPE models) and its computational efficiency (when compared to full Markov decision process implementations) suggests that the study of reinforcement learning in organisms will be best served by a feature-driven approach to comparing formal models.

Although machine learning systems can solve a wide variety of problems, they remain limited in their ability to make logical inferences. We developed a new computational model, called betasort, which addresses these limitations for a certain class of problems: Those in which the algorithm must infer the order of a set of items by trial and error. Unlike extant machine learning systems (but like children and many non-human animals), betasort is able to perform “transitive inferences” about the ordering of a set of images. The patterns of error made by betasort resemble those made by children and non-human animals, and the resulting learning achieved at low computational cost. Additionally, betasort is difficult to classify as either “model-free” or “model-based” according to the formal specifications of those classifications in the machine learning literature. One of the broader implications of these results is that achieving a more comprehensive understanding of how the brain learns will require analysts to entertain other candidate learning models.

The NMDAr antagonist ketamine interferes with manipulation of information for transitive inference reasoning in non-human primates

One of the most remarkable traits of highly encephalized animals is their ability to manipulate knowledge flexibly to infer logical relationships. Operationally, the corresponding cognitive process can be defined as reasoning. One hypothesis is that this process relies on the reverberating activity of glutamate neural circuits, sustained by NMDA receptor (NMDAr) mediated synaptic transmission, in both parietal and prefrontal areas. We trained two macaque monkeys to perform a form of deductive reasoning - the transitive inference task - in which they were required to learn the relationship between six adjacent items in a single session and then deduct the relationship between nonadjacent items that had not been paired in the learning phase. When the animals had learned the sequence, we administered systemically a subanaesthetic dose of ketamine (a NMDAr antagonist) and measured their performance on learned and novel problems. We observed impairments in determining the relationship between novel pairs of items. Our results are consistent with the hypothesis that transitive inference premises are integrated during learning in a unified representation and that reducing NMDAr activity interferes with the use of this mental model, when decisions are required in comparing pairs of items that have not been learned.

The long and the short of it: rule-based relative length discrimination in carrion crows, Corvus corone

Birds and other nonhuman animals can choose the larger of two discrete or continuous quantities. However, whether birds possess the conceptual grasp and cognitive control to flexibly switch between relative more-or-less-than judgments remains elusive. We therefore tested carrion crows in a rule-based line-length discrimination task to flexibly select lines presented on a touchscreen according to their relative length. In the first experiment, the crows needed to discriminate a shorter from a longer line, and vice versa. In the second experiment, the crows were required to choose a medium long line among three lines of different length (intermediate-size task). The crows switched effortlessly between "longer than/shorter than" rules, showing no signs of trial history affecting switching performance. They reliably chose the relatively longer and shorter line length, thus demonstrating a concept of greater than/less than with a continuous magnitude. However, both crows failed to discriminate a line of 'medium' length embedded in longer and shorter lines. These results indicate that relational discrimination exhibits different cognitive demands. While a greater than/less than concept requires only one relational comparison (with the respectively greater or smaller magnitude), the discrimination of a 'medium' magnitude demands to relate two or more comparisons, which might overburden crows and maybe animals in general.

Effects of spatial training on transitive inference performance in humans and rhesus monkeys

It is often suggested that transitive inference (TI; if A > B and B > C, then A > C) involves mentally representing overlapping pairs of stimuli in a spatial series. However, there is little direct evidence to unequivocally determine the role of spatial representation in TI. We tested whether humans and rhesus monkeys use spatial representations in TI by training them to organize 7 images in a vertical spatial array. Then, we presented subjects with a TI task using these same images. The implied TI order was either congruent or incongruent with the order of the trained spatial array. Humans in the congruent condition learned premise pairs more quickly, and were faster and more accurate in critical probe tests, suggesting that the spatial arrangement of images learned during spatial training influenced subsequent TI performance. Monkeys first trained in the congruent condition also showed higher test trial accuracy when the spatial and inferred orders were congruent. These results directly support the hypothesis that humans solve TI problems by spatial organization, and suggest that this cognitive mechanism for inference may have ancient evolutionary roots.

Transitive inference by pigeons: does the geometric presentation of the stimuli make a difference?

In studies of transitive inference (TI), nonhuman animals are typically trained with the following 5-term task: A+B-, B+C-, C+D-, D+E- where the letters stand for arbitrary stimuli and [+] indicates that choice is reinforced and [-] indicates that choice is not reinforced. A TI effect is found when, given the untrained test pair BD, subjects choose B. TI effects have been found in many nonhuman species. Although reinforcement history has been posited as an account of the TI effect, it has failed to account for a variety of conditions under which TI effects have been found. A more cognitive account of TI is that organisms are able to form a representation of the series (A>B>C>D>E). In support of this hypothesis, Roberts and Phelps (Psychol Sci 5:368-374, 1994) found that presentation of the pairs of stimuli in a linear arrangement facilitated TI performance by rats, whereas presentation of the pairs of stimuli in a circular arrangement did not. Using methods adapted from Roberts and Phelps, we trained pigeons on either a linear or a circular arrangement of stimuli with the 5-term task. Results indicated that on the BD test pair, pigeons trained with a circular arrangement did not differ from those trained with a linear arrangement. Furthermore, we found that memory for training pairs was variable and was highly correlated with degree of TI. The results suggest that regardless of how pigeons are able to represent the stimuli, choice was not affected by the spatial arrangement of the stimuli during training.

Cognitive abilities on transitive inference using a novel touchscreen technology for mice

Cognitive abilities are impaired in neurodevelopmental disorders, including autism spectrum disorder (ASD) and schizophrenia. Preclinical models with strong endophenotypes relevant to cognitive dysfunctions offer a valuable resource for therapeutic development. However, improved assays to test higher order cognition are needed. We employed touchscreen technology to design a complex transitive inference (TI) assay that requires cognitive flexibility and relational learning. C57BL/6J (B6) mice with good cognitive skills and BTBR T+tf/J (BTBR), a model of ASD with cognitive deficits, were evaluated in simple and complex touchscreen assays. Both B6 and BTBR acquired visual discrimination and reversal. BTBR displayed deficits on components of TI, when 4 stimuli pairs were interspersed, which required flexible integrated knowledge. BTBR displayed impairment on the A > E inference, analogous to the A > E deficit in ASD. B6 and BTBR mice both reached criterion on the B > D comparison, unlike the B > D impairment in schizophrenia. These results demonstrate that mice are capable of complex discriminations and higher order tasks using methods and equipment paralleling those used in humans. Our discovery that a mouse model of ASD displays a TI deficit similar to humans with ASD supports the use of the touchscreen technology for complex cognitive tasks in mouse models of neurodevelopmental disorders.

Automated cognitive testing of monkeys in social groups yields results comparable to individual laboratory-based testing

Cognitive abilities likely evolved in response to specific environmental and social challenges and are therefore expected to be specialized for the life history of each species. Specialized cognitive abilities may be most readily engaged under conditions that approximate the natural environment of the species being studied. While naturalistic environments might therefore have advantages over laboratory settings for cognitive research, it is difficult to conduct certain types of cognitive tests in these settings. We implemented methods for automated cognitive testing of monkeys (Macaca mulatta) in large social groups (Field station) and compared the performance to that of laboratory-housed monkeys (Laboratory). The Field station animals shared access to four touch-screen computers in a large naturalistic social group. Each Field station subject had an RFID chip implanted in each arm for computerized identification and individualized assignment of cognitive tests. The Laboratory group was housed and tested in a typical laboratory setting, with individual access to testing computers in their home cages. Monkeys in both groups voluntarily participated at their own pace for food rewards. We evaluated performance in two visual psychophysics tests, a perceptual classification test, a transitive inference test, and a delayed matching-to-sample memory test. Despite the differences in housing, social environment, age, and sex, monkeys in the two groups performed similarly in all tests. Semi-free ranging monkeys living in complex social environments are therefore viable subjects for cognitive testing designed to take advantage of the unique affordances of naturalistic testing environments.