A naturalistic environment to study visual cognition in unrestrained monkeys

  1. Georgin Jacob
  2. Harish Katti
  3. Thomas Cherian
  4. Jhilik Das
  5. KA Zhivago
  6. SP Arun

  • Curated by eLife

    eLife logo

    Summary: This manuscript describes a new experimental environment for training macaque monkeys to perform behavioral tasks. Using this facility, the authors trained freely moving macaques to perform a visual "same-different" task using operant conditioning, and under voluntary head restraint. The authors demonstrate that they could obtain reliable eye-tracking data and high-performance accuracy from macaques in this facility. They also noted that subordinate macaques can learn to perform basic aspects of the task by observing their dominant conspecifics perform the task in this facility. The authors conclude that this naturalistic environment can facilitate the study of brain activity during natural and controlled behavioral tasks.

    The manuscript is broadly organized along three distinct lines of inquiry. First, the authors describe a customized living space for a small group of macaque monkeys. Second, the authors train two of these monkeys to perform a cognitive task in a purpose-built room of the living enclosure. Third, the authors describe their experience training a third monkey to complete the cognitive task.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Macaque monkeys are widely used to study vision. In the traditional approach, monkeys are brought into a lab to perform visual tasks while they are restrained to obtain stable eye tracking and neural recordings. Here, we describe a novel environment to study visual cognition in a more natural setting as well as other natural and social behaviors. We designed a naturalistic environment with an integrated touchscreen workstation that enables high-quality eye tracking in unrestrained monkeys. We used this environment to train monkeys on a challenging same-different task. We also show that this environment can reveal interesting novel social behaviors. As proof of concept, we show that two naive monkeys were able to learn this complex task through a combination of socially observing trained monkeys and solo trial-and-error. We propose that such naturalistic environments can be used to rigorously study visual cognition as well as other natural and social behaviors in freely moving monkeys.

Article activity feed

  1. Version published to 10.7554/elife.63816 on eLife
  2. eLife

    Reviewer #3:

    Jacob and colleagues developed a new experimental "facility" or environment for training macaque monkeys to perform behavioral tasks. Using this facility, the authors trained freely moving macaques to perform a visual "same-different" task using operant conditioning, and under voluntary head restraint. The authors demonstrate that they could obtain reliable eye-tracking data and high performance accuracy from macaques in this facility. They also noted that subordinate macaques can learn to perform basic aspects of the task by observing their dominant conspecifics perform the task in this facility. The authors conclude that this naturalistic environment can facilitate the study of brain activity during natural and controlled behavioral tasks.

    The manuscript is doubtless a hard-fought effort. The new experimental platform introduced by the authors has the capacity to transform how researchers approach the behavioral training of monkeys for some (but not all) tasks. However, in my opinion, the manuscript would have significantly broader impact and appeal if the authors had succeeded in performing wireless neural recordings in this same environment. Without these proof-of-principle neural data, the scope of this manuscript seems more limited. If the authors can obtain these neural data, the manuscript would be substantially stronger.

    There are a few other concerns related to methodology and interpretation that should be addressed.

    Major comments:

    1. In the abstract, the authors state that macaques are widely used to study the neural basis of cognition - but in fact these animals are a valuable model organism for studying many other aspects of brain function beyond cognition. The authors seem to be missing an opportunity to highlight the broad impact of their work.

    2. A gaze window of 3 degrees is rather large for most visual-based experiments. Do the authors think that it would be possible to train animals to maintain tighter fixation windows? And have they tried to do so?

    3. Are these animals water deprived before entering the experimental environment? And how long do the animals typically work in this environment? For how many hours, and for how much fluid?

    4. How did the authors ensure that the macaques do not fight inside the facility? Are the animals continuously housed in this facility or are they moved into this facility only during testing?

    5. Line 227: the authors state the following: "Remarkably, M2 learned the task much faster using social observation and learning than M1 & M3 did using the TAT paradigm". How do the authors rule out the possibility that M2 is simply a "smarter" animal?

    6. Line 354-364: the authors describe their insights about how animals may learn to perform the task in two phases. How can the authors make these strong claims based on data from N=1 macaque?

  3. eLife

    Reviewer #2:

    The manuscript "A naturalistic environment to study natural social behaviors and cognitive tasks in freely moving monkeys" describes a large-scale system of rooms allowing for non-human primates to, potentially, freely engage in several different behaviors and neuroscientific experiments to be performed. The study is well intended, but in its current form with many claims, but few if any results does not, in my view, meet scientific standards.

    The paper presents the testing environment consisting of different rooms. Compared to earlier work (e.g. Berger et al., 2018), the main innovation is the inclusion of an eye tracking system. Data supports the notion that this works in principle. But there is no analysis of data quality and accuracy. We also do not know whether the system works on every trial, or how often the eye is not detected or the tracker loses the signal.

    The authors claim novelty of this testing environment, but similar ones have been used in behavioral research for decades and in recent years in neuroscience.

    The authors claim that it is easier to place a testing system into a separate cage then in the home cage. It remains unclear what this claim is based on. Motivation of animals in these social settings should be more difficult than in the home cage environment. So, this is a potentially interesting result. It is also a conceptually important claim for the paper's logic, if the social setting should really be beneficial for training. But the claim needs to be substantiated.

    The authors claim that natural behavior can be analyzed because a CCTV camera is mounted in the cage. There are no results or analyses to demonstrate that.

    The authors mention neural recordings on multiple occasions, but do not show any. EM shielding is neither necessary nor new.

    Automatic training appears to be a one-to-one copy of that in Berger et al. 2018, but citation is missing, except for Supplemental Information.

    The authors report an anecdote of one animal (n=1) learning socially from others. There is no indication that this subject might have performed differently without social learning. The interpretation is a just-so story and appears rather anthropomorphic.

    There are no results in the manuscript.

    The manuscript is not organized well. The Methods section reads like a Discussion, important information on methods is distributed across Supplemental Information and Results. Results, as mentioned, does not contain any results or data.

  4. eLife

    Reviewer #1:

    I'm quite enthusiastic about the care the authors have taken in designing this cutting edge hybrid environment, and the effort they've gone through to describe it in detail. I believe that this endeavor has great merit, and that seeing the advancements in animal welfare and experimentation should be of interest to the general reader. However, at present, the stated interpretations are not fully justified by the results, and this must be addressed.

    The manuscript should be amended and updated in one of two possible ways: the interpretations of the scientific result here should be tamped down significantly, or additional evidence should be presented for some of the claims in the originally submitted manuscript. I am confident that the authors should be able to carry out either of these to a satisfying degreen.

    Major issues:

    1. Throughout the manuscript, stating that the third monkey learned the task "merely by observing two other trained monkeys" is misleading. The naive monkey may have learned very important details about the cognitive testing set-up from observation. But the third monkey learned the task of a unique behavioural shaping paradigm that included -but was not limited to- watching trained monkeys. The authors trained the third monkey on the cognitive task in the absence of the other monkeys, and do not show that the third monkey learned the specific cognitive task from watching other monkeys. Over-interpreting the anecdotal observations here hinders obfuscates what is novel and notable in this manuscript.

    2. The authors repeatedly state that the third monkey learned the task faster than the previous two monkeys. It is quite difficult to parse exactly what the authors mean by this, and exactly what the data is that supports that claim.

      The authors go on to state that M2 learned the "task structure" faster than M1/M3. However, "task structure" is not defined, so it is difficult for a reader to know precisely what was learned faster under social observation. Furthermore, the data showing that M2 learned the task structure faster than M1/M3 is not clear, and it is not known how M1/M3 learned the task structure in isolation. Description of which training steps may be aided by observation of trained monkeys must be clarified. The authors allowed M2 to observe M1 and M3 during initial familiarization of the experimental set-up, but it seems that observation may not have aided M2 in learning the complex same-different task at all.

      Even though M2 may have learned the task structure faster than M1/M3, these observations are anecdotal and should not be over-interpreted. If there is a clear difference in the time to learn basic task structure, it may be due to social observation, but the authors should not favor that interpretation without considering alternatives as well. E.g., monkeys have widely varying personalities (see e.g. Capitanio 1999, Am J Primatology), and this has important implications for the curiosity, exploration behavior, and likelihood to accept and complete new challenges in training. To what extent could the differences in learning rate also be explained by these differences across these 3 monkeys? To what extent does the different training regimen in the task explain differences in learning rate across monkeys (e.g. M2 got two days of repeating correction trials, which significantly alters learning rates)?

    3. There is a vast literature in ethological settings where the gaze of nonhuman primates has been tracked using noninvasive methods that the authors do not acknowledge. Instead, authors state that most infrared eye trackers require head restraint (line 32), though this is demonstrably not the case. For review, see Hopper et al. 2020, Behav Res Methods.

    4. Some important details for introducing monkeys to the testing apparatus during Tailored Automated Training should be described. For example, were animals water-restricted, or on any sort of fluid restriction when TAT began? How did the authors entice the animals to initially explore the testing apparatus?

  5. eLife

    Summary: This manuscript describes a new experimental environment for training macaque monkeys to perform behavioral tasks. Using this facility, the authors trained freely moving macaques to perform a visual "same-different" task using operant conditioning, and under voluntary head restraint. The authors demonstrate that they could obtain reliable eye-tracking data and high-performance accuracy from macaques in this facility. They also noted that subordinate macaques can learn to perform basic aspects of the task by observing their dominant conspecifics perform the task in this facility. The authors conclude that this naturalistic environment can facilitate the study of brain activity during natural and controlled behavioral tasks.

    The manuscript is broadly organized along three distinct lines of inquiry. First, the authors describe a customized living space for a small group of macaque monkeys. Second, the authors train two of these monkeys to perform a cognitive task in a purpose-built room of the living enclosure. Third, the authors describe their experience training a third monkey to complete the cognitive task.

  6. Version published to 10.1101/2020.09.25.311555 on bioRxiv