It is always good to be reminded of the fact that we are not the only species on earth that sports advanced cognitive capacities. Equally welcome is the reminder that advanced cognitive capacity by no means is a requirement for success in the game of life, since the overwhelming majority of species do perfectly well without it, witness untold millions of successful species of bacteria (Dykhuizen, Reference Dykhuizen2005). Sheryl Coombs and Michael Trestman supply us with both of these reminders in their valiant effort to put some order regarding cognitive evolution into the bewildering assemblage of diverse forms and capacities that make up the animal kingdom.
In so doing, they have naturally faced formidable obstacles in the choice of criteria to adopt, in how to operationally define “traits of interest,” and not least in the incompleteness with which animal behavioral capacities have been mapped across phyla. The authors are to be commended for not letting the obstacles divert them from the goal of getting a systematic “first look” at the lay of the cognitive landscape on earth. In the absence of an unequivocal behavioral measure of cognitive capacity applicable across phyla, they have pursued a multitrait strategy of canvassing sets of bodily, sensory, brain, motor and cognitive traits across phyla, three of which phyla contain member species that qualify for advanced cognitive status by possessing a full set of nine “pivotal” traits whose co-occurrence identifies advanced cognition in the authors’ approach.
Space limitations force me to concentrate my commentary on a single issue, the authors’ adoption of the perspective on learning abilities introduced by Simona Ginsburg and Eva Jablonka in their search for a “transition marker” for consciousness. Learning capacity, as a purely cognitive trait in the Coombs and Trestman cognition/behavior category, comes closer to cognitive capacity itself than any of their other pivotal traits they employ, and accordingly plays a prominent role in their treatment. This is unfortunate because the Ginsburg and Jablonka treatment of learning is problematic at both ends of their spectrum of learning capacities.
To begin at its “simple” end, Ginsburg and Jablonka follow the textbook error of treating habituation as a simple form of learning because it is “non-associative”. That purported simplicity was put to rest by the seminal record of empirical studies on the habituation of the orienting response and its physiological correlates reported by Evgeni N. Sokolov in the 1960s. His detailed behavioral and physiological studies led to the inescapable conclusion that in the course of habituation the nervous system acquires an accurate and multifaceted “neuronal model” of the stimulus (whether simple or compound), since any change in any discriminable stimulus dimension, including a decrement in stimulus intensity, instantly restores behavioral and physiological responsiveness (Sokolov, Reference Sokolov and Brazier1960, Reference Sokolov1963). This means that incoming sensory afference is continually compared against this acquired neuronal model, which defines what is novel by what differs from it and therefore what requires attention. It therefore constitutes what in modern terminology is called a prior, by means of which prediction error is defined (see, for example, Clark, Reference Clark2013). In other words, depending on the type of brain in which it takes place, habituation is integral to even the most advanced cognition.
Things stand no better at the high end of the Ginsburg and Jablonka learning spectrum. The “unlimited associative learning” (UAL) they introduced and named is not “a well-studied suite of experimentally determined learning abilities,” as Coombs and Trestman suppose it to be. Rather it is a hypothetical construct Ginsburg and Jablonka proposed to be a “transition marker for sentience” in an attempt to find an analogue in the domain of consciousness (or sentience) for Tibor Gánti’s transition marker for “life”. Gánti had distilled a set of informal characterizations of “life” into a formal minimalist working model of life which he called the chemoton (Gánti, Reference Gánti2003). It features a metabolism unit in tandem with a heredity unit, tasked with cell wall maintenance and replication. To perform the feat of founding life as we know it the heredity unit must not be narrowly constrained but open to perpetual modification, for which Gánti chose the infelicitous expression “unlimited heredity” (where “open-ended” would fit the facts better).
In order to duplicate Gánti’s feat for consciousness, Ginsburg and Jablonka would have to distill the many informal characterizations of consciousness into a formal minimalist working model of sentience and then derive their transition marker from some crucial component within it. There is no such constitutive modeling in their approach. Instead, they go straight to a “transition marker” without having specified in formal terms that for which it is to be a marker, as did Gánti for life. Applied to the Coombs and Trestman evolutionary quest, the resulting speculative construct of “unlimited associative learning” is problematic.
In Ginsburg and Jablonka’s hands, UAL was proposed as a marker for consciousness or sentience, and not for advanced cognition, though on the face of it there would seem to be greater reason to make a form of learning a central feature of advanced cognition than of consciousness. Without a constitutive characterization of consciousness (for “constitutive,” see Merker, Williford & Rudrauf, Reference Merker, Williford and Rudrauf2022, pp. 11–12), there is no way of knowing whether advanced cognition and consciousness are even related, something which some of us doubt (e. g. Merker, Reference Merker2007, which holds consciousness to be compatible with the most profound dementia). The problem with deploying UAL as a pivotal trait for advanced cognition is its status as a so far hypothetical construct fashioned in a flawed (see previous paragraph) attempt to do for consciousness what Gánti did for life. Its empirical moorings are tenuous (for which see Birch, Reference Birch2020 and Birch, Ginsberg & Jablonka, Reference Birch, Ginsburg and Jablonka2020) and do not supply a secure basis for assigning evolutionary character states. It is, for example, so far unknown whether euarthropods and coleoid cephalopods possess the cluster of capacities Ginsburg and Jablonka call UAL.
As a commentary, this must suffice, and I thank the authors for providing the stimulus to these reflections.
You have
Access
It is always good to be reminded of the fact that we are not the only species on earth that sports advanced cognitive capacities. Equally welcome is the reminder that advanced cognitive capacity by no means is a requirement for success in the game of life, since the overwhelming majority of species do perfectly well without it, witness untold millions of successful species of bacteria (Dykhuizen, Reference Dykhuizen2005). Sheryl Coombs and Michael Trestman supply us with both of these reminders in their valiant effort to put some order regarding cognitive evolution into the bewildering assemblage of diverse forms and capacities that make up the animal kingdom.
In so doing, they have naturally faced formidable obstacles in the choice of criteria to adopt, in how to operationally define “traits of interest,” and not least in the incompleteness with which animal behavioral capacities have been mapped across phyla. The authors are to be commended for not letting the obstacles divert them from the goal of getting a systematic “first look” at the lay of the cognitive landscape on earth. In the absence of an unequivocal behavioral measure of cognitive capacity applicable across phyla, they have pursued a multitrait strategy of canvassing sets of bodily, sensory, brain, motor and cognitive traits across phyla, three of which phyla contain member species that qualify for advanced cognitive status by possessing a full set of nine “pivotal” traits whose co-occurrence identifies advanced cognition in the authors’ approach.
Space limitations force me to concentrate my commentary on a single issue, the authors’ adoption of the perspective on learning abilities introduced by Simona Ginsburg and Eva Jablonka in their search for a “transition marker” for consciousness. Learning capacity, as a purely cognitive trait in the Coombs and Trestman cognition/behavior category, comes closer to cognitive capacity itself than any of their other pivotal traits they employ, and accordingly plays a prominent role in their treatment. This is unfortunate because the Ginsburg and Jablonka treatment of learning is problematic at both ends of their spectrum of learning capacities.
To begin at its “simple” end, Ginsburg and Jablonka follow the textbook error of treating habituation as a simple form of learning because it is “non-associative”. That purported simplicity was put to rest by the seminal record of empirical studies on the habituation of the orienting response and its physiological correlates reported by Evgeni N. Sokolov in the 1960s. His detailed behavioral and physiological studies led to the inescapable conclusion that in the course of habituation the nervous system acquires an accurate and multifaceted “neuronal model” of the stimulus (whether simple or compound), since any change in any discriminable stimulus dimension, including a decrement in stimulus intensity, instantly restores behavioral and physiological responsiveness (Sokolov, Reference Sokolov and Brazier1960, Reference Sokolov1963). This means that incoming sensory afference is continually compared against this acquired neuronal model, which defines what is novel by what differs from it and therefore what requires attention. It therefore constitutes what in modern terminology is called a prior, by means of which prediction error is defined (see, for example, Clark, Reference Clark2013). In other words, depending on the type of brain in which it takes place, habituation is integral to even the most advanced cognition.
Things stand no better at the high end of the Ginsburg and Jablonka learning spectrum. The “unlimited associative learning” (UAL) they introduced and named is not “a well-studied suite of experimentally determined learning abilities,” as Coombs and Trestman suppose it to be. Rather it is a hypothetical construct Ginsburg and Jablonka proposed to be a “transition marker for sentience” in an attempt to find an analogue in the domain of consciousness (or sentience) for Tibor Gánti’s transition marker for “life”. Gánti had distilled a set of informal characterizations of “life” into a formal minimalist working model of life which he called the chemoton (Gánti, Reference Gánti2003). It features a metabolism unit in tandem with a heredity unit, tasked with cell wall maintenance and replication. To perform the feat of founding life as we know it the heredity unit must not be narrowly constrained but open to perpetual modification, for which Gánti chose the infelicitous expression “unlimited heredity” (where “open-ended” would fit the facts better).
In order to duplicate Gánti’s feat for consciousness, Ginsburg and Jablonka would have to distill the many informal characterizations of consciousness into a formal minimalist working model of sentience and then derive their transition marker from some crucial component within it. There is no such constitutive modeling in their approach. Instead, they go straight to a “transition marker” without having specified in formal terms that for which it is to be a marker, as did Gánti for life. Applied to the Coombs and Trestman evolutionary quest, the resulting speculative construct of “unlimited associative learning” is problematic.
In Ginsburg and Jablonka’s hands, UAL was proposed as a marker for consciousness or sentience, and not for advanced cognition, though on the face of it there would seem to be greater reason to make a form of learning a central feature of advanced cognition than of consciousness. Without a constitutive characterization of consciousness (for “constitutive,” see Merker, Williford & Rudrauf, Reference Merker, Williford and Rudrauf2022, pp. 11–12), there is no way of knowing whether advanced cognition and consciousness are even related, something which some of us doubt (e. g. Merker, Reference Merker2007, which holds consciousness to be compatible with the most profound dementia). The problem with deploying UAL as a pivotal trait for advanced cognition is its status as a so far hypothetical construct fashioned in a flawed (see previous paragraph) attempt to do for consciousness what Gánti did for life. Its empirical moorings are tenuous (for which see Birch, Reference Birch2020 and Birch, Ginsberg & Jablonka, Reference Birch, Ginsburg and Jablonka2020) and do not supply a secure basis for assigning evolutionary character states. It is, for example, so far unknown whether euarthropods and coleoid cephalopods possess the cluster of capacities Ginsburg and Jablonka call UAL.
As a commentary, this must suffice, and I thank the authors for providing the stimulus to these reflections.
Financial support
The author had no external support for the writing of this commentary.
Competing interests
The author declares no competing interests.