Animal Metacognition: A Tale of Two Comparative Psychologies

Animal Metacognition: A Tale of Two Comparative Psychologies Smith Couchman Beran

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Animal Metacognition: A Tale of Two Comparative Psychologies 2014 Journal of comparative psychology (Washington, D.C. : 1983) 10.1037/a0033105 Smith2014-animal-metacognition-a-tale-of-two-comparati Smith2014-animal-metacognition-a-tale-of-two-comparati.txt, the project's staged copy bf71bfa965e22747c33d31fc5e7e1bb8801ed760 not-redistributable psychology

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them The associative models offered to explain animals' uncertainty responses, and what adopting them costs comparative psychology theoretically. Quotation withheld because this record is not cleared for redistribution. critique Quotation withheld because this record is not cleared for redistribution. animal metacognition (uncertainty monitoring) Quotation withheld because this record is not cleared for redistribution. review review Quotation withheld because this record is not cleared for redistribution. mechanisms When deferred feedback camouflages the task's reinforcement structure, the animal builds its own subjective decisional framework and responds uncertain to whatever is near its own subjective breakpoint, not the task's objective one Quotation withheld because this record is not cleared for redistribution. mechanisms Uncertainty responses draw on executive attention and working memory, which is why occupying those resources with a concurrent task selectively degrades them while leaving stimulus-driven perceptual classification intact Quotation withheld because this record is not cleared for redistribution. mechanisms In Hampton's task no memory-relevant stimulus is present at choice time, so the animal must itself initiate and restrict a search of memory for a strong trace — an operation with no associative cue to react to Quotation withheld because this record is not cleared for redistribution. mechanisms Casting every performance as a register of activation strengths updated by reinforcement artificially equates psychologically different performances, which is what empties the label 'associative' of content Quotation withheld because this record is not cleared for redistribution. The uncertainty response is designed to carry no reward of its own, so choosing it cannot be explained by direct reinforcement of that response Quotation withheld because this record is not cleared for redistribution. In the memory task the response icons carry no trace-strength information, so there is nothing on the screen for an association to attach to - the animal must initiate its own search Quotation withheld because this record is not cleared for redistribution. The search must also be restricted to the right memory locations, and nothing in the display can tell the animal how to do that Quotation withheld because this record is not cleared for redistribution. The associative models work by mapping every performance onto one abstract register of activations updated by reinforcement, which is precisely what makes them unable to tell performances apart Quotation withheld because this record is not cleared for redistribution. The reductio: the same formalism works with a register of fully conscious metacognitive states updated by reinforcement, which shows the formalism settles nothing Quotation withheld because this record is not cleared for redistribution. The all-or-none framing is itself the error: uncertainty monitoring may exist in graded stages without the whole human package Quotation withheld because this record is not cleared for redistribution. The authors concede that not all of the models they criticise have had a fair opportunity to address the newest phenomenon. what this limits: the critique of the newest models: THERE HAS NOT BEEN TIME TO ADDRESS IT YET Quotation withheld because this record is not cleared for redistribution. They hedge their own positive interpretation: information-seeking need not imply full conscious metacognition. what this limits: the positive interpretation: ONE NEED NOT ATTRIBUTE FULL, CONSCIOUS, SELF-IMBUED METACOGNITION TO MACAQUES EVEN IF ONE ADOPTS THIS INFORMATION-SEEKING HYPOTHESIS Quotation withheld because this record is not cleared for redistribution. They acknowledge they cannot presently settle whether one criticised model succeeds in practice. what this limits: the verdict on that model: IT IS PRESENTLY DIFFICULT TO JUDGE THE MODEL'S PRAGMATIC SUCCESS Quotation withheld because this record is not cleared for redistribution. The empirical results themselves are not in dispute - the whole disagreement is over interpretation what this limits: the nature of the whole dispute: THESE EMPIRICAL RESULTS ARE NOT DISPUTED - the disagreement is interpretive, so there is no discrepant measurement to find Quotation withheld because this record is not cleared for redistribution. The animals report nothing, so the human evidence for conscious uncertainty has no counterpart in the animal case what this limits: any transfer of the human evidence for conscious uncertainty: MACAQUES HAVE ALSO SHOWN THIS PATTERN REPEATEDLY, THOUGH THEY REPORT NOTHING Quotation withheld because this record is not cleared for redistribution. The paper's own positive claim is hedged to possibly-metacognitive rather than metacognitive what this limits: the paper's own conclusion, read exactly: higher-level, cognitive, deliberate and decisional are asserted; POSSIBLY METACOGNITIVE AND CONSCIOUS is hedged Quotation withheld because this record is not cleared for redistribution. The critique is of a specific class of models, which the authors accept would be a legitimate goal if it could be made to work what this limits: the standing of the opposing programme: describing the performances in low-level associative terms is AN IMPORTANT GOAL IF ACHIEVABLE Quotation withheld because this record is not cleared for redistribution. the Sparse-Uncertain-Dense task interleaved with a Sparse-Middle-Dense task for the capuchin monkeys eliminates: the monkeys learning a fixed response to the uncertain condition - interleaving a task where the middle option is not uncertainty keeps the two from being solved as one · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. metacognition metacognition Quotation withheld because this record is not cleared for redistribution. associative - the specific sense: responding cued by a present stimulus or trained by reinforcement contingencies Quotation withheld because this record is not cleared for redistribution. associative - the stretched sense: anything expressible as a register of activations updated by reward, which covers everything and therefore distinguishes nothing Quotation withheld because this record is not cleared for redistribution. the task - with trace-relevant stimuli present (matching-to-sample): an associative account applies Quotation withheld because this record is not cleared for redistribution. the task - with memory-neutral response icons (Hampton): nothing is present to be associated with, so the same account does not carry over Quotation withheld because this record is not cleared for redistribution. metacognition - as all-or-none: the Yes/No framing the paper says the dichotomy itself may be wrong for the field's goal Quotation withheld because this record is not cleared for redistribution. metacognition - as graded: a basic uncertainty-monitoring system without consciousness or self-awareness attached Quotation withheld because this record is not cleared for redistribution. contradicts → Le Pelley (2012), associative-learning model of animal metacognition Quotation withheld because this record is not cleared for redistribution. contradicts → Jozefowiez et al. (2009a) differs in: whether the trigger is present or must be found: JOZEFOWIEZ ET AL. ARE INCORRECT WHEN THEY CLAIM THAT THE UNDERLYING PROCESSES ARE EXACTLY THE SAME WHETHER A TRIGGERING STIMULUS IS ON THE SCREEN OR THE ANIMAL MUST INITIATE ITS OWN TARGETED SEARCH FOR ACTIVE MEMORY TRACES Quotation withheld because this record is not cleared for redistribution. builds-on → Hampton (2001), macaque metamemory studies Quotation withheld because this record is not cleared for redistribution. critiques → associative models of animal uncertainty responses (Jozefowiez et al., Le Pelley, Staddon et al.) differs in: the same parameter as item 1 - a triggering stimulus ON THE SCREEN against the animal having to INITIATE ITS OWN TARGETED SEARCH FOR ACTIVE MEMORY TRACES Quotation withheld because this record is not cleared for redistribution. critiques → strict application of Morgan's canon in comparative psychology differs in: how strictly the canon is applied: THOSE MODELS REFLECT AN ULTIMATELY STRICT ADHERENCE TO MORGAN'S (1906) CANON, WHICH EVEN MORGAN DID NOT INTEND AND WHICH MANY COMPARATIVE-COGNITION RESEARCHERS DO NOT ACCEPT NOW Quotation withheld because this record is not cleared for redistribution. supports → Hampton (2001) and the memory-monitoring evidence for macaque metacognition Quotation withheld because this record is not cleared for redistribution. supports → Metcalfe's operational benchmark of stimulus absence Quotation withheld because this record is not cleared for redistribution. uncertainty response —operationalised-as→ uncertainty monitoring uncertainty response uncertainty monitoring strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the uncertainty response is shown to be more than a middle category, with monitoring mediated by a distinct network; operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. executive —causes→ uncertainty response executive uncertainty response strength of the relation: 0.9 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says UR processes are executive; causes is loose. Quotation withheld because this record is not cleared for redistribution. metacognition —part-of→ declarative metacognition declarative strength of the relation: 0.9 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span links metacognition to the declarative system; part-of is loose. Quotation withheld because this record is not cleared for redistribution. self-awareness —distinct-from→ uncertainty monitoring self-awareness uncertainty monitoring strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. reinforcement —causes→ registers reinforcement registers strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. phylogenetic —causes→ metacognition phylogenetic metacognition strength of the relation: 0.8 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says phylogeny probably gave metacognition in stages; causes states a probability. Quotation withheld because this record is not cleared for redistribution. correlation correlation Quotation withheld because this record is not cleared for redistribution. goodness-of-fit measure goodness-of-fit measure Quotation withheld because this record is not cleared for redistribution. model fitting model fitting Quotation withheld because this record is not cleared for redistribution. signal detection theory model signal detection theory model Quotation withheld because this record is not cleared for redistribution. error of prediction error of prediction Quotation withheld because this record is not cleared for redistribution. Le Pelley's associative-learning model never produces anything like the uncertainty-response pattern the macaques showed. Quotation withheld because this record is not cleared for redistribution. At its best fit the associative model mispredicts each of 120 data points by .10 on average. Quotation withheld because this record is not cleared for redistribution. A non-associative uncertainty model fits the same data more than three times more closely, with an AAD of .03. Quotation withheld because this record is not cleared for redistribution. The better-fitting cognitive model is also the more parsimonious one, using 3 free parameters against the associative model's 5. Quotation withheld because this record is not cleared for redistribution. The associative model fits best only by giving up the crucial result: the macaque's asymmetric centering of uncertainty responses at Level 14. Quotation withheld because this record is not cleared for redistribution. Capuchin monkeys essentially did not use the uncertainty response, even though they received immediate feedback after every trial. Quotation withheld because this record is not cleared for redistribution. A concurrent memory load selectively suppressed macaques' uncertainty responses while leaving their Sparse, Middle and Dense perceptual responses largely intact. Quotation withheld because this record is not cleared for redistribution. Pigeons do not express any memory-monitoring capacity. Quotation withheld because this record is not cleared for redistribution. Pigeons' uncertainty-response rate stays flat across retention intervals, though they remember less after longer delays. Quotation withheld because this record is not cleared for redistribution. With the added serial-position parameters, Le Pelley's associative model would have nine free parameters. Quotation withheld because this record is not cleared for redistribution. The authors hold that Le Pelley's published simulations must be discounted because they misstate a crucial aspect of Hampton's method. Quotation withheld because this record is not cleared for redistribution. With the delay interval controlled, macaques made far more uncertainty responses on blank trials. Quotation withheld because this record is not cleared for redistribution. Monkeys defended the same criterion of memory strength in two separate studies, a convergence across laboratories, animals and methods. Quotation withheld because this record is not cleared for redistribution. The authors argue that the associative models reflect a strict adherence to Morgan's Canon that even Morgan did not intend. Quotation withheld because this record is not cleared for redistribution. animals show uncertainty responses where no cuing stimulus is present, on first trials, and without feedback Quotation withheld because this record is not cleared for redistribution. chimpanzees checked the container before naming only on trials where they had not seen its contents Quotation withheld because this record is not cleared for redistribution. the empirical pattern is agreed by both sides; only its interpretation is disputed Quotation withheld because this record is not cleared for redistribution. a broad enough associative label can no longer separate performances that differ psychologically Quotation withheld because this record is not cleared for redistribution. the claim implies any information processing can be lumped under 'associative', which makes the term theoretically useless Quotation withheld because this record is not cleared for redistribution. metacognition metacognition Quotation withheld because this record is not cleared for redistribution. “Middle” response “Middle” response Quotation withheld because this record is not cleared for redistribution. Average Absolute Deviation (AAD) Average Absolute Deviation (AAD) Quotation withheld because this record is not cleared for redistribution. fulcrum point of URs fulcrum point of URs Quotation withheld because this record is not cleared for redistribution. subjective difficulty subjective difficulty Quotation withheld because this record is not cleared for redistribution. uncertainty response uncertainty response Quotation withheld because this record is not cleared for redistribution. Delayed-Matching-to-Sample Delayed-Matching-to-Sample Quotation withheld because this record is not cleared for redistribution. concurrent working-memory load concurrent working-memory load Quotation withheld because this record is not cleared for redistribution. penalty timeout penalty timeout Quotation withheld because this record is not cleared for redistribution. retention interval retention interval Quotation withheld because this record is not cleared for redistribution. r = 0.838 — correlation between the proportion of trials the macaque declined and the distance of the trial level from his own subjective decisional breakpoint Quotation withheld because this record is not cleared for redistribution. .10 — best-case average absolute deviation of the associative-learning model per data point Quotation withheld because this record is not cleared for redistribution. .03 — average absolute deviation of the non-associative uncertainty model on the same data Quotation withheld because this record is not cleared for redistribution. about .30 — per-data-point error of the associative model when constrained to the UR fulcrum of 14 that the real macaque showed Quotation withheld because this record is not cleared for redistribution. 5% — error of prediction of the earlier signal-detection associative model between predicted and observed response proportions Quotation withheld because this record is not cleared for redistribution. 7% — average per-stimulus-level misprediction of the uncertainty response by that associative model across levels 1-31 Quotation withheld because this record is not cleared for redistribution. 23% ... 97% — macaque Murph's uncertainty-response rate at Level 9 against his accuracy on those same trials, showing URs uncoupled from reinforcement Quotation withheld because this record is not cleared for redistribution. 19% ... 22% — the same macaque's uncertainty-response rate at Level 20 against his accuracy there, far below chance yet declined less often than the easy level Quotation withheld because this record is not cleared for redistribution. 14 and 0.12 — the real macaque's position in the simulation space (UR fulcrum, mean UR rate), which the associative model's runs never reach Quotation withheld because this record is not cleared for redistribution. 90 s from 20s — the penalty timeout increase used to raise associative pressure on capuchins to adopt the uncertainty response Quotation withheld because this record is not cleared for redistribution. five of six — capuchin monkeys who still failed to use the uncertainty response adaptively even after the timeout was quadrupled Quotation withheld because this record is not cleared for redistribution. 10% — the constant error rate macaques maintained as memory lists lengthened, by declining more tests Quotation withheld because this record is not cleared for redistribution. one of five — capuchins who searched before reaching more often on unseen-hiding trials in the food-concealment paradigm Quotation withheld because this record is not cleared for redistribution. 41-step continuum — the density scale along which stimuli are classified sparse or dense Quotation withheld because this record is not cleared for redistribution. Level 21 — the exact breakpoint, never presented, around which the uncertainty response should cluster Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1037/a0033105.