CoChat
Sign Up
Cited by 23Action Observation and SynchronizationSocial Robot Interaction and HRIEmbodied and Extended Cognition

Automatic imitation of multiple agents: A computational model

Open access
Cracco, Emiel, Cooper, Richard P.·2019·Cognitive Psychology
DOI 10.1016/j.cogpsych.2019.101224
Open this paper in CoChat — free accountOpen PDF

Abstract

There is accumulating evidence that the actions of others are represented in the motor system, leading to automatic imitation. However, whereas early work focused mainly on the effects of observing a single agent, recent studies indicate that the actions of multiple agents can be represented simultaneously. Yet, theorizing has lagged behind. The current study extends the dual-route model of automatic imitation to include multiple agents, and demonstrates, in five simulation studies, that the extended model is able to capture four critical multi-agent effects. Importantly, however, it was necessary to augment the model with a control mechanism regulating response inhibition based on the number of observed actions. Furthermore, additional simulation indicated that this mechanism could be driven by response conflict. Together, our results demonstrate how theories of automatic imitation can be extended from single- to multi-agent settings. As such, they constitute an important step towards a mechanistic understanding of social interaction beyond the dyad.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper extends the dual-route model of automatic imitation to multi-agent settings and tests it in five simulation studies against four previously reported multi-agent effects. It finds the model can reproduce those effects only after adding a control mechanism that regulates response inhibition according to the number of observed actions, and a further simulation suggests this mechanism could be driven by response conflict.

Key findings

  • Across five simulation studies, the extended model captured four critical multi-agent effects: stronger imitation for two identical observed actions, no net imitation for two different actions, asymptotic congruent RTs as identical hands increase in the mixed procedure, and reduced imitation when three hands match and one differs.
  • To capture the group-size effects, the model required a visual flux node that inhibits response nodes based on the number of moving hands; this produced an asymptote on congruent trials in the mixed procedure (50% congruent, 50% incongruent) but not in the non-mixed procedure (100% congruent).
  • Simulation 4 added a control node updated by response conflict and closely reproduced both the empirical results and Simulation 3, suggesting the dynamic inhibition process could be driven by response conflict.
  • Parameter variation studies indicated the key effects were robust across almost the entire parameter space when facilitation and interference were balanced.

Methods

  • The model is a connectionist interactive activation model extending Cooper et al. (2013): stimulus nodes (imperative, hand, visual flux) feed into response nodes; hand and imperative nodes are excitatory, the visual flux node is inhibitory, and the two response nodes inhibit each other laterally. Model A includes two hands; Models B and C include four hands.
  • Simulations are based on prior experiments: two-hand tasks from Cracco et al. (2015) and four-hand tasks from Cracco and Brass (2018b, 2018d), run with the same number of trials and participants as the original experiments. Most parameter values were taken from Cooper et al. (2013), with new parameters explored via parameter variation.
  • Main analyses include five simulation studies, parameter variation across \(w_{rr}\), \(\alpha\), and \(w_{rf}\), BIC comparisons in Simulation 3 for asymptotic versus linear fits on congruent trials and linear versus intercept fits on incongruent trials, and in Simulation 4 trial-by-trial updating of control-to-response weights based on response conflict over the past \(n = 16\) trials.

Limitations

  • Stated: the model focused on mean performance and did not model performance variability; the main model did not vary parameters across participants, though a supplementary model with interindividual variability in \(w_{ri}\) (SD = 0.70), \(w_{rh}\) (SD = 0.50), and \(\rho_h\) (SD = 0.02) reproduced the dispersion of the empirical data.
  • Stated: the model has a relatively large number of free parameters; this was mitigated by taking most values from Cooper et al. (2013) and by parameter variation studies.
  • Stated: a simpler alternative could inhibit motor activity based on hand-node activity, and additional simulations showed this could also account for why congruent RTs decreased asymptotically in the mixed but not the non-mixed scenario. Evident from the text: all evidence is simulation against previously published empirical effects, not new empirical data.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.cogpsych.2019.101224

Cites

80 of 80 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

14 of 77 cited sources checked · 10 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Imitative tendencies must be inhibited to prevent overt imitation.

    “More specifically, imitative tendencies have to be inhibited to prevent overt imitation (Brass, Derrfuss, Cramon, Matthes-von Cramon, & von Cramon, 2003;De Renzi, Cavalleri, & Facchini, 1996;Lhermitte, Pillon, & Serdaru, 1986).”
    Quote in the paperSubstantial claim2 of 3 sources checked
    • cites Brass, Derrfuss, Cramon, Matthes-von Cramon, & von Cramon, 2003 source
    • cites De Renzi, Cavalleri, & Facchini, 1996 source
    • cites Lhermitte, Pillon, & Serdaru, 1986 source
  • Interpretationreasoning

    Balancing facilitation and interference is needed for the model to produce the asymptote.

    “Together, these findings highlight the need to balance facilitation and interference (Botvinick et al., 2001;Bugg & Crump, 2012).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Botvinick et al., 2001 source
    • cites Bugg & Crump, 2012 source

      The Computational Accounts section discusses models of conflict and interference, while the item-specific proportion congruent section mentions facilitation only in a comparison with interference. Neither passage says that a model must balance the two to produce an asymptote. Searches for “asymptote” and “facilitation” found no passage establishing that requirement; page numbers were unavailable in our copy. (LLM, conf=0.74)

      Experimental — accuracy not yet measured
  • Interpretationreasoning

    The model is robust under realistic circumstances.

    “Thus, in line with Simulation 2, these findings suggests that the model is robust under realistic circumstances (Brass et al., 2000;Cracco et al., 2015;Genschow et al., 2017).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Brass et al., 2000 source
    • cites Cracco et al., 2015 source
    • cites Genschow et al., 2017 source

      The Results report automatic-imitation effects, and the Introduction discusses how imitation tasks differ in ecological validity. Neither passage evaluates a model or says it is robust under realistic circumstances (searched: the claim’s wording and model/simulation/realistic-circumstances terms). (LLM, conf=0.76)

      Experimental — accuracy not yet measured
  • Interpretationcitation

    As external input increases, inhibition has to increase accordingly, visible as larger inhibitory rebounds following stimulus offset.

    “As the external input increases, inhibition has to increase accordingly, and this is visible in the form of larger inhibitory rebounds following stimulus offset (Houghton et al., 1996;Schuch et al., 2010).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Houghton et al., 1996 source
    • cites Schuch et al., 2010 source
  • Interpretationcitation

    Various forms of social contagion may share the same functional mechanism.

    “This suggests, in other words, that various forms of social contagion may share the same functional mechanism (Raafat, Chater, & Frith, 2009).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Raafat, Chater, & Frith, 2009 source
  • May combine two claims. 26
  • Classified as probably not a substantial claim. 17
  • The quote only partly states this claim. 1

  • May combine two claims.

    Premisecitation

    Automatic imitation comprises both a facilitation effect and an interference effect.

    “Indeed, it is well-established in the literature that automatic imitation comprises both a facilitation and an interference effect (Brass et al., 2000;Cracco et al., 2015;Genschow et al., 2017).”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Brass et al., 2000 source
    • cites Cracco et al., 2015 source
    • cites Genschow et al., 2017 source

  • May combine two claims.

    Interpretationcitation

    Adding lateral inhibition between response nodes causes activation in one response node to inhibit the other, slowing response selection on incongruent trials and producing an interference effect.

    “By adding lateral inhibition to the model, activation in one response node is inhibited by activation in the other response node.This then slows down response selection on incongruent trials, resulting in an interference effect (Cooper, Catmur, et al., 2013).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Cooper, Catmur, et al., 2013 source

      In “Specific Aspects of the Model” (page unavailable), the paper explains that lateral inhibition makes more active nodes suppress other nodes, but explicitly says it was not included in the current simulations. In “Additional information” (page unavailable), it suggests that interference on incompatible trials would seem to require lateral inhibition between response nodes. It does not establish that adding it to this model slowed response selection and produced an interference effect. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Automatic imitation and spatial compatibility can be modeled by the same mechanism, and their time-course differences can be explained by different processing times for the two stimulus dimensions.

    “Moreover, they showed that automatic imitation -also known as imitative compatibility -and spatial compatibility can be modeled using the same mechanism, and that the observed differences in their time course (Catmur & Heyes, 2011) can be explained by variations in the processing time of both stimulus dimensions (Cooper, Catmur, et al., 2013).”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites Catmur & Heyes, 2011 source
    • cites Cooper, Catmur, et al., 2013 source

  • May combine two claims.

    Premisecitation

    Seeing two hands performing the same action produces stronger corticospinal excitability and stronger imitation than seeing one hand performing a single action.

    “research looking at automatic imitation of multiple agents has shown that seeing two hands performing the same action elicits stronger corticospinal excitability (Cracco, De Coster, Andres, & Brass, 2016) and therefore stronger imitation (Cracco & Brass, 2018a;Cracco, De Coster, Andres, & Brass, 2015) than seeing one hand performing a single action.”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Cracco, De Coster, Andres, & Brass, 2016 source
    • cites Cracco & Brass, 2018a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco, De Coster, Andres, & Brass, 2015 source

  • Classified as probably not a substantial claim.

    Premisecitation

    Connectionist modeling studies have shown that the dual-route model accurately describes empirical automatic imitation effects.

    “These studies revealed that the dual-route model provides an accurate description of the empirical effects (Cooper, Catmur, et al., 2013).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Cooper, Catmur, et al., 2013 source

  • Classified as probably not a substantial claim.

    Premisecitation

    Response inhibition is sensitive to variations in task context.

    “It was argued that response inhibition is sensitive to variations in task context (Bugg & Crump, 2012)”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Bugg & Crump, 2012 source

      The ‘Context-specific proportion congruent manipulations’ section reports that Stroop interference varied by context and describes rapid, context-triggered control adjustments (page unavailable). The review frames these adjustments as attentional filtering, rather than establishing that response inhibition itself is sensitive to task context. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Interpretationcitation

    Because balance between facilitation and interference is well-grounded in the literature, the model is robust under realistic circumstances.

    “Given that such a balance is well-grounded in the literature (Brass et al., 2000;Cracco et al., 2015;Genschow et al., 2017), this indicates that the model is robust under realistic circumstances.”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Brass et al., 2000 source
    • cites Cracco et al., 2015 source
    • cites Genschow et al., 2017 source

  • The quote only partly states this claim.

    Findingown data

    The model reproduced the empirical asymptote as long as the relevant parameters, including w_rf, were balanced.

    “The results of the first analysis showed that the model reproduced the empirical asymptote as long as and w rf were balanced”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In Simulation 1, the empirical effect was reproduced across almost the entire parameter space, demonstrating robustness.

    “As can be seen in Fig. 4, this revealed that the empirical effect was reproduced across almost the entire parameter space, demonstrating robustness.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Different sets of neurons code different hands according to their position.
    • The authors state this as an assumption: Conflict is restricted to the response level.
    • The authors state this as an assumption: Each individual stimulus hand provides a more salient visual trigger when there are fewer hands in total.
    • The authors state this as an assumption: Observed actions are processed more slowly than letters.
    • The authors state this as an assumption: Basic visual features such as motion are processed before complex features such as object or action identity.
    • The authors state this as an assumption: Performance in stimulus-response compatibility tasks is optimized by balancing exerted control so that interference on incongruent trials is minimized without eliminating facilitation on congruent trials.

  • May combine two claims.

    Findingown data

    The model closely captured the empirical effect, with smaller congruency effects in the 3H/1H DIFF condition than in the 3H ID and 4H ID conditions.

    “Supporting this hypothesis, Fig. 15 shows that the empirical effect was closely captured by the model, with smaller congruency effects in the 3H/1H DIFF condition than in the 3H ID and 4H ID conditions.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Simulation 4 closely reproduces both the empirical results and the results of Simulation 3.

    “As shown in Figs. 9 and 10, Simulation 4 closely reproduces both the empirical results and the results of Simulation 3.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    In Simulation 4, control-to-response weights became gradually more negative until equilibrium in the mixed scenario, but fluctuated around zero in the non-mixed scenario.

    “Therefore, in the mixed scenario, the control-to-response weights become gradually more negative until they reach a state of equilibrium.In contrast, in the non-mixed scenario, the weights fluctuate around zero (Fig. 14).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In Simulation 1, the model captures the empirical results: responses on congruent trials are faster and responses on incongruent trials are slower in the 2H ID condition than in the 1H condition.

    “As shown in Fig. 3, the model captures the empirical results, with responses on congruent trials being faster and responses on incongruent trials being slower in the 2H ID condition than in the 1H condition.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Different sets of neurons code different hands according to their position.
    • The authors state this as an assumption: Conflict is restricted to the response level.
    • The authors state this as an assumption: Each individual stimulus hand provides a more salient visual trigger when there are fewer hands in total.
    • The authors state this as an assumption: Observed actions are processed more slowly than letters.
    • The authors state this as an assumption: Basic visual features such as motion are processed before complex features such as object or action identity.
    • The authors state this as an assumption: Performance in stimulus-response compatibility tasks is optimized by balancing exerted control so that interference on incongruent trials is minimized without eliminating facilitation on congruent trials.

  • May combine two claims.

    Findingown data

    Across four simulations, the model reproduced that automatic imitation becomes stronger when seeing multiple identical actions but weaker when seeing multiple different actions.

    “Across four simulations, the model reproduced the finding that automatic imitation becomes stronger when seeing multiple identical actions (Cracco & Brass, 2018a, 2018d;Cracco et al., 2015) but weaker when seeing multiple different actions (Cracco & Brass, 2018b;Cracco et al., 2015).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Cracco & Brass, 2018a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco et al., 2015 source

  • May combine two claims.

    Findingown data

    Incongruent reaction times increased linearly across the entire parameter space, except for a small region where lateral and visual flux inhibition were both low, causing the interference effect to be absent.

    “The results of the second analysis indicated that incongruent RTs increased linearly across the entire parameter space, except for a small region where lateral and visual flux inhibition were both low, causing the interference effect to be absent (Fig. 12).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In Simulation 3 with the flux-to-response weights set to zero, reaction times no longer reached an asymptote but decreased monotonically with the number of moving hands, mirroring the empirical results.

    “Mirroring the empirical results, this caused RTs to no longer reach an asymptote but to instead decrease monotonically with the number of moving hands (Fig. 10).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Setting n to 8 or 32 also yields the same results.

    “However, the same results can also be obtained when n is set to, for example, 8 or 32.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    The cue activates short-term intentional stimulus-response associations formed for the task, while the observed movement activates long-term automatic stimulus-response associations formed through genetic disposition or learning.

    “In particular, the cue is thought to activate short-term intentional stimulus-response (S-R) associations established solely for the purpose of the task, whereas the stimulus movement instead activates long-term automatic S-R associations established as a consequence of genetic disposition or learning experience (Heyes, 2011).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Heyes, 2011 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Observing a congruent action speeds up an unrelated task response, whereas observing an incongruent action slows it down.

    “In cognitive psychology, it is studied by means of automatic imitation -the finding that observing congruent actions facilitates, whereas observing incongruent actions impedes, unrelated task responses (Brass et al., 2000;Cracco, Bardi, et al., 2018;Heyes, 2011;Sturmer, Aschersleben, & Prinz, 2000).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Brass et al., 2000 source
    • cites Cracco, Bardi, et al., 2018 source
    • cites Heyes, 2011 source

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationcitation

    When both routes activate the correct response, motor activation rises and facilitation occurs; when the automatic route activates the incorrect response, that response is inhibited and interference occurs.

    “When both routes converge on the correct response, motor activation is increased, and response facilitation occurs.In contrast, when the automatic route activates the incorrect response instead, this response has to be inhibited, and response interference occurs.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Computational research supports the idea that automatic imitation arises from associative learning.

    “Finally, by comparing different mechanisms, computational research has supported the idea that automatic imitation is a consequence of associative learning (Cooper, Cook, et al., 2013).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Cooper, Cook, et al., 2013 source

  • Classified as probably not a substantial claim.

    Premisecitation

    The congruency effect serves as an index of imitative response tendencies.

    “This congruency effect is a measure of imitative response tendencies (Cracco & Brass, 2019), and has now been reliably observed across a wide range of effectors”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2019 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The dual-route framework treats the imperative cue and the observed movement as being processed along separate routes.

    “An influential framework to interpret automatic imitation is the dual-route framework (Heyes, 2011), which assumes that the imperative cue and the stimulus movement operate via different routes.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Heyes, 2011 source

  • May combine two claims.

    Interpretationreasoning

    This threshold adjustment counteracts the decrease in response times on congruent trials but strengthens the increase on incongruent trials, thereby explaining the asymmetrical pattern.

    “This slowdown counteracts the decrease in response times on congruent trials but strengthens the corresponding increase on incongruent trials, hence explaining the asymmetrical pattern.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Existing models of automatic imitation can accommodate only one agent.

    “However, current models can accommodate just one agent.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Seeing one congruent and one incongruent action produces concurrent facilitation and interference effects that cancel each other out.

    “It was argued that seeing one congruent and one incongruent action produced a concurrent facilitation and interference effect that subsequently canceled out each other (Cracco et al., 2015).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco et al., 2015 source

  • May combine two claims.

    Premisecitation

    The disappearance of the congruent asymptote in the absence of incongruent trials is accompanied by overall faster responses, which suggests it is not due to a physical boundary but to dynamic control processes regulating motor inhibition based on the number of observed actions.

    “This change was accompanied by overall faster responses, suggesting it was not caused by a physical boundary but was instead better explained by dynamic control processes regulating motor inhibition based on the number of observed actions (Cracco & Brass, 2018d).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    A visual flux node codes the amount of visual change and inhibits response nodes; because more moving hands produce more visual change, this mechanism regulates response inhibition according to the number of moving hands and implements the dynamic inhibition process proposed by Cracco and Brass (2018d).

    “Second, we added a visual flux node that codes the amount of visual change and then uses this information to inhibit the response nodes.Because more moving hands equals more visual change, this mechanism regulates response inhibition based on the number of moving hands and as such implements the dynamic inhibition process put forward by Cracco and Brass (2018d).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationcitation

    Setting the flux-to-response weights to zero made reaction times faster, indicating that the empirical asymptote could not have been caused by physical restraints preventing participants from speeding up.

    “Importantly, in line with the empirical data, RTs also became faster when the flux-to-response weights were set to zero, indicating that the asymptote could not have been caused by physical restraints preventing participants from speeding up (Cracco & Brass, 2018d).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    Identical observed actions activate the same motor representation.

    “More specifically, it does so by implementing the hypothesis that identical actions activate the same motor representation (Cracco & Brass, 2018d, 2018a;Cracco et al., 2015Cracco et al., , 2016)).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco & Brass, 2018a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco et al., 2015 source

  • May combine two claims.

    Interpretationcitation

    A conflict adaptation mechanism might explain why response speed on congruent trials decreases asymptotically when the proportion of congruent to incongruent trials is balanced, but monotonically when congruent trials are presented in isolation.

    “Simulation 4 indicates that a similar mechanism might also explain why response speed on congruent trials decreases asymptotically when the proportion of congruent to incongruent trials was balanced, but monotonically when congruent trials were presented in isolation (Cracco & Brass, 2018d).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    In the mixed scenario, participants experience response conflict, which causes them to inhibit motor activation based on the number of moving hands; in the non-mixed scenario, they do not experience response conflict and no inhibitory control is exerted.

    “That is, in the mixed scenario, participants experience response conflict, which causes them to inhibit motor activation based on the number of moving hands.In contrast, in the non-mixed scenarios, participants do not experience response conflict, and no inhibitory control is exerted.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Imitative responses were no longer inhibited when they facilitated task performance.

    “and that imitative responses were therefore no longer inhibited when they facilitated task performance (Cracco & Brass, 2018d).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    In the model, motor inhibition from the visual flux node increases with the number of moving hands, causing a reaction-time increase that eliminates the expected decrease on congruent trials and strengthens the analogous increase on incongruent trials.

    “This is because motor inhibition from the visual flux node increases with the number of moving hands, leading to a corresponding RT increase that eliminates the expected RT decrease on congruent trials but strengthens the analogous RT increase on incongruent trials.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing).

    Interpretationcitation

    Based on the current study, larger groups may provide a stronger trigger to the motor system, leading to stronger imitative responses.

    “However, based on the current study, an alternative hypothesis might also be that larger groups provide a stronger trigger to the motor system, leading to stronger imitative responses (Cracco & Brass, 2018a, 2018d;Cracco et al., 2015Cracco et al., , 2016)).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Cracco & Brass, 2018a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco & Brass, 2018d

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Cracco et al., 2015 source

  • May combine two claims.

    Study design and conductreasoning

    The model focused on modeling mean performance and did not model performance variability.

    “A first limitation is that, following most previous computational work, we focused on modeling mean performance and did not model performance variability.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The model has a relatively large number of free parameters.

    “A second limitation of the model is that it has a relatively large number of free parameters.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Visual processing is organized hierarchically, so simple features like motion are processed earlier than complex features like the identity of objects or actions.

    “This seems unlikely, considering that visual processing is hierarchical with basic information like motion being processed before complex information like the identity of objects or actions (Van Essen & Maunsell, 1983).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Van Essen & Maunsell, 1983 source

  • May combine two claims.

    Interpretationreasoning

    Response conflict on incongruent trials increases motor inhibition, but when there are no incongruent trials, there is no response conflict and therefore little inhibition.

    “That is, response conflict on incongruent trials increases motor inhibition, but when there are no incongruent trials, there is no response conflict, and therefore little inhibition.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductreasoning

    Parameter values were not varied across participants, so no interindividual differences were included.

    “More specifically, even though noise was added to the input on each trial, we did not vary the parameter values across participants.This means that no interindividual differences were included, in contrast to real life, where such differences are the norm.”
    Quote in the paperSubstantial claim
40 of 40

  • The authors state this as an assumption.

    Different sets of neurons code different hands according to their position.

    “This formalizes the assumption that different sets of neurons code the different hands depending on their position.”

  • The authors state this as an assumption.

    Conflict is restricted to the response level.

    “Importantly, no lateral inhibition was included at the stimulus level because we assumed conflict to be restricted to the response level, where eventually a single response has to be produced.”

  • The authors state this as an assumption.

    Each individual stimulus hand provides a more salient visual trigger when there are fewer hands in total.

    “This, in turn, reflects the assumption that each individual stimulus hand provides a more salient visual trigger when there are fewer hands in total, consistent with the fact that the congruency effect produced by a single hand was stronger in the experiments with two hands (Cracco et al., 2015) than in the experiments with four hands (Cracco & Brass, 2018d).”

  • The authors state this as an assumption.

    Observed actions are processed more slowly than letters.

    “In line with Cooper et al. (2013), the h parameter was set to +80 cycles to implement the assumption that observed actions are processed more slowly than letters (Catmur & Heyes, 2011).”

  • The authors state this as an assumption.

    Basic visual features such as motion are processed before complex features such as object or action identity.

    “Similarly, the f parameter was set to -20 cycles to reflect the hierarchical nature of visual processing, with basic features like motion being processed before complex features like the identity of objects or actions (Van Essen & Maunsell, 1983).”

  • The authors state this as an assumption.

    Performance in stimulus-response compatibility tasks is optimized by balancing exerted control so that interference on incongruent trials is minimized without eliminating facilitation on congruent trials.

    “More generally, this reflects the assumption that performance in S-R compatibility tasks is optimized by balancing the amount of exerted control so that interference on incongruent trials is minimized without eliminating facilitation on congruent trials (see Botvinick et al., 2001 for a similar logic and simulations).”

  • The authors state this as an assumption.

    Setting the flux-to-response weights by hand assumes that inhibitory control is fine-tuned based on task instructions or practice trials.

    “Importantly, setting the flux-to-response weights by hand assumes that inhibitory control is fine-tuned based on task instructions or practice trials.”

  • The authors state this as an assumption.

    Motor inhibition is implemented not by visual but by prefrontal brain areas.

    “First, we added an intermediate control node, formalizing the assumption that motor inhibition is implemented not by visual but by prefrontal brain areas (Munakata et al., 2011).”

  • The authors state this as an assumption.

    Different experiments require a different response threshold or, equivalently, different levels of tonic inhibition.

    “In more general terms, the adjustment of r reflects the assumption that different experiments require a different response threshold or, equivalently, different levels of tonic inhibition.”
9 of 9

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Simulation 5 empirical data is attributed to two different papers

    Minor

    The text says Simulation 5 models a finding of Cracco and Brass (2018b), but the caption attributes its empirical data to Experiment 1 of Cracco et al. (2018c). The references list these as different papers.

  • Conflict-driven weight updates have the opposite stated effect

    Minor

    The update rule says higher-than-usual response conflict makes the control-to-response weight more positive. The discussion instead says conflict on incongruent trials increases motor inhibition, which the model implements through more negative weights.

2 of 2

Cites

Flagged first, then by citations

5 of the 85 references couldn’t be matched to a known paper.

Cited by

Sorted by citations

Related

Authors
Cracco, E., & Cooper, R. P.
Affiliations
  • Ghent University (BE)
  • University of London (GB)
Corresponding author
Cracco, Emiel
ORCID
Venue
Cognitive Psychology
Publisher
Elsevier BV
Published
2019-09
Volume
113
Pages
101224
ISSN
0010-0285 (ISSN-L)
DOI
10.1016/j.cogpsych.2019.101224
OpenAlex
W2950568652
Semantic Scholar
062a885e550f38c8253ed1c63110f3c5010ebe7c
Cited by
23
References
85
Citations by year
2026: 1 · 2025: 1 · 2024: 5 · 2023: 3 · 2022: 3 · 2021: 7 · 2020: 2
FWCI
3.92
Type
journal article
Language
en
Open access
Licence
Locations
  • Cognitive Psychology (publishedVersion)
  • PubMed (publishedVersion)
  • BIROn (Birkbeck, University of London) (acceptedVersion)
  • Ghent University Academic Bibliography (Ghent University) (submittedVersion)
Funders
Fonds Wetenschappelijk Onderzoek (FWO16/ASP_H/050)
Topics
  • Social Sciences › Psychology › Social Psychology › Action Observation and Synchronization
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
Keywords
automatic imitation, multiple agents, dual-route model, response inhibition, response conflict, action observation, motor system, social interaction
MeSH terms
Computer Simulation, Humans, Imitative Behavior, Movement, Reaction Time

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-08.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-08.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review