Re-analysis of data reveals no evidence for neonatal imitation in rhesus macaques

Abstract

Over the past decade, a growing number of publications have claimed to provide evidence for the existence and function of neonatal imitation in rhesus macaques. Here I show that there is in fact no empirical basis for these claims. Studies of the phenomenon have consistently failed to implement the gold standard cross-target analytical approach, which controls for increases in matching responses that may not be a function of the specific modelled behaviour. Critically, a pre-registered re-analysis of the entire set of existing data using this cross-target approach shows that macaque neonates have failed to produce matching tongue protrusion or lipsmacking responses at levels greater than chance. Furthermore, there is no evidence for intra-individual consistency in ‘imitative’ responses across different actions, as imitation scores for the two actions are negatively correlated with each other. Macaque tongue protrusion and lipsmacking responses may vary as a function of general factors that fluctuate over testing sessions, rather than as a function of the specific model or of between-individual variations in imitative tendencies.

Individual differences in neonatal “imitation” fail to predict early social cognitive behaviour

“…Indeed, if authors are categorizing infants as “imitators” primarily on the basis that they show greater increases in Action A in response to Modelled Action A than in response to Modelled Action B (see, e.g. studies with rhesus macaques by Simpson, Miller, Ferrari, Suomi, & Paukner, ; Simpson, Paukner, Sclafani, Suomi, & Ferrari, ; Simpson, Paukner, et al, ; Wooddell, Simpson, Murphy, Dettmer, & Paukner, ), then one would expect about 50% of infants to be “imitators” by chance alone (note also that these macaque studies do not report results using appropriate control models; Redshaw, ). We suggest that studies should not classify infants as consistent “imitators” unless it can be shown that (a) there is an overall effect of imitation in the sample, and (b) there is a positive correlation between imitation scores across time points and actions.…”

“…Under this view, one might expect individual differences in neonatal imitation to predict only very early forms of social cognition, such as the tendency to maintain mutual gaze with a caregiver (see Heimann, ), and not the forms we measured here. Of course, both of these alternative explanations presuppose that neonatal imitation is a genuine phenomenon, which recent evidence suggests is unlikely (Arcaro et al, ; Barbosa, ; Cracco et al, ; Keven & Akins, ; de Klerk et al, ; McKyton et al, ; Oostenbroek et al, ; Redshaw, ).…”

“…Indeed, if authors are categorizing infants as "imitators" primarily on the basis that they show greater increases in Action A in response to Modelled Action A than in response to Modelled Action B (see, e.g. studies with rhesus macaques by Simpson, Miller, Ferrari, Suomi, & Paukner, 2016;Simpson, Paukner, Sclafani, Suomi, & Ferrari, 2013;Simpson, Paukner, et al, 2014b;Wooddell, Simpson, Murphy, Dettmer, & Paukner, 2019), then one would expect about 50% of infants to be "imitators" by chance alone (note also that these macaque studies do not report results using appropriate control models; Redshaw, 2019).…”

“…We were compelled to conclude, therefore, that rather than providing the foundation for a definitive study of neonatal imitation's function, our data instead challenged the very existence of neonatal imitation itself. This unforeseen conclusion has since been substantiated by several other works: A comprehensive sensorimotor analysis suggesting that neonates lack voluntary, cortical control over tongue protrusion, thus making intentional imitation of this action implausible (Keven & Akins, ); A study suggesting that previously blind, newly sighted children show greatly impaired automatic imitation of manual actions, meaning that this capacity is either not innate or vulnerable to degradation in specific domains (McKyton, Ben‐Zion, & Zohary, ); A study showing that mothers’ tendency to imitate their 4‐month‐old infants’ actions predicted these infants’ own rudimentary signs of imitation (measured by EMG), suggesting that infants may learn to imitate via repeated associative pairings of their own and others’ actions (de Klerk, Lamy‐Yang, & Southgate, ); A systematic meta‐analysis suggesting that automatic imitation in adults is also best explained by an associative learning framework (Cracco et al, ), meaning that the existence of neonatal imitation would require the non‐parsimonious addition of a second, innate mechanism to explain the same phenomenon; A statistically robust re‐analysis of data previously used to support the existence of neonatal imitation in rhesus macaques (Paukner, Pederson, & Simpson, ), showing that neonatal macaques have in fact failed to produce imitative actions at levels greater than chance (Redshaw, ); A study suggesting that newborn primates are not born with an innate capacity to recognize faces, meaning they would be unable to link the actions of other faces with those of their own face (Arcaro, Schade, Vincent, Ponce, & Livingstone, ); and An unpublished PhD thesis that similarly found no evidence of neonatal imitation in a longitudinal study of 90 human infants measured at four time points (Barbosa, ). …”

“…A systematic meta-analysis suggesting that automatic imitation in adults is also best explained by an associative learning framework (Cracco et al, 2018), meaning that the existence of neonatal imitation would require the non-parsimonious addition of a second, innate mechanism to explain the same phenomenon; 5. A statistically robust re-analysis of data previously used to support the existence of neonatal imitation in rhesus macaques (Paukner, Pederson, & Simpson, 2017), showing that neonatal macaques have in fact failed to produce imitative actions at levels greater than chance (Redshaw, 2019); 6. A study suggesting that newborn primates are not born with an innate capacity to recognize faces, meaning they would be unable to link the actions of other faces with those of their own face (Arcaro, Schade, Vincent, Ponce, & Livingstone, 2017); and…”