In effect, her account of the infant data rests on the interplay between two kinds of domain-general cognitive processes. On the one hand, Heyes\u2019s low-level perceptual novelty account commits her to the assumption that infants can only encode a highly restricted<\/i> subset of the full set of properties (the low-level<\/i> properties) that are instantiated by both the test events and the earlier events perceived by the infants. On the other hand, Heyes appeals to the phenomenon of so-called retroactive interference<\/i>, whereby the infants\u2019 perception of some of the specific later<\/i> events impairs<\/i> their memory<\/i> of some of the immediately<\/i> preceding<\/i> events in the sequence<\/span>.[3]\n As Heyes (2014a, p. 647) rightly says about the infant data, \u201cthe devil is in the details.\u201d The question is: can the details of her own account perspicuously explain the relevant data? In what follows, I will proceed in two steps: I will first review her account of findings based on Onishi and Baillargeon\u2019s (2005) study and ask whether her account is consistent. Secondly, I will argue that findings based on a study by Kov\u00e1cs, T\u00e9gl\u00e1s and Endress (2010) fail to support one crucial component of her account (retroactive interference).\u00a0<\/span><\/p>\n 1. The water-melon, the green box and the yellow box<\/strong><\/p>\n Onishi and Baillargeon\u2019s study involves three temporal stages: a set of familiarization trials, a set of belief-induction trials, and a pair of test trials. In the first of three successive familiarization trials, 15-month-olds first saw a melon toy surrounded by a pair of opaque boxes (a green and a yellow box) against the background of a closed window. After the window was opened, the infants saw a human (female) agent manipulate the toy before placing it into the green box with her right hand. In the two following familiarization trials, they first saw the two opaque boxes against the background of the closed window. Then the window was opened and they saw the human (female) agent reach into the green box with her right hand (cf. Figure 1).<\/span><\/p>\n <\/a><\/p>\n In a series of four distinct belief-induction trials generating a pair of true-belief (TB) conditions and a pair of false-belief (FB) conditions, the infants saw the toy either move (by its own self-propelled motion) from the green to the yellow box or not, either in the presence of the agent or not (cf. Figure 2). In the TB-green condition, the yellow box moved towards the green box and back to its initial position (as indicated by the arrow in Figure 2a), but the toy did not move and the agent was present. In the TB-yellow condition, the toy moved from the green to the yellow box in the agent\u2019s presence (as indicated by the arrow in Figure 2b). In the FB-green condition, the toy moved to the yellow box in the agent\u2019s absence (cf. Figure 2c). In the FB-yellow condition, the toy first moved from the green to the yellow box in the agent\u2019s presence and moved back into the green box in the agent\u2019s absence (cf. Figure 2d)<\/p>\n Finally, in the test trials, the infants first saw the two opaque boxes at their previous location against the background of the closed window. The window opened and they saw the agent reach either for the green box with her right hand (green test event) or for the yellow box with her left hand (yellow test event) (cf. Figure 3).<\/p>\n Onishi and Baillargeon found that infants in the TB-green condition looked reliably longer at the yellow rather than at the green test event. Infants in the TB-yellow condition looked reliably longer at the green rather than at the yellow test event. Infants in the FB-green condition looked reliably longer at the yellow rather than at the green test event. Infants in the FB-yellow condition looked reliably longer at the green rather than at the yellow test event. According to Onishi and Baillargeon\u2019s mentalistic interpretation, infants looked reliably longer either when the agent reached to the empty location with a true rather than a false belief or when she reached to the toy\u2019s actual location with a false rather than a true belief.<\/p>\n Heyes first asks: why did infants in both the TB-green and the FB-green conditions look reliably longer at the yellow than at the green test event? On the mentalistic account, the answer to this question is: because infants expected the agent to reach for the box in which she believed (truly or falsely) the toy to be. Of course, this cannot be Heyes\u2019s non-mentalistic answer.<\/p>\n Heyes\u2019s own answer is two-tiered. On the one hand, she argues that to infants in the TB-green condition (in which the toy stayed in the green box), the yellow test event must have looked perceptually more novel than the green test event, relative to their familiarization<\/i> experience (in which on three repeated occasions they saw the agent reach for the green box). On the other hand, she must explain away the fact that infants in the FB-green condition saw the toy move to the yellow box (in the agent\u2019s absence). The way Heyes cancels the effect of infants\u2019 seeing the toy move to the yellow box in the FB-green condition is by appealing to the phenomenon called retroactive interference<\/i>. In the FB-green condition, the infants see the toy move to the yellow box, in the agent\u2019s absence<\/i>. Heyes argues that the agent\u2019s unexpected reappearance in test event retroactively interferes with infants\u2019 memory of the immediately preceding event in the FB-green condition: \u201ctheir memory for this event was impaired because it was immediately followed by a salient distractor event \u2013 the unexpected reappearance of the agent at the beginning of the test phase.\u201d In effect, she argues that retroactive interference cancels the difference between the TB-green and the FB-green condition. So in both cases, the yellow test event must have looked to infants perceptually more novel than the green test event, relative to their familiarization experience. <\/span><\/p>\n Secondly, Heyes asks: why did infants in the TB-yellow and the FB-yellow conditions look reliably longer at the green than at the yellow test event? On the mentalistic account, the answer to this question is: because infants expected the agent to reach for the box in which she either truly or falsely believed the toy to be. <\/span><\/p>\n Here again, Heyes\u2019s answer based on her low-level perceptual novelty account is two-tiered. On the one hand, she argues that in the TB-yellow condition, \u201cafter familiarization and before the test, these infants saw an event (movement of the toy-shape towards yellow), that was visually similar to the yellow test event (movement of the agent-shape towards yellow), and therefore reduced the novelty of the yellow test event.\u201d As a result, these infants looked longer at the green than at the yellow test event. On the other hand, Heyes must also explain away the fact that in the FB-yellow belief-induction trial, after seeing the toy move to the yellow box in the agent\u2019s presence, the infants saw the toy move back to the green box in the agent\u2019s absence. Here again, Heyes appeals to retroactive interference: she hypothesizes that the unexpected reappearance of the agent at the beginning of the test phase impairs infants\u2019 memory of the immediately preceding event whereby the toy moved back to the green box in the agent\u2019s absence. As a result, the difference between the TB-yellow and the FB-yellow conditions vanishes: in both cases, what matters is that after familiarization and before test, the infants saw \u201can event (movement of the toy-shape towards yellow), that was visually similar to the yellow test event (movement of the agent-shape towards yellow), and therefore reduced the novelty of the yellow test event.\u201d <\/span><\/p>\n I now want to call into question the internal consistency of Heyes\u2019s perceptual novelty account of Onishi and Baillargeon\u2019s findings. On the one hand, her low-level perceptual novelty account can only be satisfied if the novelty generating the surprise is present at a low representational level \u201cwhere the events witnessed by the infants are represented as colors, shapes and movements, rather than as actions on objects by agents.\u201d For example, it is a critical assumption of the low-level perceptual novelty account that to infants the movement of the toy<\/i> to the yellow<\/i> box (in the TB-yellow belief-induction trial) is visually similar<\/i> to the yellow test event whereby the agent<\/i> reaches for the yellow<\/i> box. In the TB-yellow belief induction trial, infants see the toy move to the yellow box; they see the agent watch the movement of the toy; but they only see the agent\u2019 head, not her full upper body parts. In the yellow test event, however, they see the agent reach for the visible yellow box with her fully visible left arm, but they do not see the toy. So it is crucial to the low-level perceptual novelty account that infants be blind to the many differences between the event of the toy moving to the yellow box in the TB-yellow belief-induction trial and the yellow test event of the agent\u2019s reaching for the yellow box. <\/span><\/p>\n But on the other hand, Heyes appeals to retroactive interference in both the FB-green and the FB-yellow conditions. So infants must be vulnerable to the relevant instance of retroactive interference whereby the agent<\/i>\u2019s unexpected reappearance in the test event impairs their memory of the immediately preceding event in which the agent was absent. Only if the infants can encode the property of being an agent<\/i> (capable of executing e.g. reaching arm movements) could they be vulnerable to relevant instances of retroactive interference. But according to the low-level perceptual novelty account, infants should not be able encode the property of being an agent. (Remember: by Heyes\u2019s own criteria, even behavior-reading<\/i> rests on domain-specific cognitive mechanisms that are unavailable to human infants who are limited to domain-general cognitive resources.) So there is a conflict between the two components of her non-mentalistic explanation of the findings reported by Onishi and Baillargeon (2005).<\/span><\/p>\n 2. The Blue Smurf and the ball<\/strong><\/p>\n Heyes (2014a, 2014b) has also tried to explain away the famous Blue Smurf studies by Kov\u00e1cs, T\u00e9gl\u00e1s and Endress (2010). In these studies with adults and infants, participants watch one of four versions of a video. In all four conditions, an agent (a blue Smurf) places a ball on a table in front of an occluder and then the ball rolls behind the occluder. From then on, the four different conditions diverge: participants see the ball either stay behind the occluder or leave and they see the agent leave the scene either before or after the ball reaches its final location. Finally, in all four conditions, the agent comes back, the occluder is lowered in his presence (cf. Figure 4).<\/span><\/p>\n In the adult study, participants were instructed to press a button as fast as possible in the final stage when the agent is back, if they saw the ball when the occluder was lowered down (which was the case 50% of the time in all four conditions). Kov\u00e1cs and colleagues measured participants\u2019 reaction times in performing this task. Not surprisingly, they found that adults were significantly faster to respond in the P+A+ condition (when both participants P and the Smurf agent A expected the ball to be behind the occluder) than in the P-A- condition (when neither participants P nor the agent A expected the ball to be behind the occluder). The surprising finding was that they were also significantly faster in the P-A+ condition (when A falsely expected the ball to be there, but participants did not) than in the P-A- condition. Kov\u00e1cs and colleagues argue that this surprising finding is evidence that participants automatically computed the content of the agent\u2019s false belief.[4]\n Heyes (2014b, p. 137) has offered a non-mentalistic alternative explanation of the adults\u2019 findings based on retroactive interference. In both the P-A- and the P+A+ conditions, the Smurf is present during the last event that is relevant to the participants\u2019 own expectations about the location of the ball. However, in the P-A+ condition, the Smurf is absent during the last event that is relevant to the participants\u2019 own expectations about the location of the ball, i.e. when the ball finally leaves the scene. Heyes argues that the \u201cperceptually salient\u201d reappearance of the Blue Smurf in the final stage of condition P-A+ (when the occluder is being lowered) is likely to have impaired participants\u2019 memory of the immediately preceding event (i.e. the last motion of the ball) by a process of retroactive interference. Heyes\u2019s non-mentalistic account makes a straightforward prediction: in the P+A- condition, when participants see the ball finally move back behind the occluder, the Blue Smurf is also absent. According to Heyes\u2019s account, the salient reappearance of the Blue Smurf when the occluder is lowered should also impair participants\u2019 memory for the immediately preceding event where the ball returns behind the occluder. So Heyes\u2019s account predicts that participants should not expect the ball to be behind the occluder and therefore be significantly slower in the P+A- condition than in the P+A+ condition. But they are not.<\/span><\/p>\n The same critique extends to Heyes\u2019s (2014a) account of the infant study, in which Kov\u00e1cs and colleagues used infants\u2019 looking time and found that when in the final stage, the Smurf is back, the occluder is lowered down and there is no ball<\/i>, 7-month-olds look longer in the P-A+ than in the P-A- condition. Kov\u00e1cs and colleagues argue that this is evidence that infants\u2019 looking time is influenced by their computation of the Smurf\u2019s false expectation that the ball should be there. Heyes on the other hand surmises that in the P-A+ condition, the Smurf was away when the infants last saw the ball leave the scene. So she hypothesizes that far from reflecting infants\u2019 computation of the Smurf\u2019s false belief, the Smurf\u2019s unexpected reappearance impaired the infants\u2019 memory of this last event by retroactive interference. As a result, infants looked longer in the P-A+ condition than in the P-A- condition because they had forgotten the ball\u2019s last motion and expected the ball to be there. <\/span><\/p>\n This account predicts conversely that in the P+A- condition, infants should also forget the last event whereby the ball rolled back behind the occluder. If so, then they should not<\/i> expect the ball to be there and should not<\/i> look longer in the P+A- condition than in the P-A- condition upon finding out that there is no ball. So far as I know (but I may be wrong), this comparison has not been tested. But it would be quite interesting to test Heyes\u2019s prediction that infants should not look longer in the P+A- than in the P-A- condition. On the assumption that the mechanisms that underlie the adults\u2019 responses are the signature of the mechanisms that underlie the infants\u2019 responses <\/span>[5],\u00a0I would expect the infants\u2019 looking time to be consistent with the adults\u2019 reaction times. Since adults were faster to detect the ball when it was there in the P+A- than in P-A- condition, I would conversely expect the infants to look longer when there is no ball in the P+A- condition than in the P-A- condition.\u00a0<\/span><\/p>\n In this short piece, I have looked at some of the details of Cecilia Heyes\u2019s non-mentalistic account of a pair of developmental findings that have been taken by some psychologists as evidence for false-belief understanding in human infancy. I have proceeded in two steps: first, I have called into question the consistency of her account. Secondly, I have drawn attention to evidence that putatively clashes with her appeal to the phenomenon of retroactive interference.<\/span><\/p>\n What is distinctive of Heyes\u2019s domain-general approach to human social cognition is that it refreshingly purports to break away from the common assumption that the major alternative to the hypothesis that human infants are endowed with implicit mindreading capacities is that they are endowed with behavior-reading capacities. She explores instead the hypothesis that human infants are genetically prepared to submentalize, i.e. to apply to social stimuli domain-general low-level perceptual processes that simulate the effects of mentalizing in social settings. If submentalizing were part of the genetic tool-kit of human social cognition, then findings based on implicit false-belief tasks in human infants should be explainable by Heyes\u2019s low-level perceptual novelty approach. Conversely, failure of the low-level perceptual novelty account to perspicuously explain the infant findings casts doubt on Heyes\u2019s hypothesis that humans are genetically prepared to submentalize in the face of social stimuli.<\/span>[6]\n References<\/strong><\/p>\n Heyes, C. (2012) Grist and mills: on the cultural origins of cultural learning. Philosophical Transactions of the Royal Society <\/em>B2012, 367, 2181-2191.<\/p>\n Heyes, C. (2014a) False belief in infancy: a fresh look. Developmental Science<\/em>, 17:5, 647\u2013659.<\/p>\n Heyes, C. (2014b) Submentalizing: I Am Not Really Reading Your Mind. Perspectives on Psychological Science<\/em>, 9(2) 131\u2013143.<\/p>\n Heyes, C. (2015) Animal mindreading: what\u2019s the problem? Psychonomic Bulletin & Review<\/em>, \u00a022:313\u2013327.<\/p>\n Heyes, C. (2017) Apes Submentalise. Trends in Cognitive Sciences<\/em>, 2017; 21(1):1-2.<\/p>\n Heyes, C. and Frith, C. (2014) The cultural evolution of mindreading. Science<\/em>, 344, 1357- 1361.<\/p>\n Kov\u00e1cs, A., T\u00e9gl\u00e1s, E. and Endress, A. (2010) The Social Sense: Susceptibility to Others\u2019 Beliefs in Human Infants and Adults.\u2028Science<\/em>, 330, 1830-1834.<\/p>\n Krupenye, C., Kano, F., Hirata, S., Call, J. & Tomasello, M. (2016) Great apes anticipate that other individuals will act according to false beliefs. Science<\/em>, 354 (6308):110-4.<\/p>\n Krupenye, C., Kano, F., Hirata, S., Call, J. & Tomasello, M. (2017) A test of the submentalizing hypothesis: Apes’ performance in a false belief task inanimate control. Communicative & Integrative Biology<\/em>, 10(4) Onishi, C. and Baillargeon, R. (2005) Do 15-month-old infants understand false beliefs? Science<\/em>, 308(5719), 255\u2013258.<\/p>\n Phillips, J., Ong, D.C., Surteees, A.D.R., Xin, Y., Williams, S., Saxes, R. & Franck, M.C. (2015) A Second Look at Automatic Theory of Mind: Reconsidering Kov\u00e1cs, T\u00e9gl\u00e1s, and Endress (2010). Psychological Science, <\/em>26(9), 1-15.<\/p>\n Samson, D., Apperly, I.A., Braithwaite, J.J., Andrews,\u2028B.J., Bodley Scott, S.E. (2010) Seeing it their way: Evidence for rapid and involuntary computation of what other people see. Journal of Experimental Psychological. <\/em>Human Perception and Performance<\/em>, 26(9), 1255\u20131266.<\/p>\n Santiesteban, I., Catmur, C., Coughlan Hopkins, S., Bird,\u2028G., Heyes, C. (2014) Avatars and arrows: Implicit mentalizing or domain- general processing? Journal of Experimental Psychological. <\/em>Human Perception and Performance<\/em>, 40, 929\u2013937.<\/p>\n Scott, R.M., & Baillargeon, R. (2014) How fresh a look? A reply to Heyes. Developmental Science<\/em>, 17:5, 660\u2013664.<\/p>\n Strickland, B. and Jacob, P. (2015) http:\/\/cognitionandculture.local\/blog\/brent-stricklands-blog\/why-reading-minds-is-not-like-reading-words<\/a><\/p>\n ***<\/p>\n[1] One alternative mentalistic possibility is that far from being submentalizers, human adults are super-mentalizers<\/i> who find it hard not to interpret their environment mentalistically. If so, then they might immediately interpret an arrow as a social cue intended by a conspecific to draw their attention to a particular dot.\u00a0<\/span><\/p>\n[2]\u00a0Krupenye et al. (2016) have reported findings, which they interpret as evidence that non-human apes can represent the contents of others\u2019 false beliefs. Heyes (2017) has proposed her submentalizing alternative account of the findings by Krupenye and colleagues (2016). Krupenye et al. (2017) have subsequently tested and offered evidence against Heyes\u2019s (2017) proposal. <\/span><\/p>\n <\/a><\/p>\n[3]\u00a0For an earlier critique of Heyes\u2019s (2014a) explanation of the infant data based on her perceptual novelty account, cf. Scott and Baillargeon (2014). <\/span><\/p>\n![\"\"](\"http:\/\/cognitionandculture.local\/wp-content\/uploads\/OnishiBaillargeon_1.png\")
<\/a>Figure 1 – From Onishi & Baillargeon (2005)<\/h6>\n![\"\"](\"http:\/\/cognitionandculture.local\/wp-content\/uploads\/OnishiBaillargeon_2-1.png\")
<\/a>Figure 2 – From Onishi & Baillargeon (2005)<\/h6>\n![\"\"](\"http:\/\/cognitionandculture.local\/wp-content\/uploads\/OnishiBaillargeon_3.png\")
<\/a>Figure 3 – From Onishi & Baillargeon (2005)<\/h6>\n![\"\"](\"http:\/\/cognitionandculture.local\/wp-content\/uploads\/kovacs.png\")
<\/a>Figure 4 – From Kov\u00e1cs, T\u00e9gl\u00e1s and Endress, (2010)<\/h6>\n
\n
\nhttp:\/\/dx.doi.org\/10.1080\/19420889.2017.1343771<\/p>\n