Leahy & Carey (2020)
To me, what is very surprising about the minimal account is the claim that children this age, cannot mark possibilities as mere possibilities, and have to make a simulation about which outcome might be true.
One of the hardest implications to accept about this has to do with the fact that it is generally widely accepted in the field that even younger learners (i.e. infants) are Bayesian reasoners, and there is much support for this in the infant literature. But reasoning in Bayesian terms necessarily implies that one has to recognise that there is a space of possibilities - and those possibilities get assigned different prior (and posterior) likelihoods. This account however, argues that learners this age, when faced with multiple mutually exclusive possibilities, stick to one (randomly) unless and until proven wrong (i.e. ‘sequentially guessing/simulating).
Although it is not impossible to interpret a lot of infant literature that provides indirect evidence to the contrary in terms of the minimal strategy (I can elaborate on this in class), what seems very odd is the fact that they claim that reasoning this way (I.e. sequentially simulating, or minimally representing possibiliites) is still compatible with reasoning in Bayesian terms. How do others feel about this claim?
Stahl & Feigenson (2024)
First of all, it is not entirely clear to me why they had a different number of subjects in each condition (e.g. 64 in the 5% probability condition, 26 in the 10%, 35 in the 2.5%) and why they collected two separate samples for the Impossible condition. There was no explanation for this.
Although there was a significant difference between proportion of children who learnt the target label (i.e. blick) after watching an Impossble condition compared to a merely Improbable or Equi-probable one, I think that the title of the paper, and what the authors set out to investigate do not capture, strictly speaking, what their design actually tested and the results they found. In my opinion, the result they find suggests that children may be better at learning a label for a novel object following an Impossible compared to an Improbable or Equivalent-probable event (though I agree with Lotta’s criticism that they might have just learnt this simply because the object appeared from nowhere and the event was highly surprising, there is other evidence showing that infants are more likely to explore and learn something about objects following certain physically impossible events where nothing appears unexpectedly). However, and in my opinion, the study does not unequivocally show that they distinguish and that they make a categorical distinction between the Impossible and the Improbable. For example, kids might only react to Impossible events without actually having a conceptual understanding of the Improbability of the events they saw (I'm not implying that I don't believe they can do this, just that the design does not seem ideal). In this latter case, the percentages of kids who do learn the label (around 40%) might simply reflect the percentage of children who learn the label after having watched something that did not violate their expectations as drastically as when one observes an impossible event. This is further supported by the fact that there was also no difference in proportion of learning between the Equi-Probable and the Improbable conditions.