Why are some dimensions integral? Testing two hypotheses through causal learning experiments

Soto, Fabián A.; Quintana, Gonzalo R.; Pérez-Acosta, Andrés M.; Ponce, Fernando P.; Vögel, Edgar

doi:10.1016/j.cognition.2015.07.001

Cited by 15 publications

(27 citation statements)

References 62 publications

(106 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Given this evidence, contemporary associative models of learning assume that the similarity between the components A and B plays a key role in whether the level of summation observed is closer to the one predicted by elemental or configural theories (Harris, 2006; McLaren & Mackintosh, 2002; Melchers, Shanks, & Lachnit, 2008; Pérez et al, 2018; Soto et al, 2014a; Thorwart, Livesey, & Harris, 2012; Wagner, 2008). A similar prediction is made by a recent normative model proposed by Soto, Gershman and Niv (2014a; Soto, Quintana, Pérez-Acosta, Ponce, & Vogel, 2015), which predicts that not only higher perceptual similarity, but also higher spatial or temporal contiguity should produce more configural processing and reduce the summation effect (see also Thorwart et al, (2012)).…”

Section: Introductionsupporting

confidence: 75%

Sub-sampling of cues in associative learning

Narasiwodeyar

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

11Theories of generalization distinguish between elemental and configural stimulus processing, 12 depending on whether stimuli in a compound are processed independently or as distinct 13 entities. Evidence for elemental processing comes from findings of summation in animals, 14 where a compound of two stimuli that independently predict an outcome is deemed to be 15 more predictive of the outcome than each stimulus alone. Configural processing, on the other 16 hand, is supported by experiments that fail to find this effect when the compound is 17 comprised of similar stimuli. In humans, by contrast, summation seems to be robust and 18 independent of similarity. We show how these results are best explained by an alternative 19 view in which generalization comes about from a visual search process in which subjects 20 process the most predictive or salient stimulus in a compound. We offer empirical support 21 for this theory in three human experiments on causal learning and formalize a new elemental 22 visual search model based on reinforcement learning principles which can capture the present 23 and previous data on generalization, bridging two different research areas in psychology into 24 a unitary framework.25

show abstract

Section: Introductionsupporting

confidence: 75%

Sub-sampling of cues in associative learning

Narasiwodeyar

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The results from these five experiments are largely unfavourable to associative models which assume that perceptual similarity should affect summation through its impact on the type of generalization strategy adopted by subjects. To recapitulate, these theories anticipate elemental processing and therefore summation with dissimilar stimuli, but configural processing and lower summation with more similar stimuli (Harris & Livesey, 2010;Soto et al, 2014Soto et al, , 2015Thorwart et al, 2012;Wagner, 2003). None of these predictions were confirmed in the present studies, suggesting that similarity-based models cannot be directly applied to summation in human causal learning.…”

Section: Discussionmentioning

confidence: 99%

“…2 Thus, unlike previous experiments exploring the extent to which stimulus properties would impact on the learning process and hence modulate responding to the components because of the accruement of different associative strengths (see, for example, Livesey & Boakes, 2004), our experiment varied the degree of similarity of visual cues to test whether this factor would have an effect on the generalization strategy after the learning stage, bringing about different levels of summation in groups extra and intra. Several contemporary learning theories predict a stronger summation effect in group extra than in group intra, both when these stimuli vary in featural similarity (Harris, 2006;Harris & Livesey, 2010;McLaren & Mackintosh, 2002;Soto et al, 2014;Soto et al, 2015;Thorwart et al, 2012;Wagner, 2003) and when they vary in others stimulus factors such as spatial separation (Harris & Livesey, 2010;Soto et al, 2014;Thorwart et al, 2012). The resulting data should allow as to probe the theoretical assumption underlying these models that stimulus similarity in general, and not stimulus modality in particular, is what controls compound generalization Experiment 1 9 SUMMATION IN CAUSAL LEARNING The stimuli used for cues A and B in Experiment 1 are shown in Figure 1.…”

Section: Summation In Human Causal Learningmentioning

confidence: 99%

Exploring the role of stimulus similarity on the summation effect in causal learning

Pérez

Martín

Soto

2017

Preprint

Self Cite

View full text Add to dashboard Cite

Several contemporary models of associative learning anticipate that the higher responding to a compound of two cues separately trained with a common outcome than to each of the cues alone -a summation effect-is modulated by the similarity between the cues forming the compound. Here, we explored this hypothesis in a series of causal learning experiments with humans. Participants were presented with two visual cues that separately predicted a common outcome and later asked for the outcome predicted by the compound of the two cues. Importantly, the cues’ similarity was varied between groups through changes in shape, spatial position, color, configuration and rotation. In variance with the predictions of these models, we observed similar and strong levels of summation in both groups across all manipulations of similarity (Experiments 1-5). The summation effect was significantly reduced by manipulations intended to impact assumptions about the causal independence of the cues forming the compound, but this reduction was independent of stimulus similarity (Experiment 6). These results are problematic for similarity-based models and can be more readily explained by rational approaches to causal learning.

show abstract

“…Once this was discovered in the 2000s, interest seemed to shift toward trying to determine under what circumstances one form of representation would be favored over the other. Several models were developed, proposing mechanisms underlying the shift (Harris, 2006;Harris and Livesey, 2010;Soto et al, , 2015a. The similarity hypothesis (Perez et al, 2018) was particularly popular among such models, proposing that elemental generalization (e.g., summation) should occur with very dissimilar stimuli, but configural generalization (e.g., averaging) should occur with very similar stimuli.…”

Section: Configural and Elemental Representation In Associative Learningmentioning

confidence: 99%

Beyond the “Conceptual Nervous System”: Can computational cognitive neuroscience transform learning theory?

Soto

2019

Behavioural Processes

View full text Add to dashboard Cite

In the last century, learning theory has been dominated by an approach assuming that associations between hypothetical representational nodes can support the acquisition of knowledge about the environment. The similarities between this approach and connectionism did not go unnoticed to learning theorists, with many of them explicitly adopting a neural network approach in the modeling of learning phenomena. Skinner famously criticized such use of hypothetical neural structures for the explanation of behavior (the "Conceptual Nervous System"), and one aspect of his criticism has proven to be correct: theory underdetermination is a pervasive problem in cognitive modeling in general, and in associationist and connectionist models in particular. That is, models implementing two very different cognitive processes often make the exact same behavioral predictions, meaning that important theoretical questions posed by contrasting the two models remain unanswered. We show through several examples that theory underdetermination is common in the learning theory literature, affecting the solvability of some of the most important theoretical problems that have been posed in the last decades. Computational cognitive neuroscience (CCN) offers a solution to this problem, by including neurobiological constraints in computational models of behavior and cognition. Rather than simply being inspired by neural computation, CCN models are built to reflect as much as possible about the actual neural structures thought to underlie a particular behavior. They go beyond the "Conceptual Nervous System" and offer a true integration of behavioral and neural levels of analysis.

show abstract

Why are some dimensions integral? Testing two hypotheses through causal learning experiments

Cited by 15 publications

References 62 publications

Sub-sampling of cues in associative learning

Sub-sampling of cues in associative learning

Exploring the role of stimulus similarity on the summation effect in causal learning

Beyond the “Conceptual Nervous System”: Can computational cognitive neuroscience transform learning theory?

Contact Info

Product

Resources

About