Humans possess the ability to transfer the learning acquired previously to new situations, enabling them to adapt and act consistently in a dynamic environment. A similar transfer of learning applies to threatening situations, in which humans generalize the fear induced by one situation to other similar situations. As a threatening stimulus can assume several forms, it is important to generalize the learning from one situation to others based on the perceptual or conceptual similarity of one event to another, which could give rise to similar adverse outcomes1. This interpretation of an event and the ability to generalize based on the similarity between two events may prevent an individual from experiencing the same negative outcome or enable an individual to avoid situations that predict negative consequences. For example, if an individual is claustrophobic and is afraid of enclosed spaces like an aeroplane, he might also be scared of elevators, crowded rooms, subways, or metro railways.
However, overgeneralization of fear may be associated with the acquisition and maintenance of anxiety disorders2–4. In experimental conditions, the differential fear conditioning paradigm is widely used to understand fear conditioning and generalization, in which a neutral stimulus (CS+) is associated with a noxious/aversive unconditioned stimulus (UCS), while another neutral stimulus is not (CS-). In this associative form of learning, the CS + acquires the properties of the UCS and elicits a threat-related conditioned response (CR), whereas the CS- serves as the safety cue.
Fear generalization is an adaptive phenomenon where individuals elicit fear responses to stimuli based on their previous learning or experiences5. A fear generalization paradigm typically comprises two phases: acquisition and generalization6–8. In the acquisition phase, differential fear learning occurs with the CS + and the CS-; in the generalization phase, conditioned responses are observed for the CS+, the CS-, and the generalization stimuli (GSs), i.e., novel stimuli similar to the CSs.
Extant literature suggests that fear generalization occurs due to perceptual2 and conceptual similarity9 between the conditioned and generalization stimuli. Generalization due to perceptual similarity occurs when the GSs and the CSs share similar physical properties, which vary on a perceptual continuum. Generalization studies on perceptual similarity have been previously conducted using geometrical shapes3, colours10, and human faces11,12. Perceptual generalization has also been observed using auditory, tactile, and olfactory stimuli13–15. The studies have indicated that fear generalization increases with increasing similarity of the GS with the CS.
Objects that are physically similar share certain characteristics that enable the generalization of acquired learning from one object to another physically related object. However, learning is also generalized to physically dissimilar objects which are conceptually related by induction. In generalization due to conceptual similarity, fear may be generalized to another object which differs in physical properties but is likely to be threatening owing to some conceptual similarity. Conceptual generalization studies have utilized object categories like animals, tools, furniture, etc.16–18, semantically related words19,20, and sounds21 as conditioned and generalization stimuli. The studies indicated that fear generalized to other stimuli which were conceptually related to the conditioned stimuli22.
The category-based fear conditioning paradigm is most widely used for studying conceptual fear generalization. In category-based fear conditioning, individuals are conditioned to trial-unique, i.e., non-repeating members of a semantic category rather than a single stimulus, and are tested with other members of the same category with which they have never been conditioned16. An aversive experience with a member of a category may elicit similar aversive responses with other objects of the same category based on higher-order cognitive processes like existing conceptual knowledge and reasoning. Even if an individual has not been physically exposed to such related stimuli, the conceptual knowledge of similarity may result in fear response and avoidance. The category-based fear conditioning paradigm may help understand the etiology of anxiety disorders like specific phobias, where fear is expressed towards members of specific categories16.
Further, fear acquisition and generalization depend on the predictability of the threat-related stimulus. Fear is related to a predictive threat, whereas anxiety is related to an unpredictable threat23. The predictability of a threat-related stimulus is affected by the certainty and frequency of occurrence of the threat-related stimulus24. In fear learning, the unexpected omission or occurrence of the UCS after the CS + results in a “surprise” element that influences the prediction of the UCS. This surprise element that gives rise to a sense of uncertainty occurs due to a discrepancy between what is expected by the organism and what actually occurs is known as prediction error25. The certainty of the UCS occurrence is dependent on the UCS reinforcement level. Continuous reinforcement occurs when the UCS is presented after each CS presentation, whereas partial reinforcement occurs when the UCS follows the CS + in some of the trials. In conditions of continuous reinforcement where the UCS certainly follows the CS 100% of the time, the organism learns to predict the UCS upon the occurrence of the CS+. However, during partial reinforcement the occurrence of the UCS after the CS + is uncertain, and the fear learning is not solely dependent on probabilistic prediction26. Therefore, the reinforcement level of the CS-UCS association plays a vital role in fear generalization. The level of UCS reinforcement determines the associative strength of the CS-UCS and may influence the extent of generalization to other similar stimuli27. Continuous and partial reinforcement levels have been observed to produce different effects in fear acquisition, generalization, and extinction.
Grant and Schipper28 studied the effect of reinforcement on conditioned responses using five levels, 0%, 25%, 50%, 75%, and 100%, and found that the conditioned responses increased with an increasing percentage of reinforcement. Further, Silver et al.29 reported that fear acquisition, measured by skin conductance responses, occurred with 75% and 100% reinforcement and not with 25% or 50% reinforcement. A linear relationship was observed between the CS-US reinforcement levels and conditioned fear responses30. Further, Dunsmoor and colleagues30 also suggested that the fear learning patterns are different with partial and continuous schedules of reinforcement. This has been reflected in a study in which partial followed by continuous reinforcement was found to be most effective in the acquisition of fear resulting in high conditioned responses and increased resistance to extinction31. Kitamura and colleagues32 reported that a higher partial reinforcement schedule of 80%, compared to a lower partial reinforcement schedule of 40%, resulted in a stronger fear memory. Partial reinforcement also results in cognitive biases due to the uncertainty of UCS occurrence. Participants have been observed to demonstrate an expectancy bias, reporting greater expectancy of the UCS when the UCS occurrence is uncertain (partial reinforcement) rather than certain (continuous reinforcement). They also show a covariation bias where they overestimate the CS-UCS contingency 33.
Therefore, previous literature suggests that partial and continuous reinforcement schedules have a varied effect on fear learning, which may also influence fear generalization. As fear generalization does not result due to a direct association between the CS and the UCS and is based on the similarity of the novel stimulus and the CS34, the uncertainty of the UCS occurrence may impact the generalized fear responses.
In this light, Zhao et al.24 were the first to compare the effects of continuous and partial UCS reinforcement levels on fear generalization using perceptually similar generalization stimuli. It was found that partial reinforcement resulted in enhanced fear responses towards the generalization stimuli and difficulty in safety learning, whereas continuous reinforcement leads to broader generalization.
The aim of the current study was to investigate the effect of reinforcement level on conceptual fear generalization due to category-based similarity. In this study, we employed a category-based fear conditioning paradigm in which exemplars from three stimulus categories (abbreviated as CAT+) CAT1+, CAT2 + and CAT3 + were followed by an UCS with 100%, 62.5%, and 37.5% levels of reinforcement, respectively, whereas the exemplars from the CAT- category were unreinforced. Therefore, the participants acquired learning about a threatening and a safe category16,17,35. In the generalization phase, participants were exposed to novel stimuli (that have not been presented to them earlier during the experiment), and the conditioned fear responses were measured. This process is different from a standard differential conditioning procedure in which a single stimulus is paired with an aversive stimulus multiple times during the acquisition phase. In the category-based conditioning paradigm, each stimulus was presented only once. The participants used the conceptual learning from each trial to generalize to other related stimuli and predict the occurrence of the UCS. We hypothesized that the conditioned fear responses would be generalized based on the conceptual knowledge about the categories and the UCS reinforcement level in each category.