As we all know, previous mental processing can influence our subsequent behaviors, such as priming. This factor has garnered significant attention in research. However, most existing studies have focused on the influence elicited by single items, meaning the impact of processing one stimulus on a subsequent encounter with another stimulus. The importance of semantic relations between individual primes and targets has been verified. However, there has been limited focus on how stimuli experienced in pairs can affect subsequently encountered pairs, particularly associative pairs. It is worth investigating whether the semantic relation can also play a significant role in the priming of associative pairs. We will first conduct a thorough examination of the multifaceted phenomena and underlying neural mechanisms associated with both priming and associative pairs. Additionally, we will outline our research objectives, shedding light on our mission to bridge the gap in our understanding of how associations shape priming patterns in both behavioral and neural activity.

Within the realm of implicit memory, extensive research has confirmed that stimuli encountered previously can impact the processing of subsequently presented stimuli. This influence can either hinder or enhance the way subsequent stimuli are processed. The hindrance or enhancement in performance can endure for a few minutes or even an extended period (Barrón-Martínez et al., 2022, Jia et al., 2023, Nie and Yu, 2021, Roelke et al., 2018, Steinhauer et al., 2017). Overall, priming is an essential aspect as it covers various forms, including perceptual priming (Hu and Liu, 2019, Wang et al., 2020) and conceptual priming (Barrón-Martínez et al., 2022, Cochrane and Pratt, 2022, Gilligan and Rafal, 2019, Nie et al., 2016, Nie et al., 2016, Nie et al., 2014, Nie et al., 2021, Nie and Yu, 2021, Roelke et al., 2018, Zhang et al., 2019). An example of perceptual priming is when the initial presentation of a word (also called prime), like “ladder,” enhances the later identification of related words (also called targets) that are presented in the same form, such as “rabbit,” compared to an unrelated word, like “RABBIT.”.

Moreover, conceptual priming can be further classified into semantic priming and repetition priming, and both of them have raised notable concerns. In the case of semantic priming, the prime and the target need to possess a shared semantic relation (Barrón-Martínez et al., 2022, Jia et al., 2023, Lee and Zhang, 2018, Roelke et al., 2018, Ruiz et al., 2018). Repetition priming is characterized by a change in an individual’s capacity to recognize, detect, or generate a stimulus. This change occurs after a prior exposure to the same prime as the target. Generally, the change is evident as an enhancement in both the swiftness and precision of performance. For instance, there is a quicker response and/or higher accuracy in identifying a frequently presented stimulus compared to a solitary occurrence (Alzueta et al., 2019, Cochrane and Pratt, 2022, Kuper et al., 2015, Nie et al., 2016, Nie et al., 2016, Nie et al., 2021, Nie et al., 2021, Lee and Zhang, 2018, Nie and Yu, 2021, Zhang et al., 2019).

We are particularly interested in semantic priming and repetition priming. These two types of priming are typically studied separately, and research has shown that they have distinct effects. In the current design, we have decided to merge them into a single study.

In research considering semantic priming, it is common to compare at least two cases by manipulating the prime and the target as semantic priming and unrelated cases. For example, words like “scale” and “range” are often used as examples of semantic priming. The performance enhancement of semantic priming is often compared to unrelated cases, such as “date” and “moose” (Barrón-Martínez et al., 2022, Jia et al., 2023, Lee and Zhang, 2018, Roelke et al., 2018, Ruiz et al., 2018, Sánchez-Casas et al., 2012, Tibon et al., 2014, Tibon et al., 2014). With the continuous advancement in the study of semantic relations between stimuli, researchers are not only focusing on distinguishing between semantic priming and unrelated conditions, but also intensifying their attention towards the influence of different semantic relations on semantic priming (Chen et al., 2014, Jia et al., 2023, Liang et al., 2020, Mech et al., 2022, Ortells et al., 2016).

To achieve this, researchers have endeavored to classify semantic relations into two distinct types: thematic and taxonomic (Kumar, 2018, Liang et al., 2020, Nie et al., 2021, Nie et al., 2021, Nie and Wu, 2023, Sachs et al., 2008, Sass et al., 2009). Thematic relation generally refers to the relation between different items that play varying roles in the same scene. This relation is based on an individual’s experience and highlights the external connections among the items. Thematic relation items should interact or complement each other functionally. For example, the relation between a “computer” and a “table” is spatial, as the computer is typically placed on the table. Similarly, the relation between a “joke” and “laughter” is one of cause and effect. On the other hand, taxonomic relations are generally determined by similarity. This similarity is identified by shared attributes and overlapping features across items, such as the relation between an “apple” and a “watermelon” or between a “pen” and a “pencil”.

Categorization is a fundamental principle for organizing semantic information. It enables us not only to identify objects and connect them to other familiar entities but also to understand how new and unfamiliar objects fit into our pre-existing knowledge structure. It is crucial to acknowledge that classical taxonomic categories, which group similar items together, differ significantly from thematic categories in various important aspects (Sachs et al., 2008, Sass et al., 2009, Nie et al., 2023). One way to interpret the differences between thematic and taxonomic relations is to classify similar items as categories, while referring to thematically related items as semantic associates. Another approach is to consider thematic and taxonomic relations as two distinct yet equally significant ways of organizing semantic knowledge into categories. This prompts us to have a particular interest in exploring the contribution of these two types of semantic relations to priming, as revealed through neural activities.

Besides behavioral enhancement, neural activity regarding semantic priming has also raised concerns. Studies that utilize techniques with a high spatial resolution, such as functional magnetic resonance imaging (fMRI) (Roelke and Hofmann, 2020, Sachs et al., 2008, Sass et al., 2009), as well as techniques with a high temporal resolution like event-related potential (ERP) (Chen et al., 2014, Jia et al., 2023, Liang et al., 2020, Mech et al., 2022, Ortells et al., 2016), have shed light on this.

In an fMRI study conducted by Roelke and Hofmann (2020), it was discovered that, in the case of semantic priming, there was an increase in functional connectivity between the left inferior frontal gyrus and various brain regions, including the fusiform gyrus, the hippocampus, the anterior cingulate cortex, and the orbitofrontal cortex. This observation suggests a connectivity pattern similar to that of the semantic layer. Furthermore, there were no differences in brain activations between low and high semantic similarity.

Furthermore, the distinction between thematic and taxonomic relations in semantic priming is further supported. Sass et al. (2009) utilized automatic processing conditions and auditory-to-visual semantic priming. Behavioral data revealed a priming effect for words that were thematically related, but not for those that were taxonomically related. On a neural level, thematically related words showed activation in the left-lateralized temporal region, while taxonomically related words elicited right-lateralized frontal activation within the hippocampus. Activation was more pronounced for thematically related words, whereas taxonomically related trials resulted in a suppression of the response in left superior temporal sulcus.

Sachs et al. (2008) found that the impact of thematic relations was greater than that of taxonomic relations in priming. Specifically, taxonomic priming resulted in significant brain activity in the right precuneus, postcentral gyrus, middle frontal gyrus, and superior frontal gyrus. In contrast, thematic priming was associated with signal changes in the right middle frontal gyrus and anterior cingulate. Notably, the signal changes in the right precuneus were more pronounced in taxonomic priming than in thematic priming. These findings indicate that thematic priming and taxonomic priming affect brain signals differently.

In addition to fMRI publications, ERP research also helps differentiate and elucidate the roles of various semantic relations in semantic priming. This can be demonstrated through the use of the N400 and other ERP components, such as LPC (late positive complex) and frontal negativity. The N400 component is a negative-going wave that peaks approximately 400 ms after the presentation of a meaningful stimulus. It is commonly associated with the relatively automatic aspects of semantic access (Kutas and Federmeier, 2011, Mech et al., 2022, Ortu et al., 2013). The LPC is a post-N400 component characterized by a posterior scalp distribution. It has been associated with enhanced explicit memory retrieval and strategic semantic processing (Kutas and Federmeier, 2011, Mech et al., 2022). A study conducted by Mech et al. (2022) found that both the N400 and LPC components were present in semantic priming. However, the study did not go into further differentiation of the semantic relations involved.

Additionally, Chen et al. (2014) and Liang et al. (2020) differentiated the effects of different semantic relations in priming. Chen et al. (2014) observed faster reaction times for semantically-related prime-target pairs compared to unrelated cases. The ERP data showed that an early N400 effect (200–400 ms) was more negative for unrelated pairs than for those of taxonomic and thematic relations. Furthermore, the study found that the activation of frontal negativity (at 400–550 ms) was reduced when priming productive relations, a subtype of semantic relations, compared to other related words. The authors suggest that the allocation of attentional resources and subsequent recruitment of additional memory processing may serve as two key indicators of thematic relations. Liang et al. (2020) also revealed both N400 and frontal negativity in their study, indicating that taxonomic and thematic relations play distinct roles in organizing object knowledge.

The investigations mentioned above show that taxonomic and thematic relations display different behaviors in priming, and also provide evidence from neural signals. However, the above fMRI investigations (Sachs et al., 2008, Sass et al., 2009) and ERP studies (Chen et al., 2014, Liang et al., 2020, Roelke et al., 2018) focus on the priming of a single item, specifically a prime item and a target item. To our knowledge, no study has specifically investigated priming related to pairs, especially pairs with distinct semantic relations. There is currently no evidence indicating whether brain activities associated with priming differ for pairs with different semantic relations. Our goal is to explore whether the priming effects are influenced by semantic relations, encompassing thematic relations, taxonomic relations, and unrelated cases.

The reason to consider semantic relations is that associative pairs have a theoretical basis. The levels of unitization framework for explicit memory (Nie and Wu, 2023, Parks and Yonelinas, 2015) provide evidence suggesting higher accuracy for pairs whose items can be easily integrated (e.g., pairs such as “black” and “board” that can be combined into “blackboard”) compared to pairs that cannot be integrated. This framework also suggests that there is a benefit for integrated pairs, as the unitization process leads to increased familiarity. There are two channels to enhance unitization between items: the top-down method involves controlling the encoding tasks, while the bottom-up method focuses on improving the semantic relations between items (Nie & Wu, 2023). This prompts us to determine whether the semantic relations obtained through the top-down and bottom-up methods would facilitate the priming of associative pairs.

Based on this, our main interest lies in priming stimulus pairs rather than single items. In addition to repeating intact pairs that are identical to the ones presented before, we also considered other pairs that have been studied in previous associative memory research. In a typical associative memory paradigm, participants are typically asked to encode a series of paired items, such as “apple-barrel,” “wood-banana,” and “phone-bicycle.” During the subsequent test phase, participants must differentiate between intact pairs (e.g., “apple-barrel”), rearranged pairs where the items come from different original pairs (e.g., “wood-bicycle”), “old + new” pairs in which only one item was old (e.g., “banana-cup”), and new pairs where the items have not appeared before (e.g., “camel-desk”) (Nie, Jiang et al., 2021; Nie and Jiang, 2021, Nie and Wu, 2023, Osth and Fox, 2019). It has been shown that intact pairs are typically memorized better than rearranged pairs, and the ERP old/new effects confirm this either (Nie & Wu, 2023). Additionally, these effects are influenced by the manipulation of semantic relations, with thematic relations resulting in a late old/new effect in intact, rearranged, and “old + new” pairs. However, this late effect is only triggered by intact pairs with taxonomic and unrelated relations.

This prompts us to investigate whether the priming effects differ among various types of pairs, including intact, rearranged, and “old + new” pairs. The priming of intact pairs refers to repetition priming, whereas the priming of rearranged and “old + new” pairs is known as semantic priming. We have decided to combine semantic priming and repetition priming into one comprehensive study. In this study, we will use picture pairs with different semantic relations, including thematic, taxonomic, and unrelated relations, as both primes and targets. Moreover, we plan to examine whether ERP waveforms, such as the N400 and LPC, are influenced by factors such as the status of the semantic relation and pair type. This can help us analyze the time latency of associative pairs in priming and determine if the latency of both semantic access and strategic semantic processing occurs later than that of single items. Furthermore, it is important to consider whether the latency of these two processes would be sensitive to the semantic relation between the items of a pair.

In summary, previous studies on semantic priming have primarily focused on the effects elicited by a single prime and a single target, with little attention given to priming effects related to pairs of stimuli. This area is worthy of further exploration as it can help determine whether the semantic relations obtained through the top-down and bottom-up methods, as suggested by the levels of unitization framework, would facilitate associative priming. To bridge this research gap, we have utilized pictures as stimuli to represent thematic relations, taxonomic relations, and unrelated relations. These stimuli were categorized into four groups: intact pairs, rearranged pairs, “old + new” pairs, and new pairs. Regarding priming, the new pairs functioned as the prime, while the other three pair types served as the target. In our current case, the task is to discern the pairs as either thematic, taxonomic, or unrelated using the top-down method, while priming between pairs is perceived as the bottom-up method. We are not only interested in the behavioral patterns elicited by priming, but also in the underlying neural signals. Therefore, we applied the ERP technique to record neural activities.

From a behavioral perspective, we assumed that the priming effect of thematic relations was significantly greater than that of taxonomic and unrelated relations. Additionally, the effect of taxonomic relations tended to be larger in comparison to unrelated relations. The priming effect would be influenced by pair type, with intact pairs demonstrating a greater priming effect compared to rearranged and “old + new” pairs. Furthermore, The interaction between the contribution of semantic relation and pair type was evident.

Regarding the neural aspects, we hypothesized the presence of reliable N400 and LPC components. Both of these components were observed to be influenced by the semantic relations of the pairs. Pairs with thematic relations tended to elicit much larger components compared to the other cases, whereas unrelated pairs elicited smaller components. We expected that the N400 would show greater susceptibility to semantic relation, especially when compared to the LPC. This assumption was based on the fact that the N400 serves as an indicator for semantic processing. Additionally, it was observed that these components were on a larger scale in intact pairs compared to both rearranged pairs and “old + new” pairs. There was a particularly significant difference in intact pairs that showed thematic relations. It should be clarified that our focus was on LPC rather than frontal negativity. This distinction is important because the LPC component reflects strategic semantic processing, which aligns with our specific focus.

In addition to the N400 and LPC components, we also examined another component, known as the N300. Previous studies have indicated that the timing and polarity of the effects observed in the 200–300 ms time window coincide with those of the N300 component. This component is believed to reflect the processes involved in object identification, specifically item identification (Truman & Mudrik, 2018). The N300 is more widely dispersed towards the frontal area and is believed to be associated with the rapid matching of visual input to stored semantic knowledge. It is only observed in response to picture stimuli (Maguire et al., 2013, Truman and Mudrik, 2018). Intriguingly, we employed pictures as materials in our investigation. Regarding this component, we hypothesized that it also functioned as the factors of semantic relation and pair type, and behaved in a manner akin to the N400 component in terms of its sensitivity to these factors. This can help us analyze the time latency of associative pairs in priming and determine if the latency occurs later than that of single items. Furthermore, it is important to consider whether the latency would be sensitive to the semantic relation between items in a pair.