The Processing Of Causal Relation And Its Asymmetrical Representations In Semantic Processing

Semantic memory is described as our long-term repertoire of world knowledge (Tulving,1972), which is very important for our learning and everyday thinking activities. Without this knowledge, we would be unable to understand the world around us and hence incapable of communicating or acting in the service of goals (Hodges&Patterson,1997). Semantic memory contains not only knowledge about categories and features that we use to represent the world, but also knowledge about relations between categories and features, including information about categories and features, as well as the complex semantic relations between them, such as "is used to,""works in,""lives in,""is made of,""is kept in,""is the outside of," and hierarchical relations (Murphy&Medin,1985; Spellman, Holyoak,&Morrison,2001). Based on this division of semantic relationships, people are constantly engaged in everyday learning and memory, classification and reasoning, problem solving, decision-making and other higher cognitive activities. Therefore, it is very important to explore how individuals process different types of semantic relationships in order to further understand the neural basis and cognitive mechanisms of a variety of higher cognitive activitiesThe existence of multiple relations within semantic memory raises a particularly interesting question that might be neglected by theories of semantic knowledge: Assuming that different types of semantic relations are relevant in different contexts, how are specific relations accessed within a network that contains various types of semantic relations? Furthermore, although previous studies have explored the knowledge about relations in semantic memory, most of them focused on taxonomic (e.g., bee and butterfly, both of them are insect) and thematic relations (e.g., bee and honey, bee produce honey). Yet although causal relation is initial regarded as examples of thematic relations (Lin&Murphy,2001), there are three features of causal relations that set it apart from other forms of semantic relations:proximity, exclusivity, and priority (Hume,1739/1978). That is, the cause and effect occur at proximal moments in time (i.e., proximity), cause-effect order has an exclusive association (i.e., exclusivity), and causes precede the effects (i.e., priority), but the effects do not typically precede causes (Fenker, Waldmann,&Holyoak,2005; Denkinger&Koutstaal,2013). The main goal of this study was to assess whether the neural basis and cognitive mechanism of causal relations was specific to the asymmetrical representations of causal relations or more generally to all asymmetric relations (e.g. unidirectional associative strength, hierarchical relation). Therefore, in order to explore the cognitive and neural mechanisms of different types of semantic relations, this study compared the asymmetrical representations of causal relations with other symmetrical and asymmetrical relations, such as noncausal associative relation, hierarchical relation, and unidirectional associative strength, by using the traditional behavior method and event-related potential (ERP) techniques with high time resolution.The whole studies are divided into four parts in details. The first part consists of Experiment1and2using ERP technique, aiming to examine the time course of how stored causal relations are represented and accessed in semantic memory via ERPs. In experiment1, participants were required to remember a task cue (causal or associative) presented at the beginning of each trial, and assess whether the relationship between subsequently presented words matched the initial task cue. We found two main differences related to the processing of causal and associative relations. The first difference was a N400effect that was more negative for unrelated words than for causal and associative related words. The second difference was a frontal-central distributed P600that was larger for causal related words than for both unrelated and associative related words. The P600effect was also elicited by the causal task cue than by the associative task cue.In experiment2, we further examined whether these differences were specific to causal relations or more generally to other asymmetric relations (e.g., hierarchical relations), and explored how stored causal relations are represented and accessed in semantic memory via ERPs. Participants were required to evaluate whether pairs of words were related in any way. The N400effect elicited by unrelated word pairs was larger than causal related word pairs, which indicated that causal relations may attract more attention resources. The P600elicited by causally related pairs was larger than noncausally associated pairs and hierarchically related pairs known to be asymmetric. These results suggested the causal and associative relations engaged distinct neural processes, and provided evidence that the processing of causal relation involved more attention resources and recruited additional higher-order executive resources.The second part consists of Experiment3and Experiment4, aiming to examine whether these differences are specific to causal relation or more generally to hierarchical relation. In experiment3, we explored the nature of causal asymmetry by manipulating the order of word presentation and type of asymmetrical relationship. In experimental3a, the participants were required to decide whether the presented word pairs were causal related or hierarchical related. The behavior results revealed that the asymmetry was found for causal related words, but not for hierarchical related words. However, the amplitude of P2elicited by superordinate-subordinate order was larger than reverse order, which might reflect the detection of the order of the stimuli. Similar P2effect was found for causal related words, and the P600effect elicited by cause-effect order was larger than vice versa, which suggested that participants appear to distinguish the specific roles of cause and effect. Thus, although participants have noticed the orders of hierarchical related words, they did not distinguish the specific roles of each word, which was different from the representation of causal asymmetry. When participants were required to evaluate whether pairs of words were related in any way in experimental3b, however, no significant different P2and P600were found between two orders of word presentation in both types of asymmetrical relationship. These results were consistent with previous studies (e.g., Fenker et al.,2005), which suggested the causal-model view seemed to be more suitable to the asymmetry representation of causal relation.In experiment4, we manipulated the spatial arrangement of word presentation and the types of asymmetrical relationship. In two experiments, the participants were required to decide whether members of a simultaneously presented word pair were causal related or hierarchical related. In experiment4a, the words were presented one above the other. The results revealed that verifying the existence of a causal or hierarchical relation is faster when the words representing cause or superordinate level are presented in the top rather than in the reverse order. However, experiment4b showed that only causal asymmetry occurred when the words were presented horizontally. People appear to represent the causal asymmetry based on temporal order when queried about a causal relation, whereas the representation of hierarchical asymmetry is based on spatial arrangement when queried about a hierarchical relation.The third part consists of Experiment5and Experiment6, aiming to examine whether the asymmetrical representations of causal relations are specific to causal relations or more generally to unidirectional associative strength. In experimental5, the participants were required to decide whether the presented word pairs were causal related or associative related. The behavior results revealed that the asymmetry was found both for causal related words and unidirectional associative related words. However, the P600effect elicited by cause-effect order was larger than vice versa, whereas the amplitude of N400elicited by S1-S2order was larger than reverse order, which might reflect the asymmetrical representations of causal relation was different from unidirectional associative strength.In experiment6, we further compared the causal relation with hierarchical relation and unidirectional associative strength by manipulating the SOA (150ms vs.1000ms) and the presentation time of S2(150ms vs.1000ms). Participants were required to decide whether the presented word pairs were causal related, hierarchical related, or associative related in separate blocks. Results were as follows:(1) No significant asymmetrical representations were found for causal relation, hierarchical relation and unidirectional associative strength in150-150ms presentation condition.(2) The asymmetrical representation was only found for causal relation in150ms-1000ms and1000-150ms presentation conditions, but not for other two conditions.(3) This asymmetrical representations were found both for causal relation and unidirectional associative strength in1000ms-1000mspresentation condition, but not for hierarchical relation.The forth part consists of Experiment7, aiming to investigate the processing of temporal and causal order information by using black-and-white line drawings depicted a diverse array of recognizable events. Participants were required to evaluate whether pairs of line drawings were causal related (Task1) or associated related (Task2). Both the behavior and ERP data have revealed a pattern consistent with that observed with the experiments with our above studies. In the causal judgment task, the RTs for cause-effect order were shorter than that for effect-cause order, and larger positive component between500ms and700ms were found for cause-effect order than effect-cause order. Furthermore, the late positivity elicited by drawing pair with small time distance was lager than drawings with large time distance. What is more, the interaction between presentation order and time proximity was significant. Specifically, the RT advantages for cause-effect order were found when the time distance is large, whereas no such RT advantages were found when the time distance is small. However, no such RT advantages and larger late positivity were found for cause-effect order in the associative related judgment task. Furthermore, the ERP data showed that difference between drawing pairs with different time distances was found between300-500ms, rather than500-700ms like causal judgment.In summary, our data yielded new insights into asymmetrical representations of causal relations. These results suggested that the processing of causal relations stored in semantic memory might recruit additional higher-order executive resources above and beyond those afforded by other asymmetric relations. Taken together, the following conclusions could be drawn in this research:(1) The processing of causal relations was different from associative relations, which elicited smaller N400effect and larger P600effect. These studies indicated the processing of causal relations recruit additional processes, such as prediction and executive resources.(2) The asymmetrical representations of causal relations are different from other asymmetric relations. The causal asymmetry was mainly affected by time order and time proximity (P600), and hierarchical asymmetry was impacted by the mental representation of spatial location (P2), whereas the unidirectional associative asymmetry was caused by the different associative strengths of two directions (N400).(3) The asymmetrical representations of causal relations exist not only in word processing, but also in the processing of picture stimuli. These results suggest that the causal asymmetry is a cognitive bias which widely exists in human cognitive activity.(4) Our results suggested that participants have noticed and distinguished the specific roles of causes and effects, which was more suitable to the causal-model view, rather than associative view. Based on previous studies, we proposed a general architecture of causal learning and reasoning, in an attempt to more accurately predict and guide our related research in the future. In the framework, the causal learning was divided into four processes:causal prediction, causal induction, abductive reasoning and the effect of causal induction and causal attribution on mental health.