priming and competition of associated memory representations: a comparison between response times and event-related potentials following lesions to left temporal cortex

Introduction The organisation of semantic memory underlies our use of language. Memory associations can facilitate behaviour (e.g., as seen in semantic priming), but can also interfere with selecting a response because of increased competition between activated representations (e.g., Piai et al., 2014). To study this mechanism, we conducted a word-picture verification task while manipulating the associative strength between the word and the picture (norms from Nelson et al., 2004). Methods Participants were four chronic stroke patients with left temporal lesions (median age = 61.4, education = 16) and six education- and age-matched controls (median age = 64.5, education = 16). Participants heard a word and subsequently saw a picture while their electroencephalogram was recorded. Word and picture were either congruent (e.g., 'apple' followed by the picture of an apple) or not. In this latter case, the word and the picture were either unrelated (e.g., word: 'banjo', picture: apple) or semantically related (e.g., word: 'banana', picture: apple). Participants were instructed to respond with a button press whether the word and the picture denoted the same concept or not. Accuracy and response times (RTs) were registered. Results Patients had more errors than controls (incorrect responses and omissions, p < .001). For correct responses (shown in the left panel of the figure), patients and controls did not differ in their RTs, but overall responses for related items were slower than for unrelated items in both groups (i.e., RT related > RT unrelated, p = .001). This effect reflects semantic interference from the association between the word and the picture. Controls showed a graded N400 event-related potential (ERP) effect, that is, a negativity between 300 and 500 ms post-picture onset following a pattern inversely related to the degree of relation between word and picture (unrelated > related > congruent, p = .002). The group-average ERP in the controls is shown in the middle panel of the figure. The graded pattern observed on the group level was also clear within each control participant. This N400 effect suggests a semantic priming mechanism. For the left temporal patients, however, this effect was not consistent (p = .09). As it can be seen in the right panel of the figure, only one patient showed a graded N400 effect (each small panel represents the ERPs of one patient). Interestingly, in the patients, the relationship between the size of the semantic effect in the RTs and in the N400 effect was strong (N = 4, adjusted r = .943, p = .037): The larger the priming effect, as indicated by the N400 effect (unrelated > related), the bigger the semantic interference in the RTs (related > unrelated). Conclusion These results suggest that associated concepts and words in memory prime each other (as indexed by the N400 effect), but also incur a stronger competition between them (as indicated by the RT effect), delaying response selection.