305 21 1296 2 4969 document 1296 Semantic Hebb Data.zip application/zip eddcb810503443c86afc43c02e7086bb MD5 45642 2023-05-09 10:26:43 https://bordar.bournemouth.ac.uk/305/1/Semantic%20Hebb%20Data.zip 305 1 1 application/zip archive Anonymised data for 3 Semantic Hebb repetition effect experiments. README file included. en public cc_by_nd_4

Semantic Hebb Data.zip

full_archive archive 9417 disk0/00/00/03/05 2023-06-30 11:40:51 2023-06-30 11:41:23 2023-06-30 11:40:51 data_collection show Legg Rebecca leggr@bournemouth.ac.uk Johnson Andrew andjohnson@bournemouth.ac.uk facsci Hebb Repetition Effect, Semantic Learning, Working Memory, Cognitive Psychology There are numerous pieces of empirical evidence (Hebb, 1961; Johnson et al., 2017; Page et al., 2013) showing that repeating a sequence of information results in better serial recall accuracy relative to non-repeated sequences (i.e., the Hebb repetition effect, HRE). Additional information, other than exact item and order repetition, can produce HREs such as motor responses (Johnson et al., 2017) and metrical patterns (Paice et al., in preparation). There has, however, been no investigation as to whether semantic information can be acquired in a similar way to item and order information, despite evidence showing semantic similarity improves recall in Immediate Serial Recall (ISR) tasks (Saint-Aubin et al., 2005). The current research therefore investigated whether a HRE for semantic information exists across three experiments using HRE procedures. Experiment 1 showed no evidence of a HRE when only a semantic pattern was repeated, (i.e., without exact item repetition). Experiment 2 replicated the canonical HRE with exact list repetitions; however, the recall advantage generated did not transfer to novel item lists following the same semantic pattern. That is, Experiment 2 showed that once learnt, sequence knowledge was not transferred to a semantically related list. Lastly, Experiment 3 adopted a typical Hebb repetition paradigm wherein participants learnt lists of category labels; however, at test participants reconstructed the lists using exemplars of the category labels. There was a significant HRE, however, the exact mechanism driving this effect is unclear. In general, results are discussed alongside two prominent models of the HRE (Burgess & Hitch, 1999, 2006; Page & Norris, 1998, 2009). Overall, findings across the three experiments suggest that semantic information is not acquired in a similar way to item and order information when learning new lists, therefore, supporting the two models in their current form as explanations of the HRE. 2023-05-09 published Bournemouth University 10.18746/bmth.data.00000305 Text Rebecca Legg bordar@bournemouth.ac.uk Principal Investigator Legg Rebecca leggr@bournemouth.ac.uk thesis https://eprints.bournemouth.ac.uk/38447/ pub Bournemouth University Data was coded as '1's for incorrect answers and '0's for correct answers, then averaged for each lists 'error' rate. Experiment 1 Per block analysis = to check for differences per block (1-5) you will need to average error rates for Hebb and filler lists for each block (five columns). This can then be analysed using a 2x5 ANOVA. Combined block analysis = average all Hebb and fillers, regardless of block number (1-5 only). Create a slope column for Hebb and filler lists, then perform a t-test comparing the slopes. Block 6 Analysis = ONLY analyse if significant learning if found in blocks 1-5. This was not analysed in our experiments. If needed though, average the 5 Hebb list error rates and the 5 filler list error rates then compare using a t-test on slopes. Experiment 2 - the two files are in the correct format for ANOVA/t-test analysis. You can run a 2x7 repeated measures ANOVA on the initial learning data, or a t-test comparing Hebb and filler slopes (create a slope for each participant across epochs). You can run a 2x2 repeated measures ANOVA on the transfer data. Experiment 3 - the file is in the correct format for ANOVA/t-test analysis. You can run a 2x7 repeated measures ANOVA on the main data, or a t-test on the slopes of Hebb and filler lists (create a slope for each participant across epochs) See README (Data Information) file for specific details about each experiment, such as column names etc. All experiments were online studies (using Gorilla) and participants were presented with lists of 6 words (auditory presentation) which they had to recall using serial order reconstruction. Every third list the same list was repeated (or the same semantic pattern was repeated with different exemplars). Items were scored as '0' if correct, and '1' if incorrect, then averaged across list types (e.g., Hebb and filler). This scoring meant that we were analysing error rates (rather than accuracy). Experiment 1 Hebb lists were different exemplars but following a repeated semantic pattern (Apple, Blue, Cow or Peach, Red, Horse). Filler lists were different exemplars and different semantic orders (Apple, Blue, Cow or Red, Horse, Peach). 70 trials overall (5 blocks of 12 trials and 1 block of 10 trials - see read me file for rationale). Experiment 2 Initial Learning - Hebb trials followed the typical structure (exact item repetition) as did the filler lists (novel items and orders). 21 trials. Transfer of Learning - 4 Novel Hebb lists (new items that follow the previous semantic pattern) and 4 novel filler lists (new items and new semantic orders) were then presented. Experiment 3 All lists were presented as categories (Fruit, Colour, Animal) but when reconstructing the lists participants used exemplars (Apple, Blue, Cow). Hebb lists were categories presented in the same serial order. Filler lists were categories presented in novel orders. All lists had novel exemplars. Sample Participants for the experiments were recruited via sona (Bournemouth University Psychology students), compensated with course credits, and Prolific (non-students), compensated with money. All participants were English speaking and aged between 18 and 50 years old. No legal or ethical issues during the generation of the data. Ethical approval was given, in line with the BPS, before data collection. All data was anonymous. 2020-09 2022-08