Word recognition from a frame and fill perspective: Unraveling the impact of spatial frequency processing and prediction generation on reading proficiency
Human Cognitive and Systems Neuroscience
Personality Psychology, Clinical and Medical Psychology, Methodology
Final Report Abstract
This project extends the scope of a neuro-cognitive model of top-down facilitation from visual object to visual word recognition. A lexical decision task (LDT) designed to induce prediction generation via magnocellular pathways (i.e., Silhouette Priming Task; SPT) elicited faster response times and lower response accuracy in a sample of German university students (N = 293). The accuracy decrease was interpreted as indicating prediction errors. Effects were especially pronounced among fast and inaccurate readers; slow but accurate participants were largely unaffected by silhouette primes. Instead of boosting prediction generation, a Case Manipulation Task (CMT) sought to reduce the impact of a word’s outline (i.e., uppercase presentation). Response times in this second LDT were slower in the uppercase compared to the lowercase condition and prominent lexicality effects in the lowercase condition (i.e., higher response accuracy for real words than for pseudowords) disappeared in the uppercase condition. Lexicality effects of response accuracy were less prominent among slow and accurate participants who were also less affected by the case manipulation. Results of both experiments support the notion that coarse word outlines convey useful information that can accelerate word recognition but can also lead to recognition errors if mismatches in detail information are disregarded. The observed individual differences were interpreted in a way that slow and accurate participants rely largely on parvocellular pathways that convey detail information whereas fast and inaccurate participants dominantly use the magnocellular path that transmits coarse shape information quickly. A subset of the original sample (n = 82) accomplished parallel versions of the SPT and the CMT again while their EEG was recorded. For the SPT, reader type and prime type effects were found in time windows 100 – 200 ms after prime and target onset. As neuronal activity in this time window is associated with visual feature extraction, it implies that slow and accurate participants already process silhouette primes differently, possibly extracting less useful information, which – in turn – does not lead to facilitation during subsequent target recognition. A large prime type effect cluster appearing 25 ms before until 100 ms after the target onset might be associated with prediction generation or transfer as proposed in Bar’s top-down facilitation model. Also, for the CMT robust reader type and case manipulation effects appeared in a time window between 100 and 200 ms after stimulus onset and were interpreted as reflecting an enforced processing of fine-grained visual information because word less outline information was available. Since reader type and case manipulation effects, appeared in similar time windows and scalp regions it seems that neuronal processes sensitive to the availability - or the lack - of coarse word outlines affect readers differently. For both tasks (i.e., SPT and CMT) source localization analyses are necessary to investigate whether the observed effects can be linked to neuronal activity in the OFC. Initially planned fMRI data collection was aborted due to the Covid 19 pandemic. Although the manipulation of word frequency created an expected pattern of slower response times and lower accuracy for infrequency words, my hypothesis that silhouettes would pre-activate the more frequent real word of an item triplet was not confirmed; there was no interaction of prime-condition (silhouette vs. zero-prime) x word frequency. A speculative interpretation of this finding is that – if predictive coding is indeed involved – predictions are not made on a word identity level but rather on a level that encodes whether a target is more likely a real than a pseudoword. Individual preferences for global versus local stimulus properties in an also applied Navon Task were expected to indicate magnocellular versus parvocellular pathway preferences. However, individual Navon Task effects did not correlate with individual SPT or CMT effects. An interpretation of this null-effect is difficult.
Publications
- (2018, June). Predicting words from silhouette primes: Contributions of global word-shape and single letters. Workshop on predictive processing. Basque Center on Cognition, Brain and Language in Donostia/San Sebastián, Spain
Korinth, S.P., & Fiebach, C.J.
- (2019, July). Individual differences in visual pathway deployment modulate top-down driven word recognition processes. In S. P. Korinth (Chair), Visual aspects of reading and word recognition: A multi-method and multi-language symposium. Symposium conducted at the 26th Annual Meeting of the Society for the Scientific Study of Reading, Toronto, Canada
Korinth, S.P., Bar, M., & Fiebach, C.J.
