This thesis investigates whether and how the human cognitive system to extract recursive nested structures from a highly simplified input where hierarchy is marked only by sequential order information. Elaboration and processing of hierarchical representations is involved in many domains of human cognition (Martins, 2012; Uddén et al., 2020), the most notable being language. Natural languages are characterized by structural dependencies in which constituents are linked to each other in such a way that sentences cannot be reduced to the linear relationships between these constituents. Therefore, to correctly interpret a sentence, the cognitive system cannot rely solely on the linear relationship between words but must go beyond this linearity and extract the underlying hierarchical structure of the sentence (Chomsky, 1957). However, demonstrating the building of hierarchical representations in sequences' processing has proven difficult (Levelt, 2020). The difficulty stems from the complexity of implementing hierarchical structure in artificial settings as well as from methodological issues associated with the conventional habituation/discrimination testing procedure. As a result, effects that have been attributed to hierarchical learning can also be attributed to the encoding of the surface properties of the input (Perruchet, 2005).
Our question is thus whether the cognitive system develops hierarchical representations when processing non-linguistic sequences. To address this question, we investigated the processing of aperiodic and self-similar binary strings generated by the Fibonacci grammar (Lindenmayer, 1968). Instead of the habituation/discrimination paradigm, we evaluated the extraction of hierarchical structures by incorporating the strings generated by this grammar into a serial reaction time (SRT) task. By leveraging the properties of the Fibonacci grammar and the SRT task, we were able to investigate the elaboration of hierarchical representations while controlling for the use of surface strategies.