Abstract: Creative thinking relies on dynamic interactions between associative idea generation and cognitive control, and semantic search processes are at the core of this system. However, traditional Remote Associates Test (RAT) paradigms lack fine-grained control over search space constraints and overlook the dynamic nature of the creative thinking process. This study aimed to explore how search space constraints modulate semantic search strategies and patterns during creative problem-solving, using a newly developed Multi-dimensional RAT (MD-RAT) to address these limitations.
We ran two separate experiments for this research. In Experiment 1, we developed and validated the MD-RAT. This task adjusts search space constraints by changing the number of cue dimensions (shape, function, pairing) that link cue words to the target word. A total of 265 healthy undergraduates participated in material validation and psychometric testing. We measured participants' task accuracy and response times, and tested how these task results correlated with their creative achievement, personality openness, and verbal fluency. In Experiment 2, we recruited 42 participants and used a within-subjects design to directly compare high- and low-constraint conditions. We analyzed both behavioral performance (accuracy, number of responses) and fine-grained process indicators, including semantic similarity, category switching, and the sequential dependency of responses.
Our results first confirmed that the MD-RAT has good reliability and validity. Across both high- and low-constraint conditions, participants' task accuracy was significantly correlated with creative achievement, openness, and verbal fluency. We also found clear differences between the two conditions: the high-constraint condition had higher accuracy and fewer total responses than the low-constraint condition, while the low-constraint condition saw far more category switching. Our semantic similarity analyses gave a clear view of the search strategies participants used: in high-constraint conditions, people adopted local search strategies centered on dominant cues, with higher intra-cue than inter-cue similarity. In contrast, low constraints led to global search with more frequent semantic jumps. Besides, we found that search patterns differed that low constraints showed progressive increases in semantic similarity to targets, while high constraints displayed abrupt similarity spikes near solutions.
These findings make it clear that search space constraints systematically shape the creative thinking process. High constraints support focused, efficient local search, while low constraints encourage open, divergent global exploration. Beyond this, the MD-RAT provides a robust tool for investigating dynamic semantic search mechanisms, and helps us better understand how associative strategies and cognitive control work together to support creative problem solving.

Key words: creative problem solving, search space, search strategy, search pattern, semantic network