This study investigated whether the strength of the link between working memory capacity and fluid intelligence differs as people use different strategies to solve fluid intelligence problems. A sample of 214 university students completed three complex span tasks measuring working memory capacity and Raven's Advanced Progressive Matrices (APM) assessing fluid intelligence. Strategic behavior was measured by both the strategy questionnaire and the eye-tracking technique. Latent profile analysis yielded three groups of participants using constructive matching, response elimination, and a third strategy, respectively. Participants adopting constructive matching spent proportionally more time on the matrix area but less time on the response bank, and exhibited longer latency to first toggle and higher rate of toggling than those using response elimination and the other strategy, consistent with the results obtained from the questionnaire data. Furthermore, strategy use moderated the relationship between working memory capacity and APM performance. The link between working memory capacity and APM scores was significantly higher for participants using eliminative strategies including response elimination (r = 0.63) and the other strategy (r = 0.54) than that for those using constructive matching (r = 0.27). Our findings suggest that the extent to which working memory relates to performance on APM varies as a function of strategy use.

这项研究调查了工作记忆能力和流体智力之间的联系强度是否随着人们使用不同的策略来解决流体智力问题而不同。214 名大学生的样本完成了三个复杂的跨度任务，测量工作记忆容量和 Raven 的高级渐进矩阵 (APM) 评估流体智力。战略行为是通过战略问卷和眼动追踪技术来衡量的。潜在概况分析分别使用建设性匹配、响应消除和第三种策略产生了三组参与者。采用建设性匹配的参与者在矩阵区域上花费的时间相应地更多，但在响应库上花费的时间更少，并且与使用响应消除和其他策略的人相比，首次切换的延迟更长，切换率更高，这与从问卷数据中获得的结果一致。此外，策略使用调节了工作记忆容量和 APM 性能之间的关系。对于使用消除策略（包括反应消除）的参与者，工作记忆容量和 APM 分数之间的联系明显更高（r  = 0.63）和其他策略（r  = 0.54），而不是那些使用建设性匹配的策略（r  = 0.27）。我们的研究结果表明，工作记忆与 APM 表现的相关程度随策略使用而变化。