社會共享式的提取引發遺忘之統合分析

Abstract

「個體內提取引發遺忘」(WI-RIF)是一種回想時的記憶抑制現象，和不同類別的訊息相較，個體會在其他相同類別卻未複習到訊息有較大的遺忘。而提取引發遺忘亦會在社會情境下發生，稱為「社會共享式的提取引發遺忘」(SS-RIF)。雖然WI-RIF和SS-RIF表面上有相似之處，但實驗數據兩者有不同的效果量和作用機制。本篇後設分析整理8篇、59個SS-RIF子實驗，發現SS-RIF實驗中會有近2倍大於WI-RIF的提取引發遺忘效果量；而SS-RIF中之所以有較大的遺忘效果量有很多心理因素如社交焦慮，但我們發現在社會提取練習階段的程序上的差異是調節效果量的主因。我們進一步將這些SS-RIF研究依有無自由交談的設計分組，發現有自由交談組相較無自由交談組大1.8倍的效果量；此外，我們發現在可自由交談組的SS-RIF研究中「遺忘效果」和「練習效果」間兩者呈現負相關，不同於WI-RIF研究中所發現的正相關。我們推論「可記憶偏差」可解釋上述可自由交談組的SS-RIF有較大的遺忘效果量以及負相關兩個現象；而我們統合回歸的結果支持這假說：SS-RIF有較大的遺忘效果量是因為「可記憶偏差」所致，而非刺激材料間更強的記憶抑制。

“Within-individual Retrieval-induced Forgetting” (WI-RIF) refers to the phenomenon that an individual’s selective retrieval of items often causes more forgetting of categorically related and yet unrehearsed items than categorically unrelated and unrehearsed items. This memory phenomenon has been also reported in social settings as “Socially-shared Retrieval-induced Forgetting” (SS-RIF). Despite their surface similarities, SS-RIF and WI-RIF studies show quantitatively and qualitatively different characteristics in experimental data. In our meta-analysis of eight SS-RIF articles with 59 studies, we found that the overall effect size of SS-RIF studies is almost twice as large as that of WI-RIF studies. Although the overall large effect size of SS-RIF studies might result from various psychosocial factors such as social anxiety, we found procedural differences during retrieval practice to be a primary factor that modulated effect sizes. Specifically, SS-RIF studies that employed a free-conversation procedure during retrieval practice showed a 1.8 times larger effect size than SS-RIF studies that did not, and they showed decreased forgetting effects as practice effects increased, which was opposite to the observed positive correlation between forgetting and practice effects in WI-RIF studies. Mechanistically, a hypothetical memorability bias can account for the large effect size and the negative correlation between forgetting and practice effects in the SS-RIF studies with free conversations. The results from our meta-regression lend support to this hypothesis and imply that the severe forgetting effects in SS-RIF studies may be due to such a memorability bias rather than strong inhibitions among memory items.