Testing Jakobson's Ｒegression Hypothesis has been one of the hot issues in the field of language attrition since language attrition was researched thirty years ago． However， the target items tested and validated have been at either morphological level or syntactic level， none at lexical level． This paper， trespassing the boundary， set out to look into the process of the foreign lexical learning and attrition among 116 Chinese college graduates by measuring their English receptive vocabulary， and this dynamic process was found to be working within a model． Specific conclusions are: ( 1) foreign lexis regresses with time in attrition just as morphological and syntactic items do， which is expected by Jakobson's Ｒegression Hypothesis; ( 2) the regression of foreign lexis in attrition can be described as“first learned， last out” ; and ( 3) the Theoretical Model for Foreign Lexical Learning and Attrition， adapted mainly from BIA-d Model， is able to illustrate both the process of foreign lexical learning and the regression of foreign lexical attrition．