Memory & Cognition 1982,Vol. 10(4), 308-317The role of arousal in memory for conversationBRIAN MacWHINNEYCarnegie-Mellon University, Pittsburgh, Pennsylvania 15213andJANICE M. KEENAN and PETER REINKEUniversity of Denver, Denver, Colorado 80210In the first experimental study of memory for natural conversation, Keenan, MacWhinney,and Mayhew (1977) found that, even after 30 h, subjects had extremely good recognitionmemory for the exact wording of statements that contained information about a speaker's"intentions, beliefs, and his relations with the listener." Such sentences were said to be high in"interactional content." One possible interpretation of the results of Keenan et al. is that it isthe immediate affective response to an utterance, rather than its interactional content, thatincreases its memorability. In the present study, the strong relationship between interactionalcontent and memory found by Keenan et al. was replicated. Subjects showed excellent recogni-tion memory for high interactional content statements from a conversation, even after a 72-hinterval. However, there was very little relationship between arousal, as determined by subjects'electrodermal response (EDR), and subsequent memory. Moreover, involvement had its great-est effect not on memory, but on subjects' arousal as measured by EDR.Keenan, MacWhinney, and Mayhew (1977) conductedthe first experimental study of memory for natural con-versation. Their study was motivated by the observationthat sentences spoken in natural conversation conveytwo fundamentally different types of information: pro-positional information and pragmatic information. Pro-positional information was characterized as informationabout the objects and events referred to by the sentence.Pragmatic information was defined as information aboutthe context of the sentence in the social interaction,including information about the speaker and "his inten-tions, beliefs, and his relations with the listener." Sen-tences that conveyed a great deal of pragmatic informa-tion referencing the speaker-listener interaction weresaid to be rich in "interactional content." Keenan et al.hypothesized that sentences with high interactional con-tent would be highly memorable because "subsequentinteractions with the speaker often depend upon thecontents of earlier interactionally significant events"(p. 551). More important, they predicted that listenerswould tend to remember the exact surface wording ofThis research was partially supported by National Institute ofEducation Grant NIE-G-78-0173 to J. Keenan. It is based in parton a master's thesis submitted by Peter Reinke to the PsychologyDepartment of the University of Denver. Joseph Campos gra-ciously made available to us the use of his laboratory, facilities,and equipment. He also provided us with valuable technical assis-tance and advice regarding the psychophysiological measurements.Without his help, the study could not have been conducted. Wealso wish to thank Suzanne Gendreau for testing the subjects inthe control study and Polly Brown for conducting the statisticalanalyses of the data in the control study. Requests for reprintsshould be sent to Brian MacWhinney, Psychology Department,Carnegie-Mellon University, Pittsburgh, Pennsylvania 15213.high-interaction sentences because it is often the par-ticular choice of words made by the speaker that is thecarrier of interactional information.To test these predictions, Keenan et al. (1977) tape-recorded a lecture and conversation at a luncheonresearch group meeting in the Psychology Department atthe University of Denver. In this meeting, BrianMacWhinney discussed a variety of issues in psycho-linguistics with a contentious and sometimes boisterousaudience made up of both students and faculty. Theresults of a recognition memory test composed ofsentences taken from that meeting showed that sen-tences with high interactional content were rememberedtwice as well as sentences with low interactional content.Furthermore, the surface form, that is, the exact word-ing used by the original speaker, was remembered threetimes better for high-interaction sentences than for low.Keenan et al. (1977) argued that subjects remem-bered the surface forms of high-interaction statements sowell because these surface forms carry information thatis relevant to the subsequent dynamics of the speaker-listener relationship. Such surface devices as hesitationplacement, lexical choice, intonation contour, para-linguistic markers, and constituent ordering all providea rich data base from which inferences regarding thespeaker-listener relationship can be drawn. By attendingto the surface form and maintaining this information inmemory, at least until the next encounter with thespeaker, the listener can more accurately assess his/hersociolinguistic status vis-a-vis the speaker and his/herinteractional commitments toward the speaker.An alternative interpretation of the results of Keenanet al. (1977) is that it is not so much that pragmatic con- Copyright 1982 Psychonomic Society, Inc.3080090-502X/82/040308-10$01.25/0THE ROLE OF AROUSAL IN MEMORY 309tent makes sentences memorable as that high pragmaticcontent induces a high arousal level that, in turn, pro-duces good memory. According to this interpretation,high-interaction statements elicit higher levels of atten-tion and arousal at encoding than low-interaction state-ments do. Because the results of list-learning experimentsshow that higher levels of arousal at encoding lead tohigher levels of memory performance on tests of long-term retention (see Craik & Blankstein, 1975, for areview), it follows that high-interaction statementswould be more memorable than low-interaction state-ments. We refer to this interpretation as the arousaleffect.There are good a priori reasons to expect the arousaleffect to be as strong as, or stronger than, the inter-actional content effect. For example, Zajonc (1980)argued that human beings can often generate directaffective responses to stimuli even before those stimulihave been categorized and evaluated cognitively.