Previous social exclusion experiments identified two factors affecting the participants’ evaluation of participation in a virtual ball tossing game (cyberball): ball reception probability and vertical position of the participant’s avatar on the screen. The P3 component in the event-related brain potentials (ERPs) indicated that both factors moderate subjective expectancies on social participation. The present research builds on an expectancy model explaining these effects and tests whether its predictions – established in a within-participant design – also hold in a between-participant design more common in behavioral cyberball studies. Participants were randomly assigned to four conditions which differed in ball reception probability (16% vs. 26%) and the avatar’s vertical position (inferior vs. superior). To track the state of expectancy of involvement online, we recorded the ERP response evoked by ball receptions of the participant. Retrospectively, social involvement and social need threat were rated in a questionnaire. As hypothesized, low ball reception probability elicited enlarged P3 amplitudes in the ERPs, increased negative mood, and threatened social needs. For participants at inferior position, ERP and questionnaire effects were less expressed. This effect of verticality can be traced back to an adjustment in the expected involvement as signaled by a differential adaptation of the P3 amplitude within an experimental run. These results confirm that the predictions of an expectancy model also apply to cyberball studies using a between-participant design. However, the comparison with the results of previous within-participant design studies suggests that the sensitivity of the adjustment processes critically depends on the choice of the experimental design.