The study used a randomized trial in a 3 (condition: individual control, partner with- encouragement, partner-without-encouragement) x 2 (performance block: Block 1 & Block 2) factorial design. Eligible participants were students who had no physical injuries that would prevent or obstruct their performance during an isometric plank exercise.

The study was conducted in a laboratory on the campus of Michigan State University in the Departments of Kinesiology and Psychology and was approved by the Institutional Review Board. Most of the data for two of the conditions (ie, 26 participants in the individual condition and 47 in the partner-without-encouragement condition) were collected as part of a study [17] completed about the time we decided to examine the effect of verbal encouragement (August 2009 to May 2010). In a new wave of data collection (August-December, 2010), we not only collected data for the partner-with-encouragement condition, but also some additional data in the two other conditions. By contrasting the latter with those collected for the aforementioned study [17], we could probe for possible history or cohort effects. We did not expect any systematic differences between the two waves of data collection because the lab settings, participant populations, and procedures for both data collection periods were identical, except for the introduction of the manipulation of verbal encouragement.

The task for this study was an abdominal plank exercise. These are body-weight resistance exercises where participants suspend their own body weight using their abdominal muscles. These exercises are also isometric in nature, require very little coordination, and are highly effort-based. Each exercise targeted the abdominal muscles, but there were slight differences between each (eg, holding a push-up position on one’s forearms vs on each side; a detailed description of all exercises is provided elsewhere [12]).

The exercises were performed as part of an exercise video game designed for the PlayStation 2 (PS2) gaming system as used in a previous study [12]. The software used was EyeToy: Kinetic, a game that offers a variety of fitness activities (eg, yoga, strengthening exercises, combat exercises). This particular software operates in conjunction with an additional accessory called the Eye Toy, designed specifically for the PS2 system. The Eye Toy is a small camera that connects to the PS2 system via a USB cable and allows images of the user to be displayed on the TV monitor and interact with objects in a virtual environment supported by the software.

A detailed description of experimental procedures is provided elsewhere [12]. In the current study, we simply describe the basic procedure and note how the encouragement manipulation was achieved.

After arriving at the laboratory, participant consent was obtained and all were ensured of the confidentiality and voluntary nature of their participation. Participants then watched a brief instructional video from the PS2-Eye Toy: Kinetic software in which a virtual trainer demonstrated the five exercises. A baseline measure of self-efficacy was then taken using an online questionnaire completed in the laboratory. Participants were asked throughout the session if they understood all instructions and, if not, the experimenter addressed his/her questions. All participants then performed the first block of exercises, holding each of the five exercises for as long as they could and with 30s rest periods between each exercise. Immediately after each exercise, the participant announced his/her perceived exertion. All participants were given veridical feedback on their performance (ie, the average of the number of seconds they held each exercise).

The condition manipulation was introduced at this point. Participants in the individual control conditions simply rested for 10 minutes. Participants in the partner conditions were told that another participant was being run simultaneously at another lab on campus and that the 2 participants would be able to see one another over an Internet video connection during future trials. The participants then met briefly with that other same-sex participant in a controlled Skype-like interaction (we will refer to that other participant hereafter as “the partner”). In reality, the partner was an experimental confederate whose side of the interaction was prerecorded. After the interaction, participants were also given bogus feedback on how well the partner had done on the first trial. That feedback score was manipulated to be 1.4 times the participant’s own actual performance. This discrepancy was chosen based on previous research that suggests that this moderate discrepancy leads to the greatest increases in exercise duration (ie, the greatest Köhler effect) [18,19].

Participants were told that they and their partners would be a 2-person exercise team. In the encouragement condition, both teammates were told that they would have the opportunity to communicate with each other during the exercises but that, due to technical problems, this capability would be provided only for the partner. Thus, ostensibly both could speak, but only the partner would be heard by the subject. No mention was made of any audio link between partners in the no-encouragement condition. For both partner conditions, it was further explained that they were working towards a team score, where the team score would be the persistence score of the first teammate to quit an exercise (ie, as soon as either partner quit, the exercise was over). This task structure is more commonly known as a conjunctive task—when the group’s performance is defined by the performance of the least-capable member (ie, the “weak link”). Following these instructions, all participants responded again to the self-efficacy measure.

Block 2 then commenced. In the individual control condition, the participants could only see him/herself on the screen, as during Block 1. In the partner conditions, the participant could see the partner’s image (which was actually prerecorded) before and during the exercise; the participant believed that the partner could likewise see his/her (the participant’s) image. The images available to the participant suggested that s/he was always the first to quit each exercise. The video link was allegedly frozen as soon as either teammate quit an exercise and until just before the start of the next exercise; hence, the participant knew only that his/her partner had been able to persist longer, but not just how much longer. In the encouragement condition, a prerecorded series of phrases of encouragement was played through a set of computer speakers controlled by the experimenter. The phrases were audible approximately every 15s (±3s) and followed the following fixed progression: “you can do it”, “you got this”, “keep it going”, “you’re doing good”, “stay strong here”, “give it your best”. After Block 2 was over, the participant completed a series of questionnaires (self-efficacy, intention to exercise, enjoyment of physical activity, and manipulation checks). S/he was then debriefed, thanked, and excused.

Following the methods of analyses done in earlier studies, exercise duration was taken as the sum of the time each participant held the five exercises within each block, producing a Block 1 and Block 2 score. There are, of course, individual differences in fitness, intrinsic task interest, and strength that we wished to control for. This can be done in different ways. In many prior studies [14,25], participants’ Block 1 performance has been used as a baseline and difference scores (ie, Block 2-Block 1) were the primary dependent variable, one that expressed each participant’s Block 2 score relative to his/her Block 1 score (the latter baseline score reflected individual differences in fitness and strength). An alternative (less vulnerable to certain problems that can arise from the use of difference scores, eg, [26]) is to use Block 1 scores as a covariate in the analysis of Block 2 scores. Here, we present the results using the former difference-score method because the mean values presented for such an analysis are more directly understood and interpreted than means adjusted for a covariate. But the reader should note that an identical pattern of results is obtained with either method.

The analyses of the exercise duration data was to proceed in two stages. The first was to check whether there were any history or cohort effects attributable to the interval between the two data collection waves. It employed a 2 (Condition: Individual vs partner-without-encouragement) x 2 (Sex) x 2 (Data collection wave: Early vs Late) ANOVA on exercise duration difference scores (ie, Trial 2 duration-Trial 1). Although overall sex differences in the magnitude of the Köhler effect are rare, some interesting sex effects have been reported, eg, [25]; hence, participant sex was included as an experimental factor. This stage 1 analysis would permit both confirming with the larger dataset that the Köhler effect originally reported in Kerr et al [17] and checking whether its magnitude differed between the participants drawn from Kerr et al [17] and those newly recruited in the same experimental conditions for the present study. The second stage was designed to examine the primary question addressed by this paper—would verbal encouragement to the weaker partner alter the Köhler effect? If there were no data-wave moderation effects, as anticipated, we planned to drop the data-wave factor and add the no-encouragement condition in a 3 (Condition: Individuals, Partner-without-encouragement, Partner-with-encouragement) x 2 (Sex) ANOVA on exercise duration difference scores.

Although our primary focus was on exercise duration, we also examined several other variables, primarily to determine if encouragement altered any of the effects previously observed [12,17] when no encouragement was provided. For those variables collected at the end of the study (ie, task effort, intent to exercise, own task ability), the analyses would employ 3 (Condition) x 2 (Sex) analyses of variance; since the individual participants had no partner, this became a 2 (Condition: Encouragement vs No-encouragement) x 2 (Sex) ANOVA for ratings of own ability relative to one’s partner. For those variables (ie, perceived effort, self-efficacy) collected during the exercise trials, a within-Ss Block factor was added to the ANOVA. Finally, a covariate (ie, the pre-exercise estimate of task self-efficacy) was included in the analysis of self-efficacy.

The initial stage 1 analyses looked for possible history or cohort effects attributable to the interval between the two data collection waves. A boost in exercise duration (ie, the Köhler effect) was evident (Condition main effect F _1,79=39.53, P<.001, partial eta squared=.33); whereas due to fatigue, Individual condition participants persisted 39.4s less at Block 2 than at Block 1, and participants in the Partner-without-encouragement condition persisted 61.0s longer, a Köhler effect of 100.4s. More importantly, there were no significant effects involving the wave factor. Thus, the magnitude of the Köhler effect in the absence of partner encouragement was comparable within each wave of data collection. Hence, the data for these two waves were combined in all subsequent analyses.

In the primary stage two analyses, the data for the Partner-with-encouragement condition was included to see whether and how adding verbal encouragement might affect the duration-boosting effect. The 3 (Condition: Individuals, Partner-without-encouragement, Partner-with-encouragement) x 2 (Sex) ANOVA on difference scores resulted in only one significant effect: the Condition main effect, F _2,105=28.79, P<.001, partial eta squared=.35. Again (see Figure 2), the Individuals showed a decline in exercise duration across blocks (estimated marginal mean of -42.2s, SD 54.2), whereas partners exercised longer in the second block, whether there was (19.49s, SD 59.30s) or was not encouragement (53.63s, SD 58.44). Post hoc Dunette tests indicated that the difference between Individuals and Partners was significant (P<.001) in both partner conditions. A Newman-Keuls post hoc test indicated that this effect was significantly smaller (34.1s) in the Partner-with-encouragement condition.

The current study corroborates a growing body of research, which shows that exercising with a moderately superior Internet partner as the weak link can boost effort and lead to longer bouts of exercise [12,13], whether or not partners communicate with one another. However, such effort-boosting effects can be mitigated when superior partners try to encourage weaker group members, especially if this encouragement undermines the weaker member’s perception of the superior ability of his/her partner. Future research should examine the effects of a wider range of messages, exercise tasks, and conditions to help inform the design of group-based, electronically mediated physical activity interventions.