First, we did preliminary analyses on verbal responses. We started by examining the number of words generated in the two conditions. We compared number of words produced in descriptions in the context of imperfective framing to the number of words produced in the context of perfective framing.3 There was no reliable difference (Perfective M = 41.15, SD = 16.78, Imperfective M = 39.09, SD = 22.64), £(130) = .59, p = .55. We then examined whether aspect would affect number of perfective and imperfective verb phrases generated. Participants in the perfective condition generated about the same number of perfective and imperfective phrases (Perfective M = 1.24, SD = .91, Imperfective M = 1.36, SD = 1.43), £(130) = 1.48, p = .14, and so did participants in the imperfective condition (Perfective M = 3.95, SD = 2.40, Imperfective M = 3.32, SD = 2.56), £(130) = -.58, p = .56. In sum, varying the aspectual form in the question did not result in notable differences in number of words or type of aspect produced in accident descriptions.
Second, we were interested in motion descriptions because they would serve as a good measure of how much action was conceptualized in a situation. We analyzed frequency of basic translational motion verbs, including drive, come, go, and £urn. Analysis here and elsewhere included finite and non-finite verbs as well as first and third person. An example of a description with the motion verb drive was, "I think a car was just driving". An example of a description the motion verb come was, "Another car came from the highway". As shown in Figure 1, participants who were asked to report what was happening (imperfective framing) produced proportionally more motion verbs in their descriptions (M = 2.32, SD = 1.38) than participants who were asked to report what happened (perfective framing) (M = 1.73, SD = .91), £(130) = -2.91, p = .004. In this case, aspectual framing resulted in reliable differences in participants' descriptions. Specifically, imperfective framing led to proportionally more motion verbs.
Figure 1. Imperfective framing resulted in more motion verbs per description (video) than perfective framing. (Error bars in this graph and elsewhere represent +/- 1 standard error around their respective means.)
Figure 2. Imperfective framing elicited fewer non-motion verbs per description (video) than perfective framing.
Third, we compared number of non-motion verbs in the two conditions. These included verbs that did not explicitly express motion, such as decide, call, think, and see. An example of decide was, "So a news lady decided to try and ride a scooter An example of call was, "The police officer called for ambulances" As shown in Figure 2, participants produced fewer non-motion verbs when asked to report what was happening (M = 3.95, SD = 2.95) than when asked to report what happened (M = 5.33, SD = 3.18), t(130) = 2.58, p = .01. These results show that aspectual framing differentially influenced the number of non-motion verbs that
Figure 3. Imperfective framing resulted in a greater number of reckless driving phrases per description (video) than perfective framing.
participants mentioned. In particular, imperfective aspect elicited fewer non-motion verbs than did perfective aspect.
Fourth, we analyzed mentions of reckless driving. Phrases were coded as reckless if they suggested dangerous driving. Examples include: "The truck was speeding", "He tried to cut off the car next to him", and "She was swerving". As shown in Figure 3, participants produced more reckless driving phrases with imperfective framing (M = 3.26, SD = 3.97) than with perfective framing (M = 1.78, SD = 2.05), £(130) = -2.69, p < .008. Once again, aspectual framing had an effect. In this case, imperfective framing biased people to focus more on reckless details of driving.
Based on our verbal data, we see that aspectual framing influenced our participants' descriptions of accidents in systematic, predictable ways. Individuals who were asked to describe what was happening (imperfective framing) generated more motion verbs and reckless driving phrases, but fewer non-motion verbs than did individuals who were asked to describe what happened (perfective framing). Importantly, there was no difference in the number of words produced overall, or in the type of aspectual form produced in the two conditions, suggesting that the aspectual framing influenced semantic content, not lexical quantity.
First, we compared number of gestures produced in the two conditions. Participants in the perfective framing condition produced about the same number of gestures as participants in the imperfective framing condition (Perfective M = 3.06, SD = 3.99, Imperfective M = 3.74, SD = 3.70), £(130) = -1.02, p = .31. No significant difference was observed.
Next, we compared number of iconic gestures generated by participants. A gesture was coded as iconic if it had semantic content, and depicted one of the following: shape of an object (e.g. two hands next to each other to show two cars side by side), shape of a path of motion (e.g. show a circular motion to show somebody spinning out), or shape of an event outcome (e.g. raise hands and arms to show an explosion). As shown in Figure 4, participants articulated more iconic gestures with imperfective framing (M = 2.65, SD = 2.63) than with perfective framing (M = 1.14, SD = 1.76), £(130) = -3.88, p < .001.
Figure 4. Imperfective framing resulted in more iconic gestures per description (video) than perfective framing.
We were also interested in how aspectual framing would influence the production of beat gestures. A hand movement was coded as a beat gesture if it carried no obvious semantic meaning, for instance, flicking the hand when stating, "Okay, in the video.. " Participants produced fewer beat gestures in the imperfective condition (M = 1.08, SD = 1.69) than in the perfective condition (M = 1.91, SD = 2.96), £(130) = 1.99, p < .05, as shown in Figure 5.
The gesture results are in line with the verbal results. They show that aspectual framing systematically influenced the way participants gestured while describing
Figure 5. Imperfective framing resulted in fewer beat gestures per description (video) than perfective framing.
accidents. Individuals responding to imperfective questions produced proportionally more iconic gestures and fewer beat gestures than did individuals responding to perfective questions. No difference was observed in the average number of gestures in the two conditions, suggesting that aspectual framing had an effect on type and form of gesture, not quantity.
-  The first and third authors coded the reckless driving phrases independently, and agreed 92 percent of the time. Discrepancies were resolved by using half the first author's codings, and half the third author's codings.
-  The second and fourth authors independently coded all gestures by type, and were in agreement about 90 percent of the time. Discrepancies were resolved by judgment of the first author.
-  Two gestures in the data set were neither iconic nor beat. They occurred when participants pointed at the computer screen. These were not analyzed because they accounted for less than 1 percent of the data.