The Taxonomic Representation of Common Events: A Research Report

Shared cognitive representations have been shown in different conceptual domains. An experiment is presented here that extends this line of research to the domain of events. Specifically, this experiment shows a cultural consensus in the participants(cid:144) event taxonomies, and suggests a shared event taxonomy based on semantic relations. Showing such a shared event representation is an important first step for learning about the role of events in cognitive processes such as memory and thought. Las representaciones cognitivas compartidas han sido demostradas en diferentes dominios conceptuales. Se presenta aqu(cid:237) un experimento que extiende esta l(cid:237)nea de investigaci(cid:243)n al dominio de los eventos. Espec(cid:237)ficamente, este experimento demuestra un consenso cultural en las taxonom(cid:237)as de eventos de los participantes y sugiere una taxonom(cid:237)a de eventos compartida basada en relaciones semÆnticas. Demostrar dicha representaci(cid:243)n de eventos compartida es un importante primer paso para aprender sobre el papel de los eventos en procesos cognitivos como la memoria y el pensamiento.

Las representac iones c og nitiv as c ompartidas h an sido demostradas en dif erentes dominios c onc eptuales. S e presenta aq uí un ex perimento q ue ex tiende esta lí nea de inv estig ac ión al dominio de los ev entos. Espec í f ic amente, este ex perimento demuestra un c onsenso c ultural en las tax onomí as de ev entos de los partic ipantes y sug iere una tax onomí a de ev entos c ompartida b asada en relac iones semá ntic as. Demostrar dic h a representac ión de ev entos c ompartida es un importante primer paso para aprender sob re el papel de los ev entos en proc esos c og nitiv os c omo la memoria y el pensamiento.   Method P a rt ic ip a n t s a n d M a t eria l s Twelve students at the University of Hamburg were paid to tak e part in this ex periment, whic h was originally done in German. Thirty-eight typewritten c ards ( 3 . 5 x 7 c ms) were used to represent the events for the ex periment. The number of partic ipants and c ards is eq uivalent to L opez et al. ' s ( 1 9 9 7 ) animal tax onomies study.
Des ign a n d P ro c ed ure partic ipants, and they were ask ed to do a free sorting by building " groups of events that somehow go together. " The nature and number of groups and members was henc e up to t h e p a r t i c i p a n t s . T h e i r g r o u p s w e r e r e c o r d e d , a n d t h e partic ipants were ask ed to do an asc ending sorting by uniting the groups into " supergroups of events that somehow go together. " Their supergroups were rec orded, and the asc ending sorting was repeated until the partic ipants indic ated that no more supergroups were possible. Then, their initial groups w e r e r e s t o r e d , a n d t h e p a r t i c i p a n t s w e r e a s k e d t o d o a desc ending sorting by separating the groups into " subgroups of events that somehow go c loser together. " Their subgroups were rec orded, and the desc ending sorting was repeated until the partic ipants indic ated that no more subgroups were possible.
Res u l ts a n d Di s c u s s i on

193
This taxonomy has six levels. Level 0 shows each of the 38 separate events such as "brushing one's teeth", and level 5 shows them all united into the class of common daily-life events. At level 1, these 38 events are sorted into 17 specific groups that can be assigned approximate labels based on most of their members. They include, for example, a group of media events ("watching television" to "listening to the radio"). These groups are further united at higher levels. By level 4 , the 17 specific groups are sorted into just four subclass groups including, for example, a group of transport events ("driving the car" to "taking the bus").
A participant's event taxonomy may have up to 38 levels and 37 specific or subclass groups.
However, the actual scope was limited to few levels and groups. Most taxonomies (5) had just six levels (range: 6-8) and three subclass groups (range: 3-6), and half (6) had less than 22 specific groups (range: 16-28). Moreover, a taxonomy may have groups of events based on different kinds of relations. But the actual membership was constrained to semantic relations. Among others, most taxonomies (7) had the same group of transport events ("driving the car" to "taking the bus"), for example. These findings suggest that the participants' event taxonomies were not idiosyncratic but tended to consent in the number and nature of groups and members.

Cultural Consensus
A distance matrix was derived from each event taxonomy by calculating the distance among the events. This distance corresponds to the lowest taxonomic level at which two events go together.
For example, in the taxonomy above, "eating" and "drinking" go together at level 1, thus their distance is 1. Analogously, the distance between "eating" and "having breakfast" is 2, "eating" and "sleeping" is 3, "eating" and "vacuuming" is 4 , and "eating" and "reading" is 5. There is no distance between an event and itself, thus the distance between "eating" and "eating" is 0. Low distance corresponds to high relatedness. Thus, in the taxonomy above, "eating" and "drinking" are highly related events, "eating" and "sleeping" are moderately related events, and "eating" and "reading" are lowly related events.
A principal components factor analysis was performed on the correlated distance matrices to test for a possible cultural consensus in the participants' e v e n t t a x o n o m i e s . A c c o r d i n g t o t h e c u l t u r a l consensus model (Romney et al., 1986), a cultural consensus is indicated by a single factor solution in which (1) the first factor accounts for most of the variance, (2) the first factor eigenvalue is several times larger than the second, and (3) the first factor loadings are all positive. Results show that the factor analysis produced such a single factor solution.
The first factor accounted for 51% of the variance (the second for 8% ), the first factor eigenvalue was six times larger than the second (6.15 to .96), and the first factor loadings were all above . 50 That is, these findings suggest that there is an event taxonomy shared by the participants.

S h ared Taxonomy
A c l u s t e r a n a l y s i s w a s p e r f o r m e d o n t h e averaged distance matrices to derive the shared event taxonomy. Figure 2 depicts this taxonomy. The shared event taxonomy has six levels. Level 5 shows the general class of common daily-life events, and level 0 shows its specific members such as "writing." At level 4 , this class is divided in three subclass groups of roughly home ("doing the dishes" to "closing the door"), nonhome ("sitting at the computer" to "phoning") and emotion events ("getting angry"). These goups are further divided at lower levels. The group of home events includes, for example, groups of household ("doing the dishes" to "cooking"), wake-up ("getting dressed" to "sleeping") and food events ("eating" to "making coffee") at level 3; clean-up ("doing the dishes" to "doing the laundry") and hygiene events ("going to the bathroom" to "brushing one's teeth") at level 2; and cloth ("getting dressed," "getting undressed") and entrance events ("opening the door," "closing the door") at level 1. The group of nonhome events includes, for example, groups of work ("sitting at the computer" to "writing"), transport ("driving the car" to "riding the bicycle") and leisure events ("watching television" to "going for a walk") at level 3; and media events ("watching television" to "listening to the radio") at level 2.
The shared event taxonomy is also limited to six     De p e n d i n g o n t h e c o n t e x t , a n e v e n t c a n b e represented at any of those levels of abstraction.
However, these levels are also limited in number which suggests some pragmatic constrains on abstracting events.
The shared event representation also shows that people have event taxonomies based on semantic relations. For example, "driving the car," "taking the bus" and "riding the bicycle" are clustered together for being transport events. Depending on a specific cognitive task, however, an event could be linked to other events based on different relations. For example, "left the office," "drove the car," and "got home" could be linked chronologically in a memory task. This suggests that the taxonomic representation of events may be a conceptual source from which events could be sampled for further cognitive processes that would relate them in non-taxonomic ways.