Categorization of spatial relations:
The flexible infant and the lexical adult
Laraine McDonough
Brooklyn College
The City University of New York
Montag, 11.06.2001, 16 Uhr c.t., Hörsaal 9
Given that languages differ widely in the way spatial relations are
lexicalized, the theory that spatial terms are mapped onto preverbal notions
of space has come under scrutiny. If spatial relations are already formed
preverbally, then the "mapping" problem infants face between linguistic form
and meaning seems complex. What are preverbal categories like, such that
they can be used for learning many different languages, each making its
own
distinctions? The spatial categorization abilities of preverbal infants as
well as English-speaking and Korean-speaking adults were examined: In two
experiments, the distinction between tight-fitting containment and
loose-fitting support was tested. This contrast is made in many languages
(e.g., 'in' and 'on'). In two other experiments, a distinction between
tight- and loose-fitting containment relations was tested. In English,
containment is not distinguished in terms of the degree to which a smaller
object fits into a larger container; however, in Korean, the contrast
between loose- and- tight fit is a salient distinction which cross cuts the
English lexical categories of 'in' and 'on'. Participants were tested using
a nonverbal preferential looking paradigm in which the videotaped relations
were demonstrated with objects that varied in shape, texture and color. For
the adults, the experimenter also demonstrated four of the relations
previously seen in the preferential looking task. Adults were asked to
select which of the four did not fit in (an oddity task) and describe the
reason for their selection. The results show that before learning language,
infants are able to categorize both spatial relations. These findings
suggest a flexible categorization system which allows infants to analyze and
represent relations among varied objects. In contrast to the findings with
infants, the categorization demonstrated by adults corresponded to the way
spatial terms are lexicalized in their own language. For example, the
Korean-speaking but not the English-speaking adults categorized the
loose-and tight-fitting containment relations (i.e., 'kkita' and 'nehta').
The adult findings were consistent across both the preferential looking and
the oddity tasks. Although these findings suggest that the language we learn
influences how we initially construe events, the results do not support the
view that language constrains how events can be interpreted. After all,
words such as 'tight' and 'loose' are also part of the English language;
yet, both terms occur at much lower frequencies than 'in' and 'on'. Language
learning apparently highlights the way in which events are construed;
however, it is unlikely that language actually prunes away
preverbal concepts (such as proposed in research on phoneme discrimination).