Melissa Bowerman and Stephen C. Levinson (2001) Language Acquisition and
Conceptual Development (Language, culture and cognition 3), Cambridge
University Press. ISBN 0-521-59358-1 hardback, ISBN 0-521-59659-1 paperback,
Eva Bar-Shalom, University of Connecticut
The introduction by the editors of the book summarizes the major themes of
the volume. As M. Bowerman and S. Levinson point out, the goal of the book is
to try to reconcile what has recently been learned about the development of
language and conceptual knowledge in young children. There is evidence that
very young children possess rich conceptual knowledge, and there is a
question of whether some of it may be innate. In addition to showing very
early conceptual development, other papers in the volume concentrate on
language specific "semantic" differences that are also acquired early on by
young children. The question arises whether children interpret the world
around them with influence from their native languages (The Sapir-Whorf
hypothesis). Since according to some, very young children possess rich
cognitive abilities before language could have appeared, how much of the
child's cognitive knowledge is shaped by a particular language?
This book contains 19 papers, not including the introduction by the editors,
and the papers are grouped into four parts. As Laura Wagner indicated in her
review of the same book, since the book consists of so many papers, each
providing so many important and rich research results, it is hard to do
justice to the entire contribution and significance of each paper.
Part I: Foundational issues.
1. J. Langer "The mosaic evolution of cognitive and linguistic ontogeny"
The author asserts that "there can be no very intimate interaction between
language and cognition in early ontogenesis - cognition leads" (p.19). The
research Langer discusses relates to both human and non-human primates. Among
several cognitive abilities both in human and non-human primates, Langer
discusses the onset of physical cognition as being similar in both human
infants and macaques and Gorilla gorilla, among others (Parker 1977; Potti
1989, Spinozzi 1989, Redshaw 1978, etc.). According to Langer, developmental
studies show that "second-order cognitive development" is well in place
before the second half of the second year, when "when linguistic production
begins to develop some power" (p.37). Langer further considers second-order
cognition to be a necessary prerequisite for developing language in humans.
In Langer (1996), he also considers the stage of second order cognition to be
a prerequisite for learning "protogrammatical language" in chimps (p.39). A
revolutionary idea is also being researched currently by specialists in
primate communication abilities is whether those chimps that have been
trained in language (e.g., Savage-Rumbaugh 1998) can achieve a higher
cognitive level of development ("third order logical classifying, p. 40). The
results so far are negative, and Langer reports that no differences have been
found in chimps trained in language and those that were not trained up to 10-
11 years of age. (For a comprehensive summary of language experiments with
non-human primates, see Wallman 1992).
2. A. Gopnik "Theories, language and culture"
Gopnik supports the Whorfian hypothesis as well as proposes a new account of
a relationship between language and thought. She mentions "theory theory"
which compares cognitive development to the development of the theory of
science, i.e., seeing a child as "a little scientist". There is a very
interesting and detailed explanation of those views in Gopnik's chapter.
Gopnik proposes using cross-linguistic studies for testing the
"interactionist hypothesis, according to which language may influence and
restructure cognition" (p. 56). She draws on studies of Japanese, Korean and
English adult and child languages. First, she describes the relevant
differences between the three languages, in particular, between English on
one hand, and Korean and Japanese on the other hand. The differences have to
do with richness of morphology in Korean and Japanese, not English, with
massive ellipsis in Korean and Japanese and not English (Clancy 1985).
Parental speech In Korean and Japanese, unlike in English, consists of many
highly inflected verbs and a few nouns. There is evidence that the same
difference is reflected in early speech of the above language groups.
Researchers found that Korean and Japanese children acquire verbal morphology
earlier than their English-speaking counterparts, but use fewer and less
varied names. The question arises whether Korean and Japanese children are
more advanced in understanding of action concepts encoded by verbs, and less
advanced in concepts encoded by nouns (at an important age 15- 21 months).
Gopnik and Choi (1990) showed in a longitudinal study that Korean-speaking
children were delayed in comparison to English speaking children in
"emergence of naming explosion and the development of exhaustive
categorization" (p.97). The analyses of further experiments and also the
analysis of input speech lead Gopnik to conclude that the parent input is in
some way responsible for the observed difference between the language groups
in their cognitive development. The exact aspects of the input, however,
As for the Whorfian hypothesis, Gopnik considers it too strong to explain the
possible effect of language on cognition, since children in these languages
eventually converge in their understanding of concepts expressed both by
verbs and nouns. The difference may be in the route or timing or
understanding these concepts.
Applying the "theory theory" explanation, it is possible to assume that
children first acquire certain concepts from the observation of the
environment only. When language appears, children acquire a "new tool", which
is a "double-edged sword" (p. 59). Among the reasons for this
characterization of language is the fact that children may receive different
linguistic information from different adults about the same concepts. This
refers to contextual variation. They also may receive more or less linguistic
information about the same category in different languages, such as verbs in
Korean and nouns in English. There are some other differences that Gopnik
examines among the linguistic information children receive in two languages
that may have a "cognitive" influence; however, we are not able to discuss it
here. On another note, the author indicates that the discussion was limited
to lexical items, and that syntactic development may increase the effect of
the language in the theory formation by the "little scientist'" - the child.
We will not address the further discussion in this paper, but the topics come
up in other papers this volume, such as the Whorfian effect in adulthood in
speakers of various languages.
3. E.S. Spelke and S. Tsivkin "Initial knowledge and conceptual change: space
In this paper a new answer is proposed to the question of "how humans build
rich and intricate systems of knowledge that are characteristic of our
species?" (p.70). The authors suggest that even though humans are endowed
with a rich set of knowledge, the systems that encode the knowledge are very
limited. Thc authors propose that the initial systems of knowledge are
modular. e.g, informationally encapsulated, task-specific, etc. Conceptual
development consists of conjoining these various domains of knowledge
systems. In fact, the ability to efficiently conjoin these modules is what
makes humans unique. As the authors point out, it is not clear how these
modules are conjoined, but it is suggested that natural language makes it
possible for these systems to be conjoined. Language then allows a richer
knowledge of concepts.
The first part of the paper concentrates on spatial representation in humans,
as well as development of this ability. The authors very correctly point out
how inferior humans are in comparison to other species in the spatial
orientation. S&T also explore spatial REorientation abilities in rats and
young children, and suggest that young children may be more flexible at it
They conduct a number of experiments with 18-24 months toddlers by adapting
Cheng's 1996 task. The studies show toddlers are unable to reorient
themselves using geometrical or color landmarks of a room in which a search
for a hidden toy took place. In another experiment, with different
conditions, they did better.
To examine developmental changes in spatial (re)orientation, Hermer and
Spelke 1994 and Hermer 1997 tested adults, and 3-7 year/old children using
the same technique. Adult subjects were asked questions as to why they
performed the object search in a certain way. The answers the subjects gave
together with the performance of 6-7 year-old children led the authors to
conclude that "developmental changes in object localization were roughly
correlated with changes in spatial language" (p.78). Some of the spatial
language used was terms like at X or in X or "to the left" or "to the right".
Further evidence for the effect of language is on reorientation is provided
by the Hermer-Vasquez's experiment, the results of which are can be found in
this paper, but are not reported here for lack of space. The main point of
the section, however, is that spatial language allows the child to represent
positions of hidden object in new ways (a Whorfian effect). In section 3 of
the paper, there is a chapter "Number", and a suggestion that prelinguistic
infant can represent numbers, e.g., noticing a difference in a number of
dots, in two different sets.
Part II: Constraints on word learning
4. L. Smith "How domain-general processes may create domain-specific biases?"
In the section on novel-word generalization, Smith reports studies (e.g.,
Landau et al. 1988) that show that kids can recognize words based on the same
shape, regardless of the size and texture, and after they see the object only
once. In a non-naming task, Landau et al. asked 2-3 year-old children to
judge whether the objects were similar . In this task children "based their
judgements on the wholistic similarity of the test object to the exemplar".
Thus, they did not take shape as the indicator of the difference. According
to Smith, this and other experiments reported and this paper show, that
children do not simply attend to one property, such as shape in naming an
object. As seen in further experiments, there is evidence that the qualities
children attend to in naming tasks "are exquisitely tailored to the specific
properties of the named object and linguistic context "(p.104). In addition,
these strategies change with age. Smith ( 1995) "proposes that the shape bias
may be the first among learned attentional biases because ".the correlation
between shape similarity and lexical category is broadest and most general in
language to children" (p.114). On the broader theoretical level, Smith agrees
with other theorists with the fact that children make domain-specific
generalization in their lexical learning process, but she argues against any
sophisticated mechanisms at work in developing domain- specificity.
5. M. Tomasello "Perceiving intentions and learning words in the second year
The author concludes that cognitive foundations for language acquisition go
hand in hand with learning to conceptualize the world the same way adults do.
In this process children become increasingly aware of "...adults'
communicative intentions, in particular, communicative circumstances" (p.
155). Tomasello argues against the "constraints" approach, e.g, syntactic or
semantic "bootstrapping", according to which young children entertain only a
limited number of hypotheses about how word "mapping" take place into the
real world. Among various studies, in Tomasello and Barton (1994) 24 month-
old children learned meanings of novel words (both nouns and verbs) in both
production and comprehension. The only cues children had were the
demonstrations of adult's intentions in searching for an object and emotional
reaction to the success in finding the object or performing the intended
action in the case of verbs. There is also, of course, an explanation in
determining by the child of what action the adult intended to perform and
what the adult intended for the child to do. There is also a summary of
studies in this paper addressing the information on the part of the adult in
letting children know what the adult's intention was in various tasks (p.149,
table 1). In section 3, Tomasello discusses the contribution of language to
the bidirectional communication of the intention between child and adult. He
also states the according to the social-pragmatic approach, children are
biased to conceptualize the world in an adult way, but ".it is the connection
of conceptualizations to language that must be learned in communicative
interactions with others" (p.153).
6. P. Bloom's "Roots of word learning"
In this paper, the author argues against innate constraints of word learning
as being a part of a specialized language acquisition faculty. Thus, he
considers lexical learning different from syntactic, phonological and
morphological skills that have and underlying neural basis, as the effects of
the "critical period" show. As opposed to these other skills, word learning,
however, is part of the child's more general cognitive capacities (p.160).
After reviewing other theories of word learning, Bloom proposes his
own. Here I will concentrate on the findings by Bloom & Markson 1996, 1998).
The results of these studies suggest the following:
A. Intent of what an object is supposed to represent is understood well by
four-year-olds. For example, in the presence of an object, if a child sees a
picture that could represent that object or a similar one, s/he will name the
object that is physically present. B. In addition three-four year old
children ".individuate their own picture in terms of intent" (p.174). C.
Three-four year old children can productively name arbitrary shapes. In one
experiment, children were shown two "blobs" of different sizes, while told
that one is a "flower" and one is a "house". The child was told that the
shapes looked "funny" because the person drawing the pictures had a broken
arm. It was also explained to the child that the person intended to draw
specific objects. The child reasonably hypothesized that the larger "blob"
was a "house", and the smaller one was a "flower".
These results seem to illustrate a child's rich conceptual knowledge, as well
as ability to understand intent and apply her own intent in deducing the
meaning of words.
Part III: Entities, individuation and quantification
7. S. Carey "Whorf versus continuity theorists: bringing data to bear on the
In the beginning of her paper, Carey states her support for the Whorfian
view, according to which the "language we speak both reflects and shapes our
conceptualization of the world". Additionally, according to the Whorfian
hypothesis there is conceptual change with cognitive development. Carey
reviews the premise of "continuity" theory according to which cognitive
structures remains the same throughout development, so that infants
prelinguistic representations draw on the same vocabulary as later
representations. "The universalist" approach assumes that all languages can
grammaticize the same set of concepts, but choose a subset of them. Extreme
Whorfianism, as Carey indicates, differs from continuity in that it assumes
that each child masters cultural and grammaticized categories anew in each
language. In this chapter Carey indicates what empirical data is needed to
discriminate among the various hypotheses. She also describes several
experiments testing infant's pre-linguistic understanding of various
concepts. For lack of space, I will just mention some of the points from the
"conclusion" section of this paper. According to Carey, in the very early
stage, infants have criteria for "individuation and numerical identity (the
sortal "object"; more specific sortals like "cup, "book") quantifiers such as
"one" and "another", the distinction between individuated entities and non-
individuated entities...etc." (p.210). Hauser and Carey (1998) argue that
these cognitive concepts have been with us for a long time and are not at all
language-specific concepts, but rather, conceptual primitives.
8. D. Gertner and L. Boroditsky "Individuation relativity, and early word
The authors take the "middle of the road position" on the cognitive and
versus linguistic dominance in the acquisition of concepts. On p. 216,
fig.8.1, they provide a Division of Dominance continuum of various word
classes. On one extreme are proper names, concrete nouns, on the other
extreme we see determiners and conjunctions, which presumably, would not have
a meaning of their own without language. Spatial prepositions and certain
"concrete" verbs such as "skate" and "enter" are in between. In terms of the
child's task in cognitive and linguistic development is to attach symbols to
concepts already understood by them from experience, such as concrete objects
and entities. However, verbs and items on the other extreme of the continuum,
are best understood through language. Thus, the particular classes of verbs
and "closed" class items would be learned later than nouns. There is
empirical evidence cited by the authors from various studies (e.g., Gilette
et al. 1998; Gleitman and Gleitman 1992; Gleitman 1994; Fisher et al. 1994)
that show that verbs are more difficult for young children than nouns. The
authors also discuss the opposing view: the controversy about nouns being
easier to acquire than verbs by young children and cross-linguistic
differences in this respect. As support for their view, G&B discuss
"relational relativity" and give examples showing that young children would
not be able to learn the meaning of some verbs with only perceptual
experience. They draw on the examples from Talmy (1985) showing cross-
linguistic differences in lexicalization of events in English and French.
Consider these two examples:
a. The duck floated past the tree
b. Le canard passe l'abre en flotant
As Talmy pointed out, there is a difference in the expression of the semantic
components between the verb and "satellites". G&B suggest that without
knowing the language-specific difference (via language), including the
relationship between verb and noun, it would not be possible to learn the
concepts in the examples above. However, some entities, (e.g., concrete
nouns) can be acquired without linguistic input from experience alone. In the
conclusion of the paper, the authors summarize the complexity of the
relationship between language and cognition. At one end of the Division of
Dominance continuum, cognition is dominant and language is subordinate to
cognitive- perceptual concepts. At the other end of the continuum,". language
influences our semantic categories" (p. 248).
9. J. Lucy and S. Gaskins "Grammatical categories and the development of
classification preferences: a comparative approach".
The authors emphasize the importance of the comparative approach in the study
of the relationship between language and thought and outline the research
criteria for successful evaluation of this relationship. The authors take
Yukatek Mayan and English as a source of comparison. Some of the linguistic
differences between the two languages are the following: Yukatek, unlike
English, optionally marks only a small number types of nouns for plural.
Also, unlike English, Yukatek uses numeral classifiers, which provide
information about properties of materials, e.g., shape ('un-tz'iit kib' - one
long thin candle, (p. 260)). The word "kib" is translated into English as
"wax " or things made out of wax, including "candles". Thus, there are no
separate English transitional equivalent for "wax" and "candle".
Interestingly enough, English also has a functional equivalent of
classifiers", such as "one cube of sugar", "one clump of dirt". These English
"functional" classifiers also describe the shape of the object. Thus,
according to the authors, shape could be more perceptually salient in
English, whereas material is predicted to be more perceptually salient in
Yukatek. These assumptions were tested in an experiment from both languages
(Lucy 1992b). Each subject was shown a "pivot" and then two more objects, one
of which resembled the pivot object by shape and the other one by material.
The subjects were given a task that would make them pick the object that they
would think most resembled the pivot. The results show a significant
difference between the two groups of speakers in the direction predicted by
L&S. To avoid the possible confounds in this study, several additional
studies were conducted by the authors. The results of the studies do not show
the same absolute differences among the English and Yukatek group, but do
show a relative tendency in the same direction as the previous study (the
experiments are discussed on pp. 263-273). The authors also discuss
developmental patterns. The pattern creates some puzzles. Children in both
language groups initially seem to favor shape as the basis for
classification. At the age of nine, they both seem to favor material as their
basis. The details and summary of the experiments are provided on pp. 273-
278. What is puzzling from the developmental perspective is the similarity of
the two groups and later reorganization of preference in classification in
middle childhood. However, some researchers think that there is also a
reorganization of grammar along adult lines at the same time, including
presuppositional and discourse factors (C. Chomsky 1969; Bowerman 1982; etc).
The details and summary of the experiments are provided on pp. 273-278.
10. W. Deutsch, A. Wagner, R. Burchardt, N. Schulz, and J. Nakath "Person in
the language of singleton, siblings, and twins"
The theme of this chapter is the effect of having immediate siblings,
including a twin, on acquisition of pronominal reference of self and
addressee. The authors begin with a section "A tribute to Clara and William
Stern", in which they review a classic study of three siblings in twenty-four
diaries (Behrens & Deutsch 1991), also available on CHILDES). The children in
this study refer to themselves for the first time in different ways: Hilde
(age 1;07.13) when looking at a picture uses the her own name "Hilde", while
her brother, Gunther (age 1;06.30) uses the pronoun "ich", when referring to
himself for the first time. Their third child, Eva, also used pronouns when
first referring to herself. The authors also note the different contextual
functions of the first use of either pronoun or proper name in self-
reference. By the age 4, however, all the siblings use the pronominal system
in a variety of contexts, just as adults do.
The Deutsch at el. longitudinal study reported in this chapter, engaged
forty-seven families - twenty-seven families with single children and twenty
families with siblings of the same sex, not differing in age by more than two
years. The sample is shown on table 10.1, p.290. The findings are reported
for both "person" naming and "possessor" naming.
Here is brief summary of the findings. Children with siblings are faster in
acquiring correct pronominal reference that those without siblings. The
broader theoretical implication, according to the authors is the fact that
different circumstances in which linguistic input is "delivered" to the child
can have an effect in what way language acquisition of concepts proceeds.
11.P. Brooks, M. Braine, G. Jia, and M. da Graca Dias "Early representation
for all, each, and their counterparts in Mandarin Chinese and Portuguese"
Two claims from previous literature are important here. One is that that
there is a mental logic (Braine et al. 1984: Braine 1978.) This theory posits
that some of this mental logical framework is innate. The child learns the
associations between the logical concepts and their linguistic
representations through learning. The authors review findings of Vendler
(1967) and Ioup (1975) of the collective and distributive interpretation of
universal quantifiers. Vender argues that there are at least two
representations of "all", "each" and "every". The first is illustrated in a
sentence "All the boys are riding an elephant", where a predicate is applied
in a collective sense. The context here would refer to all the boys riding
the same elephant. The second interpretation is distributive and would imply
and that every boy is riding a different elephant. The distributive
interpretation, according to Vendler, is synonymous with the meaning of "each
boy riding a (different) elephant". In other words, there is one-to-one
correspondence between boys and elephants. According to Vendler, the
canonical interpretation of "all" is the collective, and for "each" the
canonical interpretation is the distributive one. According to Ioup, all
languages have the canonical meaning of "all" and "every". Brooks et al.
consider the implication of the work of Vendler and Ioup in the following
way: "...the distinction between distributive and collective interpretations
between distributive and collective interpretations should be extremely
primitive, on the general basis that what that is universal is expected to be
primitive, and, perhaps, innate" (p.317). The authors further review the work
of Philip (1991) and other work by Philip and colleagues on "quantifier
spreading". This term refers to a type of error young children make in
interpreting universal quantification. In a tape with pictures showing all
the boys riding an elephant with one extra elephant in the picture, children
deny that every boy is involved in riding an elephant. According to Philip
(1991), children mistakenly interpret the universal quantifiers as modifying
whole events. The following experiments have two goals: the first one is to
see if children "obey" the "canonical" meanings of "all" and "each". The
second goal is to examine "the quantifier spreading" theory. Brooks and
Braine (1996) assumed that all the quantifier "all" would be biased toward
the collective interpretation. The task employed was a "forced picture
choice", and an example of a stimulus was "All the men are carrying a box",
and "there is a man carrying all the boxes". One of the pictures showed three
men performing an action with a single box and two extra objects. The second
picture showed one man acting on three objects with three extra men not
performing any action. Children at the age of four were able to perform
correctly on the sentences with "all". The same type of experiment was
performed with "each" to see if children understood the distributive
interpretation. Here only children as old as 9-10 showed a correct
interpretation of "each" in a pair of sentences, such as "there is a man
carrying each of the boxes", and "each man is carrying a box". The next
experiment was design to see if children preferentially associate "each" with
distributive and "all" with collecting readings. Children were presented with
depictions of all possible interpretations: collective, distributive and
exhaustive. The exhaustive interpretation example refers to a situation where
all actors are performing an action on all objects in the pictures, and there
are no extra agents or items left over. In "all of the women are carrying a
cake", adults give both collective and distributive choices (some preference
for collective one). In the first picture, all women were carrying the same
cake with two cakes left over, whereas in the second picture each women was
carrying a separate cake with two cakes left over. The exhaustive meaning
corresponded to three women carrying 5 cakes: two women had two cakes each,
and one woman had one cake. No "extra" items were depicted in the pictures.
5-6 years-old children make the same choice, but include some exhaustive
interpretations as well. "Each woman is carrying a cake" is mostly given the
distributive reading at any age. The sentence and the given representation
corresponded to the canonical distributive meaning of "each". There was no
evidence for "quantifier spreading" in this experiment, as there was in an
earlier one with "each". Cross-linguistic comparisons with Mandarin Chinese
and Portuguese provides even stronger evidence for the early availability of
collective and distributive interpretations at a young age. In their
conclusion, the authors state that the canonical semantic representations
discussed here ".... presume to underlie linguistic universals (Ioup 1975)
and seem to be likely candidates for syntax of thought serving as a
foundation for deductive reasoning across the life span (Braine 1994)". In
the last portion of the article, as well as in the discussion of the Mandarin
and Portuguese data, there is an elaboration for the support of the
"competition" model, referring to language-specific cues (Bates and
MacWhinney 1987). In fact, the age-related interpretation of sentences also
reported in this paper are interpreted as support for children's increasing
awareness between language-specific cues and the interpretations of the
12. K. Drozd "Children's weak interpretations of universally quantified
As the previous article, this paper explores children's understanding of
quantifiers, in particular, the errors children make in understanding
sentences with "every". One type of error occurs in the following context.
The children are shown a picture of three boys riding the same elephant, with
two elephants not being ridden by anyone ("exhaustive pairing error") When
asked a question "is every boy riding an elephant?", they incorrectly answer
"no". The second type of error is labeled the "underexhaustive pairing
error". In this case, in an answer to the same question children incorrectly
say "yes". This picture depicts three boys riding three separate elephants,
with the fourth boy not performing any action. According to Drozd, children
interpret universal quantifiers as weak quantifiers. Strong quantifiers
include "every", "all" and "most", whereas the weak ones include "some",
"many", and. "two".
Drozd reviews the type of errors children made in previous studies with
quantifiers and reviews the experimental designs of these studies One of the
studies that seems to have achieved successful performance is the one by
Crain et al. 1996 study with fourteen 3-5 year-old children. Crain et al,
familiarize their subjects with the intended domain of quantification before
testing them for the actual domain of quantification. On pp. 348-349, Drozd
describes an example to of this experimental manipulation by Crain et al. In
this study, children responded incorrectly only 12% of the time.
Drozd further reviews various hypotheses about the mistakes children make
with quantifiers. It is proposed that in the stimuli including an extra
"elephant' in the picture to the scenario of three boys riding an elephant,
"the condition of plausible descent" is violated. In other words, children in
their answer to the question "is every boy riding an elephant?" in this
scenario, assume that there should be the same number or elephants and boys
for the question to require a "yes" answer. However, if the children hear a
story before the actual test that tells them that not all the participants
have to be involved with all the objects, their performance improves
significantly. However, as Drozd observes, this account does not explain
(among other points) the facts why exhaustive pairing errors disappear when
cardinal or definite plural determiners replace universal quantifiers in test
Drozd further reviews other hypotheses summarizing the reasons children made
mistakes in the universally-quantified sentences. Among these hypotheses is
the one proposed by Philip (in the preceding article of this volume).
According to Philip (1995) and Philip and Coopman (1995), children apply
universal quantification of events, rather than individuals.
In the next section, Drozd proposes his solution - "the weak quantification
hypothesis". This hypothesis proposes that children assign a weak-quantifier
reading to universal quantifiers. The child's response to the question is
"every boy riding and elephant?" may be interpreted to be asking ".about
every boy who ought to be riding an elephant or every boy whom the speaker
intends to ride an elephant given the situation" (p.358). This assumption by
child, presumably, explains the exhaustive-pairing error). On pp. 359-368,
Drozd also applies the WQH to the explanation of the underexhaustive pair
reading by children, as well as the numerical strategy the child uses in
Part IV: Relational concepts in form-function meaning
13. Eve Clark "Emergent categories in first language acquisition"
Clark points out that children's first fifty words tend to be very similar in
content, and so are their first word combinations. This holds true across
languages (Clark 1970; Slobin 1970). Children start with the same conceptual
categories across languages as well. Clark points out that these early
resemblances are due to a salience of certain concepts to all children.
However, children still need to discover how the conceptual categories are
expressed by their native language. Some of this information is given to the
child by the caretakers' speech (e.g., Bowerman 1996). In regard to this
point, early acquisition phenomena suggest that children may initially
express some conceptual categories, not supported by their linguistic input.
This idea, according to Clark, supports universal conceptual representation
for children of different languages. These are the "emergent" categories. By
identifying emergent categories in a specific language is a good way,
according to Clark, to find out what categories would be likely to appear
first across languages.
These categories, according to Clark, appear temporarily, often in the second
year, but are soon abandoned or reanalyzed in accordance with the input
language. She proposes a continuum from emergent to robust categories (nouns,
adjective, verbs). Her proposal of the acquisition of language is the
following. Children first attend to conceptual categories. Around age 1;0 or
so, they begin "looking for ways to communicate about some of these
categories" (p.382). At this time, children begin to map some of these
concepts to language. Clark's view is that linguistic expressions are not
given a priori, but have to be discovered (p. 383). This would explain the
non-adult linguistic expressions children sometimes assign to certain
concepts. She also proposes criteria for discovering conceptual universals
that underlie language. The criteria consist of examining what children do
and do not map into language early on.
Clark also proposes some candidates for emergent categories across languages
looking at overextensions in early word use in English and comparing this
overextended use with the categories that are grammaticized in other
languages. Among these are classifiers, including numeral classifiers or
shape classifiers - a distinction found in some languages of South-East Asia.
These overextensions in English are observed between ages 1;0 and 2;6 and
last only a very short period of time from a day to several weeks. These
overextensions are not random: they are typically based on the shape of the
object. Attending to shape over other aspects of objects is also found in
infants (e.g., Clark 1983; Baldwin 1989). Among other overextensions created
by children, Clark lists "source", e.g., overgeneralizing the meaning of the
preposition "from" in English. On the other hand, children also grammaticize
distinctions of "inherent" vs. "temporary", not found in English, but found,
for example in Spanish verbs "ser" and "estar". Further candidates for
emergent categories are "degrees of agency" (pp. 394-398).
As for methodology that would give us the best information about emergent
categories, Clark recommends detailed diary data combined with cross-
14. Slobin "Form-function relations: how do children find out what they are".
I will begin with Slobin's final section of the paper, since in it he
indicates that he found himself going back to his original ideas about the
relationship between innateness and learning and quotes his lines from his
1966 work, some of which are repeated below: "It seems to me more reasonable
to suppose that it is language that plays a role in drawing the child's
attention to the possibility of dividing nouns on the basis of animacy; or
verbs on the basis of duration, or determinacy, or validity; or pronouns on
the basis of social status, and the like" (p.443).
In section 1, Slobin discusses grammatically specified notions. He reviews
the work by Talmy (1988, etc), who analyzed many cross- linguistic
differences in this respect. Talmy also described similarities across
languages in what they do not encode grammatically in certain "domains", for
example in verbal inflection, So we do not find "color of an event
participant" or "spatial setting" marked on a verbal inflection in any
language, but we do find tense, aspect, person. etc marked as part of verbal
inflection. In addition, there are grammaticizable and not-grammaticizable
"qualities". They include "topological" or "topology-like notions" (among
others). For the detailed discussion of Talmy's work see pp. 407-410. Slobin
raises the question as to why should precisely the notions mentioned be
grammaticizable across languages. Talmy's (1988) answer is that these notions
provide an innate "conceptual framework", or "a skeletal structure or
scaffolding, for the conceptual material that is lexically specified" (Talmy
In Section 2, Slobin lays out the "conditions and "assumptions" of the
In section 3, synchronic evidence is provided for modifying linguistic
conditions on learnability. The discussion begins with definitions and
examples of grammatical morphemes and also examples of categories which do
not clearly belong to the category of grammatical morphemes. Among these are
modals and auxiliaries, which may function normal verbs, e.g., "you had to
go" (marked for past) as) of person "you hafta, he hasta" (p.413). At the
same time, there are other modal verbs that could not be marked for tense or
person, e.g., "should" or "will". These and other examples show the
"...fuzziness of the category "closed-class item" or "grammatical category".
Also, in Spanish, modals function like regular verbs: they are marked for the
same features as regular verbs.
The next question is how do we define a "closed-class" item.
One of the proposals of identification of closed-class items by young
children was made by Gleitman and associates (e.g., Gleitman et al 1988). The
proposal is that because of prosodic property of being unstressed, these
items will be eventually recognized by the child as being of closed class.
However, as Slobin notes, this explanation does not take into account
statistical distribution or semantic, syntactic and cross-linguistic
differences in acoustic features of these elements, so it is not a sufficient
explanation of how young children acquire these items. Other questions of
"verbal class" divisions, are also discussed in this section of the chapter,
e.g., Levin 1993.
In the next section, Slobin discusses the question of "what makes a notion
grammaticizable", as well as problems surrounding this idea, when it is
examined in the light of cross-linguistic evidence. The categories discussed
are Mandarin and English classifiers, Mayan motion verbs and directionals and
motion in English and Korean (Choi & Bowerman 1991). The discussion is very
informative in showing how difficult (if not impossible), it is to come up
with the definition of content words versus closed-class items. As far as,
what is grammaticizable across language, there is empirical evidence, but not
a clear-cut explanation of this process. Slobin suggests that at the "...
present stage of our knowledge, it is premature to attribute a particular
organization of grammaticizable notions to the child at the beginning of
language acquisition" (p.427).
In adult languages, some of the psycholinguistic processes that may account
for semantic limitations on the grammaticizable notions are proposed. These
processes are the following: frequency of use and generality of meaning,
frequency of use and reduction of form, and frequency of use and online
accessibility. As for the last category, examples included in this process
are "basic words". These processes, according to Slobin, contribute to
grammaticalization in language acquisition.
In section 4.3, Slobin discusses a functionalist account of the classes of
grammaticizable and non-grammaticizable notions.
He proposes that the grammaticizable notions should be easily accessible for
online processing, they should be general, frequent and salient for the
domain in question.
In section 5, Slobin returns to the three linguistic Conditions and
Assumptions of the task of acquisition of grammatical morphemes. For example,
in discussing Condition 1, Slobin states that are many closed class or semi-
closed class items in a language (as discussed in the previous section of the
paper on English and Spanish modals and auxiliaries): however, but they
cannot be classified by the learner as clearly "grammatical" and "lexical"
(p. 437). Slobin again emphasizes his earlier point about the problems in
defining the learning task and the fact that we need more empirical sources
from various languages. This discussion also includes the role of linguistic
input to the child, which Slobin defines as a "nutrient", rather than a
"trigger" (p. 438).
In Section 5.2, Slobin proposes to "abandon the search for innate form-
function module and follow Annette Karmiloff-Smith (1992) "beyond modularity"
(p.439). In other words, there is a "developmental" approach proposed for the
acquisition mechanism that is not modular in any strict sense. Slobin
suggests that grammaticalization is included in the process of concept
In 5.2.2 the problem of constraints on hypotheses entertained by the child is
raised. It is not obvious, according to Slobin, that any existing theories,
such as "parameterization", "operating principles", etc. can explain how the
child acquired grammaticalization. The idea here is that we simply do not
have a measure of what is "too hard for a child learner "(p. 439).
Slobin also proposes a new idea of "typological bootstrapping" in section
5.2.3. According to this notion (not to be confused with the notion of
parameterization), the child builds some "expectations" while being exposed
to her/his native language. The child at some point in the development begins
to realize, for example, that Korean uses verbs to express paths of motion,
while English uses verbs and particles for locative and temporal relations.
The child will eventually discover the type of language s/he is acquiring.
This in turn ".fosters modes of "thinking for speaking" (p. 442, also Slobin
15. Heike Behrens "Cognitive-conceptual development and the acquisition of
grammatical morphemes: the development of time concepts and verb tense"
The research in this paper concentrates on the relationship of grammatical
development in the domain of tense and aspect marking. Several levels of
representation are taken into account ranging form cognitive abilities to
linguistic properties (fig, 15.1, p. 451). The author discusses ways of
linguistic marking of event types by either inflectional or lexical markers
The concept of "Aksionsarten" (lexical classification of verb classes) refers
to certain properties of actions, such as "durativity" and "telicity".
Languages differ as to how they assign Aktionsarten (Rispoli 1990, 91). One
difference is cited from Rispoli (1990) between English and Japanese in
encoding 1) the resultative activity in English and 2) putting on clothes in
English, while in Japanese both concepts are expressed by the verb "hak".
1. I am wearing pants
2. I am putting on pants
3. Watashi wa pantsu haiteru
"I am wearing pants" and "I am putting on pants" (Japanese)
The next level is considered to be tense and aspect marking. Languages can be
similar to what classes take which markers, but they may differ to, as
exemplified in the text by English and Japanese progressive markers. In
English, the progressive marker "ing" indicates ongoing action, whereas in
Japanese "tei" can also mark the resultant state of an achievement verb like
"falling" ("=has fallen", p.452).
Section 1 contains three hypotheses about the acquisition of tense- aspect
morphology. The data cited is from seven German children (ages 1;0- 4;0) from
Hilde, Gunther and Eva (Stern & Stern 1928) from Julia, Daniel and Mathias
collected by Clahsen 1982) and data of Simone (Max Muller 1976).
The analysis in this paper is based on the time period which is either right
before the onset or the actual beginning of the productive use of inflection
The first hypothesis discussed is the Cognition Hypothesis (c.f. Cromer 1974,
1988). Cognitive development is seen as a required, but not sufficient
condition for linguistic development. According to Slobin (1988 and Cromer
1973), "factors like semantic or formal complexity may influence the course
of acquisition as well." (p. 454).
As for cognitive development, Piaget considered that children are egocentric,
live in the "here and now". According to him, children fully understand the
concept of "time" by at about age 8. This view has led researchers to make a
connection between children's initial use of past tense only for events that
bear a relation to the present (resultative verbs, such as "fall", "break"
(e.g, Antinucci and Miller 76, but see Weist 1986 for a different view). This
view has come to be known as the "Aspect before Tense hypothesis". However,
Weist et al, and others showed that this view is not correct.
In the next section, the time course of acquisition is discussed. Examples in
the Behrens chapter show that children (a little over 2;0) understand and
encode past tense with activity, and somewhat later, stative verbs, rather
than just resultative verbs (having a connection to the present situation).
The most convincing case of the use of past tense without aspectual input,
according to the author, is the use of the past tense of the copula
preterite, since the copula is "semantically empty" (see example on p. 460).
In the next section, even more convincing proof is given for children's use
of past tense. This is the use of telic (verbs denoting that an action
reached a terminal end-point "change of state") verbs in the past tense.
Further, children are able to refer to remote past events, future events and
are able to connect past tense marking with other tenses. These examples are
on pp. 461-463, and they serve as very convincing empirical evidence for
children's productive use of tense early on.
In summary, Behrens states that even though, the majority of early past tense
verbs are resultative (the event they denote has bearing on the present),
this preference can be seen only in early speech. Thus the semantics of the
verb does not determine the tense children use in their initial productions.
In the next section, the "language specificity hypothesis" is discussed. Two
proposals are discussed here: "the role of the morphological and syntactic
systems in the acquisition of tense markers, and the idea that language
itself may influence the acquisition and development of conceptual and
cognitive distinctions" (p. 465). According to the Language Specificity
Hypothesis, on the semantic level, the early child language "should not clash
with the target level distinctions" (p. 465). This in turn implied that if
children obey all the language-specific adult language properties, they must
have derived these properties from the linguistic input. Indeed, the fact
that children obey the adult tense-marking system of German, and also
correctly distinguish early on between finite and non-finite verbs, supports
their language- specific knowledge. This evidence also refutes any reliance
any universal semantic notions that children may use in the acquisition of
their native language.
Behrens also cites bilingual acquisition data (Slobin 1973 and 1985) that
supports language-specific influence and argues against cognitive
prerequisites. Schluyter (1990) is cited in the study of French-German
bilingual to show that the input affects children early on in picking out
what is specific in the languages they are mastering. Here the study by Choi
and Bowerman 1991 is cited for evidence that young children pick up the
spatial terminology of the input language immediately, and this language in
turn affects their formation of concepts of space. Behrens also hypothesizes
that it is possible that tense morphology itself helps children understand
the concepts of time. However, as Behrens points out, there is no empirical
evidence to support this hypothesis.
In summary, Behrens argues against "preexisting universal conceptualizations"
(p.468) and thus against the Cognition Hypothesis or the Semantic
predisposition Hypothesis. Berhens also mentions that the German tense data
cited in this chapter shows that language- specific morphosyntactic tense
features are in place at a very early age. Behrens also points out that the
mechanism of the early extraction of this rich and rapid language-specific
information still needs to be studied.
16. M. Bowerman and S. Choi "Shaping meanings for language: universal
andlanguage-specific in the acquisition of spatial semantic categories"
Where do meanings of first words come from? Are they introduced through
language? Do they reflect concepts that "...arise simultaneously with
"spontaneously through infants' perceptual and cognitive development?" Is
there an interaction between language and cognition in introducing word
meanings? If yes, what is the nature of the interaction?" The authors argue
in favor of an interactionist view between nonlinguistic conceptual
development and semantic categories of the input language. Thus, they do no
support the view that language development consists of mapping the pre-
existing concepts onto linguistic categories. (e.g., Nelson 1974).
In section 1, B & C review evidence of the role nonlinguistic spatial
development in the acquisition of spatial words.
According to of certain spatial terms, such as "behind" and "from" are
underextended initially (Johnson 1984). Other words, such as "open" are
initially over-extended (Bowerman 1978, E.V. Clark 1993). However, since the
evidence cited in itself is linguistic, it is not clear that the source of
these errors comes from some cognitive sources, rather than linguistic
overgeneralization. Another proposed explanation of these errors is the idea
that children form universal semantic conceptualizations (e.g., Clark, this
volume; Slobin 1985).
In the next section, cross-linguistic variation in spatial terms is
discussed. The fist part of the section is concerned with motion "along a
path". Crosslinguistic differences in regard to this concept are summarized
by Talmy 1985, 1991. In Fig. 16.1 a (p.482) the authors show a diagram of
correspondences between English and Korean in the notion "in" and "on ". An
example of a difference between the two languages is the distinction made in
Korean when motion is "caused" or "spontaneous'. English, on the other hand,
does not make this distinction. So, in English "put it on the table", and
"climb on the table", or "put it in (the bathtub)" and "get n the bathtub")-
caused and spontaneous respectively, can be expressed by the same means. In
Korean, on the other hand, there is a verb ""kkita" - "to interlock, fit
tightly", but no intransitive equivalent for "moving physically into a tight-
fitting relationship. "Nehta" in Korean, means "putting something in or
around loosely". However, there is no intransitive verb meaning "move into a
relationship of loose containment of encirclement" (e.g., "get in the tub").
There is a verb in Korean "move (be) in" which does not specify tight or
The next section contains a discussion of "static spatial relationships".
Some of the notions are containment, support, encirclement, etc. In Bowerman
and Pederson (1992, in preparation ), thirty-eight languages are discussed
with respect to these notions Fig, 16.2 on p. 485 exemplifies distinctions
made in some of these languages. For example, Spanish has only one
preposition "en" for the notions shown in the figure, English has "in', and
"on", and Dutch has "op", "aan", and. "in". An English speaker learning Dutch
would have to "break up", as it were, the notions expressed by the
preposition "on" into "aan" and "op". As B&C indicate, if the contact between
objects is such that it must "counteract gravity", "aan" is used in Dutch.
Examples given on the figure "apple on a twig", "handle on a door", and
picture on the wall". Otherwise, "op" is used, e.g., "cup on the table"
(p.485). The investigation of these languages shows a systematic arrangement
of these concepts across languages, a "hierarchy".
In section 3 of this paper, the question of acquisition of these concepts
cross-linguistically is addressed. Do children first acquire the same
semantic concepts of space universally and acquire language specific
differences later? (c.f. Slobin 1985)? Do children in different languages
grammaticize spatial differences from the beginning, according to their
The authors first discuss a longitudinal spontaneous speech study in English
and Korean with children aged 1-3 years of age. Children in both groups used
spatial word productively at 16-20 months and used them following the adult
pattern of their native language. The English children's beginning uses were
particles and verbs, such as "open" and "close". "The English-speaking
children concentrated on notions of containment, and support and surface
contact, especially attachment (on and off) and vertical motion (up and
down)" p.488. They used these spatial categories (as in adult English), for
both caused and spontaneous motion. Korean children, on the other hand,
distinguished early on between caused and spontaneous motion along a path.
Children in both language groups made some errors, but their errors also
supported distinctions along their respective languages. Elicited production
of words for "separating" and "joining" objects was also conducted in both
languages with known and novel objects. In the age ranges 2;0 to 3;6 and ten
adults. The study was conducted with English, Korean and Dutch children.
Children behaved the same way as adult controls in their native languages.
Even though, young children made errors, there was no evidence for common
semantic classification in the spatial concepts across languages.
Children before age two were also tested by using the preferential- looking
paradigm (Golinkoff, Hirsch-Pasek, etc.), Thirty Korean and English -speaking
children of 18-23 month-old were tested. This task again showed that very
young children already distinguish between relevant spatial notions according
to the input language.
In section 4 of the paper, the authors raise the question of how spatial
semantic learning takes place. First of all, the authors emphasize again that
spatial learning seems to take place even before verbal productions (as their
"preferential-looking paradigm" experiment shows). However, B& also emphasize
that children bring some conceptual biases of their own to the task of
constructing spatial categories. Language then helps the learner in
strengthening the notion of what distinctions are relevant for selecting
spatial words. Thus, in the process of learning spatial terms, there is a
"...constant interaction between the learners' built-in sensitivities to
space" and the characteristics of their native language. The relevant factors
include "frequency", consistency in the word's use (rather than polysemy),
the number of words for a given semantic piece of continuum, and the degree
of overlap between the referents. This view of acquisition can be ".placed
within the framework usage-based approaches to language that stress the
dynamic properties of linguistic knowledge". This view is compatible with
Slobin's (ch.14 of this volume) of "typological bootstrapping" and other
views in (e.g., see Smith, in this volume).
In the next section, the authors theorize about the reasons of why children's
over- or -underextensions in their spatial word cannot be ascribed to
influences of some semantic universals for spatial terms, but can rather be
explained by some properties of the input languages.
Toward the end of the article B&C address the definition of "perceptual
sensitivities and conceptual biases for space" (p. 503). They review the
proposal of "constraints" on word learning proposed by some, including,
Landau and Jackendoff (1993).
17. P. Brown "Learning to talk about motion UP and DOWN in Tzeltal: is there
a language-specific bias for verb learning?"
This chapter focuses on the acquisition of the semantics the Tzeltal
terminology used for spatial orientation: adverbs of direction, verbs and
nouns. The terminology is referred the as the UP/DOWN vocabulary, and it is
used for both vertical and horizontal axes. Brown concentrates on labeling
the vertical axis, which is most "problematic" in this language. She
concentrates on the problem of universal semantic features, specifically, on
the semantic feature "vertical" in the language acquisition process. Brown
argues against built-in universal semantic, or "natural prelinguistic
concepts" (e.g., H. H. Clark 1973, Slobin 1985).
In this chapter, Brown argues that child ".do not necessarily start with a
putatively universal, perceptually based vertical meaning for ... verbs and
nouns". In fact, she argues Tzeltal children learn the semantically specific
verbs first, and also initially stick to "very specific meanings for the
UP/DOWN verbs they learn" (p.514). In their early use, Tzeltal children do
not use the UP/DOWN terminology with vertical meaning, but rather use it to
refer to the slope of the land and particular locations. Children use these
terms to refer to local slope, rather than the overall (North/South) slope of
the land up ot age 7 or 8.
First, the spatial vocabulary of Tzeltal is described. In this system ''UP"
and "Down" are used instead of front/back/left/right distinction and also
instead of north/south/west/east distinction. It is encoded in Motion +Patch
intransitive verb roots, such as "mo" (ascend/descend, i.e., move in a
South/North direction). This distinction is also included in the
transitivized counterparts of these verbs: mo-tes/ko-tes "make
ascend/descend") and in the directions adverbials derived from them, such as
uphill/downhillwards (p.515) and nouns.
In addition, this system contains and terms for "acrossway" and for "going
across". (motion along it). "Using this system, one speaks of a Figure (the
object being located) as being 'uphill/downhill/across' in relation to a
Ground (the reference point" (p. 515). The cognitive complexity is compounded
by semantic complexity. The difficulty may be seen in the ambiguity that
arises when these terms are used. e.g., 'ascending' or 'descending' may mean
motion vertically or horizontally. The ambiguity is summarized as follows:
Up/down may mean 1) vertical. 2) overall downhill slope of land and 3) the
Of course, the question comes up of which spatial distinctions the child
masters first, considering the wealth of information s/he is faced with.
Also, does the child have any prelinguistic concept of "upness" (Bloom 1973),
etc. The acquisition is examined through naturalistic and elicited production
data. The interesting finding is that in spite of the complexity involved in
spatial terms, children "... are able to postulate language-specific terms
from the beginning" (p.518). This, of course, supports Bowerman and Choi's
(1991) conclusions also discussed in this volume. Specifically, by the age of
2;0, children acquire Motion +Path verbs in four different meanings from the
start: motion toward and away from something and uphill and downhill
motion.The same can be said about relational nouns. However, Brown also
points out the lack of frequency with "ascend" and "descend" verbs, which
they interpret as a possibility that children simply memorized a list of
context in which these verbs occur. However, by age three, there is evidence
of productivity, rather than just memorization of imitation. However, it
becomes clear from further discussion that Tzeltal children even at the age
3;6 uses these terms only in relation to local topography, (or at least, in a
limited number of contexts). It is also true that children do not
overgeneralize the Tzeltal spatial terms in a way that would suggest using
'semantic" universal (not language-specific) concepts of motion in space.
There is a very detailed and interesting discussion with tables and examples
illustrating the results of natural production data. Brown includes data from
"space games" in section 3.2. These data illustrate that by even by age 7-8
child some children are too specific in using the intransitive "molko",
rather then generalizing the concept to overall North-South slope of land. It
is important to note, however, that these games are conducted indoors, a
point addressed in an interesting article by Li and Gleitman (discussed
later), in which it is shown that with available landmarks outdoors change
the results significantly.
To sum up Brown's article, children from early use the language- specific
spatial system of Tzeltal, however, in a restrictive way. There is no
evidence of "semantic universals" in Tzeltal's children conceptual system or
language. On p. 528, there is a brief discussion of how language at a very
early age could direct the understanding of the very complex conceptual
system of Tzeltal, but the mechanism of this effect is not totally clear.
One of the proposals is that children realize early on that many verb types,
e.g, verb denoting "eating" are very specific, e.g., there are different verb
roots used for eating meat, vs. "'eating" something else of or "eating" in
general. The same is true for the verb "carry/hold". This concept is
expressed by a different verb for different means of " being carried", e.g, a
distinction is made for "carrying in both arms", "in hand", "supported from
top," in "mouth, etc (p.529). Brown speculates that this specificity of
meaning expressed in different verbs may carry over to spatial terms. In
other words, children restrict their use of spatial terms until they hear all
the examples in the input language for that specific term, and this could
take a long time.
The author also points out that her proposal for a "verb specificity bias in
Tzeltal runs in the opposite direction to E. Clark's 1973 hypothesis based on
Indo-European language data". In Clark's 1993 proposal, children have a
preference for "light" verbs (the are "semantically bleached"(p. 535). In
Tzeltal, "specific'" verbs occur much more frequently than "light" verbs.
Moreover, children do not substitute more specific verbs by "light" verbs in
Tzeltal. In conclusion, Brown re-iterates her proposal according to which
children learn word meanings both conceptually and linguistically from
specific contexts (view similar to Tomasello's) in a language specific way.
Another finding is that children learn verbs before nouns. This is because
the referring information children pick out about arguments is "largely
encoded in verbs" (at least, transitive ones) (p. 536)).
18. L. de Leon "Finding the richest path: language and cognition in the
acquisition of verticality in Tzotzil (Mayan)"
The author reviews Talmy's (1985,1991) typological analysis of motion events.
Tzoltil does not fit into one type, but combines both "satellite-framed" and
"verb-framed" languages as follows. Sematically specific path verbs included
some verbs of vertical motion, such as those indicating "falling". As for the
similarity with sattelite-framed language types, Tzoltil includes
directionals that freely combine with verbs, e.g., 'pit'ochel - "jump in"
('ochel "enter"). Directionals can also combine with path-conflating verbs
and with stative predicates.
The results of a longitudinal study of two girls and two boys (aged 19-25
months) are described. Contrary to expectations, none of the four subjects
used directionals before path verbs. Also, contrary to the predictions,
children used semantically specific verbs before "pure" path verbs, such as
"ascend" or "descend", which were used very infrequently. When verbs of
"ascend" or "descend" appeared in the speech of these children, no
overgeneralizations were observed. This is contrast to English-speaking
children using "up" and "down" for any vertical motion. Among "semantically
specific verbs", Tzoltzil children use "falling verbs denoting effect of
gravity" and "verbs of posture and support". This again shows, according to
de Leon, as in Choi and Bowerman (1991 study and others), children acquire
language-specific semantic differences early on and do not overgeneralize
them. They also acquire more "general" verbs, such as "ascend" and "descend"
much later. De Leon interprets the results of the study as evidence against
cognitive influence on language, but more in an interactive way of cognitive
and language specific influence on child language development "... along the
line of "typological bootstrapping" hypothesis put forward by Slobin in this
19. S. Levinson "Covariation between spatial language and cognition, and its
implications for language learning"
In this chapter, Levinson proposes that adult cognitive operation underlying
meanings vary from language to language. He begins by giving an English and
an equivalent Tzeltal example:
(a) English: Put the bowl behind the box
Pacha-an-a bojch ta y-anil te karton-e
Bowl+put-CAUSE-IMP gourd+bowl at its-down the cardboard-DEIC
In addition to the various morphological differences seen in (a) and (b) in
the two languages, Levinson points out a lexical difference in the word
"behind" and notes that Tzeltal does not have a translational equivalent of
the English version. There a similar concept which expresses a deictic notion
of the box being placed between the speaker and the bowl.
What about the "Mapping problem" from concept to word or word to concept in
children? The discussion of adult cognition varying with language can be
found in section 2.2.
One of the differences between languages discussed by Levinson is the
Absolute and Relative spatial coordinates. The relative notion of "left" and
"right" change with the rotation of the speaker in relation to fixed
surroundings, whereas the Absolute coordinates do not change with speakers'
rotation in fixed surrounding. However, if the surroundings (reference point)
rotates, both the Absolute and Relative coordinates change. Is there any
extra computation involved in any of the two systems? The authors employed a
non-linguistic task to answer this question. In a task where an array of
three animals was presented to speakers, the speakers were turned 180 degrees
and asked to arrange the toy animals in the same array. As predicted, Dutch
speakers with a language using relative coordinates performed using the
Relative bias, while Tzeltal-speaking Tenejapans performed in accordance with
the Absolute system. This pattern remained consistent in other tasks. In
section 3, Levinson raises the question of how children approach the task of
"cracking the local linguistic code" (p.581).
Li and Gleitman (in press) repeated the experiment reported by Brown and
Levinson (1995), but with English speakers only. Their goal was to see if the
Tenejapani-like and Dutch-like spatial judgement could be induced in the
speakers of the same language by making changes in the experimental setting.
The subjects performed a similar experiment with lining up toy animals after
being rotated 180 degrees in "blinds up" and "blind down" condition. In the
first condition, the subjects could see the outdoors, whereas in the second
condition, the subjects could not see the outside. In the process of seeing
outdoors, the subjects could refer to landmarks, e.g, the local library. This
would strengthen the resemblance to the Tenejapani testing ground. With
having the "blinds down" condition, the subjects would be in the Dutch
speakers' situation, where the experiment was conducted indoors with no
landmarks available . The subjects in the "blinds down" condition performed
similarly to Dutch speakers in Brown and Levinson 1995), whereas in the
"blinds down" condition, the subjects' performance was somewhere in between
the Tenajapan and Dutch subjects (U-shaped distribution).
To see why in the "blinds down" condition, the subjects Tenejapani "absolute"
response was replicated only partially, another question was raised as to the
"strength" of the landmark cues. In this experiment, the blinds were always
up, and a toy was placed on the Stimulus Table to the right/south side of the
subject. When the subjects was rotated, the "landmark" toy remained in the
same unmoved position throughout the trials. The subjects were divided into
two groups. For the "Relative Ducks" group, the landmark toy was always on
the right of the subject, whereas for the "Absolute Ducks" group, the
landmark toy was always on the south side. With this "relative" and
"absolute" biasing manipulation, the Dutch/Tenejapani findings have been
replicated. This, of course, shows, that something other than linguistic cues
is responsible for absolute vs. relative biases of speakers of the two
different languages in previous studies, thus a "non-Whorfian result. In
section 3 of his chapter, which Levinson titled "Solving the impossible", he
addresses one of the major problems of the "mapping" problem - the fact that
the infant may be in a state of "cognitive mismatch'" with the adult. This is
because the adult has the gift of language already, which the infant lacks.
Now, if one believes that language determines thinking and affects cognition,
this is indeed a problem. What happens in the pre-linguistic stage, and how
long would it take for young children to become either "absolute" or
"relative" thinkers, for example, if they have to wait for relevant
linguistic cues to "shape' their cognition in a certain direction? The author
gives some solutions to this problem in a way of positing a temporal
succession in mapping or in considering heuristic solutions (also suggested
by others in this volume). These heuristics may reduce the demands of the
learning task by being transferred by the learner from one domain to another.
I found this book extremely valuable to me as a researcher in child
language acquisition. I highly recommend this volume to other researchers in
the field. It provides the reader with a broad range of topics on children's
conceptual and linguistic development from various viewpoints. What is also
attractive about this edited volume is the diversity of languages considered
in illustrating the authors' various theories. What I would have liked to see
added to this volume is work that argues against the Sapir-Whorf Hypothesis,
so the reader could get a more balanced and comprehensive view of the
research in this area. One such paper, for example, is Li and Gleitman (in
press). This paper convincingly challenges the Whorfian hypothesis in
experiments with "absolute" and "relative" spatial terms. All in all,
however, this book is a very useful reference book of research in the field
and a potential source of future research ideas on conceptual and linguistic
development. and the interaction between them. The research in this book also
reminds us again how much work is ahead of us in accounting for the speed and
relative accuracy of children's language acquisition.
REFERENCES (other than those appearing in the bibliography of the book under
Li and Gleitman (to appear) "Language and Spatial Reasoning".
Wallman, J. (1992) APING LANGUAGE, Redwood Press Ltd.
Weist, R. Wysocka, K. Wiokowska-Stadnik, E. Buczowska, and E. Konieczna
(1984) "The Defective Tense Hypothesis: On the Emergence of Tense and Aspect
in Child Polish", THE JOURNAL OF CHILD LANGUAGE 11, pp. 347-374.
Eva Bar-Shalom is a Research Scientist and Adjunct Assistant Professor of
Linguistics at the University of Connecticut. Her previous research includes
spoken language processing problems of good and poor readers, and her current
research in the field of child language acquisition, specializing in Russian
child language, as well as cross-linguistic studies. She teaches Second
Language Acquisition and Language and Culture, and a Linguistics course in