Preposition Use in a Speaker with Williams Syndrome: Some Cognitive Grammar Proposals

0  Introduction

    What degree of conceptual sophistication is necessary to attain adult-like
language use? This paper makes a preliminary exploration of this question 
with reference to preposition use in adolescents with Williams syndrome, a 
syndrome incorporating both cognitive deficits and, typically, certain 
physical traits.  A theory of language allowing a high degree of flexibility
in the conceptual underpinnings of language, Cognitive Grammar, is used as a
framework for explanations of non-conforming preposition use in Williams
syndrome subjects.

    The proposals made here were formulated on the basis of a specific
data sample, and are intended as exploratory and suggestive of ways to
proceed with the larger database.


1.0  Non-conformist language use in Williams syndrome subjects


     Williams syndrome (WS) adolescents show remarkable sophistication in 
language use, in spite of severe cognitive deficits in other areas.  Bellugi, 
Sabo & Vaid (1988), for instance, report that Williams subjects show signs of
mild to moderate mental retardation in many cognitive areas.  In spite of 
this, their language use shows mastery of sophisticated structures such as 
complex embeddings (Bellugi, Bihrle, Jernigan et al, 1990; Reilly, Klima
& Bellugi, 1990).

     Nonetheless, close examination of even a very limited data sample
from WS subjects shows usages that seem not to conform with standard adult 
English usage, in areas of syntax and semantics.  We examined data from several 
WS subjects (footnote 1), in which unusual usages are observable in the areas 
of vocabulary selection, pronoun use, and use of sentence and discourse 
connectors, as well as certain facets of morphology/syntax.

     The focus of this paper is unusual preposition use in a WS subject.
In the data sample we used, unconventional uses of prepositions
were most evident in data from "Crystal", with examples from other subjects,
"Ben" and "Amy". 

     We narrowed the sample to Crystal's data, consisting of two brief stories 
she recounts and a few miscellaneous sentences out of context.  Within this 
sample prepositions occur 130 times.  We judged 25 of these tokens to be 
divergent from normal usage.  This is 19% of the total preposition tokens.
The occurring prepositions and how many times each occurred are given in
Table 1 (Appendix). 

     Regarding the usages we consider to be non-deviant (footnote 2), several
prepositions (especially 'to') were used in various meanings, indicating 
possible polysemy, as is the norm in adult English usage.  Also, there are 
usages in abstract as well as concrete spatial domains.  Here we see some 
samples of conformist spatial uses:

(1) a. Frog story:
i.  ...and he is in the house...
ii. ...he goes out of the house, to the back yard...

    b. Chocolate story:
i.  And you can also have chocolate candy bars or something that has
    chocolate in it.
ii. She was on her chocolate throne...

The next examples illustrate abstract, non-spatial relations signalled by
the prepositions.  Note the multiplicity of uses of 'to': not only 
allative, as in (1), but use as an infinitival marker and purpose marker.

(2) a. Frog story:
i.  Once upon a time....
ii. ...the frog is trying to get out of the jar...
iii....the dog looked to see if the frog was there...
iv. ...he was amazed at what the frog had done...

    b. Chocolate story:
i.   Chocolate is sweet and comes in many ways.
ii.  It's fun to have when you can eat it and drink it.
iii. ...you might melt to the ground like melted butter.

These data and others (e.g. Amy's frog story) show that WS subjects' usage of
prepositions often conforms to adult English norms, and spans abstract as
well as concrete relations.

     The non-conformist uses are given in Table 2, in the Appendix.
The expected form is given in square brackets at the end of each line.
We will refer to these usages from now on with the tags given there: IF
stands for the frog story, IC for the chocolate story, and IM for
miscellaneous data.  As is clear from the data, deviant uses occur in the form 
of substitutions, i.e. choice of a different preposition from the expected 
conventional usage, and in the form of omissions -- non-occurrence of a 
preposition where one would be expected or required (always 'to' in these data).
In one case, F6, our judgment would call for no preposition; we could
call this an insertion.  In M2, the deviation is less in the preposition than 
in the grammatical class of the object -- a non-derived nominal such as job or 
position seems more in order.

     It is also clear that certain deviant uses make up a substantial portion
of the total: There are 8 deviant uses of through, all apparently having
the same meaning in the story; 2 deviant uses of for, again with similar
meanings (call for and yell for), and 5 omissions of to, 4 of these in the 
expression change a different color.  If we consider these 15 uses to be tokens of 3 deviant usage types, the number of deviant uses is reduced significantly 
to 13.

     Both the normal and non-conforming uses of prepositions by Crystal and
other WS subjects need to be accounted for.  We will next examine empirical and theoretical questions raised by these data.


1.1  Empirical questions raised by the data


     The reader will have noted that we have been careful to avoid labelling
the WS preposition usages errors, preferring to call them deviant (in
a non-pejorative sense) or non-conformist.  Several empirical questions must be answered before the exact status of the WS data can be determined.

     First, we need to discover whether this sample is a) representative of 
Crystal's overall language use and b) representative of language use in other 
WS subjects.  Second, we need to determine the systematicity (or lack thereof) 
of non-conformist uses in WS speakers.  If we find that these uses follow a 
system, we would have the intriguing possibility of an alternative semantic 
system for this subset of language in WS speakers.  Also required is comparison of the WS usages to actual speech data (rather than armchair intuitions such as those used here) from normal English-speaking adults and children.  If we find 
that either of these populations make a similar percentage of errors of similar types, we will still have an interesting language problem, but WS speakers will cease to be a special case.  Our hunch at this point is that the WS usage would 
prove to be at variance with the usage of adult normals, especially such 
persistent items as Crystal's deviant use of through.  Comparison with normal 
children may yield a different picture.  One recent study shows that children 
do use some prepositions (from, by) in non-conformist yet systematic ways in 
the process of language acquisition, in spite of 'correct' input from adults in the nurturing environment (Clark & Carpenter 1989).  Clark & Clark (1977) also 
cite other evidence that children overextend and systematically 'misuse' 
prepositions.  If the 'errors' of the respective groups (normal children and WS speakers) turn out to be similar, we may have, in WS, a case of developmental 
arrest or linguistic retardation expectable in subjects showing retardation in 
other cognitive domains.

     For the present, we shall assume that the WS data does in fact diverge fromnormal adult usage in persistent and significant ways, but we shall make no suchassumption with respect to normal children's usage.

     Having made these assumptions, we turn to the theoretical question of
the source(s) of the WS language use.


1.2  Theoretical views of Crystal's data


     As is recognized in, e.g., research on aphasia (Ellis & Young 1988, 
Caplan 1987), there are two possible sources for divergent linguistic usages: 
differences in the affected population in linguistic knowledge, i.e. linguistic representations; and differences in processing, or access to these 
representations and activation of connections between various components of 
linguistic knowledge (and possibly other kinds of knowledge such as motor
memory, etc.).  If we accept the representation/processing division, then there are several logical possibilities as to what underlies the usages of 
prepositions found in WS:

(3)
i.   WS subjects have essentially the same semantic system as normal adults,
     but processing differences induce the deviant uses

ii.  WS subjects have the same processing mechanisms as normal adults,
     but a different semantic system underlying preposition use

iii. WS subjects have mastered only an incomplete or partial version
     (some subset) of the semantic system of adults, but have normal
     processing

iv.  WS subjects have both a different or partial semantic system and
     different processing from normals 

The difference between ii. and iii. is subtle and will be made clearer below.
It is clear that we cannot settle on any one of these options here.  What we
wish to do in the sections below is consider the compatibility of these options with Crystal's data.  We will see that they are all plausible to some extent in a Cognitive Grammar framework, but some accord better with the facts of the
data than others

2.0  Cognitive Grammar


     Cognitive Grammar (CG) is a psychologically-based theory of linguistic
structure and language processing.  Modularity of language is neither assumed 
nor rejected out of hand by this framework; its major exponent claims to take noposition (Langacker 1987:13).  However, the strong claim is made in CG that the cognitive abilities that underlie semantic representation and language use are 
not essentially different in nature from other cognitive abilities of the human.
Language learning and language use are viewed as, essentially, problem solving 
activities, which exploit the same strategies as other kinds of problem solving.

     The processing claims of CG are few, but broad. Both representations and 
language use are described in processing terms.

     Mind is the same as mental processing; what I call a thought is the 
     occurrence of a complex neurological, ultimately electrochemical event;
     and to say that I have formed a concept is merely to note that a
     particular pattern of neurological activity has become established, so
     that functionally equivalent events can be evoked with relative ease.
                                                  (Langacker 1987:100)


The notion of activation of neural structures plays a central and
broad role in this theory of language.  A representation or bit of knowledge is a pattern of activation which leaves some kind of trace "that facilitates 
recurrence" (Langacker 1987:100).  The use of knowledge is the activation of 
such traces.  These representations are our store of concepts; some subset of 
our store of concepts is said to make up our semantic inventory -- the set of
concepts which happen to be the meanings of linguistic expressions.  
Phonological forms are simply cognitive routines which are associated or 
connected in some unspecified fashion with semantic representations; they serve to symbolize the concepts.  Syntax is merely the conventional patterning which 
signals relations between conceptualized entities.  Our knowledge of syntax is 
represented as schematized versions of these patterns, abstracted from
observation of many instances of use, as well as being influenced by certain 
cognitive biases or predispositions (see Clark & Clark 1977 for detailed 
discussion of some examples).  If these tenets seem overly simplistic to the 
reader, their actual complexity can be appreciated in Langacker (1987 & 
forthcoming).

     A major advantage of CG in applications to psycholinguistics is the
flexibility it allows in what may be a semantic representation, what may
be a linguistic expression, and what may be a syntactic 'rule' (actually
something more like a schematic template). A linguistic expression is any 
symbolic pairing of a phonological form and a semantic representation which has
become firmly established as a cognitive routine.  All established cognitive 
routines are called units, whether they be individual units (a particular 
concept or a particular phonological form) or paired units, called symbolic
units, in which a phonological form is associated with and symbolizes a semanticrepresentation.  A unit can be of any size which is cognitively manageable 
(probably determined by processing constraints such as short-term memory).
Single morphemes, morphemically complex words, phrases, and sentence patterns
can all be units.

     The semantic system of any normal adult speaker of a language is very rich,exploiting many aspects of the rich detail with which we perceive the world and represent it in our conceptual system.  Certain general cognitive abilities 
apply in our structuring of semantics and syntax, so that our conceptual system is not merely an image of individual past experiences.  Our capacity to 
schematize is very important here.  That is, we can extract the commonalities 
present across situations and represent them in underspecified notions such as agent or path.  Many levels of schematicity may be found in different semantic 
representations, from highly specific to highly underspecified representations.
Levels of schematicity may differ across individuals for any given expression.
The basic question we would like to raise with respect to the WS individuals
is whether they develop semantic systems of equal complexity to adult speakers, or to normal children of their mental age.  In particular, we wonder whether they schematize in the same fashion as adults, and whether they develop as rich an 
inventory of semantic representations as normals do.

     The semantics of prepositions (footnote 3) is particularly schematic,
since prepositions apply across so many situations in the real world.  In CG,
a preposition is a relational predication -- it signifies a spatial or other 
relationship between two entities.  The entity being located is referred to as 
the trajector; the entity with respect to which it is being located is the 
landmark.  There seems to be a limited inventory of relations expressed by
prepositions in human languages, including spatial relations as well as
relations such as source, association, partitive relations, etc.

     Prepositions are highly polysemous. In CG terms, we expect the senses of a preposition to be related in some fashion, even if the relation is a distant 
one.  In some cases, a single schema can be extracted which represents the 
features found in all the senses of a preposition.  Sometimes there are related senses which have some of the specifications found in other senses, but lack 
other specifications.  In English, for instance, it is common for prepositions 
to have both a PATH sense and a LOCATION sense, e.g. 'in'. In (4) the meaning 
of 'in' can either include or not include a path.

(4)  Susan fell in the water.

Susan may either have fallen into the water from a point outside of
it, or she may have been standing in the water and fallen down within
it.  The second sense includes a container, as does the first sense,
but does not include the path.  We shall suggest later that WS subjects may not
develop families of senses for prepositions as extended as those which normals 
develop.

     Another important fact about prepositions is that they often become part offixed expressions such as pick up, turn on, take off, etc.  In such fixed 
expressions, they often have a very subtle and specific meaning.  Below we raise the question as to whether WS subjects grasp these subtleties. 


3.0  Alternative explanations for Crystal's data


3.1  WS subjects have essentially the same semantic system as normal
     adults but processing differences induce the deviant uses.


     There are many possibilities that could be considered under this heading.
For example, WS speakers may suffer from difficulties in lexical access,
causing them to retrieve random substitutes -- or semantically similar
substitutes -- for a target word (cf. Pinker, 1991 for a discussion
of lexical access problems in WS speakers).  Systematicity in WS usage would 
speak against a random access hypothesis, but not against the idea that 
semantically similar words are retrieved.  The data examined here indicate some systematicity in WS preposition use, but not enough data was available for this study to determine the extent to which WS usage is systematic.

     We would like to consider here another sort of processing difference
that might induce nonconforming usage of prepositions: a difference in the 
online conceptual construal of the situation being encoded.  In any situation 
presenting relations between entities in the world, it is possible to focus on 
different aspects of the relationship.  For example, if there is a woman walkingand the location of her walking is a park, both the following are acceptable 
construals:

(5)
a. She is walking in the park.
b. She is walking through the park.

In (5)a., we focus on the park as a bounded container within which her
walking takes place.  In (5)b., we focus on the nature of her path as beginning and ending outside the boundaries of the park; or possibly, we focus on her walk
covering some more complete or more extensive path than with in.  The two usagesinvolve construing the scene at hand in different ways; in (5)a. we do not
necessarily include the endpoints of her walk, while in (5)b. this is more '
likely.  In the WS usages, what could be going on is that the speaker has a
perfect mastery of the semantics of English prepositions, but simply picks out
of the scene to be encoded available relations which accord with the meanings
of prepositions other than the ones conventionally used for such scenes.

     Crystal's usages C4 arrived to and C3 working out on a field can be 
explained in this way.  In a sentence with arrive, the goal-oriented trajectory of the subject is highly salient, due to the meaning of the verb.  Crystal may 
preserve this construal in her preposition choice, opting for to, which in at 
least some senses codes a path with endpoint focus. Crystal shows knowledge of 
this sense in her usage he goes out of the house to the back yard, in the frog
story.  The relations associated with the form to are indeed present in the 
arrival scene, and hence available for coding.

    In the case of working on a field, it is clear that a field can be construedas a bounded container-like space, sanctioning the conventional use in a field, as in a sentence like 'Along the highway we saw five cows standing in a field.'
It is also possible to focus on the two-dimensional extension of the field
-- its surfacey nature, as is common in sports, e.g. After the game, the victorsstayed on the field for a team photo.

     This explanation seems quite plausible for at least some of the usages in 
Crystal's data.  Also, it is not in disaccord with the fact that Crystal does 
not always use an unexpected preposition.  Construal of a scene is a local and 
situation-bound phenomenon, and a given speaker is free to analyze out any 
relational aspect of a scene that her imagination makes available.  On one 
occasion, Crystal may pick out the surfacey nature of a field and use on; in 
another she may well pick out the container-like nature and use in.  We would
thus be alerted to look for such alternate usages in her data.

    What remains special about her data is its unconventionality.  This may be 
due to her not having noticed the lack of such uses in adult speech -- it seems plausible that a retarded child might not notice that something does not occur.
It is also quite possible that Crystal has not been made aware of the deviance
of her usage by her interlocutors.  There are probably very few cases where her deviant usages actually disrupt communication, and people in her environment 
may also have lowered expectations, being accustomed to her cognitive 
difficulties in other areas.  We might also note the general tendency for
caregivers not to correct language use in young learners (Clark & Clark 1977:325-326).

3.2   WS subjects have undisturbed processing, but a different semantic system 
      underlying preposition use.

     In this case, we would posit different meanings for prepositions 
corresponding to the English phonological forms.  This could be compared to,
say, a foreign speaker whose language has a different way of apportioning the
set of relations usually coded by prepositions across languages.  Take as an 
example German an, which in some cases is equivalent to English on as in an der Wand 'on the wall' and in other cases expresses the notion we express with 
at, as in 'an der Universitaet', 'at the university'.  While learning English,
such an individual is likely to use some English prepositions incorrectly, 
mapping them onto her own language's semantic inventory.  In the case of WS 
speakers, their inventory of phonological forms would correspond exactly to 
that of other English speakers, but the pattern of symbolization relations from these phonological forms to the semantic structures within the inventory of 
prepositional relationswould be different.  

      To explore one example, consider the usages C3 working out on a field 
and M1 being on a [birthday] party.  It is possible to imagine a semantic 
system in which the preposition on includes, among its family of senses, a
sense of merely 'being located', similar to English 'at'.  The surfacey or 
container-like properties of the landmark are just neglected in this meaning,
remaining unspecified in the semantic structure.  Thus being on a party is 
equivalent to conventional at a party, and being on a field is equivalent to
being in a field.

     So many prepositions are used in what seems to be a fashion corresponding
to conventional English usage, that this would in any case be only a partial 
solution for the WS data.  It seems difficult to account for omissions in this
way, although we could say that some relational meanings are simply left uncoded
in the linguistic system of WS subjects.  This is still unsatisfactory, however,given that the most frequent case of omission in the data we are concerned with 
alternates with a prepositional use (change a different color vs. change to a 
different color, in the chocolate story).  To confirm this solution we would 
need to examine a large data base and map the meanings of prepositions for each speaker, in order to discover whether the English phonological forms really 
have different semantic content for WS speakers.


3.3  WS subjects have mastered only an incomplete or partial version
     (some subset) of the semantic system of adults, but have normal processing


     There are several ways we can interpret this statement.  One is that WS
subjects have a semantic inventory for prepositions which is less rich than 
that of adults.  For example, they may have fewer meanings in a family of sensesfor a given preposition.  Another possibility is that they have not grasped all of the detail of a given meaning, and hence use a preposition which matches a 
less detailed subportion of the scene.  Consider Crystal's uses of 'through'.
'Through' presupposes 'in', but not vice versa.  Thus looking through something entails looking in it, as well as seeing the region of space on the other side 
of the object, by virtue of the transparency of the landmark object (I looked 
through the window) or by virtue of holes in the object (I looked through the 
hole in the wall at the construction site).  Perhaps Crystal has not grasped 
all of the detail of the meaning of through, specifically, the mention in the 
semantics of the prepositions of the space on the other side of the object.

     This situation is similar to ii., since incomplete grasp of the full
detail of the prepositional semantic inventory would result in WS speakers
having a different semantic inventory from normal speakers.  But this system 
would overlap with the adult system rather than being essentially different 
from it; it would be a subset of the adult system rather than a set with a 
distinctly different membership. 

      There is another way of interpreting iii.  which we feel is of significant promise for Crystal's data.  This has to do with the phenomenon of verb-
preposition pairs having highly specific semantics.  The proposal is that WS 
speakers become very familiar with such pairs by virtue of frequently hearing 
them in the environment; but because of their underdeveloped cognitive 
abilities, they do not master the use of these in accurate detail.

     In CG terms, this would consist of associating the phonological form
of these verb-preposition pairs with more semantic construals than is the case 
in the adult system; i.e., WS speakers overgeneralize these pairs.

     In the course of analyzing Crystal's data, we extracted a list of
verb-preposition pairs that she would be likely to hear frequently in her
everyday environment.  Note that these pairs are not necessarily used 
'correctly' in her data; but they are common pairings she might hear.  These 
are shown in Table 3 (Appendix).

     To explore one example, let us again consider Crystal's use of through, 
especially look through.  Upon thought, we realize that this expression occurs
quite often in the context of searching for lost items. Consider sentences like those in (6).

(6)
a. I looked through the pockets of every coat in the house before I
   finally found the keys in my dresser drawer.
b. Mom was looking through a pile of magazines for a recipe.
c. We looked through all the rooms in the house for the cat.
d. I looked through my wallet for my Visa card.
e. I looked in/*through my pocket for the lost ring.
f. I looked in/*through my pocket, my shoe and the gift box
   but I couldn't find the ring.

Now this expression involves a subtle semantic specification: the search
proceeds by progressing from one to another of a series of identical locations
((6)a., c.) or from one to another part of a single object ((6)d., the 
compartments in a wallet); or from one to another of a series of similar objects
(e.g. credit-card like objects in a wallet, (6)d.; magazines, (6)b.).  It does 
not work for an individual object ((6)e.) or a series of non-identical 
objects ((6)f.).  It is possible that Crystal has not picked up on this
subtlety and equates look through with look in, which can be used to describe a
visual search proceeding from one to another of a series of non-identical
locations.

     Several other unusually used pairs are amenable to such an explanation.
The semantic differences between listen for and listen to and call and call for are quite subtle and may also be missed by WS subjects.

     The pairing phenomenon could be used to explain pairs including material 
other than verbs as well.  Table 4 (Appendix) shows additional, non-verbal 
expressions including a preposition which might occur with high frequency in 
Crystal's language experience.  In the CG view, such pairs can easily achieve 
unit status.  And given the flexibility of semantic systems across individuals, it is very plausible that an individual could arrive at a different, less
specific semantic structure for a frequently-heard form.

     The incidence of such prepositional pairings in this sample of data is veryhigh.  There are 77 tokens of such pairings, constituting 59% of the total 
number of tokens of prepositions in Crystal's data, including many conformist 
uses.  If this explanation is correct, it could account for a good subset of 
non-conformist preposition use in WS speech.

     Interference from multi-word units without prepositions may also be
occurring.  Consider the alternations change __/to/into a different color.
Change color, change to, and change into are all likely to be units in English.
The latter two involve a change in the properties of the entity changing;
so does the first, though the property is specifically color.  It is easy to 
see how the semantic subtleties of the differences here could be missed.
It seems that the notion of overgeneralization of the meanings of multi-word
units would be a promising avenue to explore in examining WS data.


3.4 WS subjects have both a different or partial semantic system and
    different processing from normals


     We shall not explore this option here.  Differences in 
both representational 
structure and, say, associations to phonological forms, or activation patterns 
and activating connections between subparts of the linguistic system, would be 
likely to produce strongly aberrant uses of prepositions that would disrupt 
communication and be much more noticeable than what we find in this small data 
sample.  The data actually point to fairly subtle differences, and not gross 
differences of this nature.  So many of the WS usages of prepositions conform 
to convention, that we must assume a body of overlap in some sense.


4  Conclusion


     It is likely that several explanations will be required to account for the deviant preposition use found in the WS data.  Unusual use of prepositions in 
WS data put us on the alert for other ways in which their language use may miss subtle points of semantics/syntax.  CG provides a framework in which speakers 
having a lesser degree of cognitive sophistication than normals of their 
chronological age can still have a functional semantic system.  It remains to beseen whether examination on a large scale of WS language use will bear out the
'semantic deficit' explanations explored here.  It may turn out that WS subject have greater impairment in their linguistic capabilities than was previously
thought, and that the 'language faculty' has not been completely spared in this case.

Acknowledgements

This research was supported in part by the National Institutes of Health
Grants HD 26022 and DC 00146 to Ursula Bellugi and Edward S. Klima at
The Salk Institute for Biological Studies, La Jolla, California 92037.

Footnotes

1:  The WS subjects referred to in this paper were adolescents with IQ's
    between 40 and 60.

2:  Use of the term deviant here is without pejorative connotation.  We use it      to mean simply divergent from the norm of adult English usage.

3:  For a detailed discussion of the semantics of English spatial prepositions 
    in CG, see Hawkins (1984).

References

Bellugi, U., Bihrle, A., Jernigan, T., Trauner, D., & Doherty S. 1990.
        Neuropsychological, Neurological, and Neuroanatomical Profile of
        Williams Syndrome.  American Journal of Medical Genetics, 6, 115-125.

Bellugi, U., H. Sabo & J. Vaid. 1988. Spatial deficits in children with
	Williams Syndrome. In J. Stiles-Davis, M. Kritchevshy, & U. Bellugi, 
	eds., Spatial Cognition: Brain bases and development, 273-298. 
	Hillsdale, NJ: Lawrence Erlbaum. 

Caplan, David. 1987. Neurolinguistics and linguistic aphasiology.  Cambridge, 
	UK: Cambridge University Press.

Clark, E. & K. Carpenter. 1989. The notion of source in language acquisition.
	Language 65:1, 1-30.

Clark, H. & E. Clark. 1977. Psychology and language: An introduction to
	psycholinguistics. New York: Harcourt Brace Jovanovich.

Ellis, A. & A. Young. 1988. Human cognitive neuropsychology.  Hillsdale, NJ: 
	Lawrence Erlbaum.

Hawkins, Bruce. 1984. The semantics of English spatial prepositions.  Ph.D.
	thesis, University of California, San Diego.

Langacker, Ronald. 1987. Foundations of Cognitive Grammar, vol. I: Theoretical 
	prerequisites. Stanford, CA: Stanford University Press.

-----, forthcoming. Foundations of Cognitive Grammar, vol. II: Descriptive
	Application. Stanford, CA: Stanford University Press.

Pinker, Steven. 1991. Rules and Associations. Paper presented to the Seventeenth
	Annual meeting of the Berkeley Linguistics Society, February 15th, 1991.

Reilly, J., Klima, E.S., & Bellugi, U. 1990. Once More with Feeling: Affect
	and Language in Atypical Populations. Development and Psychopathology,
	Vol. 2, 4, 367-391.

 
				Table 1
	    Prepositions & frequency of occurrence in Crystal's data

Prep		times occurring		Prep		times occurring

about		     1			off			1
after		     1			on			7
as		     1			on top of		1
at		     7			out of			4
behind		     1			over			3
by		     2			through			9
for		     3			to			40
from		     4			under	 		1
in		     12			up			3
inside out	     1			upon			2
into		     2			with			5
like		     1 			of			18


				Table 2
		Deviant preposition uses in Crystal's data

F1	So, he looked everywhere, he looked through his boots,	[in]
F2	he looked through his slippers,				[in]
F3	he even looked through the table,			[in? under?]
F4	and the dog even looked through the jar, but he		[in]
	couldn't find his frog that he catched.
F5	There the dog has the jar in his face.			[on,or his face
						 		in the jar]
F6	and he's calling for the frog but he still can not 
	find him.						[zero]
F7	So what he did is he looked through an anthole		[in]
F8	He looked through the hollow tree			[in]
F9	he didn't see any frog through there.			[in]
F10	Then when suddenly he was yelling for the frog		[to?]
F11	standing on the tall rock, and then looking 
	through the trees					[in?]
F12	he got up from the water and the dog too		[in]
F13	he was trying __ listen for sompin			[to]
F14	he was trying listen for sompin				[to]
F15	He looked over [the log] and inside out it		[inside]
C1	You can be saved if the sun changes __ a 
	different color						[to]
C2	see if he can change the sun into a different color	[to]
C3	One of the leaders that has been working out on 
	a field.						[in]
C4	The Princess arrived to one of the leaders and told him	[at]
C5	If you want me to change the sun __ a different color	[to]
C6	What can I do to help you change the 
	sun __ a different color?				[to]
C7	use my magic powers to change the 
	sun __ a different color				[to]
C8	So they stand to each other.				[near, next to]
M1	This is your first experience of being 
	on a [birthday] party.					[at]
M2	He retired from working in that office.			[that job]





				Table 3
	   High-frequency verb-preposition pairs in Crystal's data


		look at			listen for
		wake up			look under	
		try to			climb over
		take care of		be amazed at
		call for		decide to
		have to			look in
		come after (chase)	get in
		run away from		be going to
		yell for		come in (colors,flavors)
		stand on		have fun with         
		look through		bow to
		lay on			come to see
		bark at			say to
		keep on VERB-ing	want to
		fall off		change to
		get up			change into     
		go to see


			  Table 4
      Other high-frequency expressions in Crystal's data

Once upon a time			one of
at all					a family of
all of a sudden				the end of
out of nowhere				experience of
on top of				retire from
inside out				selection of
as a family				story...about
used to be