Adjective order in English: A semantic account with

cross-linguistic applications
 

Enrica Rosato

Linguistics

Department of Philosophy, Carnegie Mellon University

Senior Honors Thesis, May 2013

Advisor: Dr. Thomas Werner  

  1  
 Adjective order in English: A semantic account with
cross-linguistic applications
Enrica Rosato

Table of Contents

1. Introduction …………………………………………...…………………………….………….. 4

2. Previous Studies ………………………………………………………………………………. 6

3 Proof of an underlying adjective order in English: “big red” vs. “red big …………..10

4. A semantic theory for the description of adjective order in English ………………...15

Sphere 1: The intrinsitivity of material and color and the optimization of cognitive efficiency in
the restriction of subsets ……………………………………………………………………………18

Sphere 2: Age and shape -- Internally comparative adjectives and internal relativity ……… 22

Sphere 3: Quality and size -- Externally comparative adjectives and speaker relativity …….24

Outside of the spheres: the scope-taking adjectives ……………………………………………26

Tricky cases: using multiple adjectives from the same category ………………………………29

5. Application of the theory cross-linguistically …………………………………………... 30

6. Conclusions …………………………………………………………………………………… 35

7. Acknowledgements ………………………………………………………………………….. 38

  2  
Abstract
 
Which is more correct, the “big fat cat” or the “fat big cat?” Why is a particular order

preferred? In English, established Phrase Structure rules place no limit on the number of

adjectives before a noun. The adjectives, however, cannot occur in just any order, and

native speakers of English have very particular intuitions about what order is more correct,

even if they have never been explicitly taught ordering rules. In this study, I seek to describe

the mechanics of an underlying adjective order in English and explore if the same principles

operate cross-linguistically. After outlining some previous work across disciplines on the

subject, I prove the existence of a preferred order using the results of searches from the

Corpus of Contemporary American English and the British National Corpus. Additionally, I

briefly discuss the prosodic differences between a given order and its alternative. Secondly,

I develop a semantic theory that describes how pre-nominal adjectives are ordered based

on their semantic properties, with adjectives that depict “intrinsic” properties closer to the

noun, and adjectives that are “speaker relative” in a more distant position. In the theory, the

use of multiple adjectives is described as being equivalent to a sequential series of

restrictions placed on the set of properties for a given noun. This allows for a change in

adjective order to affect the way in which we conceptualize of a noun, while also

establishing an underlying order that is the most cognitively efficient. Lastly, I apply the

theory to languages from different families, namely Italian, Sakha, Hebrew, and Welsh. My

findings demonstrate that a uniquely semantic theory is successful in describing what native

speakers perceive as the “proper” order of adjectives in a diverse group of languages.

  3  
1. Introduction

Native speaker intuitions, corpus searches, and prosodic features of ordered

adjectives in English all provide substantial evidence that the order of adjectives within a NP

is subject to certain restrictions. This idea has been explored by previous work on the

subject of adjective order beginning in the 1960s and heavily in the 1990s, and has been

reaffirmed in recent years by studies in Cognitive Science. The extent to which an

underlying order exists cross-linguistically has also been explored previously; however,

linguists have remained pretty consistently within the domain of syntactic analysis in the

cross-linguistic context. These syntactic descriptions seek to describe cross-linguistic

ordering variation by invoking syntactic movement principles, and have been successful in

doing so. However, to my knowledge, a semantic analysis for English adjective order has

not been revisited since the 1970s and has not been applied cross-linguistically. My goal is

not to simply describe the syntactic variation among adjective order across languages, but

to establish a unifying semantic account for how adjectives restrict a noun’s set of

properties, and the order in which they do so. My theory intends to focus on the mechanics

of underlying adjective ordering (AO) rules, and explores how these mechanics operate

cross-linguistically.

I have chosen English as the primary language of analysis for many reasons. Among

the most prevalent of these reasons is the fact that it is my native language, and I can

therefore trust my intuitions as a native speaker and the intuitions of my English-speaking

peers to be the basis for proper assessments of “correctness.” Secondly, English places

very few restrictions on adjectives in addition to the ordering restrictions; for example,

adjectives in English are pre-nominal only, with no declension system or gender and number

agreement rules. This makes English AO rules more important in terms of the relationship

between an adjective’s placement and the semantic properties it contributes to the NP, and

  4  
there are fewer variables to control for when analyzing the effects of an alternative ordering.

With a theory that accounts for AO and only AO, it is possible to see how a semantic

account behaves in languages where there are additional restrictions on the adjectives. The

additional languages I choose to explore are not thought to be historically related, and on

the surface, all treat the syntax of adjectives within a NP differently. Sakha (Yakut) follows a

similar structure to English. In Italian, however, adjectives occur both post-nominally and

pre-nominally, and in Welsh, all adjectives occur post-nominally. Hebrew is unique in that it

embeds the determiner, and the adjectives occur post-nominally. It is also important to note

that in this paper, I avoid discussing the role of the determiner or of quantifying adjectives

(many, few, etc.) and number. I also choose to refer to a string of adjectives and the noun

they modify as a NP, not a DP. This is done so that the emphasis of the discussion is on the

semantic value of the noun and attributive adjectives rather than on the role of the

determiner or any quantifiers.

I begin the discussion by citing previous work on the subject in the domains of

semantics, syntax, cognitive science, and from the perspective of functional grammar. In

part 3, I begin my own analysis by providing additional evidence for the existence of an

underlying AO in English using corpus data and a brief discussion of prosody. In part 4, I

begin my semantic description of the underlying order of English, where I establish my own

semantic categories for adjectives and place those into even broader categories, which I

refer to as “spheres.” I then attempt to explain, using a concept I call “cognitive efficiency,”

why altering an underlying order is inefficient both within a sphere and across sphere

boundaries without being given the proper contextual information. I discuss the semantic

categories of adjectives in terms of the “intrinsitivity” or “speaker relativity” of the properties

they denote, and I establish that adjectives denoting more intrinsic properties must occur

closer to the noun while those denoting relative properties occur farther away. Lastly, in part

  5  
5, I apply what I establish as the underlying order for English to the aforementioned

languages and compare it to previous analyses for AO in those languages. When I apply the

theory, I am mostly looking to see if the underlying order that I establish for English has

cross-linguistic applications.

2. Previous studies

The work that has been done to study the sequence of pre-nominal adjectives in

English was at its most prevalent in the 1960s, when it was first noted by psycholinguists

that speakers have a preference for certain orders. At the time, one of the most popular

strategies of contemporary psycholinguistics had been to interpret various laboratory

phenomena in terms of constructs derived from linguistic theory. At the forefront of this work

were three psycholinguists by the names of J.E. Martin, (1969), J.H. Danks, and S.

Glucksberg (1971). In his 1969 paper, Martin proposed a semantic rule for explaining

adjective ordering that was based on the qualities referred to by the adjectives- definiteness,

absoluteness, or intrinsicalness1. According to Martin, in comparison to size, color is more

definite in meaning, changes less from object to object (absoluteness), and is considered a

more intrinsic property of the object. Additionally, Martin claimed that the semantic

dimensions of definiteness and absoluteness determine the “accessibility” of adjectives,

which in turn determine their ordering. He defined “accessibility” in terms of how quickly

participants could produce an adjectival description of a physical stimulus. He found that the

response speed was correlated with ordering: The nearer an adjective was preferred to the

noun, the more “accessible” it was.

                                                                                                               
1  In this paper, I attempt to adapt and expand upon Martin’s “intrinsicalness” idea; instead I

refer to it as “intrinsitivity.”
 
  6  
 In their study, Danks and Glucksberg (1971) considered violations of adjective

ordering constraints by having participants complete a ranking test with six possible

permutations of three pre-nominal adjectives. The results showed that the position of the

adjective that was most closely related to an intrinsic property of the noun was the primary

determinant of acceptability: the closer it was to the noun, the higher the sentence was

ranked. This demonstrates a speaker preference to place adjectives that denote intrinsic

properties of a noun closer to the noun.

In 1985, M.A.K Halliday published his first edition of An Introduction to Functional

Grammar, where he provides a theory for adjective order that describes how adjectives

“decrease in specifying potential” and become “increasingly permanent as attributes” as

they approach the noun. Halliday’s concept of permanence is a way of re-framing Martin’s

idea of “intrinsicalness” from the functional perspective rather than from the perspective of

meaning. He also makes the claim that “the more permanent the attributes of a thing, the

less likely it is to identify in context.” This is similar to the parallels that Martine draws

between intrinsicalness and accessibility, although Halliday did not do a psychological test

to determine the validity of his assumptions about permanence.

In the 1990s, several studies were done on restrictions on the syntax of adjective

ordering cross-linguistically (Sproat and Shih 1991, Svenonius 1993, Bernstein 1993,

Cinque 1994, and Bouchard 1998, to name a few). These studies focused primarily on

describing the variation of the internal placement of constituents within DPs across

languages. Sproat and Shih (1991) looked at adjective ordering in Chinese, and found that it

was relatively free in comparison to English when the adjectives were marked by particles.

In general, their conclusions lead them to lean against the idea of a universal semantic

description for adjective order. Despite this conclusion, they still established a general

ordering hierarchy, Quality > Size > Shape > Color > Provenance. Cinque (1994) modified

  7  
this order to be slightly different, establishing broader semantic categories and claiming that

they exist in the following order: Possesive > Speaker-oriented > Subject-oriented >

Manner/Thematic. Cinque focused on the relationship between the ordering restrictions on

Germanic and Romance languages, giving a syntactic explanation for the differences. He

introduced the idea that in Italian (a language with pre and post-nominal adjectives), it is the

noun that moves within the DP, with the adjectives remaining in the same order relative to

each other as they occur in English.

The idea of internal movement of the noun within a DP was controversial, and many

linguists published critiques of Cinque’s work into the late 1990s. Most recently, in 2006, a

linguist named David Willis published a critique of N-raising movement theory as it would

apply to the Celtic languages, namely Welsh. Willis cited examples of Welsh NPs

demonstrating both “mirror image” and “universal” adjective orders, making it nearly

impossible for a coherent syntactic explanation to work consistently in the language. In

general, the discussion within the linguistics community about AO in recent years has been

focused more on cross-linguistic inconsistencies regarding the placement of the noun in

multi-adjective NPs, with most of the questions being about the usefulness of movement

theories in explaining these inconsistencies. In my own paper, I seek to return to the

development of a uniquely semantic theory that explains the underlying ways in which

language functions to describe objects in the physical world. With a more complete

understanding of these basic concepts, I believe movement theories will be minimally

necessary when attempting to explain cross-linguistic variation in adjective use.

In recent years, neurologist David Kemmerer at Purdue University has used more

modern methods to explore some of the same questions about the cognition of AO

restrictions that were being asked in the early 1970s. In 2000, he tested the supposition that

the features of adjective meaning that constrain their linear order reside at a different level of

  8  
mental representation than the features that are invisible to syntax. “If this is true,” he

claimed, “then it is possible that the two components of meaning could be impaired

independently of each other by brain damage (Kemmerer 2000).” In his experiment, six of

16 brain-damaged subjects failed a test that assessed their knowledge of the semantic

principles that determine pre-nominal adjective order. However, all of the subjects

performed within normal limits on a test that assessed their knowledge of the grammatically

irrelevant perceptual and conceptual features of the same adjectives that were used in the

first test, and all of the subjects performed well on a test that assessed their knowledge of

the basic syntactic structure of English NPs. This study therefore provides support for the

view that there is an independent level of representation in the mind/brain for grammatical

semantics. In 2007, Kemmerer used ERPs (event-related brain potentials) to assess how

participants reacted to “proper,” reversed, and contradictory adjective pairs when reading

aloud. Increased brain activity was recorded for both the reversed and contradictory

pairings. Kemmerer believed this to be a reflection of semantic and syntactic aspects of a

temporary reanalysis of the adjective order construction, providing further evidence that

certain adjective orders require more cognitive effort.

The topic of proper adjective ordering has spanned across disciplines for many

decades. All of the work I have mentioned has informed my desire to develop my own

semantic account for the proper ordering of adjectives in English, and attempt to apply this

across languages. While the work of semanticists in the late 1960’s has inspired me to

develop my own semantic theory for proper AO, Kemmerer’s work has inspired me to use

this semantic theory to develop my own ideas for why certain pairings are optimally

cognitively efficient. The work of Cinque, Willis, and other linguists in the 1990s and 2000s

has inspired me to attempt to apply my theories across languages from multiple language

  9  
families. Additionally, I have chosen to incorporate prosodic analysis and some methodology

from corpus linguistics as proof for the existence of an underlying order.

3. Proof of an underlying adjective order in English: “big red” vs. “red big”

The existence of a preferred underlying adjective order of English can be proved by

both corpus analysis and prosodic analysis. Let us, for the sake of simplicity, take a

common case of two adjectives depicting size and color, “big” and “red.” Case (a.) below

shows the adjectives occurring in their more common underlying order and case (b.)

demonstrates the inverse, which is perceived by native speakers of English to be less

correct. This would make SIZE > COLOR the proper underlying pre-nominal order. Using a

corpus, we can determine and compare the ways in which the cases are actually used.

(a). the big red house

(b). # the red big house

My hypothesis for the corpus analysis is that “big red” will occur at a very high

frequency, both in written and spoken language. I hypothesize that “red big” will only occur

in contexts of spoken language where the speaker is giving an off-the-cuff description of an

event or object. Within this description, I would imagine the speaker to pause often and

insert discourse makers such as “um” or “like.” Additionally, this type of description may elicit

such an order because the object being described is being conceptualized or recalled at the

same rate at which language is being produced. If the object were to be already

conceptualized and simply being described using language, I predict that the underlying

order would surface.

  10  
 Using COCA, the Corpus of Contemporary American English, I searched for

occurrences of “big red [noun]” versus occurrences of “red big [noun].” “Big red” yielded over

382 tokens in many contexts, and “red big” yielded 0 tokens. I then moved on to the BNC

(British National Corpus). When I searched for “red big” on the BNC, I found only one

occurrence of “red big.” It was from a speaker categorized as “FLP,” which is defined in the

corpus as a speaker within a group of 10 Scottish women, who are having a discussion

about weddings. The token was recorded on an unknown date and was transcribed as

follows:

Example i.
FLP: And their bride has to wear er, all in red and er the bridegroo er groom
has to wear a long costume with a red big flowers in front i aha and then they
get married and there's erm band, the Chinese traditional band with drums and
trumpets blowing all the time and er, all the guests have a very nice time.

Much as I expected, the speaker is in the process of describing a wedding. Her

utterance is filled with what the transcriber writes as “er,” the British equivalent to the

modern American “um.” In this context of off-the-cuff description, the Scottish woman says

that the groom has to wear a “long costume with a red big flowers.” Unfortunately, we do not

have a recording of this utterance and the transcriber did not annotate it with the prosodic

elements that would provide it with some spoken context. Additionally, it is possible that the

previous mention of the word “red” in the phrase “all in red” made it contextually reasonable

on a discourse-relevance level to list “red” first in the order. Another possible explanation for

why this speaker may have ordered her adjectives in this manner is that in her dialect of

Scottish English, it is possible that this is acceptable, although no other Scottish speakers in

the corpus produced an inverse order and I found no evidence for this fact.

All other occurrences of these adjectives adjacent to one another in the BNC yielded

case (a) order. Below are just some of the contexts in which case (a) was recorded. None of

  11  
the instances in which “big red” occurred were uttered by a Scottish woman, so there is a

possibility that my dialect theory may not be entirely invalid.

ii. (2619) Now why would Kurt be running around Soho with a big red woolly
jumper on and having intense conversations with lampposts?
iii. (303) Inside were all the things they had asked for, and some they had not-
some wine, two chickens, twelve big red roses.
iv. (901) Love does not involve giving fancy parcels tied up with big red bows.
v. (5406) And it, and it like, it run down me face there and I had a big red mark
on me face.

Only one of the above cases involves the detailed description of something, and that

is example ii. (303). According to the BNC, this example is from a book called The Railway

Children published in 1993. Because this is written data, it cannot be considered as

contextually similar to the Scottish example. Example v. (5406) was uttered by an Irish

woman and recorded in 1992. It contains commas, which indicates pause, and in addition,

the transcriber decided to include the discourse marker “like.” This case of spoken data,

although not a case of improvised description, clearly shows the preferred adjective order.

While it was clear that color consistently occurred before size in both COCA and the

BNC, with the only example of the inverse as my hypothesis predicted, I decided to

investigate the other semantic categories of the adjectives to be sure it wasn’t a unique

case. I did searches for a variety of categories and their respective inverses, and found that

although the inverses of the preferred order do exist in the corpus, they are fewer in number

and can be explained in context. Figure 3.1 below displays the occurrences of certain

adjective pairs when searched in COCA, and demonstrates that an order is in fact preferred

by speakers in both spoken and written contexts, while the inverse occurs consistently fewer

than 15 times in the entire corpus. This is evidence that the preferred order is an underlying

order, and the inverses likely have a different semantic meaning as a consequence of

switching the order. The semantic differences, as well as other consequences, can be

  12  
explained when the inverse cases are analyzed in their contextual environments. My theory

will attempt to explain why an inverse order may produce a certain semantic shift, and why

this shift is to be expected, in section 4.

Figure 3-1. Adjective pairs and their occurrences in COCA

Adjectives COCA occurrences

color material inverse
red brick 217 7
white cotton 262 4
age color
old red 114 1
new white 157 6
quality age
beautiful old 101 2
wild new 24 16
size age
big old 247 17
small young 8 2
size quality
big beautiful 33 13
small fat 3 0

I would now like to return again to the analysis of “big red” vs. “red big,” and discuss

the prosodic differences between these cases; that is, if asked to read both cases out loud,

what prosodic differences would an average English speaker, selected at random, produce?

To answer this question, I created sentences with both case (a) and (b), which are designed

to be read aloud in sequence without the reader having seen the paragraph before. This

eliminates an improvisation variable, but will control for prosody and hopefully illuminate an

underlying distinction between the two cases. Below is the paragraph to be read aloud. I

asked a participant to read the paragraph aloud into Praat. The spectrogram for her reading

is shown on the next page. I will look specifically at intonation and the temporal spaces

between words to determine how the participant may have reacted to the adjective orders.

“And in the morning, I walked down the road to the big red house on the
corner. I knocked on the door, and a woman wearing a red big hat entered.
She smiled and said good morning.”

  13  
 On the above spectrogram, the participant’s utterance of “big red house” remains

relatively consistent in pitch with each word (pitch is displayed by the blue line). There is a

longer pause between red and house than there is between big and red, but it is hard to

determine the significance of this. Later in the paragraph, when the participant utters “red

big hat,” the pitch pattern is quite different. “Red” is at the same basic pitch level that it was

at in the utterance of “big red house,” which may suggest that certain adjectives have

common pitch levels based on their estimated distance from the noun, which would be

interesting to investigate. However, in the second case, the adjective “big” drops slightly in

pitch when it occurs after “red”, creating a pitch contrast. This signifies intent to distinguish

the adjectives from one another, which may suggest that the speaker is adjusting for an

adjective ordering error by making the adjectives audibly separate entities.

In addition to the creation of pitch contrast, the speaker also pauses for a significantly

longer amount of time between adjectives in case (b). This suggests, once again, the

speaker’s attempt to either adjust for the ordering error or create temporal distance between

the two adjectives in an attempt to express their “list-like” nature. In the orthography, we

  14  
represent this “list-like” nature using commas, which allow us to freely order adjectives by

making them independent syntactic entities that modify the noun equally and simultaneously

rather than sequentially according to ordering restrictions.

In conclusion, it is possible to find proof of the prescribed adjective order not only in

corpus data of American and British English, but also in the prosodic features of both cases

when they are read aloud by an unsuspecting reader. These are two very different

supporting cases for the existence of a “correct” order of adjectives in English, which

strengthens the argument that a correct order indeed exists.

4. A semantic theory for the description of adjective order in English

A semantic account of adjective ordering requires placing adjectives into semantic

categories, which has been done and made available to learners of English by people who

have developed grade-school textbooks and instructional materials. Grammarians have

worked on this topic before, but adjective order seems to have been taught only in recent

years to grade-school students and people acquiring English as a Second Language. This

suggests that classifying adjectives into semantic categories is also a rather recent teaching

tool for language instruction. The Internet, for example, has numerous sites dedicated to the

instruction of English and English grammar which all refer to adjective ordering as “rule-

based,” claiming that there is a specific order which must be learned in order to have a full

grammatical understanding of the English language. According to the British Council2, a

British organization for cultural relations, educational opportunities, and the instruction of

                                                                                                               
2  http://learnenglish.britishcouncil.org/en/english-grammar/adjectives/order-adjectives

  15  
English for children, proper adjective order in English (based on the semantic categories of

the adjectives) is as follows, increasing in proximity to the noun from 1-83.

1. general opinion
2. specific opinion
3. size
4. shape
5. age
6. color
7. origin (nationality)
8. material
(noun)

There are issues, however, with the order of the first three categories. The first issue lies

in understanding the distinction between “general” and “specific” opinion. What is the

difference? According to the British Council, a “general opinion” adjective would be

something like “nice” or “interesting,” whereas a “specific” opinion would be an adjective like

“beautiful” or “curious.” What makes “interesting” a more general opinion than “beautiful?”

It’s also difficult to classify “nice” as a general opinion, when it seems to have semantic

variation; it can mean “pleasant”, or it can be more specific and mean “kind” or “pretty.” The

distinction is theoretically difficult to grasp, especially when we attempt to apply this order to

real world examples, as I will attempt to do. The second issue is with the relative order of

shape and age; in my modified order, I will switch them. Additionally, “size” seems to be

incorrectly placed. According to the British Council, the order in 1(a) would be correct, and

the order in 1(b) would be incorrect.

1. (a) the nice beautiful big house

(b) # the nice big beautiful house

                                                                                                               
3  
This order is modified by the British Council from Grammarian R.M.W. Dixon, who discusses the
semantic properties of adjectives and their proper order in his paper, "Where Have all the Adjectives
Gone?" Studies in Language 1, no. 1 (1977): 19-80.
 
  16  
 My intuitions as a native speaker of English (and the intuitions of my English-speaking

peers) tell me that 1(b) is actually the correct order. I affirmed these intuitions when I

conducted a corpus search for “big beautiful” versus its inverse (see figure 3-1 on page 8),

which yielded 33 occurrences of “big beautiful” and only 13 of its inverse. Additionally, “nice”

is an interesting case, as it changes its semantic value depending on its relationship to the

size adjective. For example, “nice” in “the nice big house” carries the meaning of general

appeal, with the house’s size is contributing to that general appeal. However, “nice” in “the

big nice house” seems to mean something more specific, perhaps in reference to the

house’s aesthetic appeal. Additionally, “the big nice house” seems to imply that there is a

group of aesthetically appealing houses, and the speaker is attempting to identify the big

one. Other adjectives, which Cinque (1994) refers to as “operators” behave similarly to

adjectives such as “nice.” These adjectives include “former” and “alleged,” and are able to

move more freely within a multiple-adjective NP, depending on the desired semantic

outcome. The British Council’s order does not include operators.

We therefore have two similar categories, one that is not mentioned by the British

Council (“operators”) and the other (“specific opinion”), which seems to include a larger

variety of adjectives. These two categories seem to have two things in common: a scope-

taking quality and an ability to move more freely, making them different from the other

categories. With this generalization, as well as the changes I made to the ordering of age,

shape, and size, I would like to suggest the following modification to the British Council’s

order.

1. scope-taking (take all adjectives) (alleged, former, nice)

2. size (big, small, fat, skinny, tall, short)
3. quality (formerly “specific opinion”) (beautiful, ugly, silly, *little, *young)
4. age (old, new)
5. shape (square, round, rectangular)
6. color
7. origin (nationality)

  17  
 8. material
(noun)

*Adjectives such as “little” and “young” can be a part of multiple categories, although not
simltaneously. This will be discussed later.

Let us take this modified order to be the proper underlying order. We then have to

explain the semantic shifts that occur when adjectives are switched, as well as account for

why this order can be taken to be the underlying one. At first glance, the order is seemingly

arbitrary, but the above semantic categories can be farther consolidated into larger

categories. These larger categories can be thought of as spheres in which the sub-

categories lie, with the noun at the core of these spheres. If we begin with the noun and

work backwards, we can make observations that can help us to establish our innermost

sphere. Firstly, the categories closet to the noun, color, origin, and material, are very rarely

found out of order (this can be affirmed by the COCA search). Secondly, these three

categories attribute intrinsic properties to a noun, that is, they denote properties that are

inherent in the object’s physical existence (For example, an object cannot exist without

being made of something). Lastly, as these three categories increase in distance from the

noun, it is apparent that they become less intrinsic. We can therefore take color, origin, and

material to make up our innermost sphere containing the adjectives that attribute intrinsic

properties to the core noun.

Sphere 1: The intrinsitivity of material and color and the optimization of cognitive

efficiency in the restriction of subsets

To explore these observations, let us look at a simple case that makes use of the

“material” and “color” categories, “red brick house.” Let us first establish that “red brick

house” is the preferred order, where “brick red house” is not. If we look at the properties that

each of these adjectives attribute to the house, we can see that the material (“brick,” in this

  18  
case) is a slightly more intrinsic property of a house, or any noun, for that matter. To test the

intrinsitivity property of adjectives that attribute “material” to a noun, we can attempt to

imagine a house without the property of being made of something. This seems impossible to

do, because a physical object (such as a house), by nature, must be made of something. If

someone were to ask a group of people to picture a house, one person may picture a

wooden house, while another person may picture a brick house, but it would be impossible

for either person to picture a material-less house. Therefore, we can consider material an

intrinsic property of a physical object. Additionally, both color and material require no

additional contextual information to be determined; however, we do require context to

determine if something is big, small, beautiful, or ugly.

If we are to be in keeping with the order I have suggested and the observation that

the adjectives denote decreasingly intrinsic properties as they get farther from the noun, the

question then becomes, “What makes color a less intrinsic property than material?” If we

use the same diagnostic above and attempt to picture a “colorless” house, some would

argue that it is impossible to do so. However, it seems more plausible to imagine a

“colorless” house than a “material-less” house. “Colorless” is a word in English that is used

to describe something that is lacking color, so it seems that we are in fact capable detaching

the property of color from an object successfully without losing our entire concept of it.

Detaching the property of material from a physical object seems much less plausible. In fact,

if I were to ask you to picture a “colorless house,” it is likely you picture a glass house. The

house is still made of something, but it does not have a color.

In an attempt to explain why to ordering rules are so strict within this first sphere, let

us think of the order of the adjectives as the order in which we restrict the noun’s properties

into subsets, starting from the adjective closest to the noun and working outward. If we do

this, then when we conceptualize of the NP “the red brick house,” the order in which we

  19  
restrict the set of houses is indicative of the order in which we attribute properties to the

noun, and thus represents the order of the adjectives. We begin with the noun, which

denotes the set of all houses. Then we begin to restrict this set with increasingly non-

intrinsic properties. First, we create a subset of houses that are made of brick. Then, we

take brick houses of varying colors and isolate the red ones. By doing this, we are being

more efficient in how we describe the object we wish to refer to. If we were to first isolate the

red houses, we would not only conceptualize houses that are red, but also be cognitively

required to assign some material (or materials) to them, since it is impossible to

conceptualize of a material-less house. In short, it seems much more plausible for our brains

to isolate a set of colorless brick houses than a set of red material-less houses. To be

maximally cognitively efficient, it makes sense to begin by making a subset of brick houses

of no particular color, and then make yet another subset of the red ones. When we reverse

the order, the result is a less cognitively efficient concept; this is likely why reversing the

order within sphere 1 is so rare.

When the order is reversed, however, what are the semantic consequences?

Additionally, how can we conceptualize of a “brick red house” without being cognitively

inefficient? If we think of a context in which saying “the brick red house” is acceptable, it is

required that the discourse context makes it previously clear that we were already

discussing a set of red houses. It seems plausible for someone to say, “We passed several

red houses on our tour of Brussells. My favorite was the brick red house.” Basically, saying

the “brick red house” in a grammatically correct fashion simply requires that you begin with a

set of red houses before attributing other properties, such as material, to the set. In doing

this, you also avoid the problem of cognitive inefficiency. Diagrams 1, 2, and 3 below

demonstrate this process. In diagram 1, the context-free underlying order, which is also the

most cognitively efficient, is depicted. In diagram 2, the context-free reversed order is

  20  
shown, demonstrating the cognitive inefficiency of conceptualizing of the adjectives in that

order. Lastly, diagram 3 demonstrates how in context, this problem can be fixed by altering

the starting point form which we begin restricting subsets of properties. For the purpose of

including another contrastive semantic category from another sphere, the phrase shown

below is “big red brick house.”

  21  
Sphere 2: Age and shape -- Internally comparative adjectives and internal relativity

Color is the final semantic subcategory in Sphere 1, making age and shape the next

two semantic categories as we work outward from the noun. Age and shape, in terms of

their relationship to sphere 1, are semantically less intrinsic than properties such as material

and color, but this is not what places them in a unique sphere. What distinguishes age and

shape from the categories in sphere 1 is their ability to be internally comparative. Unlike

material, origin, and color, age and shape can be comparatives; for example, we are

capable of saying that one thing is “rounder” or “older” in comparison to another thing, or to

itself at a previous point in time. We are not capable, however, of saying something is “more

brick.” We can’t really say that something is “more French” than something else; if we do,

we mean French-like in quality rather than the property of being from the country of France.

Something that distinguishes age and color from the remaining semantic categories

is that they are not semantically deictic. This is what I have intended to convey by

describing these adjective as internally comparative and internally relative; such adjectives

do not require information from an external context to be determined as true or false, and

additionally, they require no information about the speaker’s beliefs. For example, if a man

were to be alone in an empty room with no other people or things to compare him to, one

would still be able to assess his age. Signs of age can exist independent of any context,

allowing us to be able to say a man is “old” without needing any comparison class. The

same is true for shape. If a table were to be sitting alone in an empty room, we would still be

able to determine if it were square, rectangular, or round.

Unlike in sphere 1, sphere 2 adjectives do not possess the same necessary intrinsic

properties and it is therefore not obvious that reversing the underlying order within sphere 2

is cognitively inefficient. For example, saying “the old round table” is not necessarily more

cognitively efficient than saying the “round old table,” it just elicits a different cognitive

  22  
process in terms of the sequence of which we denote the set of properties of the table. How

do we know then, that age comes before shape in an underlying pre-nominal adjective

sequence? Speaker intuitions tell us that “it depends on what you are trying to say,” but a

COCA search for “old square (noun)” yields 9 occurrences, whereas the inverse yields 1.

This shows us that whether speakers are aware of it or not, they are demonstrating a slight

preference for saying age before shape, so this must be what feels natural.

How then, do we explain this preference? Because sphere 2 adjectives are

characterized by their internal relativity, I believe that it is logical to assume that like the

intinsitivity property in sphere 1, the internal relativity property is also scalar. It appears that

as we get farther from the noun, the adjectives decrease in intrinsitivity and increase in

relativity. Shape is a less relative property than age, because it requires no temporal scale.

Age is a property that is impossible to determine without the presence of a time, but it is

internally relative (rather than externally relative) because it relates an object to the quantity

of time that the object itself has been in existence. Determining shape requires no such

information, and is therefore a more intrinsic property, placing it closer to the noun than age.

If we were to cross the sphere boundary, we would see that reversing the order

seems much less natural. If we take a sphere 1 adjective, for example, “red,” and inverse it

with a sphere 2 adjective, “old,” the result is a non-preferred NP. Saying the “old red table” is

clearly preferred to saying the “red old table,” and this is likely due to the fact that the

reversed order involves placing a sphere 1 (intrinsic) adjective farther from the noun that a

sphere 2 (internally comparative) adjective. The semantic consequences of reversing the

sphere order are similar to the semantic consequences of reversing the order of the

subcategories within sphere 1. If reversing the order, it is necessary to have a context in

which the starting set is pre-established before additional properties restrict it. We have thus

far established the following order of adjectives before a noun.

  23  
 ((Sphere 2 AGE > SHAPE) (Sphere 1 COLOR > ORIGIN > MATERIAL) (noun))

Sphere 3: Quality and size -- Externally comparative adjectives and speaker relativity

Continuing outward from the noun, the next two semantic categories are quality (what

the British Council referred to as “specific opinion”) and size. I made the decision to

eliminate matters of opinion; however, adjectives that denote “quality” can, in fact, be words

that denote the opinion of the speaker. Unlike the adjectives in the first two spheres, quality

words such as “beautiful,” “funny,” and “interesting,” require a speaker who has access to an

external comparative class. Words such as this are difficult to define without also

understanding their opposites. How can a speaker understand if something is beautiful,

funny, or interesting, without first being informed by the context of what makes something

beautiful, funny, or interesting and what does not? To a large degree, the definition of quality

adjectives are fixed by societal norms and are established a community of speakers; for

example, what is “beautiful” to Americans is very different than what is “beautiful” to other

cultures. Quality adjectives, to a degree, are a matter of the speaker’s relationship to the

object, and the speaker requires an external context to determine the quality of that object.

This need for an external context is what makes these adjectives externally comparative

rather than internally comparative, making them different then sphere 1 or sphere 2

adjectives. Additionally, we can say that these sphere 3 adjectives are speaker relative; that

is, they depend on the context of the speaker and rely on the speaker’s observations to be

determined as true or false.

Even more externally comparative (and therefore speaker relative) than quality is

size. Let us return briefly to the “alone in an empty room” test. If an observer saw an object

alone in a colorless, empty room, he would be able to assess its material, color, and shape

  24  
without any other information. Age would be more difficult to determine, since the observer

would need some reference to time and the object’s internal “lifespan.” Assessments of

quality would require the speaker to contribute some information from his own external

context from outside of the empty room, and size would be nearly impossible to assess.

Without another object next to it, or an idea of how “big” the room is, the size of that object

would not be able to be determined. The speaker would need not only his own information,

but any assessment of size would require an external comparison class. If that object were

an elephant or a mouse, we would have no way of knowing if either creature were big or

small without having other creatures to compare it to or the speaker itself. We can therefore

say that size, like quality, is also externally comparative and speaker relative. This would

result in it being the most distant adjective from the noun in almost every context-free case.

According to COCA, size is almost always preferred in the external-most position as

possible from a noun. To assess an object’s size prior to assessing its other more intrinsic

properties would be cognitively inefficient.

This leaves us with a final explanation for the underlying order of pre-nominal

adjectives. The noun is preceded by those adjectives that contribute the most intrinsic

properties, and as we work outward from the noun, the adjectives become increasingly

comparative, first internally (requiring no or minimal outside information), and then externally

(requiring information from an external context). Additionally, intrinsitivity decreases as we

get farther from the noun. The closer the adjectives are to the noun, the less they have to do

with the speaker and his or her external world and the more they have to do with the intrinsic

properties of the noun’s existence. The diagram on the next page displays this relationship,

while also demonstrating the order of the spheres and their subcategories. The final

category, the scope-taking adjectives, will be explained in the next section.

  25  
Outside of the spheres: the scope-taking adjectives

The final subcategory, which the British Council referred to as the “general opinion”

category, lies outside of the spheres. I have chosen to call this category the “scope-taking”

category, because adjectives of this sort have the capability of moving freely between sub-

categories depending on the intended meaning. These adjectives, when context-free, can

usually be found as far from the noun as possible, since they are capable of modifying the

  26  
entire NP and all of the adjectives within it. In some ways, they are as “adverbial” as

possible without being actual adverbs. “Former and alleged” fit well into this category but are

operators, and have a stricter semantic value. “The former obese mayor” means that he is

one in a series of obese mayors, and he is the previous one in the set to which the speaker

is referring. However, “the obese former mayor” means that there are a series of former

mayors, and he is the obese one. Non-operator scope-taking adjectives operate a similar

way, although they have the ability to shift semantically, as demonstrated below.

(1) the nice big long French baguette

(2) the big nice long French baguette
(3) the big long nice French baguette

“Nice,” when in the left-most position, almost has an adverbial-like behavior, in that it

seems to modify not just the baguette, but also all of its other qualities. Anywhere it goes, it

seems to modify all of the adjectives that come after it, making it scope-taking. The farther it

gets from the noun, however, the less intrinsic the properties that it denotes become. In its

most distant position, “nice” has a more general externally relative meaning, in that it

compares the baguette not just to other baguettes, but other objects in the speaker’s

context. As it approaches the noun, it takes on a more noun-relative scope, assessing its

aesthetic quality. Many “quality” adjectives can be removed from their subcategory and be

brought outside of the sphere for the purpose of comparing the noun to a broader external

comparative class. In some cases, doing this actually changes the semantic value of the

adjective. “Interesting,” and “curious” are some of these, as demonstrated below, when

interacting with the other categories from spheres 1, 2, and 3. In the below examples, I have

used size, a sphere 3 category.

(4) the interesting small paper

(5) the small interesting paper

  27  
 In case (4), the paper is small, and is interesting in comparison to other things in the

world. In case (5), the paper is interesting for a paper, and is small in size.

(6) the curious little man

(7) the little curious man

Between cases (6) and (7), curious has a more distinct semantic difference. In case

(6), curious means odd or bizarre, which is an assessment that relates the man to a broader

class of things in the world. Case (7) prompts a different understanding of curious, meaning

inquisitive or questioning, which are properties that are more person-relative. He is still small

in stature in both cases; it is only the semantic value of “curious” that changes, suggesting

that “curious” is the one switching semantic categories, rather than the size adjective.

In some cases, an inverse order is actually the convention, because certain

adjectives have come to take on a popularized different semantic value depending on the

context in which they appear. If searched in a corpus, these conventional adjective pairs

would likely occur with more frequency. These cases, on the surface, appear to be

exceptions to the underling AO, but in reality, such an inverse order actually results in a

semantic shift, making the out-of-place adjective more fitting with a different semantic

category. Examples of this usually involve the size adjective “little,” because “little” is used to

mean more than just small in size; it has the additional meaning of “cutesy” or “child-like,”

and appears to only occur in the inverse position in cases such as “pretty little liars,” or “silly

little boy,” where it does not entirely mean physically small in size. Another case that

demonstrates a semantic shift is the use of “old” in “big old dog.” This case, unlike the

previous case, is not a surface violation of the underlying AO, but “old” does display a

conventional semantic shift from the “age” category to the “quality” category. When it occurs

after “big,” old does not mean old in age, but tends to retain a more quality-oriented meaning

of familiarity and strength. If we contrast this with “small old woman,” we see “old” shift back

  28  
into the age category. We will be able to explore semantic shifts such as this with regards to

placement cross-linguistically when we attempt to apply the theory to Italian.

Tricky cases: using multiple adjectives from the same subcategory

The next question becomes, “Are there ordering restrictions when we use multiple

adjectives from the same subcategory?” First it is important to make note of the fact that we

can only use multiple adjectives from the same category if they are age, quality, or size

adjectives. It does not make sense to say “the red blue car,” the brick glass house” or the

“rectangular circular table.” To summarize, sphere 1 subcategories do not have the ability to

be repeated, but sphere 2 and sphere 3 adjectives do. Let us look at the effects of this in the

following examples below.

Age:
(8) my new old car
MEANING: The car is old, but I have recently acquired it.
(9) my old new car
MEANING: I once had a car that was new, and now I do not.

(10) the young old man

MEANING: The man is old, but young for an old man.
(11) the old young man
MEANING: The man is young, but old for a young man.

Size:
(12) the big small dog
MEANING: The dog is small, but big for a small dog.
(13) the small big dog
MEANING: The dog is big, but small for a big dog.

Quality:
(14) the lazy crazy man
MEANING: the man is mentally disturbed, and generally lacks motivation
(15) the crazy lazy man
MEANING: the man is by nature unmotivated, and does things that are bizarre, but is not
necessarily mentally disturbed.

In the above examples, there does not appear to be a preferred AO, therefore, there

are no apparent AO restrictions on the use of two adjectives from the same subcategory.

  29  
There are, however, semantic differences between the pairs above, which can be

summarized as such: the adjective closest to the noun attributes internal properties to the

noun, and the adjective farther away from the noun distinguishes it from a broader external

context or set of nouns. This is consistent with my theory that as an adjective gets farther

from the noun, intrinsitivity decreases and speaker relativity increases. The adjectives

farther from the noun are externally comparative, while the adjectives closer to the noun are

internally comparative. In sphere 1, the adjectives are not comparative at all.

5. Application of the theory cross-linguistically

Sakha (Yakut)

While I have very limited resources with regards to the Sakha language, I have spent

a few months studying it rather closely and have had interactions with a native speaker.

Sakha, or Yakut, is a Turkic language with around 360,000 native speakers, spoken in the

Sakha Republic of Russia in northeastern Siberia. It is an agglutinating language with a

SVO word order. It demonstrates vowel harmony, and has a very rich case system, but no

grammatical gender.4 In my study of the language, I have seen adjectives primarily in

predicative contexts. However, when I asked a native speaker to tell me how to say, “small

white clouds” she told me the following.

1. Кыра үрүҥ былыбыпар

/kɯra yryŋ bɯlɯbɯpar/
“small white clouds”

The order demonstrated above is the same as it would be in English (SIZE > COLOR

> noun). To make sure that this order is actually also preferred in Sakha, I asked the native

speaker to tell me what it would mean if you said “white small clouds (in Sakha),” instead.

She said it to herself a few times, and immediately concluded that “No, it sounds wrong,”
                                                                                                               
4
http://en.wikipedia.org/wiki/Sakha_language#cite_ref-Krueger_3-0
  30  
just as a native speaker of English would react to such a change. While this is only one

native speaker’s opinion and one example, it does provide a small amount of evidence that

the same semantic restrictions in place in English have the ability to be in place in other

languages, even if the language has absolutely no genetic relationship to or history of

contact with English. This suggests that these semantic restrictions are underlying.

Italian

Italian is a Romance language, and was the primary language of study (in

comparison to English) for Cinque (1994) when he developed his internal N-raising theory.

Italian has both pre and post-nominal adjectives, but as we will see, they exhibit similar

semantic shifts that English adjectives do depending on our placement. One thing that

makes Italian interesting is its “mirror image” post-nominal ordering. For example, I came

across the following text on a wine bottle. The bottle was imported from Italy and

consequently translated from Italian to English.

The English on the label reads “Red Dry Wine,” which demonstrates an incorrect

adjective order. In Italian, this order (if read from left to right) is technically correct. Let us

compare the correct order in English with the correct order in Italian.

a. English: dry red wine

b. Italian: vino rosso secco
English gloss: wine red dry

  31  
 In correct Italian, we see that “red dry,” or “rosso secco,” is actually the correct order

when literally translated from Italian to English and read from right to left; however, the

adjectives follow the noun rather than preceding it. If we then consider the adjectives’

placements in relation to the noun in terms of their distances from it, it appears that in this

case, we have contrasting headedness rather than contrasting adjective order. More

specifically, we see a head-final noun phrase (in English) versus a head-initial noun phrase

(in Italian). This difference would result in a “mirrored” order for all words in the noun phrase

when translating from English to Italian. We can then conclude that the placement of the

adjectives in relation to the noun in a noun phrase is in fact identical in English and Italian,

with the word “red” (denoting the sphere 1 color adjective) occurring closest to the noun, and

the word for “dry” (denoting the sphere 3 quality adjective) occurring farther away.

If this is indeed the case, the wine bottle above demonstrates how the translator

made an error such as this. While the translator seems to understand the difference in noun

placement relative to the adjectives, he or she did not understand the difference in

headedness, which would imply a “reversal” of every element in the NP, not just the

movement of the noun. The fact that the term denoting color (“red,” in this case) occurs

closest to the noun in both English and Italian supports the idea that N-movement may not

be necessary to explain the different in adjective ordering. We do not have to assume an

underlying position of the noun to see that the adjectives have the same relationship to it as

they would in English. Let us consider another example, from Cique (1994).

a. La sola possible invasione romana della Tracia

the only possible invasion Roman of the Tracia
“the only possible Roman invasion of Tracia”

b. # La sola possible romana invasione della Tracia

the only possible roman invasion of the Tracia

  32  
 In this example, we see that while “sola (only)” and “possible” occur before the noun,

while “romana (Roman)” must appear after it. This does not necessarily have to be a due of

noun movement. I would like to suggest that it is actually the post-nominal placement of

spheres 1, 2, and 3, that make Italian different than English syntactically. For example, the

sphere 1 adjective “roman,” designating origin, occurs after the noun, while the one that

takes all adjectives, “possible,” occurs before. The case of “sola” is a little more complicated.

Although I did not focus on quantifying adjectives, I believe that “sola” would fall into such a

category. Traditionally, quantifiers are analyzed as being in the most distant position, and

since they are not attributive, they do not have a place in my theory. If I were to hypothesize,

however, I would guess that quantifiers, much like determiners, would remain pre-nominal

and distant form the noun. In addition, they would not be part of any particular sphere. We

can therefore conclude form our data that in Italian, adjectives in spheres 1, 2, and 3 are

post-nominal, where they are analyzed in mirror image order from English, while the

determiner, numerals, and scope-taking adjectives are found consistently pre-nominally.

Hebrew

Like Italian, Hebrew exhibits mirror-image order for spheres 1, 2, and 3, as

exemplified in the following example from Willis (2006). Hebrew also imbeds the determiner

into its nouns and its adjectives, and is a right-to-left language, which makes it unlike any

language we have analyzed thus far. Despite these differences, it still demonstrates an

order that is in keeping with my theory. The example below Willis’s example (example b)

was given to me by a native speaker of Hebrew. I was unable to come up with a successful

Romanization, since I do not trust my knowledge of transcribing from the Hebrew to

theLatin alphabet. However, the native speaker did provide me with a literal translation.

  33  
 a. ha-mexonit ha-amerika’it ha-aduma
the-car the-American the-red
“the red American car”

b. ‫והישן היפה הגדול הבית‬

The-house the-big the-beautiful and the-old

With these adjectives, we see the same phenomenon that we see with Italian, where

adjectives in the first three spheres are post-nominal and demonstrate mirror image ordering

to English. Unfortunately, I do not have data containing adjectives from other spheres, so I

am unaware of how Hebrew treats scope-taking adjectives and quantifiers. It would be

interesting to investigate, especially considering that Hebrew embeds its determiners.

Welsh

Among all the languages I have examined in the search to find one that does not

conform to the ordering rules, Welsh (and other Celtic languages) has been the only one

that cannot survive my theory without some form of noun movement. Noun movement in

Celtic has been a very established topic in syntax (Willis 2006), however, in his paper, Willis

seeks to disprove it. The data in Welsh shows how a simple noun-movement theory is

seemingly inefficient; while noun phrases are noun-initial, as in Hebrew, the adjectives

alternate between demonstrating a mirror-image order and a non-mirror image order. The

example below, from Willis’s paper, demonstrates a noun-initial NP that demonstrates

English AO rather than mirror-image AO.

a. cwpan mawr gwyrdd Sieineaidd

cup big green Chinese
“a large green Chinese cup”

  34  
 As you can see, the example demonstrates how n-movement would be the easiest

way to account for this. It is important to note, however, that the adjectives remain in the

same order relative to one another as they do in all other languages I have discussed. This

further motivates n-movement for this case and makes it the most understandable account.

However, in his paper, Willis provides ample data demonstrating that Welsh does in fact

behave with mirror image order. One example of this is below.

b. acen Rwsieg ysgafn.

accent Russian mild
“a mild Russian accent”

In this case, Welsh demonstrates a mirror-image order, with origin closer to the noun,

and quality further away. If n-movement were to be occurring here, the underlying form prior

to movement would violate the semantic restrictions placed on adjective placement as

outlined in my theory. It then seems that the only conclusion that we can draw about Welsh

and the Celtic languages is that they contain both mirror image ordering and n-movement.

The different contexts in which the two occur would be interesting to investigate, and

perhaps a topic of future study.

6. Conclusions

In this paper, I have established a uniquely semantic theory for the AO restrictions

that are in place in English. I demonstrated that these restrictions can be proved using

corpus data, as well as prosodic data and the ways in which native speakers react to

inversely ordered adjectives. I discussed the other work that has been done on the topic,

citing research from many disciplines spanning across several decades. My theory

discusses the “cognitive efficiency” of optimally ordered adjectives, where those adjectives

that are most intrinsic occur closest to the noun and those that relate the noun either to

  35  
another point in time, another object, or the external world, occur farther away. I discuss the

order in which adjectives are placed as the order in which they restrict the set of properties

of a given noun.

In my theory, the semantic subcategories of adjectives (shape, color, size, etc.) are

further categorized into broader “spheres,” where sphere 1 contains adjectives that are

decreasing in intrinsitivity as they increase in distance from a noun. Sphere 1 adjectives are

also less likely to be found out of order. Sphere 2 adjectives are internally relative, and

internally comparative, and when found out of order, result in a semantic shift from the

underlying order. Sphere 3 adjectives are externally comparative, and involve the speaker’s

opinion and perspective. Sphere 3 adjectives are more likely to be found in inverse position,

and I discuss the semantic shifts that occur when this is done. Lastly, I discuss the category

of scope-taking adjectives, which move more freely and have a more “adverbial” quality.

Scope-taking adjectives are capable of taking the noun and all of the adjectives that modify

it and modify all of them further. I briefly discussed how adjectives from identical semantic

categories behave when they interact and the semantic differences between ordered pairs

and their inverses.

I attempted to take my theory about the semantic restrictions of AO and apply it

across languages, which was successful in some languages, but not as much so in Welsh.

My cross-linguistic analysis is clearly very limited; I did not have extensive data or access to

many native speakers, so naturally, it was very difficult to test the semantic effects of

inversing the underlying AO in the other languages, as I was able to do in English. In the

future, I believe this would be an interesting study to do.

Additionally, I would be interested in exploring in more detail the relationship between

scope-taking adjectives and the adjectives within the other spheres; these adjectives seem

to behave in an interesting way that is easy to understand in English, but may be having an

  36  
affect on how other languages treat their AO. The free nature of these adjectives makes

them naturally more likely to vary cross-linguistically, and I would be very interested to

analyze this behavior.

In general, however, one thing that is never seen cross-linguistically, despite the

controversy about the placement of the noun, is a clear violation of the underlying order I

have established. The adjectives consistently remain in the same order in relation to one

another, whether in mirror image or not, which is a very robust finding. Having consistent

cross-linguistic relative adjective order as predicted by a semantic theory may be evidence

that a semantic theory will also suffice in explaining cross-linguistic variation among the

syntactic relationships between the placement of a noun and the adjectives that modify it. I

hope to have given some insight into this process during the brief application of my theory in

Sakha, Hebrew, Italian, and Welsh.

References

Cinque, G. 1994. On the evidence for partial N movement in the Romance DP. In G. Cinque,
J. Koster, J.-Y. Pollock, L. Rizzi and R. Zanuttini (eds.), Paths Toward Universal Grammar,
85–110. Georgetown: Georgetown University Press.

Danks, J. H., & Glucksberg, S. 1971. Psychological scaling of adjective orders. Journal of Verbal
Learning & Verbal Behavior, 1.10,63-67.

Halliday, M.A.K. 1994. An Introduction to Functional Grammar (2nd Ed). London: Edward Arnold.
Kemmerer, D. 2000. Selective impairment of knowelsge underlying pronominal adjective order:
Evidence for the autonomy of grammatical semantics. Journal of Neurolinguistics. 13-57-82.

Kemmerer, D., Weber-Fox, C., Price, K., Zdansczyk, C., & Way, H. 2007. Big brown dog or Brown
big dog? An electrophysiological study of semantic constraints on adjective order. Brain and
Language, 100, 238-256

Martin, J. E. 1969. Semantic determinants of preferred adjective order. Journal of Verbal Learning &
Verbal Behavior, Vol 8(6), 697-704. doi

Sproat R. and C. Shih. 1991. “The Cross-linguistic Distribution of adjective ordering restrictions”. In
C. Georgopoulos and R. Ishihara (eds.), Interdisciplinary Approaches to Language. Essays in
Honor of S.-Y. Kuroda, 565-593. Dordrecht: Kluwer Academic Publishers.

Willis, David. 2006. Against N-raising and NP-raising analyses of Welsh noun phrases
(abstract) Lingua 116: 1807-39.

  37  
Acknowledgements

First and foremost, I would like to thank Dr. Thomas Werner for advising this very
challenging topic. I was never at a loss for challenges and advice. Thank you also to Dr.
Mandy Simons for her guidance.

I would like to give thanks to my mother and Akira for constantly giving me their intuitions
and insights, and to Sara for lending her voice to my brief foray into prosodic analysis.

I would also like to thank Sardana, Danny, Tamar, Todd, and Professor Maroney for giving
me their inputs on the strange concoction of languages I chose to analyze.

  38