As I say, it has taken me a very long time to get my head round this one.
The distinction made in Quirk and others (as opposed to Quirk et al.) between pre-central, central, post-central and pre-head seemed rather arbitrary to me. Gradually it began to dawn on me that many of the items in the pre-head category were items which, when combined with a particular head had the potential to result in a distinct entry in a dictionary, for instance (tea + caddy, mud + hut, ...). This would seem to be particularly true when the pre-modifier is nominal. Various sources indicate that items which immediately or closely
precede the head identify crucial qualities such as domain, material, style, purpose, and so on.
There is broad agreement that, among "normal" (central) adjectives, there is an identifiable sequence.
Many of the simpler models ignore much of the material in the pre-head category, but the material is largely unproblematic.
The category post-central caused me considerable problems, but I now think there is a good argument for this group, and I think I can see cognitive motivations for the four groups. However, I find the nomenclature misleading and will introduce new names. For the sake of the logical argument I shall be pursuing, I shall also describe the groups in reverse linear order, i.e. right to left, away from the head).
Starting from the back, with no attempt to organize the items within each of the groups, we have the following:
- Isolating (pre-head)
The items in this group strongly define (rather than merely describe) the head. They isolate a particular subcategory of the head. The resulting concept is very liable to be distinctly identifiable within the semantic space with which we structure our experience. Thus, "a tea caddy" forms a distinct category of objects in our mind in a way that "a hot day" does not conjure up a category encompassing all days that are hot. The resulting composite items often have high collocative strength, and, if they do, they are likely to appear as separate headwords in a dictionary, for instance.
This is not to say, however, that every instance of ahead pre-modified in this way will form a previously known conceptual category, but it does instantiate a particular type of the head. Although pre-modification of this type is frequently achieved with nominals and denominals (social life, political life, coal miner, tea clipper, Queen Anne chair, ...), other word types are also possible, particularly participles (rocking chair, reading light,…).
- Defining (post-central)
Pre-modifiers in this group also serve to define (rather than describe) the head, but they do not have the isolating quality of the pre-head group. Most of the items in this group are participles, but again, it is their function rather than their form that determines membership in the group.
Unlike the items in the isolating group, these items do not conjure up or create a distinct type of the head ("a leaning tower" does not conjure up a category of all leaning towers, neither is the item "leaning tower" likely to appear as a headword in a dictionary). It seems to me that the items in this category are also likely to be more absolute than items in the central category. Thus, "burnt toast" is more absolute than "hot toast", with "hot" belonging to the descriptive/judgmental category and "burnt" belonging to the defining category. It is not that we cannot conceive of different degrees of "leaningness" of a tower, or "burntness" of toast, but rather that we are more interested in the fact that the tower is leaning or the toast is burnt. Hence my opening contention for this category that the items "serve to define rather than describe". This approach also accounts for the membership of colors in this group. We are interested primarily in the fact that the dog is brown, and not in the degree of brownness of the dog when we refer to “the brown dog”.
Items in this group have an objectively definitive quality about them. They not only refer to quantifiable qualities, but expressly identify the quantity. Thus, “18th-century” is an objective, definitive point (range) within the quality “age”. This is an important aspect distinguishing items in this category from items in the “descriptive” category below.
- Descriptive (central)
The items in this group serve to describe, rather than define or isolate, the head. Typically, but not exclusively, items in this group are adjectives rather than other parts of speech acting as adjectives, and they also regularly represent a point on a continuum (often top or bottom) of the quality they identify. Thus, we find things like "hot" and "cold" (as two points on a temperature continuum), but also items such as "warm" or "tepid". Other items such as "urgent" do not initially appeared to be on a specific, easily identifiable, continuum, but they nevertheless identify a point towards the top of the "urgency" continuum in a way that "leaning" does not identify any particular point on the "inclination" continuum. (In the same way, "blue" does not identify any point on the "blueness" continuum, and is hence usually found in the defining group rather than the descriptive group.)
In contrast to the “defining” category, we find adjectives such as “old” and “new” in this category rather than “18th-century”, which belongs to the “defining” category. In general, I contend that the pre-modifiers in the “defining” category are objectively quantified (“18th-century”), whereas items in the “descriptive” category are subjective and not quantified (“old”).
- Peripheral (pre-central)
The items in this group appear to have no defining or descriptive quality, and I am happy to accept the term "peripheral" as given in Quirk et al. to cover "emphasizers", "amplifiers" and "downtoners". Pre-modifiers such as “entire”, “total”, “slight”, “unique” and so on are found here.
Thus, across the four categories, we have a continuum from (right to left) "isolating" -> "defining" -> "descriptive" -> "peripheral".
This gives us an underlying ordering criterion and allows easy categorization without resorting to parts of speech (participles, for instance, can occur anywhere).
Some notes on category membership
This level of ordering appears to me to be almost trivial compared with the ordering within the categories themselves, particularly within the descriptive category, but before I make a stab at that, a word about my little dog. The more I thought about it, the more I became aware that the word "little" appeared, rather inexplicably, to behave differently from the word "small". It then occurred to me that it does so in certain circumstances. To give just a couple of examples:
- A little, painted cottage
- A small, painted cottage
- *A painted, little cottage
- *A painted, small cottage
- A small, frightened boy
- A little, frightened boy
- ? A frightened small boy
- A frightened little boy
It seems to me that when we use "little" to qualify many inanimate things, we are concerned only about the size, and hence the word behaves in the same way as the word "small". It is descriptive and falls into the corresponding category. However, when we use the word "little" to qualify animate objects in particular (but also other items), it bears an extra meaning associated with cuteness, lovableness and so on. (In conjunction with the word "job", for instance, it can also bear the meaning "insignificant".) In such cases, the word is no longer merely descriptive, but takes on a defining character that would lever it into the defining category. This would appear to explain the noted discrepancies. In the case of "little dogs", for instance, I would even contend that it moves to the isolating category and (for me) clearly isolates a particular type of pampered, overfed, ill-trained, yappy, snappy animal that is at best only a peripheral member of the "dog" class.
It is important to remember that the isolating/defining/descriptive/peripheral distinction is the criterion according to which pre-modifiers are assigned to categories. It is merely coincidental that defining and isolating pre-modifiers are often nominal, participial or denominal. The form of a pre-modifier (nominal, participial, …) does not exclude it from any given category; neither does it provide a reliable test of category membership. To take a simple example, the participial "well-behaved" belongs to the central category in "the well-behaved old dog". There is a sense in which we are regarding "well-behaved" as gradable and placing the dog towards the top of a continuum of good behavior.
Similarly, it has been noted that colors usually fall into the defining (post-central) category, where they are not regarded as gradable. We can, however, force colors to be regarded as gradable and hence change their category. Compare the following:
- A threatening, black cave
- A very black, threatening cave
This phenomenon also occurs without explicit marking of the gradability of the color:
- An old, black kettle
- A black old kettle
In the first phrase, we are merely identifying that the kettle is black. In the second, we are saying that on a continuum of grubbiness/blackness, the kettle rates pretty highly (in our judgment). Clearly, this highlights the aspect of subjectivity and speaker intention that will occasionally produce an order that may appear infelicitous to another speaker.
In the same way, a pre-modifier usually found in the descriptive category can be promoted to the isolating category, as in "hot seat", "hot plate", "high society". In this event, the pre-modifiers lose their gradable aspect ("*the hotter seat", "*the very high society").
This implicit shift can, of course also be applied to the “black cave” example:
Ordering within categories
- A black, threatening cave
Having established what appears to be a robust and usable sequence for pre-modifiers according to functional type, it remains to identify any "rules" that govern the ordering of pre-modifiers within their functional categories.
Ordering within the isolating category
Typically, the "isolating" category will only contain one element. I am at present not entirely clear as to whether this is a strong rule or not. It is clear that, in the phrase "the beautiful, old tea clipper", the word "tea" belongs to the isolating category. If, however, we expand the phrase to form "the beautiful, old China tea clipper", it is unclear whether the word "China" also belongs to the isolating category. There are a number of possible analyses:
- Both "tea" and "China" belong to the isolating category and are ordered by their isolating strength.
- "Tea" belongs to the isolating category and "China" belongs to the defining category.
- "Tea clipper" is an atomic concept, and "China" is in the isolating category in relation to the head "tea clipper".
Each of these approaches has its advantages and disadvantages, and it may well be that one analysis may be preferred for a given head. My own intuition is to prefer the second analysis unless there are good and clear reasons to prefer one of the other two. If the second approach is generally correct and the isolating category always or typically only contains a single item, ordering is of no significance. If the first approach is applied, ordering is by isolating strength. The third approach is recursive until the first or second approach is applied.
In any event, this issue does throw up the rather imprecise boundary between the defining and isolating categories. The issue can also be approached as follows: If we can identify ordering criteria for the defining category (see below), and we identify an element or elements immediately preceding the head that do not conform to the ordering criteria of the defining category, this is an indicator that the element or elements belong to the isolating category, in which case the speaker perceives them as "isolating".
Essentially, however, it remains my contention that the isolating category will usually only contain one item, in which case ordering is of no significance.
Ordering within the defining category
In order to initially develop a theory governing ordering within the “defining” category, I shall build and manipulate a phrase into which multiple defining elements can be inserted. To do so, I shall make use of the functional groups widely used in existing descriptions (shape, color, provenance, age, and so on).
As a starting point, let us take “steam engine”. This is clearly an instance of the head “engine” being preceded by an item from the “isolating” category, namely “steam”. It should not be possible to insert any further defining element between “steam” and “engine” (without fundamentally changing the underlying conceptualization). If we further add the word “old”, which belongs to the “descriptive” category, everything we insert from the “defining” category should appear between the words “old” and “steam”.
Our first element will belong to the functional group “style/design”:
“The old reciprocating steam engine”.
Now the fun starts. Let us add a word from the functional group “age”:
- The old Victorian reciprocating steam engine
- *The old reciprocating Victorian steam engine
It would thus appear that “age” precedes “style/design”. (Note that we have two items from the “age” domain, but they are from different categories, namely descriptive and defining.)
Now from the “weight” functional group:
- The old 2-ton Victorian reciprocating steam engine
- *The old Victorian 2-ton reciprocating steam engine
We can now pursue this method with further functional groups without creating a single monstrosity. Let us try “shape”:
- *The old 2-ton square reciprocating steam engine
- The old square 2-ton reciprocating steam engine
Proceeding in this way, we can establish the following hierarchy:
Note 1: “Physically identifiable properties” can be extremely difficult to distinguish from opinions, which rightfully belong in the “descriptive” category. Thus, we can have:
- Shape (square)
- Color (black)
- Physically identifiable features such as “crumbling”, “leaning”, often participial, can occur here, but see Note 1 below.
- Weight (2-ton) / Size (30-foot). Other quantifiable physical properties such as temperature also occur here, with the sequence being variable.
- Age (Victorian)
- Provenance (Belfast)
- Material (steel)
- Style/design (reciprocating), location (colliery), domain (agricultural), purpose (irrigation). See Note 2. Order is variable, but tends to go with defining weight, with the most defining characteristics closer to the head.
- The crumbling old cottage
- The old crumbling cottage
In the first instance, “crumbling” has been positioned in the “descriptive” category (preceding “old”, where it denotes a subjective opinion or judgment on the part of the speaker (see below). In the second instance, it has been placed in the “defining” category, where it denotes a physically identifiable, verifiable property. As in this case, the distinction can be very subtle, but indicates the degree of subjectivity of the speaker.
Note 2:The items in group 8 in particular are potential candidates for the “isolating” category, and in the absence of any item the clearly belongs to the “isolating” category (for instance an item which would not normally have this final position in the sequence), it is unclear, but of no significance, whether an item from group 8 should be assigned to the “defining” category or to the “isolating” category.
At first glance, this sequence appears to be relatively arbitrary, but it can be argued that there is a cognitive motivation for the sequence in the ease with which the relevant characteristics can be identified with the human senses (visual perception appears to take priority here).
Ordering within the descriptive category
Adopting the same procedure, and using the stop word “square” to demarcate the “defining” category, we can again work with a phrase such as “the big, old square house”. This procedure yields the following sequence:
Note 1:The position of gradable characteristics other than those listed separately appears to be unstable:
- Subjective opinion (beautiful). If there are multiple subjective opinions, the order is undefined.
- Size (big)
- Other gradable qualities (welcoming, solid). See Note 1 below.
- Weight (not in this example, but “a big, heavy, old anvil”)
- Age (old)
- Temperature (cold)
- Humidity (damp)
- The big, welcoming, old house
- The big, old, warm, welcoming house
- ? The big, old, welcoming, warm house
A possible explanation for this is given below.
In general, this sequence appears to be much more susceptible to variation than the sequence for the “defining” category. It appears to be subject to prosody, discourse and pragmatic effects to a greater extent. If stressed correctly, it is perfectly possible to say “a welcoming, old, big house” or “a damp, cold, unwelcoming house”. All other things being equal, the sequence given above appears to hold, and it is possible to argue that there is also a continuum in operation here, namely the immediacy of the subjective impression. Our first perception is an overall impression, often associated with an extremely general opinion. The primarily visual impression of size is also extremely immediate, whereas impressions of temperature and humidity are less likely to be so immediate. This could well explain the extreme variability of the position of the “other gradable qualities”. (Only after I have identified that the house is big, damp and cold do I deem it to be unwelcoming as well.)
Ordering within the peripheral category
As with the “isolating” category, this category has a strong tendency only to admit a single item, so ordering within the category is of no significance. It should, however, be noted that this category is preceded by other pre-modifiers whose order is already well-documented elsewhere. These include various determiners, articles, demonstrative adjectives, and so on.
The general rules governing the order of pre-modifiers appear to be based on a continuum of four categories, two of which contain distinct ordering sequences.
There is a cognitive motivation for the sequence of the four categories to be found in the (objective) defining power of the items that appear in each category.
Within the “defining” category, there is a sequence which also appears to have a cognitive motivation, namely the ease of perception (with our human senses) of the given characteristic.
Equally, there appears to be a cognitive motivation for the sequence in the “descriptive” category, namely the immediacy of the perception of the given characteristic. This is similar, but not identical to the motivation for the sequence in the “defining” category. Because this motivation is particularly subjective, it leads to greater variation in sequence than within the “defining” category.
As it stands, I have only looked at a couple of examples based on inanimate, concrete heads. Characteristics that qualify animate or abstract heads in all probability form separate groups within the categories, but it is to be expected that they will be ordered in accordance with the general principles for ordering of the four categories themselves and for ordering within the “defining” and “descriptive” categories.
That’s my theory, anyway. I would appreciate it if you “kick the tires”.