Saturday, July 5, 2014

Supplementation (from CGEL, by Rodney Huddleston & Geoffrey K. Pullum)


Supplementation
We turn now to supplementation constructions, illustrated in such examples as:

[1]
i   Pat — the life and soul of the party — had invited all the neighbours.
ii  The best solution, it seems to me, would be to readvertise the position.
iii Jill sold her interest shares in January — a very astute move.

The underlined expressions are supplements, elements which occupy a position in linear sequence without being integrated into the syntactic structure of the sentence.

General properties of supplementation
In the clear and central cases, supplements have the character of interpolations or appendages. An interpolation, as in [1i–ii], is located at a position between the beginning and end of a main clause: it represents an interruption to the flow of the clause. An appendage is attached loosely at the beginning or end of a clause. In speech, supplements are marked as such by the prosody: they are intonationally separate from the rest of the sentence. In writing, they are normally set off from the rest of the sentence by punctuation marks — commas, or stronger marks such as dashes, parentheses, or (in the case of appendages in end position) a colon. Punctuation allows for different degrees of separation, as described in Ch. 20 §§4–5.

Supplementation in relation to dependency constructions and coordination
It is the lack of integration into the syntactic structure that distinguishes supplements from dependency constructions and coordination. But supplementation is like coordination in being non-headed: since the supplement is not integrated into the structure it cannot function as a dependent to any head. The three types of construction are thus distinguished as shown in:

[2]                                                            INTEGRATED?              HEADED?
i   DEPENDENCY CONSTRUCTION                    Yes                 Yes
ii  COORDINATION                                         Yes                  No
iii SUPPLEMENTATION                                    No              .   No

It should be noted, however, that expressions introduced by a coordinator can have the status of supplements rather than coordinates in an (integrated) coordination construction:

[3]
Jill — and I don't blame her — left before the meeting had ended.    [supplement]

In spite of the and, the underlined clause is an interpolation, and is clearly not of equal syntactic status with the clause Jill left before the meeting had ended. We thus treat [3] as an instance of supplementation, not coordination, such as we have in Jill left before the end of the meeting and I was sorely tempted to follow her.

Supplements and anchors
Although supplements are not syntactically dependent on a head, they are semantically related to what we will call their anchor69. In [1i] the anchor is the NP Pat, while in [1ii]iii] and [3] it is a clause — the clause which the supplement interrupts or follows. Other possibilities are shown in [4], where bold marks the anchor, underlining the supplement:

[4]
i   When the patient closed his eyes, he had absolutely no spatial (that is, third-dimensional) awareness whatsoever.
ii  The goal is to produce individuals who not only possess ‘two skills in one skull’, that is, are bicultural, but can also act as human links between their two cultures.

In [1] the anchor is the adjective spatial (which functions as attributive modifier to the noun awareness); in [ii] it is the VP possess ‘two skills in one skull’ (the first coordinate in a VP-coordination).
A supplement must be semantically compatible with its anchor. Compare, for example:

[5]
1   This stipulationthat the amount of damages not be divulged — was ignored.
ii  #This stipulationwhether the press could be informed — was ignored.

The supplement in [1] is a declarative clause and as such can appropriately combine with the anchor this stipulation. The anomaly of [ii] stems from the fact that the supplement is an interrogative clause and hence is not semantically compatible with its anchor.

69Some writers use the term ‘host’, but we have avoided this because we use it elsewhere in its primary sense, where it applies to the word to which a clitic is attached (see Ch. 18, §6.2).

Supplements vs dependents
Semantic compatibility vs syntactic licensing of complements
The restriction illustrated in [5] is comparable to that which holds between a complement and the head nominal in NP structure:

[6]
1   This stipulation that the amount of damages not be divulged was ignored.
ii  *This stipulation whether the press could be informed was ignored.

This time, underlining indicates the complement, bold the head. The noun stipulation licenses a declarative content clause as complement, but not an interrogative, so [ii] is inadmissible.
There is a significant difference between [5] and [6], however. The integrated construction shown in [6] requires that the complement by syntactically licensed, whereas in supplementation it is, as we have said above, a matter of semantic compatibility. Compare:

[7]
i   a. The stipulation that Harry could not touch the money until he was eighteen annoyed him enormously.
ccb. *The codicil that Harry could not touch the money until he was eighteen annoyed him enormously.
ii  a. This stipulation that Harry could not touch the money until he was eighteen — annoyed him enormously.
ccb. The codicil in the willthat Harry could not touch the money until he was eighteen — annoyed him enormously.

The examples in [1] belong to the integrated head + complement construction. Stipulation licenses a declarative complement, but codicil does not: hence the ungrammaticality of [ib]. In [ii] the content clause is a supplement, interpreted as specifying the content of its anchor NP. And this time the codicil example is acceptable: the NP it heads denotes an addition to a will and hence has propositional content which can be specified by a declarative content clause.
As a second illustration of the difference between the integrated and non-integrated constructions, consider:

[8]
i   a. The question (of) where the funding would come from wasn't discussed.
ccb. *The thing (of) where the funding would come from was rather more important.
ii  a. The second questionwhere the funding would come from — wasn't discussed.
ccb. The thing they didn't discusswhere the funding would come from — was rather more important.

Here the content clause is interrogative. In [ia] it is a dependent within the NP headed by the noun by the head noun: , so [ib] is ungrammatical. In [ii] the interrogative clause is a supplement and is subject to the weaker constraint that it be semantically compatible with its anchor. Example [iib] is therefore admissible because the anchor NP as a whole denotes a potential topic of discussion, so that the content of this topic can be specified by means of an interrogative clause supplement.

Form and interpretation of supplements realised by clauses
A further important difference between supplements and dependents is that the former may be realised by main clauses with their own illocutionary force:

[9]
Sue felt — can you blame her? — that she was being exploited.
The supplement here has the form and interpretation of a main clause: there is no change in form or loss of independent illocutionary force such as is found with clauses realising a dependent function.

Supplements and non-restrictiveness
By virtue of not being integrated into the syntactic structure, supplements are necessarily semantically non-restrictive. Compare, for example, [8ia–iia]. In the former, the integrated construction, the content clause is semantically restrictive, distinguishing the question being referred to from other questions. It provides the identifying information that makes it appropriate to use the definite article the. By contrast, in [iia], the supplementation construction, the second question by itself constitutes a definite referring NP. The supplement doesn't serve to distinguish one second question from other second questions: it doesn't restrict the denotation of the head nominal.
The same contrast between dependency and supplementation constructions is commonly found with relative clauses and appositives:

[10]
i   a. The necklace which her mother gave to her was in the safe.             i i  [modifier]
ccb. The necklace, which her mother gave to her, was in the safe.               [supplement]
ii  a. They are working on a new production of the opera ‘Carmen’.                [modifier]
ccb. Bizet's most popular opera, ‘Carmen’, was first produced in 1875.     [supplement]

In [ia] the relative clause is a modifier of the head noun necklace and serves semantically to identify which necklace is being referred to, but in [ib] it is a supplement to the anchor NP the necklace, which is assumed to be identifiable independently of the information given in the relative clause. Similarly, in [iia] the appositive Carmen is a modifier of opera, identifying which opera is being referred to, while in [iib] it is a supplement to the anchor NP Bizet's most popular opera, and since there can be only one entity satisfying that description the supplement is again non-restrictive.
However, we have noted in our description of relative clauses and appositives that the integrated construction is not necessarily semantically restrictive — see Ch. 12, §4.2, and Ch. 5, §14.3, respectively. Compare, then:

[11]
i   The father who had planned my life to the point of my unsought arrival in Brighton took it for granted that in the last three weeks of his legal guardianship I would still act as he directed.
ii  This is my husband George.

In [1] the relative clause doesn't distinguish one father from another: the narrator has only one father, so the modifier provides non-restrictive information about him. And [ii] does not convey that the speaker has more than one husband.
It is for this reason that we have departed from the traditional account of relative clauses, in which the two main constructions are distinguished as ‘restrictive’ and ‘non-restrictive’. A distinction in terms of integrated versus supplementary reflects the semantic difference more accurately and also matches the prosodic difference that distinguishes them in speech. It enables us, moreover, to capture the similarity between the unintegrated relatives and other elements that are semantically, prosodically, and syntactically unintegrated with the rest of the sentence: these can all be subsumed under the concept of supplement.

Syntactic representation of supplementation
A supplement, as we have seen, requires a semantically appropriate anchor: it cannot occur, as a supplement, without the anchor. Thus if we drop the anchor from [10ib], for example, the result is ungrammatical: *Which her mother gave to her, was in the safe. And if we drop it from [10iib] ‘Carmen’ takes on the status of an integrated dependent, the subject: ‘Carmen’ was first produced in 1875. For this reason, we take the anchor and its supplement to form a construction: a supplementation construction. But the lack of integration of the supplement into the syntactic structure means that there is no good reason to treat the supplementation as a syntactic constituent. We propose, therefore, that in the syntactic representation supplements should be kept separate from the tree structure, related to their anchors by some different notational device, as in [12]:

[12]




In [12i], representing the structure of example [1iii], the supplement a very astute move has the clause Jill sold her internet shares in January as its anchor: this is shown by the broken line leading from the functional label ‘Supplement’ to the category label ‘Clause’. Similarly in [12ii] the broken line shows that the relative clause is a supplement to the NP the necklace.

Prepositions and preposition phrases (from CGEL and A Student’s Introduction to English Grammar, by Rodney Huddleston & Geoffrey K. Pullum)


from The Cambridge Grammar of the English Language:


The category of prepositions

This book employs a definition of the category of prepositions that is considerably broader than those used in traditional grammars of English. The purpose of this section is to give an overview of the items that are assigned to the category of prepositions under our analysis.

Traditional definition
The general definition of a preposition in traditional grammar is that it is a word that governs, and normally precedes, a noun or pronoun and which expresses the latter’s relation to another word. ‘Govern’ here indicates that the preposition determines the case of the noun or pronoun (in some languages, certain prepositions govern an accusative, others a dative, and so on). In English those pronouns that have different (non-genitive) case forms almost invariably appear in the accusative after prepositions, so the issue of case government is of less importance, and many definitions omit it.

In our framework we substitute ‘noun phrase’ for the traditional ‘noun or pronoun’. With that modification, the traditional definition can be illustrated by such examples as the following, where the preposition is underlined.

[1]
i   Max sent a photograph of his new house to his parents.
ii  They are both very keen on golf.

In [i] the preposition of relates the NP his new house to the noun photograph (we understand that the new house is depicted in the photograph), while to relates the NP his parents to the verb send (we understand his parents to have been the recipients of the photograph). Similarly, in [ii] on relates the NP golf to the adjective keen (the semantic relation is like that between direct object and verb in They both very much like golf: the semantic role associated with golf is that of stimulus for the emotional feeling).

Prepositions as heads
In this book, in keeping with much work in modern linguistics, we adopt a significantly different concept of prepositions. We take them to be heads of phrases — phrases comparable to those headed by verbs, nouns, adjectives, and adverbs, and containing dependents of many different sorts. This change in conception leads to a considerable increase in the set of words that are assigned to the category of prepositions. Before turning to a description of the full membership of the category, we will explain, by reference to words that are uncontroversially prepositions, why they should be regarded as heads of phrases taking various kinds of dependent.

Modifiers
Note first that some prepositions can take modifiers like those found in other phrases:

[2]
i   She died [two years after their divorce].
ii   She seems [very much in control of things].
iii  It happened [just inside the penalty area].

These modifiers, marked with underlining, are found also in AdjPs (two years old), NPs (very much a leader), and VPs (She [just managed to escape]).

Prepositions followed by constituents that are not NPs
Secondly, it is not only ‘nouns or pronouns’ (NPs in our terms) that occur after prepositions:

[3]
The magician emerged [from behind the curtain].                                          [PP]
ii   I didn't know about it [until recently].                                                       [AdvP]
iii  We can't agree [on whether we should call the police].       [interrogative clause]
iv  They took me [for dead].                                                                            [AdjP]

In [1], one PP (underlined) is embedded inside a larger one (enclosed in brackets). This parallels the way one NP is embedded inside a larger NP in a house that size, or one clause is embedded inside another in That she survived is a miracle. In [ii], until has an AdvP as complement, instead of the NP that it has in examples like until last week. In [iii], on takes an interrogative clause complement rather than an NP, as in We can't agree [on a course of action]. And in [iv], dead is an AdjP, which has a predicative function, with me as predicand.
It is important to note that different prepositions license different types of complement. Until can take an AdvP but not an interrogative clause, while on can take an interrogative clause but not an AdvP, and so on. This is entirely parallel to the way verbs, nouns, and adjectives select particular types of complement. And the fact that the AdjP in [3iv] is predicative means that in the structure of the PP, as in that of the VP (or clause), we must make a distinction between objects and predicative complements. Compare:

[4]           OBJECT                                        PREDICATIVE COMPLEMENT
i   a. She consulted a friend.                        b. She considered him a friend.      [clause]
ii  a. She bought it [for a friend].                   b. She took him [for a friend].           [PP]

Extending the membership of the preposition category
Once it is recognised that prepositions head phrases comparable in structure to those headed by verbs, nouns, adjectives, and adverbs, we need to take a fresh look at what words belong in the category. When we do this, we find there are strong grounds for including a good number of words beyond those that are traditionally recognised as prepositions.

Traditional grammar's subordinating conjunctions
We have noted that prepositions take complements that are not NPs, such as the PP, AdvP, interrogative clause, and AdjP in [3]. This conflicts with the general definition for ‘prepositions’ given in most traditional grammars and dictionaries, though in practice traditional grammarians would have no hesitation in classifying from, until, on, and for in [3] as prepositions. Traditional grammarians thus tacitly accept that there can be PP, AdvP, or AdjP complements of prepositions. They do not, however, allow declarative content clauses. A word otherwise similar to a preposition but taking a declarative content clause complement is traditionally analysed as a ‘subordinating conjunction’. This is not a policy that can be justified. Consider the analogy with verbs that take both NP and declarative content clause complements:

[5]        NP COMPLEMENT                       DECLARATIVE COMPLEMENT
i   a. I remember the accident.          b. I remember you promised to help.
ii  a. He left [after the accident].       b. He left [after you promised to help].

No one suggests that the difference in the category of the complement between the [a] and [b] examples requires us to assign remember to different parts of speech in [i]. It would traditionally be treated as a verb in both cases. There is no reason to handle after in [11] any differently: it can be analysed as a preposition in both cases. Or take the following pairs, where the complement clause is declarative in [a], and interrogative in [b]:

[6]  DECLARATIVE COMPLEMENT        INTERROGATIVE COMPLEMENT
i   a. I assume he saw her.                       b. I wonder whether he saw her.
ii  a. the fact that he saw her                    b. the question whether he saw her
iii a. glad that he saw her                         b. unsure whether he saw her
iv a. He left [after he saw her].                 b. It depends [on whether he saw her].

The head words in [i-iii], those in boldface, belong uncontroversially to the same category in [a] as in [b] (verb, noun, and adjective respectively). We are proposing that the same applies in [iv]. The difference in the type of complement between [a] and [b] in [iv] no more justifies a part-of-speech distinction in the head than the similar difference in [i-iii].
We therefore include in the preposition category all of the subordinating conjunctions of traditional grammar, with three exceptions. The exceptions are, first, whether; second, those occurrences of if that are equivalent to whether (as in Ask him if he minds); and, third, that when it introduces a subordinate clause. These items we take to be markers of subordination, not heads of the constructions in which they figure: see Ch. 11, §8.1, for detailed discussion of this issue.

A subset of traditional adverbs
The traditional account does not allow for a preposition without a complement, but within a framework where prepositions function as heads of phrases, like verbs, nouns, adjectives, and adverbs, there is again no principled basis for imposing such a condition. Compare:

[7]      WITH COMPLEMENT                                         WITHOUT COMPLEMENT
i   a. She was eating an apple.                                       b. She was eating.
ii  a. She's [the director of the company].                      b. She's [the director].
iii  a. I'm [certain it's genuine].                                        b. I'm [certain].
iv  a. I haven't seen her [since the war].                         b. I haven't seen her [since].

The presence or absence of a complement has no bearing on the classification of the head in [i-iii], where in both the [a] and [b] members of the pair eating is a verb, director a noun, and certain an adjective. There is no reason to treat [iv] any differently, and we accordingly take since as a preposition not only in [a], but also in [b], where traditional grammar analyses it as an adverb.

[…]

Position of the preposition relative to its complement
The traditional definition specifies that prepositions usually precede the NP they govern. Simplified versions often omit the qualification ‘usually’, but it is indispensable for two reasons. In the first place, a very small number of English prepositions can follow the complement: compare notwithstanding the weather (head + complement) and the weather notwithstanding (complement + head) — we deal with these items in §4.2. Secondly, we have to allow for cases like What are you looking at?, where the complement appears in prenuclear position in the clause and the preposition is said to be stranded — this matter is discussed in §4.1. Preposition stranding is restricted to various kinds of non-canonical construction such as open interrogatives, relatives, etc.: in canonical constructions traditional prepositions (with the minor exception of the notwithstanding type) always do precede their complements. But so do verbs, adjectives, and adverbs. The location of prepositions before their complements is thus not a distinguishing feature of the category. Moreover, we have argued that not all prepositions have complements, and where they don't the issue of relative position obviously doesn't arise.


[…]

The ability of many prepositions to head phrases in complement function is an important property distinguishing them from adverbs, which can appear as complements only under extremely restricted conditions. Particularly useful from a diagnostic point of view are cases where the complement is obligatory.
One such case is the goal complement of certain transitive verbs such as put or place and a few intransitives such as dart and slither:

[6]
i   a. I put it in the drawer.                       b. *I put it.
ii  a. He darted behind the curtain.         b. *He darted.

The [a] examples here have prototypical PPs consisting of preposition + NP complement. But other forms can be assigned to the PP category on the basis of their ability to occur in this position. Compare:

[7]        PREPOSITION                                      ADVERB
i   a. I put it in/downstairs/away.                 b. *I put it adjacently.
ii  a. He darted off/indoors/home.              b. *He darted immediately.

Phrases consisting of such words as in, downstairs, away, off, indoors, home by themselves are distributionally like uncontroversial PPs such as in the drawer or behind the curtain, and are accordingly assigned to the same category. Prototypical adverbs, those formed from adjectives by suffixation of ·ly, do not occur in these positions.
A second case of an obligatory complement is in clauses with the verb be as head:

[8]    
i    a. Jill is in the office.                          b. *Jill is.
ii   a. The proposal is without merit.       b. *The proposal is.

The [b] examples are admissible if elliptical, with a complement recoverable from the preceding text (Max isn't in the office, but Jill is __), but otherwise be normally requires an internal complement. Leaving aside the specifying use of be, which allows complements of just about any category (cf. Ch. 4, §5.5), adverbs cannot in general function as complement to be. Where we have morphologically related adjectiveadverb pairs it is the adjective that is required in this function: Jill is sad, not *Jill is sadly. The fact that the underlined words in [7ia/iia] can occur as complement to be is thus further support for their classification as prepositions: compare Jill is in/downstairs/*locally.

[…]

There are a small number of adverbs such as right and straight which occur with a certain sense as modifiers of prepositions but not (in Standard English) of verbs, adjectives, or adverbs:

[9]
i   They pushed it [right under the bed].                         [preposition]
ii   *They were [right enjoying themselves].                              [verb]
iii   *I believe the employees to be [right trustworthy].        [adjective]
iv *The project was carried through [right successfully].      [adverb]

Not all prepositions accept these modifiers — they occur primarily with prepositions indicating spatial or temporal relations. But they are not restricted to phrases containing preposition + NP:

[10]
i   They pushed it [right in/inside].
ii   She ran [straight upstairs].

The occurrence of such highly restricted modifiers with words like in, inside, and upstairs as well as in uncontroversial PPs thus provides further evidence for recognising PPs without complements.

[…]

Among the words that do not license NP complements there are a fair number belonging to the spatial domain that occur as goal complement with such verbs as come and go,  and also, in most cases, as locative complement to be.

[29]
i   a. They went ashore.                          b. They are ashore.
ii  a. I'll take them downstairs.                 b. They are downstairs.
iii a. Kim is coming home.                      b. Kim is home.
iv a. Let's put everything indoors.           b. Everything is indoors.

There are good grounds for putting words of this kind in the preposition category. In the clause, prototypical adverbs generally occur in adjunct rather than complement function. They are not entirely excluded from functioning as complement (cf. They treat us appallingly, etc.: see Ch. 8. §2.1), but this usage is relatively exceptional. No adverb in ·ly could substitute as goal complement for the underlined words in the [a] examples in [29]. Leaving aside its specifying use, be does not license adverbs in ·ly as complement, so none could substitute for the underlined words in the [b] examples either. Notice, moreover, that ashore, downstairs, etc., in these [b] examples cannot reasonably said to ‘modify’ the verb. They no more modify the verb than does young in They are young. Thus although they are traditionally analysed as adverbs, it is arguable that they do not in fact satisfy the traditional definition.
These words are syntactically very like the prepositions in [27] except that they cannot take NP complements. Compare, for example, They went/are aboard and They went/are ashore. Some of them can be modified by right and straight, as in They are right downstairs or We went straight indoors. We accordingly include these too in the preposition class.

from A Student's Introduction to English Grammar:

Prepositions vs subordinators
The traditional class of subordinating conjunctions contains (among others) the words in [5]:

[5]
i        after                 before          since      till         until
ii   a. although       because        if(c)      lest     provided     though      unless
.    b. if(i)                  that                whether

We need to distinguish two words with the shape if.
One has a conditional meaning, as in I'll help you if I can: we show this above as if(c).
The other occurs in subordinate interrogative clauses like See if there are any vacancies, corresponding to main clause Are there any vacancies?: we show it as if(i). This is a variant of whether: compare See whether there are any vacancies.

The words in [i] traditionally belong to the preposition class as well, whereas those in [ii] do not. We have argued against a dual classification treatment of the [i] words, analysing them simply as prepositions that license different kinds of complement. But once we reconsider the distinction between prepositions and subordinators we find there are good reasons for reassigning the words in [iia] as well to the preposition class. This leaves a very small subordinator class, with that, whether and if(i) as its main members. The major argument for drawing the boundary between prepositions and subordinators between [iia] and [iib] is that that, whether and if(i) function as markers of subordination whereas the other words in [5] function as heads of the constituents they introduce. Consider the following examples:

[6]
i   a. I think [(that) she's probably right].
.   b. I don't know [whether they have received our letter yet].
ii  a. She stayed behind for a few minutes [after the others had left].
.   b. They complained [because we didn't finish the job this week].

In [1] the bracketed constituents are subordinate clauses with that and whether simply marking the subordination: the main clause counterparts are She is probably right (declarative) and Have they received our letter yet? (interrogative). In this context the that is optional (as indicated by the parentheses): the clause is in the position of complement to think, so it is not obligatory to mark its subordinate status in its own structure. Whether is not omissible because it marks the clause as interrogative as well as subordinate: it is just with the default declarative type that the subordinator is often optional.
After and because in [ii] by contrast are not grammatical markers of subordination. They have independent meaning, and it is by virtue of this meaning that we interpret the bracketed constituents as adjuncts of time and reason respectively. This makes them like heads — just as after is head in the time adjunct after the departure of the others. They are not themselves part of the subordinate clause; rather, the subordinate clauses are just the others had left and we didn't finish the job this week, and these function as complement within the phrases headed by after and because.

Prepositions vs adverbs
We begin the task of redrawing the boundaries between prepositions and adverbs by looking further at words like before, which can occur either with an NP complement or without a complement. There are a fair number of words of this kind; a sample are listed in [7]:

[7]
aboard      above       across     after      along     behind       below
beneath    beyond       by          down          in           off         outside
over            past        round       since     through     under    up

As we have noted, these are traditionally analysed as prepositions when they have an NP complement but as adverbs when they have no complement:
[8]
TRADITIONAL PREPOSITION                              TRADITIONAL ADVERB
i   a. She went aboard the liner.                               b. She went aboard.
ii  a. He sat outside her bedroom.                           b. He sat outside.

It has often been suggested that the traditional adverb category has something of the character of a classificatory wastebasket, a dumping ground for words that don't belong in any of the other more clearly defined categories. This criticism certainly seems valid in the present case. Aboard in [ib] and outside in [iib] don't, on the traditional account, qualify as prepositions because prepositions are defined in such a way that they require NP complements. They are obviously not nouns, verbs, adjectives or conjunctions, so there is nowhere to put them except in the adverb category.
We put it this way because it is important to see that these words do not in fact satisfy the definition that traditional grammar gives to the adverb category:An adverb is a word that modifies a verb, an adjective or another adverb’. They typically occur, for example, in the three functions given in [3] for PPs:
[9]
DEPENDENT OF NOUN         DEPENDENT OF VERB         COMPLEMENT OF BE
i   the conditions aboard              She went aboard.               She is still aboard.
ii   the temperature outside          He sat outside                     He is outside

(Contexts for the first example in each set might be I won't attempt to describe [the conditions aboard] and [The temperature outside] was over 40°.)
The first and third of the functions illustrated in [9] are characteristic of prepositions, but not of adverbs. Let us consider them in turn.

(a) Dependent of nouns
Adverbs do not normally occur as dependents of nouns: in related adjectiveadverb pairs it is the adjective that appears in this function. No such restriction applies to prepositions. Compare:

[10]                      PP                                                         ADVERB
i   a. She criticised them with tact.                   b. She criticised them tactfully.
ii  a. [A manager with tact] is needed.              b. *[*A manager tactfully] is needed.

The underlined expressions in [i] modify the verb, and we see that both PP and adverb are admissible.
In [ii], however, they modify the noun manager and here the PP is admissible but the adverb is not; instead we need an adjective: a tactful manager.

(b) Complement of the verb be
Adverbs cannot normally function as complement to be in its ascriptive sense: here we again have adjectives, in their predicative use. As with (a) above, there is no comparable constraint applying to prepositions. Compare:

[11]   PP AS COMPLEMENT OF BE                    ADVP AS COMPLEMENT OF BE
i   a. The key is under the mat.                          b. *Lucy was enthusiastically today.
ii  a. The meeting is on Tuesday.                       b. *Rain is again.

The [a] examples, with a PP functioning as complement of be, are impeccable, but the [b] ones, with an adverb in this function, are ungrammatical.

Instead of [ib] we have Lucy was enthusiastic today, with the corresponding adjective.
Since the adverb again has no adjective counterpart we cannot correct [iib] in the same way; for this particular example we could have It is raining again, with again now functioning as modifier to the verb rain.

The classification of words like aboard and outside as adverbs is thus inconsistent with the traditional definition of that category. The best way to remove this inconsistency is to amend the definition of prepositions so that they are no longer required to have an NP complement. Aboard, outside and similar words will then be prepositions both when they have NP complements and when they occur alone. This revision simultaneously gets rid of the complication of a dual classification for these words and removes them from the adverb class words which differ radically in their syntactic properties from genuine adverbs, thus making it a significantly more coherent class. Notice in particular that with our more restricted class of adverbs, but not with the larger class of traditional grammar, all functions that can be filled by adverbs accept some of the most central type, and the type formed from adjectives by adding ·ly.
This revision of the traditional analysis is not an original idea of ours. The core of it was first put forward as early as 1924 by the great Danish grammarian Otto Jespersen, and it is adopted in much work in linguistics since the 1970s.
One reason why traditional grammarians have not taken it up may have to do with the etymology, or historical source, of the term 'preposition'. This suggests a word placed in front of another word — the traditional preposition is a word placed in front of a noun (or NP, in our analysis). It may therefore seem undesirable to apply the term to a word which is not positioned in front of an NP. But there are three points to be made in favour of doing so:
First, prepositions do not always precede their complements: in What are you looking for? the preposition for does not precede its complement what (see §5 for more discussion of this construction).
Second, no one worries that the etymology of ‘adverb’ suggests a word dependent on a verb, although the term applies also to words modifying adjectives, other adverbs, and so on.
Third, the term 'preposition' is so deeply ingrained in the grammatical tradition that there would inevitably be a great deal of opposition to a newly invented replacement; it is better just to recognise that words often change their meanings, and to accept a change in the meaning of ‘preposition’. The property of being placed before an NP will still apply in central cases, but not in all.