Some clear cases of modification of directional adpositional phrases can be seen in (65). Note that the reference objects (i.e. the internal arguments of the postpositions in and op) can either precede or follow the modifier. However, since reference objects can also precede VP adverbials such as snelquickly, as in Jan wandelde snel het bos inJan quickly walked quickly into the woods and Jan wandelde de berg snel opJan quickly climbed the mountain, it seems plausible that the order in which the reference object precedes the modifier of the directional PP is derived by leftward movement of the noun phrase.

In (66), we see similar examples with an adjectival modifier; the choice of the modifier depends on the nature of the reference object.

The modifications in (65) and (66) do not relate to the orientation or the magnitude of the vectors involved, but to the implied path: what is expressed is that a subpart of the path has been traversed. For example, the examples in (65) presuppose an itinerary that goes into the woods/up the mountain, and it is claimed that the located object has traversed 3 kilometers of this path. The difference between the directional constructions in (65) and the non-directional constructions in (67) is that in the latter Jan may have returned to his starting position after he has finished walking 3 kilometers, while in the former Jan must be located in the woods/up the mountain.

The adjectives in the non-directional constructions in (68) differ from the directional ones in (66) in that they only specify the place where the eventuality takes place: (68a) expresses that the activity of walking took place deep in the forest, and (68b) that the activity of climbing took place high in the tree.

Modifications of the type in (65) and (66) are only possible if the path is not intrinsically bounded, i.e. if the beginning and end of the path are not fixed. In (65a), this condition is satisfied; even though the path must be at least partly located in the forest, it is left implicit where it starts and ends, i.e. any position outside the forest is a suitable starting point and any position inside the forest is a suitable endpoint of the implied path. Something similar holds for the path referred to by the postpositional phrase in (65b). That the boundedness of the paths is relevant is also shown by the contrast between the use of directional PPs headed by the prepositions vanfrom, naarto and totuntil in (69a) and the use of the directional PPs headed by the phrasal preposition in de richting vantowards in (69b). The intuition is that the length of the path in (69a) is determined by some implied anchoring point, e.g. the position of the speaker. This means that the length of the implied path is contextually fixed, and therefore cannot be modified. In (69b), on the other hand, the beginning and the end of the implied path are not specified (either explicitly or implicitly), so that the length of the implied path is not contextually determined and modification is possible.

Other cases of directional phrases that denote paths that are inherently bounded and therefore cannot be modified are given in (70). For example, in (70a-c) the length of the implied paths is largely determined by the dimensions of the reference object; the beginning and end of the implied paths are bounded by two positions adjacent to and on opposite sides of the field/tunnel/house. Note that it is not clear whether the degraded example in (70d) can be accounted for in a similar way.

Finally, note that both the nominal and the adjectival modifiers can be extracted from the adpositional phrase by wh-movement. This is demonstrated in (71) by the interrogative counterparts of the examples in (65) and (66).

The boundedness of the implied path is also relevant in the case of modification by helemaal/gedeeltelijkcompletely/partly, which indicates whether the implied path is completely or partly covered. If the path is not inherently bounded, as in (72a), the use of these modifiers makes no sense and therefore results in unacceptability. Given the discussion of (65b) in the previous subsection, the grammaticality of (72b) may come as a surprise, but the difference in acceptability between (65b) and (72b) goes hand in hand with a difference in interpretation: in (65b) the endpoint of the implied path is left implicit in the sense that it can be situated anywhere on the mountain, whereas in (72b) the endpoint must be the top of the mountain. In other words, in (65b) op is interpreted as “onto”, whereas in (72b) it is interpreted as “on top of”. If the “on top of” reading is not possible, as in (72b'), the judgments are as expected. Note in passing that (72b) differs from (65b) in that the modifier must follow the reference object (under neutral intonation of the sentence).

The pattern in (73) is more or less what we expect to find: the PPs in (73a) denote bounded paths and modification by helemaal is correctly predicted to be possible, although it is surprising that the use of gedeeltelijk gives rise to a somewhat marked result; the PPs in (73b) denote unbounded paths, and modification by helemaal/gedeeltelijk is correctly predicted to be impossible.

The examples in (74) all involve inherently bounded paths, and the use of helemaal/gedeeltelijk is possible, as predicted. Note that in these cases, the modifier also follows the reference object (under neutral intonation of the sentence).

The modification possibilities in (72) and (74) correlate nicely with the distribution of the adjective heelwhole as an attributive modifier or predeterminer of the noun phrase expressing the reference object, as in (75). However, there seems to be a subtle difference in meaning between the two sets of constructions: whereas the examples in (72) and (74) suggest that the path proceeds in a more or less straight line, the examples in (75) suggest that the path proceeds in a more disordered (e.g. crisscross) manner. This might also explain the contrast between (75c&d) and (75e&f); while it is certainly possible to go crisscross across a field/through a tunnel, it seems less easy to pass a house/follow a canal in this way.

The intuition that the implied paths in (75) are of a more disordered nature may be related to the fact that heelwhole forces a distributive reading if used as an attributive modifier of the head of a reference object in a locational construction. Consider the examples in (76).

While (76a) is compatible with one family living in the castle (or with only a part of the castle being inhabited), example (76a') implies that the castle is divided into separate living units; people are more or less evenly distributed throughout the castle. Similarly, (76b) could be used to express that there is a pile of poles lying by the side of the road, while (76b') implies that the poles are placed at specific intervals along the road. Similarly, the examples in (75) could express that the implied path is more or less “evenly distributed” on the reference object; cf. Section N20.2 for a more detailed discussion of the predeterminer/attributive modifier heel.

Most of the examples in the previous subsections involve postpositional phrases. Since circumpositional phrases can also be used directionally, they have the same modification possibilities as the postpositional phrases. If the denoted path is not inherently bounded, as in (77a), the circumpositional phrase can be modified by a nominal measure phrase. If the path is bounded, as in (77b), the modifiers helemaal and gedeeltelijk can be used.