1. Framing the disagreement

Silvennoinen (Reference Silvennoinen2025) presents a careful corpus study of going forward, tracing its rise from business jargon in the 1970s to widespread use across registers by 2019. Taking a Construction-Grammar (CxG) approach, his central claim is that going forward is a phonologically fixed stored sequence. He treats it ‘as a whole’ as bearing the lexical category adv; the construction, in his view, inherits the morphosyntactic behaviour of the adverb class. He briefly considers the option of an intransitive preposition and declines it (fn. 7), chiefly on the ground that this ‘extends the notion of preposition’ and that ‘the majority of linguists do not seem to have adopted’ the analysis.

Those are sociological and terminological reservations, not empirical diagnostics; the question isn’t who adopts which labels, but which expressions occupy which slots. The flexible-complement view of prepositions – allowing them to take various complements or none – is developed in Jespersen (Reference Jespersen1924: 87–90) and Huddleston & Pullum et al. (Reference Huddleston and Pullum2002: §7.1, 610–17, §5.1, see also 58, 272; henceforth CGEL) and recurs across frameworks including Head-driven Phrase Structure Grammar (Sag, Wasow & Bender Reference Sag, Wasow and Bender2003), Lexical Functional Grammar (Bresnan et al. Reference Bresnan, Asudeh, Toivonen and Wechsler2015) and CxG (Michaelis Reference Michaelis2017), as catalogued by Reynolds (Reference Reynolds2025).Footnote ¹

Working within Silvennoinen’s CxG assumptions, I agree that going forward is entrenched and listed. I separate storage from external category and internal headedness. Externally, the construction patterns with PP adjuncts rather than with clausal adverbs: it’s systematically unavailable in the integrated-medial adverb slot and surfaces clause-initially or clause-finally (see section 2). Internally, going heads the construction as a deverbal preposition licensing a directional complement forward(s); complement drop is disfavoured in the futurate adjunct use, aligning the item with established deverbal prepositions such as according [to …], owing [to …] and depending [on …] (see section 3). On this basis, the inheritance prediction fails: a construction that neither occupies core adverb slots nor forgoes head–complement structure isn’t well analyzed as adv; it projects PP headed by going.

Storage isn’t category; category isn’t storage. The analysis therefore turns on two diagnostics – external distribution and internal complement licensing – and on their convergence.

2. External distribution: integrated medial vs. PP-like positions

Like typical PPs, going forward occurs only as a supplement clause-initially (1a) and clause-finally (1d), and in the subject–auxiliary corridor. The corridor position is truly ambiguous between a clausal supplement (1b) and an NP-internal supplement (1c).

The same is not true with going forward as an integrated modifier. The seeming ambiguity of (2c) dissolves under the pronoun probe – replacing a lexical head noun with a pronoun blocks NP-internal modification, forcing a clausal reading:

Under Silvennoinen’s coding, 59 of 83 ‘medial’ tokens are ambiguous between clause-medial and NP-internal readings. Under the pronoun probe, these resolve as NP-internal post-head modifiers. The remaining 24 tokens represent an upper bound on corridor supplements – at most 1.6 per cent of all tokens or 1.9 per cent of clausal tokens.Footnote ²

The key finding: what Silvennoinen codes as ‘medial’ conflates two distinct constructions – NP-internal post-head modifiers and clausal corridor supplements. Once separated via the pronoun probe, no clear cases of integrated clause-medial going forward remain. The construction patterns with PP adjuncts rather than with core adverb positions.

3. Internal structure: complement licensing and headedness

CGEL distinguishes prepositions and adverbs in part by complementation: prepositions commonly license an obligatory or optional complement; adverbs normally don’t (Huddleston & Pullum et al. Reference Huddleston and Pullum2002: 610–17). In the construction at hand, going licenses a selected obligatory complement.

Beyond forward, the same head occasionally combines with other directional complements in the same clause-adjunct frame.

These alternants differ in frequency and in conventionalized discourse use (e.g. managerial/policy-commitment), but they share the same structural profile: going as head, a directional complement, and adjunct extensionality. The availability of a small paradigm supports a head–complement analysis rather than a fused adverb.

Adverbs that license complements are exceptional, and when putative adverbs appear to take complements, they are generally better analyzed as prepositions under the flexible-complement view (see fn. 2), which countenances prepositions with a range of complements or none at all. The deverbal class is well established: according [to …], owing [to …], depending [on …], barring obj, concerning obj, regarding obj, among others (Kortmann & König Reference Kortmann and König1992; Huddleston & Pullum et al. Reference Huddleston and Pullum2002: 610–16). Going aligns with this group in licensing a complement and in disallowing complement drop in the construction under discussion.

4. Subsidiary diagnostics and potential counterarguments

Before examining the theoretical implications, I briefly address several potential objections and subsidiary diagnostics that, while less probative than external distribution and complement licensing, merit discussion.

4.1. Status of forward and a neutral XP label

Nothing in the argument requires taking forward to be a PP head. Some traditions treat it as an intransitive preposition; others label it an adverb. I use a neutral XP label for the complement inside the construction and keep the theoretical weight on going as the head that licenses a complement. The contrast that matters for external category is between PP and AdvP at the level of the whole construction, not between alternative internal labels for the complement.

For expository clarity:

The structural parallel in (7) shows how a deverbal preposition can select a complement and project PP without committing to a particular inner label for the complement in (7a).

The category decision is made at the level of the stored construction: going heads a PP that selects a directional complement, and the internal label for forward (P vs. Adv) is orthogonal to the argument, which turns on external distribution and complement licensing rather than sublabels.Footnote ⁵

4.2. Coordination and NP-internal occurrence

Going forward coordinates naturally with undisputed temporal adjuncts of a variety of categories, providing no clear evidence one way or the other.

NP-internal post-head uses exist, as Silvennoinen notes. That distribution is also compatible with either analysis, since English allows both PPs and (much more rarely) AdvPs in post-head modifier position (Payne, Huddleston & Pullum Reference Payne, Huddleston and Pullum2010).

The fact that going forward occurs post-head in NPs is therefore not probative for AdvP over PP, though on a frequency basis, PP is clearly favoured in this position.

4.3. Orthography and intervenability

Orthography does not diagnose category, but the spacing behaviour of lexicalized expressions often tracks resegmentation. Items that have lexicalized as single words or tight compounds tend toward fused or hyphenated spellings in running prose. Going forward overwhelmingly appears as two words, with hyphenation largely confined to attributive uses such as on a going-forward basis. This aligns with the retention of internal structure under storage and sits comfortably with the present analysis.

Intervenability is likewise weak evidence on its own. Clear futurate readings with modifiers between going and forward are rare, though they can be found.

The rarity is compatible with either a strongly selected complement inside a PP or a fused adverbial unit; the other diagnostics carry the load here.

5. Why adverbial inheritance is not supported

Silvennoinen’s claim that the construction, ‘as a whole, has the lexical category of adverb’ and ‘inherits the morphosyntactic behaviour of this wider category’ predicts robust access to core adverb slots and no need for head–complement structure. Neither prediction holds: externally the integrated-medial slot is unavailable beyond supplements, and internally deverbal going selects a directional complement.

6. Consequences for the analysis of prepositions and constructional change

Under complement-based diagnostics and the flexible-complement view of prepositions, the present analysis requires no category extension beyond independently motivated deverbal prepositions. It meshes with constructional approaches in which expressions become entrenched and stored as multi-word units while preserving their internal syntactic structure, rather than fusing into single lexical items like nevertheless or nowadays (Croft Reference Croft2001; Goldberg Reference Goldberg2006; Traugott & Trousdale Reference Traugott and Trousdale2013). The case study therefore supports two broader claims. First, external distribution is a better guide to category than category assumptions inherited from the adverb schema; diagnostics must target the slots a category actually occupies. Second, constructionalization frequently yields stored, semantically specialized phrasal units whose internal headedness remains visible and whose external distribution aligns with their heads.