1. INTRODUCTION The modification patterns of “compound indefinite pronouns” in English like something, everybody, and nowhere (Quirk et al. 1985, cf. Wu 2021) are intriguing because unlike other pronouns, this word class can take adjectival modifiers, but unlike nouns, these modifiers cannot occur PRE-(pro)nominally (Quirk et al. 1985, Kishimoto 2000, Larson & Marušič 2004, Leu 2004, Wu 2021). For example, something interesting is grammatical but *interesting something is not. 1 Many theorists have explained indefinite pronoun modification by comparing it to nominal adjectival modification. The two main nominal adjectival modification constructions in English are the attributive construction, in which adjectives come before a noun, and the predicative construction, in which adjectives come after a predicate that relates the adjective to a noun (Bolinger 1967). However, there are a limited set of scenarios in which an adjective can occur in a postpositive position, directly after a noun, in English. First, adjectives with complements or coordination can come after nouns, as in an actor suitable for the part and soldiers timid or cowardly (Quirk et al. 1985). Second, a specific class of attributive adjectives are able to follow nouns without complementation, construing a relationship with the noun that is temporary or “stage-level” (Carlson 1977): visible, navigable, responsible, and other -ible/-able adjectives, ablaze, afloat, and other a- adjectives, as well as adjectives such as present, concerned, and involved (Bolinger 1967, Quirk et al. 1985, Larson & Marušič 2004). The existence of what I call the POST-INDEFINITE PRONOUN MODIFICATION CONSTRUCTION (PIPM) has stumped many proponents of the generative syntactic perspective: how, if at all, are examples like something interesting related to post-NOMINAL modification? From what underlying structures do such examples originate? Do they involve attributive adjectives, predicative adjectives, postpositive adjectives or something else entirely? Various analyses have been proposed to explain post-indefinite pronoun modification. Kishimoto (2000) argues for a movement-based analysis of these structures, in which everything interesting is a transformation of the NP deep structure every interesting thing. Larson and Marušič (2004) critique this analysis, articulating that the adjectives in this construction must originate “in place” because of certain behavioral similarities between postnominal adjectival modification and post- indefinite pronoun adjectival modification. For example, like all the stars visible, everything visible has a stage-level (temporary/episodic) interpretation, while prenominal modification is ambiguous between stage-level (temporary/episodic) and individual-level (inherent/intrinsic) interpretations. However, Larson and Marušič’s (2004) arguments rely on comparison to a different set of postpositive adjectives (e.g. visible, navigable) than those that are most typical of post-indefinite pronoun modification (e.g. interesting, unusual). Wu (2021) adds on to Larson and Marušič’s analysis, arguing that adjectives like unusual and interesting are “coerced” into postpositive position, while visible and navigable are inherently postpositive. However, like Larson & Marušič (2004), Wu’s analysis assumes that all types of post-modification, whether nominal or pro-nominal, have the same types of interpretations. I suggest that there are key differences between the semantic construal associated with postnominal modification and PIPM that set PIPM apart. In this paper, I draw on the framework of Construction Grammar (CxG; Michaelis 2012; Hilpert 2014; Goldberg 1995, 2006) to demonstrate that post-indefinite pronoun modification is best understood as a separate construction from postnominal modification. Although it is true that post-indefinite pronoun modification shares various qualities with postnominal modification, tokens like everything interesting have individual-level construal rather than stage-level construal. For example, while everything visible means everything that is currently visible, everything interesting means everything that is inherently interesting. This difference is evidence of a separate schematic form-meaning pairing from other cases of postpositive modification. In PIPM, gradable evaluative adjectives follow indefinite pronouns, and pick an indefinite entity or event out of a class of entities or events prototypically socially evaluated as the adjective in question for the situation being described. For example, something weird happened construes a set of prototypically weird events that can happen and identifies the referent as one of these events. This paper will be structured as follows: In §2, I will give several examples of the PIPM construction. In §3, I will review previous analyses of this construction that compare post- indefinite pronoun modification to postnominal modification, involving what I call “restricted access” adjectives. In §4, I will present preliminary semantic arguments for why PIPM should be understood as a construction separate from not only attributive and predicative modification, but also from postnominal modification involving restricted access adjectives. I also discuss why the framework of CxG is perfectly suited to describe the unique construal associated with PIPM. In §5, I will follow up with additional syntactic evidence from the Contemporary Corpus of American English (COCA) 2 that PIPM is a separate construction from postnominal modification. The remaining sections of the paper will outline the form and meaning of PIPM (§6) and demonstrate some coercion effects (§7), before concluding remarks in §8. 2. EXAMPLES OF PIPM In this section I provide several examples of PIPM from COCA. This construction consists of any compound indefinite pronoun, and any gradable, evaluative adjective. Compound indefinite pronouns are those made up of two morphemes, a determiner/quantifier morpheme (e.g. every-, some-, any-, no-) and a nominal morpheme (e.g. -one, -body, -thing), and thus do not include indefinite pronouns like some and one (Quirk et al. 1985). Based on the corpus analysis I conduct in §5, common evaluative adjectives in this construction are as follows: new, wrong, different, unusual, cool, good, bad, suspicious, strange, interesting, funny, special, nice, stupid, similar, negative, terrible, unexpected, & amazing. See examples of PIPM in 1: 1) a. And that’s where we see something new and potentially something very promising b. Do you see anything wrong in that? c. They are willing to spend more to get something special. d. He saw nothing unusual at first e. Everything bad was over f. The deceased must have been somebody important g. Just find someplace nice h. One thing good about late spring is… While I agree with Larson & Marušič (2004) that transformational analyses that involve movement of thing from post-adjectival to pre-adjectival position are not appropriate, thing is still relevant to this construction. When preceded by a quantifier, thing and other semantically “light” nouns play a similar semantic role as compound indefinite pronouns and therefore fit into this pattern (see 1h). 3. PREVIOUS ANALYSES Kishimoto’s (2000:558-559) analysis outlines the problem of post-indefinite pronoun modification with a set of contrasts between attributive adjectival modification of nouns vs. indefinite pronouns, reproduced below in Examples 2-7. It should be noted that these adjectives must PRECEDE nouns (as in 2a, 3a, & 4a) and must FOLLOW indefinite pronouns (as in 5b, 6b, & 7b): 2) a. every interesting book b. *every book interesting 3) a. a delicious dish b. *a dish delicious 4) a. cold rooms b. *rooms cold 5) a. *interesting everything b. everything interesting 6) a. *delicious something b. something delicious 7) a. *cold someplace b. someplace cold To account for these distinctions, Kishimoto (2000) proposes an N-raising analysis in which the “light nouns” thing and place can move from a location in an NP, after the adjective, to a position in a Num Phrase, before the adjective, as outlined in 8a (pre-movement) and 8b (post- movement): 8) a. [DP every [NumP [NP interesting thing]]] b. [DP every [NumP thing [NP interesting________]]] Larson & Marušič (2004) argue that the movement analysis cannot be correct, by drawing comparisons between post-indefinite pronoun modification and the behavior of certain postnominal adjectives, including visible, navigable, and responsible. They build off of Bolinger’s (1967) argument against a movement analysis that ties attributive and postnominal uses of these adjectives to the same underlying structure. Bolinger instead analyzes postnominal adjectives as reduced relative (predicative) clauses, as in 9b, reduced from 9a: 9) a. the stars that are visible b. the stars visible Although Larson & Marušič (2004) provide many reasons why a movement analysis of post- indefinite pronoun modification cannot be correct, here I will review only two of them: 1) attributive-only adjectives do not occur with indefinite pronouns and 2) indefinite pronoun modification involves stage-level construal, like postnominal modification. Firstly, post-indefinite pronoun modification cannot be underlyingly attributive because indefinite pronouns cannot be modified by the adjectives that Bolinger (1967) identifies as only occurring attributively. Bolinger (1967) points out that there are specific adjectives, such as live and mere, that occur attributively, as in 10, but do not occur predicatively or postnominally, as in 11 and 12 respectively. Larson & Marušič (2004) points out that these attributive-only adjectives also do not occur following indefinite pronouns, as in 13 (Larson & Marušič 2004:273, Wu 202:826): 10) a. live animal b. mere idea 11) a. *this animal is live (cf. this animal is alive) b. *no idea is mere 12) a. *an animal live b. *no idea mere 13) a. *something live (cf. something alive) b. *nothing mere Therefore, Larson & Marušič (2004) suggest that post-indefinite pronoun modification cannot be underlyingly attributive, otherwise indefinite pronouns would be able to be post-modified by attributive-only adjectives. Secondly, and crucial to the current analysis, Larson & Marušič (2004) draw on Bolinger’s (1967) observation that while attributive modification can either be interpreted as stage-level (temporary/episodic) or individual-level (inherent), postnominal modification NECESSARILY has a stage-level construal. Larson & Marušič (2004) argue that like postnominal modification, indefinite pronoun modification also only has a stage-level construal. Larson & Marušič (2004:274) contrast attributive modification in 15a & 16a with postnominal modification in 15b & 16b: 15) a. List all the visible stars, whether we can see them or not. b. ??List all the stars visible, whether we can see them or not. 16) a. List all the responsible individuals, whether they were involved or not. b. ??List all the individuals responsible, whether they were involved or not. Attributive modification, in 15a, can refer to stars that are in general visible to the naked eye – or “inherently” visible (individual-level) – or it can refer to stars that are currently (temporarily/episodically) visible (stage-level). Because visible stars can have both readings, it can occur with a continuation that directly references the two possibilities of currently visible or currently invisible. Similarly, in 16a, responsible individuals can be interpreted either as individuals who are responsible in general, in other words, trustworthy (individual-level), or responsible for a specific situation (stage-level), and thus can occur with a similar continuation that denies the possibility of current involvement. Nominal post-modification on the other hand, does not have both possibilities. Stars visible in 15b necessarily means the stars that are temporarily currently visible (stage-level), and thus the same continuation sounds odd (denoted by ??). Similarly, individuals responsible necessarily refers to individuals responsible for some current situation (stage-level). Larson & Marušič’s (2004) then compare these construals to 17, examples of post-indefinite pronoun modification: 17) a. ??List everything visible, whether we can see it or not. b. ??List everyone responsible, whether they were involved or not. Like 15b, everything visible in 17a necessarily means everything that is CURRENTLY visible (stage- level) and cannot mean everything that is visible in general (individual-level), and like 16b, everyone responsible in 17b necessarily means someone who is responsible for some current act (stage-level). Therefore, both of these cannot take the continuation that contrasts two possibilities. Larson & Marušič (2004) propose that if everything visible was underlyingly every visible thing, that it would have the same semantic construal as the attributive modification pattern in 15a & 16a – that it would be ambiguous between stage-level and individual-level. They thus provide convincing evidence that post-indefinite pronoun modification does not originate prenominally (attributively). However, while Larson & Marušič (2004) begin their discussion by citing Kishimoto’s examples (2-7) of evaluative adjectives that can occur after indefinite pronouns but not after nouns, their argumentation relies solely on comparison to adjectives that CAN occur after nouns, the -ible/-able adjectives, such as visible and responsible. Many of these adjectives convey that there is restricted access to the noun they modify, in other words, that only a limited number of the noun is visible or responsible (and that others are non-visible, or not responsible). Therefore, I’ll call this group of adjectives “restricted access” (RA) adjectives, for ease of reference.3 This is in general, a different set of adjectives than those that can only follow indefinite pronouns and which typically occur with PIPM (see §2), which I’ll refer to as PIPM adjectives. Larson & Marušič (2004:270) admit that their account is not able to explain, if post-indefinite pronoun adjectives “originate postnominally” in a similar fashion to postnominal modification, what prevents *every book interesting, *a dish delicious, and *rooms cold in 2-4. An analysis of the PIPM pattern must address the distinction between adjectives that can generally follow nouns (RA adjectives) vs those that cannot (PIPM adjectives), and cannot rely upon the syntax and semantics of one modification pattern to explain the other. Along these lines, Wu (2021) expands upon Larson & Marušič’s (2004) analysis, proposing a syntactic explanation for why *every book interesting doesn’t occur but something interesting does. While Wu (2021) agrees that tokens like stars visible are reduced from relative clauses (stars that are visible), he claims that only certain adjectives (RA adjectives) are inherently postpositive and thus can undergo this reduction. Adjectives that typically (except for in PIPM) occur prenominally (e.g., interesting) are instead coerced to postpositive position specifically when occurring with indefinite pronouns (not nouns). Wu (2021) argues that this coercion process occurs because the “prenominal” modifier position is not available, as the “determiner” (e.g., every) and “noun” (e.g., thing) pieces of compound indefinite pronouns cannot be broken up by the insertion of a prenominal modifier. Therefore, the modifier needs to occur after the compound indefinite pronoun, since modifiers cannot occur before determiners. This explanation is favorable because it focuses on the unique morphosyntactic properties of indefinite pronouns – on their properties that fall between phrase-hood and word-hood. However, Wu (2021:836) claims that as a corollary of this coercion process, “the placement of potential attributive adjectives in postposition will restrict them [to] ‘temporariness.’” In other words, he again explains the semantics of PIPM modification in terms of RA modification, even while successfully separating the syntactic patterns involved. While this construal effect is often true for RA adjectives like visible, it is not true for PIPM adjectives. I will demonstrate evidence for this assertion in the next section. 4. PIPM: A CONSTRUCTION IN ITS OWN RIGHT Some theorists have suggested that post-indefinite pronoun modification is underlyingly attributive and undergoes movement to occur in postpositive position (Kishimoto 2000). Others have suggested that it is underlyingly postpositive, possibly reduced from a predicative relative clause (Larson & Marušič 2004), or coerced to postpositive position because of the specific morphosyntactic structure of compound indefinite pronouns (Wu 2021). In this section, I review the evidence for each position, detailing in what ways post-indefinite pronoun modification is similar to attributive modification, and in what ways it is similar to RA postnominal and predicative modification. Ultimately, I argue that post-indefinite pronoun modification shares particular features with each of these types of modification and is therefore best described as a separate construction in its own terms. As discussed in the last section, previous analyses have demonstrated that post-indefinite pronoun modification is similar to postnominal and predicative modification patterns in that attributive-only adjectives such as live and mere cannot modify indefinite pronouns. PIPM is clearly not the same modification pattern as attributive modification. Previous approaches (Larson & Marušič 2004, Wu 2021) have also argued that post-indefinite pronoun modification produces a temporary (stage-level) construal. However, I argue that this is often true for RA adjectives but not for PIPM adjectives. For example, in 18a, an example with a PIPM adjective, goodness is characteristic of everything (individual-level), not a temporary quality. Similarly, in 18b (repeated from 1f), importance is a characteristic quality of the deceased (individual-level), that does not disappear after their death. Lastly in 18c (repeated from 1d), unusualness is a characteristic quality of the possible entities or events to be seen (individual-level). What is temporary in this case is the period in which one has not seen one of these items. 18) a. You remind me of everything good. b. The deceased must have been somebody important. c. He saw nothing unusual at first. Due to the individual-level construal involved in PIPM, in addition to this pattern showing conflict with attributive-only adjectives, it is in fact also rare with predicative-only adjectives, like afraid and sorry, that tend to convey temporary feelings rather than general characteristics. This is not predicted by Larson & Marušič’s (2004) and Wu’s (2021) claims that all postpositive adjectives have a temporary construal. While the man is afraid is felicitous, someone afraid is distinctly odd. In COCA, although there are 54 hits for INDEF-PRONOUN afraid, many of these involve secondary predicates or adjectival complements. There is only one true example of PIPM, in 19: 19) Mr-HAMILL: They -- I learned that, that they -- they dwell on fear. If you -- if you give into them, you know, that -- that -- that gives them a big high. They want to see somebody afraid. And -- and -- and I really wasn't -- didn't have to -- to try to act like I wasn't. Being afraid is typically not something that characterizes somebody, but in 19, Hamill construes fright as a characteristic quality of someone the individuals in question are looking for. This is a marginal example; in general, afraid doesn’t tend to occur in PIPM. Similarly, there are no true examples of PIPM with sorry. Thus, while post-indefinite pronoun modification shares some features with predicative and postnominal modification (including conflict with attributive-only adjectives), it shares individual-level construal with attributive modification. This makes examples of PIPM different from reduced relatives – someone who is sick doesn’t mean the same thing as someone sick, because someone sick construes sickness as an individual-level property, while the predicative relative clause construes it as a temporary stage-level one. This is very different from what we see with post-indefinite modification involving RA adjectives, like visible. As discussed in the previous section, something visible does mean the same thing as something that is visible. Thus, PIPM is a different constructional pattern than attributive, predicative, or postnominal modification, and should not be explained in terms of any of these other syntactic patterns. Instead, the framework of Construction Grammar (CxG) can be utilized to explain the idiosyncrasies associated with post-indefinite pronoun modification. Within CxG, syntactic patterns are assigned specific meanings and/or functions, just as words are (Michaelis 2012, Hilpert 2014, Goldberg 1995, 2006). Whereas generative perspectives find it difficult to explain cases in which the same schematic syntactic structure is associated with two different functions or meanings, CxG can account for such patterns. For example, Goldberg (1995:204-209) identifies two different meanings associated with the “way” construction, a “means” version and a “manner” version labeled in 20: 20) a. Means: In some cases, passengers tried to fight their way through smoke-chocked hallways to get back to their cabins to get their safety jackets. b. Manner: …he was scowling his way along the fiction shelves in pursuit of a book. The means version of this construction involves interpretations of the verb (e.g. fight) in which the associated action is the means of creating a path, while the manner version involves interpretations of the verb (e.g. scowl) in which the associated action is an activity occurring at the same time as traversing a path. In the present analysis, we are dealing with a similar case, in which the same surface form, postpositive modification, is associated with two separate functions, restricted stage-level modification (RA modification), and evaluative individual-level modification (PIPM). If there are two separate constructions associated with the same surface form, we should be able to find examples in which the same adjective is used in both constructions, with two different construals. Indeed, these cases can be found. While Bolinger (1967) says that “the man responsible is unambiguously ‘to blame’ and the responsible man is almost unambiguously ‘trustworthy’” (p. 4), and Larson & Marušič (2004:273) claim that post-indefinite examples always share the episodic (to blame) reading with postnominal modification, we do see post- indefinite examples like those in 21, in which someone responsible means someone that is trustworthy: 21) a. That's great, " she says. " The landlord is definitely looking for someone responsible. " " I'm that person, " I say. b. If you could pair him with someone responsible - - maybe a girl. These examples, although they include RA adjectives, appear to be examples of the PIPM construction, with individual-level construal, rather than examples of the typical pattern that Bolinger (1967) and Larson & Marušič (2004) describe for postnominal RA modification, involving stage-level construal, as in 22: 22) I did nothing... not my fault. Nobody says it is. Take it easy, right? We will find someone responsible. Thus, rather than post-indefinite pronoun modification following the phrase structure rules for nouns or requiring movement or coercion to produce, the modification pattern that occurs with indefinite pronouns is specific to this lexical class – it is a formal idiom (Michaelis 2012), associated with its own form and function. In the next section, I aim to demonstrate that while there are examples of RA adjectives that occur in the PIPM construction, in general, PIPM adjectives and RA adjectives are separate sets of adjectives that occur in separate syntactic patterns. 5. CORPUS ANALYSIS In the previous section, I demonstrated that PIPM and RA modification have different semantic construals. In the current section, I demonstrate that although their immediate surface structure is the same – they both occur postpositively – they occur in different larger syntactic patterns. Specifically, PIPM occurs in at least two larger syntactic patterns that RA adjectives are rare or do not occur in: after verbs of perception (e.g., see something unusual) and before verbs of occurrence (e.g., something unusual happened). The general goal of this corpus case study is to show that adjectives common in previous arguments that compare postnominal modification to post-indefinite pronoun modification – the RA adjectives – don’t occur in the same syntactic contexts as PIPM adjectives. For RA adjectives, I selected a small group of adjectives that consistently have come up in the literature: visible, navigable, available, possible, and present (a non-ible adjective, that also participates in the same restricted access pattern). For PIPM adjectives, I first selected the larger syntactic patterns that appear to be specific to PIPM (after verbs of perception and before verbs of occurrence), in order to isolate examples of PIPM, and then identified adjectives that occur most commonly in these patterns. This led to the group of adjectives similar, strange, terrible, unusual, different, and new. Although the adjective wrong is also very frequent in PIPM, it is often used as a secondary predicate or adverbial, and its use in PIPM is therefore difficult to isolate. The need for an analysis involving larger syntactic structures becomes apparent when examining the results of an initial corpus search in Table 1 below. Many more nouns occur before RA adjectives than indefinite pronouns, but only some PIPM adjectives have more indefinite pronouns than nouns preceding them. In order to show that PIPM occurs in unique syntactic patterns, additional examination of these examples is needed. Restricted Access Adjs Indef-Pronoun ADJ something visible Noun ADJ man visible visible 102 2201 navigable 0 19 available 351 31845 possible 1210 9416 responsible 402 7538 present 236 11825 PIPM Adjs similar 2664 9486 strange 1278 676 terrible 1336 325 unusual 1848 275 different 6684 5291 new 14498 11489 TABLE 1. Restricted Access Adjs vs PIPM Adjs after Nouns & Indefinite Pronouns (COCA) In Table 1, there are several types of examples of PIPM adjectives following nouns that aren’t true examples of post-modification. For example, included in these numbers are secondary predicates as in 23a, adverbials as in 23b, and examples with adjectival complements as in 23c-d. Examples with adjectival complements are not examples of PIPM because all adjectives can occur postpositively if they occur with complements, as discussed in §1. 23) a. it's harder to make those jokes and make comedy funny if you don't have any profanity b. After completing the book I saw things different c. …creating a self funded program similar to those in at least 10 other states… d. …we'd found 400 species new to the park… Thus, corpus work with this construction requires manual attention to weed out irrelevant examples. I thus examined larger syntactic patterns for two reasons: 1) I thought I would be more likely to isolate examples of PIPM within these patterns, and 2) to create a smaller set of data to manually remove examples of secondary predicates, adverbials, adjectives with complements, as well as idioms like brand new. For both analyses below, I manually examined all tokens, in other words, examples with both RA and PIPM adjectives, and those after both nouns and indefinite pronouns, removing all examples of these irrelevant patterns. Although coordinated postposed adjectives are another pattern that license post-modification, many valid examples of PIPM involve coordination (as in 1a). Therefore, I retained these examples and discuss an example of this pattern below. 5.1 VERBS OF PERCEPTION Table 2 shows the results of searches in COCA for patterns involving post-modified nouns and post-modified indefinite pronouns with RA & PIPM adjectives, following the perception verbs see, hear, taste, smell, and touch. Two separate noun searches were needed to account for singular & plural tokens. During manual review of the data, in addition to the types of examples discussed above, I also removed tokens in which the constituency wasn’t clear. For example, in 24, a Wh-question constituency test is awkward: 24) a. I don’t see anything unusual about this. b. ??What do you see _____ about this? Restricted Access Adjs See/hear/taste/smell/touch Indef-Prn ADJ See something visible See/hear/taste/smell/touch NOUN ADJ See men visible See/hear/taste/smell/touch * NOUN ADJ See a man visible visible 1 0 0 navigable 0 0 0 available 2 2 16 possible 1 0 4 responsible 0 0 1 present 0 0 2 PIPM Adjs similar 73 1 0 strange 119 1 0 terrible 22 0 0 unusual 187 0 0 different 191 0 0 new 276 0 0 TABLE 2: Restricted Access Adjs vs PIPM Adjs in the “Perception-verb X ADJ” Pattern (COCA) As expected, indefinite pronouns with PIPM adjectives are much more common after perception verbs than both indefinite pronouns with RA adjectives or nouns with either type of adjective. Examples of PIPM adjectives following indefinite pronouns in this pattern are shown below in 25: 25) a. Did you hear anything unusual last night? b. Because whenever he saw something new and interesting, or new and ridiculous, he always wondered what she'd have to say about it. While RA adjectives following indefinite pronouns are less common (at least after perception verbs), they are still grammatical, as the examples in Larson & Marušič (2004) show. One example of this type is given in 26. This example has the stage-level construal predicted in previous analyses, while PIPM examples in 25 are individual-level. 26) we need a lefty in the outfield and I don't see anyone available. Lastly, while it is expected that RA adjectives would also follow nouns, it is unexpected that PIPM adjectives would follow nouns. Therefore, these tokens, in 27, require further discussion: 27) a. we saw visions strange and foreboding, but we kept them to ourselves, because heaven blesses the meek b. But just hearing and seeing videos similar on the Internet, it just made me uncomfortable. 27a is an example that involves coordination, and thus is a predictable example of post- modification (§1). However, 27b is not easily explainable. While “on the Internet” is a secondary predicate of see rather than a complement of similar, perhaps this example may also be licensed due to the “heaviness” of this clause (Bolinger 1967, Larson & Marušič 2004). Overall, such examples appear to be marginal. Examples of PIPM are common after verbs of perception, while examples of restricted access modification are less common after verbs of perception. Crucially, the adjectives identified as PIPM adjectives only felicitously occur after indefinite pronouns. 5.2 VERBS OF OCCURRENCE Table 3 shows the results of searches in COCA for patterns involving post-modified nouns and post-modified indefinite pronouns with RA & PIPM adjectives, preceding the occurrence verbs happen, occur, begin, start, and be going on. Restricted Access Adjs Indef-Pronoun ADJ happen/occur/start/begin/be going on Something visible happened NOUN ADJ happen/occur/start/begin/be going on Thing visible happened visible 2 0 navigable 0 0 available 0 1 possible 0 0 responsible 0 0 present 0 3 PIPM Adjs similar 140 0 strange 183 2 terrible 186 0 unusual 106 1 different 36 0 new 37 0 TABLE 3. Restricted Access Adjs vs PIPM Adjs in the “X ADJ Occurrence-verb” Pattern (COCA) As in the previous section, it is clear that indefinite pronouns with PIPM adjectives are more common preceding verbs of occurrence than both indefinite pronouns with RA adjectives or nouns with either type of adjective. Examples of PIPM adjectives following indefinite pronouns in this pattern are shown below in 28: 28) a. Did anything strange happen when you were living there? b. it becomes more apparent that something terrible is going on inside Kosovo. There are three postnominal examples of PIPM with verbs of occurrence, two examples involving the same phrase (shown in 29a) from different parts of the movie Killer Tomatoes Eat France, and one additional example in 29b: 29) a. When these four things strange occur as one... " The True King of France shall return with the sun. b. …that he will go away satisfied and not report back to the authorities that some thing unusual is going on in that household 29a is an example of quantified “light noun” thing, therefore an example of PIPM. Since this example also involves a rhyme, this suggests that the ordering of this token is creative and playful. 29b is also an example of PIPM with a space between the “determiner” and “noun” parts of the compound indefinite pronoun. Therefore, while these examples are marginal, they are explainable. In this short corpus case study, I have shown that despite the same surface structure of post- modification, there are larger syntactic contexts in which PIPM is relatively common but restricted access modification is rare. By isolating PIPM examples after verbs of perception and before verbs of occurrence, I have provided further evidence that PIPM is a separate modification construction than restricted access modification, adding on to the semantic evidence in the previous section. The PIPM adjectives that I examined rarely post-modify nouns in these patterns, and when they do, they are typically examples of the PIPM pattern, with the quantified light noun thing or with a space between parts of the indefinite pronoun. Furthermore, despite comparisons to RA adjectives in the literature on the PIPM pattern, in the syntactic patterns analyzed here, RA adjectives rarely occur. 6. CONSTRUCTIONAL ANALYSIS OF PIPM PIPM is a formal idiom, or construction, because its form cannot be licensed by traditional phrase structure rules, it has a special interpretation, and because there are semantic constraints on the words that can occur in the construction (Michaelis 2012). The formal structure of the PIPM construction is as follows: [compound indefinite pronoun | quantifier thing(s)] (Degree or Essence Adverb) Evaluative Adjective The meaning of the construction is a gestalt construal of an indefinite entity or event that is selected from a backgrounded larger category of entities or events, evaluated by prototypical societal norms as falling somewhere along the scale of the adjective in the typified context of use. Context is important to this construction because something unusual means something different in say something unusual, taste something unusual, and something unusual happened: a different larger category of actions or entities (unusual things that are said, things that taste unusual, unusual things that happen) is construed for these different phrases. Therefore, the larger category from which an indefinite pronoun is selected can be abstract or concrete, and made of entities or events, since indefinite pronouns can refer to many different things. The fact that similar and different are common in this construction emphasizes the social situatedeness and intersubjectivity of this construction, as social actors commonly employ it to discuss situations that are similar or different from what they are currently jointly focused on. However, despite a contextual construal, the categories evoked by this construction are contingent on conventional ideologies and social stereotypes about what kinds of things can be described as the adjective in question (or can be evaluated as similar or different from the entity in question). Thus, while the adjectives in this construction are subjective and gradable, the construction identifies a class of entities or events that rely on conventional societal norms to establish what is terrible, unusual, strange, new, similar, or different in a given context. The construction doesn’t just identify one such thing, but a whole category of things, which emphasizes socially sanctioned or “typified” understandings of social action (Gal & Irvine 2019). PIPM’s referent is a gestalt – rather than emphasizing the referents’ indefiniteness or the quality of the adjective, the construction construes both as equally important to its meaning. This distinguishes something unusual from similar ways of to say the same thing like something that is unusual, or some unusual thing, which have far fewer tokens in COCA. As discussed in §3, while Larson & Marušič (2004) & Wu (2021) argue that postnominal and post-indefinite modification construes adjectives as only temporarily or episodically modifying the nouns they describe, the kind of social meaning that PIPM imparts is necessarily individual- level, since it deals with stereotypes of social action. This can be shown by the fact that you can modify PIPM with “essence” adverbs, such as fundamentally, essentially, and inherently, as in 30. It does not appear that these adverbs can be used in RA modification. 30) This is an exciting result that suggests something fundamentally different about what processes play a key role in the generation of Mercury's magnetic field…. Degree adverbs can also modify the adjectives in this construction. This includes adverbs such as so, very, slightly, and totally, which push the category in one direction on the scale construed by the adjective. Finally, comparatives also are sanctioned, since this merely construes a category that is compared to another category. But superlatives, like ??something most different, are not attested in COCA, since superlatives refer to the highest end of a scale rather than a categorical class associated with a region on a scale. Something should be said about the effect of different indefinite pronouns in this construction, since the construction has a slightly different meaning with indefinite pronouns that start with the quantifiers no, some, any, and every. In these different cases, the backgrounded category against which the indefinite meaning is construed is the same, but the foregrounded part of the category, the referent of the overall construction, is different. With indefinite pronouns that start with no, as in the phrase nothing new, a category is construed of things prototypically socially evaluated as new, and the referent of the phrase is associated with none of these things. With some, we have seen that the referent is identified as an indefinite thing out of the larger category. Any acts similarly to some, but is often used in negative contexts, questions and subjunctives. Lastly, when the indefinite pronoun begins with every, the referent of the construction is the entire construed category. 7. COERCION EFFECTS AND SOCIAL MEANING One of the demonstrations of a meaningful and productive constructional pattern is the observation that its constructional meaning can coerce a new meaning from a lexical item that doesn’t match its selectional restrictions, in other words, the types of lexical items with which it normally combines (Michaelis 2012). As I discussed in §6, PIPM generally occurs with gradable, subjective adjectives that are associated with socially proscribed categories in a particular context. However, occasionally, non-subjective adjectives (31a), non-gradable domain adjectives (Sullivan 2013; 31b-c), and proper nouns or adjectives formed from proper nouns (31d-g) can occur in PIPM: 31) a. I was waiting for him to say something drunk like This baby needs us. This baby didn’t need us. b. Nothing sexual happened beside a few erotic kisses. c. Does that mean a Christian may not say anything Christian in an Islamic state? d. And they stopped paying Sarah Palin to come into the office once every three months and say something Sarah Paliny e. When I get home from work, I build a fire and chuckle to myself that Mr. Murphy saw something Thoreauvian in my nature f. I almost wanted to do a Vegas theme for this but alas nothing Vegas happened on this day in hockey history. g. And Now for Something Completely… Obama (Gonzálvez-García 2014:281) In these examples, adjectives (or nouns) that typically are not considered gradable, subjective, or associated with social evaluation receive a construal as such. In 31a, an adjective that is not typically used to describe a subjective quality, drunk, is construed as a subjective evaluative term that describes a category of things one might say while drunk. In 31b-c, non-gradable adjectives that normally designate a domain under which types of activities or actions can be categorized, such as sexual and Christian, are also construed as socially evaluated gradable categories. Crucially, these tokens rely on shared social knowledge of what types of things can be evaluated as sexual or Christian. Lastly, there are several examples of PIPM tokens that involve proper nouns or adjectival forms of proper nouns, in 31d-g. These examples rely on metonymic inferencing (Gonzálvez-García 2014, Danygier 2011) to evoke the quality associated with a category of socially evaluated entities, and thus presuppose expertise with the domain associated with the proper noun. Since expertise often signals ones’ engagement with particular activities or communities, this construction may be leveraged in the construction of social meaning (Eckert 2008, Silverstein 2006). For example, something Thoreauvian indexes one as both elite (possessing academic knowledge) and as a lover of nature. 8. CONCLUSION In this case study I have shown that the Post-Indefinite Pronoun Modification, or PIPM, construction is a separate adjectival construction than postnominal modification with restricted access adjectives. While the “light” noun thing can participate in this construction, other nouns cannot. In addition, while PIPM involves individual-level construal, postnominal modification involves stage-level construal. Thus, PIPM should be analyzed on its own terms, and not as a form of postnominal modification as generative analyses have done (Larson & Marušič 2004, Wu 2021). Using the framework of Construction Grammar (Michaelis 2012), I have shown that this construction has a specific form and semantic interpretation that cannot be predicted from its parts. Its meaning plays into and (re)constructs ideologies about particular social qualities, and it even coerces adjectives and nouns that are not usually evaluative to evaluate entities and events along social scales. Analysis of this construction can thus complement work at the intersection of syntactic analysis and the construction of social meaning (Moore 2009). REFERENCES Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18.1–34. Carlson, Greg. 1977. A unified analysis of the English bare plural. Linguistics and Philosophy 1.413–456. Dancygier, Barbara. 2011. Modification and constructional blends in the use of proper names. Constructions and Frames, 3.208–235. Eckert, Penelope. 2008. Variation and the indexical field. Journal of Sociolinguistics, 12.453– 476. Gal, Susan and Judith Irvine. 2019. Signs of difference: Language and ideology in social life. Cambridge University Press. Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure. The University of Chicago Press. Goldberg, Adele. 2006. Constructions at work: The nature of generalization in language. Oxford University Press. Gonzálvez-García, Francisco. 2014. “That’s so a construction!”: Some reflections on innovative uses of “so” in Present-day English. Studies in Functional and Structural Linguistics, ed. by María de los Ángeles Gómez, Francisco Ruiz de Mendoza Ibáñez, and Francisco Gonzálvez-García, 271–294 (Vol. 68). John Benjamins Publishing Company. Hilpert, Martin. 2019. Construction Grammar and its Application to English (2nd ed.). Edinburgh University Press. Kishimoto, Hideki. 2000. Indefinite Pronouns and Overt N-Raising. Linguistic Inquiry, 31.557– 566. Larson, Richard and Franc Marušič. 2004. On Indefinite Pronoun Structures with APs: Reply to Kishimoto. Linguistic Inquiry, 35.268–287. Leu, Thomas. 2005. Something invisible in English. University of Pennsylvania Working Papers in Linguistics, 11.143–154. Michaelis, Laura. 2012. Making the Case for Construction Grammar. Sign-Based Construction Grammar, ed. by Hans Boas and Ivan Sag, 31-67. CSLI Publications. Moore, Emma and Robert Podesva. 2009. Style, indexicality, and the social meaning of tag questions. Language in Society, 38. 447–485. Quirk, Randolph; Sidney Greenbaum; Geoffrey Leech; and Jan Svartvik. 1985. A Comprehensive grammar of the English language. Longman. Silverstein, Michael. 2006. Old Wine, New Ethnographic Lexicography. Annual Review of Anthropology, 35.481–496. Sullivan, Karen. 2013. Frames and Constructions in Metaphoric Language. John Benjamins Publishing Company. Wu, Zhen. 2021. Compound pronouns in English. English Language and Linguistics, 25.825– 849. END NOTES 1 Wu (2021:845) explains an exception that some indefinite pronouns can occur with articles and prenominal adjectives, however then they are construed as nouns, as in a very special someone. 2 All examples cited in this paper are from COCA. 3 While it is likely the case that these postnominal adjectives occur in a particular restricted access construction, I stay agnostic on the constructional status associated with this postnominal modification since it is not the focus of this paper.