Remarks on the Lexicographical Treatment of Metaforms

This paper examines the way dictionaries describe meta forms, items that refer to the linguistic system itself (e.g., pronouns and indexicals). The paper first shows the inconsistent grammatical classification of metaforms, which are usually and incorrectly categorized as adverbs, conjunctions, prepOSitions, or interjections. It is then argued that metaforms should be classified in their currently known grammatical categories: discourse marker (now), focus particle (even, like), quotative (like), and so on. Discussion then turns to the definitions of meta forms, with illustration of their incorrect or misleading semantic characterization and suggestions for capturing their denotations consistently. This examination of the lexicography of metaforms is ultimately a case study in how dictionary making might be informed by judicious use of current grammatical, semantic, and pragmatic theory.


Introduction
This paper explores the lexicographical treatment of fonns that refer to the Lexikos 6 (AFRILEX-reeks/series 6: 1996): 1-13 informational system of language itself -either the formal linguistic code or the conditions of exchange in which the code operates.I refer to this class of reflexive forms broadly as metaforms.
This investigation has two main purposes: A. to show the need for careful form-class judgments of metaforms (and, by.implication; forms in general) and the enhancement of these judgments by the use of current grammatical theory; B.
to show the way consistent definitions can be constructed for terms that defy definition (or seem not to merit it, in some cases) if defining practice is informed by current semantic and pragmatic theory.

2.
What are metaforms and do dictionaries include them?
Metaforms denote, all or in part, some aspect of the linguistic system or context of exchange.Typically included in the list of metaforms are indexicals, focus particles, evidentials, discourse markers, certain interjections, quotatives, metalinguistic forms, and so on (see Konig 1991, 5chiffrin 1987, and Lucy 1993 for illustrations).For example, an essential part of the meaning of indexicals is reference to properties of the system of exchange.There means 'some spatial position away from the source of speech,' and so crucially involves the linguistic and contextual system itself.Indexicals are fairly obvious and uncontroversial metaforms, but there are many other items that are similarly reflexive, and not often understood as such: focus particles, for example.Even gets its meaning by signaling that the hearer must add information to that currently within the scope of the particle and consider that new information more important or stronger than any presupposed in the context (Konig 1991: 69ff.).Even is thus a metapragmatic signal to the hearer to examine both the proposition asserted and the context and then explicitly a,dd information that exceeds that encompassed by the particle itself.
'While metaforms are semantically productive and determinate, their lexicographical treatment has not followed suit, perhaps most significantly in their very inclusion in the dictionary.As long as the metaform dovetails with an established form-class or can be (erroneously, I will argue) forced into one, it is likely to be included.50 no dictionary excludes even (adverb -wrongly categorized, I will argue) or 1 (pronoun).But few include the discourse-level metaphrase you know (1 was, you know, surprised by that).i Furthermore, form-class recognizability frequently interacts with usage judgments.Metaforms are sometimes excluded because they are incorrectly assigned to certain registers or varieties (e.g., really and huh), or attributed unsystematic or idiosyncratic distributions (e.g., like: see below).
We can see this conspiracy of uncertain• grammatical class and usage judgments in the uneven lexicographical treatment of a long-debated form -like.It has a notorious hist~ry ~ English lexico~raphy, the ~ost obvious illustration of which is its substitution for the ostensibly more refmed as.The AHD (1992: 1042) puts this amusing history nicely: Writers since Chaucer's time have used like as a conjunction, but 19thcentury and 2Oth-century critics have been so vehement in their condemnations of this usage that a writer who uses the construction in formal style risks being accused of illiteracy or worse.
In modem American English, like has undergone a fairly established, but only recently noticed (and "horrifying," to purists), innovation to a metaform in certain meanings and distributions.Like now has two meta-uses: as a focus marker and as a quotative verb.
In the former, like is a free particle, indicating either attenuation (la) or significantly new information (lb) (examples from Underhill 1988: 236): (la) But it's like a five, ten minute hike to the cabin (b) She like paid for the whole thing ... In (la), like means 'something like,' and hedges the commitinent of the speaker to the information that follows.In (lb), however, like means 'and here's the real new information or the point of this exchange, namely .. .'; in this sense, like is a focus particle.Interestingly enough, these two meanings are closely related both conceptually and historically.But even more important, they are formally unifiable since they are both instances of focus (though Underhill 1988 reserves focus for only the latter usage).
In the latter use, as a marker of quotation, like has also drawn some attention (Romaine and Lange 1991: 227): (2) And she's like, "Urn ... Well, that's cool." In this function, like signals reported speech or thought (or even behavior: "So I was like: [gesture or behavioral demonstration].").This like appears to be a grammatical innovation of an alternate for say; express, or even do (Cf.colloquial go: Butters 1980).It is thus a metaform, reflexively signaling the fact of speech, thought, or behavior.Both uses of like are quite widespread.Underhill (1988) reports instances of focal like across social class, gender, and age.A literature professor is reported as saying about E.M. Forster: "Then, all of a sudden, he like stopped producing completely" (Underhill 1988: 238).Romaine and Lange (1991) observe a similarly wide distribution of quotative like.Although they find it more common"among young females (like many grammatical changes), they also report written instances (e.g., from the New York Times and Washington Post) and some usage in British English.Indeed, the New York Times Magazine (May 5, 1996: 54) quotes Christine Whibnan, the current Governor of the state of New Jersey, as follows: "I was like, 'Please don't let my Dad see me ......The linguistic and social distributions of like as a metaform should recommend it for lexicographical treabnent, but it has been picked up only sporadically.The AHD (1969AHD ( , 1992) ) reports only the second focal use, calling it nonstandard.This is a curious usage label because the attenuating sense has a long history and is closer to standard.In fact, depending on how the form's semantic history is to be interpreted, it could be argued that the OED CD-ROM (1993) reports related uses and meanings back to the 13th century (Cf. the first sense in the OED CD-ROM (1993)}.The BONE (1973BONE ( , 1990)) COD (1995: 789) reports the second focal sense (as 'so to speak'), but classes it as an adverb and slang.And if you look closely in the OED CD-ROM (1993), you find both focal senses well attested for over 1 000 years (like adv., B.7).
Quotative like, unfortunately, loses out completely.It rarely appears anywhere, as far as I can tell, not even in the BONE.One exception is Spears' (1989) dictionary of American slang, which records both quotative and focal/ike.The appearance of these uses in a dictionary of slang is a striking illustration of the socio-political conditions on choice of entries (see Willinsky 1988 for some interesting discussion in this respect).
The lexicographical success and failures with like open a window on dictionary-making itself.Why is there such uneven treatment of metafonns?

The form class of metaforms
Part of the problem with metaforms is their grammar.Certainly the present variation and grammatical innovations of forms such as like make form-class judgments difficult.But when you look at how metaforms have been treated in dictionaries as a whole, a number of puzzles about their part of speech immediately surface, suggesting that they have never been properly classed and that such errors have been passed down out of lexicographical tradition.does likewise, and cites, as supporting evidenc~, even, which if also classes as an adverb, as do the other dictionaries.1But there is little to recommend this construal of focus particles as adverbs.Admittedly, some derive from adjectives, and the adjectival base of some adverbs is well known.Just and even fit this pattern, but this does not mean that they currently are adverbs.Many other focus particles have nonadjectival and non-adverbial sources, like numerals and verbs, and others are unquestionably nonadverbial in their current manifestation (such as the modal particles: Konig 1991: 163ff.) .In fact, the relevant counter-evidence to the construal of focus particles as adverbs can be found in the dictionaries' own constructed examples.The AHD (1992: 979) exemplifies the focal meaning of just ('merely') with just a scratch.
The RHO (1971: 775) has a similar exemplification: just a clerk and just one of those things, The COD (1995: 737) illustrates the focal meaning 'no more than' with just a minute.These cannot possibly be adverbial uses, as the failure of the substitution of other adverbs attests: (3a) ?? happily a scratch (b) ?? quickly a clerk (c) ?? fast one of those things (d) ?? eXcitedly a minute (Cf.ran happily, quickly, fast, excitedly) While adverbs are known to have a fairly unrestricted cooccurrence, there are limitations (McCawley 1988: 192).But focus particles have unrestricted categorial distribution because their communicative purpose is to mark the status of pieces of propositions (or propositions as a whole) with respect to context.In this function, they behave like all scope-bearing items (adverbs included) and have their effects left to right.Even or just can thus modify a determiner (l saw even THAT dog) or a preposition (Donna went just INTO the street, not ACROSS it), and like can modify a whole sentence (Like, Bob caught a fish) or part of one (Bob caught a, like, FISH).
Similar conclusions hold for discourse markers, forms that signal speaker and hearer position with respect to the discourse context (now, well, oh, and then, e.g.Schiffrin 1987): Now, 1 was going to say that ... , Well, 1 think I'll go, Oh, ] was going to say that ... , So you're leaving, then?Dictionaries usually class these as adverbs, conjunctions, or interjections.The AHD (1992AHD ( : 1239) ) reports now in this use as an adverb.Chambers (1983: 866) has it as both an adverb and interjection.Curiously, the OED CD-ROM (1993) has it as an adverb, but it cites the very form that Chambers classes as an interjection (now then, which the OED CD-ROM (1993) traces back to the year 10q as an adverbial use).WNWCD (1996: 929) has it as both an adverb and interjection, and the COD (1995: 932) has it as only an adverb.
Similar alternate categorizations can be found for the other discourse markers.The COD (1995: 1445, 3c andd) classes two uses of discourse-level then as an adverb.But other dictionaries exclude these uses altogether.The RHO does not cite then in any of its discourse forms, nor does WNWCD.Per-haps by this practice, the RHO and WNWCO mean to indicate that the discourse-marker sense is subsumed by, or is indistinct from, the adverbial or conjunctive sense and so not necessary to report.However, this would overlook the unique surface distribution of the discourse marker: We decided on going, then ... Like focus particles, discourse markers must be differentiated from adverbs, conjunctions, or interjections (whatever this latter category might be!).They are, to put it baldly, discourse markers.Why not call them what they are?
Perhaps most confusing and hence (ironically) illustrative is again the case of like, which has not only a checkered history of inclusion but an equally checkered history of categorization, both in its "standard" uses and in its metafonn innovations.
Almost without divergence, dictionaries list a standard use of like as a preposition, as in the following expressions (AHO 1992: 1042): It's not like him to take offense, lived like royalty, looks like a bad year for farmers.The RHO, WNWCO, and COO echo this classification, as do Romaine and Lange (1991: 244ff.) in their study of the quotative.But I have my doubts about this (as does the OEO CD-ROM (1993), which, as far as I can determine, never classifies it as a preposition).
In all its ostensible prepositional uses, like fails the standard tests for constituency as the head of a prep.ositionalphrase (PP).A PP can take a pro-fonn: (4) Bob sat on the chair, and Bill sat there, too This holds even for apparent conflations of prepositions with conjunctions when prepositions take sentential objects: (5) Bill ran after Bob shot the gun, and Sam ran then, too But no such substitution is possible for the like-phrases that supposedly exemplify prepositional uses: (6a) ?? They lived like royalty, and we lived such, too (AHO) ??He works like a beaver, and she works such, too (RHO) (c) ??She acted like an idiot, and he acted such, too (COO) Another test is movement.PP's allow object extraction and fronting for topicalization, sometimes leaving behind a resumptive pronoun.Like-phrases do not: .The fence -he looked over it for his father Like cannot be a preposition in these standard uses.It behaves much more like a complementizer, signaling a full or reduced clause.(Quirk et al. 1978: 727-28, 754-55 call it simply a subordinator and note its peculiar behavior, though they also give it prepositional uses (323).)But no dictionary I know of uses the cate~ gory complementizer.Indeed, that, uncdntroversially a complementizer, is called a conjunction in WNWCD (1996WNWCD ( : 1386)), COD (1995: 1444), and RHO (1971: 1470).But complementizers have quite restricted distributions because they mark certain types of clauses (McCawley 1988) while conjunctions link categories of any type.Note that the only possible pro-form for the like-phrases in ( 6) is so (9a); which is also an acceptable substitute for a noun clause marked by that (9b), but not acceptable as a pro-form for either a clause with a conjunction (9c) or a preposition (9d): (9a) They lived like royalty, and we lived so, too. (b) Tom believes that Tom was here, and Donna believes so, too, (c) ?? Tom cried because Bob was here, and Donna cried so, too.(d) ?? Bob sat on the chair, and Bill sat so, too.
These clear problems with the categorization of standard like make me worry about its classification in the focal and quotative -i.e., metaform -uses.The BONE (1973: 260) classes focal like as a conjunction, citing a number of written instances.The OED CD-ROM (1993) does likewise, giving citations that reach back two centuries.The AHD (1969AHD ( , 1992) ) calls the focal use an adverb; WNWCD (1996: 782) puts the focal use under the verbal use (like to died, which it astutely calls intransitive), but it is said to be syntactically empty even though it precedes words phrases and clauses.The COD (1995: 789) has focal like as an adverb.But all these categorizations fail the standard constituency tests, and do so in the very illustrations the dictionaries use to exemplify them.A conjunction must conjoin grammatical objects of like category: boys and girls, not?? boys and with.The BONE cites the following as illustrative of conjunctive like: And I thought like wow, this is for me.While one might think that this exemplifies the conjunction of two clauses -and I thought + CONJ + this is for menote that no other conjunction can appear in the CONJ slot: ??And I thought since wow / and wow / if wow, this is for me.The OED CD-ROM's (1993) examples are equally misleading: As we say pragmatically in Huddersfield: "C'est la vie, like!" Here, nothing is "conjoined" by like; in any case, no conjunction appears in sentencefinal position in English.Like is not a conjunction here, but a focus particle.
These examples underscore the kinds of form-class choices made' by dictionaries and the way lexicographical tradition reproduces outdated and misleading grammatical theory.Perhaps the most radical lesson of the uneven grammatical treatment of metaforms is the suggestion that form-class judgments should be eliminated from dictionaries altogether.After all, users infrequently consult the dictionary for grammatical class anyway, so why not ter-minate the practice and thus avoid mischaracterization?This would surely evoke protests from the public, who want to see grammar in the dictionary even if the public does not use the dictionary for such a purpose.It would also evoke protests from the publisher's marketing division, which wants the public to see exactly what the public wants to see (Willinsky 1988 has some revealing comments about the interaction of marketing and editorial decisions in dictionaries).
Still, if form-class categorizations are to be included in the dictionary, they should at least be accurate.Modem linguistics has added to the list of formclass categories, so why not class things as they are?At least four new parts of speech should be included in dictionaries as a consequence of an examination of the lexicographical treatment of metaforms: complementizer, discourse marker, focus particle, and quotative.There is no need to use the term metaform, which would be too inclusive and too abstruse to the user.The prefatory material in the dictionary could be easily emended to include these categories -without fancy abstract explanations but with clear illustrations.Some suggestions to this effect are in the last section of this paper.

4.
The definition of metaforms Just as curious as the gramui.aticaltreatment of metaforms in dictionaries are their definitions.Sometimes metaforms are defined as if they uncontroversially denote, just like nouns, verbs, and prepositions.At other times, they are defined as.if they defectively denote or, worse, are meaningless.These discrepancies can be found both within and across dictionaries, suggesting an uneven treatment of metaforms both in-house and across the profession as a whole.There appear to be three strategies to define metaforms: A. define them• in the regular defining formulas of the dictionary and so treat them like all other (normally) denoting items; B.
define them via a usage or distribution statement, not a formulaic definition, and so call atterition to them as a special class; C.
avoid defining them or define them as empty.
The first strategy, normal form, typically applies only to those metaforms that are also easily (and wrongly, see above) put into traditional form-classes.Thus, the AHD (1992: 979) defines the focus particle (n=nee adverb) just as "precisely; exactly ... barely ... merely ... simply ... " The OED CD-ROM (1993) does likewise, .subdividingsenses by distributions: "exactly, precisely ... used freely before a demonstrative ... as an emphatic expletive, strengthening an assertion ... " WNWCD (1996: 734) follows suit and elegantly captures, in a single definition ("neither more nor less than"), the way the particle focuses both sides of a gradient: 'no less than,' just great, and 'no more than,' just five.
Similar results emerge from an examination of the definitions of even, though again the WNWCD has the most semantically sophisticated treatment.
WNWCD (1996: 470) initially defines even as "though it may seem improbable; moreover; indeed," and goes on to give clear variations of this essential meaning, all in a strikingly cle.ar de~ti~nal vocab~ary.From bOth.a ~guistic and lexicographical standpomt, this 15 an admiiable charactenzation of even because it is semantically accurate in casting the form as a focus marker, and hence a comparative expression, yet it remains simple and faithful to the dictionary's entire denotational style. (This practice is one of the features that makes the WNWCD an excellent book.)Discourse markers and other metaforms fare less well in terms of normal form and often evoke strategy #2.Chambers (1983: 866) defines the discourse marker now as "used ... in remonstrance ... or taking up a new point .. , expressing admonition, warning ... " The AHD (1992: 1239) reports: "Used to indicate a change of subject or to preface a remark ... " The OED CD-ROM (1993) notes: "in sentences expressing a command or request, with the purely temporal sense effaced ... Used to introduce an important or noteworthy point ... Inserted parenthetically, or at the end of a clause, with similar force."Even the WNWCD cannot resist distributional statements, which it uses to precede both the excellent definition of even and discourse markers like well (1996: 1516): "an exclamation used to express ... " While this defining style suggests an awareness of these forms as metaforms since they are defined by their conditions of use, the practice has two problematic consequences.First, despite (or because of?) the gesture to pragmatics, this sort of definition often obscures the meaning of the form.What does it mean to say that now means 'used in taking up a new point'?All points?Whose points?Only points?Why are the following disallowed?(lOa) ?? Now, hello.(but Cf.Now, what's your name?)(b) ??She said that, now, she wants to take up a new point, in remonstrance.
Second, this style unnecessarily brackets and calls attention to the meta-use.Indeed, in some dictionaries, metaforms are defined literally in bracketed form.
WNWCD (1996: 470) begins the definition of even with "used as an intensive or emphatic particle meaning," with the usage statement italicized, presumably because all usage labels in WNWCD are in italics.But, in fact, this is a statement of the form-class, not of usage.
The RHO's (1971: 987) definition of now also illustrates the practice of bracketing: "(used to introduce a statement or question)."2The discourse marker now means, in fact, 'upcoming discourse focused on the speaker with the current moment of speech as an initiation point and so this is a break in the flow of information' (Schiffrin 1987: 228-46).This definition resolves both problems in a single sweep -it is in the spirit of normal definition and incor-porates the meta-use into the definition proper.Furthermore, this definition can be constructed only with the help of current pragmatic and discourse theory.
Bringing the definitions of metaforms in line with existing defining formulas and removing all trace of special bracketing might improve definition in other ways.Consider the first two senses of the RHD definition of I (1971: 704): -pron. 1. the nominative singular pronoun used by a speaker in referring to himself ... -no 2. (used to denote the narrator of a literary work written in the first person singular.).
These senses seem unnecessarily confusing to me, tnlXmg grammatical, semantic, and pragmatic facts and bracketing special literary uses.For one thing, I is classed as a pronoun elsewhere in the entry before the first definition, so why repeat this grammatical classification in the first sense?(And why change the classification to a noun in the second sense?)For another, the second sense is really a type of the first.
A simple definition for I could be given that solves all these problems in a single sense: 'speaker self-reference, whatever the ultimate source of the speech itself, sayan author (1) writing the speech of a first person narrator (1) speaking it.'(See Levinson 1989: 68).Narrator self-reference is a type of recursive speaker self-reference, and both can be accommodated in a straightforward denotation, without mixing criteria and without bracketing metafunction.(Cf.WNWCO 1996: 666, which has a remarkably similar definition.) The third defining strategy is to avoid definition or to call metaforms meaningless.Chambers (1983: 866) defines the metaform now as "used meaninglessly .. ."(sic) and then goes on to characterize its meaningful uses!The OED CD-ROM (1993) is also guilty of empty definition for now: "used elliptically in various ways .. ."This practice, quite obviously, is not much help, though it is really not much different in spirit, if not letter, from a distribution statement, which fails to pinpoint the meaning of the term and defers the definition to its collocation.Imagine defining dog as if it had no content: "used to identify, describe, or solicit certain animals." It turns out that all three definitionalpptions can be found in lexicographical treatments of focal like.In the BONE's (1973: 260) initial treatment, like is defined in bracketed and empty fashion -"used without a definite meaning" but then given a determinate characterization of its meaning -"to understate or de-emphasize the word or phrase that follows or precedes it."The later BONE (1990: 289) defines it in normal fashion .. The 1969 AHD brackets it as a pausal or emphatic and assigns it expletive usage (1969: 757); the laterAHO (1992: 1042) eliminates the expletive characterization but keeps the same bracketing.3Even the WNWCO defines focal like as "used without meaning or syntactic function," but the same dictionary takes on equally non-standard metaforms and straightforwardly defines them: the discourse phrase you know is defined as "you understand" (748).
These disparities both within and across dictionaries return us to our earlier lessons on form class.Why not treat metaforms as they are?Focal like denotes 'attenuated, intensified, or very new or very salient information.'The COD (1995: 789), in fact, captures the first part of this denotation by defining focal like as 'so to speak,' a definition also fotind in BONE (1990: 289).If focal even can be defined simply, accurately, and consistently as a focus particle, why should other metaforms be given different treatment?

Conclusion
The foregoing arguments lead to two recommendations: A.
Increase the number of part-of-speech labels to be more sensitive to the variety of grammatical categories in a language.B.
Write definitions that define, even if they are to characterize discourselevel or pragmatic forms.Empty definitions -or defining an item as meaningless -can be misleading.
With respect to the first, dictionaries might include the following categories and descriptions in the prefatory material: Complementizer: a form that precedes and signals only the presence of a clause.Unlike a coordinating conjunction (and) or subordinating conjunction (because), which may optionally mark clauses, a complementizer can signal the presence of a subject clause, and, when marking an object, can sometimes be deleted.These form classes require modem syntactic and discourse theory for proper identification and description.Importing current linguistics and its categories into lexicography is a difficult, uphill battle against the medieval, part-ofspeech tradition.But editorial decisions should at least be sensitive to the negative tradeoffs in accuracy of adherence to such a tradition.It is entirely possible to continue to please the public without overdoing the prefatory material by adding abstruse form-classes.
The second recommendation has a motivation similar to the first's.Current semantic and pragmatic theory can help dictionaries be consistent and complete in definitions.It is one thing to exclude metaforms from the list of entries.But it is another thing to include them and then inaccurately capture the denotation.I know of no "meaningless forms" in any language.Even those forms that appear to be entirely grammatical in their meaning and hence "empty" (e.g., of in a picture of Bob) have determinate senses that are captured by the traditional defining formulas of dictionaries: Cf.WNWCD on of (1996: 940).
Why not define all metaforms instead of focusing on a subset and characterizing those in terms of distribution statements?
COBUILD has shown that it is possible to have a simple and consistent defining style.It also is possible to have a defining vocabulary that accepts form classes beyond what leXicography has traditionally propagated without overcomplicating the attendant definitions.Lexicography is very up-to-date in its method -witness the prevalence of machine-readable databases.Why, then, perpetuate an outdated semantic meta theory?
By excluding certain metaforms or, more importantly, by including them and treating them as they do, dictionaries make choices that sometimes quietly reproduce social and political sympathies contrary to the otherwise expressed leanings pf the dictionaries themselves.Thus a dictionary that includes focal like, calls it nonstandard or colloquial, and then defines it as meaningless sends very mixed signals.The dictionary satisfies its expressed obligation to record actual speech, but then it simultaneously devalues that speech by calling it empty.Notes 1.
This judgment is not restricted to lexicography.The most recent theoretical semantics text lists some focus particles as adverbs (Larson and Segal 1995: 301).2.
I have been unable to find an explanation of this pr~ctice in the RHD's guide to the dictionary.
Reproduced by Sabinet Gateway under licence granted by the Publisher (dated 2011) http://lexikos.journals.ac.za Focus 'particles are almost uniformly classed as adverbs.The AHD (1992:   979), for instance, categorizes just this way, andChambers (1983: 686), theCOD   (1995: 737), and theRHO (1971: 775)  follow suit.The OEDCD-ROM (1993) (7a) He looks like a good prospect for the job (RHO) (b) ?? A good prospect -he looks like it for the job.(Sa) He looked over the fence for his father (b) Quotative: a form that indicates the presence of direct or imitated speech, thought, or behavior.Examples of quotatives are say, do, and, colloquially, go and like: She said / went / was like, "What are you doing?" Examples of complementizers are (the fact) that, if, and whether: That John bought a new car surprised me (subject); I know that/i//whether/O John bought a new car (object and deletion).Cf. ?? Since John bought a new car surprised me and ?? I know sincelbecause/and John bought a new car.somehow made salient.Examples of focus particles are even and just: Even the DOG ate that caSserole, The dog even ATE that casserole, The dog ate even THAT casserole: