No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Long-standing issues in adjective order and corpus evidence for a multifactorial approach
Abstract
In this paper, we introduce the issue of adjective order and show that different approaches vary in their answers to the question of how fine-grained the semantic categories determining adjective order are. We report on a corpus study that we conducted and that illustrates that a clear answer to the question of what general factors exactly determine adjective order is elusive, given the multifactorial nature of the problem. We then present the individual contributions to this special issue, and how they attempt to add new observations from Germanic languages to the general issues revolving around the topic of adjective order.
... While these hierarchies are powerful predictors of adjective order, they are not really explanatory: the question remains how such hierarchies are formed by the underlying properties of adjectives. Furthermore, corpus studies have shown that many counterexamples against various versions of these hierarchies can be found (Truswell, 2009), demonstrating the need for a more fine-grained analysis (Trotzke and Wittenberg, 2019). ...
... Finally, various papers highlight the need for a multi-factorial account of adjective order: it is unlikely adjective order is driven by a single factor and therefore it's the result of various competing pressures (Wulff, 2003;Trotzke and Wittenberg, 2019;Dyer et al., 2023). Relatedly, a recent position paper (Levshina et al., 2023) argues for a fundamentally gradient theory of word order that would be able to capture its hybrid and multi-faceted nature. ...
... We present the predictive accuracy of these relative n-gram occurrences in Table 4.1. %(a 1 ) predicts adjective order with 60.5%, which provides weak evidence that more frequent adjectives tend to occur first (Martin, 1969;Trotzke and Wittenberg, 2019). %(a 1 a 2 ) is the strongest predictor of adjective order-except for the LMs-at 90.3%. ...
When we speak, write or listen, we continuously make predictions based on our knowledge of a language's grammar. Remarkably, children acquire this grammatical knowledge within just a few years, enabling them to understand and generalise to novel constructions that have never been uttered before. Language models are powerful tools that create representations of language by incrementally predicting the next word in a sentence, and they have had a tremendous societal impact in recent years. The central research question of this thesis is whether these models possess a deep understanding of grammatical structure similar to that of humans. This question lies at the intersection of natural language processing, linguistics, and interpretability. To address it, we will develop novel interpretability techniques that enhance our understanding of the complex nature of large-scale language models. We approach our research question from three directions. First, we explore the presence of abstract linguistic information through structural priming, a key paradigm in psycholinguistics for uncovering grammatical structure in human language processing. Next, we examine various linguistic phenomena, such as adjective order and negative polarity items, and connect a model's comprehension of these phenomena to the data distribution on which it was trained. Finally, we introduce a controlled testbed for studying hierarchical structure in language models using various synthetic languages of increasing complexity and examine the role of feature interactions in modelling this structure. Our findings offer a detailed account of the grammatical knowledge embedded in language model representations and provide several directions for investigating fundamental linguistic questions using computational methods.
... (7) SIZE>LENGTH>HEIGHT>SPEED>DEPTH>WIDTH>WEIGHT> TEMPERATURE>WETNESS>AGE>SHAPE> COLOR>MATERIAL> NP As a currently highly prominent model in the domain of complex NP research, the cartographic model has attracted a lot of criticism, ranging from the problem of innateness, origin, and functional projection proliferation to the problem of rigidity (e.g., Svenonius 2008, Truswell 2009, Scontras et al. 2017, Leivada and Westergaard 2019, Larson 2021. Given the conclusions based on large databases (cases of actual use of multiple adjective strings), the concerns of corpus studies focusing on the rigidity problem, or the so-called empirical undergeneration problem, seem particularly relevant (Wulff 2003, Truswell 2009, Kotowski and Hartl 2019, Trotzke and Wittenberg 2019. Since a rigid order is assumed as a direct output of adjectival functional hierarchy within the cartographic model, the data that do not attest to restrictions in ordering among the proposed semantic categories could be problematic for this strictly syntactic approach. ...
... And finally, although much attention has recently been paid to testing adjective order constraints with the corpus approach, no corpus study has really directly examined the predictions of the cartographic model, even though the cartographic model is currentlydominant in the domain of adjective order research, and Scott's (2002) proposal, moreover, is one of the most fine-grained of all offered so far. Generally, corpus studies test the predictions of Scott's cartographic model either indirectly, against other semantic classifications, or within a limited set of semantic categories and nouns (Wulff 2003, Truswell 2009, Scontraset al. 2017, Kotowski and Hartl 2019, Trotzke and Wittenberg 2019. ...
... & lemma="..."] [tag="P.*" & lemma="..."] 6 Frequency and length have been shown in previous studies to be factors that can influence the adjective order in multiple adjective strings (e.g., Wulff 2003, Scontras et al. 2017, Kotowski and Hartl 2019, Trotzke and Wittenberg 2019. However, it is not entirely clear why frequency and length affect adjective order in some cases more than in others. ...
... AOPs are assumed to be present whenever the speaker uses two or more attributive adjectives; the opposite order is regarded as marked or strongly dispreferred. AOPs have been investigated for many languages since the 70s (Dixon 1977) and, in recent years, the interest in understanding AOPs has revived (e.g., Scontras et al.2017, Trotzke & Wittenberg 2019, Kotowski & Härtl 2019. The current debate asks i.a. ...
... Comparing three hypotheses (ordering hierarchies, subjectivity and positional input frequency) in a probabilistic learning study of adjective orders in child corpora, Bar-Sever (2019) found that all three hypotheses were equally likely. Accordingly, it does not come as a surprise that some approaches explicitly postulate multifactorial explanations (Trotzke & Wittenberg 2019, Wulff 2003. ...
... The finding stated above -that less subjective adjectives in Hebrew, from two different semantic classes, were not preferred at positions in close proximity to the noun -is incompatible with the conceptual accounts. To explain our results, we turn to accounts that suggest that several properties of the adjectives, not solely conceptual ones, influence the ordering of adjectives in multi-adjectival phrase (Trotzke & Wittenberg, 2019;Wulff, 2003;Wulff & Gries, 2015). These could be semantic or conceptual factors (e.g., availability or absoluteness) as suggested by several accounts, but also lexical (e.g., length or frequency; see also Scontras et al.,2017, for an examination of these factors), or phonological (e.g., syllable structure) factors. ...
... These factors interact when determining the ordering of the adjectives. Corpus analyses and experimental results have shown that length and frequency contribute to the ordering preferences in English (Scontras et al., 2017;Trotzke & Wittenberg, 2019;Wulff, 2003;Wulff & Gries, 2015). Specifically, shorter adjectives tend to precede longer ones, and more frequent adjectives also tend to precede less frequent ones. ...
Referring to a specific object sometimes requires using multiple adjectives. The ordering of the adjectives is assumed to be constrained by universal hierarchies (grammatical or conceptual). It is therefore predicted that different languages will present similar ordering preferences. The ordering in languages where adjectives appear after the noun is further expected to mirror the ordering observed in languages where adjectives appear before the noun. We investigated these predictions in prenominal (English) and postnominal (Hebrew) languages, using three different tasks: production, naturalness rating, and forced-choice. English speakers showed a robust ordering preference. The preferences in Hebrew were significantly weaker. Moreover, for some of the adjective strings, the weakly preferred orders in Hebrew did not mirror the preferred order in English. We argue that constraints on adjective ordering must include additional factors beyond the suggested hierarchies, and discuss this implication in relation to differences between prenominal and postnominal languages.
... At the same time, the latter remains influential in current accounts of adjective ordering. A number of recent works follow standard cartographic assumptions in accepting the adjective hierarchy to be innate and universal (Sheehan, 2017;Huang, 2017), in acknowledging it as one of the possible explanations on the topic of adjective ordering (Payne, 2018;Trotzke & Wittenberg, 2019), or in developing proposals that put forth less fine-grained hierarchies that are, however, quite similar to those explored in cartography (Pereltsvaig & Kagan, 2018). In sum, it has been recently argued that it is still unclear which theory is the most adequate to explain the attested data (Trotzke & Wittenberg, 2019), while several attempts to experimentally verify the basic claims of cartography in relation to adjective ordering are currently in progress (Mišmaš, Marušič & Žaucer, 2018;Plesničar, 2018;Dolenc, 2018). ...
... A number of recent works follow standard cartographic assumptions in accepting the adjective hierarchy to be innate and universal (Sheehan, 2017;Huang, 2017), in acknowledging it as one of the possible explanations on the topic of adjective ordering (Payne, 2018;Trotzke & Wittenberg, 2019), or in developing proposals that put forth less fine-grained hierarchies that are, however, quite similar to those explored in cartography (Pereltsvaig & Kagan, 2018). In sum, it has been recently argued that it is still unclear which theory is the most adequate to explain the attested data (Trotzke & Wittenberg, 2019), while several attempts to experimentally verify the basic claims of cartography in relation to adjective ordering are currently in progress (Mišmaš, Marušič & Žaucer, 2018;Plesničar, 2018;Dolenc, 2018). Fitting the purposes of the third group of studies, the present work aims to add to the several recent works that have produced results that call the cartographic claims into question (e.g., Scontras, Degen & Goodman, 2017;Kotowski & Härtl, 2019). ...
Background:
Linguists and psychologists have explained the remarkable similarities in the orderings of linguistic elements across languages by suggesting that our inborn ability for language makes available certain innately wired primitives. Different types of adjectives, adverbs, and other elements in the functional spine are considered to occupy fixed positions via innate hierarchies that determine orderings such as A>B>C, banning other permutations (*B>C>A). The goal of this research is to tap into the nature and rigidity of such hierarchies by comparing what happens when people process orderings that either comply with them or violate them.
Method:
N = 170 neurotypical, adult speakers completed a timed forced choice task that featured stimuli showing a combination of two adjectives and a Spelke-object (e.g., 'I bought a square black table'). Two types of responses were collected: (i) acceptability judgments on a 3-point Likert scale that featured the options 'correct', 'neither correct nor wrong', and 'wrong' and (ii) reaction times. The task featured three conditions: 1. size adjective > nationality adjective, 2. color adjective > shape adjective, 3. subjective comment adjective > material adjective. Each condition had two orders. In the congruent order, the adjective pair was ordered in agreement with what is traditionally accepted as dictated by the universal hierarchy. In the incongruent order, the ordering was reversed, thus the hierarchy was violated.
Results:
In the first experiment, the results of n = 140 monolinguals showed that across conditions, both congruent and incongruent orders were generally accepted as correct. For 2/3 conditions, the difference in acceptability ratings between congruent and incongruent orders did not reach statistical significance. Using time as a window to processing, reaction times showed that incongruent orders do not take longer to process than congruent ones, as should be the case if the former were treated as being licensed under some type of special condition (e.g., contrastive focus) that reverses the unmarked order and legitimizes the violation of the hierarchy. In the second experiment, the results of n = 30 bidialectals, tested in both language varieties, corroborated the findings of the first experiment.
Conclusions:
Our findings do not provide evidence for an innate hierarchy for adjective ordering that imposes one rigid, unmarked order. We discuss the importance of notions such as subjectivity and inherentness, and show that for some conditions, not only is there no evidence for a hard constraint that bans incongruent orders, but even simple preferences of congruent orders over incongruent ones are hard to discern. Capturing the bigger picture, given that both the hierarchies and their legit permutations have been described as innate, our results reduce the amount of primitives that are cast as innate, eventually offering a deflationist approach to human linguistic cognition.
... Frequency Several authors have shown adjective frequency to be a reliable predictor of adjective order, with more frequent adjectives appearing earlier (Martin, 1969;Wulff, 2003;Scontras et al., 2017;Trotzke and Wittenberg, 2019;Westbury, 2021). This effect of frequency is consistent with a broader finding that more frequent words appear earlier in sentences. ...
The literature on adjective ordering abounds with proposals meant to account for why certain adjectives appear before others in multi-adjective strings (e.g., the small brown box). However, these proposals have been developed and tested primarily in isolation and based on English; few researchers have looked at the combined performance of multiple factors in the determination of adjective order, and few have evaluated predictors across multiple languages. The current work approaches both of these objectives by using technologies and datasets from natural language processing to look at the combined performance of existing proposals across 32 languages. Comparing this performance with both random and idealized baselines, we show that the literature on adjective ordering has made significant meaningful progress across its many decades, but there remains quite a gap yet to be explained.
... In Scott's (2002) words, almost "all writers claim that AO[P]s can be adequately accounted for using broad "psychological" criteria, yet none of them are able to provide a convincing argument-which is, moreover, consistent with the data-for a psychological basis to AO[Ps]". More recent studies have re-affirmed that a clear answer to the question of what factors drive AOPs remains elusive (Trotzke and Wittenberg, 2019). Filling this knowledge gap requires (i) connecting linguistic structures observed at the phenotypic level to their cognitive underpinnings and (ii) bringing into the picture environmental triggers that may result to variation within or across linguistic communities. ...
When people are asked to create a phrase with the elements {blue, earrings, beautiful}, they produce ‘beautiful blue earrings’. Several theories have been proposed about the origins of this universal tendency to order multiple adjectives in a specific way: an innate universal hierarchy with designated positions for each category of adjectives, sensitivity to the definiteness of the adjectival denotation, availability and psychological closeness of the adjective attributes to the speaker, the encoding of subjective vs. objective properties, and the adjective’s phonological weight. Although these theories have strong descriptive power, they often focus on what happens at the phenotypic level without explaining what cognitive needs trigger this behavior. Through a timed task that measures acceptability in ‘Adjective-Adjective-Noun’ sequences that either comply with the universal order or violate it, we adduce evidence for the high acceptability of the violating orders, whose processing did not take longer than that of the compliant orders, as should have happened if the former were non-canonical. The results suggest that ordering preferences exist but are not invariable, as one would expect if a strong linguistic universal was involved. We track the origin of adjective ordering preferences to the synergistic interplay of three cognitive biases: Zipf’s Law, Intolerance of Ambiguity, and Novel Information Bias. Last, we show that the linguistic manifestation of these preferences is sensitive to the statistical distribution of the input data, resulting to variation even among speakers of the same language.
... This necessity for multi-causal explanations has led to the rise of multifactorial quantitative analyses of linguistic data (Bresnan et al., 2007). Ordering phenomena have increasingly become an object of quantitative study (Trotzke & Wittenberg, 2019). While many scholars (e.g., Lohmann, 2014;Lohmann & Takada, 2014;Morgan & Levy, 2016;Wulff, 2003;Wulff & Gries, 2015) have adopted the multifactorial method in examining word order, there have been no multifactorial studies on antonym ordering. ...
This study presents a contrastive analysis of antonym order in English and Chinese coordinate structures using a multifactorial corpus method. The analysis yields the following findings. First, antonym ordering in coordinate structures is driven largely by the same ordering constraints across the two languages. Chronology and Positivity are the most prominent semantic motivating constraints, and Morphology is the most important formal motivating constraint. Second, there are some differences in constraints on antonym ordering between English and Chinese. Age and Gender exert a significant effect on Chinese antonym ordering, but show no effect on English antonym ordering. Hierarchical Superiority significantly affects Chinese antonym ordering, but has only a marginal effect on English antonym ordering. Furthermore, this study explores the motivations underlying these similarities and differences and argues that there is a clear correlation between iconicity and antonym order. The exploration reveals that their similarities in semantic constraints may be attributed to the iconicity of closeness and the iconicity of temporal sequence, while their differences may be ascribed to the iconicity of cultural values and norms. Their similarities in formal constraints stem from cognitive accessibility. These findings provide further evidence that antonym order is determined by general cognitive principles.
... Р. 230-239]. Неоднозначный характер прилагательных приводит к тому, что их не всегда легко распознать в тексте, верно употребить и, в частности, выстроить порядок их следования в цепочке определений [Abd Trotzke, Wittenberg, 2019]. ...
This paper presents an analysis of the deep language factors that predetermine polysemy of English adjectives denoting moral and mental qualities of human beings. In line with a well-established point of view in cognitive linguistics, this study treats the semantics of a word as a two-level phenomenon possessing the semantic (external) level and the conceptual (internal) level. Given polysemy belongs to the external level, this study aims to reveal the internal language factor allowing for umbrella adjectives to develop meanings of moral and mental qualities. This is the first research that has analyzed English adjectives from this perspective; it is proposed to unearth the deep language foundation of polysemy by modeling the conceptual foundation of polysemantic adjectives, which is undertaken via analysis of their etymological data. The choice of the adjectives encoding moral and mental qualities is substantiated by the following reasons: first, these words name the major human characteristics, whose recognition and verbalization can be traced back to the Pre-Old English period; second, they denote abstract qualities unperceivable by senses but estimated due to their indirect manifestation in individuals’ judgments, conduct and activity; third, since these adjectives convey evaluation of the quality, they reflect cultural axiological standards. The findings show that the semantics of the English adjectives in question is governed by a certain set of conceptual metaphors. The commonality of the adjectives’ conceptual basis seems to be the internal language factor that accounts for polysemy, i.e. an ability for an adjective to comprise meanings of mental and moral characteristics. In addition, the results demonstrate that the unearthed concepts form oppositions, namely, LIFE - DEATH, MOTION - STILLNESS, FRIEND - FOE. The opposed concepts are endowed with the positive or negative value that appears to determine the evaluative meaning of the adjectives. Besides, the research has shown that, while participating in the formation of adjectival semantics, the concepts can demonstrate ambiguous value, which enables a concept to underlie both the positive and negative evaluative meanings of an adjective; therefore, an adjective may comprise meanings of mental and moral characteristics that are opposite in their evaluation.
... Unlike the ordering of color and pattern, the relative ordering of other attributes has been shown to be semantically or syntactically restricted (e.g., Culbertson, Schouwstra, & Kirby, in press;Dixon, 1982;Hetzron 1978;Quirk, Greenbaum, Leech, & Svartik, 1985;Scontras, Degen, & Goodman, 2017;2019;Scott, 2002;Trotzke & Wittenberg, 2019;Whorf, 1945). Thus, one may wonder to what extent the current findings depend on the fact that the ordering of color and pattern attributes can be varied more freely than the ordering of, for example, size and color. ...
Using eye-tracking, we examined if over-specification hinders or facilitates referent selection, and the extent to which this depends on the properties of the attribute mentioned in the referring expressions and the underpinning processing mode. Following spoken instructions, participants selected the referent in a visual display while their eye movements were monitored. The referring expressions were presented either simultaneously with the displays, so the attributes could be incrementally processed in sequence, or before the display presentation, so the attributes could be processed in parallel from the outset of search. Experiment 1 showed that when the attributes were processed incrementally, how quickly an earlier-mentioned attribute discriminated determined whether a late-mentioned, over-specified attribute contributed to discrimination: When color was mentioned first and fully discriminating, the referent was selected fast regardless of the second-mentioned pattern, whereas when pattern was mentioned first and fully discriminating, the second-mentioned color facilitated discrimination. Experiment 2 found that under incremental processing, color mention after a fully discriminating pattern increased fixations but delayed referent selection relative to a pattern-only description; under parallel processing, however, color mention immediately eliminated alternatives and sped up referent selection. Experiment 3 showed that pattern mention after a fully discriminating color delayed referent selection and tended to reduce fixations relative to a color-only description in both processing modes. Hence, additional attributes can speed up referent selection but only when they can discriminate much faster than alternative attributes mentioned in a more concise description, and critically, when they can be used early for referent search.
... For instance, it has been shown extensively that the more frequent word tends to appear first in binomial expressions in English (Morgan and Levy, 2015;Benor and Levy, 2006;Fenk-Oczlon, 1989;Gustafsson, 1976). Other studies have demonstrated similar patterns for adjective orderings in English (Trotzke and Wittenberg, 2019;Wulff and Gries, 2015;Wulff, 2003). Here I asked whether lexical frequency predicts PP ordering typology, i.e., whether the PP with more frequent individual words tends to appear first. ...
Why are languages the way they are? In this dissertation, I take up this question with
a focus on crosslinguistic constituent orderings. Specifically, borrowing insights from
language processing and language evolution, I ask what abstract constraints as well as
idiosyncratic biases govern language users’ choice among grammatical alternatives of the
same syntactic constructions across genres and languages. Adopting a data-driven ap-
proach, I explore three directions in particular. First, from Chapter 3 to Chapter 6,
taking advantage of large-scale multilingual corpora, I investigate and quantify the roles
of numerous factors that are motivated by long-standing linguistic theories as well as previ-
ous empirical findings in word order preferences. I show that while the effect of individual
factors depends on the ordering structures of different languages, generally the predictive
power and direction of these constraints are more dependent on whether the orderings
are in the preverbal or the postverbal domains. In addition, besides these abstract con-
straints that yield probabilistic typological tendencies, in Chapter 7 I ask why language
users have idiosyncratic ordering preferences and how regularization of this idiosyncrasy
arises diachronically, using Bayesian iterated learning models that simulate the process
of language change. Lastly, I adopt the theoretical framework of dependency syntax to
develop a dependency treebank for Hupa, an endangered Dene language of northwestern
California, as a way to formalize and model the syntax of indigenous languages.
... Recent years have seen similar discussions; in particular, the distinction between adjectives and determiners has been justified for Serbian (LaTerza, 2015), a double semantic value of the Gothic adjective sama has been stated (Ratkus, 2018), and a mixed category combining morphemic and syntactic peculiarities of nouns and adjectives has been singled out in the Tungus language (Nikolaeva, 2008). The ambiguity of adjectives leads to difficulty in their identification in texts, proper use, and proper adjective ordering (Abd Rahim, 2013;Sosnina, 2015;Berg, 2019;Trotzke & Wittenberg, 2019). Although the dominant opinion is that adjectives are not a universal language category, researchers agree that "every language seems to be equipped with adjectival concepts" (Pustet, 2006, p. 61). ...
This paper focuses on the cognitive foundation of the semantics of English adjectives that denote mental and moral characteristics of human beings. Research into these adjectives seems a challenging task because they denote abstract qualities that cannot be perceived through vision, hearing, or touch; and here a question arises: How are abstract qualities interpreted in English encoded through adjectives? To answer it, this study follows the idea of two-level semantics, i.e. word semantics is treated as a two-level phenomenon that comprises the semantic (external) level and the conceptual (deep) one. This study is the first to address adjectival semantics from this perspective. Here a novel approach to revealing the cognitive foundation of adjectives is introduced: given that adjectives originated from old syncretic items and a word cognitive model forms at the moment of word creation, cognitive models underlying adjectives' semantics are unearthed via analysis of their etymological data. Our contribution is two-fold. First, the approach has revealed that the image schema CONTAINER guides semantics of an array of various adjectives independent of their morphemic structure or date of origin. The examples demonstrate that abstract human qualities are interpreted via the following container features: boundary, container substance, size, hardness/softness of a container shell, etc. The semantics of affixed or compound adjectives appear to stem from the integration of concepts represented by an affix and a root or two roots, respectively. Second, the findings show that the value given to every container feature appears to predetermine the evaluation conveyed by an adjective. Container features tend to possess ambivalent value, realizing the positive or negative one due to the interaction with a frame in which the CONTAINER is incorporated, therefore the same polysemantic adjective may develop both positive and negative meanings. To reveal the whole inventory of cognitive models that govern adjectival semantics in English, further research needs to be conducted.
Adjective ordering preferences have been addressed by theoretical and empirical studies. Some accounts propose that the distance of an adjective from the head noun depends on its semantic/conceptual features such as subjectivity. Subjectivity has been observed to reliably predict adjective ordering cross-linguistically, albeit with variation in strength. We propose that cross-linguistic variation might stem from lexical factors, which might operate differently in pre- and post-nominal languages. Frequency, for example, may affect ordering linearly with frequent words appearing earlier in the string, rather than based on distance from the noun. Our study aimed at examining this hypothesis, using a binary forced-choice task contrasting two adjective orders in a post-nominal language (i.e. Hebrew). Our results suggest that subjectivity is indeed a strong predictor for ordering preferences, but its effect interacts with lexical factors. Our findings highlight the importance of studying a diversity of languages, where linguistic phenomena might manifest differently.
Adjective ordering preferences stand as perhaps one of the best candidates for a true linguistic universal: When multiple adjectives are strung together in service of modifying some noun, speakers of different languages—from English to Mandarin to Hebrew—exhibit robust and reliable preferences concerning the relative order of those adjectives. More importantly, despite the diversity of the languages investigated, the very same preferences surface over and over again. This tantalizing regularity has led to decades of research pursuing the source of these preferences. This article offers an overview of the findings and proposals that have resulted.
Expected final online publication date for the Annual Review of Linguistics, Volume 9 is January 2023. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.
We ask whether word order preferences for binomial expressions of the form A and B (e.g. bread and butter) are driven by abstract linguistic knowledge of ordering constraints referencing the semantic, phonological, and lexical properties of the constituent words, or by prior direct experience with the specific items in questions. Using forced-choice and self-paced reading tasks, we demonstrate that online processing of never-before-seen binomials is influenced by abstract knowledge of ordering constraints, which we estimate with a probabilistic model. In contrast, online processing of highly frequent binomials is primarily driven by direct experience, which we estimate from corpus frequency counts. We propose a trade-off wherein processing of novel expressions relies upon abstract knowledge, while reliance upon direct experience increases with increased exposure to an expression. Our findings support theories of language processing in which both compositional generation and direct, holistic reuse of multi-word expressions play crucial roles.
The need to distinguish two syntactic sources for adnominal adjectives (a direct modification and a relative clause one), with the associated interpretive properties, turns out to have certain implications for the semantic classification of adjectives. We will see that it provides evidence for the existence of a class of truly privative adjectives, and for the conclusion that subsective adjectives of the tall/big type and of the skillful/good type cannot be reduced for the direct modification class to (context-sensitive) intersectivity. Furthermore, the available syntactic evidence appears to converge with Partee's more recent position that the principal divide is between predicative and non predicative adjectives, which correspond in the adnominal case to the relative clause and direct modification sources, respectively. © 2014 by T.T. Ng Chinese Language Research Centre, Institute of Chinese Studies, The Chinese University of Hong Kong.
Why is it preferable to say salt and pepper over pepper and salt? Based on an analysis of 692 binomial tokens from online corpora, we show that a number of semantic, metrical, and frequency constraints contribute significantly to ordering preferences, overshadowing the phonological factors that have traditionally been considered important. The ordering of binomials exhibits a considerable amount of variation. For example, although principal and interest is the more frequent order, interest and principal also occurs. We consider three frameworks for analysis of this variation: traditional optimality theory, stochastic optimality theory, and logistic regression. Our best models—using logistic regression—predict 79.2% of the binomial tokens and 76.7% of types, and the remainder are predicted as less frequent—but not ungrammatical—variants.
This paper is concerned with the question of which factors govern prenominal adjective order (AO) in English. In particular, the analysis aims to overcome shortfalls of previous analyses by, firstly, adopting a multifactorial approach integrating all variables postulated in the literature, thereby doing justice to the well‐established fact that cognitive and psychological processes are multivariate and complex. Secondly, the phenomenon is investigated on the basis of a large corpus, rendering the results obtained more representative and valid of naturally occurring language than those of previous studies. To this end, corpus‐linguistic operationalizations of phonological, syntactic, semantic and pragmatic determinants of AO are devised and entered into a Linear Discriminant Analysis, which determines the relative influence of all variables (semantic variables being most important) and yields a classification accuracy of 78%. Moreover, by means of the operationalizations developed in this analysis, the ordering of yet unanalyzed adjective strings can be predicted with about equal accuracy (73.5%).
A new analysis of adjectives, supported by comparative evidence.
In The Syntax of Adjectives, Guglielmo Cinque offers cross-linguistic evidence that adjectives have two sources. Arguing against the standard view, and reconsidering his own earlier analysis, Cinque proposes that adjectives enter the nominal phase either as “adverbial” modifiers to the noun or as predicates of reduced relative clauses. Some of his evidence comes from a systematic comparison between Romance and Germanic languages. These two language families differ with respect to the canonical position taken by adjectives, which is prenominal in Germanic and both pre- and postnominal in Romance. Cinque shows that a simple N(oun)-raising analysis encounters a number of problems, the primary one of which is its inability to express a fundamental generalization governing the interpretation of pre- and postnominal adjectives in the two language families. Cinque argues that N-raising as such should be abandoned in favor of XP-raising—a conclusion also supported by evidence from other language families. After developing this framework for analyzing the syntax of adjectives, Cinque applies it to the syntax of English and Italian adjectives. An appendix offers a brief discussion of other languages that appear to distinguish overtly between the two sources of adjectives.
This monographs investigates into the influence of the individual-/stage-level distinction (IL/SL) on order restrictions of multiple prenominal adjectives (AORs). It rejects the restriction regularly postulated-across different research frameworks-that SL-adjectives are being realized farther from the head noun than IL-adjectives, relegating the alleged constraint to an epiphenomenon of more general principles. While formal-theoretic hypotheses on AORs are formulated and put to the test empirically via a large corpus as well as two rating studies, the book also addresses adjective classification, modification patterns, and the IL-SL-debate in general. The preferred prenominal positions of typical SL-adjectives are argued to follow from their nature as absolute-gradable adjectives as well as from the distinction between object- and kind-modification. The empirical studies corroborate these considerations. The book critically discusses and opposes several well-established hypotheses on AORs, sketches a flexible and parsimonious syntax of adjectival modification, and will be of interest to syntacticians and semanticists working on DP-structure, the IL-SL-debate, and adjectival modification. © 2016 Walter de Gruyter GmbH, Berlin/Boston. All rights reserved.
From English to Hungarian to Mokilese, speakers exhibit strong ordering preferences in multiadjective strings: “the big blue box” sounds far more natural than “the blue big box.” We show that an adjective's distance from the modified noun is predicted not by a rigid syntax, but by the adjective's meaning: less-subjective adjectives occur closer to the nouns they modify. This finding provides an example of a broad linguistic universal—adjective ordering preferences—emerging from general properties of cognition.
English displays well-known restrictions on the ordering of multiple prenominal adjectival modifiers (see Bloomfield, 1933; Whorf, 1945; Lance, 1968; Vendler, 1968; Quirk et al, 1972 among numerous others). Most descriptions include a hierarchy such as the following: QUALITY > SIZE > SHAPE > COLOR > PROVENANCE.1
Traditional analysis of variance (ANOVA) software allows researchers to test
for the significance of main effects in the presence of interactions without
exposure to the details of how the software encodes main effects and
interactions to make these tests possible. Now that increasing numbers of
researchers are using more general regression software, including mixed-effects
models, to supplant the traditional uses of ANOVA software, conducting such
tests generally requires greater knowledge of how to parameterize one's
statistical models appropriately. Here I present information on how to conduct
such tests using R, including relevant background information and worked
examples.
The view that the syntactic description of a sentence can provide a model of certain habits involved in sentence encoding has not been extensively tested. A theory of encoding was proposed in which some aspects of order in the base were taken to correspond to the psychological order of choice of morphemes for production. The theory was tested for English noun phrases with prenominal adjectives and for Indonesian noun phrases with post-nominal adjectives. The results were consistent with predictions based on the theory. An analysis-by-synthesis hypothesis of decoding was proposed, which contained the proposed encoding model as a subcomponent. Experimental results did not confirm the predictions based upon the decoding hypothesis.
Cinque (1994) and Scott (2002) propose conceptions of nominal functional structure in which a rigid and highly elaborated series of functional heads regulates the distribution of attributive adjectives according to the class of property that they denote. The following are their proposed adjectival hierarchies:
(1)
a. Serialization of adjectives in event nominals poss[essive] > cardinal > ordinal > speaker-or[iented] > subj[ect]-or[iented] > manner > thematic
Serialization of adjectives in object-denoting nominals poss[essive] > cardinal > ordinal > quality > size > shape > color > nationality
(Cinque 1994:96)
b. determiner > ordinal number > cardinal number > subjective comment > ?evidential > size > length > height > speed > ?depth > width > weight > temperature > ?wetness > age > shape > color > nationality/origin > material > compound element > NP
(Scott 2002:114)
The main interest of these heterogeneous and apparently arbitrary hierarchies stems from the hypothesis that the heads regulating adjective order may form part of the linear functional sequence often assumed to regulate nominal morphosyntactic and semantic properties, a connection that Cinque (1999) investigated with respect to adverbials and clausal functional structure. Viewed in this light, these hierarchies make a clear and testable prediction: movement aside, and disregarding exceptional word orders linked to marked information structures, if two adjectives belong to different classes, only one relative order of the two should be possible. Section 1 tests this prediction against data gathered from www.google.co.uk. The data attest to a far greater freedom of order among multiple adjectives than predicted by the models in (1). Section 2 scrutinizes the wider project of relating adjective orders and nominal morphosyntax and semantics through a rigidly ordered functional sequence.
The choice of Google as a huge, but unregulated, corpus was made on the basis of the finding that, although the British National Corpus contains 262,838 tokens of pairs of attributive adjectives, over 76% of adjective pairs occur only once, and therefore trivially show only one order (Malouf 2000). For a study that aims to investigate consistency of ordering of adjective pairs, then, even the BNC is not large enough to attribute much significance to the results.
Searches were carried out on Google for prototypical members, in English, of six classes of adjective—namely, the five lowest classes of Cinque's object-denoting hierarchy, plus the material class from Scott's hierarchy. In addition, a number of modal adjectives, which do not have a clear place in either of the taxonomies, were considered. The search terms consisted of pairs of adjacent adjectives from these classes in a specified order. As this squib is concerned solely with canonical Adj-Adj-N constructions, many examples returned by Google were irrelevant, including many cases where, for example, the Adj-Adj sequence does not modify a noun or is disrupted by punctuation (arguably indicating marked prosody and information structure), or the Adj-N sequence is idiomatic (for example, big top or new potatoes). This means that Google is highly problematic as a reliable indicator of relative frequency of different orders: too many of the items obtained must be judged ungrammatical or discarded as irrelevant, and statistical patterns are expected to be too approximate for confidence. Instead, Google was used heuristically in the study reported here, as a potential source of positive evidence for the existence of grammatical examples of certain adjective orders. All examples given below are from Google, representing the tokens judged to be most acceptable from the first 100 hits returned. Furthermore, the sites from which the examples were taken were consulted, to verify that there was no evidence for marked information structures.1 The grammaticality judgments given do not reflect absence of a sequence of adjectives on Google, then; rather, they reflect absence of grammatical NP constituents. I use "??" to indicate that all attested Adj-Adj-N constituents found on Google are judged unacceptable by the native speakers I consulted, and "*" to indicate that no Adj-Adj-N constituents were found.
One clear division shown in the data is between the two subsective, and the four intersective, classes examined.2 The four intersective classes considered are shape (illustrated here with circular), color (red), nationality (French), and material (wooden). Pairs of adjectives drawn from these classes are well attested in any order (see (2)). The same is true...
Most of the attributive adjectives in French occur in a postnominal position. Some adjectives are found in a prenominal position when they are quantificational, focused or short (weak) forms. On the basis of an articulate DP-structure, rich in functional projections, we will explore various types of DP-internal movement (cyclic, remnant and pied-piping XP-movement), which can account for the placement of attributive adjectives not only with respect to the noun they modify, but also with respect to postnominal PP complement/adjuncts. We will also take adjective ordering into consideration following a specifier-based analysis of adjective positioning [Paths Towards Universal Grammar, Georgetown University Press, Georgetown, p. 85]: co-occurring adjectives are merged as the specifiers of their corresponding semantic category in accordance with a strict hierarchy of DP-internal functional projections. The various types of XP-movement with the DP-structure and the occurrence of right-hand adjectives in a predicate position affect the linear placement of adjectives, not only with respect to the noun (pre/postnominal placement), but also with respect to each other (mirror image of adjective ordering).
Osherson and Smith (1981, Cognition, 11, 237-262) discuss a number of problems which arise for a prototype-based account of the meanings of simple and complex concepts. Assuming that concept combination in such a theory is to be analyzed in terms of fuzzy logic, they show that some complex concepts inevitably get assigned the wrong meanings. In the present paper we argue that many of the problems O&S discovered are due to difficulties that are intrinsic to fuzzy set theory, and that most of them disappear when fuzzy logic is replaced by supervaluation theory. However, even after this replacement one of O&S's central problems remains: the theory still predicts that the degree to which an object is an instance of, say, "stripped apple" must be less than or equal to both the degree to which it is an instance of "striped" and the degree to which it is an instance of "apple", but this constraint conflicts with and the degree to which it is an instance of "apple", but this constraint conflicts with O&S's experimental results. The second part of the paper explores ways of solving this and related problems. This leads us to suggest a number of distinctions and principles concerning how prototypicality and other mechanisms interact and which seem important for semantics generally. Prominent among these are (i) the distinction between on the one hand the logical and semantic properties of concepts and on the other the linguistic that between concepts for which the extension is determined by their prototype and concepts for which extension and prototypicality are independent.
Introduction Montague (1970a) presented a semantic treatment of adjectives which he credited to unpublished work done independently by Hans Kamp and Terence Parsons; that work, and similar independent work of Romane Clark, was subsequently published (Clark 1970, Kamp 1975, Parsons 1970). The central claim in that work was that adjective meanings should be analyzed as functions from properties to properties. Among adjective meanings, some might satisfy further constraints such as intersectivity or subsectivity, but no such constraint can be imposed on the class as a whole, the argument goes, because of the existence of adjectives like false, ostensible, alleged. The strategy of "generalizing to the worst case", followed by Montague in order to have a uniform assignment of semantic types to syntactic categories, called for giving all adjectives the type of functions from properties to properties. More restricted subclasses of adjectives, such as the subsective<F
Paths towards universal grammar: Studies in honor of Richard S. Kayne
- Guglielmo Cinque
Where have all the adjectives gone? The Hague: Mouton.Google Scholar
- Robert M W Dixon
On the relative order of adjectives
- Robert Hetzron
- Hansjakob Seiler
Hetzron, Robert. 1978. On the relative order of adjectives. In Hansjakob Seiler (ed.), Language
universals, 165-184. Tübingen: Narr.
The semantic classification of adjectives: A view from syntax. Studies in Chinese Linguistics 35. 1-30
- Guglielmo Cinque
Adjectives and nominalizations. The Hague: Mouton
- Zeno Vendler
Vendler, Zeno. 1968. Adjectives and nominalizations. The Hague: Mouton.
- Leonard Bloomfield
Bloomfield, Leonard. 1933. Language. Chicago, IL: Chicago University Press.
Prototype theory and compositionality
Little lexical learners: Quantitatively assessing the development of adjective ordering preferences
- Galia Bar-Sever
- Rachael Lee
- Gregory Scontras
- Lisa Pearl
- Anne B. Bertolini
- Maxwell J. Kaplan
Kotowski, Sven. 2016. . Berlin: Mouton de Gruyter.
- Sven Kotowski
Bache, Carl & Niels Davidsen-Nielsen. 1997. . Berlin: Mouton de Gruyter.
- Carl Bache
- Niels Davidsen-Nielsen
Feist, Jim. 2012. . Cambridge: Cambridge University Press.
- Jim Feist
Jackendoff, Ray. 1972. . Cambridge, MA: MIT Press.
- Ray Jackendoff
Vendler, Zeno. 1968. . The Hague: Mouton.
- Zeno Vendler
Bloomfield, Leonard. 1933. . Chicago, IL: Chicago University Press.
- Leonard Bloomfield
Cinque, Guglielmo. 2010. . Cambridge, MA: MIT Press.
- Guglielmo Cinque
Dixon, Robert M. W. 1982. The Hague: Mouton.
- Robert M. W. Dixon
An information-theoretic explanation of adjective ordering preferences CogSci
- Michael Hahn
- Judith Degen
- Noah D. Goodman
- Dan Jurafsky
- Richard Futrell