A Study on Metaphoricity of Duck across Word Classes with Wmatrix and BNCweb Combined — Implications for Learning of Polysemous Words

This paper combines Wmatrix with BNCweb to look at the grammar of linguistic metaphors formed by concrete nouns and their derivatives across word classes in order to shed light on learning of polysemous words. The methodology is to search different word classes of such a noun in BNCweb, use the concordances to construct corpus data and upload the data to Wmatrix for semantic domain tagging, choose the self-contained corpus in Wmatrix — British English 2006 — for a reference corpus to do the frequency comparison, then study the resultant key semantic tags as clues to identify the source domains of metaphor. The underlying idea of Cognitive Metaphor Theory, the Great Chain of Being 2 and the semantic makeup of a concept will help with the identification of a metaphor. The case study on duck shows the detailed process and feasibility of the research method. The research results reveal how different meanings of a concrete noun are related by metaphor or metonymy, which facilitates the learning of polysemous words.


Introduction
The study of metaphor in discourse is at full swing; however, the study at the lexical level is still necessary and useful, and is more feasible given the availability of extremely large corpora like British National Corpus (hereafter BNC). Since metaphor involves systematic semantic mapping, any corpus tool that only provides grammatical tagging is insufficient to identify the source and target domains in a metaphor. Wmatrix [1] can help 1  Turner for details about the concept. It categorizes different beings in the world into five different levels which in turn are put into a hierarchy, with human beings on the top and natural physical things at the bottom, and a being of a higher order having all the defining features of lower beings. with the identification by virtue of its semantic tagging function; however, it runs on self-uploaded files, which is why we need the data from BNC. This paper finds an effective interface between Wmatrix and the web edition of BNC-BNCweb [2]-in the exploration of possible linguistic metaphors by a concrete noun and its derivatives across word classes.

Studies on the Grammar of Metaphor
The research in this paper belongs to the field of the grammar of metaphor in terms of word classes. Studies on the grammar of metaphor can be found in the works of Brooke-Rose, Andrew Goatly, Lynne Cameron, Alice Deignan, Gerard Steen, and so on. The earliest researcher is B. Rose (quoted from A. Deignan [3]145-46) who studies the grammar of metaphor in literature. She finds some subtle semantic variations with different linguistic forms, e.g. the different effects of noun metaphors with or without the definite article. L. Cameron [4]89 researches the question of metaphor and part of speech in her classroom data, finding that verbs account for nearly half of the metaphors in her data, while adjectives and adverbs together account for less than five per cent of the total, hence the tendency of English to put metaphoricity on the verb. She also finds that "metaphors with single nouns as vehicle terms, often the form used to exemplify metaphor theory, account for only 5 percent of the total." But in my opinion, she may have ignored the fact that English is a noun-dominated language. Even noun metaphors only account for 5 percent of the total number of metaphors in her data, the number of noun metaphors in the whole language can still be larger than that of verb metaphors, and still be the most prototypical in the common sense of the general public. And G. Steen [5]194-207 gets different results from Cameron's in terms of metaphor and word classes, which shows that different research contents may lead to different results sometimes. The more evidences we get, the closer we are to the realities. A. Goatly [3] 146-47 makes some interesting observations about metaphor and word class, proposing that noun metaphors are inherently more marked than verb metaphors. His reason is that the primary function of a noun is to refer, and therefore an unconventional referent is very noticeable. According to him, verbs used metaphorically may also be marked, but that depends on our perceiving an unusual context rather than an unusual referent. To this, I disagree. Just because of the context that we feel verb metaphors more marked in that they are more vivid and economical, but need less processing effort to be understood. This conforms to the Relevance Theory by Sperber & Wilson [6]. A. Deignan [4]147-67 investigates the carrying-over relation between different parts of speech of a word, like the semantic relation between the noun dog and the verb dog in metaphor; different semantic prosodies between different inflectional forms, like the positive prosody from the singular noun rock and the negative prosody from the plural noun rocks; different linguistic collocations in the literal use and metaphoric use of a word, like the transitive verb build to be used literally, while the intransitive verb build followed by the particle for to be metaphoric. Deignan is the only scholar who has dedicated a book to the grammar of metaphor. She proposes that many linguistic metaphors seem to be restricted grammatically either at the macro-level of word class, or at the more concrete level of syntactic patterns. Under the inspiration of these scholars' research, esp. Deignan's, this paper attempts to study the grammar of linguistic metaphors by a concrete noun and its derivatives across word classes in order to find how different meanings of a word are related. The relation may be based on metaphor, metonymy, metaphor-based metonymy, metonymy-based metaphor, and so on. Since the focus in this paper is on metaphor, the mention of metonymy is not much detailed. The target words come from the 2-5 levels in the Great Chain of Being [7]. The findings are intended to shed light on the learning of polysemous words.
There are semantically content words and empty words [4]49. The former include nouns, verbs, adjectives and adverbs, and the latter include pronouns, conjunctions, prepositions, auxiliaries and articles. Content words usually constitute the most prototypical metaphors, while empty words, especially pronouns, articles and conjunctions are seldom metaphoric. It is well known that content words make the majority of the vocabulary of any language, and they can be changed into other word classes by conversion or derivation. It is quite common for a noun to be turned into a verb, adj. and adv by (zero) fixation, and normally there are semantic carry-over between different meanings of the word within a certain word class or across word classes. And that is the scope of study in this paper.
Due to the limit of space, the paper will take an animal word, duck, from the second level of the Great Chain of Being as example to look at its metaphoricity as noun and across word classes. In addition, this paper mainly studies the situation when such a word acts as the source domain in metaphor, and ignores personifications where it acts as the target domain.

Studies on Metaphor Identification
Collins English Dictionary (21 st Century Edition) [8]978 defines metaphor as a figure of speech in which a word or phrase is applied to an object or action that it does not literally denote in order to imply a resemblance. It is widely accepted that metaphor is no longer merely regarded as a figure of speech; instead it is found pervasive in human thought and behavior [9]3, and its nature is to experience or understand one thing in terms of another [9]5, which is the fundamental idea behind the Cognitive Metaphor Theory (hereafter CMT). The thing to be experienced and understood is the target domain, while the thing used to understand and experience the target domain is the source domain. Although there are other metaphor theories, like the mental space blending theory by G. Fauconnier [10], CMT is by far the most influential paradigm in metaphor studies.
Metaphor identification is the most critical step for any metaphor studies. There are some well-known metaphor identification procedures, like MIP proposed by the Pragglejaz Group [11], and MIPVU by Steen et al [5] for corpus-based metaphor studies. However, because this paper exploits a semantically tagging corpus tool, a different way of metaphor identification is suggested instead. This paper follows the design idea of Wmatrix [1][12] [13] [14] that the semantic domains tagged by the tool are roughly equivalent to the source or target domain in metaphor. Rather than marking up every word as metaphoric or non-metaphoric, this paper only looks at those semantic tags that are related to the search word and are likely to be the source domain (in this paper, only source domains are targeted at); as a result, unrelated words, even metaphoric, will be dismissed as irrelevant.
Normally three aspects are considered when searching for the source and target domains. The first is disparateness between the two conceptual domains. This is determined by the nature of metaphor to experience and understand one thing in terms of another. If two things are too alike, they will constitute an ordinary comparison, or a model relation [15]24. The second standard is semantic incongruity resultant from the disparateness between the source and the target domains. And the third is similarity perceived between the two disparate and seemingly incongruous concepts or propositions, without which metaphor will become willful.
There are two things to clarify. One is about the nature of the source domain. It is well-known that CMT assumes the source domain concept to be dominantly concrete, while the target domain concept to be mainly abstract. Most of the time this is the case, reflecting the cognitive economy of human thinking; but theoretically and practically there are metaphors where the source domain terms can be abstract 364 A Study on Metaphoricity of Duck across Word Classes with Wmatrix and BNCweb Combined -Implications for Learning of Polysemous Words or difficult in order to de-automatize too familiar concepts, especially in literary language. Another thing is that similes are treated as metaphors and included in the scope of study here. In addition, metaphoric idioms are also included.
The process of metaphor identification is a combination of both top-down and bottom-up procedures. On the one hand, from the schematic makeup of a concept as the source domain we can guess what can be mapped onto a target domain, and this is the top-down part. According to Lakoff & Turner [7]63-64, the schematic structure of a concept includes the slots, attributes, relations and knowledge. Take duck as noun for example. Its slots can be the bird itself and its body parts; its attributes include its appearance, habits, behavior, etc.; relations may include its instinctual habit of swimming in water; knowledge is the causal relation(s), or the logic, e.g. a duck loving water suggesting intense fondness. Relations and knowledge can form analogy-based metaphors: the knowledge that ducks are born to swim can be mapped to convey easy excellence of somebody in doing something. All the above information will guide us in telling what is relevant and irrelevant in a metaphor triggered by the noun duck as the vehicle. On the other hand, what may happen may not really happen. In other words, there may be accidental gaps. This is the very meaning of corpus-driven studies on naturally occurring data. We check the concordances from BNC about duck to look for what about duck is really mapped to form metaphors. This data-driven part constitutes the bottom-up part of the procedure.
The identification should be guided by rough pre-knowledge about what can be the candidate source domain(s) or the target domain(s). Knowing one part will lead to an oriented search for the other. Semino [14] and Koller [13] use Wmatrix to search for candidate source domains based on prior knowledge of candidate target domains. Stefanowitsch [17]63ff looks for possible target domains based on possible source domains. As to metaphor use at the word level, some words tend to be the source domains, like war; others tend to be the target domains, like life; and some can be both, like cancer. In case a word belongs to the source domain, the task is to look for candidate target domains, and the situations for the other two cases are inferable. Since the research in this paper only concerns metaphor and personification is not included, so only the source domains will be looked for.
The identification would be easier if we refer to the Great Chain of Being [7]160ff which has five levels: humans, animals, plants, complex objects and natural physical things, with each level having its own nature, i.e. its defining features and behavior. Higher-order attributes and behaviors, like thought and character are the defining features of humans; instinctual attributes and behavior are the defining features of animals; biological attributes and behavior are the defining features of plants; structural attributes and functional behavior are the defining features of complex objects; natural physical attributes and physical behavior are the defining features of natural physical things. When concepts in levels 2-5 are mapped onto humans, metaphors occur; when concepts concerning humans are mapped onto concepts of all the other levels, personifications are produced. Mappings between levels 2-5 are seldom, if any. According to the theory, only the defining features of one level can be mapped onto another in metaphor. For example, when humans are understood in terms of an animal, it is the animal's instinctual attributes and behavior that are mapped onto human beings, so anything related to human beings will belong to the target domain, while anything related to the animal will form the source domain. So lexical items referring to human character, like confident, considerate, wise, etc., can only belong to the target domain human beings, rather than to the source domain animal. This helps us put related lexical items into proper source or target domains, and exclude impossible ones.

Methodology
Before the appearance of a semantic tagging corpus tool, the most prevailing tagging was grammatical tagging, represented by POS tagging, i.e., part of speech tagging. Then what a corpus researcher generally did was key in the syntax in the query box of the corpus, get the concordances, look at the hits to find evidences for their research. And so was the corpus research into metaphor. The researchers would look at the concordances to tell the metaphoricity of the search word, group those that are metaphoric, and then elicit the cognitive metaphor(s) [3]160, [16]191ff.
However, since Dr. Paul Rayson [1] developed and ran Wmatrix under the UCREL infrastructure of Lancaster University, things have somewhat changed. Given that metaphor involves semantic mappings, grammatical tagging can not be of much help in metaphor identification. Wmatrix was originally developed for thematic studies in discourse analysis; however, it was creatively borrowed by some scholars in Lancaster University for metaphor research. Hardie et al [12], Koller et al [13] and Semino et al [14] try some methods to pin down candidate source domains of metaphor and suggest some calculations, including tag-lemma ratio and lemma-token ratio, to look at the density and variation of metaphors in a corpus. Their studies have proved that Wmatrix can effectively extend the researchers' instincts in finding more relevant tokens that might prove their pre-assumptions, or lead them to new findings. Screencast 1. The folder page before the keyness analysis list pops out This paper combines BNCweb with Wmatrix, with the former for data, and the latter for semantic tagging. The method is to key in the search word in the query block of BNCweb, click on "Start query", get the concordances, use them to build a word document, delete filenames and numberings to reduce unreadable words and useless information. Since all of the hits in a BNC concordance are cut on the sentence boundary, the new file is made up of complete sentences, so insufficiency of context can be ignored to a certain degree. Re-save the word document in the format of a plain text, upload it to Wmatrix by clicking on "tag wizard" button on the top of its homepage, allow some time, and POS tagging and semantic tagging will be finished automatically. And then a folder will be produced and added to the file "My folders". Click on "My folders", choose and click on the target file, and the following page will appear. Under "keyness analysis" area, go to "Key concepts compared to:", click on the dropdown arrow, choose "British English 2006" in the menu, click on "Go", and a keyness analysis list in terms of semantic tagging with the target file and British English 2006 compared will pop out. Refer to Screencast 1 and 2 to get a glimpse into how this works.
The main task of the study is to look into the keyness analysis list, esp. the items whose loglikelihood value is above 6.63, which means the significance of the statistical item is above 99%. According to the procedures about how to identify metaphors, open the concordances of those items that may serve as the source domain in metaphor, list them, and try to find the regularities about how different meanings of a word are related to each other. Although metaphor is the focus of study, sometimes metonymy is unavoidable. And we all know that metonymy involves substitution of one thing for another within a single conceptual domain based on proximity. When the duck is used to refer to the duck egg, it makes a metonymy. Sometimes, there are more complicated cases: metaphor-based metonymy and metonymy-based metaphor.

Case Study on the Word Duck
Due to the limit of space, this paper only picks the animal word duck from the second level of the Great Chain of Being for illustration. First, key in *duck* in simple query mode in BNCweb, click on Start query, choose Frequency breakdown in the dropdown menu with New Query at the top, and 99 types (2982 tokens) will be obtained. Among them, the noun (duck, ducks), the verb (ducks, ducked, ducking) and the adjective (ducky) forms of duck can be found, and no adverbs are returned. The compounds like duckbilled, duck-toed, duck-like are all dismissed except various forms of lame duck which is highly lexicalized, with forms of lameducked, lameducking, lame-duck and lame duck. Obviously, lame has become a frequent collocation of duck. The type duckie is a noun, a diminutive term, and it is also dismissed.

Duck as Noun
Key in {duck/N} (N for noun) in simple query mode in BNCweb, and 1,893 hits in 585 different texts will be returned. Save the concordances in the word format, deleting the columns containing the numberings and the filenames respectively. Merge all the words, and a file with 40,683 words, punctuation marks included, will be obtained. Re-save the content in the text format. Browse the resultant document to delete unnecessary blank spaces, paragraph markers and other symbols that will affect the accuracy of tagging (cf. http://ucrel.lancs.ac.uk/wmatrix3.html). Then upload the file to Wmatrix by pressing the Tag Wizard button on top of the homepage of Wmatrix, where POS tagging, USAS (UCREL semantic annotation system) tagging and frequency counting will be conducted automatically. The results will be saved in the folder entitled duck_N. Open the folder, and in the keyness analysis area, go to the third search box Key concepts compared to, choose the folder British English 2006 (BE06) in the dropdown menu as the reference corpus, click on Go, and a list of key semantic domains with log-likelihood value (hereafter LL) in decreasing order will appear. Take all the items whose LL value is equal to or higher than 6.63: Since duck is a water bird, if it functions as the vehicle of a metaphor, it must be the defining features and behavior of the bird that are mapped onto the target domain. To be specific, they include the slots, features, relations and knowledge about a duck. Go through the above list, it is not difficult to guess that there should be words under Items 1, 7, 11 and 14 that belong to the source domain in metaphor. Since a duck is a bird, it must be under Item 1. Open the concordance of Item 1, one can find that it is dominated by the word duck. However, to one's surprise, duck alone seldom makes metaphor except as an endearing term, like "my love", which must be grounded on the assumption that a duck is something lovely. Item 7 is chosen because swimming is an instinctual feature or action of a duck. But in the concordance, only one metaphor is found: 1) He writes as a duck swims. Item 11 is very conspicuous, because common sense tells us that ducks are closely related to water, and water is a liquid. Open the concordance, and the assumption can be proved: the page is occupied conspicuously by water: Among all the 51 occurrences, 40 are water. Read each hit carefully, such analogy-based metaphors from different collocations of duck with water are found: Like a duck out of water: metaphoric meaning "very uncomfortable" Take to something like a duck to water: metaphoric meaning "be good at or fond of something very much" Water off a duck's back/pour water on a duck's back and watch it run off: metaphoric meaning "be in vain; without effect". Item 14 is chosen because the body parts of a duck must be included under the item "Anatomy and physiology". The neck, the feet, the toe, etc., may be borrowed for description of a person. They may also form such compounds as duckbilled, duck-toed, duck-webbed, etc., or similes like "his neck is like a duck's".
As to other items that are not chosen, there are ample reasons. For example, Item 2, "color and color patterns": even the color of a duck is metaphoric, it must be very rare, because the color of a duck is not its defining feature. Item 3, "Time: ending", is obviously irrelevant to duck. Item 4, "Discourse bin", since it only includes discourse markers, hence is irrelevant, too. Item 5, "Unlikely", irrelevant; Item 6, "Food", should also be irrelevant in that it at most contains the word duck referring to the meat of this bird, and that does not make a metaphor, but a metonymy. The non-metaphoricity of other items can be inferred according to our knowledge about the bird duck in combination with our prior study on metaphor theories.
To sum up, most metaphors formed by the noun duck come from the relation or knowledge concerning the habitual behavior or living environment of the bird. They are most analogy-based metaphors. And the results also prove L. Cameron's observation: metaphors from single nouns are quite few. Now let's look at the different meanings of duck as noun and the possible relations between them. Look up duck in Collins English Dictionary [8], and the meanings are as follows: 1 any of various small aquatic birds of the family Anatidae, typically having short legs, webbed feet, and a broad blunt bill: order anseriformes. 2 the flesh of this bird, used as food. 3 the female of such a bird, as opposed to the male (drake). 4 any other bird of the family Anatidae, including geese, and swans. 5 Also: ducks. Brit. informal. dear or darling: used as a term of endearment or of general address. see also ducky. 6. A person, esp. one regarded as odd or endearing. 7.
Cricket. a score of nothing by a batsman. 8. like water off a duck's back. Informal. without effect. 9 take to something like a duck to water. Informal. to become adept at or attracted to something very quickly.
Here the first meaning of duck is the most basic; the second meaning is related to it based on the metonymy that DUCK MEAT IS DUCK; the third and fourth meanings are narrowed and broadened meanings respectively, and no metaphors are formed; the fifth and sixth meanings are metaphor-based metonymies, coming from the metaphor that AN ENDEARED (ODD) PERSON IS A DUCK; the seventh meaning is a metaphor-based metonymy in which the metaphor is ZERO IS SHAPED LIKE A DUCK EGG, and the metonymy is BREAKING ONE'S DUCK IS BREAKING ZERO, meaning scoring; the eighth and ninth meanings are analogy-based idioms whose meanings can be inferred by referring to the cognitive knowledge about what happens in the real situations about a duck.

Duck as Verb
Key in {duck/V} (V for verb) in the query box of Wmatrix, and 585 hits in 320 different texts will be returned, showing duck as verb is not as highly used as its noun. Build a new corpus duck_V with all the concordances, upload it to Wmatrix, compare its frequency list with that of British English 2006, and the following results will be returned:

Screencast 2: Key semantic domains in duck_V in comparison with British English 2006
Since duck now is a verb, signifying an action, only those semantic tags of action can subsume it. Browse the above list, and only Items 1, 6, 8, 20 and 25 are such tags. To decide which one or ones can serve as the source domain, we have to refer to the concept duck again. Here, it must be the typical behavior or action of the bird duck that is relevant. Before making the decision, I refer to the online edition of Collins English Dictionary [8] again, and find the following meanings of duck as v: 1. to move (the head or body) quickly downwards or away, esp. so as to escape observation or evade a blow 2. to submerge or plunge suddenly and often briefly under water 3. (when intr, often foll by out) informal to dodge or escape (a person, duty, etc.) 4. (intransitive)bridge to play a low card when possessing a higher one rather than try to win a trick Among the four meaning, the second one must be the most basic, because it signifies the typical instinctual action of a duck. The first one is a little bit away from the second one: it signifies a similar physical action, but takes the meaning "to evade something physical, like a blow". The extension is based on the intention. The third one is still further away: it carries over the "evade" part of the first meaning and signifies a completely abstract action, and its objects can be such words as duty, question, responsibility, etc, which are completely unrelated to the action of a duck. The fourth entry is also far from the second meaning: it carries over the "taking a low posture" part of the duck action and refers to an attitude in playing the bridge.
Given the above results, it is very clear which items are relevant: only Item 1 and Item 6. Open their concordances, I find that the concordance of Item 1 is dominated by various verbal inflections of duck. I even find one example in which duck is used literally 3 . It is quite easy to find that the target domains in the above list are becoming more and more abstract. The extension of meaning between the meanings is quite clear. And the most abstract is the ones when duck is followed by completely non-physical objects, like issue, duty, quarrel, etc. And that is the commonest type of metaphor: to use something concrete to metaphorize something abstract.

Duck as Adjective
Search {ducky/A} in Wmatrix, and 28 hits in 13 different texts will be returned. Among them, 20 are an addressing term, four are capitalized either as the name of a person, or of a pony. Only three are used as an ordinary adjective, among which two are found in the expression 'chucky duchy lover', where ducky means 'lovely', and one in its literal sense, as in 3) … watching the ducks do ducky things in the water.
The situation with ducky in (3) is quite similar to that of ducked in (2) where the logical subject of both is duck the bird. They can be regarded as literal, or metaphoric if the attributes or actions of a duck are extracted and generalized.
The frequent occurrences of ducky are dominantly a noun, a highly conventionalized metaphor, and equivalent to duck/ducks when used as an address term. So duck, ducks and ducky can all convey the concept of a 'lovely' human metaphorically. Of course, they can also be called metaphor-based metonymies because their major function is to refer, or address.

A glimpse into the Collocation of Lame with Duck
The frequency breakdown for *duck* gives such compound forms as: duck-like, duck-billed, duck-egg, duck-quack, duck-toed, diving-duck, duck-egg-blue, duck-shaped, duck-jumping, duck-talk, duckboard, duck-pond, etc., which fall into two word classes: nouns and adjectives. It seems that almost all adjectives, including duck-shaped, duck-billed, duck-toed, duck-like, duck-egg-blue, etc. are metaphoric, describing the properties of something else, while most nouns tend to be literal.
Among all the compounds, lame is found to be a frequent collocation of duck. Search lame*duck* in BNCweb, we can get 17 hits in 8 different texts in the form of four types: lame-duck (8 hits), lameduck (2 hits), lameducked (3 hits), lameducking (4 hits). Besides, the search a lame duck gets 13 hits; and lame ducks, 10 hits; lamest duck, 1 hit. All of them are metaphoric. Almost all the noun forms refer to people/organization/project... that are losing (have already lost) their current power, vigor, benefit, etc., like a President (or his administration) that has failed in the election into the next term. There is one exception in which a lame duck actually refers to a lame person who limps out of the room. Of course, that is an example of direct mapping between a lame duck and a lame person. However, all of the verb forms mean to sympathize or comport a person who has become a lame duck, as in 3) I am the one who needs lameducking.
So here, the meaning of the verb lameduck is obtained not from metaphor but from a metonymy THE RESULT IS THE REASON. To be specific, since a lame duck is a person who needs sympathy, to lameduck somebody is to show sympathy to the lameducked person because of his disadvantageous position. The situation of the verb lameduck is quite different from that of the verb duck where the meanings of the verb duck all come from the typical action of the bird duck based on metaphor.

About Other Scholars' Studies
The research results in this paper shed some light on prior studies. If we compare the results from duck with Goatly's [3] study, we can come to a different conclusion: nouns used in metaphor seem to be less marked than verbs, since the verbal metaphors in the case of duck are more eye-catching, more vivid and easier to be understood.
Cameron's studies can also be justified to a certain degree: that is, English tends to put metaphoricity more on verbs. This can be seen from the frequency of the verbal duck metaphors. But further generalization must be based on much more evidences.
As to Deignan's [3] studies, the study of duck across its word classes shows that the nominal metaphoric meanings and verbal metaphoric meanings come from different parts of the concept duck: the former are from duck's features, relations and logic, which form descriptive and analogous metaphors; while the latter mainly from its behavior, describing the actions of human beings. This further proves that metaphoric mappings are more of conceptual than linguistic.
And the study also proves Deignan's above-mentioned assumption that many linguistic metaphors seem to be restricted grammatically either at the macro-level of word class, or at the more concrete level of syntactic patterns. Duck as a verb almost always makes metaphor; however, as a noun, it often makes metaphor in some fixed collocations, or syntactic patterns.
The results also reveal that although the noun duck alone seldom makes metaphor, it often makes analogy-based metaphors by collocating with other things essential in its existence, like water.

About Duck and Lameduck
In 4.1 -4.3 the relations of different meanings of the noun duck, the verb duck and the adjective ducky have been examined, showing that all meanings of the word duck and its derivatives are related: Some are evolved around its basic meanings; others are evolved from its extended meanings. The defining features and behavior of the bird duck determine the extension relations between these meanings. And the extensions are realized mainly by metaphor or metonymy. And the most basic meanings of the verb duck and the adjective ducky come directly from the defining features and behavior of the bird duck.
The finding about lameduck as verb is a little bit different from that of duck as verb. The meaning of duck as verb comes from the bird's typical action, so the extension of meaning is based on metaphor; however, the meaning of lameduck as verb comes from how a lame duck is treated because of the bird's lameness; therefore, the extension of meaning is based on metonymy.
It is important to find regularities and rules governing language phenomena. It is even more important and useful to mind the differences.

Implications for Learning of Polysemous Words
The nouns from the five levels of the Great Chain of Being make a bulk part of the vocabulary of English. They are polysemous as noun, and many of the nouns can be converted or changed into verbs/adjectives/adverbs, and most of these derivatives are polysemous, too. The most important is, there is obviously connection between the meanings of the word across different word classes. There are many polysemous words like duck in English. Since they have many meanings, it is not easy for language learners, esp. beginners to memorize them one by one. If we can find out how different meanings of such a word, within a word class and across word classes, are related by virtue of metaphor/metonymy, the learning process will be easier.
Of course, different words may have different ways of meaning extension. For example, when the word fish is converted into a verb, it is often used in such quite fixed patterns as go fishing and fish for, in which fish the animal becomes the target of catching. This is because fish as a creature is a kind of important sources of food for common people, hence the extension of meaning4. That is to say, besides the defining features and behavior of the referent in the real world, the ways human beings influence it also exert great impact on the extension of its meaning. And this is more obvious with a word bottle from the fourth level of the Great Chain of Being. When it is converted into a verb, its functional behavior of holding something is carried over, and to bottle wine can only mean to pour wine into a bottle for keeping.
To sum up, the Great Chain of Being is helpful in guiding us in how to understand and remember the extended meanings of a noun and its derivatives. The defining features and behavior of the correspondent referent of a word, and possible human influence on it can give us clues as to the possible extended noun meanings, derived verb meanings, adjective and adverb meanings.
But one thing must not be forgotten. To any rule, there may be exceptions. The context of language use must be borne in mind. For example, to bottle wine is easy to be understood, but to bottle a person must mean differently: how big must a bottle be to hold a person? And here, its meaning may be changed into "to attack a person with a bottle" [18]. The daily container bottle now becomes a weapon in a different context.
Rules and regularities can save us a lot of efforts in learning polysemous words, but remember the variable-context-all the time.