A Functional Approach to the Choice between Descriptive, Prescriptive and Proscriptive Lexicography

In lexicography the concepts of prescription and description have been employed for a long time without there ever being a clear definition of the terms prescription/prescriptive and description/descriptive. This article gives a brief historical account of some of the early uses of these approaches in linguistics and lexicography and argues that, although they have primarily been interpreted as linguistic terms, there is a need for a separate and clearly defined lexicographic application. Contrary to description and prescription, the concept of proscription does not have a linguistic tradition but it has primarily been introduced in the field of lexicography. Different types of prescription, description and proscription are discussed with specific reference to their potential use in dictionaries with text reception and text production as functions. Preferred approaches for the different functions are indicated. It is shown how an optimal use of a prescriptive, descriptive or proscriptive approach could be impeded by a polyfunctional dictionary. Consequently arguments are given in favour of monofunctional dictionaries.


Introduction
The use of the terms prescription and description became popular in American linguistic studies that followed the era of Bloomfield's behaviouristic approach and American structuralism. Within this domain, these terms had a strong linguistic application. Prior to this use, the notion of prescribing, describing and even recommending had been used in scientific discussions, albeit primarily directed at textbooks. However, it is important to note that the notions of prescribing and describing did not initially result in establishing prescription and description as terms in fields outside linguistics. Although the notions of prescribing and describing were prevalent, the use of these words was of a nonterminological and general language nature. In a language like German, general language words like beschreiben ("describe") and vorschreiben ("prescribe") were rather used to carry these meanings. Unfortunately English does not have equivalents for the German beschreiben and vorschreiben other than describing and prescribing to help with the distinction between the general, i.e. the nonterminological use, and the specialised use, i.e. as terms especially in the field of linguistics. In this article, various aspects of the processes of describing and prescribing will be discussed, with the eventual main focus on the use of these words, in a non-terminological sense, in the field of lexicography. However, it will also be shown that lexicography is not the only field benefiting from an unambiguous use of these words, clearly distinguishing their use from the use of the terms describing/description and prescribing/prescription as primarily found in linguistics.
One of the real problems in the use of the contrasting pair descriptive/prescriptive is the fact that the distinction has primarily been in linguistics where it had a specific use and in the course of time acquired terminological status. Although the notions of describing and prescribing, used in both lexicography and some other scientific fields in a non-terminological sense, were integral components of earlier lexicographic work as well as scientific discussions regarding language criticism and textbooks, the terms description and prescription had not been integrated into the terminological collection of these fields. In the early phases of the development of lexicographic theory, prior to the emergence of lexicography as an independent discipline, many linguistic concepts, including the dichotomy descriptive/prescriptive, were imposed upon dictionaries and used and interpreted from a linguistic perspective. Linguists often failed to realise that a dictionary is not an instrument that should respond to linguistic rules and criteria but is a practical tool that should be compiled in accordance with the specific needs and reference skills of a clearly identified target user.
Where linguists might have found it unproblematic to characterise their linguistic endeavours as descriptive or prescriptive, it was not so uncomplicated to describe, for instance a general language dictionary as either descriptive or prescriptive, especially not by the user of such a dictionary. Although it might have been possible to detect certain prescribing entries in a dictionary, the dictionary as such did not necessarily follow a prescriptive approach. Both description and prescription are processes. Dictionaries mostly displayed single occurrences of, for instance, prescription, because the approach had been directed at single phenomena and not at the entire dictionary. It would have been equally difficult to classify a given dictionary as descriptive, because such a classification depends on the way a lexicographer decides to present data to ensure that a function identified for the specific dictionary can be achieved. This would once again be the application of a process the user can hardly identify by merely looking at isolated instances. What one user may regard as prescriptive, another user may regard as descriptive. To illustrate this, one can look at the "grammar study note" (intended to provide information about areas of grammar tested in many examinations) attached to the article of the lemma sign homework in the Longman Exams Dictionary: "Homework is an uncountable noun and has no plural form. Use a singular verb after it …" Where one user may interpret this note as descriptive because it objectively states something about the system of the language, another user may regard it as prescriptive, feeling that the lexicographer is telling the user how this word should be used.
In this article, a vital point of departure is the conviction that lexicography is an independent discipline. As a result, lexicographic practice needs to be dominated by lexicographic theory and not linguistic theory. Attention will therefore be given to ways in which the concepts of prescription and description prevail in lexicographic practice and whether they need to be introduced as fully-fledged lexicographic terms. The use of a complementing notion, i.e. proscription, and its possible terminological introduction, will also be discussed.
Before coming to the use of the concepts of prescription and description in lexicography, it is necessary to take a look at the historical development and some aspects regarding the use of these terms in both linguistics and in a more general scientific discussion.

Description and prescription as an old tradition
Some linguists assert that all dictionaries published prior to 1820 had been prescriptive (cf. e.g. Greimas and Courtes 1979). However, this is done without a mutual understanding of either prescription or description. When one uses a definition comparable to that given for the German verbs beschreiben ("describe") and vorschreiben ("prescribe") the assertion is false. Statements about language, especially in grammars and dictionaries, have since their early beginnings, been primarily descriptive. Very seldom, only in a few instances, one finds items that are primarily prescribing. Nevertheless, the focus has been on these few exceptional cases, although all types of information tools usually try to reflect actual language use, as is noticeable in the following definition from Wikipedia (http://en.wikipedia.org/wiki/Description_(linguistics) (24.2.2010)): In the study of language, description, or descriptive linguistics, is the work of objectively analyzing and describing how language is spoken (or how it was spoken in the past) by a group of people in a speech community.
Here one should replace the word "spoken" by the word "written", because written texts, especially as reflection of the competence of the compilers of dictionaries and grammars, are authoritative. One can see in definitions like this one that the starting point is, without reservation, the observation and presentation of the observer. The purpose of this observation is not stated here. In other definitions, one can observe that such descriptions can offer help when the user of a textbook has reception problems. Others emphasise the documentary value of the description of a given synchronic section. The issue is not how someone has to speak or write in future. Yet, this remains the issue when one looks at the prevailing definitions of prescription (http://en.wikipedia.org/ wiki/Prescription_(linguistics) (24.2.2010)): In linguistics, prescription can refer both to the codification and the enforcement of rules governing how a language ought to be used.
A reflection of the language is not at issue here; the sole issue is future language production. However, one can hold the point of view that one should speak and write in the same way as it has been done up till now or as it is most frequently done. Whether such a demand can be regarded as descriptive is questionable, because it actually perpetuates the prescription of the prevailing description. This applies, today as in the past, to different approaches and different types of needs that grammarians and lexicographers try to satisfy with their respective books. The following discussion presents a few examples taken from German grammar history, which is typical of European history. Examples of other countries and languages can be found in Haßler (2009), who gives lengthy quotations from French, Italian, Spanish and German literature of the 17th and 18th centuries. Examples from lexicographic textbooks are deliberately not used, because the purpose is to present a general view of the status of description and prescription that can then be applied to lexicography. The first German grammars were empirical and synchronic, taking the then current German language as basis for their language description. Regarding text production, the prevailing language use was in most instances also the example for new texts. In the earliest grammars, this was the only approach, but in the next centuries, occurred only in exceptional cases. An analysis of texts can then also lead to a criticism of certain sections of these texts. The best example comes from the most comprehensive German grammar (in five volumes with more than 3 247 pages) of Bauer (1827Bauer ( -1833. Bauer understood his grammar as one that directs itself at observations. These observations are actually a presentation of his own language competence, and many of his own constructed examples are used to illustrate his rules of language production. His many citations from contemporary literature merely contribute to his criticism of individual language use. Calling them "language adversities" and "sin indices", he presented them in footnotes together with lengthy comments, e.g. a four-page footnote (Bauer 1833: 333-337) with 96 citations from 58 sources, especially from Goethe and Schiller (the most famous authors of that time). Passages like these can be regarded as contributions to language criticism or simply as critical text analysis. Whether criticism can be regarded as being descriptive is a central question for the theme of this article. This question is affirmed, because it is believed that there can be no observations without judgements, including critical judgements. Bauer understood his rules as instructions for written language use: rules are deduced from the actual use. But not every form of usage has been included in the main section, with language use that can be criticised given only in the footnotes. This relation between rules and language use has since then been a basis for discussion.
A good example of arguments for and against pure description, more specifically dictatorial prescription, are known from the debate between the professors Richey and Fabricius, in papers from the year 1723. It should be noted that they did not discuss the issue as linguists. In addition, they did not provide any examples but argued quite abstractly as in philosophical discussions. Richey was an educated person and a poet. He studied Theology, Natural Science, Mathematics and History and was professor of Greek and History at a gymnasium in Hamburg. Fabricius studied Medicine and Theology. He also worked as librarian, became a doctor of Medicine and later held the chair of Theology and Ethics at the University of Kiel. Richey, whose paper has the title "Usus Tyrannus precario imperans, oder Vernunft-mäßige Sätze von der Gewalt des Gebrauches in den Sprachen / insonderheit in der Teutschen, ingleichen ob und wie weit man denselben durch Regeln der Sprach-Kunst Einhalt thun könne", argues from the assumption that no language in the world is inherent to people. From this follows that not all people learn to know a good "Mund-Ahrt" (dialect/variety) during their youth, that many do not hear the true, actual and stable use and therefore are compelled to uncertain, faulty or false language use. Richey asks whether this would not lead to a situation where the inane masses with false ways of expressing themselves gain the upper hand. Instead of following the "usus tyrannus", Richey proposes that one should carefully undertake certain changes that each reasonable person would have to employ.
It is clear that Richey wants to change or preserve the prevailing language use in order to enable reasonable discussions. Fabricius replies to these ideas in his "Usus imperium sine exceptione assertum oder die unumschränkte Gewalt des Gebrauchs / gleichwie in andern Sprachen / also auch in der Teutschen, behauptet wider die weit aussehenden Neuerungen der SprachLehrenden Malcontenten", working with the following thesis: There exists no language that has been made according to grammars or that has been improved in a uniform way. In the course of time, language has much rather been arranged by nature and all types of coincidences. Such a natural and not regulated development of a language, that Richey certainly regards as possible but also as not tenable, forms the basis for the following thesis of Fabricius and for the subsequent demands made on future linguistic work: 1. Each language is constantly subjected to changes. 2. Linguists should not restrict the deviations and changes. 3. The description of a language should rather be changed by the changed language use.
Fabricius admits that his demands on linguists imply major efforts, because the extent of language use can hardly be grasped. Fabricius regards the demand for comprehensive investigations of actual language use as the only solution. In addition the language of educated as well as uneducated people from all components of the population should be collected. When a word, construction or expression is frequently used in written and spoken language, Fabricius regards himself as not qualified to reject the form or simply condemn it. It is clear that Fabricius adheres to the famous American thesis of "leave your language alone", and therefore one should endeavour to describe real language use. It is not quite clear why he desired this. He apparently rejected every form of language criticism and as a linguist he did not want to preserve or change the structure of the language. Fabricius writes in a fluent, elegant and especially faultless language. One can therefore assume that Fabricius would have discussed obvious mistakes. But it had not been his topic. When anyone does not write in German Sprache ("language") but rather Spache (without the r) every teacher, including Professor Fabricius, would regard it as a mistake that needs to be corrected. Yet, the border between correct and incorrect remains unclear. When someone writes or speaks a foreign language, the mother tongue sometimes has a definite influence that leads to mistakes. Textbooks are used to avoid such mistakes and also to be used as materials for language teaching. La Forêt (1760) is an example in this regard. This book promises in the preface to help those Danes who want to avoid the typical mistakes made when they write in German. It is not a dictionary but rather a special German grammar for Danes. The third chapter treats "corrupt words and partially bad constructions", with comments in alphabetical order. It is a monofunctional information tool, as stated in the subtitle of the work: "Lingva Germanica In Ore Danico. Das ist: Unvorgreiffliche Anweisung, Wie Ein Teutsch redender Däne Unterschiedliche Danismos in einer Teutschen Rede zu vermeiden habe: Wobey Von dem Genere der Teutschen Substantivorum gründlich und ausführlich, Von Der Declination und Conjugation zulänglich, von Der Wort-Fügung beyläuffig gehandelt wird." The brief alphabetical part of the book can be regarded as a small dictionary of false friends, as shown by the following entry: At gaae i Caution for nogen, heisset nicht auf Teutsch, in Caution für einen gehen; sondern Bürgschaft leisten; Bürge seyn; gut für einen sagen. (At gaae i Caution for nogen is not in German "to be cautious for someone" but "to guarantee something, to be guarantor, to vouch for someone".) Almost everybody would then have said, and will still say today, that "in Caution für einen gehen" does occur. And almost everybody would have said and will still say today: This is a mistake that can be explained as occurring under the influence of Danish. But when it comes to stopping and reversing a change that has already entered the language, the question is no longer whether it is right or wrong but rather whether one is trying to prevent the development of a language into something new, as was the case in the dictionary of Changuion (1844). In the preface to this restricted general language dictionary Proeve van Kaapsch Taaleigen, Changuion states his intent explicitly by saying: "The main purpose of the following selection … has been to rid Dutch, in so far as the language spoken in this colony can be called that, from partially completely foreign, partially mutilated words and expressions, or at least show this direction …" The lemmata in this dictionary represent a selection of Afrikaans words that differ from their Dutch counterparts and the treatment focuses on presenting the proper Dutch form. By doing so, Changuion consistently condemns the Afrikaans forms and prescribes the Dutch forms that he prefers and that he would want to preserve. It is important to note that his dictionary, as was the case with that of La Forêt, was not an independent publication but rather a supplement to a comprehensive grammar De Nederduitsche Taal in Zuid-Afrika Hersteld. The ideas Changuion discussed in the grammar were applied in the dictionary. The types of problems in the demise of Dutch in South Africa, as identified by him, were exemplified in the accompanying dictionary, which did not present an unbiased or neutral selection of items but rather those items regarded by him as contaminated. This was a clear and explicit case of language policy put to practice in lexicographic format.
A different type of language politics consists in describing specific words, expressions or syntactic constructions as wrong or ugly, and to compile a dictionary so that these wrong or ugly expressions can be substituted, e.g. Wustmann (1912). Similarly purism can be seen where foreign words are prohibited and substituted by new ones, e.g. Engel (1918).
In the 17th, 18th and early 19th century, textbooks were aids. This is more evident from the introductions of real textbooks than in the above-described theoretical discussions. The observations can be summarised as follows from different assignments which endeavoured to find: Description of the previous language use as a set of rules for the language system (b) Description of the previous language use for single phenomena (c) Criticism of the previous language use with real texts as examples (d) Suggestions for the preservation or change of the set of rules for the language system (e) Suggestions for the preservation or change of the use of specific single phenomena (f) Information as aid in text reception problems (g) Information as aid in text production (h) Information as aid in translation (i) Execution of language and communication policy The last point has not been deduced from the analysis of older grammars. It has been added here to be discussed later in the article. So far, it can be determined that description had been the approach in points (a), (b), (c) and (f), whereas prescription is prevalent in (d) and (e). For language production (g), the option was either a prescriptive or a descriptive approach. Language production is therefore the point where the problem can be detected most clearly. This is not the only solution and perhaps not even the best that is found in, among others, Bauer (1827-1833) when he presents different variants but elevates one which he recommends as rule.

Prescription, description and proscription
As stated in the first sections of this article, the notions of prescription and description have prevailed in lexicographic practice, although for many years the words prescription/description did not acquire terminological status within the field of lexicography. To ensure clarity in future lexicographic discussions, it is necessary that prescriptive/prescription and descriptive/description as general language words but also as potential terms should be used and understood in an unambiguous way. Whereas the terms prescriptive/prescription and descriptive/description are currently used in both the field of linguistics and of lexicography, lexicographic practice has shown the need for a complementing term not previously used in linguistics but relevant in modern-day theoretical lexicography, i.e. the term proscriptive/proscription. Few theoretical lexicographers have given attention to the issues regarding prescription, description and proscription. Important guidance can be found in Bergenholtz (2003). In this section, reference will be made to some of the suggestions of Bergenholtz (2003), and these suggestions will serve as basis for further discussion. Bergenholtz (2003: 70, 71) points out the differentiated perspectives on description in lexicography. According to him, descriptive lexicography refers to the use of data from a linguistic survey and a text investigation in practical lexicographic work. He also argues that a more precise explanation is needed, because provision has to be made for different types of description. This depends on the empirical basis where the following possibilities can be distinguished (Bergenholtz 2003: 71):

Description
analysis of a linguistic survey, (c) involvement of descriptions in existing dictionaries, grammars, monographs, articles, etc., (d) analysis of a number of examples randomly chosen from random texts (corresponding with the practice of dictionary making before the age of computers), (e) analysis of a specifically constructed text corpus, and (f) analysis of usage found in texts in the examined language in all available websites on the internet.
In order to deal with these possibilities, Bergenholtz distinguishes the following types of description, i.e. open and hidden description, total and partial description, strong and weak description, and explicit and implicit description.
Within the dichotomy open and hidden description, open refers to a situation where the outside matter of a dictionary informs the user about the empirical basis of that dictionary, whereas no such information occurs in the case where hidden prevails. Although many users might not be interested in it, critical users do take an interest in an open approach where the guidance in a front matter text will assist them to compare the treatment in the dictionary with their own experience of language. They need a clear indication from the lexicographer concerning the empirical basis or the investigation of the text corpus.
With regard to the distinction total and partial description, total implies that a combination of all the description possibilities ((a)-(f) given above) are used (cf. Bergenholtz 2003: 71). Few dictionaries employ total description. Partial demands that only some of the possibilities are used. Partial description can also refer to situations where the description is not directed at the dictionary article as a whole, but only at an item presenting a specific data type, e.g. an item giving pronunciation. As an example, one can refer to volume XI of the Woordeboek van die Afrikaanse Taal which is partially descriptive by listing all the possible pronunciation variants of a given word. As far as this one data type is concerned, this dictionary also follows an approach of explicit description, e.g. the pronunciation variants given for the lemma sign onkonstitusioneel: With strong description an analysis of a broad empirical basis is a prerequisite, e.g. an exemplary corpus or sample, in contrast to an analysis of a narrow empirical basis in the case of weak description. Explicit description includes all results from the empirical basis in the dictionary, also, for example, obsolete words or mistakes, whereas implicit description includes only certain results in the dictionary and excludes, for example, obsolete words, mistakes and also informal and colloquial words and expressions. In the standard variety of English, advertisement is an established word. In informal English, the variant ad is often found. Where implicit description prevails, the form advertisement will be included but not the variant ad. Explicit description will result in the inclusion of ad as a lemma. Explicit description will include both the correctly spelled and the frequently incorrectly spelled forms. If the English book had frequently been misspelled as boook, a typical user of the dictionary coming across the form boook will need guidance to help him/her finding the correctly spelled form. Explicit description will include the form boook, and its treatment will refer the user to the correct form book. Bergenholtz (2003: 74) gives the following three possibilities of prescription:

Prescription
(a) a specific linguistic variant is explicitly prohibited, (b) one or more linguistic variants are explicitly prescribed, thus prohibiting all other non-mentioned variants, and (c) a specific linguistic variant is explicitly prescribed (as opposed to prescription (b), this involves a new word, new spelling, new pronunciation, new inflection or neologism).
It should be noted that prescription is typically employed in specific dictionary articles, and therefore all these possibilities are not used in every article of a dictionary with a prescriptive approach. The distinction between a descriptive and a prescriptive approach is not always absolutely clear, and dictionaries opting for a prescriptive approach will employ this approach in certain articles but will also include articles similar to, for example, those in weakly descriptive dictionaries. As is the case with description, prescription can also be explained with reference to different dichotomies, i.e. open and hidden prescription, total and partial prescription, strong and weak prescription, explicit and implicit prescription. With open prescription the outside matter contains a note in which the intention of the dictionary to influence and eventually change the language use is indicated, whereas such a note does not occur in the case of hidden prescription. One of the best known "declarations of intent" in this regard is Samuel Johnson's passion for "fixing the language", as expressed in The Plan of a Dictionary of the English Language (1747) that preceded the publication of A Dictionary of the English Language (1755). It has to be added that Johnson later admitted that he eventually had to abandon the idea of his dictionary succeeding in fixing the language .
In total prescription a combination of the prescription possibilities (a)-(c) is used, while in partial prescription only one or two of the possibilities can be used. Partial prescription can also refer to situations where the prescription is not directed at the dictionary article as a whole, but only at an item presenting a specific data type, e.g. an item giving pronunciation or an item giving the orthographic form of the lexical item. The original meaning of the Afrikaans word sondebok ("scapegoat") was "person unfairly blamed for wrongdoings". In everyday language use, this word acquired a second sense, i.e. "person who trespasses". From a puristic point of view, linguists and lexicographers warned against this new sense, and a partial prescriptive approach could have seen a treatment directed at indicating the negative attitude towards this second sense.
Strong prescription prohibits and only allows certain variants in cases with a clear difference from the normal language, whereas with weak prescription the dictionary articles have items that prohibit and only allow certain variants in certain cases without showing such a clear difference from normal language use. With explicit prescription one variant is explicitly allowed and another explicitly prohibited, whereas implicit prescription allows one variant but prohibits all other variants by omitting them.

Proscription
Proscription allows the same possibilities for the empirical basis as description, cf. the options (a)-(f) given above in section 3.1. However, the results of empirical analysis are dealt with in a different way compared to a descriptive approach. In this regard the most salient distinction lies in the fact that the lexicographer does not only provide the results from the empirical analysis but goes further by indicating a specific variant that he/she regards as the recommended form. Bergenholtz (2003: 77) sees it as "selective description", a formulation which has to be interpreted as referring to weighed description. The following distinctions apply to proscription: open and hidden proscription, total and partial proscription, strong and weak proscription, and exact and non-exact proscription.
Within the dichotomy open and hidden proscription, open refers to a situation where the outside matter of a dictionary informs the user about the empirical basis of that dictionary, whereas no such information occurs in the case where hidden prevails. With regard to the distinction total and partial proscription, total implies that a combination of all the proscription possibilities ((a)-(f) given above) are used, whereas partial demands that only some of the possibilities are used. With strong proscription an analysis of a broad empirical basis is a prerequisite, in contrast to the analysis of a narrow empirical basis or, in the case of weak proscription, without any reference to a broad empirical basis. Exact proscription demands that only one variant is recommended although other variants may be mentioned, whereas non-exact proscription recommends more than one variant, while other variants may be mentioned. Volume XIII of the Woordeboek van die Afrikaanse Taal shows a change in approach with regard to the presentation of items giving the pronunciation of a word. In the previous volumes, a descriptive approach resulted in all the variants being given without any recommendation from the lexicographer. Volume XIII still offers all the variants, but the user guidelines state quite emphatically that where more than one pronunciation variant is given, the first is the variant recommended by the lexicographer. This is a type of exact proscription. Complementing the notion of exact proscription where other variants may be mentioned (Bergenholtz 2003: 77), Gouws and Potgieter (2010) make provision for an approach of exclusive proscription according to which only the recommended form and not the other variants will be included in a given dictionary. However, unless a form of marking or labelling is used in the dictionary, the user will be unable to distinguish between exclusive proscription and implicit prescription. Important in a proscriptive approach is an acceptable way to deal with the non-recommended forms. Tarp and Gouws (2008) discuss various aspects regarding the implementation of a proscriptive approach, including different ways in which the non-recommended forms can be presented.
In the existing literature on proscription, the defining feature has been identified as the giving of a recommendation. This remains the unique characteristic nature of proscription. However, the application of a proscriptive approach can be enhanced if the recommendation is supported by a motivation in which the lexicographer gives his/her reasons for opting for the specific variant or form and not for another. Such a motivation may give the user the needed assurance regarding the appropriateness of the specific form.

Realising ideological issues
The use of prescription, description and proscription in dictionaries should not be seen in isolation. As the motivation for these approaches often lies on a dic-tionary-external level, these applications can be part of the attempted realization of an underlying ideological issue. Two of the real issues coming to the fore in the application of, for example, a prescriptive, descriptive or proscriptive approach are language policy and language criticism. Bergenholtz and Gouws (2006) have argued that every lexicographic decision may be regarded as the result of a political decision, more precisely a language and communication policy decision. Dictionaries can be employed to support a specific language and communication policy, and a prescriptive, descriptive or proscriptive approach may be chosen to realise such an assignment.

Description, prescription, proscription: Opting for the best practice approach in general language dictionaries
Although Bergenholtz (2003) has indicated that in total description/prescription/proscription, a combination of description/prescription/proscription possibilities are used, he did not make provision for a combination of subtypes of description/prescription/proscription. However, before discussing this possibility, it is important to know exactly when prescription/description/proscription could be used in dictionaries. Yet again the determining role of lexicographic functions needs to be acknowledged. Within a user-driven approach to lexicography, every aspect of the contents and structures of dictionaries have to be determined by the envisaged function(s) of the specific dictionary. The central question the lexicographer needs to ask him-/herself remains: "What do I want my user to be able to do with this dictionary?" A functional approach demands that notions like prescription, description and proscription should be seen as ways to assist in achieving the function identified for a given dictionary. Consequently the decision of a lexicographer should not be to compile, for example, a prescriptive, a descriptive or a proscriptive dictionary, but rather to compile a dictionary to achieve for example a text reception, text production or cognitive function. This function of the dictionary will then determine a process that might lead to a stronger prescriptive, descriptive or proscriptive approach.
Having identified the function of a dictionary, the lexicographer may decide on a process with a stronger prescriptive, descriptive or proscriptive approach. The lexicographer will be aware of the specific process employed in the given dictionary, but, although other lexicographers may endeavour to detect the approach, they will find it difficult to make an unambiguous classification, although they may be able to identify certain cases of prescription, description or proscription in the treatment presented in some individual articles. Where text production is the prevailing function, the lexicographer may do well to introduce some prescriptive or proscriptive entries in order, for example, to ensure innovative language use, to preserve existing variants or to bring about changes. Where text reception is the envisaged function, a prescriptive approach serves no purpose. The lexicographer could assist the user much better by introducing a descriptive approach based on a thorough text analysis.
When opting to compile a dictionary based on lexicographic functions, the lexicographer is not primarily interested in the different subtypes of prescription, description or proscription, valid as they may be. He/she much rather tries to employ a combination of these subtypes that can best help to achieve the identified lexicographic function. This demands the application of open, total, strong, explicit description, prescription or proscription. Such a combination helps the lexicographer better to achieve the desired function.

Reception
The principle that all lexicographic decisions should be determined by the envisaged lexicographic functions, also applies to the methods of selection, i.e. whether a descriptive, prescriptive or proscriptive approach is followed. Selection according to functions at the same time assumes that a quicker and better access of the data in the dictionary can be obtained in an unproblematic way. There is, for example, no access possible when an orthographic variant does not belong to the dictionary data. Until 2001, two Danish equivalents for the English word line were used and officially recognised by the language commission, i.e. linie (with -i-) and linje (with -j-). But since November 2001, only linje (with -j-) has been permitted. In new dictionaries compiled after this date, and this applies to both printed and internet dictionaries, only the permitted variant has been included. The newly prohibited form does not form part of the dictionary data. Because this variant previously was and still remains the most frequently used form, many users are looking for guidance regarding this spelling. This type of prescription therefore has negative consequences for the usability of a dictionary. A prescriptive approach that omits the prohibited variants, does not give adequate solutions when users are looking for an aid that can really assist in solving reception problems. Proscriptive solutions are also unsatisfactory and not recommended, because information regarding the meaning of words, word combinations or sentences is needed when reception problems are experienced.
For a general language reception dictionary, an open, total, strong and explicit description would usually be the optimal procedure. For specialised reception dictionaries, a text corpus is not required as empirical basis, because only partial and not total description is needed. After it has been established that reception dictionaries need to be descriptive, the specific consequences still have to be resolved. The types of items required in a reception dictionary that have to be selected and described according to descriptive methods still have to be established. Therefore the following suggestions are given albeit that the last two ((c) and (d)) occur in general language dictionaries, but very seldom in specialised dictionaries: Item(s) giving the meaning of the lemma (c) Idioms and other word combinations whose meaning cannot be determined by merely looking at the meaning of the individual words (d) Item(s) giving the meaning of idioms or other word combinations Regarding the selection of lemmata and idioms, it should especially be noted that the frequency criterion, so popular in modern-day lexicography, does not have priority. In prevailing log file analysis (cf. Bergenholtz and Johnsen 2007), no investigations have been made to determine whether the words with the highest usage frequency in texts are also those consulted the most frequently for text reception problems in internet dictionaries. It is presumed that the words most frequently looked up in general language dictionaries for reception problems belong to the following groups: (a) infrequently used words, (b) obsolete words, and (c) polysemous words (where the user is not quite sure how to distinguish between the different senses). But frequently used words are certainly not looked up so often, at least not by mother-tongue speakers. Amongst foreign users, it would be different. In specialised dictionaries, there would be big differences between lay persons, semi-experts and experts. When consulting a reception dictionary, lay users would typically behave as foreign users, i.e. they would often look up the frequently used terms. However, although these are still only suppositions, they seem to apply more likely when selecting lemmata and idioms, than the highly acclaimed frequency criterion (cf. Bergenholtz 1992Bergenholtz , 1994. There is an additional selection possibility resulting from a late survey of test persons. A survey is regarded as one of the empirical possibilities to gather data for description. This possibility is familiar from the experience with printed dictionaries where interested users write to the publisher that they looked up a word or expression but could not find it in the dictionary, constituting a so-called lemma gap. In electronic dictionaries, this method can be optimally employed because the log files can be examined every day for the inclusion of those words and expressions, looked up but not found, as lemmata in new dictionary articles.
Those words and idioms presented in reception dictionaries have to be provided with an explanation. This is the item that the data users look-up must contain to ensure a successful consultation. But how does a descriptive item giving the meaning look? In the first instance, all senses of a lemma or idiom should be given and no "non-fitting" senses should be omitted, e.g. the second sense of the Danish verb hustle, meaning both "to make an untrue and deceitful statement" and "to have sex with a man for money or other favours". This might perhaps not be controversial. But a descriptive item presenting meaning may also consist of critical remarks regarding specialised terms, e.g. from a scientific perspective it is false for a music dictionary to explain the term Vienna Classicism as referring to Mozart and Haydn, because they belong to the classi-cal period. This then is not a prescriptive item but a describing addition that aims to correct the scientific classification.
It can certainly be argued that collocations, examples or synonyms can also be helpful when a reception problem occurs. It has to be stressed that the user finds the collocations and examples in the text where the reception problem occurs. The user therefore needs an appropriate item giving the meaning that will solve the reception problem. It here concerns the principle of a monofunctional dictionary containing as much data as necessary but as little as possible to guarantee a rapid access that is not impeded by unnecessary data or that leads to information stress or even information death.
Finally, by the use of the predicate descriptive, it has not been indicated how comprehensive the item giving the meaning in a reception dictionary should or could be. The question really is whether a descriptive item giving the meaning should only contain so-called essential items in order to be an optimal entry in a reception dictionary. This does not imply a direct distinction between semantic and encyclopaedic data -a distinction that in any case cannot be maintained from a purely scientific perspective (Haiman 1980: 351 andKaufmann 1996). The question of what is needed and of what is redundant is important, because a pure descriptive description of the meaning of a word can easily amount to the extent of a complete book. The distinction between essential and non-essential items giving meaning was a central theme in the journal Dictionaries (1993). Wierzbicka (1993) distinguishes and believes to be able to distinguish clearly between meaning and knowledge. She argues that much space is wasted when encyclopaedic knowledge is also presented in a dictionary article. She criticises the following article: sugar a sweet substance that consists wholly of sucrose, is colourless or white when pure, tending to brown when less refined, is usually obtained commercially from sugar-cane or sugar-beet, and is nutritionally important as a source of carbohydrate as a sweetener and preservative of other foods. and proposes instead a lexicographic definition that is much easier to understand: sugar something that people add to things they drink or eat when they want to make them taste sweet; it comes from some things growing out of the ground; it is white.
To both articles various entries could be added, e.g. that Wierzbicka's proposal does not contain "sugar-cane" or "sugar-beet" but rather (a) "it comes from some things growing out of the ground" (b) "it is white". The latter statement (b) is false, because brown sugar is also found. The former part (a) is difficult to comprehend. Even a young school child would understand "plants" and "sugar-cane" rather than the formulation the author obtained by means of a linguistic theory of semantic primitives and not a the-ory of dictionary functions. The real question in this regard is that of description. Both these entries can hardly be called descriptive as far as descriptive completeness (cf. Bergenholtz and Schaeder 1977) is concerned. Comprehensive books have been written about sugar. The real issue is not whether it is comprehensively descriptive or in one way or another prescriptive, but rather for whom and for what purpose the meaning has been presented in the dictionary. It can be maintained that for receptive needs a descriptive approach is the only appropriate one. But there are different ways of presenting an appropriate description.
Whether long or short is not the question, but rather for which user group and which type of usage situation the entry has been planned. Also the distinction between semantic and encyclopaedic is not really the issue, because this distinction is scientifically not tenable (cf. Bergenholtz and Kaufmann 1996). The question is rather whether assistance is needed for reception, for obtaining as much knowledge about a given matter, word or term, and also whether for lay persons or semi-experts. Compare in this regard two different articles, both planned for a reception dictionary. The first is aimed at lay persons in the specific field: bacteriophage Bacteriophages belong to a group of viruses that infect bacteria.
The second entry comes from a proposal for a reception dictionary for semiexperts in the given field: bacteriophage Bacteriophages, or phages, are viruses that infect bacterial cells. The size of a phage particle is 20-200 nm (1 nm = 10 -9 m). A phage usually consists of two components, a chromosome of DNA or RNA and a protein coat, the capsid, which serves as a protective shell containing the genome and which is involved in the infection process.
The extent of the articles is not the question, but rather which items are needed for a reception of different user groups of a given dictionary. Bergenholtz (1998) also contains the following test entry for a gene technology dictionary for lay people and semi-experts. Here the entry for semi-experts is quite brief, because this term is not often used, compared with the terms gametes, gameteproducing cells and bacterial spores: germ cell The expression germ cells is sometimes used for gametes, gamete-producing cells, or bacterial spores.
The test entry for lay people is in principle not only longer, but also more informative, because an item giving the meaning is presented corresponding to the general language use, i.e. how gamete is used: germ cell Germ cells are the base for sexual reproduction; a fusion of a male and a female germ cell causes the fertilization of the egg.

General remarks regarding text production
In the previous section on text reception, it has been mentioned that this article is directed at monofunctional dictionaries. This needs be emphasised yet again, because of the real implications a monofunctional approach also has for text production. Too many existing general dictionaries give no indication regarding their specific communication function. The default approach of many lexicographers is that they are producing a polyfunctional dictionary which should assist the user in satisfying at least a cognitive function, a text reception function and a text production function. Consequently dictionary users do not really know what a general language text reception or text production dictionary should look like, because they have not encountered such dictionaries in their usual dictionary consultation procedures. Where such monofunctional dictionaries have been produced, users responded in a positive way and utilised these dictionaries as practical instruments to assist them in solving specific problems. When text production is focused on from a descriptive/prescriptive/proscriptive approach, it is important to have an unambiguous interpretation of the implications of text production for the content and presentation of dictionary articles. One of these implications is that the dictionary will not necessarily include all the high-frequency lexical items from a given language nor will the dictionary necessarily display a homogeneous article structure or a comprehensive treatment of the meaning of a given word. A detailed discussion of the choice of lemma candidates and the data types to be included will not be given in this article -it belongs in an article on text production. However, some aspects of these issues should be dealt with here.
With regard to almost all aspects of text production, a distinction needs to be made between dictionaries compiled for mother-tongue speakers of the treated language and dictionaries compiled for non-mother-tongue speakers of the treated language. The scope of the data included in a text production dictionary will also be determined by the medium of the dictionary, i.e. whether it is planned as a printed or an electronic dictionary. This aspect will not be discussed in this article.
The following types of items could be included in a general language text production dictionary: the lemma sign, pronunciation data, grammatical data, a brief explanation of the meaning of the lemma or a mere translation equivalent, one or more example sentences, one or more collocations, synonyms, and antonyms. Where idioms and fixed expressions are included as treatment units, a text production function compels the lexicographer to present items indicating the typical use of the idioms and fixed expressions.
It is important to pay attention not only to the data types that have to be included in, but also to those items or indicators that should be omitted from a text production dictionary. A popular feature especially in English learners' dictionaries is the use of non-typographical structural indicators to mark usage frequency of words, e.g. a system where five stars indicate that a word belongs to, say, the top 500 words of the language, whereas one star indicates that it belongs to, say, the top 3 000 words of the language. From a text production perspective, these indicators have little value. The user consults the dictionary to find text production assistance regarding a specific word. When he/she finds this word in the dictionary, the information that needs to be retrieved is related to the use of the word. A mere indication of the usage frequency data, i.e. usage frequency for the sake of usage frequency, does not help with text production. The only value it may have could be if it is used in a relative way, i.e. to support other entries. For example, in the articles of words marked as infrequently used, a cross-reference is given to an appropriate synonym with a higher usage frequency that could enhance the communicative success of the text production procedure. Another value of frequency indications could be to support the motivation for a given recommendation when applying a proscriptive approach. For example: Danish has the following sets of variants: bevislig x beviselig and ubevislig x ubeviselig. The proscriptive Danish internet text reception dictionary (The Danish Internet Dictionary) recommends the use of bevislig and ubeviselig. Following an enquiry from a user about the choice of recommended forms, an investigation of the usage frequency of these forms indicated that the recommended forms show 9 180 and 2 208 occurrences respectively, whereas their non-recommended counterparts show a usage frequency of 1 900 and 448 respectively. This support from a corpus enabled the lexicographer to add a motivation for the recommendation in which usage frequency is given as a criterion for the specific recommendations.
In a text production dictionary, the treatment of the lemma has to be planned in terms of the needs of the user of the specific text production dictionary. The criteria for the selection of lemma candidates for a text production dictionary have to differ from the criteria applicable to text reception dictionaries. The typical needs of mother-tongue speakers regarding the selection of lemmata will differ from those of the non-mother-tongue speakers. Lexicographers can work with the assumption that mother-tongue speakers will have a better knowledge of high usage frequency words compared to non-mothertongue speakers. However, where this will play a determining role in the selection of lemma candidates for a dictionary for text reception, it will play a diminished role in text production dictionaries. The user of a text production dictionary does not primarily need to obtain the meaning of the common word but rather information regarding its productive use. High usage frequency words may have unpredictable uses or collocations for a given polysemous sense that are not so well known. In order to present the user with these example sentences or collocations, even high usage frequency words may qualify as lemmata in text production dictionaries. The selection of lemmata must be determined by the text production assistance the inclusion and treatment of a given lexical item as lemma can give to the target user of the specific diction-ary. This implies that not only correctly spelled words should receive lemma status. Words that are frequently misspelled should be considered for inclusion as lemma candidates. A user might want to use a word in a text production situation, but is uncertain of the spelling. Looking for the incorrectly spelled form (cf. the comments on the form boook in section 3.1) and finding it in a dictionary with a cross-reference to the correctly spelled form, the user will be assisted in proper text production. The following can serve as another example: The German equivalent for sixteen is sechzehn. Users often think the spelling is sechszehn. This incorrectly spelled form can be included, with a cross-reference to the lemma sign sechzehn. Having found this lemma and the cross-reference in the dictionary, the user can use the correct spelling sechzehn when producing new texts.
For text production purposes, pragmatic guidance can be very helpful. A text production dictionary should therefore employ a well-defined set of pragmatic labels to guide the user with regard to, among others, stylistic, geographic and subject-specific restrictions of a given word, expression, sense or other item in a dictionary article. The use of stylistic labels like informal, colloquial, obscene and vulgar constitutes a type of guidance non-mother-tongue speakers of the treated language especially need to avoid communicative embarrassment.
Data on pronunciation is valuable for spoken text production. Where a dictionary has oral language use in its text production scope, items giving pronunciation should be included. This is primarily as a response to the needs of users who are non-mother-tongue speakers of the language treated in the dictionary. For mother-tongue users, more limited pronunciation guidance could be given. The treatment in text production dictionaries is often focused on written texts, and in these dictionaries pronunciation guidance is not needed. Where the focus is on written texts, grammatical data, e.g. entries presenting pluralisation, degrees of comparison, and the tenses of verbs, can play an important role. The user might be familiar with, for example, the singular form of a noun, but needs to use the plural form in a given text. In a dictionary for mother-tongue speakers, items giving the plurals that are formed systematically are not so important, because the typical user should be familiar with the systematic and predictable morphological patterns of the language. Again, however, it is important that the lexicographer has to work with a clearly identified target user and has to be familiar with the needs and reference skills of this user. The absence of items giving the plural forms of some nouns should not imply that no plural form may be given. Where the lexicographer regards it as important for the target users in their text production endeavours to have access to a specific, typically unsystematically formed plural, such an item needs to be included. It should be noted that, for example, all nouns do not have to be treated in exactly the same way in such a dictionary. The lexicographer should employ the users' guidelines text to give an account of the different approaches in the dictionary. However, the lexicographer should never use the opportunity to give explanations in a users' guidelines text as an excuse to present the data in the dictionary in an ambiguous way or in such a complicated or condensed way that the user can only comprehend the system of the dictionary by using the guidelines text. A good dictionary will include a users' guidelines text, but the presentation in the articles has to be of such a nature that the typical target user should be able to use the dictionary, albeit not necessarily in an optimal way, without consulting the users' guidelines text.
Where users have text reception problems, the typical information they need to retrieve from a dictionary article is of a semantic nature. They need a relatively comprehensive explanation of the meaning of the word. Contrary to this, the users of a text production dictionary merely need to confirm the meaning or a specific polysemous sense of a word. A full explanation of the meaning is not needed. If the dictionary is compiled for non-mother-tongue speakers, the semantic data may be restricted to a translation equivalent in the mother-tongue of the target user. This will confirm the meaning of the word to the users and enable them to proceed with their text production activities. In addition to the explanation of meaning and the presentation of translation equivalents, an indication of semantic relations, especially synonymy and antonymy, holding between lexical items is a typical part of the semantic assignment of dictionaries. The inclusion of synonyms in a text production dictionary assists users in finding a more varied selection of words. Because the user is consulting the dictionary by looking at the treatment of a specific lemma, the synonyms should be included as lemmata in their own alphabetical article stretches and each article needs a cross-reference to the other synonym(s). The explanation of meaning will typically be given in the article of one of the synonyms, with the other article(s) containing a cross-reference to this article. In the distinction between absolute and partial synonyms, it becomes clear that partial synonyms occur much more frequently than absolute synonyms. The proper use of partial synonyms is more demanding than that of absolute synonyms. Compared to non-mother-tongue speakers, mother-tongue speakers are better equipped to use them appropriately. Consequently a text production dictionary for mother-tongue speakers of the treated language may embark on a more comprehensive presentation of partial synonyms than its non-mother-tongue speaker counterpart. In text production, the negation of a contrasting form is often employed to express a given meaning. Instead of saying the door is open one would say the door is not closed. In order to enable users to express themselves in this way, it is important that text production dictionaries should also include some antonyms.
One of the most salient features of a text production dictionary is its presentation of example sentences and collocations. It has to be accepted that only a limited number of examples and collocations can be included in a printed dictionary. An electronic dictionary has less space restrictions, but can also not include everything. In dictionaries claiming to have a text production function, it is often found that the examples illustrate the most general and typical use of a given word, and the collocations are the most frequently encountered ones. Again the lexicographer should negotiate the target users. Where non-mothertongue speakers of the treated language are the target users, some of the general collocations and example sentences giving typical uses can be included, but the mother-tongue speakers are usually familiar with the most typical uses and collocations and are rather in need of lesser known and used forms.
In trying to satisfy a text production function, the lexicographer should pay careful attention to the application of an approach characterised by either description, prescription or proscription or a hybrid application in which more than one of these approaches can be combined. This decision should not be made in a haphazard way.
Where prescriptive and proscriptive approaches are employed, a user consulting a dictionary for text production purposes should preferably know whether the lexicographer is prescribing or recommending a given form and by doing so signalling that there are other non-prescribed and non-recommended variants. Where a given form has no variants, e.g. regarding pluralisation or orthography, the form given in the dictionary article is not only the only appropriate form but often also the only existing form. Users need to be informed accordingly so that they can know the entry does not represent an application of implicit description, implicit prescription or exclusive proscription where only one variant is included and the others excluded. But even when only one recognised form exists, one often finds non-recognised forms that are the result of, for example, spelling mistakes. As indicated earlier in this article, a text production dictionary should make provision for the inclusion of these forms to assist the user in not using them. When one looks at the different types of description, prescription and proscription (cf. Bergenholtz 2003 and section 3 of this article), it becomes clear that description is not a viable option for text production if more than one variant prevails. The user in need of text production assistance does not want to make choices, but is rather looking for the best form for a given context. This applies to the choice of a word or a grammatical form, e.g. the choice between two plural forms, synonyms, collocations, etc. This is especially true in the case of users who are non-mother-tongue speakers of the treated language and do not have the ability to evaluate the different variants and their appropriateness for a given situation of use. The success of their dictionary consultation procedure should not rely on an uninformed choice they make between different variants. Mother-tongue speakers of the treated language who have a good command of their language will also have to make a choice when confronted by a descriptive approach in which more than one variant is presented. Even then description will not be the ideal solution, especially not for a quick answer to their text production problem, because negotiating the different variants can be time-consuming. However, where cross-references between lemmata representing different variant forms are given, a knowledgeable user may benefit from the descriptive approach by being made aware of different variants from which a choice could be made. If these variants are not all equally suited for all contexts, they have to be labelled to indicate, for example, style and usage differences. Although the presentation of variants can be seen as a form of description, the added labels bring a proscriptive or even prescriptive nuance to the fore, because they indicate the environment where the use of the specific variant is prescribed or recommended. This can be regarded as another form of hybridisation between description, prescription and proscription.
For text production purposes, prescription can be a viable option. A prescriptive approach influences future text production activities and gives the user one prescribed form to use. The success of prescription in a text production dictionary depends on the type of prescription employed. Implicit prescription which allows one variant but prohibits all the other variants by omitting them is of little assistance, because the users will not find the lemma they are looking for if it is a prohibited variant. Explicit prescription with one variant explicitly allowed and another explicitly prohibited might be more useful. A cross-reference entry should guide the user from the prohibited to the prescribed form. Prescriptive entries are often guided by ideological and language political motivations. They do not always represent the actual and default language use. Following the prescribed advice can lead to a user producing a text in which unnatural language is used, e.g. superstandard, puristic or idiolectal forms. Where a user relies on prescriptive dictionary entries, it may impede the communicative success of a text production process. However, if the lexicographer wants to introduce an innovative form and assist users in using such a form in a proper way, the prescriptive approach has its advantages.
Because proscription relies on an empirical basis, the actual language will be reflected in a proscriptive approach, especially in an application of strong proscription where an analysis of a broad empirical basis is a prerequisite. Explicit proscription gives the lexicographer the opportunity to include different variants, but also to indicate the recommended form. If a proper system of cross-referencing is employed to link the non-recommended variants to the recommended form, proscription can be the best option in a text production dictionary. This applies to all the different data types where variation occurs and a recommendation is made.
A proscriptive approach can add value to the items presented in the article slot for collocations. A given word can combine with different words in different collocations in order to express the same meaning. In Afrikaans the noun antwoord ("answer") can be used in the collocations 'n antwoord gee/'n antwoord verstrek ("give an answer"). As both are correct and good Afrikaans, a descriptive approach will include them as variants. The collocation 'n antwoord verstrek is slightly more formal than 'n antwoord gee, but not so formal that it needs to be labelled as such. The latter form can be used in both formal and informal situations and should be recommended as the best option for general use. Here the application of exact proscription will not only allow the inclusion of the recommended form, but also that of 'n antwoord verstrek as a variant form.
A major advantage of a proscriptive approach lies in the fact that the lexicographer as someone acutely aware of the needs of the intended target users of the specific dictionary, can make a recommendation that should suit the intended target user in the best possible way in his/her text production endeavours. In this regard it is important to note that a proscriptive approach does not imply that the same recommendation for the same function will necessarily be given in different dictionaries. The recommendation in a text production dictionary for learners in primary school may differ from the recommendation in a text production dictionary for adult users. The needs of the user will determine the nature and extent of the recommendation to ensure the best text production possibility.

Mono-and polyfunctional dictionaries
This article is not exhaustive, because only two of the four main types of dictionaries with communication functions have been discussed. The remaining functions of text correction and translation demand a separate discussion. Even more important would be the discussion of the cognitive functions, which have not been treated here. Such a contribution is urgently needed because of all the uncertainty and vagueness that prevail. However, some aspects in this article do apply to the cognitive functions of a descriptive dictionary (e.g. a documentary function), some to that of a prescriptive dictionary (e.g. when aiming to achieve a fundamental or partial change of existing norms or uses) and some to that of a proscriptive dictionary (e.g. as an aid in language learning). A main reason why the discussion of description versus prescription up till now has been less than productive and that it has, for example, led to faithrelated controversies, is to be found in the fact that only strong polyfunctional dictionaries had been targeted for investigation. It might be that for small languages and for many specialised languages only polyfunctional dictionaries could have been compiled, because the market would not allow the full spectrum of possible and necessary monolingual dictionaries. But this does not change the fact that such an argument does not apply to electronic dictionaries, where monofunctional dictionaries could be extracted from a huge mutual databank.