Kirkeby's English–Swahili Dictionary

Kirkeby's English–Swahili Dictionary is a bilingual dictionary of more than 50 000 entries. The most laudable feature of the dictionary is its attempt to be user-friendly especially in the way the entry words have been arranged and the amount of information given. However, a clear objective for the compilation of the ditionary is lacking. The compilers do not seem to know the lexicographical gap they want to fill, the users they are targeting, and their dictionary-using skills. In discussing the strong and weak points of the dictionary, the article will refer to theories of dictionary criticism. Three criteria set by McMillan (1949) will guide this review article: (1) the quantity of the information in the dictionary; (2) the quality of the information presented; and (3) the effectiveness of the presentation of the information. Questions posed in the course of this article will include: Does the dictionary give the information required by the user? Is the information transparently accessible? How is the information presented?


Opsomming: Kirkeby se English-Swahili Dictionary. Kirkeby se English-Swahili
Dictionary is 'n tweetalige woordeboek met meer as 50 000 inskrywings. Die mees prysenswaardige kenmerk van die woordeboek is sy poging om gebruikersvriendelik te wees, veral deur die manier waarop die trefwoorde gerangskik is en die hoeveelheid inligting wat verskaf word. 'n Duidelike doelwit vir die samestelling van die woordeboek ontbreek egter. Die samestellers is skynbaar onseker oor die leksikografiese leemte wat hulle wil vul, en die gebruikers vir wie dit bedoel is en hul woordeboekgebruikersvaardighede. In die bespreking van die sterk en die swak eienskappe van die woordeboek sal die artikel verwys na teorieë van woordeboekkritiek. Die kriteria gestel deur McMillan (1949) sal as gids dien vir hierdie resensieartikel: (1) die hoeveelheid inligting in die woordeboek; (2) die gehalte van die inligting aangebied; en (3) die doeltreffendheid van die aanbieding van die inligting. Vrae wat in die loop van hierdie artikel gestel sal word, sluit in: Gee die woordeboek die inligting wat deur die gebruiker verlang word? Is die inligting deursigtig beskikbaar? Hoe word die inligting aangebied? Introduction Throughout its long history, the dictionary has always been an object of criticism and it is this practice that helped it to grow from its beginning as a glos-sary to a book of multi-linguistic and multi-cultural information like James Murray's Oxford English Dictionary. Some of the early recorded dictionary critics as quoted in Landau (1984: 48-56) include Addison who suggested an English dictionary with quotations from literature. Many critics of 18th century English dictionaries such as Swift, Defoe, Pope, and others, proposed a dictionary attempting to survey and record the English language. Johnson (1755) took into account proposals made by these critics and introduced into his dictionary illustrative quotations as evidence of how a word is used in its different contexts and meanings. Johnson had his critics too. Richardson and Trench, as quoted in Mathews (1933: 63-65) criticized Johnson and others for not adhering to the principles of historical lexicography. It is on the basis of these comments that James Murray with others compiled the Oxford English Dictionary according to the principles of historical lexicography. Modern dictionary critics include Lemmens and Wekker (1986) who reviewed grammar in English learner's dictionaries in which they pointed out flaws and inconsistencies in entering grammar in these dictionaries. Recommendations made by the critics helped to improve the treatment of grammar in the Oxford Advanced Learner's Dictionary of Current English, the Longman Dictionary of Contemporary English, the two volumes of the Oxford Dictionary of Current Idiomatic English and the Longman Dictionary of Phrasal Verbs. Critical reviews of the coding system of the dictionaries and proposals for an adequate coding system helped much to improve the coding system used in presenting various kinds of grammatical information.
The value of a dictionary is not based on the size of the dictionary alone (Ilson 1988: 10) but also on other more important factors. Jackson (1998: 28) puts these in question form: Is the information required by the user present in the dictionary and is it transparently accessible? How is the information presented?
The following three criteria for evaluating dictionaries set forth by McMillan (1949) will be used in discussing Kirkeby's English-Swahili Dictionary: (1) The quantity of information the dictionary offers. This includes the number of entries, meanings, new words (when compared with its competitors), synonyms, pronunciation, etymology and the use of subject and usage labels.
(2) The quality of the information presented in the dictionary. The aspects to examine here are: (a) the accuracy of the information, (b) the completeness of the information, (c) the clearness of the information, (d) the simplicity in presenting the information, i.e. the information should be rendered in such a way that it can easily be deciphered, and (e) the modernity of the dictionary, i.e. the number of current words in the language which has been included in the dictionary.
(3) The effectiveness of the presentation of the information. This includes the order of arranging entries, the placement of etymology, the ordering of senses, and the presentation of pronunciation. Also important is the typography.
McMillan notes that any thorough review should also examine the essays in the front matter and the appendices in the back matter. It is important that a dictionary review should further answer the following questions: What are the purposes of the particular dictionary? What are its properties ? How is its content presented and how are the information categories structured?

Kirkeby's dictionary
If a dictionary were to be judged by the number of its printed pages alone, the number of entries it purportedly contains according to the blurb, the information categories it has and the size of the paper used in printing it, Kirkeby's dictionary would undoubtedly have been the largest English-Swahili dictionary. It has 1 069 pages in comparison with the 924 pages of TUKI (1996). The largest Swahili dictionary is the Swahili-French dictionary of Sacleux (1939) with 1 112 pages. Kirkeby (2000) is the first English-Swahili dictionary that offers a pronunciation guide for English words. Even the paper size is also the largest. Unlike TUKI or Johnson or many English dictionaries whose page numbering in Arabic numerals only begins with the main text, Kirkeby starts off with Arabic numbering from the first page so that the main section of the dictionary only begins on page 13. Kirkeby (2000) is a replica of an English-Norwegian dictionary by the same compiler. The project started in 1997 when Willy Kirkeby and his assistants/ consultants began to replace the Norwegian text with a Swahili text (see Preface). The dictionary was completed in two and a half years. In addition to several assistants, the compiler had the help of a Swahili novelist, a teacher and a Swahili expert who acted as proof reader. Kirkeby is a prolific dictionary compiler. He has to his credit ten other dictionaries which were compiled between 1979and 1998and which were published in 1979, 1983, 1986, 1988, 1991, 1992, 1993, 1996and 1998. Two dictionaries appeared in 1988. Of these dictionaries, six were Norwegian-English while four were English-Norwegian, one of which was used in compiling this English-Swahili dictionary.
Kirkeby does not explicitly identify the user profile of his dictionary except mentioning nurses complaining that they had no dictionary to look up medical terms they came across in textbooks (see Preface). It is important for any dictionary project to identify the users, the information the users need and if they can understand such information. The lexicographer has therefore to answer the following questions: Who will use the dictionary? What will they use it for? What are their reference skills and what are their expectations? It is important to recall the observation of Barnhart (1962: 161): "It is the function of a popular dictionary to answer the questions that the user of the dictionary asks, and dictionaries on the commercial market will be successful in proportion to the extent to which they answer these questions of the buyer." Dictionary writing should not be driven by commercial factors alone, e.g. the potential size of the market and current competitor dictionaries in the market, but should be designed with a special set of users and their specific needs in mind. Apart from being guided by the three aspects of evaluation outlined in par. 1, the following discussion will especially consider the needs of the two main user groups for whom this dictionary is possibly meant: English learners of Swahili and Swahili learners of English.

Entries in Kirkeby
Kirkeby estimates that his dictionary contains 60 000 entries within the 1 047 pages of the main dictionary text. But a rough estimate shows that the dictionary may contain a little more than 50 000 entries. The entries are arranged in alphabetical order. An entry in this dictionary could be a basic form, a derived or inflectional form of the basic form, a compound word or verbal phrase and fixed expressions whose first member is the headword. For example: (1) accident, accidental, accidentally, accidental death, accident write-off.
All these entries are in bold-face. In addition to this, some phrasal terms in bold-face have also been entered as run-ons. For example: (2) tablet … n 1. … kidoge (vi-); sleeping ~ …; suckable ~ … I. track … n. 1. alama (-) (of ya); cover one's -s …; colloq: stop dead in one's -s …; off the beaten ~ … It seems as if the dictionary has entered as many word combinations (compound words) of the headword as possible. Although this could be seen as a good practice, the question remains: Which of these are essential in a dictionary of this size? Of the word family 24 combinations are given (see p. 339), all of which are entered as headwords.
(3) family allowance, family bliss, family business, Family Division, family doctor, family estate, family feeling, family friend, family grouping, family heirlooms, family house, family income supplement, family influence, family law, family likeness, family lines, family man, family name, family planning, family responsibilities, family room, family saga, family silver, family tree.
Are these combinations sufficient? And could further ones not be added indefinitely? For example: This question is asked, not because the inclusion of indefinite word combinations is propagated, but because there seems to be too many superfluous entries.
By giving every lexical item as a main headword, too much valuable space has been wasted. Furthermore, the so-called "senses" of a headword are all moved to the left side of a column, each beginning on a new line regardless of whether the last line of the definition of a "sense" is completely filled or not. So entry words like get with 53 "senses" or make with 38 "senses" plus 23, the socalled "various connections", add up to 61 unfilled lines for make alone. The entry word take has a total of 128 cm of unused space. If one line of a column of the dictionary covers 6.5 cm this means that the unused space for the entry take equals 19 (i.e. 128 cm ÷ 6.5) lines. A random selection of two pages of the dictionary reveals that page 437 has 130 cm and page 928 has 108 cm of unused space which equals 17 lines per page. Statistically the dictionary has 17 799 (i.e. 1 047 pages x 17) unused lines. Translated into pages the unused space equals 59 (i.e. 17 799 lines ÷ 150 lines per 2 columns of a page = 118 columns ÷ 2) pages. So if all the space had been utilized, the dictionary would have been reduced to 988 (1 047 -59) pages.
The dictionary was made unnecessarily voluminous. Although by giving headword status to all derived, inflected and compound words, the compiler might have attempted to user-friendliness, this was done at the expense of wasted space whose cost the user will have to bear, because the bigger the dictionary, the more expensive it is.
It is not necessary to enter every combination a word can have but only a few combinations, especially those whose meanings are opaque, i.e. those whose meanings cannot be deduced from the meanings of the words that form them. Combinations like those of family given under (3) and even those of home listed under (5) should not all have been entered in the dictionary because the meanings of many can easily be deduced from the meanings of their component parts. The word home has 66 independent headwords which are either derivatives of home or compound words whose one member is home (see p. 461-462).
Compared with Kirkeby, TUKI (1996) has almost the same lexical stock. The only difference is that TUKI has been more prudent in selecting and arranging entries. As a result TUKI has fewer main entries because most of the word combinations are entered as run-on entries as shown by the following Kirkeby, on the other hand, lacks the following lexical items found in TUKI (1996) between Aa-advocate: (7) abb, ablate, abnegation, A-bomb, abscise, absentmindedly, absent oneself, absorption, abusively, abusiveness, abysmal, Abyssinia, academically, academicals, academician, accelerando, accidence, accordant, accretion, accrebities, acriflavine.
As regards the method of presenting entries which share a common basic form, TUKI treats all such items as run-on entries to the main entry word as shown in (8).
Suffixes ~able, ~ability and ~ance represent the verbal derivatives of accept, namely acceptable, acceptability and acceptance.

Lexicographical categories
After the selection and presentation of headwords have been looked at, the lexicographical categories given under these headwords can now be examined. The reference needs of foreign learners are basically two: receptive and productive, and any information given should try to meet these needs. The information usually found in a dictionary entry is of four types: 1. Internal information which presents facts about the lexicographical unit: the orthography, pronunciation and form of the headword. The form includes derivational and inflectional aspects as well as compound forms which make up one lexical item. 2. External information provides facts about its relationships with other lexicographical units. In this information, the syntax of a word is shown, especially the obligatory combinations. 3. Usage information gives facts about the way the lexicographical unit is used. This aspect deals with the context in which the entry word is used in ordinary linguistic discourse, revealing its grammatical and semantic meanings. 4. Etymological information provides facts about the history of the headword which includes the historical development from its origin to its current form, and the development of its senses. Kirkeby makes available the following lexicographical information for the headwords: orthography, pronunciation, grammatical information, usage labels which include status labels, subject or domain labels, regional labels and time labels. Other information categories are meaning and illustrative examples. Etymological information is not provided. Kirkeby does not disclose the function of his dictionary, but from its content it can be deduced that it has decoding and encoding, i.e. receptive and productive functions respectively. It also gives technical language from some specialized fields such as botany, zoology, physics, etc. All this information is condensed in a pedagogical bilingual dictionary for English learners. There is no doubt that the compiler wanted to meet all the needs of different categories of prospective users of the dictionary. But the desire to serve the diverse needs of these users makes the different categories of this dictionary virtually incompatible, often leaving the users very dissatisfied. The information categories can now be examined separately.

Orthography
The dictionary gives variant spellings of some headwords and variations in British and American English, e.g. haulier (US: hauler), and hardware shop (US: hardware store). This is normal practice in lexicography, and the examples have been treated properly except that the variants and the English American variations which are on a par with the headwords, are given in italics instead of the bold type used for the headwords. Many word combinations or compound words are entered as headwords. Most of them are written as two separate words while others are joined by a hyphen and some are spelt as a single word. For example: If the word combinations in (10)(a)-(c) are studied, it is unclear why the compound words are presented in different spellings. Why, for example, are mail sack, mail carrier and mail van written as two words while mailbag, mailbox and mailman as one word, especially when bearing in mind that mail sack and mailbag are synonyms as is the case with mail carrier and mailman, both meaning postman. Likewise the adjectives male-dominated and male-chauvinist are joined by a hyphen while the nouns male chauvinist, male dominance and main line are written as two words, and mainframe and mainstream are written as one word without a hyphen. Although some adjectival forms are joined by hyphens as can be seen in (10)(b), nominal forms like man-eater and man-hour are also joined by hyphens. Forms written as one word without a hyphen cut across all grammatical categories: adjectives (homesick and manhandled), nouns (homelover and honeybee) and verbs (manhandle and mainline).

Pronunciation
Pronunciation in a dictionary is rendered in phonetic symbols which represent the sounds of a language. Provision of the symbols is based on the assumption that the user has already mastered the sound inventory of the target language and is able to produce the acceptable sounds as they occur in familiar words.
Transcribing phonetic symbols to represent the sounds of a word is difficult to a non-phonetician, but to interpret them is much harder because many dictionary users do not have the background in theoretical phonetics needed to interpret the transcription system. Kirkeby uses the phonetic symbols of the IPA to represent the different sounds in English words, e.g. /A:/, /ae/, /ø/, /e/, /θ/, /ð/, /ŋ/, /Z/, /S/, etc. To be able to decipher these symbols one needs a pronunciation guide. Kirkeby's guide is not very useful in helping to pronounce the words. One would have expected that, since this is a dictionary intended to help Tanzanians whose "knowledge of English has greatly deteriorated" (see Preface), the compiler would have made a serious study of Swahili words in order to find suitable examples for producing certain sounds instead of referring to English, German or French words as shown in (11). (11) x in words like 'loch' and 'ugh' is pronounced like ch in the German word 'nach'. ~ above a vowel indicates nasalisation (only in some French words).
How can Tanzanian English learners be expected to use German or French to help them with the pronunciation of English words? One would even question whether a non-linguist Tanzanian can really make any distinction between the following set of vowels: /A:/ and /ae/, /e/ and /´/, /o/ and /ou/, or /u/ and /ø/ as in the following example: Likewise it is difficult for them to decipher phonetic symbols such as /ŋ/, /θ/, /ð/, /S/, /tS/, /Z/ and /dZ/.
A lexicographer intending to provide this information has to think of how best to help the users. Indeed he has to be sure whether dictionary users will actually need this information. Lexicographical research has shown that pronunciation is information that users look up the least in a dictionary. However, the guide would have been meaningful if examples of Swahili were used. For example: ng' as in ng'ombe could have been used instead of the phonetic symbol /ŋ/, th as in thamani instead of /θ/, dh as in dhana instead of /ð/ and kh as in khatibu instead of /x/. The Swahili phonetic equivalents would have been more meaningful to the Swahili learner than the IPA. Although this would have been breaking away from the tradition, it is userfriendly as far as the Swahili English learner is concerned. It should be recalled that the English dictionary had also used the respelling method before the IPA was adopted. This can be done if one is willing to involve consultants in the field. The native speakers, journalists and amateur lexicographers used as consultants by Kirkeby have limitations. When one looks at the pronunciation guide and the pronunciation symbols in the entries, it becomes obvious that the information was copied from another dictionary disregarding its relevance for this specific dictionary.
Further examination of the other symbols in the guide reveals some pho-nological information such as main stress, secondary stress and nasalisation which does not form part of the general knowledge of the anticipated users and cannot be deciphered by them. This is known only to students of phonology. Provision of a pronunciation guide requires the relevance of the symbols for the anticipated users, the users' ability to interpret these symbols and the lexicographer's innovation of the pronunciation guide to make it suitable for the users.

Grammar
Grammar is one of the important information categories in a dictionary, especially one designed for learners of a language.

Grammatical categorization
Grammatical information is more useful to the learner of a foreign language than to the native speaker. Pedagogical and bilingual dictionaries provide more grammatical information than those for native speakers (Landau 1984). Information given in a dictionary should be helpful to the user and the user should not be confused with obscure conventions. More pragmatic information is needed about how people communicate with each other and more attention must be given to the structure of the spoken language . Most of the headwords in this dictionary have been classified and labeled with their word-class markers, viz. noun, verb, adjective, adverb etc., but some headwords have not been marked: (13) tax arrears, tax avoidance, tax bill, tax burden, tax certificate, tax concession, tax consultant, tax collector, tax dodger, tax evader, tax expert, jack plug, jack point, jack tar, jack towel, jet engine The following items with similar forms have, however, been classified and labelled: (14) tax-exempt n, tax-finding n, tax-free adj, rag-and-bone-man n, jack-in-thebox n, jack-in-house n, jack-of-all-trades n, jet-propelled adj, jet-black adj There are no linguistic explanations that discriminate (13) from (14). Although one could argue that the examples under (14) (with a hyphen) tend to develop towards (becoming) one-word compounds, it could still be argued that the two sets (with or without a hyphen) do not differ because all of them are lexical units representing one concept just like single lexical units.

Subgrammatical categorization
Subgrammatical categorization is very important for a pedagogical dictionary. It is not enough to classify words into seven or eight categories as if words in each category behave the same morphologically and syntactically, especially the major categories viz. nouns, verbs and adjectives. Language learners need to be informed whether a noun is countable or uncountable and whether it has a plural and what the plural is. Users need to be told whether a verb is transitive or intransitive and what its conjugation forms are. Information is also needed about attributive and predicative adjectives or their comparative and superlative forms. Morphological aspects of lexical items such as inflections and derivational affixes is furthermore important information that should be given.
Kirkeby's dictionary lacks information about verbal conjugation in its entries. This information is necessary if it is to be used efficiently for production purposes. Kirkeby does indicate plural inflections for the irregular nouns, e.g. wife (wives), thief (thieves), but this has not been done consistently. Some nouns e.g. directory, fly, family etc., do not have their plural inflections indicated. Moreover, the compiler does not mark nouns which are countable and those which are uncountable, cf. furniture, sympathy, butter, water, etc. This information is useful to learners of English because it helps them to select the appropriate agreement between a noun and the verb.
As for the comparative and superlative forms of adjectives, the dictionary is very inconsistent. At some adjectives like big and bad this information is given while many adjectives such as fast, poor, pretty, quick, safe, tall, thin etc. lack this information. Even the irregular adjectives which do not take -er and -est but are preceded by more and most for the comparative and superlative forms such as beautiful, handsome, competent, etc. have also not been provided with this information. Even the so-called regular forms undergo morphological changes when the comparative and superlative inflections are suffixed to them, hence it is important to indicate them. For example: (15) tall, taller, tallest; safe, safer, safest; pretty, prettier, prettiest; thin, thinner, thinnest; beautiful, more beautiful, most beautiful; competent, more competent, most competent; little, less, least.
The adjectival examples show forms which take ~er and ~est; others take ~r and ~st or ~ier and ~iest while still others double the last consonant and then add ~er and ~est. Because all this information is important for the English learner, a pedagogical dictionary should supply it. Kirkeby completely neglected this information.

Swahili grammatical information
Kirkeby has attempted to enter some Swahili grammatical aspects in the Swahili equivalents of the English headwords. These are the plural affix of nouns. For example: (16) friend n rafiki (ma-); hand n kiganja (vi-) cha mkono The affixes in brackets after the Swahili equivalents (see rafiki and kiganja under (16)) are the plural affixes of the headwords. This means that the plural of rafiki is marafiki and that of kiganja is viganja. Legere (1990) has done the same in his dictionary. This is relevant only when dealing with single lexical units like rafiki and not with multilexical units like kiganja cha mkono because, when this lexical unit changes into plural every item in it is affected, as shown in (17).
If the compiler felt that it was necessary to show the plural form of the Swahili noun, he should have considered how best to deal with this type of lexical unit. Should all the plural affixes and concordial agreements for the headword be indicated as in (18)? (18) kiganja cha mkono n (vi-vy-mi-) Or should he rewrite the whole multi-lexical unit with its plural prefixes as in (19) if he feels that the affixes are not communicative enough?
(19) kiganja cha mkono n (viganja vya mikono) Alternatively, the plural prefixes can be eliminated and instead an elaborate front matter with a sketch of Swahili grammar, giving the classification of nouns can be included. Hopefully, with such an alternative, the user will consult the front matter for clarification when in need. What is important here is that any information in a dictionary should be communicative.

Word combinations and collocations
Syntax of the headword is usually shown through verbal phrases, collocations and illustrative examples. This information is important because it shows words found in combination with the headword. In Kirkeby's dictionary many verbal phrases, most of which are idiomatic in character, have been arranged alphabetically as headwords and translated. Because English is rich in verbal phrases and idiomatic expressions, many verbs have a significant number of these multilexical items. For example look has 30 verbal phrases such as: Whereas these word combinations are important in encoding and decoding, not all deserve a place in the dictionary because their combinations are neither obligatory nor have an idiomatic connotation. For example, look at, look across, look down, look into or look up are phrases that should not have been listed because the meaning of each phrase is the total sum of the meaning of the words that form the phrase. Indeed one could replace look with another verb and get the meaning of the two. For example: (21) wave at sby, walk across, go down, put into, climb up.
Although collocations and especially verbal phrases are important in a dictionary, Kirkeby should have selected and entered only those which have fixed meanings. Words which take certain prepositions, should be indicated, e.g. look (~ at, ~ up).

Illustrative examples
Illustrative examples are given to support and in some cases to supplement definitions. They help foreign learners to understand the meaning of a word. They serve to show the usual collocations or contexts in which each sense is used, thus providing a variety of grammatical information such as whether a verb takes an indirect object or is usually used in the passive voice, or whether an adjective is predicative or attributive. Examples are therefore given under headwords which have collocations that provide specific syntactical information or can help to distinguish different senses of a polysemous lexical item. Illustrative examples are also given to function words which have no lexical but only grammatical meaning. Familiar words are usually not exemplified because their meaning and usage is known unless they have developed new senses. For an example sentence to be useful, it must give a clear idea of the meaning (which has become polysemous). It must be short and to the point but sound authentic. Familiar words must also have few collocations and a fair amount of grammatical information. Kirkeby provides many examples of usage that cut across different word classes. For example:

Meaning
Meaning of an entry word in a bilingual dictionary is rendered by an equivalent lexical item in the target language. In the absence of exact equivalents in the two languages, glosses can be used as translation equivalents (Bartholomew and Schoenhals 1983). Thus the number of senses in the source language are usually reflected in the target language in the dictionary. Likewise, the phrases and fixed expressions in the source language are given their equivalent translations in the target language. In this section, the senses given as Swahili equivalents of the English headwords, the manner in which they have been presented and the degree of their comprehension to learners of English will be examined. The entry words in the dictionary have one or more than one Swahili equivalent. For example: (25) alcohol … n: alikoholi; kileo; pombe; kilevi; … alcohol abuse (= abuse of alcohol) unywaji wa pombe kupita kiasi. I. alcoholic … n: mlevi (wa-) sugu. II. alcoholic adj: -a kulevya; -a kileo; -enye asili ya kulevya.  (25) it can be seen that alcohol has only one sense which is translated by four Swahili synonyms: alikoholi, kileo, pombe and kilevi. The headword alcoholic n has only mlevi sugu while alcoholic adj has the synonymic equivalents: -a kulevya, -a kileo and -enye asili ya kulevya. The Swahili equivalent alikoholi unlike its other synonyms is not a common but a technical word, hence it needs to be marked in order to alert the user that it has a restricted usage. Although the compiler has extensively marked words with a special usage, the marking of such words has not been consistent. Moreover, alikoholi should appear last in the list of synonyms because it is a rare word. Common synonyms should appear first because they are the ones which users may most often want to look up in the dictionary. The presentation of Swahili equivalents has in some cases been preceded by synonymic paraphrases of the senses. This can be illustrated by alcohol abuse and the senses 1 and 2 of alert under (26). The paraphrases are meant as elaborations (glosses) of the headwords before the Swahili equivalents are given. With the help of these paraphrases, the reader is expected better to understand the headword in the source language before coming to the Swahili equivalents. This is probably possible for the speaker or learner who already knows English and understands fixed expressions such as wide awake or fully awake. A Swahili speaker learning English will find the paraphrases useless because he does not know what the fixed expressions mean. Even for the English speaker, 'wide awake' in 1 and 2 would not make any difference. Indeed, -liyemacho can be used figuratively as well. Moreover, the illustrative sentence in alert sense 2 (see (27)) is neither helpful to the reader nor a good example of a well-constructed grammatical English sentence.
The Swahili translation is also confusing. Do the phrases following 'mtu mwenye akili' qualify this phrase or are they different interpretations of the English sentence: 'he's an alert person'. Another feature in the entry alert which is also found throughout the dictionary, is the Even if the words in brackets were elaborating the headwords, such words are often more difficult than those being elaborated. The words celebrity and renown are no more common than the headword fame. Likewise silver heirlooms is more difficult than family silver and child benefit than family allowance. A critical examination of the Swahili equivalents reveals a number of shortcomings in the equivalents selected and the manner of presenting them. In the following section, the Swahili equivalents and how they have been presented will be examined.

Sentences used as Swahili equivalents
The main objective of a bilingual dictionary in rendering the lexical meaning of the headword is to provide equivalent words in the target language. Phrasal equivalents are accepted as translation equivalents in the absence of one-to-one lexical units. The significance of using a lexical or multilexical unit is that such items can easily be applied in translations from one language to another. Definitions of headwords are confined to monolingual dictionaries because their main objective is to paraphrase the headword in such a way that one can understand its meaning. This is in contrast with a bilingual dictionary where one is expected to know the meaning of a lexical item in one of the languages and only wants the equivalent word. Kirkeby has in many cases used full-fledged definitions as Swahili equivalents for English headwords. For example: (30) buffet 2. in restaurant or hotel: chakula (vya-) kilivyowekwa mezani ili walaji wajihudumie call loan mkopo (mi-) unaoweza kudaiwa wakati wowote call mark namba (-) ya kitabu inayoonyesha mahali kitabu kinapopatikana call meter mita (-) ya kuhesabu dakika katika matumizi ya simu call sign fungu (ma-) la maneno na tarakimu zinayotambulisha utangazaji redioni cable television huduma ya televisheni kwa njia ya kebo bila antena callboy mtu (wa-) aliye pembeni ya jukwaa ili kuwahimiza waigizaji chaps US: ovaroli za ngozi zivaliwazo na wachunga ng'ombe katika tambarare za Amerika There is no reason for such long descriptions which could have been shortened. If the language does not have words for the headwords in the source language, simple coinages would have been better. For example: (31) buffet: chakula cha kujihudumia call loan: mkopo wa kulipa haraka call mark: alama bainishi call meter: mita ya simu call sign: ishara ya kutangaza cable television: televisheni ya kebo callboy: mhimiza waigizaji chaps: US: bwelasuti la ngozi

Superfluous equivalents
The dictionary has some examples of good Swahili equivalents for certain English headwords. Unfortunately these are followed by further paraphrases of these shorter equivalents. In most cases such descriptive paraphrases are repetitive and long-winded.

Ambiguous or misleading Swahili equivalents
Some Swahili equivalents are ambiguous or misleading to users, not representing exactly what the headword denotes. They share some semantic features of the headword but cannot qualify to be the equivalent of the headword.
(33) hard currency fedha za kigeni hard palate the ~ kaakaa (ma-) la juu hard sell ugumu wa kuuza off camera nje ya kamera The Swahili equivalent of hard currency is not fedha za kigeni (foreign currency) but convertible currency, i.e. currency whose value cannot fall suddenly (Oxford Advanced Learner's Dictionary of Current English 1988: 568). The term fedha za kigeni was coined from the Tanzanian perspective, because foreign currencies such as the US dollar, British pound or European Union's euro are to the Tanzanians fedha za kigeni (foreign currencies). The term cannot be used in the same way by Americans, British or Europeans when referring to their own currencies.
To them the currency is not foreign currency, but hard currency because it is convertible.
Hard palate is a term contrasted with soft palate. In Swahili the equivalents of these phonetic terms are kaakaa gumu, and kaakaa laini respectively. It is therefore not correct for hard palate to be translated as kaakaa la juu (the upper palate) as if there is kaakaa la chini. Moreover, the dictionary lacks the term soft palate.
Hard sell means "difficult to sell". This is used when referring to goods which sell very slowly. The Swahili neologism cannot be ugumu wa kuuza because this is a literal translation of the words hard (ugumu) and sell (kuuza). The headword in English is sell and the qualifier is hard. The headword in the Swahili equivalent should have been kuuza and the qualifier gumu. So the Swahili equivalent could be kuuza kugumu which is better than kuuzika kwa shida.
Off camera is incorrectly translated as nje ya kamera, literally meaning "outside the camera" while it actually means "away from the camera" or "not to be photographed". An appropriate equivalent could have been bila kupigwa picha.
Other headwords which have not been given appropriate Swahili equivalents are given under (34).

Incorrect description of headwords
Some descriptions of headwords provide false information. For example: (35) drama drama; riwaya hair root kisiki cha nywele; mzizi wa nywele hair piece kipisi cha nywele I. Gambian 2. language: Kigambia Senegalese 2. language: Kisenegal Swahili has adopted drama as the equivalent of the English word drama also given as headword. The term riwaya was coined to represent the English word novel. It is therefore misleading to translate drama with riwaya as well.
Hair root cannot be defined as kisiki (stump) cha nywele (of hair) because kisiki is neither the equivalent of root in Swahili nor the appropriate translation of hair root. The appropriate equivalent should have been mzizi wa nywele. The translation kipisi cha nywele as equivalent of hair piece cannot be accepted because kipisi is not a Swahili word. Kigambia and Kisenegali as languages of Gambia and Senegal respectively are questionable. There are no such languages.
Many African countries were created and given names which do not conform with the name of the people and languages spoken in these territories, cf. Tanganyika, Kenya, Uganda, Sudan, Namibia, Mozambique, etc. This is in contrast with many European countries, e.g. Ireland (Irish), Spain (Spanish), England (English), etc.

Long descriptions as Swahili translation equivalents
The dictionary has some headwords with long descriptions which are appropriate for a monolingual dictionary but not for a bilingual dictionary whose objective is to help the reader to translate from a source language to a target language.
Compare the examples under (36)

Use of Swahili proverbs to describe headwords
Proverbs are fixed expressions with figurative meanings which cannot be deduced from the meanings of the words which form them. It is therefore unwise to use them to describe headwords because their meanings would not be understood. For example: flattery n: kilemba cha ukoka; … kupakwa mafuta kwa mgongo wa chupa be made feel big meat 5. colloq: easy meat (= an easy target) kuku mweupe These headwords could simply be described as follows: (39) flatter vb ~ sby sifu mno; sifu kupita kiasi praise falsely flattery sifa za uwongo, kusifu kupita kiasi false praise The glosses in (38) are correct except that they are proverbs which cannot easily be understood. Therefore equivalents in plain English as in (39) are proposed.

Unacceptable Swahili equivalents or translations
Some Swahili equivalents cannot be regarded as appropriate translations because they are ungrammatically constructed and do not accurately represent the meaning of the English headwords.

Miscellaneous errors
Some typographical and grammatical errors appear in the dictionary. There are also factual errors as well as omissions of some entry words. For example: (42) alert adj 1. (= wide-awake; fully awake) -liyemacho cajole vb raiai booking n 3. the player received a ~ (= the player was booked) mchezaji aliyeandika jina kwa kuvunja sheria chandalua n; (in East frica) (= mosquito net) chandarua half-bound adj; bookbinding liojadiliwa kwa ngoz disappointment n … a bitter ~ masikitiko makali; it was a great ~ ilikuwa jambo la kukatisha tamaa sana; a bitter ~ masikitiko makali; it was a great ~ ilikuwa jambo la kukatisha tamaa sana; ... her ~ was obvious from her face kukata tamaa kwake kulionyesha bayana usoni mwake Some typographical errors are the following: The equivalent of alert should be written as two words -liye and macho and not one word -liyemacho as given in the dictionary. The entry words above show examples of typographical errors such as raiai (instead of rairai under cajole), frica (instead of Africa under chandalua) and ngoz (instead of ngozi under halfbound). Under the entry disappointment two illustrative examples are repeated: a bitter disappointment and it was a great disappointment. Some grammatical errors include the following: For the wrong Swahili translation, e.g. the player received a booking mchezaji aliyeandika jina kwa kuvunja sheria (the player who wrote his name for breaking the rule), the proper translation is mchezaji aliandikwa jina kwa kuvunja sheria (the player's name was written for breaking the rule). Likewise the translation of the illustrative sentences: Her disappointment was obvious from her face kukata tamaa kwake kulionyesha bayana usoni mwake (his disappointment was shown clearly on his face) and it was a great disappointment ilikuwa jambo la kukatisha tamaa sana, should have been kukata tamaa kwake kulionekana bayana usoni mwake (his disappointment was obvious on his face) and lilikuwa jambo la kukatisha tamaa sana (it was very disappointing).
Omissions include the entries between headwords great and grow.
Usually errors in such a large-scale work are inevitable. However, some simply show that the compiler was not careful enough in constructing the entries and editing the dictionary manuscript. Given the time frame for this project, two and a half years, it is obvious that the work was done rather hurriedly.

Dictionary and standardization
A dictionary is a historical document and a standardizing tool. It records the language as it was used when it was first written and/or as it is used today. The dictionary also helps to standardize a language and show or teach the user the accepted (standard) form of the language if that has been determined. This explains why people always make reference to the dictionary whenever they are in doubt about the spelling, meaning or usage of words. A lexicographer should be able to determine and show the most accepted form of a word where there may be variants. Kirkeby, however, instead of determining the most accepted form and giving it as standard, has entered all variants: (43) cake n US: apple ~ keki ya tofa(h)a I. calm 2. naut: utulivu; ush(u)wari calumny n kash(i)fa calumnious adj -a kash(i)fa flight instructor mwalimu (wa(a)limu) anayetoa mafunzo ya kurusha ndege These entries show different variants of Swahili words: tofaha and tofaa, ushwari and ushuwari, kashifa and kashfa, and walimu and waalimu. Any Swahili scholar knows that walimu, tofaa, kashfa and shwari are the most accepted forms and the compilers of Kirkeby should have been aware of this and not introduced littleused variants. Since Swahili has dialects, variant pronunciations are inevitable. A pedagogical dictionary should aim at the standard form and avoid the inclusion of variants unless there is evidence that all variants are regarded as accepted standard forms, or unless the stated policy of the dictionary is to record all variants. Kirkeby's dictionary does not have such a policy.

Sense discrimination and arrangement
Senses of an entry word are usually introduced by numbers, letters or semicolons. Numbers show that the senses are more distant in relation to each other, while letters and semi-colons suggest that the senses have a closer relationship. Kirkeby uses numbers and semicolons to distinguish different senses of the headwords. Example (44) shows four senses of depress as found in Kirkeby.
The headword devil is said to have 16 senses but in reality it has only two, 1. and 2. in the example above. The rest (3. → 16.) are fixed expressions in which the headword is used. It is therefore misleading to indicate that the word has sixteen senses. It is important to differentiate senses of an entry word, and fixed expressions, idioms or proverbs in which the headword occurs. A dictionary can introduce the senses with numbers, and then give the fixed expressions, idioms or proverbs under a subtitle.

The use of labels in the dictionary
Labels are used in a dictionary to draw the attention of the user to the limited use of some words. The labels could be categorized into the following: (a) currency label which indicates the dimension of time, e.g. obsolete/archaic; (b) domain label which shows the field to which the word belongs, e.g. law/music; (c) evaluation label which indicates the speaker or writer's attitude, such as pejorative/appreciative; (d) figuration label which shows the type of meaning, e.g. literal/figurative; (e) regional label which shows where an item is mainly used (US/Brit. or Tanzania/Kenya); (f) register label indicates the manner of speech and writing in which the word is used, such as formal/informal; and (g) status label which shows whether a word belongs to the standard language or to a subset (slang/colloquial/dialect/offensive, etc.). Kirkeby applies labels to mark the usage restriction of words. These labels are important and useful to the user. However, they need to be presented consistently. When the headwords in the dictionary are looked at more closely, it becomes obvious that some words of the same status have been marked and others not. For example: (46) machine-cut engin: -liokatwa kwa mashine machining allowance engin posho ya kufanyakazi kwa mashine machine-finished of paper -liomalizwa kwa mashine machine-glazed of paper -liong'arishwa kwa mashine (liongarishwa) machine part kipuri cha mashine machine-readable -enye kusomeka kwa mashine mediocre derog. -a hivi. hivi, -a kadiri; sio-zuri sana mediocrity n: uduni; uhafifu; ukadiri social democracy polit: demokrasia ya jamii social democratic polit -a demokrasia ya jamii socioeconomic -a kijamii na kiuchumi social intercourse (formal) mwingiliano wa kijamii social ladder mfumo wa kitaba katika jamii Words whose domain seems to be easily deduced have been labeled, e.g. machine allowance or social democracy, but those such as machine part or social ladder whose domain is less easily determined, have not been marked. Words of the same status like machine-cut on the one hand and machine-finished and machine-glazed on the other have been treated differently. The former has been labeled but not the latter. It seems as if the compilers could not make a distinction between words with restricted usage and ordinary words. Almost every word has a domain, even the everyday words we use. Cf. cook, boil, fry, roast, bake, etc. (cookery); ugali, nyama choma, biriani, makande, ndizi choma, etc. (foods); dengelua, buza, mbege, tembo, kangara, chimpumu, ulanzi, gongo, etc. (drinks); or plane, gauge, Tsquare, square, mullet, hammer, screw driver, chisel, etc. (carpentry). Such words need not be labeled because they are ordinary words. Users know their domain and that they are no longer restricted to specific users. Kirkeby has also labeled words of this type. For example: (47) solar eclipse astr kupatwa kwa jua lip anat mdomo (mi-) meeting house rel jengo (ma-) la mikutano melon bot tikiti (ma-) maji Looking at the examples under (47), one is tempted to ask whether users really need to be told that solar eclipse, lip, melon and meeting house are specialized words with restricted usage. Actually, many of the labeled words in this dictionary are no longer specialized. Cf. (48): (48) machine accounting bookkeeping uwekaji mahesabu kwa kutumia mashine machine-cut engin -liokatwa kwa mashine machining allowance engin posho ya kufanya kazi kwa mashine social democracy polit: demokrasia ya jamii social democratic polit -a demokrasia ya jamii A general dictionary like this one should rather mark status, region, currency, figuration or register.

Discussion
Kirkeby's dictionary has most of the properties required of a dictionary. It contains more than 50 000 entries which are arranged alphabetically within 1 047 pages. It is indeed the first English-Swahili dictionary which provides pronunciation of the English headwords. However, the first major shortcomings of the dictionary are: its failure to identify (a) the users for whom the dictionary has been compiled, (b) the dictionary needs of that group and (c) their dictionary-using skills. The compilation of a dictionary of this magnitude should have been preceded by a limited survey to try and find out the information categories suitable for the expected users. After having examined different aspects of the dictionary, some of the questions raised at the beginning of this article can now be answered.
(a) Is the information required by the users present in the dictionary?
Although the purpose of Kirkeby is not stated, its structure shows that it is a pedagogical dictionary. It contains illustrative examples and grammatical categorization of the headwords. However, it lacks subgrammatical information necessary for the learner of English (see par. 4.3.2). As far as grammatical information for the learner of Swahili is concerned, it is not very useful either because even this is not fully rendered (see par. 4.3.3). As for the Swahili equivalents, the dictionary fails the user here too because many of the headwords are defined instead of being given equivalent lexical or multilexical units (see parr. 5.1 and 5.5). Furthermore proverbs have been used to define entry words (see par. 5.6).
(b) Is the information transparently accessible?
Although pronunciation is important for the user, its presentation in this dictionary is very cumbersome, especially for one who is not well versed in the IPA. Moreover, reference to German or French words to help pronounce English words as shown in the pronunciation guide does not assist a Swahili learning English. It seems that the compilers did not try to find out the dictionary-using skills of the expected users of the dictionary before giving lexicographical information through graphic symbols. As a result, a user cannot decipher these symbols to reproduce the sounds. The provision of pronunciation information considered as an advantage to this dictionary in comparison with other similar dictionaries, therefore reflects negatively on its design because it is information not easily accessible.
(c) How is the information presented?
The information is orderly presented except when other information inserted in between distorts the sequence of presentation. For example: (i) A paraphrase of the headword has sometimes been inserted between the grammatical/subject label and its equivalent: (49) merchant n (= wholeseller) mfanyabiashara ya jumla merchantable adj (= saleable; marketable) -a kuuzika, -a kununulika. merchant ship n (= merchant vessel) meli (-) ya bidhaa (ii) At times additional English illustrative sentences introduced by ex (meaning example) which are not translated into Swahili are inserted between the Swahili equivalent and another illustrative example: (50) meet: ~ the eye -ona (ex a bleak landscape met the eye); our eyes met macho yetu yalikutana measure n 1. kipimo (vi-) (ex a glass measure for holding liquids); common ~ kipimo cha kawaida This practice is contrary to the dictionary-making principle in bilingual lexicography which requires a sentence in the source language to be translated into the target language.
Sometimes such examples are not sentences but phrases which indicate the context in which the headword can be used. This juxtaposition of extra information overloads the entry to such an extent that it becomes cumbersome for the user to decode it and thus to use the dictionary efficiently.
(iii) As a rule, senses of a headword are usually introduced by a number, and users can easily tell the number of senses of a word. In this way, they can scan through the entry to find the sense they want. Kirkeby's dictionary applies this principle, but mixes senses with phrases. Distinction is not made between senses of a word and word combinations/fixed expressions, hence all are introduced sequentially by numbers without differentiating senses from word combinations/fixed expressions (see par. 8).
The quality of the information presented and the effectiveness of the presentation of the information can now be critically examined. The work has some typographical errors (see par. 6). The compilers failed to determine forms which are more accepted than others, especially where there are variants (see par. 7) or headwords with similar characteristics (compound words) spelt differently, as one word, two hyphenated words or two separate words (see par. 4.1). The dictionary also contains inaccurate information (see parr. 5.3, 5.4 and 5.7). These all show that the compilers often did not seriously attempt to collect correct data.
The omission of the entries between the headwords great and growth is considerable: about 80 entries if the missing entry words are counted in TUKI (1996), but compared with Oxford Advanced Learner's Dictionary of Current English (1990) the omitted words are 180. The incomplete grammatical information (see parr. 4.3.2 and 4.3.3) is also a serious shortcoming because this information category is important for a productive dictionary. Very little is known about the consultants and advisers of this dictionary because it is just a replica of an English-Norwegian dictionary. But for the Swahili part, the consultants and advisers are identified as two native speakers of the language, a journalist, a school teacher, a banker and a computer expert. The compilers should have consulted more people than these, especially linguists and lexicographers.

Conclusion
Having been compiled according to the principles of lexicography, the dictionary is a good contribution to the development of Swahili. However, its quality and effective presentation of the lexicographical categories show some flaws. The collection and processing of the data have not been carried out carefully and rigorously. This is partly because it was done in a limited timespan and without qualified experts. This work, if improved, can be a good dictionary for the learner of English.