1 The concept of reference guide

A review is a summary with critical evaluation, so that at least the evaluation component depends on standards that are specific to the genre. A clear example of the difference in evaluation criteria is the review of a research monograph versus a textbook. Whereas for a monograph the main object of evaluation is the theory presented, for a textbook criticism of the theory introduced by it would be misdirected. As reference guides are not an established genre we first have to determine which criteria can be used for their evaluation.

Obviously, a reference guide is neither a research monograph nor a textbook. A better starting point for a comparison is a handbook. Handbooks constitute a type of publication that is increasingly frequent, perhaps because when an edited volume is labelled a handbook, academic libraries are more likely to feel obliged to buy it. As a side effect, handbooks become ever more specialized and it is more than likely that at some point we will see a Handbook of English morphology. In such a handbook, we would expect to find an overview, not of English morphology, but of theoretical approaches to English morphology. In a reference guide the balance between these is reversed.

It is therefore more reasonable to think of a reference guide along the lines of grammars and dictionaries. What they have in common is their priority to present data about a particular language. Most of the users of grammars and dictionaries are not linguists. However, linguists are heavy users of both and they are expert users. Perhaps the main difference between a reference guide to English morphology and grammars and dictionaries of English is that non-expert users are much less likely to use the reference guide.

A problem that grammars and dictionaries share, and that is more prominent when we consider expert users, is the relationship between the underlying theory and the presentation of data. This problem is equally relevant for a reference guide on morphology. It is a generally accepted insight in the philosophy of science that observation is inherently theory-dependent. Which of the possible observations are worth retaining as an actual observation depends on the theory we have. Moreover, in presenting the data, they have to be organized in some way. This organization adds another layer of theory-dependence.

Another problem that a reference guide to morphology shares with grammars and dictionaries is that they take as their subject (an aspect of) a particular language. As Chomsky (Reference Chomsky1980) observes, languages in the sense in which English is a language are not empirical objects. A speaker's knowledge of language is an empirical object as are texts and utterances, but they are not reflections of an object corresponding to English. English as a language only arises because we classify speakers as speakers of English and texts or utterances as being written or spoken in English.

The reference guide under review has to address these issues at least at the practical level. The authors (abbreviated as BLP from now on) have to select and group the data, which can only be done by using a theory, and determine the limits of what they consider English. We will come back to these issues after a brief sketch of the background and structure of the work.

BLP delimit the scope of their reference guide so as to include both inflection and derivation. There are not many other examples of reference guides, but they tend to focus on word formation only. For English, the most notable example is Marchand (Reference Marchand1969). Marchand's book is not called a reference guide, but the label would not be out of place. It still serves as the first reference work for individual affixes in English, although its theoretical outlook and data collection methods are now obsolete. Especially in his description of compounding, Marchand uses a modified version of Lees's (Reference Lees1960) transformational approach that was no longer pursued even in the 1970s. Especially for the description of affixes, Marchand's source is the Oxford English dictionary, which at the time was still the first edition in ten volumes completed in 1928 and the supplementary volumes.

BLP's reference guide is not an updated version of Marchand (Reference Marchand1969), but represents a conceptually different approach to the subject matter. This is by no means a necessary consequence of the problems that make Marchand's work look dated to current users. De Haas & Trommelen (Reference de Haas and Trommelen1993), in their reference guide to Dutch morphology, are inspired by Marchand (Reference Marchand1969) and maintain many elements of his approach, while removing the methodological components that are outdated.

The innovations that BLP bring to the genre of the reference guide are, first, that they treat inflection as well as word formation, second, that they make systematic use of large corpora to check various observational hypotheses, and third, that they provide a more systematic than alphabetical presentation of the data. With reference to the last point, if Marchand (Reference Marchand1969) is like a dictionary, we might consider BLP's guide as more of a thesaurus.

2 Overview of the book

The reference guide is divided into six parts, together containing twenty-nine chapters. Part I is the introduction and contains four chapters. Chapter 1 (pp. 3–6), ‘Aims and structure’, is the actual introduction. Chapter 2 (pp. 7–39) gives a preliminary discussion of the terminology used in the reference guide, presenting the terms in relation to each other and giving references. Chapter 3 (pp. 40–5) makes the point that corpora were used as a basis for the data, in particular the Corpus of Contemporary American English (COCA; Davies Reference Davies2008–). Chapter 4 (pp. 46–57) is about orthography. Compared to the introduction of a typical dictionary, this one is long. Introductions are the most underused parts of dictionaries and it is quite likely that these chapters will have a similar fate. Not all of the material in the introduction is entirely relevant to the rest of the volume. Thus, more than half of chapter 4 is devoted to the orthographic representation of vowel length. Although interesting, because of its overview of the historical development of orthography and pronunciation, the relation to morphology is not always clear.

Parts II, III and IV are devoted to inflection, derivation and compounding, respectively. They constitute the heart of the reference guide. Part II is divided into chapters on verb inflection (pp. 61–102), adjective and adverb inflection (pp. 103–20), noun inflection (pp. 121–49) and function words (pp. 150–8). For a language as poor on inflection as English it is surely remarkable to find almost a hundred pages of discussion of inflection. There are very few issues (if any) that are not discussed here in detail. In chapter 5 on verbs, for instance, we find discussions of the different semantic, phonological and etymological properties of regular and irregular verbs as contrasting sets, different vowel pattern alternations, and contraction patterns of auxiliaries and infinitival to, as well as a twelve-page list of irregular verbs.

As might be expected, part III, on derivation, is the most extensive. It takes up 267 pages and is divided into ten chapters. In this part, the thesaurus nature of the work, as opposed to Marchand's (Reference Marchand1969) alphabetical treatment, is most prominent. Chapter 9 (pp. 161–94) is devoted to the interaction of phonological processes with derivation. It looks at allomorphy, stress effects, etc., in a general way, so that they do not have to be repeated for each derivational process. In chapters 10–15, the syntactic category of the output is the first organizing principle. There are three chapters on nouns, then one each on verbs, adjectives and adverbs. Among the chapters on nouns, the division is determined semantically. Chapter 10 (pp. 195–215) treats the formation of nouns expressing event, state, or result, chapter 11 (pp. 216–44) the formation of nouns designating a person or other participant in an event, and chapter 12 (pp. 245–66) the formation of abstract nouns, including names for qualities and collectives. These chapters do not cover all derived nouns. This difference is even clearer in the chapters on verbs (pp. 267–87), adjectives (pp. 288–321) and adverbs (pp. 322–32). The reason is that the last three chapters of this part are devoted to semantically determined output categories and they take priority over the chapters ordered by syntactic output category. Chapter 16 (pp. 333–53) covers locatives, both temporal and spatial, chapter 17 (pp. 354–84) negatives, and chapter 18 (pp. 385–427) expressions of size, quantity and attitude, including diminutives. This means, for instance, that un- and in- attached to adjectives to express the opposite of the quality designated by the base are not treated in chapter 14 (derived adjectives) but in chapter 17 (negatives).

With the exception of chapters 15 (adverbs) and 18 (size, quantity and attitude), all chapters presenting derivational processes have the same structure. They consist of three sections. Section 1 (‘Prospectus’) gives a brief overview of the issues treated in the chapter; section 2 covers the formal and section 3 the semantic considerations for the entire set of data treated in the chapter. This means that for a particular affix, the questions of which sets of bases it attaches to, how productive it is, which allomorphs it has, and how it affects the stress pattern and other aspects of the base it attaches to are not discussed in the same place as its meaning. One might see this as a disadvantage, because the information about an affix is divided between two places. However, at least for some aspects, it can be argued that it also offers advantages. Affixes with the same output category and the same broad area of meaning are often in competition. The reference guide gives some very useful overviews of this competition in tables. Thus, table 11.1 gives an overview of the nominalizing suffixes and conversion expressing thematic roles such as agent, instrument, patient, etc. It is only to be deplored that there is no list of tables in the volume, so that the only way to find such tables is by leafing through the entire chapter. Another problem is the absence of gridlines. Thus, table 14.1 gives 11 columns and 22 rows for suffixes. Without gridlines, it is often almost impossible to track which row corresponds to a particular tick, especially because the absence of a tick results in a blank and there are many blanks in the table.

The different structure of chapters 15 and 18 is motivated by the different range of issues they deal with. Chapter 15 on adverb formation is mainly concerned with the questions of whether adjectives and adverbs are different categories and whether adverb formation is inflection or derivation. Chapter 18 is much more heterogeneous. It treats diminutives, attitude (anti- and pro-), quantification (e.g. -ful and re-) and cardinal and ordinal numbers. Rather than organizing each of these sections into formal and semantic considerations, BLP organize them by affix.

Part IV is devoted to compounding. Here we return to the pattern of separating formal and semantic considerations, but this time they are each assigned a chapter of their own. Chapter 19 (pp. 431–62) treats three issues, first the delimitation of the domain of compounding, then the internal structure of compounds, and finally a typology based on formal properties. Chapter 20 (pp. 463–90) deals with semantic considerations. It starts with an unusually long prospectus that motivates the structure of the chapter. The chapter treats a smaller set of compounds than chapter 19, restricting its scope to those compounds in which the meaning of the components plays a role in their interpretation. The main distinction made among these compounds is between what is called argumental and non-argumental compounds. In an argumental compound, one component can be interpreted as an argument of the other one. The division cross-cuts with the one between endocentric and exocentric compounds. The boundary between argumental and non-argumental compounds depends of course on the definition of argument, which brings its own set of borderline cases. Finally, blends, neoclassical compounds, phrasal compounds and reduplicative compounds are briefly discussed as minor categories of compounding.

Part V describes interactions between different processes. It consists of three chapters. Chapter 21 (pp. 492–508) discusses combinations of affixes, chapter 22 (pp. 509–17) affixation on compounds and phrases, and chapter 23 (pp. 518–30) processes that should rather be described as paradigmatic than syntagmatic. This part is quite short, with the three chapters together shorter than some of the chapters in part III. The chapters each have a character of their own. Chapter 21 gives brief discussions of different types of affixes that may be combined. Chapter 22 gives a classification of each affix as to whether it combines frequently, occasionally, or never with compounds and with phrases. The data are presented as tables with examples and followed by a brief discussion of factors that may influence the behaviour. Chapter 23 introduces analogy first theoretically, then as a factor in inflection, derivation and compounding, and finally as the central component of splinters, for which three pages of exemplification are given.

Finally, part VI is entitled ‘Themes’. It consists of six chapters devoted to theoretical issues. Chapter 24 (pp. 532–44) is devoted to the distinction between inflection and derivation. BLP argue that in English it is not difficult to maintain the distinction, but it does not contribute to understanding English morphology. As we cannot help but observe, of course, the distinction nevertheless provides one of the principal divisions used in this reference guide. Chapter 25 (pp. 545–67) presents and evaluates various views of conversion. BLP propose to maintain the term conversion but see it as a prototype concept with more and less typical instances rather than a precise delimitation. In the reference guide, conversion and affixation are treated together in the part on derivation. Chapter 26 (pp. 568–82) discusses the cluster of phenomena that involves competition, blocking and productivity. BLP show that competition can develop over time or remain stable and that blocking is at most a weak factor because it has so many counterexamples. For productivity, they adopt a division into availability and profitability and note that the latter can be measured in various different ways. Chapter 27 (pp. 583–615) discusses the stratification of affixes and bases into different levels. BLP provide twenty-six pages of tables with combinations of affixes with stems and with other affixes to demonstrate that the original idea of two levels of affixation is not supported by the data. Chapter 28 (pp. 616–27) looks at English morphology from a typological perspective. BLP argue that broad classification schemes of languages into, for instance, isolating, agglutinating and inflectional offer little insight. Instead, they propose a finer-grained approach comparing English inflection, derivation, conversion and compounding separately to French and German. Chapter 29 (pp. 628–42) considers how the corpus data collected in the course of compiling the volume can be used to settle theoretical claims. BLP argue that several claims such as the Righthand Head Rule are refuted by these data. In conclusion they list five claims they call ‘uncontroversial’ and four somewhat less uncontroversial claims that are supported by the data they found while compiling the reference guide.

3 Corpora and the nature of English

One of the questions that comes up persistently in a work such as this reference guide is how the authors interpret the nature of English. Traditionally, language as a non-countable noun has been opposed to a language. However, Chomsky's (Reference Chomsky1965) distinction between competence and performance does not leave any room for an empirical object corresponding to named languages such as English. As explained in ten Hacken (Reference ten Hacken2007: 274–81), Chomsky only discovered this consequence in the course of the 1970s. The fact that English is not an empirical object does not mean that we can no longer say, for instance, that this review is written in English. It does mean, however, that there is no empirical way to determine whether an expression belongs to English or not.

Both competence and performance are empirical objects. Competence is the sense of language as found in a speaker. It is implemented in the speaker's brain, but not in a way that we have direct access to it. The most direct way we have that gets us information about the competence is by means of judgements. Performance is the sense of language as found in utterances and texts. Performance does not directly reflect competence, but rather the interaction of competence with other knowledge components in specific situations, filtered through a channel (e.g. acoustic or digital) that can introduce errors.

A corpus is a large collection of performance data. The size is often thought to even out any bias to individual speakers. Still, a corpus does not select itself. Decisions are involved in determining whether to include certain texts or utterances. Whereas it might be thought that in principle it is possible to find examples of rare phenomena by increasing the size of the corpus, there remains a problem with performance errors. Errors occur. The only way to remove them is to let somebody's competence override the corpus data.

As explained in section 1, BLP's reference guide can be seen as the equivalent for morphology to what an English dictionary is for the vocabulary. As I argued in ten Hacken (Reference ten Hacken, Fjeld and Torjusen2012), a dictionary such as the OED cannot be ‘the definitive record of the English language’ as it claims to be on its website. This is not due to any flaws in the dictionary, but because there is no empirical entity corresponding to the English language. Instead, in accordance with the analysis in ten Hacken (Reference ten Hacken2009), I propose to reinterpret dictionaries as tools. Speakers can use the information in the dictionary to support their competence in solving certain problems relating to language. For BLP's reference guide, the only difference to dictionaries is the nature of the problems and, as a consequence, the typical nature of the user.

In their introduction, BLP present as their aim ‘a comprehensive, up-to-date overview of contemporary English morphology that is based on large collections of data that reflect the usage of words by contemporary speakers of English’ (p. 3). The emphasis on corpora, and in particular on the Corpus of Contemporary American English (Davies Reference Davies2008-), sometimes almost gives the impression that BLP believe that there is an empirical object English, of which the corpus is a faithful representation.

An example where this comes up is the variability of irregular verbs. In their introduction to the list of irregular verbs, BLP (p. 90) take a very cautious stand, stating that ‘[w]e make no claim about the status of the forms we list beyond that we believe that they are used’. Their table has a column ‘Also regular’, which is marked ‘Yes’ not only for well-known cases such as burn and kneel, but also for, for instance, come, drink, give, go, know, make, sell, teach, tell, throw. In the preceding discussion, BLP (p. 76) note that knowed has ‘hundreds of attestations in COCA’. Presumably, the basis for the other decisions recorded in the table is similar.

In my search in COCA (14 Sept. 2014), I found 182 instances of knowed as opposed to 138,390 instances of knew. This surely shows that knowed has been used in the material that Mark Davies and his team decided to include in COCA. However, knew is 760 times more frequent than knowed and the comparison is biased against knew because part of the occurrences of knowed are for the past participle. The BNC only gives 2 instances. The important question is whether (certain) speakers of English have knowed in their competence. This is difficult to find out, because the norm is so strong that many speakers will base their explicit judgement on the norm rather than their competence. Moreover, the question of who is a native speaker of English is not an empirical one. It depends on a judgement that in borderline cases is quite difficult to make in a consistent way.

The treatment of knowed by BLP and the data from COCA indicate the general approach BLP take to the issue of what I would call norm-violating corpus data. It should be noted that knowed is much more frequent than any of the other nine verbs listed above. For gived, COCA gives 6 occurrences, for maked 3, and for drinked and selled only 1. The BNC gives at most 4 occurrences for any of the regular past-tense forms for the ten verbs listed above. With such numbers, it is far less unlikely that the occurrences reflect performance errors. In any case, this treatment of irregular verbs makes the reference guide inappropriate for a user interested in finding out what is accepted or widely used in English.

Whereas in the case of irregular verbs, BLP present data that conflict with the norm of English, in other cases, they use corpus data to argue against theoretical statements from the linguistic literature. In chapter 29, seven often-repeated theoretical generalizations are discussed briefly (five pages together) and refuted on the basis of data mostly presented earlier. An example is the constraint that verbal or synthetic compounds cannot take the subject as the non-head. This claim, though already refuted by Allen (Reference Allen1978), who gives examples such as population growth, has been repeated for decades, so it is good to find it rejected here again. A case that is treated in much more detail is level ordering, to which all of chapters 22 and 27 are devoted. In chapter 22, BLP give corpus data showing that many affixes can attach to compounds and phrases and in chapter 27 they give corpus evidence against a stratification among affixes. The dependence on corpus queries is particularly visible in section 22.4, when BLP do not want to draw any conclusion as to the influence of lexicalization (i.e. an expression being item-familiar) as a factor in affixation to compounds or phrases, because there is no obvious way to extract the full set of relevant examples from the corpus.

Overall, BLP use the corpus in a way that is reminiscent of its use in lexicography. Sometimes it looks as if the corpus is seen as the authority, which of course neglects the fact that the composition of a corpus is the result of conscious decisions that are not (and cannot be) determined entirely by the data. In other places, this image is somewhat corrected. Thus, when discussing corpus examples of double comparisons (e.g. most beautifulest), BLP (p. 107) state that ‘the difference, if any, between double and simple comparison for those speakers who use both is not clear’. Here at least the speaker's competence is given its due weight.

4 Theory and the presentation of data

It is well known that observations are necessarily driven by theory. Even if we do not consciously refer to a particular theory, we use existing knowledge and expectations of what is interesting to make a selection of the possible observations. Against this background, it is reasonable for BLP to declare they aim to make their reference guide ‘as theory-neutral as possible . . . but . . . theoretically informed’ (pp. 3–4).

The main reason to aim for a theory-neutral presentation is to make the data immediately useful for researchers working with different theories. The degree of perceived theoretical bias depends, of course, on the distance between one's own theory and BLP’s. The fact that there are three authors reduces the potential for idiosyncratic viewpoints, even though their theoretical backgrounds are very similar. As a reviewer, I will try to resist the tendency to highlight my own theoretical positions here, but I would like to point out some possibly controversial decisions. I will not discuss part VI here, because it overtly addresses theoretical questions directly, but concentrate on the presentation of data in parts II to V.

An obvious case of a theoretical viewpoint adopted in the presentation of data is found in the discussion of stress assignment in derived adjectives in chapter 14 (p. 296):

Another case where an approach from one of the authors is adopted is in chapter 21 on the combination of affixes:

In fact, chapter 21 is a summary of Plag's results supplemented with results obtained by applying the same method in the course of compiling this reference guide.

A different type of bias can be found when decisions on how to present data are based on assumptions that seem to be far from obvious or necessary. In chapter 5 on verb inflection, the verbal paradigm is presented as having six forms in English (p. 62). Only be has more forms. In fact, two of the six forms are identical for all verbs except be, viz. ‘non-past non-3rd sg’ and ‘infinitive/subjunctive/imperative’. There are no reasons based on the data discussed by BLP to separate these two forms.

In the same chapter, when discussing the extensibility of the class of irregular verbs, BLP (p. 72) state that ‘[c]laims that it is a closed class are certainly too strong, as new verbs can in principle enter the irregular class by compounding, derivation, or back-formation’. BLP's claim depends on the definition of closed class in morphology. If we assume that word formation rules may propagate the inflection class of the base to the output, the alleged counterexamples can be explained as instances of inheritance. This is surely not the same phenomenon as new instances arising through other means. Thus, there is no verb formation process assigning its output to a class of irregular verbs independently of its input.

A point that continues to mystify me, although I have seen it stated elsewhere, is BLP's (p. 124) classification of linguistics and other such formations as plurals. To me, *linguistics are is simply ungrammatical. COCA lists 7 occurrences of the string, but all of them are in contexts of the type ‘researchers in linguistics are . . .’, where it is not the plural number of linguistics that triggers the plural number of the verb. For *these linguistics, COCA does not give any instances. Given that for lexical items like linguistics there is no morphological contrast with a singular form, they trigger no plural agreement, and their meaning is that of a field of knowledge, calling them plural seems rather odd.

In presenting data, theory can act as an enabling factor, making it possible to come up with the right questions leading to a systematic overview, but it can also create blind spots, making certain insights and perspectives inaccessible. Needless to say, there are many examples of the former in the reference guide. However, I would like to point to two aspects that seem to be neglected as a result of the general theoretical orientation chosen by BLP.

It is generally known that word formation does not create words in a vacuum. We do not produce a new word and then start thinking of what it might mean. Instead, words are formed as a way to name new concepts. These concepts occur in a particular communicative context and the names chosen for them are meant to be understandable to the recipient of the communication. This insight is often neglected in generative morphology, which instead takes the form of the word as its starting point. In BLP's (pp. 29–32) section on ‘[t]he historical development of individual words’, this is obvious. I do not argue that it is always wrong to start the analysis from the perspective of the word, but in some cases the onomasiological perspective, starting from the concept, gives more insight.

In section 17.3.4, BLP discuss competing negative prefixes and note that there are sometimes two or more negative adjectives for the same positive one. They use this observation as an argument against blocking. What an onomasiological perspective would predict is that we find on the one hand synonyms, where two (or more) forms are in use because people do not know the other existing one(s). This should happen for negative adjectives that are geographically constrained or generally rare. On the other hand, we will find contrasting meanings. Here, a new concept needed to be named and a word that was not in use for another concept was chosen for it. The description by BLP seems to be in accordance with this prediction, but they do not frame the question in onomasiological terms. This makes it look as if a different constellation would have been equally possible.

Another type of data where an onomasiological perspective is helpful is cases where an intermediate form is not attested. BLP (p. 226) assume that there is a suffix -arian, even though most of the relevant cases can be analysed as combining -ary and -an. In this way, authoritarian can be derived from authority directly, because authoritary is unattested. (BLP use a hand symbol to mark unattested possible words.) In an onomasiological perspective, such cases are entirely straightforward. There is no obligation to go exactly one step further in a derivational path and then lexicalize the result. In the case of authoritarian, the concept to be named required the application of two word formation rules on the base authority. The non-occurrence of authoritary can be explained because there was no need for this word.

While in some cases, an explanation along such lines may give the impression of circularity, a good example of the way this perspective avoids rather awkward problems is the formation of decaffeinated. BLP (pp. 502–3) discuss the formation of the verb decaffeinate and suggest three possible analyses. In [[de-caffein]-ate], we would have a structure that is not supported by the meaning. In [de-[caffein-ate]], the meaning may be reasonably transparent, but there is no verb caffeinate. Finally, [de-[caffein]-ate] would be a parasynthetic formation, which is not allowed in a binary branching framework and rather undesirable if the two affixes have meanings in isolation that can easily be used in a paraphrase of the combined verb. In fact, I would suggest that decaffeinate is a back formation from decaffeinated. COCA gives 100 occurrences of the adjective as against 2 for the verb, BNC has 26 against 1. What happened was that a name was needed to describe the property of a coffee or tea from which caffeine had been removed. The availability of de- and -ate and -ed was exploited to coin the adjective decaffeinated. There was no need for any intermediate stages to be lexicalized in the process, but once the adjective was there, the verb was already latent and could be used without giving the impression of creating a new word. The idea that word formation is necessarily the application of a single process to an existing word generates problems that simply do not arise in the onomasiological perspective.

Another apparent blind spot in the theory adopted by BLP is the competition between word formation and borrowing. This is particularly obvious where BLP assume a bound stem which typically occurs in a single formation. There are so many examples of this that I will just select a few. BLP give ornament as an example of ‘-ment on bound roots’ (p. 198), grotesque as ‘-esque found . . . on [a] bound base’ (p. 293) and bracelet as an example of a specialized meaning of the diminutive suffix -let (p. 387). To my mind, and according to COED (Reference Stevenson and Waite2011), they are rather borrowings from French. To what extent current speakers of English would analyse them is doubtful. It seems that BLP want to maximize the domain of word formation, even beyond the limits of where it can provide insight.

A case where the contrast between borrowing and word formation seems more systematic is the opposition between -ar and -al. BLP (p. 298) give a phonological generalization, based on the occurrence of an /l/ in the base, which favours -ar especially when the /l/ occurs closer to the end of the base. Their counterexamples (e.g. clausal) and doublets with different meanings (e.g. familiar vs familial) point to another possible explanation. Most of the examples with -ar involve a bound stem. It is therefore quite plausible that -ar is not the reflection of any English word formation process, but rather a frequent component of borrowings from Latin or French. BLP do not mention this possibility, probably because their theoretical background does not cater for such an analysis.

5 Accessibility of information and references

A review of a reference work is not complete without an evaluation of the means by which the information included in it is made accessible. In this respect, BLP's reference guide is rather disappointing. As mentioned in section 1, BLP take an approach that can be characterized as more typical of a thesaurus than of an alphabetical dictionary. A good thesaurus has an index part that is about the same length as the main part with the grouping of words. In this case, however, the way in which access to details of the information in the reference guide is provided is more typical of a research monograph than of a reference work.

The table of contents only lists the six parts and twenty-nine chapter titles. There is no place where the internal structure of a chapter can be seen at a glance. The chapters are often structured up to the fourth level, e.g. 6.6.2.1, but the only way to retrieve the structure is by leafing through the entire chapter. This is a pity, because the section headings would provide a welcome and rich additional source of access to the content.

There are a number of insightful tables in the volume. Their presentation could be improved by adding gridlines and making sure the labels are properly explained (thus, on p. 290, table 14.1 has ‘bb’ without any explanation, presumably meant as bound base). The main problem, however, is that the only way to find them is to stumble across them in leafing through the chapters, because no overview of tables is given in the volume.

There are three indices, one for affixes, one for names, and one for subjects. The index for names is more typical of a research monograph, but there is no harm in that, of course. In the index for affixes, we often find unstructured entries of such length that their use is much reduced. Are we expected to check all forty-four page references for -ation or guess which ones are relevant? Among the affixes listed under a alone, we find eight with more than twenty page references. No typographic means is employed to guide the user to the most important ones.

Some of the cross-references are less than optimally helpful. In chapter 9, BLP (p. 190) refer to chapters 18, 19 and 20 for a detailed analysis of clippings, hypocoristics, blends and expletive insertion. These chapters do treat the phenomena, but only in very specific sections. Section 18.2.3 is on hypocoristics, 18.2.4 on clipping, 18.2.7 on expletive insertion, and blends are treated in 19.4.8 (form) and 20.5.1 (meaning). In the absence of a detailed table of contents, at least references to the individual sections could have been expected.

There are also some indications of incomplete adaptation of the structure to modifications. BLP (p. 267) state ‘there are three other prefixes that derive verbs: a-, be-, for-, en-’. Then they give examples for only three of the four. Section 18.2.2.2 is headed ‘Non-native affixes: -ette, hypo-, mini-, micro-, nano-’, but hypo- is only discussed in section 18.4.8. This is in fact the seventh subsection of 18.4, because 18.4.5 is skipped.

A moot point is the question of the selection of references. Obviously, BLP refer to many of their own works and those of their collaborators. This is not a point of criticism, because if they were not such prominent morphologists they would surely not have been asked to compile this reference guide. It is also obvious that not all publications can be known to BLP and that they have the right to select among them. It is strange, however, not to find a reference to Mühleisen's (Reference Mühleisen2010) book-length study of -ee, whereas Barker (Reference Barker1998) is called ‘the most thorough analysis of the semantics of this suffix to date’ (p. 237).