Multi-access keys are very appropriate for flower, fruit and other seasonal and other optional data, which may be missing when the plant is observed in the field.
An extra dimension for diagnostic information can be added to books (or bound sets of cards) by colour coding or by using simple marks or icons on the edges of pages in a book. In Ribeiro et al (1999) the pages have such marginal marks and the individual pages hold simple multi-access keys, with the leading questions also colour-coded in the centre of the page spread and photographs inserted around the page centre. This
Identification 105
Source: Hawthorne (1990)
Figure 5.1 Simple multi-access key to groups based on latex (top row) and leaf form arrangement of colour-coded information can take a significant time for the novice user to digest, and although more stylish than simpler keys, such use of colours is not necessarily easier to use than straightforward black-and-white indented or tabular keys, especially in the dim light of a rainforest understorey. Full-colour books are also much more expensive than guides with many non-colour pages.
Apart from the simplest types of table and few-species guides (for instance, Westfall et al, 1986), multi-access keys are perhaps most suitable where they are in a dynamic format – that is, where they can be mechanically manipulated or are electronic. The
106 Plant Identification
simplest dynamic keys are a small step beyond static keys. For instance, in a static, circular multi-access key, the arrangement can be made dynamic in the sense that the key has moving parts, like a slide rule (see circular key in Johns, 1978). Indeed, simple ‘identification wheels’ may have a sliding cover with a window in, whose only function is to hide information except for one species at a time (see Table 5.1i).
Most users can easily learn to use a basic indented key; ‘clever’ designs, and multi-access keys with several dimensions, are unlikely to be easier to use even if they save space; but there is endless scope to improve all keys by clearer, professional design and careful control of key content and logical structure. However, clever designs that avoid the old-fashioned schoolbook look of dichotomous keys probably encourage more non-botanists to take an initial interest in a field guide.
Some principles of good access design in keys
Although most principles of key design are general, some are particularly important in static keys, where there may be no way round a particularly badly designed section.
Here are some points that are useful especially for (indented) field guide keys:
•
Each pair of statements (or questions) in a dichotomous key should be mutually exclusive (that is, opposites, or at least strict alternatives) – for example, ‘flowers with five petals’ versus ‘flowers with four petals’ is fine; but ‘flowers with five petals’
versus ‘blue flowers’ is not fine, even if it is strictly correct. However, it can be useful to mention correlated characters to help users spot where they have gone wrong.
The latter pair might be rephrased as ‘flowers with five petals, not blue’ versus ‘blue flowers with four petals’. A user with a flower of five blue petals would therefore know they had probably made a wrong choice earlier.
•
Use simple, easily observed and more reliable characters first in the identification process. There may be a conflict between easily observed and reliable characters; but it is important to try and reach a workable compromise. For tropical plant field guides, a good compromise for the ‘first cut of the cake’ is the leaf arrangement as outlined in Chapter 6.
•
Combinations of characters can be used to make a key question more precise and reliable; but try to avoid complex logic in questions. In your key questions or statements, strive to avoid ‘and’ and ‘or’ altogether, unless the logic is really basic, in which case you should limit yourself to a maximum of one ‘and’ or ‘or’. For example, ‘flowers red or pink’ is alright; but ‘flowers red and white, or blue and pink’ may already stretch the overheated field botanist’s patience. Avoid constructs such as ‘statement 1, or if not statement 1 then statement 2’ .
•
Never include more than two optional answers if you are using characters combined with ‘and’ or ‘or’. When checking for errors of logic, note that if you combine two characters in one question, one option will probably include ‘and’ and the other should then include ‘or’. Make the ‘and’ and ‘or’ stand out – for example, with bold or italics.
Although you may think your logic is easily comprehensible at your desk, even very simple questions can appear as an impenetrable fog at the end of a long hot day in the rainforest, when users are being pestered by mosquitoes and failing light. In other
Identification 107
words, this need for logical simplicity is more of a constraint on field guide design than on herbarium keys, where generally well-ventilated users will have more time to work through such statements. It will usually be possible to restructure your key to simplify.
Bad, over-compact key
Leaves hairy, or if glabrous then not serrated
Examplus tomentosus
Leaves without hairs and serrated
E. serratus
Longer but clearer key
Leaves without hairs
Leaves serrated
Examplus serratus
Leaves entire
E. tomentosus (below)
Leaves hairy
E. tomentosus
Limit the number of choices of answer for any one question to two (the classic ‘dichotomous design’) unless there is a good reason for more. Users may not notice the third or subsequent options, and even if they do, choosing one of three is often far more taxing than choosing one of two. There are cases where this point does not apply, as follows.
One situation that justifies more than two options is where more than two options are exactly ‘compatible’, referring to a categorical character with a few distinct states (such as leaf arrangement):
Slightly awkward design
Flowers not red
Flowers blue
Exampla coerulea
Flowers white
E. albida
Flowers red
E. rubra
BOX 5.9 EFFICIENT KEY CHARACTERS FOR A GIVEN POINT IN A KEY
The most efficient key (MEK) characters for a given point in the key are those that split the remaining species into groups of equal number, ultimately enabling identification with the fewest steps. This is best illustrated for dichotomous keys, where MEK characters split the remaining taxa into two halves. So, if a dichotomous key is made of successive MEK
characters, one character is needed to differentiate two taxa, two characters for four species, three characters for eight species, and so on. Define ten characters correctly, and you find your plant among 1024. In this context, the worst key characters overall are those that are only found in a single taxon – we call these ‘fingerprint characters’ if they uniquely define species and occur in all of a large set of them: 1023 fingerprint characters might have to be examined (or answered ‘yes/no’) to differentiate 1024 species in the worst case.
Infrequently, unless very common species are listed first, users might hit the right species after considering just one or two fingerprint characters, but the 10 MEK character route is much faster, on average. Fingerprint characters (for example, bark slash pattern) can, however, be ideal as the focus for structured browsing if they can be arranged in a reasonable sequence (for example, more to less fibrous). In this sense, browsable characters could be thought of as the 1024 possibilities for a single key question of a polychotomous key. Compromises between two, or more, choices, browsing pictures and reading about characters, analysis or recognition have to be considered together.
108 Plant Identification
Better design
Flowers red
E. rubra
Flowers blue
E. coerulea
Flowers white
E. albida
Arrange for key steps to divide into subgroups of similar size (see Box 5.9); but learn when picking off one species at a time is the best solution.
Here are the three occasions when halving the species with each choice is far from ideal – that is, when you should consider ignoring the two rules of dividing into equal groups and of keeping the key dichotomous:
1 To make the subsequent subdivisions clearer, you may want to get species that are particularly distinctive, yet variable, ‘out of the way’ so that they do not have to appear in both branches of subsequent questions. In this example, losing a species with very variable petiole lengths makes the subsequent divisions easier:
Plant with red leaves
(petiole 1–10cm long)
Species c
Plant with green leaves
Petiole < 2cm long
Species d
Species e
Petiole > 5cm long
Species f
Species g
2 In ‘difficult’ groups, species may differ by subtle differences in several characters combined. In these cases, it may be safer and easier, however tedious, to ‘pick off’
species one by one: in effect, the user reads a list until they reach an answer.
Similarly, you may choose to list ‘fingerprint characters’ where every species is unique in some distinctive character, especially if associated pictures that are opposite the key can be browsed in tandem (for example, ‘Choose shape from the list and from the accompanying illustrations of shapes A–Z’, rather than ‘shape A: not shape A, shape B: not shape B, etc.’).
3 Try and pick off commoner species first: users may then rarely have to answer the subsequent questions. If one species is much commoner than all the rest, it is worth using the otherwise awkward construct of the type in the key below.
Large canopy tree with steep buttresses;
Triplochiton scleroxylon
leaf lobes divided to halfway … more
details … very common …
Not Triplochiton: description otherwise
(Key to less common species starts)
Species a
Species b
Species c
Most beginners will correctly identify their (common) tree with the first question, and most experienced users will know Triplochiton, and will therefore ignore the first question.
Identification 109
Make retracing steps (to previous questions) easy in a key; facilitate trial and error, and support the ‘what if’ approach to identification that most users use when confronted with difficult questions in keys.
In the case of electronic designs, a web browser or document with hyperlinks, this simply requires a ‘back’ button: a major advantage of electronic media over their printed equivalent, one would assume, although they are not always used well. For static keys, a strictly hierarchical arrangement of questions – where a question can only be arrived at from one previous question, rather than a web-like arrangement, with questions being arrived at from more than one possible previous question – will greatly help users to retrace their steps if they need to. Indented keys are better in this sense than numbered ones because it is easy to find the previous, less indented option. Although you can put the ‘return to’ numbers in numbered keys, they are awkward and often confusing (see Table 5.1d).
Make the structure of the keys and guide logical and memorable so that users can learn to recognize and jump straight to particular groups or sub-groups. If the groups correspond to plant families or genera, then there will be educational benefit to be gained through regular use of static formats.
In some types of guide book the keys are fully integrated with the species descriptions (see, for example, Hawthorne, 1990; Ribeiro et al, 1999; Hawthorne and Jongkind, 2006), rather than the keys in one place and the descriptions and pictures in another. This saves space by reducing cross-references and the need to repeat the key features in the descriptions; above all, it facilitates the ideal mixture of browsing (recognition) and key use (analysis). The keys lead directly into the descriptions of species at the end of the keys, and the species order (nature order) follows the structure of the key.
The pictures that relate to the species follow the same sequence, ensuring structured browsing and, indeed, reducing the need to illustrate key questions separately. This design aspect can be planned in consultation with users (see Chapter 3).
Note that the sequence of species even in an indented key can be manipulated within reason by swivelling the sequence (usually a pair) of options at each level.
A common complaint of indented keys is that the number of levels is limited by how far you can indent the text, and that the heavily indented text is a waste of space.
However, this is only a problem if you try and arrange keys to tens or hundreds of species in one block. Keys with more than one page of questions can be broken up into hierarchical groups, so the entire structure for each sub-key can be seen on one page.
Hawthorne (1990) breaks indented keys to 674 trees in Ghana into groups of about 5 to 12 species, each group on one page, all illustrated on the facing page. The indentation is never more than five steps deep.
Dynamic guides and associated access methods
In dynamic guides, the order of species and other aspects of the content of a guide can be rearranged during use, or at least after initial publication, to suit different users, seasons or places (see Boxes 5.10 and 5.11).
If only the order or species content is dynamic, as in a basic ring-bound file, users can remove species that do not occur in the current patch of forest, or life forms they are not interested in, perhaps leaving a small enough number to browse (as with Hawthorne
110 Plant Identification
BOX 5.10 CARDS WITH PUNCHED HOLES AND/OR
OTHER INFORMATION
There are two main types of access methods based around sets of cards (which can be reshuffled) with holes punched in them. Both have novelty or educational value for a field key, although they are largely superseded by computerized multi-access keys today.
In the ‘species-per-card’ type, each card represents a species (or other taxon), and holes around the margin represent different characteristics, perhaps ‘fleshy fruit’. If the species represented by the card does not make a fleshy fruit, the relevant hole is clipped to make it into a notch. When a pin is passed in this position through a pile of cards and lifted, only the cards that represent species with fleshy fruits are left dangling on the pin.
In the other ‘character-per-card’ type a single master card has all the species written out at their specific point in a matrix and each punched card represents a character state, say fleshy fruit. Holes are made in the middle of each card, at positions corresponding (on the master card) to all those species with fleshy fruits. By collecting together several cards for different character states applying to the specimen being identified, it is possible to see through the holes on the combined cards only those species that have that character combination.
Punched hole cards have been used in the past for certain types of plant identification guide (see, for example, Hansen and Rahn, 1969; Simpson and Jones, 1974; Hyland, 1982), with the species-per-card type more appealing for many field guides because species can be added as time goes by, and other notes or photographs about species can be added to the card. Keay et al (1960) appended to their tree field guide a ‘numerical key’, which, although usable in the static form in the book, was in a format designed to be used in a reader’s home-made card index.
Although these can be satisfying to use, like a game, especially with large sets of species, one problem is that each hole can only represent a binary character. This is efficient for some characters, but less useful for multi-state characters, such as flower colour, or smell for which separate holes are needed – for example, flower blue (yes/no); flower white (yes/no). At least it is possible for a single species to have
more than one possible flower colour; but there is usually only a limited amount of space for characters around the card.
Hence, species-per-card packs are best when the number of index characters is limited, and then some other method – possibly just browsing of the pictures of the species
– is used for further refining the search within the selected set.
Alternatively, you could leave out the punched holes altogether. The suits (hearts, etc.) of classic playing cards can be seen as index icons, and you could design similar icons to represent leaf arrangement, for instance, with other details to separate species on each card. Such playing cards have been used to stimulate interest in plant identification rather than for field guides per se. For field use, we would recommend some binding to prevent the pack from becoming dispersed (see Box 5.7 and Case study 8.1, page 184).
and Gyakari’s Photoguide to the Larger Trees of Ghana’s Forests (2006) in Box 5.6). In other dynamic guides, access involves the guide responding to user input. In the most dynamic guides on computer, access and other aspects of the ‘behaviour’ of the guide can be indistinguishable and such programmes, customizable in many ways, will no doubt become more common.
Identification 111
BOX 5.11 MAIN TYPES OF DYNAMIC FORMAT GUIDES
The main types of dynamic (field) identification guides are as follows:
•
Sets of cards – for example, with access information represented by punched holes, and/or with various types of illustration; see Box 5.10.
•
Pages in ring-bound files. Ring binders hold together pages, while allowing for additions to or removal from the species dealt with in a field guide – for instance, to include only species for one forest, or to add newly discovered species. For the pages to be independent, a single page with two faces has to cover one species – hence, probably pictures on one side, text on the reverse – so the pictures cannot face the text without having half the guide as redundant white space. Where this is not the case (as with Thikakul, 1985), it is hard to see any advantage of a ring binding.
Plant Identification Page 19