Click below for larger images and slideshow more...

A transection of a Saururus cernuus stem shows a bundle with cambial activity producing secondary xylem and secondary phloem, but a cambium is not really developing in the primary ray area. Fibers occur at both inner and outer faces of the bundle.

A transection of a Houttuynia cordata (Saururaceae) stem shows a bundle in which, if there is any secondary growth at all, the cambium can have contributed only a few cells of xylem and phloem. The bundle is enclosed with a sheath of fibers, indicative that either secondary growth is absent or is quite minimal. The sclerenchyma links bundles laterally, there is no interfascicular cambium. Something very much like a monocot stem structure has been achieved.

Although Hedyosmum is a woody plant, it tends to have a cluster of stems is like the canes of a Piper plant. Each of these stems has wood that remains juvenilistic, with upright ray cells, as shown in this radial section. The dark contents in some cells suggest accumulation of compounds that work in deterring herbivores—but all the ray cells are not involved in that program: there can be division of labor within a ray (and often there is!).

A tangential section of Hedyosmum wood shows (in conjunction with the radial section) that all of the ray cells are upright. Such rays are often indicative of secondary woodiness.

The very short fusiform cambial initials of Lactoris fernandeziana wood make limits between ray cells and fascicular cells difficult to determine. The wood of this plant begins rayless, and shifts to rays with upright cells only.

A transection of the secondary xylem of Crassula arborea shows that it lacks fibers, and consists of vessels in a parenchyma background. The parenchyma thus serves for mechanical support (through turgor) and water storage.

Division of labor can be considered a tradeoff between conductive efficiency and mechanical support.

The wood of Myrica hartwegii has a very brief earlywood in which few vessels are produced. Most of a growth ring consists of tracheids. This seems like an excellent formula for providing earlywood conductiveness while maximizing the conductive safety that the ground tissue of tracheids assures.

In the wood of Arbutus texana, the latewood contains only a few vessels, and only small ones. Thus, conductive safety is achieved to an appreciable extent by the tracheids in the latewood.

Cercocarpus intricatus wood consists mostly of tracheids, and is optimally “safe” conductively. A few vessels, relatively small in diameter, are scattered in the tracheid background. Narrower vessels embolize less readily than wide vessels.

A portion of a perforation plate of a vessel of Ascarina philippinensis. The long axis of the perforation plate is oriented horizontally in this picture. The holes in the pit membranes occupy less than half of the area of those pit membranes—so is this a vessel element or a tracheid with vessel-like characteristics?

The lateral wall of a tracheid of Trochodendron wood, from a radial section. The warts on the tracheid wall are probably related in some way to the fact that Trochodendron grows in areas where frost occurs.

The inside of a tracheid of Tasmannia lanceolata (Winteraceae). Warts on the wall are probably related to frost occurrence where this species occurs. Most species of Winteraceae, however, do not experience frost.

The inside of a tracheid of Tasmannia insipida (Winteraceae). The sculpturing on the tracheid wall (more nearly horizontal than helical) bears some relationship to conduction in conditions where frost occurs.

A transection of Artemisia filifolia wood shows an abundance of narrow vessels, grouped together in infinite numbers. This construction offers safety. Also contributing to conductive safety are the layers of very narrow vessels at the end of the growth ring. Some of these may be so narrow that they lack perforation plates and are thus vascular tracheids.

This tangential section of Grubbia wood reveals numerous rays, which are very tall and narrow, with upright cells only in the uniseriate rays. This ray configuration would provide a very large number of potential points of contact with an axial parenchyma system. In Grubbia, axial parenchyma is diffuse.

A tangential section of wood of Bursera simaruba shows that rays are relatively few in number, are vertically rather short, and are composed of procumbent cells. Such rays potentially provide few but large points of contact with an axial parenchyma system. In the case of this particular species, the fibers are septate and living, and effectively substitute for an axial parenchyma system (in fact, axial parenchyma strands are rare in Bursera). The presence of a secretory canal in the center of the ray of Bursera provides compounds retarding herbivory.

Diffuse axial parenchyma is evident in this wood transaction. Such a parenchyma distribution provides a large number of contacts between vertical and horizontal parenchyma, provided that the rays are tall and are relatively abundant—as they are in Grubbia.

Axial parenchyma cells often are in contact with vessels. Sometimes only a few such cells surround vessels but, here, in Amburana, the parenchyma forms a sheath several cells thick.

In the wood of Brosimum (Moraceae), the axial parenchyma forms bands that extend around much of the stem. Large points of contact between rays (several of which traverse the axial parenchyma in the portion shown here) and the axial parenchyma are present. Water storage and starch storage occurs in axial parenchyma of Moraceae.



   With my interests in wood anatomy beginning with studies of Asteraceae, one might think that I would continue working in groups supposedly related to composites (“Asteridae”).  And I did—Acanthaceae, Hydrophyllaceae, Boraginaceae, Buddleiaceae, Convolculacaeae, Fouquieriaceae, Gentianaceae, Gesneriaceae, Goodeniaceae, Hydrophyllaceae, Myoporaceae, Polemoniaceae, Solanaceae, Stilbaceae, and others.  But I didn’t see any reason not to study woods of angiosperms that seemed, in phylogenies from the mid-20th century to the present, to be “primitive.”  Because systematists have been very interested in these families, I collected them on my many trips.  And the southern hemisphere is rich in those families.  In fact, because many species of primitive angiosperms are not notably woody, they weren’t collected by wood collectors, so wood collections were poor in them.  So I collected Amborella and various Winteraceae and Chloranthaceae in New Caledonia, Austrobaileya in Queensland, Pseudowintera in New Zealand, Tasmannia in Sabah and Australia, Drimys in Chile, Sarcandra in the Ryukyu Islands, Euptelea in Japan,  and Illicium in Malaya, Japan, and Florida (list incomplete!).  If one carefully accumulates wood samples of particular groups over a period of years and takes advantage of any opportunity to collect them, one can have remarkable bases for wood studies.  It takes patience, and also a little disbelief (I probably won’t study these, but just in case, I’ll collect them).  Del Wiens sent me a wonderful sample of Lactoris, and Carol Todzia very generously sent me materials Hedyosmum (Chloranthaceae).  So I have published papers on all of these.  I showed that Lactoridaceae were close to Piperaceae before others thought so (wood anatomy does show relationships quite well sometimes, but we can’t be confident that it always does…).  In my summary of woods of Chloranthaceae, I showed how wood of Chloranthaceae suggests monocot origins.  Of course, those who worked with DNA in order to derive phylogenies did not refer back to these efforts.  And in a sense, that’s fine, because the DNA evidence is potentially so powerful that evidence from plant anatomy in general and wood anatomy in particular is no longer a primary tool for discerning relationships.  But assuming that the more recent DNA phylogenies will be confirmed in their essential outlines, isn’t there an interest to see how wood evolves when studied in the light of DNA-based phylogenies?  That brings me to the next theme….are there numerous resemblances between these woods, and what do they show about ecology of early angiosperms?  What do we know about the modes of early diversification with respect to wood anatomy? 
    Why does wood evolve into more specialized anatomical expressions?  One important question that needs to be answered is: why are there primitive woods still around today?  Wood can evolve rather rapidly, so primitive configurations must represent structures still very functional under some circumstances.  What are those circumstances, and under what circumstances are woods with apparently specialized features optimal?  What’s the advantage of the monocot type of structure?  Textbooks never answer any of these questions, and—more importantly—they don’t indicate that such questions exist.  The answers we have at present may not answer these questions completely, but better to ask them than to pretend that understanding xylem ends when one has described it.  All too often, we present students in science with knowledge to be learned, stuff that they should know.  A lot of this stuff goes out of date—so what is the value in their learning it?  Some basic knowledge is necessary, but if there is any value of an instructor in front of a classroom, it’s to convey enthusiasm and thereby keep the students interested in learning and—if possible—teaching themselves.  Telling students there is more to learn, and giving them examples of what is not yet understood should be inspirational to them, yes?  Telling them a lot of facts that have already been learned—do students need that?  They could find those facts on the internet any time that they need them. 
   How to make a monocot from a dicot.  A separate section on vessel element evolution in this website tells about the advantages and disadvantages of the primitive vessel element.  This section tells about wood as an entire entity, because no one cell type exists in isolation.  Wood is a compromise between two main functions: water conduction and mechanical strength.  Both have to be satisfied in some way.  Monocots use fibers around vascular bundles as the main source of mechanical strength, and dicots can use fibers in the bark for strength, too.  But in woody dicotyledons, conduction and mechanical strength are the two main functions of secondary xylem.  They are functions that are opposed to each other where cell structure is concerned (little wonder that monocots have tracheary elements in xylem for conduction and have fibers outside of the xylem for mechanical strength: no compromises are needed.  Saururaceae and Chloranthaceae show interesting examples of how that happened. 
   In two genera of Saururaceae, Houttuynia and Saururus, we can see how the monocot story happened.  I’m not saying that Saururaceae are ancestral to monocotyledons, please understand!  Piperales/Chloranthales are probably not very far from monocots, however.
   In Saururus cernuus, there is normal cambial activity, although not a great deal of it.  More about that in a minute.  In Saururus chinensis and in Houttuynia cordata, cambial activity in stems is very reduced.  In some of the bundles, there may not be any secondary growth at all.  The mechanical tissue of the bundle is outside of the phloem.  To change to a monocot stem from a condition like this, one needs only shifting in the course of bundles so that some are displaced into pith regions along part of their length.  Piperaceae have done that (perhaps not precisely the same way that monocotyledons have, but you get the idea). 
    The growth habit of Saururaceae is sympodial: branching rhizomes, the branches of which turn upwards and form canelike stems.  Not tall ones, and not long-lived. And the older parts of the horizontal rhizomes die while the system grows from the tips.  The proximal portions of each branch are horizontal and bear adventitious roots, the distal portions are the upright canes.  This pattern is precisely that of a monocotyledon [ PDF ].. 
    Saururaceae isn’t alone in this architecture.  Piperaceae and Chloranthaceae share this, as do the Aristolochiaceae that seem more basal in that family (Thottea).  Nymphaeaceae could be said to have it, although their rhizomes stay rather horizontal and don’t form upright branches extending  above the water. 
    Some Chloranthaceae, like Chloranthus, Sarcandra, and even Hedyosmum, have a canelike habit, although roots and stems can have indefinite amounts of secondary growth.  Interestingly, ray cells in Hedyosmum are predominantly upright, an indicator of paedomorphosis [ PDF ].  This suggests that Hedyosmum is secondarily woody.  The upright cell pattern does not give way to procumbent cells to any appreciable degree with growth in stems, a fact that is probably correlated with the relatively limited amount of wood produced.  Paedomorphosis in rays characterizes rays in Aristolochiaceae (a 1993 paper) and Piperaceae: they tend not to branch, and they are wide and composed mostly of upright cells.  This is true also in Lactoridaceae [ PDF ].  Lactoris has storied wood, which probably is related to its stature as a small shrub, and so fiber-tracheids are short.  The short fiber-tracheids of Lactoris are indistinguishable from ray cells in early wood, which looks rayless, but rays are identifiable as the stem grows in size.
     Dicots like Saururaceae, Chloranthaceae, and Piperaceae tend to grow in places that are moist or inappropriate for taproots (epiphytic or epilithic Piperaceae).  The monopodial habit, with a taproot, of most woody dicots, represents an alternative to the monocot pattern: using roots as a way of going to deeper soil levels for more abundant moisture and the trunklike stems as ways of reaching higher levels for a photosynthetic advantage.  Monopodial dicots mostly achieve both depth of roots and height of stems by means of woody cylinders, and thereby a division of labor between fiberlike cells and vessel elements is necessary.  There are some exceptions: plants that have fibers in bark instead of in wood, like Caricaceae (see my 1998 paper on that family) and plants that use succulence as a support mechanism.  Moringaceae show a nice range of transitions in that respect [ PDF ], as do Crassulaceae and Cactaceae.
   The adventitious nature of roots of monocotyledons has a consequence: there can be a difference between the adaptations of the root xylem and those of the stem xylem.  Cheadle found an organographic progression in specialization as one goes from roots upwards in the plant.  If one thinks about it, the nature of this adaptation becomes obvious: monocot roots tend to be much more short-lived than the stems, and so conductively more efficient xylem (e.g., simple perforation plates) is an obvious correlation.  This becomes very pronounced in bulbs, which have vessels with simple perforation plates in the roots, which are rather ephemeral, but tracheids only in the leaves and stems.  A tracheid-only condition maintains the conductive system cell by cell, and prevents a collapse of an entire vertical series of conductive cells.   Thus, the tip of a leaf can be dead while the remainder of the leaf stays alive in a bulbous monocot.
   Division of labor in dicot woods. In a primitive dicot wood, vessels are not much larger than tracheids in diameter.  The end walls are long, like those of tracheids, and the numerous perforations per perforation plate, although different in shape, are not so different in number from the bordered pits on end walls of tracheids.  The greater width of vessels makes possible perforations that are wider.  Pits on the end walls of tracheids of conifers have torus-and-margo structure and can close off if air invades one tracheid, isolating it from the rest of the conductive system.  This conifer tracheid is not only rather effective in conduction (thanks to wide pit membranes and their margo holes) but it excels in conductive safety.  Little wonder that conifers do not have vessels.  The Gnetales did develop vessels, but, interestingly, still retain tracheids in addition to their vessel elements (wood-and-bark-anatomy-of_1989) [ PDF ], [ PDF ], [ PDF ], and [ PDF ]. 
   The perforation plate of a dicot wood is not maximally designed for either conduction or safety.  If perforations are small enough, they could trap bubbles, perhaps sieving out bubbles when frozen water thaws. (Most dicotyledons with long perforation plates do not grow where freezing is likely, however). But the perforations cannot close off (aspirate) as can conifer tracheid pits.  Tracheids are better than vessel elements at preventing spread of air bubbles. The point that needs emphasis is that many dicot woods still do retain tracheids—the safe conductive system—in addition to vessels.  Such a system retains water columns during extreme conditions when vessels fill with air.  Tracheid water columns in such a case may be in latewood and may permit survival of foliage and stems, and the earlywood the following year can restore the advantages of vessels. A latewood composed entirely of tracheids in a wood that has vessels in earlywood, like that of Myrica hartwegii or some species of Ephedra would seem to satisfy all possibilities.  But does it?  Such woods are infrequent.  Woods with relatively few but narrow vessels embedded in a background of tracheids, as in Cercocarpus, are also infrequent. 
    Why do we have primitive vessels in extant plants at all?  Recent work (by Uwe Hacke) has shown that primitive vessels offer only a small increase in conductivity, compared with specialized vessels.  The perforation plate, in this case, is a more of a blockage (especially considering the pit membrane remnants) than a gateway.  The famous Hagen-Poiseuille equation, which states that conductivity is equal to the fourth power of the vessel radius, isn’t applicable in that case—even though a long scalariform perforation plate might offer spaces collectively equivalent to the transactional area of the vessel.  A long scalariform perforation plate offers friction.  If this idea is operative, then vessel-bearing plants should be restricted to wet habitats.  And in fact, to a remarkable extent, they are.  The opposite idea, that scalariform perforation plates don’t offer much of an impedance compared to simple perforation plates, has difficulties, because in group after group, as in Saururaceae [ PDF ], simple perforation plates apparently evolved rather rapidly.  Some species have scalariform perforation plates in latewood but simple perforation plates in earlywood (Styrax) or similar configurations (fewer and thinner bars in earlywood), which seems to indicate that there is selective pressure for simplication of the perforation plate.
    Were vessels lost? Are we always sure what vessels are?  The “perforation plates” in some dicotyledons and contain such extensive pit membrane remnants that functionally the vessel elements are more like tracheids: the few holes left in the perforation plate can’t enhance conduction very much.  One sees such perforation plates in Aextoxicon [ PDF ], Ascarina [ PDF ], Chloranthus [ PDF ], Hedyosmum [ PDF ], and Illicium [ PDF ].  “Perforations” that have imperforate membranes can coexist with those that are perforate in a single wood section of wood in Illicium and Chloranthaceae.  Thus, those who chart vessel presence or absence (Evolution 56:466) may be dealing with a nonexistent contrast.  Certainly the contrast is questionable in many ferns and monocots (see the Fern and Monocot Xylem section of this website).  Cladistic reconstructions suggest that vessels may have been lost prior to origin of Winteraceae and Trochodendraceae, but perhaps pre-vessels were merely retained and thus true vessel elements were never present.  The idea of vessel loss on physiological grounds (tracheids better when freezing occurs than vessel elements) has been offered for Winteraceae (Evolution 56:464-478).  However, if that was an advantage for Winteraceae, they should be better represented in cold areas (the habitats of most species do not freeze: many are in cool montane cloud forest, but there is a difference between cool temperatures and freezing temperatures), and they should have radiated well since that loss (they do not occupy large areas of the world).  The same is true of Trochodendraceae. Incidentally, tracheids of Winteraceae in areas where frost does occur have warted surfaces facing the tracheid lumina—a good indication of the occurrence of frost. [ PDF ].  And if loss of vessels, with simultaneous retention of tracheids were an advantage in montane areas where frost occurs, there should be an abundance of woods with an all-tracheid formula in tropical mountain elevations (but there is not).  Conifers are abundant in high north-temperate latitudes, but so are such vessel-bearing plants (without tracheids) as Populus and Salix. We now know that narrow vessels embolize much less readily than wide ones, and therefore are common in latewood of temperate trees (experimental evidence also confirms this capability).  So why not woods constructed of tracheids plus narrow vessels in areas prone to frost—would a few narrow vessels disable an entire conductive system?  Obviously not. There are some unanswered questions here.  Such mechanisms as vasicentric tracheids, and grouping of vessels in woods that have libriform fibers offer safety mechanisms.  Each wood has its own story…. [ PDF ].  The various familial monographs I have done are not to help anyone in wood identification or phylogeny.  Their purpose was to uncover patterns of diversification in wood anatomy within a natural group and to attempt to explain the nature and probable reasons for the patterns uncovered.
       The shift away from tracheids—a mistake ?   Division of labor in wood cells would seem to involve not only better conduction in vessels, but mechanically stronger fiberlike cells.  A surprising number of woods do have tracheids and vasicentric tracheids.  Oaks (Quercus) are an example of a single such genus.  There are very few species in which the imperforate tracheary elements teeter on the border between tracheids and fiber-tracheids.  Tracheids are apparently conductive cells, which deter vessel grouping.  A shift from tracheids to fiber-tracheids to libriform fibers has occurred.  Smaller simple pits with slitlike pit apertures that run parallel to the helix in which cellulose microfibrils are laid down offer minimal loss of strength (libriform fibers).  The species that have fiber-tracheids mostly have such minimal pit borders that loss of strength is, theoretically, negligible.  Conductively efficient vessels with simple perforation plates, plus libriform fibers (or fiber-tracheids) with little or no conductive ability: this a great formula in tropical rain forest trees, where conduction of large volumes of water (in tree trunks that are mechanically strong) to canopies high in the air is required.  In these situations, the loss of tracheids as a conductive safety mechanism isn’t important.  That would explain explosion of woody dicotyledons into rain forests, but what about seasonally dry areas?  As Asteraceae indicate, there are alternative methods of achieving safety [ PDF ].  The grouping of vessels into large aggregations and the formation of very narrow vessels are two tactics seen in woody Asteraceae (the widespread Great Basin sagebrush, Artemisia is a wonderful example).  However, some Asteraceae have very narrow vessels that are equivalent to tracheids in conductive safety, and some have vascular tracheids (end of growth ring) as well as vasicentric tracheids (in contact with vessels)—not many, but just enough.  Grouping of vessels is a very common strategy in woods of various families that lack tracheids—in fact, few such families are without it [ PDF ].
   The other wood functions.  Division of labor in tracheary elements is a familiar phenomenon and one easily understood, what about parenchyma in wood? There is a division of labor between axial parenchyma and ray parenchyma.
    In a young dicot stem, the rays are composed of upright cells only or predominantly, and the rays are often narrower.  As the stem increases in diameter, the rays become wider and procumbent cells become more abundant in them.  But more importantly, this happens phylogenetically, too.  Rays become horizontal shafts of tissue increasingly more efficient at radial transport of solutions containing photosynthates.  Remember that conductive cells are elongate.  Something that never seems to get mentioned when plant anatomy is taught—it’s so obvious!  Conductive cells are elongate because there are fewer impedances in the direction of flow (the cross walls).  Ideally, there would be no cross walls, and the cells would form a tube, which is what happens in vessels with simple perforation plates.  But fewer impedances are better than more numerous impedances, so one elongate cell is better designed for conduction (in the direction of elongation) than an equivalent series of five short cells with the total length of the elongate cell. 
   In a primitive wood, rays tend to be narrow, tall, and with procumbent cells only in the center of the ray—the rays otherwise composed of upright cells.  Tall, narrow rays, relatively frequent (abundant), offer numerous points of contact with the axial parenchyma, which in primitive woods, is composed of diffuse strands of cells.  The two types of parenchyma have evolved simultaneously and conjunctively in dicot woods.  In primitive woods, the more numerous points of contact and greater dispersion of ray and axial parenchyma tissue are a configuration that works well.  If storage and conduction are the main functions of parenchyma in wood, having these cells dispersed more seems like a good strategy.  But in woods where there is massive flow of photosynthates related to seasonality, having fewer points of contacts between between sheets of horizontally-oriented cells (the rays) and big sheets of vertically-oriented cells (the axial parenchyma) makes sense.  Wood is where photosynthates are stored, for the most part—where else in a plant can they be stored?  And they are stored in axial and ray parenchyma of a wood. 
As rays become fewer and larger in size and consist more predominantly of procumbent cells, axial parenchyma changes from diffuse strands into larger sheets and aggregations.  The points of contact between the vertical and the horizontal systems become fewer but those points of contact are larger, involving more numerous cells at each contact point.  The occurrence of bordered pits on ray cells (see Bordered Pits on Ray cells in the Wood Evolution section of this website) shows that flow of photosynthate solutions is going on.  It also shows that ray cells have secondary walls, a cellulose investment that indicates that mechanical strength of ray cell walls definitely is of importance.  More than half of the dicot woods one encounters have secondary walls on ray cells, and investment in cellulose and associated compounds would not be made by the plant unless wall rigidity were achieved.
    Notice that in the above, I have tried to suggest why more primitive character expressions are still in existence.  Botanists always seem to stress the advantage of specialized character states, but why are the primitive ones still around?  The answer is, of course, that they still work very well, at least under some circumstances.  So we should explain those circumstances, as well as the ones in which specialized character states occur. 
    Parenchyma cells in woods also have the function of accumulating distasteful and toxic substances—bitter, resinlike compounds that tend to stop chewing beetles and other organisms from attaching the wood.  Probably even fungi are deterred by some of these compounds.  If these are functions of parenchyma, having more numerous strands of axial parenchyma, diffusely distributed, and rays that are taller, slenderer, and more numerous would provide a better distribution of cells with such contents.  So how do the woods with more specialized rays and axial parenchyma configurations deter predation?  One possibility is that the more specialized woods have libriform fibers and fiber-tracheids rather than tracheids, and thus deterrent compounds can be deposited in libriform fibers and fiber-tracheids, which are nonconductive cells.  Tracheids, which many woods have (contrary to the 1989 IAWA glossary which misdefines tracheids and vasicentric tracheids) are conductive cells, and thus cannot serve as sites for deposition of deterrent compounds
    Parenchyma cells in wood can serve in water storage, a phenomenon that is not limited just to succulents [ PDF ].  Even living fibers can serve in water storage in wood.
    Incidentally, parenchyma cells adjacent to vessels tend to be a very common distribution pattern.  If maximizing intersections between axial and ray parenchyma were the only factor in play, axial parenchyma distribution with respect to vessels would be random.  It isn’t.  Axial parenchyma relates to conduction in vessels.  We know that in the sugar maple, sugars are released into the conductive stream from parenchyma cells, and what happens in the sugar maple probably happens very widely in dicots, it just hasn’t been documented yet.  And there are other possible values of having living cells in contact with a dead conductive cell.  Such cells could block intrusion of air bubbles into pits of the vessel.  And since a vessel element isn’t a living cell….think about the consequences of having conduction in a system of dead cells. 
    Woods are trying to tell us stories about function, we shouldn’t be just describing what the woods look like, we should be trying to tell their stories.
   Why did the Bailey-Frost-Kribs correlations show what they did?  In 1918, Bailey and Tupper published a paper that shows that tracheary elements have shortened during the evolution of flowering plants, and perhaps during the evolution of other woody vascular plants as well.  They left it at that, as though some inexorable power was at work.  The reason for long tracheary elements is that fewer impedances per unit length, as mentioned above for ray cells, is more efficient.  So why should flowering plants have abandoned that good plan?  The division of labor between vessel elements an imperforate tracheary elements allows length in the two systems to drift independently.  Both are governed, to be sure, by length of fusiform cambial initials.  Vessel elements tend to be about the same length as a fusiform cambial initial, but the imperforate tracheary elements can be longer because the derivatives destined to become imperforate tracheary elements can undergo intrusive growth. 
   What’s so good about a short vessel element?  First, if it has a simple perforation plate, the perforation plate no longer serves as an impedance, and a short vessel element can be as good at conduction as a long one.  Secondly, the constrictions in a vessel formed by the perforation plates can serve to confine air bubbles to individual vessel elements should air embolisms occur (a hypothesis originated by Slatyer), and thus smaller portions of the conductive system are disabled.  And thirdly, perforation plates may serve as strengthening devices, countering deformation of vessel elements when the water columns are under high tension. 
     All of these ideas I mentioned in “Ecological Strategies of Xylem Evolution” in 1975, but Martin Zimmermann didn’t believe them—he said the length of a vessel element has no functional significance.  But Martin Zimmermann’s thinking about wood physiology was primarily in terms of conductive efficiency, whereas wood structure is a compromise between conductive efficiency and conductive safety (with some considerations for mechanical strength, photosynthate management, etc. thrown in).  If “Ecological Strategies of Xylem Evolution” had any merit, it was to show that wood structure is a compromise among these selective pressures, which often run counter to each other.  Textbooks always mentioned growth rings as having better conductive efficiency in the earlywood, but they never told us what the advantage of latewood was!  And the other message of my book was that all wood structures have functional significance (evolution is a rigorous process!) and must be considered conjunctively.
    Frost, a student of Bailey, mapped the shortening of the vessel element in angiosperms and the changes in end walls and in lateral walls—using vessel element length as a measuring stick without considering why the length should shorten.  Kribs used the same measuring stick for plotting how axial parenchyma specialized and, independently, how ray parenchyma specialized.  Kribs did not think of the correlations, mentioned above, between the two conductive systems.  If he had, he probably would have said that if evolution changes the characteristics of the radial sheets of parenchyma tissue with respect to functional design, the characteristics of the vertical parenchyma system tend to change with it.
    Wood evolution is a pattern of coherent changes.  The changes in wood character states tend to be directional and take place independently in different clades not because reversion in a character state can’t take place in wood (it can), but because shifting backwards in one character (say, changing from banded parenchyma to diffuse axial parenchyma) without changing other functional designs of the wood in accord with the optimal functioning of diffuse parenchyma is highly unlikely.  In addition, genetic information for reconstituting a lost character state may be lost. 
   The picture that emerges is that wood of a relative primitive nature was present in a number of the major clades of dicotyledons when those clades first diverged.  The functional advantages of the changes outlined above applied not to one clade, but to all of them.  Therefore, similar changes took place independently (often in a kind of “parallel” or “convergent” way) numerous times. And when one of the changes takes place, there is a likelihood that many of them will because of functional interrelationships among them, although at least some of the character states can drift independently to appreciable degrees. If homoplasy (evolution of a character state independently more than once) hadn’t happened abundantly in dicot woods), Bailey, Frost, and Kribs would have found groupings of particular specializations in particular clades.  Asterids would have one kind of wood, rosids quite another.  But the ways in which one can construct a vessel-bearing wood aren’t infinite.  The constraints must be very considerable, so that what works successfully in wood of a rosid is much the same as what works in the wood an asterid.  The data mapped by Bailey, Frost, and Kribs were not mapped on any phylogenetic system, nor was any functional basis for the changes offered.  Today we can see that there have been a series of homoplastic changes, so that both primitive and specialized grades of structure and everything in between (including some character state reversions) can exist within various clades.  But the functional explanations must be sought for any and all changes.  Adaptation to ecological situations drives those changes, as it always has. 
    In order to understand wood evolution, does one have to have a global knowledge of wood anatomy, be acquainted with the ecology of the plants studied, understand wood physiology, and comprehend phylogenetic systematics, all at once?  It couldn’t hurt.