Introduction
The goal of phytolith field and laboratory work, to transfer the information contained in the soil horizon to the microscope slide accurately (and efficiently), is not a trivial task. Phytolith analysis consists of various complex parts, from field sampling and extractionâwhich encompasses several steps of pre-treatment as well as separation of bio-opal from the mineral fractionâto slide preparation, and finally counting. Each of these steps may affect accuracy in representation of the source assemblage. Thus far, a considerable work effort among phytolith researchers has centred around on methods of extraction and the biases that pre-treatment and separation steps may cause (Carbone 1977, Fredlund 1986, Hart 1988, Lentfer and Boyd 1998, 1999, 2000, Madella et al. 1998, Parr 2002, Pearsall 2000, Piperno 1988, Rovner 1971, Twiss et al. 1969, Zhao and Pearsall 1998). However, there are other methodological aspects of phytolith analysis that are less often discussed, including choice of sample size, level of magnification, and count size (but see Strömberg 2003, in prep.). These procedural steps commonly show great variation among studies, and it is not always clear how and to what degree the differences influence the resulting phytolith count and vegetation inference.
One such âoverlookedâ step is slide preparation, that is, how phytoliths are transferred onto microscope slides for viewing under the scope. Why is choice of slide preparation method important? As will be elaborated below, a particular slide-making protocol can be hypothesized to bias phytolith assemblages due to sorting of phytoliths with different mechanical properties. Compared to for example palynomorphs, phytoliths represent a highly heterogeneous group of particles that exhibit great variation in a number of parameters, namely:
- size, where phytoliths range from 2 to 250 micrometers and up to 2000 micrometers (Piperno 1988, Runge 1996, 1999);
- shape, as phytoliths include everything from spherical bodies to rods and sheets);
- specific gravity, which varies between 1.5 and 2.3 due to differences in amount of occluded carbon in bio-opal (Jones and Beavers 1963).
Biases in size distribution are of particular concern. This is because size distributions in phytolith assemblages have been shown to be ecologically meaningful. More specifically, grasslands produce assemblages with on average smaller phytoliths than do forests (mean = ca. 25 micrometers and ca. 39 micrometers, respectively; Hansen et al. 1998). As a consequence, procedures that lead to different size biases can potentially change or obscure the vegetation signal contained in the sample. It is not clear whether there is also systematic variability in shapes and specific gravity between grassland and forest phytolith assemblages (e.g., a higher abundance of sheet elements or large mesophyll complexes in forest assemblages; see Dinan and Rowlett 1993), but it cannot be ruled out. The current study will focus on the effect that different slide protocols may have on size distribution in an assemblage.
What methods for slide preparation are used and how may they uniquely influence assemblage size distribution? The range of different approaches to slide making among workers is difficult to assess, because the exact procedure is not commonly reported or described in the phytolith literature (e.g. Alexandre et al. 1997b, Kondo et al. 1994, Strömberg 2002). However, it appears that the two most common slide preparation techniques are either to, a) transfer a portion of a phytolith-alcohol suspension using a pipette onto a slide, letting it dry and then adding mounting medium (Huber 1982 [in Pearsall 2000], J. Jones 2000 pers. comm., Kaplan and Smith 1980 [in Pearsall 2000], Lawlor 1995, Mulholland 1989, Parr 2002, Strömberg 2002, 2004, Thomson 1982 [in Pearsall 2000]), or b) dry the phytolith residue, transfer part (or all) of it to a slide and mix it with the mounting medium on the slide (e.g. Bozarth 1996, Carbone 1977, Pearsall 2000, Rovner 1971, Runge 2001). These âwetâ and âdryâ methods may not necessarily lead to differential size sorting. If phytoliths of relatively narrow size fractions are extracted (particularly if the size fraction is e.g., 2â30 micrometers) or if the entire phytolith yield is used to make one slide, the two slide-making procedures are likely to result in assemblages with comparable size distributions under the microscope. However, most commonly this is not the case. Phytolith researchers regularly chose to study a wider size range (e.g. Donohue and Dinan 1993, Lawlor 1995). Also, processing of standard amounts of sediment (e.g., 5â20 grams) is likely to produce phytolith yields that are far too rich to fit on a single slide (e.g. Piperno 1993). Under these circumstances, the two approaches to slide preparation can potentially make a difference in terms of the resulting size composition on the slide. This is because particles suspended in a solution such as water (or alcohol) behave fundamentally different from particles in a âgrain flowâ (resulting from stirring of the dry phytolith residue).
Particles suspended in water settle at different rates depending on size, shape, and specific gravity (Carver 1971), but also to a minor degree due to roughness of the surface (Arnold 1911). Even though a sample is thoroughly agitated prior to pipetting, sorting due to differential settling velocities for particles of different size/shape/density will ensue within nanoseconds. The exact settling times for different size fractions are impossible to calculate exactly for phytoliths in an ethanol suspension contained in a centrifuge tube (or a similarly narrow storage vial), since the assumptions for Stokesâ law are violated in this situation (see Carver 1971 for discussion). This is because:
- particles are mostly non-spherical (Wadell 1934);
- particles are commonly above the limit for Stokesâ law (in theory >50 micrometers but in practice above 140 micrometers for settling in water) causing turbulence during settling (Rubey 1933);
- the fluid medium is not infinite (the centrifuge walls influence friction; Krumbein and Pettijohn, 1938);
- particle concentration is >1% above which fluid viscosity changes due to interaction between particles (at least in water), which can cause particles to clump and settle together (Carver 1971, Cook 1969).
Nevertheless, applying Stokesâ law can give a rough idea of settling velocity. Using 2.2 g/cm3 as the density of phytoliths, 0.789 g/cm3 as the density of alcohol, and 1.19·10â3 Ns/m2 as the viscosity of alcohol, the velocity of particles in suspension are estimated to vary from 16.2 ÎŒm/s for 5 micrometer size particles, to 1.62 ÎŒm/s for 50 micrometer size particles, to 14.5 ÎŒm/s for 150 micrometer size particles (see also Lentfer et al. 2003). In theory, this means that in order to collect particles >150 micrometers from the centrifuge tube using a pipette, a sample has to be taken >14 mm below the surface of the suspension within one second after agitation. Samples taken from the surface of the suspension should be relatively depleted of larger size particles, and samples taken from the bottom could potentially be enriched in coarser particles. Note again that differential sorting due to varying shape and specific gravity is likely to also occur, further complicating the issue; as mentioned, these differences will be disregarded for now.
On the other hand, there is some degree sorting in dry sediment as well. Bagnoldâs theory predicts that dispersive pressures developed in a grain flow (during mixing) should be proportional to particle size, forcing larger sized particles to regions of least shear, that is, the top surface (Bagnold 1941, Blatt et al. 1980). In addition, smaller particles might work their way down to the bottom through a âk...