2.1. Till fabric

Grains rotate in till during its deformation, resulting in partial alignment of the long axes of non-equant grains to form a fabric. Fabric strength and direction should be related to the three-dimensional state of strain in the till, ideally as expressed by a second-rank tensor (Reference OertelOertel, 1996). Till-fabric characteristics have also been related to the geomorphic context (e.g. Reference BoultonBoulton, 1976; Reference BennBenn, 1995; Reference Yi and CuiYi and Cui, 2001), till-layer thickness (Reference HartHart, 1994, Reference Hart1995; Reference Hart and RoseHart and Rose, 2001) and the degree to which till may have displayed fluid-like behavior, as indicated by its porosity, water content or strength (Reference Dowdeswell, Hambrey and WuDowdeswell and others, 1985; Reference BennBenn, 1995, Reference Benn2002; Reference Benn and EvansBenn and Evans, 1996; Reference Evans, Phillips, Hiemstra and AutonEvans and others, 2006). However, the state of cumulative strain, by directly influencing grain rotation, is the most obvious first-order control on fabric, as emphasized by structural geologists interested in preferred clast orientations in rocks (Reference NicolasNicolas, 1992; Reference Tikoff and TeyssierTikoff and Teyssier, 1994) and fault gouge (Reference CladouhosCladouhos, 1999).

There is widespread but not universal agreement that high cumulative strains in till result in strong pebble fabrics parallel to the direction of shear (Reference BennBenn, 1995; Reference Benn and EvansBenn and Evans, 1996; Reference Piotrowski, Mickelson, Tulaczyk, Krzyszkowski and JungePiotrowski and others, 2001; Reference Carlson, Jenson and ClarkCarlson and others, 2004; Reference Evans, Phillips, Hiemstra and AutonEvans and others, 2006). This relationship is supported by experiments in which till containing pebblesized inclusions was sheared to known strains and fabric development was measured (Reference Hooyer and IversonHooyer and Iverson, 2000a). These experiments also showed that steady fabric strengths were attained at moderate shear strains (∼20), and provided no support for the postulate that particles in shearing till assume transverse or poorly aligned orientations at sufficiently high strains (e.g. Reference Clark and MartiniClark, 1997; Reference Lian, Hicock and DreimanisLian and others, 2003; Reference HartHart, 2006). Such transverse or weak fabrics at high strains are expected from the model of Reference JefferyJeffery (1922), but slip of matrix grains across the surfaces of pebbles in till clearly violates the no-slip condition of the Jeffery theory. This slip at the surfaces of pebbles, as indicated by their commonly striated surfaces (e.g. Reference BennBenn, 2002), keeps them aligned in the direction of shear (Reference Hooyer and IversonHooyer and Iverson, 2000a; Reference Iverson, Hooyer, Thomason, Graesch and ShumwayIverson and others, 2008).

There are several problems with using pebble fabrics. The relatively low density of pebbles in many tills, including the Batestown till, limits the spatial resolution of pebble fabrics to zones in the bed that may be thicker than zones where shear deformation occurred (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008). Human subjectivity and error in measuring pebble fabrics are well documented (Reference DrakeDrake, 1977; Reference Klein, Davis and HansonKlein and Davis, 2002). Perhaps most importantly, little useful information is contained in the orientations of the intermediate and short axes of pebbles; they provide, for example, no information about the three-dimensional state of strain.

Two other kinds of till fabric measurements are also made: fabrics formed by directions of maximum magnetic susceptibility in intact till specimens (Reference FullerFuller, 1962; Reference Gravenor, Stupavsky and SymonsGravenor and others, 1973; Reference Stupavsky, Symons and GravenorStupavsky and others, 1974; Reference Gravenor and StupavskyGravenor and Stupavsky, 1975; Reference BoultonBoulton, 1976; Reference Eyles, Day and GavicanEyles and others, 1987; Reference Stewart, Bryant and SweatStewart and others, 1988; Reference Principato, Jennings, Kristjánsdóttir and AndrewsPrincipato and others, 2005) and those formed by the long axes of sand particles measured in thin section (e.g. Reference Ostry and DeaneOstry and Deane, 1963; Reference Evenson and GoldthwaitEvenson, 1971; Reference Yi and CuiYi and Cui, 2001; Reference Carr and RoseCarr and Rose, 2003). These studies have yielded important insights regarding till genesis, but until recently the relationship between fabric characteristics and the magnitude and direction of strain had not been demonstrated. Ring-shear experiments with basal tills, including the Batestown till, have provided this relationship for the case of simple shear (Reference Thomason and IversonThomason and Iverson, 2006; Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008; Reference Iverson, Hooyer, Thomason, Graesch and ShumwayIverson and others, 2008) and hence serve as a rational basis for interpreting our field measurements.

2.2. AMS and sand fabrics caused by simple shear: past experimental results

The Iowa State University ring-shear device shears a water-saturated till specimen that is ∼75 mm thick. Shearing speeds are glacial (10–800 m a⁻¹), and effective normal stresses are similar to those measured beneath glaciers (5–400 kPa). The rotary design and associated annular specimen chamber place no limit on strain magnitude. Deformation within the central 10–35 mm of the till specimen is relatively uniform and closely approximates simple shear. Specimens for fabric analysis are collected from within this zone, so the distribution of strain outside this zone, which reflects lateral boundary effects (Reference Iverson, Baker and HooyerIverson and others, 1997), is irrelevant (Reference Thomason and IversonThomason and Iverson, 2006). Prior to an experiment, wet till is mixed in the specimen chamber to minimize preferred particle orientation. Consolidation of the specimen by ∼5–15%, after application of the normal stress, causes discernible but minimal preferred orientation of sand grains (e.g. Reference Thomason and IversonThomason and Iverson, 2006) and of axes of maximum magnetic susceptibility (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008). The till is then sheared to a predetermined displacement. Shear strain is calculated as the shearing displacement along the specimen center line divided by the shear-zone thickness, which is indicated by the relative displacement of marker beads that spanned the shear zone.

On completion of experiments conducted to various shear strains, intact till samples are collected from the shear zone for fabric analyses. Some samples are impregnated with epoxy and thin-sectioned for optical measurements of sand-particle fabrics. Sets of 25–75 samples (18 mm cubes) are also collected to measure anisotropy of magnetic susceptibility (AMS). These measurements yield orthogonal directions of maximum, intermediate and minimum magnetic susceptibility (k ₁, k ₂ and k ₃, respectively) for each sample. Ancillary tests in which blocking temperatures and magnetic hysteresis (Reference Tarling and HroudaTarling and Hrouda, 1993) are studied allow the source of the AMS to be determined (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008). More details regarding the fabric analyses and procedure of these experiments can be found in Reference Thomason and IversonThomason and Iverson (2006), Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others (2008) and Reference Iverson, Hooyer, Thomason, Graesch and ShumwayIverson and others (2008).

Rather then reproducing data from these studies, herein we simply summarize the conclusions from this work that guide our field interpretations. Although numerical values differ slightly among different tills, the following generalizations apply to all of the tills that we have subjected to slow, simple shear:

Therefore, where tills have sand-particle and AMS fabric strengths less than steady-state ring-shear values, bearing in mind their uncertainty, a reasonable inference is that these tills have not been subjected to unidirectional simple shear to strains as high as 7–30. In addition, fabrics will be oriented in the direction of shear. Moreover, for bed-parallel simple shear, fabrics should plunge up-glacier, k ₁ and k ₃ should both lie in a vertical flow plane, and k ₂ should be horizontal and oriented perpendicular to the shearing direction (Fig. 2).

Fig. 2. Relationship between simple shear and directions of principal magnetic susceptibilities observed in ring-shear experiments that were carried out to strains greater than the critical value. The steady-state fabric formed by sand-grain long axes in experiments was similar to the k ₁ fabric (Reference Thomason and IversonThomason and Iverson, 2006).

We acknowledge that subglacial conditions are almost certainly more complex than those of our ring-shear experiments, but that is why the experiments are valuable: they show how fabric develops for the idealized case of bed-parallel simple shear. Isolating more complicated states of strain from till fabrics measured in the field seems unlikely without first understanding this simple case.

3.1. Friday 3

Near Henry, Illinois, about 20 km up-glacier from the terminal margin of the Batestown advance, the Batestown Member and underlying deposits of the earlier Tiskilwa advance are exposed along a cut-bank of Clear Creek, a tributary to the Illinois River (Fig. 3a and b). This exposure was also described by Reference Carlson, Jenson and ClarkCarlson and others (2004), who studied the Tiskilwa Member. Four glacigenic facies are associated with the Batestown advance at this site (Fig. 4a). Lowermost fluvial sands are overlain by lacustrine silts, both of which were presumably deposited proglacially. Overlying these sediments is a homogeneous loam diamicton (13% gravel, 59% sand, 28% silt/clay) that is about 3.6 m thick and has been interpreted to be a basal till in nearby outcrops (Reference Johnson, Hansel, Mickelson and AttigJohnson and Hansel, 1999). This till is overlain by thin beds of oxidized diamicton and sand and gravel, which are likely supraglacial, ice-marginal sediments.

Fig. 3. (a) Topographic setting of the Friday 3 exposure. (b) Friday 3 exposure after trenching. (c) Oblique aerial view of the Wedron silica quarry and profiles W1–W3. (d) Wedron exposure of the Batestown Member and members deposited by earlier (Tiskilwa) and later (Yorkville) advances.

Fig. 4. Schematic stratigraphic columns at (a) Friday 3 and (b) Wedron. Photomicrographs show variability of till texture among sampling sites.

3.2. Wedron

At Wedron, Illinois, the Batestown Member is exposed in a large silica quarry, approximately 70 km up-glacier from the Batestown ice margin at its maximum extent (Fig. 3c and d). Reference Johnson and HanselJohnson and Hansel (1990, Reference Johnson, Hansel, Mickelson and Attig1999) described in this quarry four facies of the Batestown Member: a massive silty clay, two overlying mostly homogeneous diamictons and an uppermost layered diamicton containing sorted silts and sands (Fig. 4b). They interpreted these facies as deformed lake sediments, basal tills and a supraglacial till, respectively. The lake sediments contain occasional slickensides indicating deformation, and the overlying diamictons are compact, with moderately strong pebble fabrics (S ₁ > 0.7) in the direction of regional ice flow (60° southwest) and scattered deformed sand bodies.

The quarry sections studied by Reference Johnson and HanselJohnson and Hansel (1990, Reference Johnson, Hansel, Mickelson and Attig1999) are no longer accessible, but a different exposure in the quarry was located that differed only in that the uppermost facies consisted of bedded sand and gravel without diamicton. Our focus was the two basal till facies, which Reference Johnson and HanselJohnson and Hansel (1990) first interpreted to be lodgment tills but later considered in the context of a deforming-bed model (Reference Johnson, Hansel, Mickelson and AttigJohnson and Hansel, 1999). The lower till is approximately 0.8 m thick and is fine-grained (8% gravel, 27% sand, 65% silt/clay) (Fig. 4b) with a moderately sharp to gradational contact with the underlying lake sediments. The protolith for this till may be lake sediments that were sheared subglacially (Reference Johnson, Hansel, Mickelson and AttigJohnson and Hansel, 1999). The upper contact of this facies with the overlying till facies is sharp. This overlying facies is coarser (15% gravel, 47% sand, 38% silt/clay) (Fig. 4b) and approximately 1.0 m thick.

5.1. Friday 3 k ₁ and sand-grain fabrics

Laboratory experiments indicate that both k ₁ and sand-grain fabrics strengthen in the flow direction with shear deformation, so fabric strengths help indicate deformation magnitude. Figure 5 shows how k ₁ fabric strengths at Friday 3, as indicated by S ₁ values, vary with depth. S ₁ values are referenced to the steady-state, experimental fabric strength (S ₁ = 0.94) developed at moderate strains (7–30). Regression of the experimental data (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008) allows a 95% lower confidence limit (S ₁ = 0.81) to be specified for the steady-state experimental S ₁ value (Fig. 5). S ₁ values are larger than this lower limit near the top (0.5–1.3 m depth) and bottom (3.1–3.3 m) of the section. Over a broad range of depth in the middle (1.5–2.9 m) and at the very top (0.1–0.3 m) of the section, fabrics are weaker. The strongest fabrics measured have S ₁ values close to but not in excess of the steady-state ring-shear value. Depth variations in sand-grain fabric strength (Fig. 6) mimic variations in k ₁ fabrics to the extent that fabrics stronger than the 95% lower confidence limit occur near the top and bottom of the section. However, they occur over ranges of depth that are smaller than for the k ₁ fabrics (Fig. 6). The lower confidence limit in this case is calculated from regressions of ring-shear sand-grain data (Reference Thomason and IversonThomason and Iverson, 2006).

Fig. 5. AMS fabric strengths and directions as a function of depth at Friday 3. Horizontal and vertical marks in lower-hemisphere stereo plots are k ₁ and k ₃ orientations, respectively. Their respective eigenvectors (V ₁) are shown with open circles and squares. Great circles indicate planes of shear, as inferred by assuming only simple shear and directly applying relationships from ring-shear experiments (see text). Solid dots within those planes indicate the direction of shear.

Fig. 6. Sand-grain long-axis fabric strength and direction as a function of depth at Friday 3. Full-circle diagrams display particle measurements in a horizontal plane. Triangular tick marks at circle perimeter indicate orientations of V ₁ (azimuths). Half-circle diagrams display particle measurements in a vertical plane parallel to V ₁. Two-dimensional S ₁ values are based on orientations of particles in vertical thin sections, to allow direct comparison with measurements from ring-shear experiments.

Fabric directions, which are described herein using the principal eigenvector V ₁ to characterize the direction of clustering, yield important information about strain orientation. Azimuths of k ₁ at Friday 3 vary from the bottom to the top of the section: k ₁ azimuths shift from ∼90° (east–west) to ∼0° (north–south) over the 3.5 m thickness of the section (Fig. 7a). Azimuths of sand-grain long-axis fabrics shift similarly (Fig. 7a). Plunges of k ₁ cluster around the steady value obtained in ring-shear experiments (28° up-glacier) (Fig. 7b). The mean plunge at Friday 3 is 23.5° up-glacier (northeast), with a standard deviation of 14.1°. The mean plunge of the sand-grain long axes is similar but with more variability, 25.1° northeast ±34.4° up-glacier (Fig. 7b). AMS measurements provide k ₂ values, which also reflect the state of strain. The mean plunge of k ₂ orientations is near zero (mean = 10.8°), but plunges change smoothly with depth and locally exceed 40° (Fig. 7c).

Fig. 7. (a) Azimuths (V ₁ orientations) of k ₁ and sand-grain fabrics at Friday 3. Dashed bold line is a regression of both the k ₁ and sand-grain data. The regional flow direction was estimated using the orientation of the terminal moraine of the Batestown advance, which is ∼20 km from the exposure. (b) Plunges of k ₁ and sand-grain fabrics. Ring-shear results for k ₁ are shown; sand grains plunged 12° up-glacier in ringshear tests. (c) Plunge of k ₂ compared with the result from ring-shear tests.

The relationship between the state of strain during simple shear and orientations of principal magnetic susceptibilities determined in ring-shear experiments provides a means of estimating the orientation of shear planes from AMS data. If simple shear is assumed, results from ring-shear experiments tell us that the plane of shear should be oriented normal to the longitudinal flow plane, as defined by k ₁ and k ₃ orientations. Also, this plane should be rotated ∼28° down-glacier around an axis defined by k ₂ orientations to account for the 28° up-glacier plunge of the k ₁ fabric. Stereonets of Figure 5 show the resultant shear-plane orientation implied by the AMS data from each depth, as well as the direction of shear within that plane. If simple shear dominated strain, then shear planes over some ranges of depth were close to horizontal (0.3–1.5 m, 2.3–2.7 m), but elsewhere they were commonly inclined more than 20°. Dip directions of these shear planes are locally at a high angle (>45°) to the inferred shear direction (e.g. at depth 1.7 m).

5.2. Wedron k ₁ and sand-grain fabrics

Depth-averaged k ₁ and sand-grain fabrics at Wedron are generally stronger than those at Friday 3 (Figs 8 and 9). The lowermost 70% of profiles W1 and W2 (0.5–1.7 m), which were only 2 m apart, have k ₁ fabric strengths with S ₁ values in excess of the lower 95% confidence limit for till sheared to moderate strains (7–39) (Fig. 8). There is no abrupt change in fabric strength across the facies contact at ∼1.0 m, but the fabric weakens notably over the uppermost 0.5 m of the section. Depth variation of the strengths of sand-grain fabrics collected at W1 is somewhat similar: low in the section, S ₁ values exceed the lower 95% confidence limit, although in this case these strong fabrics are restricted to the upper three-quarters of the lower facies (Fig. 9). At profile W3, which was ∼100 m from W1 and W2, k ₁ fabrics are somewhat weaker in the lowermost three-quarters of the section (Fig. 9) than at W1 and W2. Sand-grain fabrics are uniformly weak: S ₁ values nowhere exceed the lower 95% confidence limit. As at Friday 3, the largest S ₁ values for both k ₁ and sand-grain fabrics are close to or equal to the steady-state ring-shear value but do not exceed it.

Fig. 8. AMS fabric strengths and directions as a function of depth at W1–W3. Horizontal and vertical marks in lower-hemisphere stereo plots are k ₁ and k ₃ orientations, respectively. Their respective eigenvectors (V ₁) are shown with open circles and squares. Great circles indicate planes of shear, as inferred by assuming only simple shear and directly applying relationships from ring-shear experiments (see text). Solid dots within those planes indicate the direction of shear. W1 and W2 were about 2 m apart; W3 was ∼100 m away.

Fig. 9. Sand-grain long-axis fabric strength and direction as a function of depth at W1 and W3. Full-circle diagrams display particle measurements in a horizontal plane. Triangular tick marks at circle perimeter indicate orientations of V ₁ (azimuths). Half-circle diagrams display particle measurements in a vertical plane parallel to V ₁. Two-dimensional S ₁ values are based on orientations of particles in vertical thin sections, to allow direct comparison with measurements from ring-shear experiments.

Similar to Friday 3, azimuths of k ₁ fabrics at the three profiles display systematic depth variation (Fig. 10a). Although there is considerable variability among the profiles, moving up-section, the profile-averaged azimuthal variation is from ∼40° (northeast–southwest) to ∼90° (east–west). Azimuths of the sand-grain fabrics indicate a similar up-section variation in azimuth (Fig. 10b). This variation in fabric direction has the opposite sign to the variation observed up-section at Friday 3, where azimuths rotate up-section from approximately east–west to north–south.

Fig. 10. (a) Azimuths (V ₁ orientations) of k ₁ fabrics at W1–W3. Dashed bold line is a regression of the data. The regional flow direction was assumed to be the clast-fabric direction measured by Reference Hansel and JohnsonHansel and Johnson (1996). (b) Azimuths of sand-grain fabrics at W1 and W3. (c) Plunges of k ₁ fabrics at W1–W3 and ring-shear result. (d) Plunges of sand-grain fabrics at W1 and W3 with ring-shear result. (e) Plunges of k ₂ at W1–W3 compared with the result from ring-shear tests.

Although at Wedron fabric strengths and azimuths do not change abruptly across the contact between the two till faces, fabric plunges do (Fig. 10c). Within the lower till facies, k ₁ fabrics generally plunge up-glacier, consistent with results from Friday 3. However, in the upper till facies, k ₁ fabrics plunge uniformly down-glacier at an average of 21°. Sand grains indicate a similar change in fabric plunge, although the change to down-glacier plunging fabrics occurs high within the upper facies, rather than across the contact with the upper facies (Fig. 10d). As at Friday 3, plunges of k ₂ cluster around 0 (mean = 1.0°), vary with depth and at some depths are large (Fig. 10e); k ₂ plunges do not vary systematically across the facies contact.

If, as at Friday 3, we use directions of principal magnetic susceptibility to determine the orientation of shear planes, assuming that simple shear dominated strain, then most of the inferred shear planes, unlike at Friday 3, dip at angles in excess of 20° (Fig. 8). Many of the largest dips (>40°) are in the down-glacier direction and in the upper till facies: a result of the down-glacier k ₁ plunges in that facies (Fig. 10c).

5.3. Fabric shape

Thus far, in discussing fabric strengths, we have considered only S ₁ values, but there can be additional information contained in S ₂ and S ₃ values (e.g. Reference BennBenn, 1994; Reference Bennett, Waller, Glasser, Hambrey and HuddartBennett and others, 1999). Our focus here is the AMS data, since they are fully three-dimensional (there would be no point in studying S ₂ variation for the two-dimensional sand-grain data because in that case, S ₁ = 1 − S ₂). If S ₁ values based on k ₁ orientations for all depths at both Friday 3 and Wedron are plotted against S ₃ values for these orientations, 49% of the values plot within the 95% confidence limit for fabrics attained at strains that exceed the critical value (7–30) (Fig. 11a). This confidence limit was determined from ring-shear data in which S ₁ and S ₃ values were regressed against known strain magnitudes. The ternary diagrams of Reference BennBenn (1994) provide additional insight into fabric strengths by including all of the principal eigenvalues. Again, 49% of the data fall within the 95% confidence interval for fabric shapes attained at strains that exceed the critical value (Fig. 11b).

Fig. 11. (a) Bivariate and (b) Reference BennBenn (1994)-style plots based on k ₁ orientations. Numbers next to experimental values indicate the strain at which the fabric was measured. Till with fabric strengths that plot outside the shaded regions has less than a 5% chance of having been sheared to the critical strain.

5.4. Comparison of directions of k ₁ and sand-grain fabrics

Comparison of directions of k ₁ and sand-grain fabrics indicates that these two methods of fabric characterization yield results that are correlated but with significant variability (Fig. 12a and b). The average angular distance along great circles between k ₁ and sand-grain principal eigenvectors is 36.6°, with a standard deviation of 26.8°.

Fig. 12. Fabrics directions (V ₁) based on k ₁ orientations (black triangles) and sand-grain orientations (circles) on lower-hemisphere stereo plots. Sand-grain data reflect measurements in horizontal and vertical thin sections as described in the text.

5.5. Bulk densities and pebble fabric

At Wedron there is a clear difference in bulk density between the lower till facies (2.0 ± 0.06 g cm⁻³) and upper till facies (2.3 ± 0.05 g cm⁻³). Bulk-density values at Friday 3 are the same as those of upper facies at Wedron (2.3 ± 0.04 g cm⁻³) and relatively uniform throughout the section (Reference ThomasonThomason, 2006, p. 136). Pebble fabrics, measured over very broad areas (∼40 m²) owing to the sparseness of pebbles, plunged to the northeast at both locations (22°) and were weaker (S ₁ = 0.60 at Wedron; S ₁ = 0.55 at Friday 3) than pebble fabrics reported at Wedron by Reference Johnson and HanselJohnson and Hansel (1990) (S ₁ ∼0.7).

6.1. Fabric strength and direction

Large, relatively smooth depth variations of S ₁ values from k ₁ and sand-grain fabrics indicate that strain magnitude in the Batestown till was insufficient over some depth ranges to be consistent with the bed-deformation model (Figs 5, 6, 8 and 9). Fabric-strength indicators that are less than the lower 95% confidence limits of Figures 5, 6, 8, 9 and 11 indicate a less than 5% chance that these till samples were sheared to the critical shear strain (7–30). Thus, 25–100% of a given profile and 51% of the total number of till samples have fabrics too weak to reflect shear to even moderate strains. These conclusions are essentially the same, regardless of whether we consider AMS fabric strength based only on S ₁ eigenvalues (Figs 5 and 8) or on fabric indicators that involve the other eigenvalues (Fig. 11).

Significant zones in the till, however, do have fabrics sufficiently strong to be consistent with strains greater than or equal to the critical value. For example, k ₁ fabric strengths from 0.5 to 1.3 m at Friday 3 (Fig. 5) and the lowest 75% of the two facies at Wedron (Fig. 8) indicate that such strains could have been attained or exceeded. With the exception of profile W3, sand fabrics also point to such strain magnitudes, although over more limited ranges of depth (Figs 6 and 9). The observation that in six of the seven fabric profiles of Figures 5, 6, 8 and 9, S ₁ values approach closely but do not exceed the steady-state k ₁ and sand-grain S ₁ values from ring-shear experiments indicates that the experiments seem to have accurately predicted the maximum fabric strengths that were developed subglacially.

Nevertheless, in inferring strain magnitude from these data, several caveats apply. Fabrics necessarily provide information about only the most recent deformation to have set the fabric. Similarly, although we have no evidence for post-shear disturbance of fabric, we cannot preclude it. Most importantly, the strong fabrics that fall within the confidence limits for till sheared to the critical strain do not strictly demonstrate that such strains occurred. For example, if fabric was partly inherited from oriented particles in basal ice upon lodgment, then the critical strain required to reach the steady-state fabric strength would have been smaller than in the experiments. Moreover, the critical strain determined in ring-shear experiments (7–30) is less than that required by the bed-deformation model (>100). Thus, in the absence of post-shear disturbance, weak fabrics rule out the high strains of the bed-deformation model, but strong fabrics do not demonstrate that such strains were attained (Reference Thomason and IversonThomason and Iverson, 2006; Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008; Reference Iverson, Hooyer, Thomason, Graesch and ShumwayIverson and others, 2008).

Both k ₁ and sand-grain fabrics indicate that shearing direction changed systematically up-section within the till: from approximately east–west to north–south at Friday 3 (Fig. 7a) and in the reverse direction at Wedron (Fig. 10a and b). This observation is not consistent with the till layer shearing simultaneously over its full thickness. Reference Carlson, Jenson and ClarkCarlson and others (2004) reached a similar conclusion for the underlying Tiskilwa till based on depth variation of pebble fabric azimuths, although depth resolution of that study was limited to 1.0 m. The depth scale of fabric azimuthal variation in our study indicates simultaneous deformation over thicknesses of no more than a few decimeters.

Shear-plane orientations inferred from directions of principal susceptibilities at Friday 3 indicate some depth ranges where shear planes were essentially horizontal (0.3–1.5 m, 2.3–2.7 m) and hence consistent with bed-parallel simple shear (Fig. 5). These depth ranges correspond to those where both k ₁ and k ₂ plunges are close to those expected from ring-shear tests (Fig. 7b and c). Cluster of k ₁ and k ₂ plunges about means that agree with the results of the ring-shear tests suggests again that the experiments provide a valid basis for field interpretations.

Deviation of k ₁ and k ₂ plunges from ring-shear values over other depth ranges at Friday 3 (Fig. 7b and c) and more conspicuously at Wedron (Fig. 10c–e) must reflect deviations from bed-parallel simple shear. These deviations could be interpreted, as expressed in the stereonets of Figures 5 and 8, as simple shear inclined to the horizontal, or as significant components of pure shear superimposed on bed-parallel simple shear. A well-established precept of structural geology is that any state of strain can result either from pure shear or from simple shear inclined at an appropriate angle (Reference Hobbs, Means and WilliamsHobbs and others, 1976, p. 32). Thus, for example, k ₂ plunges in Figure 7c that deviate significantly from zero can be interpreted either as shear that occurred in a plane with a large transverse dip component or as bed-parallel simple shear with transverse shortening superimposed on that shear. More enigmatically, at Wedron there is an abrupt shift in k ₁ plunge from largely up-glacier (east) in the lower till facies to uniformly down-glacier (west) in the upper till facies (Fig. 10c and d). Inferred shear planes in the upper facies, therefore, tend to dip steeply down-glacier. This could reflect a major deviation from simple shear, associated perhaps with longitudinal shortening or till accretion from basal ice on a paleoslope inclined down-glacier (personal communication from D. Evans, 2008).

An ancillary objective of this study was to compare k ₁ fabrics with sand-grain fabrics. There is considerable correspondence between k ₁ and sand-grain fabric azimuths (Fig. 12), in agreement with the experimental finding that k ₁ fabrics align in the direction of the long axes of particles (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008). However, the correspondence is far from perfect. We do not know the reason for this difference in k ₁ and sand-grain fabric orientations but suspect that it is due primarily to some combination of the measurement uncertainty, inherent variability, and two-dimensionality of sand-grain fabrics. Optical measurement of sand-grain long axes in thin sections involves human error and subjectivity that is largely absent in the measurement of AMS principal susceptibilities (Reference Iverson, Hooyer, Thomason, Graesch and ShumwayIverson and others, 2008). In addition, sand-grain fabric signals will be inherently more variable than k ₁ signals because a single k ₁ measurement reflects the volume integration of the effects of many aligned magnetic grains in a sample, rather than the orientation of a single grain (Reference Hooyer, Iverson, Lagroix and ThomasonHooyer and others, 2008). Also, additional uncertainty is introduced by the difference between true sand-grain orientations and apparent orientations measured in thin sections.

6.2. Interpretation of till deformation

These fabric measurements leave little doubt that the Batestown till was deformed (e.g. Reference Johnson, Hansel, Mickelson and AttigJohnson and Hansel, 1999), but the picture that emerges is that this deformation commonly deviated from the high strains required of the bed-deformation model and the deep, simple shear of the bed that is frequently assumed. The up-section shift in fabric azimuth at both locations (Figs 7b and 10a and b) indicates that the full till thickness did not shear simultaneously. Rather, till likely sheared in a layer a few decimeters thick near the glacier sole, as till progressively accreted to the bed from ice and the shear direction in the till changed with time. Variability in fabric strength with depth (Figs 5, 6, 8 and 9) therefore suggests temporal variability in the magnitude of bed deformation near the glacier sole: strain during some periods equaled or exceeded the critical strain required for steady-state fabric development and during other periods fell short of the critical strain. In addition, bed-parallel simple shear of the bed did not occur over many depth ranges, as indicated by inferred shear planes that dip substantially (Figs 5 and 8). This implies either that shear planes were indeed inclined or that there were significant components of pure shear. In either case, there were apparently three-dimensional effects that locally caused major deviations from simple shear. This inferred state of strain is consistent with till deformation focused in three-dimensional patches (e.g. Reference Piotrowski and KrausPiotrowski and Kraus, 1997), either causing shear zones to dip near the margins of patches or alternatively causing significant local shortening or extension of the till layer.

What caused shear direction to change with time? Two timescales of variation are evident: the secular azimuthal variation across the till thickness as indicated by the slopes of the regression lines in Figures 7a and 10a and b, and the higher-frequency azimuthal variability about the regression lines over depth ranges of decimeters. The secular azimuthal shift likely reflects how flow direction changed as the Lake Michigan lobe retreated. Moraine geometry, which indicates the lobe broke into smaller sub-lobes during retreat, provides support for this hypothesis at Friday 3 (Fig. 13a–c). The terminal moraine of the Batestown advance, the Eureka moraine, is oriented north–south, consistent with ice flow from the east during the Batestown maximum and the associated east–west-trending fabric near the base of the Friday 3 section (Fig. 7a). Inboard of the Eureka moraine and within 10 km of Friday 3, the recessional Arispie moraine is oriented east-southeast, consistent with the north-northeast fabric azimuth near the top of the section (Fig. 7a). Near Wedron, on the other hand, recessional moraines are oriented closer to north–south, such that, during glacier retreat in that area, flow would have been from approximately the east, consistent with fabric azimuth near the top of the Wedron section (Fig. 10a and b). The up-section northeast-to-east rotation of the fabric at Wedron could thus be explained if prior to retreat in this area flow was dominantly from the northeast, as seems likely given the general southwest flow of the lobe out of the Lake Michigan basin (Fig. 1). The higher-frequency azimuthal variations superimposed on the secular shifts are consistent with heterogeneous shallow deformation: temporally variable, patchy deformation of the bed would have resulted in zones of till convergence and divergence and associated shifts in shear direction as till accumulated and the geometry of deforming patches changed. Some of this high-frequency variability may be random noise (e.g. Reference Benn and RingroseBenn and Ringrose, 2001), but some is not. For example, from 1.7 to 2.6 m depth at Friday 3, variability about the secular shift in azimuth is nearly identical for both k ₁ and sand-grain azimuths (Fig. 7a).

This picture of shallow heterogeneous deformation during till accretion from basal ice agrees with the interpretations of Piotrowski and colleagues based on diverse field observations of basal tills of the Fennoscandian ice sheet (Reference Piotrowski and KrausPiotrowski and Kraus, 1997; Reference Larsen and PiotrowskiLarsen and Piotrowski, 2003; Reference Larsen, Piotrowski and KronborgLarsen and others, 2004, Reference Larsen, Piotrowski and Menzies2007; Reference Piotrowski, Larsen and JungePiotrowski and others, 2004). Heterogeneous, temporally variable bed deformation was also stressed by Reference Evans, Phillips, Hiemstra and AutonEvans and others (2006) in their review of basal till characteristics and genesis. This interpretation is also supported by measurements beneath modern soft-bedded ice masses. These measurements indicate that steep spatial gradients in water pressure at the bed are common (e.g. Reference Murray and ClarkeMurray and Clarke, 1995; Reference Engelhardt and KambEngelhardt and Kamb, 1997; Reference Fischer and ClarkeFischer and Clarke, 2001), and measurements of till deformation and sliding indicate that basal water pressure modulates the extent of bed deformation (Reference Blake, Fischer and ClarkeBlake and others, 1994; Reference Iverson, Hanson, Hooke and JanssonIverson and others, 1995, Reference Iverson, Baker, Hooke, Hanson and Jansson1999, Reference Iverson2003, Reference Iverson2007; Reference Fischer and ClarkeFischer and Clarke, 1997, Reference Fischer and Clarke2001; Reference Fischer, Clarke and BlatterFischer and others, 1999; Reference Boulton, Dobbie and ZatsepinBoulton and others, 2001) by regulating both till shear strength and the coupling between till and the glacier sole. Thus, heterogeneous deformation of the bed, with only patches of the bed deforming, is expected (e.g. Reference Fischer and ClarkeFischer and Clarke, 2001) and is indeed more consistent with subglacial measurements than the spatially pervasive simple shear of the bed that has often been assumed.

Fig. 13. (a) Maximum extent of the Lake Michigan lobe during the Batestown advance. (b) Possible configuration of the lobe during its retreat from the Eureka end moraine. (c) Final moraine configuration of the lobe. End moraine of the Batestown advance is indicated by dark gray. All other Wisconsin episode moraines are light gray.