Recommendations concerning the identification of Carex retroflexa and Carex texensis (Cyperaceae; section Phaestoglochin Dumort)
Downer, Robert G
A statistical analysis of characteristics of Carex retroflexa Willd. and Carex texensis (Torr.) L. H. Bailey is provided. This analysis confirms the distinctiveness of these two species, which some authors have considered as a single taxon in the past. Analysis reveals Carex texensis perigynia average less than 1.3 mm wide, with a spongy portion length of less than 1.1 mm. Carex retroftexa perigynia average 1.3 mm or more wide, with a spongy portion 1.1 mm long or longer.
The genus Carex L. (Cyperaceae) causes headaches for even professional botanists. Kartesz (1999) reports that the genus contains 484 species and 634 taxa (including varieties and subspecies) in North America, north of Mexico. Identification of many species depends heavily on reproductive characters.
Section Phaestoglochin Dumort. contains 16 species in the southeastern United States. This paper will address the identity of C. retroflexa Willd. and C. texensis (Torr.) L. H. Bailey (Figure 1), and will provide a key for determining specimens of these two species. These two species differ from other species in this section in three ways. They lack serrulate teeth on the beak of the perigynia. The have relatively narrow leaves, 3 mm or less wide. They also have corky or spongy material only on the ventral side of the perigynia.
During Hyatt’s (1998) study of the genus Carex in Arkansas, A. A. Reznicek identified duplicate collections and suggested hints for identification of species. In 1993, Reznicek (pers. comm.) commented that “once learned” C. retroflexa and C. texensis could be told apart “at a glance.” Armed with this information, we have attempted to learn the visual cues and habitat preferences of these species in order to facilitate their identification. These observations, supported by various comments of Reznicek, provide the basis for the following descriptions. Field experience is based on work in the southeastern United States, so may not accurately reflect conditions further north.
The two species do occur together in several settings, including along streams, as well as in disturbed habitats such as lawns, roadsides, parks, etc.
MATERIALS AND METHODS
The authors studied collections from MICH, LAF, and from Hyatt’s personal herbarium. Each specimen was initially examined in order to reject those with immature perigynia for use in statistical analysis. An attempt was made to select specimens distributed throughout the geographic range of the species involved. When a specimen was reviewed for potential study, and its state or province of collection was already well represented by other previously measured collections (two to four sheets) the specimen was rejected for inclusion in statistical analysis. This method resulted in a study covering most of the range of both species. Full sized perigynia from the middle or lower (but not base) portion of the spikelets were selected for measurement.
Initial observations led to selection of the following characters being measured for this study: perigynium length, perigynium width, length of the spongy portion of the perigynium, width of the widest leaf at its widest point, length of the longest bract, length of the tallest culm. The above data were recorded by individual measurements; averages were used only in production of a key to the species. The spongy portion can be seen ventrally at the base of the perigynium. A series of somewhat parallel lines are usually visible on this spongy layer (Figure 1).
Perigynium length was measured from the base of the perigynium to the tip of the teeth. All measurements for all characteristics were taken to the nearest 0.05 mm.
Culm length varies widely from specimen to specimen, by date of collection, and by location of collection, for both species. We used the longest culm available. Season of collection was ignored during the data collection process.
With species identification as the main objective, analysis was conducted in a manner consistent with the process of identification of a single plant by one observer. As a result, the data were reduced to a single measurement for each perigynium.
Sample mean, median, standard deviation and range per plant were investigated in the prediction techniques and none exceeded the prediction success of the sample average. Using sample averages eliminated any difficulty in analysis that would result from having five perigynia measured for some plants and ten for others.
In discriminant analysis (Johnson and Wichern 1998), a criterion is developed to classify an observation into one of the possible groups or populations, which in our case are the classifications of Carex retroflexa (n = 30) and Carex texensis (n = 30). The sample means, variances and correlations for the predictor variables define statistical distance D^sub j^^sup 2^ for each observation. We initially assume the probability of classification into either species is 0.5. After observing the set of predictor variables x, the probability of a plant being from group one as opposed to group two is estimated to be exp(-0.5D12(x))/(exp(-0.5D^sub 1^^sup 2^(x)) + exp(-0.5D^sub 2^^sup 2^(x)). The probability of falling into group two is 1 minus this probability. The higher of the two estimated probabilities defines the classification for an observation. In the cross-validation prediction of plant i, the discriminant function is computed using all plants other than observation i, and the value of the discriminant criterion changes for each plant. Discriminant analysis was performed using PROC DISCRIM in Version 8.0 of SAS (SAS 1999).
In logistic regression (Collett 1996), one models the probability of observing one of two possible responses (species in this case). Rather than the continuous response y as in standard linear regression, the log odds, (the logarithm of the probability of observing response one rather than response two) is modeled as a linear function of the k predictor variables X^sub ?^..X^sub k^. Hence, we are directly modeling the probability of either species as a linear function of the plant characteristics (perigynium width, etc.). Significance of a given predictor indicates whether or not it significantly impacts whether the species is C. retroflexa or C. texensis. This effect of a given variable is considered in the presence of all others. Cross-validation prediction of observation i involves fitting the model without plant i, and predicting the species for plant i from its characteristics using this fitted model. All logistic regression models were fit using PROC LOGISTIC in SAS version 8.0 (SAS 1999).
Exploratory analysis revealed that the distribution of variables for each species differed most for the measurements of the perigynium width and length of the spongy layer. The entire Carex texensis perigynium width distribution is less than the minimum of the corresponding C. retroflexa distribution (Table 1). The third quartile of the C. texensis spongy layer length distribution is less than the minimum of the corresponding C. retroflexa distribution. In fact, only three out of 30 C. texensis specimens had an average spongy layer length exceeding the minimum sample average (1.17) of C. retroflexa. Considerable species overlap occurs in the distribution of each of the other variables (Table 1). For example, leaf width was not a dependable character, despite the fact C. retroflexa sometimes has leaves wider than the widest leaves of C. texensis.
Preliminary exploratory analysis suggested a focus on the perigynium width or the spongy layer length for the identification of the species. Prediction techniques verified this result. With all predictors included, discriminant analysis predicted 29 of 30 C. retroflexa correctly and 30 out of 30 C. texensis correctly (overall 59/60 = 98.3%). The prediction performance of 59/60 was matched when the perigynium width alone was used as a classifying variable (i.e., other the input variables including spongy layer length were not required). This predictive ability was matched by a discriminant function with both the perigynium width and the spongy layer length as input variables. Perigynium width and the length of the spongy layer had a correlation of 0.77. The separation of the species according to these two variables is clear from Figure 2. The single classification error is also evident. This C. retroflexa specimen was very unusual in its shorter spongy layer length and in its narrower than typical perigynium width of 1.5 mm. A classification error was quite possible, but review of the specimen confirmed the presence of a subjective characteristic, which clearly shows this specimen is C. retroflexa. That is, the perigynia from this plant has darkened striations on the ventral side of the spongy layer which is typical of C. retroflexa. These striations appear on mature material of this species, but are not always obvious in immature material. They do not appear on C. texensis perigynia. Spongy layer length on its own was able to correctly predict 29 out of 30 C. retroflexa and 29 out of 30 C. texensis.
Logistic Regression gave complimentary results to the discriminant analysis (without any distributional assumptions for the input variables). With all plant characteristics as predictors included in the model, only the perigynium width was significant at the 0.05 level (p = 0.0107). In cross-validation prediction with only the perigynium width as a predictor, the exact classification results of the discriminant analysis were obtained (29 out of 30 C. retroflexa, 30 out of 30 C. texensis).
We recommend a primary focus on the width of the perigynium for the species identification of any single plant. As a secondary identifying characteristic, we recommend using spongy layer length. Exploratory data analysis, discriminant analysis, and logistic regression all consistently suggest the importance of these characteristics. Perigynium width on its own may be sufficient for identification. If the average perigynium width is less than 1.3 mm, then it is most likely Carex texensis. Additionally, if the average spongy layer length is less than 1.1 mm then it is very likely C. texensis. The combination of these two measurements will be very accurate in species identification.
These statistical recommendations are based on either 5 or 10 observations per plant, and input to the discriminant analysis was reduced to the sample average of each characteristic. In a separate analysis, a random sample of two perigynia was taken from the full data set and the average of these measurements for each plant were subsequently used in the discriminant analysis. Twenty-four out of thirty C. retroflexa were correctly identified while 23 out of 30 C. texensis were correctly identified. Hence some predictive success is possible with fewer observations per plant. However, as seen from the complete analysis, multiple measurements per plant are advantageous.
Several concerns arose during data collection. The primary concern is the selection of mature material. Most herbaria house morphologically unidentifiable immature collections of Carex. Collectors of C. retroflexa and C. texensis should collect mature material with fully formed perigynia, when possible. In collections with barely mature perigynia, dried perigynia tend to shrink and wrinkle slightly making the edge of the spongy layer indistinct. Mature perigynia do not exhibit this shrinking, and have a clear line marking the end of the spongy layer.
A spongy or corky material fills the base of the perigynia in these and related Carex species. The spongy layer runs from the base of the perigynium to about 0.25-0.60 times the length of the perigynium itself. The distal end of this spongy layer forms a straight or slightly arched line across the ventral face of mature perigynia. Measurements of various observers might vary by about 1 mm (measuring from a point along this arch to the perigynium base as described above) depending on the point along the arch selected for measuring. In such cases (where one might wonder exactly where to measure the spongy layers end at the middle of the perigynium), an average can be safely used. For example, if the measurement of the length of the spongy layer appears to be about 1.05-1.15 mm, depending on exactly where along the edge of the spongy layer one measures, it is safe to use a medial figure of 1.1 mm for that perigynium.
A different problem exists with overly mature material. The perigynia are lost rapidly after they mature, so specimens exist in herbaria with no perigynia left. These specimens may be morphologically unidentifiable to species with any degree of certainty. Such specimens obviously could not be used in this study. Detached perigynia, which had been placed in envelopes were not used in this study, so we were certain the measured perigynia came from the specimen at hand. We also realized that the culms of plants tend to elongate as the growing season progresses. Collections from early in the season tend to have shorter culms than late season collections (pers. obs.).
Distributions of the sample averages of some plant characteristics deviated slightly from normality. This issue clearly did not impact the predictive ability of the discriminant function. Normality of the input variables was not required for the logistic regression, and similar results were obtained. It is also recognized that a repeated measures logistic model could be applied to the raw data (i.e., on the original measurements). Such an analysis would not lose any information on the variability within experimental units. Practical simplicity and the success of the techniques with the sample average suggested that such an investigation was unnecessary.
Some authors have used leaf width to differentiate these species, such as Diggs et al. (1999), Jones (1994), Mackenzie (1935), and Yatskievych (1999). While field observations and our data confirm that C. retroflexa sometimes has wider leaves, the difference in width is not substantially different from leaves of C. texensis (Table 1). We recommend against using leaf width as a key character. For example, in our sample we found C. texensis to have leaves up to 2.5 mm wide and C. retroflexa leaves up to 3.4 mm wide with considerable overlap in the lower end of the range of widths (Table 1).
We suggest the following key to these two species:
Average perigynium width less than 1.3 mm (measure 5-10 perigynia); average spongy portion of the perigynium less than 1.1 mm long . . . . . . . . . C. texensis (Torr.) L. H. Bailey
Average perigynium width equal to or greater than 1.3 mm; spongy portion of the perigynium 1.1 mm long or longer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. retroflexa Willd.
Carex retroflexa tends toward more upland habitats than C. texensis; habitats for the two species do overlap, especially in disturbed areas. Carex retroflexa may occur in riparian zones, but it occurs more often in uplands. It ranges through the Ozark and Ouachita regions, the rolling hills of Louisiana, and similar habitats across the eastern United States. Carex retroflexa ranges from Texas to Kansas to Iowa, east to Indiana and Michigan, north to Quebec and Maine, and south to the gulf coast states (Kartesz 1999).
Carex texensis has a strong affinity for riparian habitats, from small streams to riverine systems. Waller (1929) indicated that C. texensis becomes a major component of the grassy vegetation of lawns. Indeed, the species appears as the dominant vegetation in shaded lawns on clay soils in some areas and at localized sites in the lower Mississippi valley and related drainages. It is definitely absent on dry rocky hillsides in undisturbed sites inhabited by C. retroflexa in the Ozarks. Carex texensis has a similar range to that of Carex retroflexa, but extends north only to Nebraska, Missouri, and through the Ohio River Valley states to New York and also occurs in California (Kartesz 1999) with the northern most records and California records representing adventive occurrences.
All specimens are housed at the MICH unless otherwise noted.
Carex retroflexa Willd. Canada: ONTARIO. Essex County: Harrow vicinity, 2 km NE, wet-mesic Carya-Quercus-Fraxinus forest with scattered grassy and sedgy openings, clayey soils with occasional sandy pockets, 19 Jun 1983, Reznicek et al. 7156. Kingsville vicinity, 6 km W on S side of Hwy. 18, Gosfield S. Twp., Arner Point Conservation Area, 22 Jun 1984, 7156 Oldham 4257.
U. S. A.: ARKANSAS. Overcup community, 4.2 km S of Solgohachia Post Office on Ark. Hwy. 9, T7NR16WS27NW4, 29 Apr 1992, Hyatt 4806 (peh, personal herbarium of Philip E. Hyatt, willed to MO). Lee County: T2NR4ES30SE4NW4, upland deciduous woods, loess soil, 09 Jun 1992 Hyatt 4634.39 (peh). Marion County: Buffalo National River, T18NR13WS31SW4SW4, old field above Buffalo River bluffs, silt loam on limestone, 20 May 1993, Hyatt 5461 (peh). Searcy County: Buffalo National River, T16NR15WS4SW4NE4, 0.6 km S of Buffalo River on Ark. Hwy. 14 on N roadside, mixed oak woods, seepy creek, regenerated area (old fire?), on limestone, 17 Jun 1993, Hyatt 5677 (peh). Yell County: T4NR21WS13/24 border, 5.8 km NW of Perry Co. on Ark. Hwy. 7, ridgetop roadside park, shortleaf pine, mixed oak (stunted) on silty soil, 20 May 1998, Hyatt 8049 (peh). DELAWARE. New Castle County: Wilmington, no date, Canby s.n. GEORGIA. Putnam County: Eatonton, 19.3 km E, moist rich woods, 2 Apr 1938, Pyron 2386. KENTUCKY. Carter County: Carter Caves State Park, parking area by Cascade Trail, in earth filled crevices in limestone outcrop, 20 May 1991, Brunton & McIntosh 10,298. Lyon County: Lake City, ca. 4.8 km E and off of Hwy. 641, barren-like roadside or mesic to subxeric woodland, 28 Apr 1994, McKinney 6144. ILLINOIS. Alexander County: Shawnee National Forest, Ozark Hill Prairie Research Natural Area, Mill Creek quad, upland disturbed open site, 29 Apr 1992, Phillippe 19,910. Saline County: Harrisburg, about 11 km ESE on Shawnee National Forest, T9SR7ES34SE4, common along creek with Plantago cordata in deciduous woods, 20 May 1992, Hyatt 4352 (peh). Union County: Atwood Ridge Research Natural Area, T12SR2WS4SW4NE4, Jonesboro quad, dry upland woods, 7 May 1991, Phillippe & Ketzner, 18,670. LOUISIANA. De Soto Parish: Mansfield, ca. 3.2 km ENE, rather dry hillside with shagbark hickory and pines, 12 Apr 1968, Thieret 28,565 (LAF). MASSACHUSETTS. Hampshire County: Mt. Holyoke, Jun 1872, Jesup s.n. MICHIGAN. Washtenaw County: Ann Arbor, Mitchell-Scarlett Woods south of Scarlett Middle School, dry oak-hickory woods, 19 Jun 1995, Walters 494. MISSISSIPPI. Tallahatchie County: Tutwiler, S at roadside park at junction of U.S. Hwy. 49E and 49W, T25NR2WS29, in open areas and under widely scattered Quercus spp. on well-drained sandy loam, 11 Apr 1990, Bryson 8780. Lowndes County: Steens vicinity, 4.0 km N, T17SR17WS10, on Vaughn Robertson Rd, steep slopes under mixed deciduous forest, rich, sandy soil with thick leaf litter, 20 Apr 1986, Bryson 4219. MISSOURI. Madison County: Cedar Bottom Church vicinity, ca. 1.3 km N., or ca. 1.3 km N. of Co. Rd. 508, rich mesic mature woodland, 8 Jun 1994, McKenzie 1408. NEW JERSEY. Mercer County: Somerset, dry wooded bank of streamlet, 4 Jun 1933, Hermann 4278. NEW YORK. Tompkins County: ravine between Renwick and McKinneys, dry open woods, 17 Jun 1916, Wiegand 6005. NORTH CAROLINA. Rockingham County: Price, about 7.5 km WSW, xeric, thin hardwoods on lower, south facing slope along Buffalo Creek, ca. 0.3 km W of confluence with Mayo River, 2 May 1994, Widboldt 8886. OHIO. Eire County: Old Woman Creek Sanctuary, edge of woodland trail, 31 May 1985, Oldham & Allen 4876. OKLAHOMA. Bryan County: Colbert vicinity, ca. 8 km SE on River Road, 1.9 km E. of Chickasaw Rd at small stream (Hendrix P. O. vicinity, 4.0 km NW on River Road), elm-hickory-deciduous woods adjoining pastures, highly disturbed area on silty soils, 20 Apr 1998, Hyatt 7965. Marshall County: Kingston, ca. 8 km S & 0.5 km E, on N shore of Lake Texoma at Cross Point Methodist Camp, T7SR6ES19SE4SW4, grassland and woodland areas, loam over limestone, 19 Apr 1997, Magrath, et al. 19,775. PENNSYLVANIA. Lehigh County; Hosensack Post Office, about 0.6 km WSW, rather open grassy scrubby slope along NW side of streamlet, 30 May 1919, Pretz 9616. TEXAS. Houston County: Crockett vicinity, 2.4 km SW on Texas 21 from its junction with Loop 304 at South Fork of Spring Creek, Open to wooded mesic roadside with reddish sandy loam soil, 14 Apr 1999, Jones & Wipff 6423, Sabine County: Toledo Bend Reservoir vicinity, just south of Shelby Co. line NE of N end of U.S. Forest Service Rd. 121 N of Bennett’s Cemetery in dry-mesic to mesic slope and ravine forests, dominated by upland mixed hardwoods & pine, with few beech, 12 Apr 1988, Orzell & Bridges 6209. VIRGINIA. Fauquier County. Big Cobbler Mountain, moist thicket at foot of mountain, 17 May 1937, Allard 2770.
Carex texensis (Torr.) L. H. Bailey. U.S.A.: ALABAMA. Madison County: Huntsville, lawns along Mastin Lake Road, 23 May 1983, Bryson 3536. Tuscaloosa County: Tuscaloosa vicinity, ca. 9.7 km NE, Rocky Branch Recreation Area, E of entrance, off of Ala. Hwy. 216, tributary stream to Black Warrior River, 11 Apr 1992, Horn & Hill 5051. ARKANSAS. Desha County: Reed vicinity, about 4.8 km S, (or 11.2 km N of Chicot Co. line on U.S. Hwy. 165, T12SR3WS23, deciduous woods roadside in dense brush and heavy shade, along road and railroad rights-of-way; silty soils, 24 May 1996, Hyatt 7110 (peh). Izard County: Sylamore, roadside ditch by railroad under Ark. Hwy. 9 bridge at White River, ruderal old town site, 27 May 1993, Hyatt 5551 (peh). Marion County: Buffalo National River, from Ark. Hwy. 14 go E about 150 m on S side of riverbank, deciduous woods, especially in floodplain, sandy soil on limestone, 14 May 1993, Hyatt 5333 (peh). Sevier County: DeQueen, several km E, at the junction of U.S. Hwys. 70 & 71, roadside park, on clay soil, mixed deciduous floodplain woods, Cossatot River floodplain at T8SR30WS28SE4NE4, 23 Apr 1995, Hyatt & Sheila Hyatt 6316 (peh). INDIANA. St. Joseph County: Notre Dame, top of Grotto of our Lady, 3 Jun 1953, Herbert 4182. Hendricks County: Plainfield vicinity, rest area at mile 65 on I-70 eastbound, 1.6 km SW of Rt. 267 to Plainfield, partially shaded mesic sedge meadow under widely spaces Carya, 24 May 1993, Rothrock 2949. KENTUCKY. Edmonson County: Mammoth Cave National Park, campground by headquarters, Red Maple-Hickory-Oak woods with sparse undergrowth on silty sand, 21 May 1992, Brunton & McIntosh 10,303. Greenup County: Edgington, 0.8 km S, W of U.S. Rt. 23, moist, shaded ground at Bethlehem Church, Portsmouth Quad, 3 Jul 1990, Cusick 28,988. Pendleton County: Falmouth, by road 3.2 km ESE of eastern edge, S side of Bishop Ridge Road, 1.1 km E of Marquette Rd, dry-mesic Quercus muehlenbergu-Juniperus-Carya forest, upslope from stream, 3 Jun 1994, Naczi & Thieret 4048. ILLINOIS. Montgomery County: Hillsboro, 340 Church St., lawn, 21 Jun 1982, Colin C-13168. INDIANA. Hendrix County: Plainfield vicinity, rest area near Plainfield and 1.6 km SW of route 267 exit on I-70 eastbound at mile marker 65, partially shaded mesic sedge meadow under widely spaced Carya, 24 May 1993, Rothrock 2949. MARYLAND. Baltimore County: Towson vicinity, east of town, Foppa and Providence roads, abandoned field, 16 May 1957, Balters 913. MISSISSIPPI. DeSoto County: Walls vicinity, north at north junction of highways U.S. 61 and Miss. 602, E of U.S. Hwy. 61, T1SR9WS22SW4, mesic W facing slopes, hardwood forest or along woods edge, loess bluffs region, 16 Apr 1994, Bryson 13,346. Marshall County: S of Holly Springs, 8 km S of the junction of highways U.S. 78 and Miss. 7, T5SR3WS1, edge of scattered oak-pine woods, sandy soil, 21 Apr 1995, Bryson 14,802, Bryson & Newton 8780. MISSOURI. Wayne County: T27NR5ES15NE4, open canopy above disturbed sinkhole pond ca. 0.6 km W of Co. Rd. 406, 20 May 1994, Brant 2787. Benton County: Big Buffalo Creek Marsh Natural Area, T41NR20WS12SW4NE4, path to fen from entrance, along stream, 26 May 1987, Castaner 9760. NEW YORK. Onondaga County: Syracuse, large cemetery adjacent to Univ. campus, weed, in sandy soil, 8 Jun 1988, Oldham 8161b. NORTH CAROLINA. Orange County: Chapel Hill, south side of Univ. of N. Carolina campus, roadcut through oak-hickory forest, 9 May 1967, Logue 992. OHIO. Ottawa County: South Bass Island in western Lake Erie, grassy lawn surrounding Ohio State University building, 31 May 1985, Oldham 4783. OKLAHOMA. Adair County: Chewey vicinity, ca. 6.7 km S. and W. at Tate Ranch, T18NR24ES7SW4, oak-hickory woods and associated prairie pasture, N facing slope in woods, gravelly soil, 6 May 1990, Magrath, Norman, & Hamilton 17,812, Reznicek 9364. PENNSYLVANIA. Bradford County: Towanda community, Riverside Cemetery’s back edge, in moist shaded lawn, edge of woods, 16 Jun 1994, Naczi 4247. SOUTH CAROLINA. Edgefield County: N side of SC Rd. 230 at Fox Creek, just N of I-20, rich woods with Trillium reliquum, Trillium maculatum, on clay soil, 11 Apr 1992, Nelson, Pittman, & Fairey 12,275. McCormick County: Sumter National Forest, Long Cane District, Folks Analysis Area, immediately NW of junction of S-204 and S. Carolina Rd. 28, 3.2 km W of Edgefield Co. line, shady woods in springy area at headwater of small unnamed tributary to Stevens Creek, delicate patches on boggy ground, with Arisaema quinatum, 18 Apr 1996, Nelson 17,143. TENNESSEE. Cannon County: Short Mountain Sanctuary (ca. 2.4 km NE of Sugar Tree Knob Church and 4.8 km W of Tenn. Hwy. 146), mesic calcareous ravine forest, on N side of North Short Mountain Road, 3 May 1989, Orzell & Bridges 9445. Humphreys County: Cuba Landing, N of I-40, just north of Cuba Landing Marina, E of Tennessee River, edge of mesic wooded slopes above river, sandy loam soil, 18 Apr 1995, Bryson 14,732. TEXAS. Houston County: Ratcliff vicinity, 13.4 km N (6.8 km NE on Davy Crocket National Forest Rd. 526 from its junction with Rd. 227, mesic hardwood/pine forest on sandy and clay loam soil, creek side and associated dry uplands, 10 Apr 1990, Jones 4324. San Augustine County: on S side of Texas Hwy. 21, 0.2 km E of PaIo Gaucho Bayou, ca. 1.0 km W of Sabine Co. line, mesic calcareous slope forest above seepage swamp forest, 11 Apr 1989, Orzell & Bridges 9160. Tyler County: Steinhagen Lake, E facing slope (E of Hooper Drive, 0.5 km NE of road 92 at a point 1.4 km N of intersection with road 1746 at Town Bluff), mesic calcareous slope forest with spring seeps, 4 Apr 1989, Orzell & Bridges 8987.
COLLETT, D. 1996. Modeling binary data. Chapman Hall, London, England.
DIGGS, G.M. JR., B.L. LIPSCOMB, and R.J. O’KENNON. 1999. Shinner’s and Mahler’s illustrated flora of north central Texas. Botanical Research Institute of Texas, Fort Worth, Texas.
HYATT, P.E. 1998. Arkansas Carex (Cyperaceae): a briefly annotated list. Sida 18:535-554.
JOHNSON, R.A. and D.W. WICHERN. 1998. Applied multivariate statistical analysis, fourth edition. Prentice Hall, Upper Saddle River, New Jersey.
JONES, S.D. 1994. A new species of Carex (Cyperaceae: Phaestoglochin) from Oklahoma and Texas; typification of section Phaestoglochin, and notes on sections Bracteosae and Phaestoglochin. Sida 16:341-353.
KARTESZ, J.T. 1999. A synonymized checklist and atlas with biological attributes for the vascular flora of the United States, Canada, and Greenland. First edition. In: Kartesz, J.T. and C.A. Meacham (eds.). Synthesis of the North American flora, Version 1.0. North Carolina Botanical Garden, Chapel Hill, North Carolina.
MACKENZIE, K.K. 1935. Carex. N. Amer. Fl. 18:9-478.
SAS. 1999. SAS/STAT user’s guide, version 8.0. SAS Institute Inc, Gary, North Carolina.
WALLER, A.E. 1929. Origin of cultivated plants as illustrated by a sedge. Ecology 10:415-419.
YATSKIEVYCH, G. 1999. Steyermark’s flora of Missouri. Volume 1, revised ed. Missouri Department of Conservation, Jefferson City, Missouri, in cooperation with The Missouri Botanical Garden Press, St. Louis, Missouri.
Received July 11, 2002; Accepted November 15, 2002.
ROBERT G. DOWNER1 and PHILIP E. HYATT2*
1 Department of Experimental Statistics, Louisiana State University and Louisiana State University Agricultural Center, 161 Agricultural Administration Building, Louisiana State University, Baton Rouge, Louisiana 70803-5606;
2 U.S.D.A. Forest Service, 1720 Peachtree Road NW, Atlanta, Georgia 30309-2417
* Corresponding author’s email address: email@example.com
Copyright Southern Appalachian Botanical Society Sep 2003
Provided by ProQuest Information and Learning Company. All rights Reserved