Danielson Framework for Teaching: Rubric & Scoring

The Danielson Framework for Teaching is an observational evaluation and research architecture developed by Charlotte Danielson in 1996 and updated in 2013 and 2022. It translates complex classroom teaching into a criterion-referenced measurement system. Understanding the mathematical aggregation of its rubrics, its psychometric properties from the Measures of Effective Teaching (MET) project, and the variance structures governing observation reliability is essential for educational researchers and administrators designing defensible evaluation systems.

Definition in one paragraph

The Danielson Framework for Teaching is an observational evaluation architecture that organizes pedagogical practice into four overarching domains—Planning and Preparation, The Classroom Environment, Instruction, and Professional Responsibilities—encompassing 22 distinct components assessed across a four-level ordinal rubric consisting of Unsatisfactory (Level 1), Basic (Level 2), Proficient (Level 3), and Distinguished (Level 4). Grounded in constructivist cognitive theory, the framework establishes a standardized qualitative rubric and quantitative rating continuum used by school systems, state education agencies, and researchers to evaluate classroom performance, guide instructional coaching, and measure teacher practice over time.

Formula and Scoring Model

Transforming qualitative observations into quantitative metrics requires a multi-stage aggregation model. Individual rubric ratings are scored on an ordinal integer scale from 1 to 4, which aggregate into domain subscores and an overall composite evaluation metric.

Let X_ik represent the rubric score assigned to teacher i on component k, where X_ik ∈ {1, 2, 3, 4}. The 22 components span four domains: Domain 1 (Planning and Preparation, K₁ = 6), Domain 2 (Classroom Environment, K₂ = 5), Domain 3 (Instruction, K₃ = 5), and Domain 4 (Professional Responsibilities, K₄ = 6).

The mean domain score D_id across K_d components is:

D_id = (1 / K_d) ∑_{k ∈ Cd} X_ik

The overall composite score S_i applies domain weighting coefficients w_d:

S_i = ∑_d=1⁴ w_d D_id    where    ∑_d=1⁴ w_d = 1.00

Equal weighting sets w_d = 0.25. Observation-weighted models assign higher weights to observable practice in Domains 2 and 3 (e.g., w₂ = 0.35, w₃ = 0.35, w₁ = 0.15, w₄ = 0.15).

Kane and Staiger (2012) in the MET project decomposed score variance via Generalizability Theory:

σ²_total = σ²_T + σ²_L + σ²_R + σ²_E

where σ²_T is teacher variance, σ²_L is lesson variance, σ²_R is rater variance, and σ²_E is measurement error. Single-visit reliability is given by the intraclass correlation coefficient:

ICC(1) = σ²_T / (σ²_T + σ²_L + σ²_R + σ²_E)

For an evaluation design pooling M lessons and R independent observers, the generalizability coefficient G is:

G = σ²_T / [σ²_T + (σ²_L / M) + (σ²_R / R) + (σ²_E / (M × R))]

Worked example with real numbers

Consider an evaluation cycle for Teacher A incorporating classroom observations for Domains 2 and 3 and artifact portfolios for Domains 1 and 4.

Step 1: Component Rubric Ratings

• Domain 1 (6 components): 1a = 4, 1b = 3, 1c = 3, 1d = 3, 1e = 4, 1f = 3. Sum = 20.
• Domain 2 (5 components): 2a = 3, 2b = 3, 2c = 4, 2d = 3, 2e = 3. Sum = 16.
• Domain 3 (5 components): 3a = 3, 3b = 3, 3c = 4, 3d = 3, 3e = 3. Sum = 16.
• Domain 4 (6 components): 4a = 3, 4b = 4, 4c = 3, 4d = 3, 4e = 4, 4f = 4. Sum = 21.

Step 2: Calculate Domain Subscores

• D₁ = 20 / 6 = 3.3333
• D₂ = 16 / 5 = 3.2000
• D₃ = 16 / 5 = 3.2000
• D₄ = 21 / 6 = 3.5000

Step 3: Calculate Composite Scores

Equal weighting (w_d = 0.25):

S_equal = 0.25(3.3333 + 3.2000 + 3.2000 + 3.5000) = 3.3083 ≈ 3.31

Observation weighting (w₁ = 0.15, w₂ = 0.35, w₃ = 0.35, w₄ = 0.15):

S_weighted = 0.15(3.3333) + 0.35(3.2000) + 0.35(3.2000) + 0.15(3.5000) = 3.2650 ≈ 3.27

Step 4: G-Theory Reliability Estimation

Using MET empirical variance parameters (Kane & Staiger, 2012): σ²_T = 0.12, σ²_L = 0.10, σ²_R = 0.04, σ²_E = 0.08 (total variance = 0.34). Single observation visit (M = 1, R = 1):

ICC(1) = 0.12 / 0.34 = 0.3529 ≈ 0.35

Pooling M = 3 lessons across R = 2 observers:

Var(Error) = (0.10 / 3) + (0.04 / 2) + (0.08 / 6) = 0.0667

G = 0.12 / (0.12 + 0.0667) = 0.12 / 0.1867 = 0.6427 ≈ 0.64

Expanding to M = 4 lessons across R = 2 observers:

Var(Error) = (0.10 / 4) + (0.04 / 2) + (0.08 / 8) = 0.0550

G = 0.12 / (0.12 + 0.0550) = 0.12 / 0.1750 = 0.6857 ≈ 0.69

Interpretation

Danielson Framework composite scores map directly to qualitative performance tiers established by institutional policy or state regulatory statutes.

Composite Range	Rubric Level	Classification	Supervisory Action & Interpretation
1.00 – 1.99	Unsatisfactory	Ineffective	Instruction exhibits significant pedagogical misconceptions or disorder. Mandates an intensive Teacher Improvement Plan (TIP).
2.00 – 2.74	Basic	Developing	Teacher demonstrates basic content knowledge but inconsistent classroom routines and superficial questioning. Targeted coaching recommended.
2.75 – 3.49	Proficient	Effective	The benchmark standard for accomplished practice. Instruction is coherent, rigorous, and student-centered with active participation.
3.50 – 4.00	Distinguished	Highly Effective	Exemplary student-directed learning community where students take ownership of learning routines and self-assessment.

In educational measurement, researchers must distinguish between score consistency within a single rater and across multiple observers. Establishing baseline scorer consistency requires assessing intra-rater reliability across calibrated video scoring sessions. Evaluating whether score variance reflects day-to-day fluctuations in student behavior versus differences in rater strictness requires contrasting test-retest vs. inter-rater reliability.

Researchers must also examine the tension between reliability vs. validity. High inter-rater agreement ensures measurement consistency, but construct validity requires confirming that rubric ratings correlate with student learning gains. When evaluating whether observation rubrics perform consistently across schools, investigators must address internal vs. external validity trade-offs. Furthermore, when evaluating instruction for diverse student cohorts such as English learners, observers must evaluate how component 1b (Demonstrating Knowledge of Students) and component 3a (Communicating with Students) accommodate linguistic scaffolding.

Assumptions and common mistakes

Rigorous application of the Danielson Framework rests on foundational assumptions:

• Constructivist Orientation: Assumes that effective teaching emphasizes student-led inquiry and conceptual understanding over direct lecturing. In curricula requiring phonics drills or algorithmic practice, scoring must be calibrated to avoid penalizing necessary direct instruction.
• Observer Calibration: Assumes raters are formally certified and calibrated against national anchor videos to prevent idiosyncratic scoring drift.
• Interval Properties: Treats ordinal rubric ratings (1, 2, 3, 4) as interval data for mathematical averaging, assuming equidistant thresholds between performance levels.

Common methodological errors include:

• The Single Snapshot Fallacy: Relying on a single annual 45-minute drop-in observation to make high-stakes employment decisions. With an ICC of ~0.35, nearly 65% of score variance reflects lesson-specific noise and observer subjectivity rather than true teacher quality.
• The Distinguished Ceiling Trap: Expecting master educators to score Distinguished across all 22 components in every lesson, when Danielson explicitly intended Level 4 for student-initiated classroom dynamics.
• Central Tendency Bias: Observers clustering ratings at Level 3 (Proficient) to avoid supervisory conflict, artificially compressing score variance.
• Equating Compliance with Rigor: Scoring classrooms highly simply because students sit quietly, ignoring low cognitive demand in assigned tasks.

How to report it (APA example)

When reporting Danielson Framework results in empirical studies, researchers should document framework edition, observer training, reliability metrics, and domain-level descriptive statistics.

Example Narrative:

Classroom instructional quality was evaluated using the 2013 edition of the Danielson Framework for Teaching (Danielson, 2013). Two state-certified observers independently conducted double-scored ratings across three 45-minute classroom observations per teacher during the academic year. Inter-rater reliability for single observations demonstrated modest agreement, ICC(2,1) = .37, 95% CI [.28, .47]. Aggregating ratings across the two raters and three observation cycles yielded composite reliability of ICC(2,k) = .72, 95% CI [.63, .79]. Overall composite scores were computed using equal domain weighting across Domains 1 through 4. For the educator sample (N = 128), mean composite effectiveness was M = 3.26 (SD = 0.39, range = 2.15–3.92). Domain-level means were: Planning and Preparation, M = 3.32 (SD = 0.41); Classroom Environment, M = 3.35 (SD = 0.44); Instruction, M = 3.16 (SD = 0.46); and Professional Responsibilities, M = 3.42 (SD = 0.38). A paired-samples t test indicated that scores in Domain 3 (Instruction) were significantly lower than scores in Domain 2 (Classroom Environment), t(127) = 5.24, p < .001, Cohen’s d = 0.46, indicating that instructional questioning represented the primary area for targeted coaching.

How to calculate it in Excel, R or Python

The following tested snippets compute domain averages, composite scores, and G-theory reliability projections.

1. Excel Spreadsheet Implementation

# Domain Subscores:
Domain 1 (X2): =AVERAGE(B2:G2)
Domain 2 (Y2): =AVERAGE(H2:L2)
Domain 3 (Z2): =AVERAGE(M2:Q2)
Domain 4 (AA2): =AVERAGE(R2:W2)

# Equal Weight Composite (AB2):
=AVERAGE(X2:AA2)

# Weighted Composite (AC2):
=SUMPRODUCT(X2:AA2, {0.15, 0.35, 0.35, 0.15})

# G-Theory Reliability (AD2, M=3 lessons, R=2 raters):
=0.12 / (0.12 + (0.10/3) + (0.04/2) + (0.08/(3*2)))

2. R Script Implementation

calc_fft <- function(d1, d2, d3, d4, w = c(0.25, 0.25, 0.25, 0.25)) {
  sub <- c(mean(d1), mean(d2), mean(d3), mean(d4))
  return(list(domain_scores = sub, composite = round(sum(sub * w), 2)))
}

g_rel <- function(vt, vl, vr, ve, m = 3, r = 2) {
  err <- (vl / m) + (vr / r) + (ve / (m * r))
  return(round(vt / (vt + err), 3))
}

d1 <- c(4, 3, 3, 3, 4, 3); d2 <- c(3, 3, 4, 3, 3)
d3 <- c(3, 3, 4, 3, 3); d4 <- c(3, 4, 3, 3, 4, 4)
res <- calc_fft(d1, d2, d3, d4)
print(res)
cat("G-Reliability (M=3, R=2):", g_rel(0.12, 0.10, 0.04, 0.08, 3, 2), "n")

3. Python Implementation

import numpy as np

def score_fft(d1, d2, d3, d4, weights=None):
    w = np.array(weights if weights else [0.25, 0.25, 0.25, 0.25])
    means = np.array([np.mean(d1), np.mean(d2), np.mean(d3), np.mean(d4)])
    return {"domain_means": np.round(means, 4), "composite": round(float(np.sum(means * w)), 2)}

def g_rel(vt, vl, vr, ve, m=3, r=2):
    err = (vl / m) + (vr / r) + (ve / (m * r))
    return round(float(vt / (vt + err)), 4)

d1, d2 = [4, 3, 3, 3, 4, 3], [3, 3, 4, 3, 3]
d3, d4 = [3, 3, 4, 3, 3], [3, 4, 3, 3, 4, 4]
print("Scores:", score_fft(d1, d2, d3, d4))
print("G-Theory (M=3, R=2):", g_rel(0.12, 0.10, 0.04, 0.08, 3, 2))

FAQ

What are the four domains of the Danielson Framework for Teaching?

The four domains are Domain 1: Planning and Preparation (instructional outcomes, lesson design, assessment); Domain 2: The Classroom Environment (respectful rapport, culture for learning, managing procedures, student behavior); Domain 3: Instruction (questioning techniques, student engagement, communication, responsive assessment); and Domain 4: Professional Responsibilities (reflection, records, family communication, professional ethics).

How is an overall composite score calculated in the Danielson Framework?

Component rubric scores (1 to 4) are averaged within each domain to establish four domain subscores. These subscores are multiplied by domain weights (either equal 25% weights or weighted higher toward observable classroom practice in Domains 2 and 3) and summed to produce a composite score from 1.00 to 4.00.

How many observations are needed for a reliable evaluation score?

Findings from the Measures of Effective Teaching (MET) project demonstrated that a single observation has low reliability (ICC ≈ 0.35). Achieving an acceptable reliability coefficient of 0.65 to 0.70 requires pooling ratings across at least three to four distinct lessons observed by at least two independent evaluators.

What distinguishes Proficient from Distinguished practice?

Proficient practice (Level 3) is teacher-directed and represents successful, consistent implementation of rigorous pedagogy. Distinguished practice (Level 4) is student-directed, where students take active ownership of classroom routines, participate in instructional pacing, and conduct self-assessment.

Can Danielson Framework scores be combined with student test score gains?

Yes. State evaluation frameworks and research studies frequently combine observation composite scores with Student Growth Percentiles (SGPs) or Value-Added Models (VAM). Kane and Staiger (2012) found that combining observations with student surveys and growth metrics produces greater predictive validity and reliability than relying on observations alone.

Sources

• Danielson, C. (2013). The Framework for Teaching Evaluation Instrument (2013 Edition). The Danielson Group. https://danielsongroup.org/the-framework-for-teaching/ (Accessed: 2026-11-09).
• Kane, T. J., & Staiger, D. O. (2012). Gathering Feedback for Teaching: Combining High-Quality Observations with Student Surveys and Achievement Gains. Measures of Effective Teaching (MET) Project, Bill & Melinda Gates Foundation. ERIC ED540958. https://eric.ed.gov/?id=ED540958 (Accessed: 2026-11-09).
• New York State Education Department. (2024). Teacher Evaluation and APPR Guidance Under Education Law §3012-d. Office of Educator Quality and Professional Development. https://www.nysed.gov/educator-quality/teacher-student-data-linkage-and-evaluation (Accessed: 2026-11-09).