LBH Cult-evo agent based models emergent properties oct-17

Simulation parameters

• A description of the original long-billed hermit (LBH) data can be found here
• “Survival” probabilities (or more acccurately, the probability of showing up at the lek the next year) were estimated using the Cox Proportional-Hazards Regression (see results at the end of this file), which generates different probs for each age class
• When a bird “dies” a new bird takes its slot (territory) and either
  • produces a brand new song type (with a prob of 0.1848), or
  • learn a song from the song-type pool in the previous year, and that’s when the learning rules are applied:

Learning rule	Description	Algorithm
Anti-conformity	Adopt the least common trait in the previous generation	Assign the song type sung by the smallest number of individuals in previous year, if several then random sampled from those
Conformity	Adopt the most common trait in the previous generation	Assign the song type sung by most individuals in previous year, if several then random sampled from those
Content bias	Choose a cultural trait from the previous generation based on intrinsic characteristics of the cultural trait itself (e.g. good transmission)	A “content value” score was randomly assigned to each song type. Copying individuals get the highest score song type from previous year pool. New song types have a 0.25 probability of becoming the highest score song type
Independent decisions	Adopt a cultural trait from the previous generation at random ignoring trait frequency	random sampling of previous year unique song types
Model bias	Choose a cultural trait from the previous generation based on intrinsic characteristic of the individual bearing it (e.g. high mating success)	A “model value” score was randomly assigned to each individual. Copying individuals get the song type sung by the highest score individual from previous year. New individuals have a 0.25 probability of becoming the highest score model
Random copying	Choose another member of the previous generation at random and copy their cultural trait	random sampling of previous year song type pool

• All leks started with a single song type
• Each lek simulation was run for 550 years
• The first 50 years were discarded
• Lek size (# of individuals) did not vary through time
• Lek sizes (4-15) represent the observed range

• Each combination of lek size and learning rule was replicated 50 times:

  4	5	6	7	8	9	10	11	12	13	14	15
  Anti conformity	50	50	50	50	50	50	50	50	50	50	50	50
  Conformity	50	50	50	50	50	50	50	50	50	50	50	50
  Content bias	50	50	50	50	50	50	50	50	50	50	50	50
  Independent decisions	50	50	50	50	50	50	50	50	50	50	50	50
  Model bias	50	50	50	50	50	50	50	50	50	50	50	50
  Random copying	50	50	50	50	50	50	50	50	50	50	50	50

Preliminary results:

Cultural evolution parameters by learning rule-lek size

* Gray horizontal lines represent mean +/- SD

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]